A General Strategy for Rapidly Optimizing Wearable Resistive Pressure Sensors

Due to the complicated relationship between wearable electronics performance and various parameters, months even years are needed to obtain a desired sensor by random and time‐consuming trial‐and‐error methods. Herein, a general analytic model based on the micro‐element division and equivalent circuit is presented to guide a rapid optimizing strategy for wearable resistive pressure sensors, which is like the method always used in the traditional design of the metal‐oxide‐semiconductor field‐effect transistor for integrated circuits. The quantitative relationship between the sensitivity and related parameters is declared in the presented model, and the optimized parameters are achieved to design a sensor. The demanded ultra‐highly sensitive pressure sensor is successfully designed and optimized in minutes based on the built model, and the fabricated sensor is applied in a voice real‐time recognition system to obtain 100% recognition accuracy. The on‐demand and agile development strategy paves a promising way to greatly accelerate the transition from random and time‐consuming to the controllable design of wearable electronics.


Introduction
Over the past decade, wearable pressure sensors are capable of converting collected human body signals into electrical signals and providing visual output as a human-readable response, which is of value in health monitoring, [1][2][3] smart medical treatment, [4][5][6] artificial intelligence, [7][8][9] human-computer interaction, [10][11][12][13][14] and so on. Resistive pressure sensors have been widely studied benefit from their simple structure, easy signal acquisition, fast response, etc. [15][16][17][18][19][20][21] In these studies, most sensors DOI: 10.1002/aelm.202300201 are prepared with pure conductive sensitive materials, which realize pressure sensing by the conductive path changing caused by the opening and closing of micro/nano-pore structures between nanomaterials under external stimuli. Many inspired achievements have been gained to improve the sensors' performance in each specific application field. For example, Song et al. [22] fabricated an ultrathin sensor based on a 45 μm thick reduced graphene oxide (rGO) film by a flame-induced foaming process to form a 2D micro-porous structure. The proposed pressure sensor based on such porous film displayed a sensitivity of 12.55 kPa −1 in a wide detection range (0 -20 kPa). Chen et al. [23] achieved a sensor with low pressure detection limit. The MXene/carbon nanotube films with high flexibility were prepared by water evaporation induced self-assembly and used as sensitive layers. The sensor exhibited a sensitivity as high as 165.4 kPa −1 in the detection range from 0 to 6 Pa. Fan et al. [24] fabricated MXene/single-wall carbon nanotube composite films through vacuum-assisted filtration followed by thermal shrinkage. The sensor based on the proposed film shows good sensitivity (116.15 kPa −1 below 40 kPa and 12.7 kPa −1 at 40-130 kPa) and long-term stability. However, the tedious trial and error process (the material preparation, device preparation, device testing of several or dozens of samples with different experimental parameter combinations) requires months or even years to obtain satisfactory sensor performance. So far, most technical solutions are still based on such traditional trial-and-error methods, which are uncertainty and timeconsuming.
To obtain the desired characteristics of the sensor, many factors should be considered comprehensively, such as the electrical properties (resistivity), mechanical properties (stress-strain curve/ elastic modulus), and structural parameters (volume and thickness). Although finite element simulation software can supply support in some sense, it is complicated to build a 3D model as the real structure because there are always abundant nano-scaled morphologies of sensitive materials composed in the structure. In the nearest report, Xu et al. [25] gave a hyper elastic model that can guide design for the pressure sensor. However, it was just suitable for the sensor based on the ordered micro hierarchical structure, and only considered the influence of geometric parameters of the structure. It is difficult to build a general analytic model to predict the relationship between the aimed performance of the sensor and the various related electrical/mechanical/geometrical parameters. Thus, the need for agile development of wearable pressure sensors to improve the design effectivity like the method always used in the traditional design of the MOSFET or MEMS is significantly important, which can promote it to be applied in the industrial fields.
