Design Strategies for Strain‐Insensitive Wearable Healthcare Sensors and Perspective Based on the Seebeck Coefficient

Large healthcare markets have been created in highly developed economies to improve the quality of life. Wearable healthcare sensors are attracting considerable interest because of their 24 h real‐time monitoring capability, which make them useful in the detection of potential diseases. To guide the diagnosis, these sensors are designed to monitor various physical (e.g., pressure, temperature, strain, touch, bioelectricity, etc...) or chemical (e.g., glucose, oxygen, bacteria, viruses, proteins, etc...) quantities. In order to be comfortable to wear for a longer period of time, the sensors must be made with good stretchability to conformably deform with human organs. However, high stretchability always brings the problem that the measurement is very often polluted by the deformation of the substrate, making the data unreliable. According to each the sensor mechanism, multiple strain‐insensitive design strategies compatible with large deformations of the human body are discussed and the performance of these strategies are comprehensively analyzed. Then, how the intrinsic strain insensitivity of the Seebeck coefficient of nanomaterial percolation networks can define an alternative promising strategy is demostrated. Finally, the outlooks for future research and challenges in realizing strain‐insensitive sensors by applying the Seebeck effect are reported.


www.advelectronicmat.de
In this perspective, we would like to present a novel technique that, rather than relying on advanced structural design, helps use the Seebeck effect to create strain-insensitive sensors. First, we discuss extensively applied strategies to achieve straininsensitive structural designs and explain their advantages and disadvantages. Then, based on our previous work on the design of stretchable temperature sensors with high strain insensitivity, we demonstrate a novel method of using the Seebeck effect of percolated networks to create strain-insensitive sensors. [143] The physical mechanism for the Seebeck coefficient of percolated networks' inherent strain insensitivity is described. [144] Finally, we provide our perspective on this Seebeck effect-based straininsensitive sensor design and discuss future challenges.

Strategies for Strain-Insensitive Sensors
Wearable health sensors must be designed as per certain principles, such as ultra-thinness, lightweightness, high flexibility, and high stretchability, to ensure their compatibility with human skin or organs. Strain insensitivity is an important design requirement for wearable sensors because induced deformation should not affect the sensor performance. We herein discuss the influence of strain on wearable health sensors as per their working mechanisms. Then, for achieving strain insensitivity, the most commonly used structural designs are presented.

Influence of Strain on Physical Sensors
Mostly, strain sensors work by monitoring conductance or capacitance; however, strain outside the working direction disturbs conductance-based strain sensors by changing the section of the sensing layer. The distance between these two electrodes in capacitance-based strain sensors is easily affected by strainrelated noise. Similarly, conductance-based temperature sensors fabricated from temperature-sensitive materials, such as metals and composites, are sensitive to noise strain. Conductance in metal sensors can easily be perturbated by strain caused by a change in the cross-section, whereas that in composite sensors is affected when the deformation induced changes the percolation networks. Finally, temperature-based sensors based on pyroelectric materials are influenced by strain-induced changes in capacitance between the gate and semiconducting layer. The application of strain decreases the thickness of the insulating layer between the gate and the semiconducting channel because of the Poisson effect, hence perturbating capacitance. [145] Mostly, touch and tactile sensors are based on monitoring changes in capacitance. The capacitance of 2D and 3D capacitors is affected by multiple factors, including the effective distance between the electrodes, effective areas of the electrodes, and dielectric constant of the dielectric material. [146,147] The effective distances and areas of electrodes are closely related to geometric parameters, which are sensitive to strain.
Strain influences pressure sensors. Noise strain can affect the microstructure of the piezoelectric materials of piezoelectric pressure sensors, thereby distorting the output voltages used to determine pressure. Capacitance-based pressure sensors are affected by strain in a manner similar to touch and tactile sensors.

