Highly Sensitive Piezoelectric E‐Skin Design Based on Electromechanical Coupling Concept

Stretchable electronic skin (e‐skin) paves the way for applications that exceed the scope of intrinsic rigid devices and hard‐to‐stretch sensors. The broad application range of flexible e‐skins benefits from device architectures that can simultaneously provide mechanical flexibility and superior sensitivity. Classic fractal design provides a simple architecture to achieve the desired flexibility through structural design for improved wear comfort, but at the expense of sensor sensitivity. In this study, the proposed method addresses the trade‐off between stretchability and sensitivity in fractal design. A high‐sensitivity e‐skin is obtained by eliminating the effect of negative charge on the output by applying the concept of electromechanical coupling. This concept for designing e‐skins with high sensitivity is validated through the delicate patterning of hard‐to‐stretch functional materials. Further, human speech signals are acquired through the integration of e‐skin with signal processing circuits, and speech pattern recognition is realized using machine learning. The stretchable e‐skin with an enhanced gauge factor illustrates the wider application of this concept for improving the sensitivity of stretchable electronic functional materials.


Introduction
Stretchable electronic skin (e-skin) has attracted the attention of an increasing number of researchers owing to its seamless fit onto complex surfaces and wearing comfort. These properties make e-skins suitable candidates in the fields of physiological signal monitoring and healthcare. Achieving sensation-free attachment to the area of interest and precise monitoring of subtle mechanical vibrations not only poses a core challenge to www.advelectronicmat.de simultaneously offer advanced electronic properties and unusual compliant mechanics. Theoretical analysis and experimental verification have proven that Peano, Greek cross, Vicsek, and other fractal constructs can effectively reduce the modulus in stretchable electronic devices, providing the ability to co-integrate multiple high-performance material platforms with high spatial resolution and engineering control. [7,29,30] In this way, electrophysiological sensors, precision monitors, actuators, and radio frequency antennas have been illustrated owing to proper electrical contacts of e-skin that is applied internally and externally to the system. [7,28] In particular, threedimensional (3D) structures can spontaneously form through fractal-designed 2D architectures. These structural designs can flexibly control the modulus of e-skin and achieve a better skin fit. However, because fractal architectures are processed using subtractive manufacturing of functional materials where hardto-stretch material is partially removed, providing better stretchability due to smaller modulus. At the same time, the strain sensitivity of the e-skin is reduced accordingly due to removal of the functional material. However, in practical applications, such as physiological signal monitoring, speech signal recognition, and monitoring weak mechanical signals caused by subtle vibration of organs, e-skins with higher sensitivity are required without restraining daily activities. [31][32][33][34] Therefore, it is of great significance to explore approaches to improve sensor sensitivity with typical architectures that can provide e-skins with high mechanical properties.
To address the trade-off between stretchability and sensitivity, this study proposes a strategy based on the concept of electromechanical coupling. Due to the e-skin based on the piezoelectric effect can generate electrical signals when its functional material is stretched or compressed. Moreover, the output voltages of e-skin are related to the stress on the functional material. Thus, a stretchable piezoelectric strain sensor was chosen to demonstrate the e-skin. Areas of the e-skin that form a negative charge owing to compressive stress during health monitoring are removed, which improves the sensitivity of the e-skin under the premise of reducing the modulus and enhancing the durability of the stretchable e-skin. Through theoretical analysis and experimental verification results based on the fractural structure of the e-skin, the proposed method can increase the strain sensitivity of the e-skin by 260%. This design enables the realization of a high-sensitivity and conformable e-skin. Subsequently, the adhesion properties of the e-skin with different encapsulations are investigated in this study to ensure that the e-skin can be in conformal contact with human skin. In addition, human voice signals were monitored using a signal-processing circuit. Using the Support Vector Machine (SVM) algorithm, the results show that the e-skin proposed in this study can recognize voice signals. Figure 1A presents a schematic illustration of the basic working mechanism of the e-skin. The structure of the functional layer was based on a fractural architecture consisting of serpentine unit cells of the same width, arc angle, radius, and span, along the vertical and horizontal directions. Using the cutting method (see Experiment Section for details), a piezoelectric poly(vinylidene fluoride) (PVDF) layer (28-µm-thick) sandwiched between two metal layers (Cu-Ni, 70-nm-thick) which served as the bottom and top electrodes, was processed to form the serpentine architecture. As indicated by the finite element analysis (FEA) results (Figures S1-S3, Supporting Information), an obvious inhomogeneity in the stress distribution can be observed in the stretched piezoelectric film with the serpentine layout, which affects the distribution of the charge generated by functional materials. This leads to difficulties in evaluating the output electrical voltages based only on the global tensile strain. Thus, it is necessary to further investigate the influence of the stress distribution on the piezoelectric output properties (There is out-of-plane deformation, but it has little effect on output voltages because the functional material used in this study is uniaxially polarized).