In this work, a method to guide an agile design strategy of wearable resistance pressure sensors through a general analytic model is presented. The model was built by means of microelement division to simplify the disordered and complex porous structures of the sensor, then the analytic model was established based on the equivalent circuit. The model declares the quantitative relationship between the sensitivity and related properties (resistivity, stress-strain curve/elastic modulus, volume, thickness) of the sensor. With the guidance of the presented analytic model, a pressure sensor was designed with the adopted micro-porous structural film by vacuum filtration and thermally induced foaming in confined spaces processes. The proposed pressure sensor exhibits extremely high sensitivity in the pressure range of 0-0.9 kPa (0.918 kPa −1 ), which can be successfully used in real-time monitoring subtle pressure signals such as human pulse and voice signals. Most importantly, the guidance of the analytic model greatly saves the time for parameter optimization from months or years to several minutes, by the calculation method with the analytical model instead of the trial-and-error method. The proposed strategy of building the general model to realize the on-demand and agile design of the wearable sensors opens a new path for the controllable design.

Model Establishment of Porous Pressure Sensor
To quantitatively declare the sensitive sensing mechanism of the porous pressure sensor, a simplified pore-closure model and equivalent circuit were built. The modeling process of the singlelayered porous pressure sensor is shown in Figure 1a. Under normal conditions, the porous structure can be represented as a resistor matrix with resistors on both rows and columns. The resistance between any two points can be calculated using the recursive transformation method proposed by Tan et al. [26] However, in our model, more attention is paid to the influence of resistance changes between electrode plates (surface to surface). The upper and lower electrode plates were divided into countless points through the micro-element method. The resistance can be expressed as the parallel connection of conductive paths between different points (the red path in Figure 1a). In the nopressure state, because the equivalent resistance value of multiple conductive paths in parallel is always limited by the shortest one (as shown in Figure S1, Supporting Information), the porous sensitive layer can be simplified as countless (N) conductive paths with minimal and equal resistance (R p ) (The minimum length of the conductive path can be approximated as the distance between the electrodes). The countless vertical conductive paths can be merged into one (The parallel connection of N identical resistors can be equivalent to an overall resistor with a cross-sectional area N time of the single resistor). Therefore, the initial resistance R O of the sensitive layer can be written as: where , h, w p , and l are the resistivity, length, width, and height of the conductive paths, and w is the width of equivalent resistance of N parallel conduction paths, respectively. The V is the volume of the skeleton material. As shown in Figure 1a, when the sensor is under pressure, the change of conductive paths can be simplified as the changes in the cross-sectional area and thickness of the overall resistance, while the volume is unchanged. Under the applied pressure P, the strain generated by the porous layer film is (P), and the length of the conductive paths changes to h' = h × [1-(P)], which is equal to the film thickness under pressure. The resistance R(P) under the applied pressure P can be written as: where N' is the number of the conductive paths under pressure, and w' is the width of equivalent resistance of N' parallel conduction paths. The simplification procedure can be extended to the multilayered composite porous structure, as shown in Figure 1b, the resistance of the n layers of the porous structure can be described as n resistors in series: where R O1 , R O2 , and R On are the initial resistance values of the first porous layer, the second porous layer, and the n th porous layer, respectively (the contact resistance between the layers is ignored). The initial resistance and resistance under the pressure of different layers can be written as: where 1 and V 1 are the resistivity and the volume of the skeleton material used in the first layer, respectively. 1 (P) represents the strain of the first layer under the applied pressure P (The second and n th layer is similar). The relationship between relative resistance variation ratio ΔR/R 0 and applied pressure P can be written as: Equation (7) is the simplified multi-layered pore closure model of the porous pressure sensor which declares that the sensitivity is affected by both the electrical properties (resistivity ) and mechanical properties (h and (P), which are related to the elastic modulus E) of the sensitive materials.