Influence of Strain on Chemical Sensors
Solid-state LiCl is rigid and lacks stretchability; first-generation humidity sensors based on conductance monitoring are not strain-insensitive but they cannot tolerate large strains. Moreover, humidity sensors based on LiCl-doped nanomaterials are affected by strain in the same manner as physical sensors based on nanomaterials comprising percolation networks. Capacitance-based humidity sensors using metal oxides (e.g., SnO 2 / MoS 2 ) as dielectric materials are rigid and cannot be stretched. Although porous Si has good stretchability, when used it is used as a dielectric material, noise strain can affect the output capacitance by changing the geometric properties of the capacitor.
The positive electrode of a Clark electrode-based gas sensor lacks stretchability because it is composed of a metal and a metal salt (e.g., Ag/AgCl). These sensors lack strain insensitivity as strain can disturb the stability of the solution and thus influence the gas diffusion from the solution to the electrode surface. Because of the hardness and fragility of their sensing materials, gas sensors using γ − Fe 2 O 3 /Fe 3 O 4 lack stretchability. Semiconductor-based gas sensors are influenced by strain in two manners. First, strain affects the resistance of the semiconducting layer by inducing elongation/compression induced by strain. Second, strain changes O − amount attached to the semiconducting layer surface, thus disturbing electron absorption and release. A Schottky diode-based gas sensor is composed of a nonstretchable metal and an N-type semiconducting material, and such sensors lack stretchability. Moreover, the catalysts in catalytic flameless combustion-type gas sensors are normally not stretchable. Even if the catalyst has excellent stretchability, deformation significantly affects temperature stability.
The ion-selective electrodes in ion sensors require a stable solution environment to detect ion concentration. The reference electrode is designed having two bridges to prevent the shaking of the solution. Thus, any deformation of ion-selective membrane can influence ion migration, thus altering membrane potential on which the ion-selective electrode depends.
For biosensors based on highly selective materials, strain has a considerable effect on transistor geometry and affects the evolution of the I-V curve.

Structural Design Toward Strain-Insensitivity
Structural design strategies are an important tool for creating wearable electronics that are stretchable and strain-insensitive. Such strategies can be classified based on whether the functional component is stretched or not. The first approach, including buckling, origami, and kirigami, involves soft functional materials that can deform under strain. In the second method, the functional components are placed in isolated rigid islands that are connected by stretchable bridges; these bridges can have serpentine, spiral, arc, and helix shapes.

Deformable Functional Components
Deformable functional components are essential for wearable electronics to perform their intended functions made of www.advelectronicmat.de flexible/stretchable materials. Even with a thin enough thickness, rigid materials can be used to create deformable functional components. To increase stretchability and limit strain effects, the most common techniques, such as bucking, origami, and kirigami, use in/out-of-plane bending rather than in-plane stretching. Bucking, origami, and kirigami fold the functional layer with different levels in the initial stage. Under stretching, the folded layer is expended, during which process the principal strain of the functional layer burdened is bending.
Buckling structures or waves/wrinkles resemble waves on a water surface; they transform the stretching-induced in-plane strain in out-of-plane bending. A buckling structure is obtained by coating a functional thin film on a prestrained substrate and then releasing the pre-strain caused by stretching or thermal expansion. [148][149][150][151] In this method, the functional material requires to have a large modulus and be thinner compared to the substrate. As shown in Figure 2a, a stretchable temperature sensor based on a buckling structure offers stable performance at strains of 0-10%. However, a buckling structural design only partially solves the strain-dependence problem because it cannot provide strain insensitivity out of the prestrained direction. To solve this limitation, Hyun et al. [152] formed a 2D buckling structure that releases a pair of orthogonal strains ( Figure 2b). The morphology of the 2D buckling structure is dependent on the release sequence of orthogonal strains; isotropic (zigzag) buckling is generated when strains are simultaneously released (sequentially).
Origami, an ancient art of paper folding, is adopted to transform a 2D planar sheet in a compressive or stretchable 3D structure having predefined hinge crease patterns (Figure 2c). Origami structures have been used in multiple applications to bear large deformations, including Li-ion batteries, [157,158] solar cells, [159] photodetectors, [153] and thermoelectric generators. [160] In morphological terms, origami is a conceptual Figure 2. a) Stretchable temperature sensor on an elastomeric substrate and the influence of strain on sensor performance. Reproduced with permission. [148] Copyright 2009, American Institute of Physics. b) Isotropic buckling and zigzag buckling generated by simultaneous and sequential release, respectively, of two orthogonal strains. Reproduced with permission. [152] Copyright 2011, Wiley-VCH. c) Classic origami style for achieving high stretchability. Reproduced with permission. [153] Copyright 2017, American Chemical Society. d) Bent and twisted origami paper. e) Stretched and released kirigami paper. Reproduced with permission. [154] Copyright 2018, Elsevier. f) Kirigami paper absorbing stretching by the rotation of rigid islands. Reproduced with permission. [155] Copyright 2014, United States National Academy of Sciences. g) Compressible kirigami design with cut units at different compressive strains. Reproduced with permission. [156] Copyright 2017, Elsevier. www.advelectronicmat.de extension of 1D or 2D buckling, which transforms in-plane strain to out-of-plane bending but offers additional design possibilities. Traditional manufacturing methods, whether manual or machine-assisted, cannot assemble origami structures on a small scale for advanced materials. Thus, self-folding methods based on active stimuli-responsive hinges, [161][162][163] capillary forces, [164,165] or residual stresses [161,166,167] have been developed to develop small-scale origami structures. Moreover, many types of origami, such as Miura-ori, Yoshimura, diagonal, zigzag, square-twist, diamond, waterbomb, and eggbox, have been proposed. [168][169][170][171][172] The most extensively used origami structure for stretchable electronics is the Miura-ori pattern, which realizes large deformation based on in-plane deformation and out-of-plane bending and twisting (Figure 2d).
Similar to origami, kirigami is an art of paper folding and cutting that has recently been applied for designing stretchable electronics. [140,141,173] Figure 2e shows a classic kirigami pattern comprising stripes and slits. When the paper is stretched, the slits open, enabling the strips to bend and twist out-of-plane. The transformation of in-plane stretching into out-of-plane bending and twisting yields thin-film structures high stretchability while eliminating the effect of in-plane strain. [174] Furthermore, kirigami designs based on in-plane rotation have been introduced to wearable electronics design to make outof-plane bends and twists conform to substrates. [154] As shown in Figure 2f, a thin film is cut in a rigid "island" connected by rotatable "hinges." When the film is stretched, the rigid islands shift and rotate around the hinges, and the only parts that deform are hinges. Unlike origami, which enables a material to be stretched and compressed, kirigami only allows a pattern to be stretched; it lacks compressibility too. Therefore as shown in Figure 2g, circular cutouts that enable 50% compressibility have been introduced. [