Principle of Electromechanical Coupling-Based Design
Based on the piezoelectric material constitutive equation and mechanical analysis model of the serpentine structure, we propose an expression for calculating the open-circuit output voltage of the serpentine e-skin. The expression begins with a piezoelectric constitutive equation. The serpentine architecture belongs to the category of piezoelectric plate theory and plane strain because of the thickness of the e-skin and absence of out-of-plane loads. [35,36] Thus, the out-of-plane shear strains e 13 and e 23 have no effect on the open-circuit output voltage and the electric displacement component D 3 is independent of the out-of-plane stress of electronic skin. [36] Based on the above theory, the piezoelectric constitutive equation can be expressed as Equation (1) because σ 13 = σ 23 = σ 33 = 0. [36] where e 11 , e 22 , and e 33 are the strains in the x, y, and z directions, respectively. c ij (i, j = 1, 2) are elements of the compliance matrix. d 31 , d 32 , and d 33 represent the piezoelectric constants of the functional layer in the x, y, and z directions, respectively. σ 11 and σ 22 represent the stresses in the x-and y-directions, respectively. E 3 represents the electric field strength in the thickness direction of the functional material and D 3 is the electric displacement induced by the piezoelectric effect. The functional material used in this study was elastic and isotropic. Thus, the elements of the compliance matrix can be expressed as in Equation (2): where µ and Y represent Poisson's ratio and Young's modulus of the functional material, respectively. We assume that the electric potential of the bottom electrode is zero and that of the top electrode is equal to the e-skin's output potential U. Therefore, the electric potential inside the functional layer is related to the parameters of the normal rather than the www.advelectronicmat.de parameters in the x-y plane. Verification details are provided in the Supporting Information. Because the normal component of the electric field E 3 is constant inside the material, the normal component of the electric displacement D 3 depends on the parameters of the x-y plane rather than its counterpart in the normal direction. Therefore, the electric field can be expressed by Equation (3).
where U, U 0 are the electric potential, electric potential of the top electrode; h is the thickness of the functional material; E 1 , E 2 , and E 3 are the electric fields in the x, y, and z directions, respectively; and D 1 , D 2 , and D 3 are the components of the electric displacements in the x, y, and z directions, respectively. Because the free charge in the electrode is zero during the open-circuit voltage test, the relationship between the output electrical potential and strain can be obtained by combining Equations (1-3). The detailed calculation process is illustrated in the calculation section of the Supporting Information.
where A s and dA denote the area of the serpentine structure and unit area of the functional material, respectively.
∧ 11 e represents the local strain in functional materials in the x-direction.
From Equation (4), the open-circuit voltage of the e-skin with thickness h can be obtained by integrating the local strain. The corresponding top electrode generates a positive electrical potential when the local strain of the functional layer is positive which strengthens the average electrical potential. Conversely, the average electrical potential weakens when the local strain of the functional layer is negative. Figure 1B shows the electrical potential distribution analysis of the e-skin with a fractural architecture under tensile strain. Local positive and negative electrical potentials coexist when the e-skin is stretched owing to the corresponding local tensile strain, and compressive strain appears in different parts of the functional layer (see Figures 1B and Figure S2, Supporting Information). This result is in agreement with the theoretical analysis described above. However, the electrical potential measured by the e-skin during the actual monitoring process is the average of the positive and negative electrical potentials generated by functional materials. Thus, it is necessary to eliminate the influence of negative electrical potential to improve the voltage output of the e-skin. Based www.advelectronicmat.de on electromechanical coupling, this work improves the output voltage of the e-skin by removing the corresponding electrodes of the negative potential. Figure 1C shows an e-skin diagram for speech signal monitoring. Studies have confirmed that muscle motion patterns in the lower jaw contain useful voice information related to speech. [37,38] In addition, different voice signals cause different motion patterns in human and silent speech. Therefore, it is possible to distinguish between internal speech information and useful voice information according to the motion patterns caused by the muscle group of the lower jaw. [39] Here, the e-skin was attached to the lower jaw to capture the weak strain during the volunteers' speech. Owing to the piezoelectric effect, the e-skin converts the mechanical strain into electrical signals, which are output by an overamplifying circuit.