Design and Fabrication of the Pressure Sensor under the Guidance of the Model
Taking the typical application field of subtle pressure sensing such as pulse detection [27][28][29] or intelligent sound recognition [30,31] as an example, the sensor needs to have high-sensitivity sensing characteristics. In the meanwhile, the attachment and comfortability of the wearable devices to be used in subtle pressure sensing scenes require the sensor should have the characteristic of thinner thicknesses, which help the sensor to capture subtle pressure changes. [32][33][34] Therefore, it is aimed to design and fabricate an ultra-thin and ultra-sensitive wearable pressure sensor to verify the proposed strategy. Figure 2a displays the structure and fabrication process of the pressure sensor. The sensor is composed of three layers including the upper electrode, the reduced graphene oxide/MXene/ reduced graphene oxide (RGMRG) sensitive film, and the lower electrode. The RGMRG sensitive film is a porous-based film consisting upper rGO layer, MXene layer, and lower rGO layer, which is fabricated by thermally foaming the graphene oxide/MXene/graphene oxide (GOMGO) film in the confined space processes. There are two purposes of introducing the top and lower rGO layers. First, in the fabrication, because of the poor film-forming performance of MXene, it is difficult to completely peel off the MXene film obtained through suction filtration from the filter paper without the underlying rGO. Second, in the performance of stability, MXene is prone to deterioration and loss of performance in water vapor atmospheres, while the presence of upper and lower rGO layers can provide good protection for the middle MXene layer. The GOMGO film is obtained by vacuum filtration of graphene oxide (GO) and MXene nanosheets layer by layer. The thermally foaming process reduces GO to rGO and induces porous structures in rGO and MXene layers. The upper and lower electrodes are assembled to the upper and lower surfaces of the RGMRG sensitive film to form the pressure sensor.
For the pressure sensor designed in this study, the resistance value mainly consists of the resistances of the middle MXene layer, the upper and lower rGO layers. Since the thickness and the mechanical properties of each layer are related to the thermal reduction time of the film. The thickness of the MXene and the rGO layers can be corrected to h MXene (t) and h rGO (t), respectively. The functional relationship between stress and strain can be modified as MXene (P, t) and rGO (P, t). Due to the poor filmforming properties of MXene, it is difficult to peel off a complete film for mechanical property testing. The thickness of the MXene layer under different pressures can be obtained by subtracting the thickness of the rGO layer from the total thickness, which can be written as: The sensor resistance change rate ΔR/R 0 as a function of thermal reduction time and applied pressure can be written as: The cross-sectional morphologies of the RGMRG-sensitive film were investigated (the weight ratio of rGO/MXene/rGO = 1/1/1). As shown in Figure 2b, the thickness of the RGMRG film is ≈11.4 μm. The MXene layer was covered inside the upper and lower rGO layers. The partially enlarged views are also obtained in Figure 2b. It can be observed that the rGO and the MXene layers exhibit two different micro-porous structure morphologies. Compared with the disordered packing of MXene nanosheets, rGO nanosheets formed a more ordered layered porous structure. The formation of the micro-porous benefits from the thermally foaming process: Under the condition of the reduction temperature, the elimination of oxygen groups on GO and MXene nanosheets leads to the production of CO, CO 2 , and H 2 O. The gaseous species release rapidly and establish high pressure between nanosheets, which conquer the van der Waals forces that hold the nanosheets together. Eventually, a porous structure is constructed in the RGMRG film. The SEM cross-sectional morphologies of the composite film containing various porous structures are in good agreement with the established model, which lays the foundation for the use of this model in pressure sensors. To further explore the effect of different thermal reduction times on porous structures, samples with different reduction times were prepared (RGMRG@0.5 h, RGMRG@1 h, RGMRG@2 h). Figure 2c shows the SEM cross-sectional of different samples. With the increases of reduction time, the RGMRG film thickness increases and the porous structure becomes fluffier (Corresponding to reduction times of 0.5, 1, and 2 h, the thicknesses of the film are 8.24, 11.4, and 14.5 μm, respectively, and the thickness of pure rGO film is 7.96 μm). The thickness of each layer of different samples can also be obtained from SEM cross-section images. Figure 2d shows the stress-strain curves of different samples. The functions Total (P, t) and rGO (P, t) can be obtained by fitting the curves in Figure 2d. By substituting the relevant data into Equation (9), the theoretical sensitivity curves of samples with different reduction times can be obtained. As shown in Figure 2e, the sensitivity increases first and then decreases with the reduction time and shows the best sensitivity at t = 1 h.