Rigid Functional Components
The island-bridge design is a most commonly used structure in stretchable electronics; it offers stretchability and strain insensitivity because of mechanical isolation. A functional part is assembled on rigid islands to avoid the influence of strain, and these islands are connected by conductive bridges, which absorb the strain. Different bridge patterns, such as serpentine, self-similar, spiral, arc-shaped, non-coplanar serpentine, and helix patterns, have been designed to absorb strain.
The serpentine which is composed of periodically distributed units connected in series, was the first bridge pattern developed. Each unit comprises two semicircles connected by straight lines, as shown in Figure 3a. During stretching caused by applied strain, serpentine units can rotate or exhibit in-plane bending and out-of-plane buckling to reduce stretching tension and thus promote strain insensitivity. Moreover, serpentine patterns can be fabricated by lithography, [177] inkjet printing, [178] and laser ablation. [122] Figure 3. a) Structure of planar serpentine interconnects for stretchable electronics. Reproduced with permission. [120] Copyright 2016, Elsevier. b) Self-similarity strategy applied to planar serpentine interconnects. Reproduced with permission. [176] Copyright 2014, Elsevier. c) 2D silicon spiral and its application to island-bridge design at different stretching levels. Reproduced with permission. [125] Copyright 2014, American Institute of Physics.