Design and Fabrication of Highly-Sensitive E-Skin
The functional layer of the e-skin consists of periodic arc-shaped units ( Figure S4, Supporting Information,). The unit cells are all oriented in the same way, maximizing uniaxial stretchability along the direction of the unit cell. The e-skin contains cells with alternating orientations, balancing the maximum strain supported along the x-and y-axes, respectively. The improvement iterations are due to the combined effects of the geometric scaling of the arc section, the increased length of the lines, and spring-like constructions. The cells' sizes and serpentine architecture are consistent with those of the fractal e-skin electrode, as shown in Figure 2A. For further exploration of the influence of the electromechanical coupling concept on the e-skin sensitivity, the length of the e-skin is designed larger than the width, i.e., length and width are 3.5 and 1.5 cm, respectively. The tensile stress in the x direction is larger than the counterpart in the y direction when e-skin is stretched transversely. Figure 2A-C shows the diagrams of the electrodes that are not etched, electrodes with only the central part removed, and electrodes with the one that generates negative electrical potential removed, respectively. [40] As already highlighted, e-skin sensitivity can be improved by removing the electrodes corresponding to the compressive strain (stress) regions.
Cut-and-paste and wet etching technologies are used to process the e-skin ( Figure 2D). A (PVDF) film (with a thickness of 28 µm) sandwiched between two 70 nm copper-nickel layers was patterned according to the designed serpentine structure using a mechanical cutting machine. The overall size of the patterned PVDF was 3.5 × 1.5 cm. Subsequently, the sandwiched structure was wet-etched to etch the PVDF electrode using a UV film as a mask. Then, the e-skin was transferred to a glass Petri dish using transfer printing technology to prepare for connection with external electrodes. Finally, the e-skin was encapsulated in elastomers (with a thickness of 100 µm). To prevent leakage current in the e-skin during speech signal monitoring, the length and width of the encapsulation layer were 5 mm greater than those of the functional layer. The detailed processing is shown in Figure S5 (Supporting Information) and the Experiment Section.