In addition, the effect of material ratio on the sensor performance was also investigated. Set = V rGO /V MXene , the relative resistance variation ratio ΔR/R 0 can be written as: When t = 1 h, the sensor curves of the samples with different can be obtained by substituting the relevant data into Equation (10). As can be seen in Figure 2f, the sensitivity of the sensor increases with the increase of MXene proportion. In addition, the sensitivity curves of pure MXene and pure rGO films were also extrapolated from the obtained data. Pure MXene films exhibited the highest sensitivity but were not suitable for practical preparation due to the difficulty of suction filtration to form films and complete peeling. Pure rGO has the lowest sensitivity, which also proves that the introduction of MXene is beneficial to improve sensitivity. More derivation details and results can be achieved in the Supporting Information.
Through the built analytic model, the influence of various preparation parameters (thermal reduction time and material ratio) of the porous film on the sensitivity of the sensor was revealed. By entering different parameter combinations of the sample into the analytic model programmed in MATLAB, the sensitivity curves of different samples were rapidly obtained (as shown in Figure S2 (Supporting Information), the calculating time consumption of a single sample is less than 0.05 s), and the optimal parameters were selected by comparing the sensitivity curves. The whole time spent on the decision of the optimized parameters was within minutes. To verify the effectivity of the model, the parameter optimization process was executed by the traditional trial and error method. The time spent on the material preparation, device fabrication, and testing of a single sample was taken nearly two weeks, and the whole optimization process took more than half a year. The presented method greatly saves the time and workload of device optimization. According to the results of the above theoretical analysis and the actual situation, the process parameters of the device are preliminarily selected as thermal reduction time t = 1 h, the mass ratio of rGO:MXene:rGO = 1:1:1.

Electrical Properties and Sensing Performance of the Pressure Sensor
To cross-validate the experiment results with the analytic model, the sensitivities of the pure rGO sensor and the sensor with different thermal reduction times were tested and analyzed in Figure 3a. The sensitivity S is defined as S = -(ΔR/R 0 )/ΔP, where ΔR/R 0 is the relative resistance variation ratio under the applied pressure and ΔP is the loaded pressure range. R 0 is the initial resistance value of the sensor (The initial resistance R 0 of each sample is shown in Figure S3, Supporting Information). The sensor with RGMRG@1 h film exhibits the highest sensitivity among all the tested sensors. Especially in the 0-0.9 kPa detection range, the sensitivity S of the sensor is S RGMRG @1 h = 0.918 kPa −1 , and the sensitivities of other samples are S RGMRG @2 h = 0.789 kPa −1 , S RGMRG @0.5 h = 0.426 kPa −1 , S rGO @1 h = 0.412 kPa −1 , respec-tively. The sensitivity of the sensor first increases and then decreases with the increase of reduction time, which is consistent with the results of theoretical analysis in Figure 2f. Additionally, as predicted in Figure 2e, the sensor with RGMRG@1 h film displays higher sensitivity than the pure rGO sensor, which proves the positive effect of MXene layer on the sensor sensitivity. Figure 3b presents the comparison of theoretical and experimental sensitivity of different samples. The results are consistent with the above statement. The error between theoretical and experimental values mainly comes from the neglect of the influence of contact resistance between electrodes and sensitive materials during the modeling process. It is difficult to establish a theoretical model for this part, as the differences in surface morphology of different films and the non-uniformity of electrode adhesion can affect the initial value of contact resistance. To realize the optimization of performance, the RGMRG@1 h sample was selected for further testing, whose dynamic range is 0-45 kPa (the sensitivities in 0-0.9, 0.9-2, and 2-45 kPa are 