www.advelectronicmat.de
A self-similar pattern was introduced to obviously improve the stretchability of the island-bridge design. [176,179,180] This pattern contains subunits having structures similar to those of a central unit in the unit cells (Figure 3b). [176] Here, we present a self-similar serpentine pattern as an example of this concept. The first-order serpentine pattern is based on the initial serpentine pattern; however, it changes the semicircle to a straight line and straight line to two serpentine units. The nth-order selfsimilar serpentine pattern is modified from the (n − 1)th-order pattern and follows the aforementioned changing principle. Another bridge pattern that can achieve high stretchability and strain insensitivity is the 2D spiral, which is characterized by a curve winding around a fixed point (Figure 3c). [181,182] The 2D spiral unwinds and straightens under stretching, and the spiral arm absorbs the stretching strain and ensures the islands are free of strain. [125] During the unwinding process, the spiral is only loaded by bending strain, which has a limited influence on the conductance of the bridge. The serpentine structure is combined with a 2D spiral to enhance stretchability and strain insensitivity. [138] Similar to buckling structures, freestanding or partly bonded arc-shaped bridges are formed by releasing a prestrained substrate to which rigid islands are strongly bonded (Figure 4a-c). [130,131] An arc-shaped bridge can be naked or protected via embedding in an elastomer such as PDMS. [133,184] Embedded arc-shaped bridges have slightly less stretchability compared to naked bridges, and they have weak strain insensitivity because of an encapsulation constraint.
The helix, a classical structure used in telephone cables to achieve stretchability, has recently been applied to soft electronics. [134,183,188] When stretched or compressed, helices can absorb the applied strain by twisting and bending, which exert only a small influence on electric properties (Figure 4g-i). Similar to buckling and non-coplanar serpentine patterns, helices can be fabricated via pre-stretching and releasing with the precise control of bonding and freestanding points. [189] Multiple other fabrication methods, such as 3D additive printing, [135,190] direct laser writing, [191] manual winding, [192] microcontact printing, [193] mold processing, [194,195] and chemical synthesis, have been proposed. [196,197]

Stability of Seebeck Coefficient of Nanomaterials-Based Percolation Networks under Stretching
In previous sections, we discuss the sensing mechanisms of multiple wearable healthcare sensors and review strategies available in the literature for achieving stretchability by eliminating the influence of strain. However, certain strategies require complex fabrication processes, whereas some are only suitable Figure 4. a-c) Fabrication process and images of an arc-shaped bridge on a pre-stretched PDMS substrate. d-f) Non-coplanar serpentines and the different morphologies on various types of pre-stretching. Reproduced with permission. [133] Copyright 2008, United States National Academy of Sciences. g-i) The linear helices used for stretchable electronics. Reproduced with permission. [183] Copyright 2018, American Association for the Advancement of Science.

www.advelectronicmat.de
for specific applications. Wearable healthcare sensors based on an inherent strain-insensitive working mechanism are necessary. We present here a path introduced recently, leveraging the properties of the Seebeck effect in nanoparticle networks.

Seebeck Effect
The Seebeck effect refers to a temperature gradient-induced migration of charge carriers (e.g., electrons and holes) in a material that causes a voltage potential. Physically, charge carriers on the hot side have higher potential energy; therefore, they tend to migrate to the cold side, thus resulting in a difference between the charge carrier concentrations on cold and hot sides. The relation between the Seebeck voltage V C , Seebeck coefficient α, and temperature difference ΔT can then be expressed as follows: The Seebeck coefficient α is the open-circuit voltage across a material under a per-centigrade temperature gradient. It is an intrinsic material property that is related to temperature; however, it can be considered to be constant in a narrow temperature band. [198][199][200][201] Multiple factors such as conductivity, carrier concentration, and Fermi energy affect the Seebeck coefficient, for example, the Seebeck coefficient of organic materials can be described as follows [202] 1 d where σ is the conductivity, K B is Boltzmann's constant, e is the elementary charge, E F is the Fermi energy, T is the temperature, and f is the Fermi function. The function σ E is related to the energy E, temperature T, density of states (DOS), carrier velocity, and relaxation time. [202] The Seebeck voltage V C linearly changes with the temperature difference in a narrow temperature range. Based on this phenomenon, thermocouples having high linearity and sensitivity (0.5 °C) have been fabricated by welding two metal cables with different Seebeck coefficients. [203][204][205] However, metal-based thermocouples lack stretchability. Our previous work shows that the Seebeck coefficient of percolated nanowires/nanotubes is intrinsically insensitive to strain. [143] Based on this phenomenon, we designed a strain-insensitive temperature sensor with high stretchability (>50% strain). Then, we examined the physical mechanism of the strain insensitivity of the Seebeck coefficient of percolated nanowires/nanotubes in terms of electron tunneling and a conductance-weighted model. Approaches that adopt the Seebeck effect have a considerable potential as a simple, effective method of designing strain-insensitive sensors. We describe below the fundamental results of our previous works in this direction.