Performance Characterization of E-Skin
To explore the effect of electrode removal on the sensitivity improvement of the e-skin based on the concept of electromechanical coupling, the output voltages of the three types of e-skins, namely e-skin with the electrodes generating negative electrical charge and central parts removed (ESNECPR), e-skin with only the central parts removed (ESCPR), and e-skin without etched electrodes (ESWEE) were tested ( Figure S6, Supporting Information). Figure 3A shows a schematic of the setup for the cyclic tensile test of the e-skin. The two ends of the e-skin were clamped onto the clips of the tensile tester, and the initial state was maintained in a tensioned state to prevent compression of the e-skins during cyclic stretching. The electrical potential generated during the cyclic stretching of the e-skin was measured using an electrometer through the bonding wires on both sides of the PVDF. Figure 3B shows Because the piezoelectric coefficient d 31 in the x-direction of the PVDF used in this study is more than 10 times larger than its counterpart in the y-direction, removing the central part of the electrode can further reduce the influence of the strain in the y-direction. Figure 3C shows a comparison of the local strain for the three types of e-skins under the same global strain. The local strain generated by the ESNECP is smaller than that generated by the other two e-skins. Maximum local strains of the ESCPR and ESWEE are the same because maximum local strain occurs at the place inside and outside the arc segments, where the device is subjected to tensile stress. Here, the global strain is expressed as ΔL/L, where ΔL and L represent the change in the length and total length of the e-skins, respectively. The relationship between the displacement and stretching force of the three e-skins is shown in Figure 3D. Compared to the other two e-skins, the figure shows that the ESNECP requires less stretching force applied under the same displacement, and has a smaller modulus. Figure 3E shows the voltage-displacement curves of the three types of e-skins. The output voltage of the ESNECPR is the highest of 2.6 times higher than that of the ESWEE, followed by the ESCPR of 1.5 times higher than that of the ESWEE at the same displacement. Figure S7 (Supporting Information,) shows the voltage-strain curves of three different sensors. It can be seen that the strain sensitivities of the sensors are 0.33, 0.21, and 0.128 mV µε −1 according to the slopes  Figure 3F shows the output voltages of the three e-skins under cyclic stretching, with a maximum displacement of 3 mm. Based on the results of Figure 3E,F, electrode removal technology based on electromechanical coupling can effectively improve the sensitivity of fractal e-skins.
Stretchability and robustness are critical metrics for the physiological monitoring of e-skin. Therefore, it is necessary to study these two characteristics after electrode removal. Because the modulus of the electrode is larger than that of the flexible functional material in this study, the resistance change of the electrodes measured by the multimeter can account for the stretchability of the e-skins during the stretching process. This is because the electrode often cracks or even breaks during stretching before the functional material experiences fatigue damage. Figure 4A shows the normalized resistance changes of the e-skins with straight ribbons, ESWEE, ESCPR, and ESNECPR under different strains. This indicates that the fractal architecture can significantly improve the tensile properties of the hard-to-stretch e-skin. Meanwhile, given critical strain-to-rupture as the tensile strain applied to the substrate at which R/R 0 = 6(R 0 and R represent the resistances before and after stretching), [41] ESNECPR has a stretching limitation of 60%, which is larger than that of the other two e-skins (52.5% and 55% for ESWEE and ESCPR, respectively). Figure 4B shows images of the e-skin at 0% and 25% tensile strains. The stretchability of e-skin exceeded the stretch limit of painless human skin (approximately 23%). Figure 4C shows the output voltages of the ESNECPR when 5 mm uniaxial stretching was applied 760 times and reveals that the output voltages of the e-skin remained stable during the stretching cycle, which confirms the robustness of the fabricated e-skin. Figure 4D shows an enlarged view of Figure 4C. The output signals of the e-skin had a stable period during the cyclic stretching test, with no distortion. This further verifies the reliability of the fabricated e-skin during the long-term strain-monitoring process.

Adhesion Study of E-Skin with PDMS and Ecoflex Packages
Merely exhibiting excellent electrical performance is not sufficient to capture the weak mechanical vibrations; this is because it is difficult to guarantee that the mechanical vibration can be well transmitted to the surface of the e-skin when the e-skin is layered within the area of interest. This weakens the detection accuracy and even leads to failure of the monitoring process. Therefore, the adhesion toughness at the area of interest is very important for the monitoring accuracy of the e-skin system to carry out long-term stable and reliable monitoring, aimed at ensuring high electrical sensitivity and mechanical flexibility of the e-skin. [42] Furthermore, additional mechanical fixing devices or tapes should be avoided as much as possible to reduce restrictions at the attached area. [43] Thus, it is necessary to study the adhesion between e-skin and human skin, which is dominated by van der Waals forces. [44] Figure 5B illustrates the strain-stress curves of Ecoflex with a mixing ratio of 1:1 and PDMS with different mixture ratios (A:B = 5:1, 10:1, and 15:1). (The modulus is shown in Table S3, Supporting Information). To verify the robustness of the adhesion ability of the e-skin on human skin and guarantee the repeated use of the e-skin, peeling experiments were performed 15 times using PDMS and Ecoflex at different mixing ratios. Figure 5C summarizes the results of the measured peeling energies. The adhesion energy of PDMS with a mixture ratio 15:1 is larger than that of other elastomers. However, the 15 th peelings' adhesion energy does not dissipate significantly compared to the first peeling for PDMS and Ecoflex. Figure 5D shows the determination of conformal contact with the skin for 100 µm-thick encapsulation materials with different Young's moduli. The detailed calculation process is illustrated in the calculation section of the Supporting Information. The Work of adhesion increased with increasing modulus of the encapsulation layer. When the modulus of the encapsulation layer was greater than 40 kPa, the critical adhesion increased slowly. Tough conformal contact can be realized when the selected elastomer's adhesion is distributed above the curve. In contrast, the e-skin cannot be in good conformal contact with the skin. From the above comparison, it can be concluded that PDMS with a mixture ratio of 15:1 has the best adhesive properties to human skin, and it is selected for encapsulation in the following section.