0.918,0.122, and 0.00843 kPa −1 respectively). Figure 3c shows I-V curves of the proposed sensor from −1 to 1 V at different pressures. The high linear behaviors of the curves demonstrate that the sensor has excellent ohmic contact. The average resistances under different pressure of 0, 0.5, 1.0, 1.5, and 2.0 kPa are 1950, 515.5, 321.5, 175.4, and 143.6 Ω, respectively. Figure 3d illustrates the sensitivity curves of the proposed sensor after 1, 7, and 14 days. The performance of the sensor remains consistent during 14 days in the air environment. The inset is R 0 of the sensor, which is stable for 14 days (≈1950 Ω). It is further proved that the upper and lower rGO layers can protect the middle MXene layer and prevent its decomposition and oxidation from the air and water vapors. To prove the real-time reliability of the pressure sensor, ΔR/R 0 under various pressure (100 Pa, 500 Pa, 2 kPa, 5 kPa, and 10 kPa) are displayed in Figure 3e. Each loading-unloading process was repeated 4 times, and the applied pressure is maintained for nearly 20 s to determine whether the output signal is stable under continuous pressure. It can be observed that ΔR/R 0 kept stable under different applied pressures. With the increasing pressure, ΔR/R 0 also increased gradually. Furthermore, due to the highly tough porous structure, the sensor exhibits good dynamic performance under different loading-unloading frequencies as well. As shown in Figure 3f, ΔR/R 0 of the sensor is stable with the frequency of 0.17, 0.83, 1.6, and 3.3 HZ under 10 kPa pressure. Figure 3g illustrates that the response and recovery times are ≈161 and 80 ms, respectively, which is mainly benefit from the low hysteresis properties of the porous structure (The applied pressure is 1 kPa). Figure 3h shows ΔR/R 0 of the sensor during 5000 cycles durability test. The sensing performance was stability retained and showed no obvious degradation during the whole durability test. More details of the stable signal output can be seen in the inset. As a wearable sensor, bending is a common condition. The bending sensitivity curve of the proposed sensor is shown in Figure S4 (Supporting Information). The experiment results display that the resistance of the sensor changes slightly at a small bending angle (0-45°), while significant increase at a large bending angle range. The low sensitivity to small curvature surfaces of the proposed sensor indicates that is suitable for wearable fields such as pulse detection.
The pressure sensor demonstrates a high sensitivity (0.918 kPa −1 in the range of 0-0.9 kPa) and a thin thickness (Overall thickness after encapsulation is 150 μm) compared to the state of art (Figure 3i, more performance details in Table S1 and Figure S5, Supporting Information). The advantages make it suitable for small signal detection applications in wearable devices.

Applications in Real-Time Pulse Wave Monitoring
Among the pressure signals collected by various wearable applications (such as various joint bending, plantar pressure, sound, and pulse, etc.), the pressure signals generated by pulse or sound belong to subtle pressure signals. It not only requires the sensor to have a low detection limit, but also requires the sensor with high sensitivity for many detailed detections (subtle signal changes) during the monitoring. [49][50][51] Considering the outstanding subtle performance of the proposed sensor, as a proof-ofconcept, the sensor was used as a pulse monitoring system for collecting human pulse waveform before and after exercise. The characteristic peaks of the three human sphygmic waveforms relevant to PTD waves (percussion waves, tidal waves and diastolic waves) can be distinguished successfully in Figure 4c,d. [52,53] Figure 4a,b illustrates the response signals of the pulse for 1 min before and after exercise, respectively. The Fast Fourier transform (FFT) results are shown in Figure 4e, which demonstrated pulse rate before and after exercise was 1.11 and 1.73 Hz, respectively. (72 times min −1 before exercise and 98 times min −1 after exercise). The results are consistent with the pulse rate of healthy adult males, [54] which further confirms the application prospect of the proposed sensor in sports health monitoring.