Previous Works
In our previous studies, we designed temperature sensors based on AgNWs, SWCNTs, and MWCNTs. [143] We printed two different nanoparticles (AgNWs/SWCNTs, AgNWs/MWCNTs, or SWCNTs/MWCNTs) in a "U"-shape on a soft substrate to form a nanomaterial-based thermocouple (Figure 5a). The "U"shaped printed circuits generated a Seebeck voltage in a nonhomogeneous temperature field because AgNWs, SWCNTs, and MWCNTs had different Seebeck coefficients. This temperature sensor demonstrated high linearity, strain insensitivity, and consistency between different samples. Figure 5b,c shows the relation between the Seebeck voltage and applied temperature difference. The Seebeck voltage linearly changes with the applied temperature difference and had a goodness of fit of >0.98. Then, we tested the temperature sensors at different strains from 0% to 40%, and the results are shown in Figure 5b. The Seebeck voltage-temperature difference curves at various strains almost overlap, which indicates that strain did not influence sensor performance. This strain insensitivity could be attributed to the stability of the Seebeck coefficient of the percolation networks. To examine this stability, we compared the performance of multiple samples (Figure 5c). No notable differences were observed, which indicates that the Seebeck coefficient of the percolation networks was insensitive to the nanomaterial coverage. We successfully applied this temperature sensor to an artificial skin on a prosthesis to monitor the temperature field activated by finger touching or ice cube contact ( Figure 5d). As shown by our study, the Seebeck coefficient of percolated networks has an intrinsic strain insensitivity, which we used to design a stretchable temperature sensor; [143] however, the mechanism of the stability of the Seebeck coefficient under strain was unclear. To examine this mechanism, we developed a theoretical model that combines percolation theory and electron tunneling theory. [144] First, we simplified random percolated networks in a series-connected model (Figure 6a,b). This simplification was based on the equivalent conductance of the percolated networks before and after simplification, and its rationality has been reported in multiple studies. [206][207][208][209] Then, a conductance-weighted model was used as a bridge to relate the Seebeck coefficient to the conductance between the percolated networks and pathways. [210] The conductance-weighted model is expressed as follows where α and G are the Seebeck coefficients and conductances, respectively, of the percolation networks, respectively; α j and G j are the Seebeck coefficients and conductances, respectively, of the pathway.
Once α j and G j are known, α and G can be determined based on the conductance-weighted model (Equation (4)). The morphology of the percolation networks under strain was separated in three regimes, as shown in Figure 6c. To simplify the expression, we used three pathways to demonstrate each regime; each pathway comprised two nanotubes. In regime 1, all pathways were well percolated. In regime 2, certain pathways www.advelectronicmat.de formed disconnected nanomaterials; however, the other pathways were well percolated. In regime 3, all pathways formed disconnected nanomaterials.
The electrons in the well-connected nanoparticles could easily transfer through the junction. Thus, each well-connected pathway could be seen as a long nanowire (Figure 6d) with the same Seebeck coefficient as that of the material(α m ).
For two NWs separated by a gap, electrons could not easily transfer via the gap, inducing a zero current because of a higher energy barrier (Figure 6e). Therefore, we assumed that the Seebeck coefficient and the conductance of the gapped pathway were zero.
Our inferences were as follows: In regime 1, all pathways were percolated well and most junctions were in-contact (no-gap). Thus, the conductance of each pathway only slightly decreased because of the equivalent elongation by stretching. The Seebeck coefficient of the pathway was all equal to α m because there were no gap. As per the abovementioned analysis and the conductance-weighted model, the conductance of the nanoparticle network was high and only slightly sensitive to stretching; moreover, the Seebeck coefficients were constant and equal to those of nanoparticles.
In regime 2, the pathways were partially connected, and certain gaps began to appear, causing the conductances of the percolation networks to rapidly decrease. The gapped pathways did not have any visible influence on Seebeck coefficients. To verify this conclusion, we assume that the first h pathways were wellconnected with Seebeck coefficient α m , and h + 1 to n pathways are gapped with zero Seebeck coefficient. These assumptions were plugged in the conductance-weighted formula.
Finally, in regime 3, all pathways were disconnected because of the large applied strain. The conductances progressively dropped to zero. Because of the lack of well-connected Figure 5. a) Ink printing and transfer processes of printed temperature sensors. b) Outputs of two temperature sensors, namely, multiwalled carbon nanotubes (MWCNTs)/AgNWs and single-walled carbon nanotubes (SWCNTs)/AgNWs, at different strains; the strain has a negligible effect on the temperature-sensing. c) Change in voltage with respect to the temperature difference between various samples. The samples show no notable difference. d) Application of these temperature sensors as components of artificial skin and successful detection of touched fingers and ice cubes based on an abnormal temperature field.Reproduced with permission. [143] Copyright 2019, Royal Society of Chemistry.
www.advelectronicmat.de pathways, the Seebeck coefficients sharply decreased from their nominal values to zero.
The main outcome of this previous work is that, while the conductance of a network depends on the morphology and number of percolated paths, the Seebeck coefficient does not. Doing so, the Seebeck coefficient remains almost constant even when experiencing high stretching, as long as one percolated path is still active. That makes Seebeck physics based sensors promising candidates for strain-insensitive measurement systems.