Stretchable and Sensitivity-Enhancing E-Skin for Speech Monitoring
The monitoring of speech signals can be achieved by integrating ESNECPR with a signal processing circuit. A schematic of the signal processing circuit is shown in Figure S14 (Supporting Information,). By attaching the e-skin to the lower jaw, different electrical output patterns can be obtained, owing to the different dynamic strains received by the e-skin while the volunteer says different words. Figure 6 shows that when the volunteers repeat words such as "let," "me" "Introduce," "Tianjin," "university," "MEMS" etc., there is a strong correlation between the collected voltage signals and speech pattern. Meanwhile, each specific word has its corresponding unique voltage signal pattern, with good repeatability, including the voltage signal amplitude and duration. These characteristics are attributed to different micromotions of the skin on the lower jaw, suggesting that muscle movements also have recognizable signatures. Figures S15 and S16 (Supporting Information) show www.advelectronicmat.de the measurement setup and dimensions of the e-skin electrodes for speech monitoring. Figures S17 and S18 (Supporting Information) show the voltage signals captured from ESWEE and ESCPR, respectively, when volunteers say the same words. As can be seen from the figures, the amplitudes of the speech signal measured by ESWEE and ESCPR are lower than that measured by ESNECPR with increased sensitivity. Meanwhile, the periodicities of the voltages measured by the two devices with lower sensitivity are irregular. When volunteers repeatedly uttered words, ESCPR sometimes could not capture the generated microstrain well, so that the output voltage signals could not be distinguished. ESWEE failed to capture the signal more times than the other two sensors. It shows that ESNECPR is superior over the other two e-skins in speech signal monitoring.
Using machine learning to classify speech, the core of the support vector machine is to seek the optimal hyperplane of the feature space to classify the target. The SVM is a novel few-shot learning method with a solid theoretical foundation. The inner product kernel function can be used instead of nonlinear map-ping to a high-dimensional space. It offers the advantages of a simple algorithm and strong robustness. Therefore, this study uses the support vector machine to classify and learn the 11 520 signal segments corresponding to different English words, as shown in Figures 6 and 7. Before recognition, 80% of the signal segments were randomly selected from the overall signal for the SVM model training. The training process involved 20 features of a signal segment. The remaining 20% were used as the test group. Figure 7A shows the number of real-type and prediction signals after model training, and Figure 7B shows the ratio of the true positive rate (TPR) and false negative rate (FNR). The results show 89.5% recognition accuracy of the trained data.

Conclusion
This study proposes a highly sensitive fractal piezoelectric e-skin enabled by a novel electromechanical coupling concept. Figure 6. Voltage signals captured from e-skin when the volunteer says the different words. A-I correspond to the words: "Let," "Me" "Introduce," "MEMS" "Group," "Tianjin," "University," "Strain," and "Let me introduce myself", respectively.

www.advelectronicmat.de
The fundamental mechanical and electrical response dependence on the strain distribution were demonstrated using a combination of theoretical analysis, FEM simulations, and high-precision electromechanical measurements. Through investigating the electromechanical coupling concept, a novel fractal-based e-skin in stretchable electronics was designed to monitor macromotion. In addition, e-skin, fabricated using the concept mentioned above, can reliably monitor muscle movement caused by speech when attached to the skin of the lower jaw. Combined with machine learning, speech signals can be recognized well. Additionally, this study illustrates the broader application opportunities of the electromechanical coupling concept not only in piezoelectric materials but also in conventional hard-to-stretch high-performance functional materials in the design of novel, flexible, and stretchable electronic devices. Furthermore, the study proposes a method that can be applied to other e-skins to achieve the balance of flexibility and sensitivity. We believe that the novel design shows great potential for medical treatment, biomonitoring, and sensing.