Applications in Sound Signal Collection
The pressure sensor can also be used as a sound sensor like a microphone to detect the tiny vibrations of the air caused by sound for voice recognition. The Bluetooth speaker was used to repeat the pronunciation of "two", "happy", "wonderful", "Happy new year", "Merry Christmas", "Have a good time" and the song 800 times each, and a total of nearly 480 000 datapoints were obtained (the waveform of each sound signal is shown in Figure S6, Supporting Information). According to the chronological order of datapoints, 26 features were calculated for every 400 datapoints during the feature extraction, and 1200 feature samples were extracted after feature extraction (80% as training set and 20% as the test set). The data preprocessing process is shown in Figure 5a. The original signals of different pronunciations displayed different waveforms and distinguished feature points. The signals were low-pass filtered by a third-order Butterworth filter with a cut-off frequency of 20 Hz, and then processed by the Z-Score Normalization. The processed signals were divided into frames with a 128-frame length and a 32-frameshift. The Hann window function was used to window the signals. The windowed signals were subjected to the FFT algorithm to extract 26 audio features, including Chromagram, Spectral Centroid, Spectral Bandwidth, Spectral Roll-off, Root Mean Squared Error (RMSE), Zero Crossing Rate, and 20-dimensional Mel Frequency Cepstral Coefficients (MFCCs). The audio features were used as inputs to the Bi-LSTM model, which shows great advantages in the fusion of contextual semantic information. As shown in Figure 5b, the Bi-LSTM model consists of two LSTM layers including a forward LSTM layer and a backward LSTM layer. The LSTM layer can selectively forget and retain information through the gating mechanism, effectively solve the gradient disappearance or explosion problem of RNN, and is suitable for the classification of timeseries signals. [55,56] The Bi-LSTM model can fuse forward and backward semantic information from two LSTM layers to provide complete contextual state information, which is beneficial for sound recognition. The Bi-LSTM output was used as input to the fully connected layer and was recognized and classified by the SoftMax activation function. The ROC curve of sound recognition is shown in Figure 5c. Classification using audio feature extraction results in an FPR of 0, TPR of 1, and AUC of 1. Figure 5d,e shows that the confusion matrix of the sound recognition accu- racy was improved from 61.79% to 100% by using audio feature extraction in classification. A real-time voice display system was designed, and the collected data was uploaded to the host computer through Bluetooth. After being classified by the machine learning model, the identified result was displayed on the host computer screen in real-time as shown in Figure 5f. The system can help quadriplegics enter text into a computer by voice, instead of typing with their hands. A more detailed presentation is available in Video S1 (Supporting Information), the text "one world one dream", "S" and "Micro-Electro-Mechanical System" were successfully identified and displayed on the host computer screen in this demonstration. Additionally, Figure S7 (Supporting Information) demonstrates the ability of the pressure sensor to be used in monitoring the signals of human throat vocalization by the movements of muscles in the larynx, which are different from Bluetooth audio vocalizations generated by the air vibration from the sound transmission. Informed consent was obtained from the participant who volunteered to perform all the experiments and studies (i.e., wearable testing and image publication). All testing reported conformed to the ethical requirements of Southeast University. And there are no animal and medical experiments included in the article.