Outlooks
The abovementioned analysis suggests that the Seebeck coefficient of percolation networks is stable at strains less than the threshold point and then decreases to zero at higher strains.
The high stability of the Seebeck coefficient of percolation networks under strain makes this microstructure potentially applicable to the design of strain-insensitive sensors. In addition to temperature sensors, a stable Seebeck coefficient makes percolation networks suitable for multiple chemical sensors, including gas, ion, humidity, and biomaterial sensors. As shown in Equation (3), the Seebeck coefficient of organic materials is affected by the Fermi energy E F , Fermi function f, DOS, carrier velocity, and relaxation time. All these factors are highly dependent on the microstructure of the nanomaterial. A stretchable chemical sensor with strain insensitivity can be designed by modifying the surface microstructure of a nanomaterial, for example, the Seebeck coefficient of CNTs with percolation networks increases with the oxygen density attributed to the CO bond. [208] This mechanism can be used to design strain-insensitive stretchable oxygen sensors whose the structure comprises two covered reference electrodes and one naked counter electrode, as shown in Figure 7. Due to the stretchable substrate's cover, the Seebeck coefficients (α 1 and α 2 ) of the Figure 6. a) Physical model of percolation networks. b) Serial model of the percolation networks after simplification. c) Three regimes assumed for the percolation networks under an applied strain. Relative conductance and Seebeck coefficient in d) a well-connected pathway and e) a disconnected pathway. Reproduced with permission. [144] Copyright 2022, Elsevier.
www.advelectronicmat.de two reference electrodes are well-known and unaltered. Based on the output voltage, it is able to calculate the temperature difference ΔT between the hot and cold sides. Once the output voltage between the counter electrode and one of the reference electrodes is known, the counter electrode's Seebeck coefficient (α 3 ), which is related to the target materials' concentration, can be calculated by Equation (3). Similarly, the nanomaterial can be modified with a specific functional group that can be bonded with the measured gas or H 2 O to design gas and humidity sensors. Ion sensors can be designed with a combination of ion-exchange polymers that selectively trap ions of interest as well as release other ions. [211][212][213] Multiple ion-exchange polymers have been developed, including cation-exchange polymers [214][215][216][217] and anion-exchange polymers. [218][219][220] Enzymes, antibodies, and aptamers with high specificity to biomaterials can be used to design Seebeck coefficient-based biosensors. However, several challenges must be addressed to realize the abovementioned applications, including robust doping/modification methods, highly selective doping/modification materials, and in-depth research on the physical mechanism of the Seebeck coefficient. Presented from this perspective, this study can be applied to stretchable electrodes in human healthcare and other fields of interest.

Conclusion
In this perspective, we summarize how strain affects the functionality of wearable sensors, including physical and chemical sensors. The developed strategies are then discussed, most of which require complicated fabrication to counteract the strain's influence. We discover that the Seebeck coefficient of percolated nanomaterials is extremely stable based on our prior research, which can be used to design strain-insensitive sensors. From a physical perspective, we are able to pinpoint the mechanism underlying this stability, which is dominated by the wellconnected pathway. Finally, we provide a prototype design that demonstrates how to use the Seebeck coefficient to create other varieties of stretchable sensors without the influence of strain.