Experimental Section
Fabrication Process of E-Skin: The fabrication process of the e-skin is divided into three steps. (1) Functional layer patterning. First, a suitably sized PVDF film with metallized electrodes was pasted onto a temporary substrate of the UV film. Then, the PVDF film was patterned using a cutting machine based on the designed fractal architecture. It is worth noting that the cutting speed and pressure are set to 20 mm s −1 and 50 mN, respectively. This ensured the integrity and smoothness of the edge of the e-skin. (2) Electrode etching. First, the structure of the UV film was designed according to the results of mechanical analysis. Because the UV film is sensitive to UV light, processing of the UV film should be avoided when exposed to UV light. The patterned UV mask was then attached to the surface of the processed PVDF. The functional layer sandwiched between the UV film was etched using a wet etching solution (the composition of the wet etching solution Figure 7. A) Numbers of real and prediction type signals (Here, "swallowing" represents "strain sensor and swallowing"). B) Ratio of true positive rate (TPR) and false negative rate (FNR).
www.advelectronicmat.de used in this study was H 3 PO 4 :C 2 H 4 O 2 :HNO 3 :H 2 O = 32:1.5:1:28). The electronic skin was completely immersed in the etching solution for 1 min under compressive strain to remove the electrode parts. The e-skin was then placed under an ultraviolet lamp for degumming and the UV film was slowly peeled off using tweezers. Finally, the etched e-skin is transferred to a temporary glass substrate for packaging. (3) Encapsulation. A 0.5 mm wide Cu ribbon was bonded to the top and bottom electrodes of the PVDF film using conductive silver glue, serving as the lead-out electrode during the cyclic stretching process. The e-skin was encapsulated with an elastomer. The details of the etchant ratio and the UV film pasting process are shown in the Supporting Information.
Encapsulation's Fabrication Process: To fabricate PDMS encapsulation with different mixing ratios, prepolymer A and cross-linking agent B were mixed in mass ratios of 15:1, 10:1, and 5:1, followed by stirring for 20 min using a magnetic stirrer to ensure uniform mixing. A vacuum pump was then used for 30 min to eliminate air bubbles in the mixture. Finally, the appropriate volume according to the thickness of the encapsulation was poured into a petri dish and then baked in an oven until it solidified. The preparation process of the Ecoflex encapsulation was the same as that for PDMS. The detailed fabrication process can be found in Supporting Information.
Cyclic Stretching Test: To verify the robustness of the e-skin, a tensile cyclic stretching test was conducted. The e-skins were stretched to a set displacement using a tensile tester (Model ESM303H, Mark-10). An electrometer (DMM5614; Keithley) was used to record the output voltage during the cyclic stretching. To verify the tensile limitation, the change in resistance of the top electrode was measured during the stretching process of the e-skin. The tensile limitation measurement was taken at the end in the last part because e-skin microcracks appeared in the top electrode.
Peeling Force Test Process: The standard 90-peeling test was used to measure the peeling force of the prepared elastomer sample (15 × 10 × 0.2 mm) bonded to the human forearm. The measurements were taken using a tensile tester (ESM303, Mark-10). To record the peeling force, one end of the elastomer was clamped such that the elastomer was laminated to the forearm. The force applied was recorded using a dynamometer when the tester pulled the elastomer from the clamped end. It is worth mentioning that the top surface of the elastomer was attached to a thin Polyimide (100 µm thick, Figure S7, Supporting Information), which prevented elongation of the elastomer along the peeling direction. The peel energy was determined by integrating the peel force over the sample length.

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