Conclusion
In summary, a general analytic model of the wearable resistive pressure sensor was established to study the effects of material and process parameters on the device sensitivity, and a demanded pressure sensor was designed. The proposed sensor was fabricated by thermally induced foaming in confined spaces to form an ultra-thin RGMRG film with micro-porous structures. The cross-validation of experimental results and theoretical analysis shows that with the increase of the thermal reduction time of the RGMRG sensitive film, the sensitivity of the sensor increased and then decreased, and there was an optimal value of the thermal reduction time (1 h). In addition, the sensitivity also increased with the proportion of the MXene. The sensor with the optimized parameters exhibited extraordinary performance such as ultra-high sensitivity (0.918 kPa −1 ) in the detection range of 0-0.9 kPa, fast response times (response time 161 ms, recovery time 80 ms), and long-term stability. It demonstrates an effective way to achieve ultra-high sensitivity to satisfy the increasing demands for subtle pressure sensing in real-time health monitoring, voice recognition systems, and so on. As the most significant contribution of this work, the presented agile design strategy with the analytic model can significantly shorten the optimization time compared with the traditional method. It paves a promising way to accelerate the research and transition from random to the controllable design of wearable electronics. However, due to the nonstandardized fabrication process of sensors, it is difficult to consider including more materials and sensor structures to verify the model in one article. In future research work, we will continue to design new sensors under the theoretical guidance of the model to further verify the model.

Experimental Section
Materials: The conductive metal tape was brought from MaoYe Inc. (Shenzhen, China). The GO nanosheets with diameter of 0.5-5 μm, an av- Figure 5. Applications of the pressure sensor in sound recognition. a) Schematic illustration of the data preprocessing. b) Schematic illustration of the bidirectional long short-term memory (Bi-LSTM) deep-learning algorithm. c) The ROC curves of the algorithm with feature extraction and without feature extraction. d) Classification confusion matrices without feature extraction. e) Classification confusion matrices with feature extraction. f) The schematic illustration and photograph of the real-time voice display system. erage thickness of 4 nm, and MXene with an average diameter of 0.05 μm, an average thickness of 4 nm was purchased from XFNANO Inc. (Nanjing, China). All the materials above were utilized without any further purification.
Preparation of the RGMRG Films: The preparation processes of the RGMRG film are schematically displayed in Figure S8 (Supporting Information). In the typical experiment, 100 mg GO nanosheets were mixed in 100 mL deionized water (DI). The mixed solution was sonicated and stirred for 30 min to obtain 1 mg mL −1 GO suspension at room temperature (20°C). The MXene suspension (1 mg mL −1 ) could be achieved by the same procedure. After that, the GOMGO film was prepared by vacuum filtration as shown in Figure S8b (Supporting Information). First, 10 mL GO suspension was filtrated to form the first GO layer on filter paper. Second, 10 mL MXene suspension was filtrated to form the MXene layer. Third, 10 mL GO suspension was filtrated to form the second GO layer. The filtration process continues for 12 h and natural drying for 24 h. The MXene layer can be well encapsulated in two GO films and not easily oxidized. Finally, the high-temperature reduction method was used to prepare the RGMRG film. The RGMRG film was peeled off from the filter paper and clamped with two quartz plates. Figure S8c (Supporting Information) displays the optical photo of the GOMGO film with a diameter of 4 cm. After heating and reducing at 300°C for 1 h in a vacuum oven, the RGMRG film was successfully prepared (The optical photo could be seen in Figure S8d, Supporting Information).
Preparation of the Pressure Sensor: The pressure sensor includes the upper and lower PI encapsulation layers, the upper and lower electrodes prepared with metal conductive tape, and the RGMRG film as the pressure-sensitive layer. The upper and lower electrodes were obtained by cutting the conductive metal tape into the required shape by laser cutting (as shown in Figure S9 (Supporting Information), the size of the electrode area is 1 cm * 1 cm), and two copper wires were connected to the pad to realize electrode extraction. The pressure sensor was assembled with the RGMRG film between two electrodes and encapsulated by PI tape. The RGMRG film was cut into 1.2 cm * 1.2 cm and sandwiched between the electrodes.
Characterization: The morphology of the RGMRE films was observed with Helios 5 CX Scanning Electronic Microscopy at a 15 kV acceleration voltage. The Mark-10 machinal testing machine with ESM303 force test stands and SERIES 5 digital force gauges was used for mechanical measurements, and a semiconductor parameter analyzer (Keithley 4200-SCS) was used for electrical measurements.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.