Review of flexible microelectromechanical system sensors and devices

Inorganic semiconductors such as silicon, gallium arsenide, and gallium nitride are by far the most well-established routes to high-performance electronics/optoelectronics.¹ However, their rigid and brittle format suffers from disadvantages when used in intimately wearable or bio-integrated applications ranging from representative applications such as electronic eyeball cameras² to advanced surgical devices and “epidermal” electronic monitoring systems.³ There is significant interest in technologies that enable microelectromechanical system (MEMS) devices to sit directly on flexible foils or plastic sheets.^4–9 In addition, particular attention has been paid to electronic devices based on elastomeric substrates that are not only bendable but also stretchable and can thereby provide both versatility and functionality.^10–20 Such devices allow intimate, conformal, and seamless contact with complex structures and nonplanar surfaces, such as those found in health-monitoring systems,^21–25 that are subject to high strain levels. These characteristics cannot be realized using conventional rigid structures formed on silicon wafers or glass panels. Flexible electronic devices offer great potential for applications such as flexible displays, electronic textiles, sensory skin, artificial muscles, radio-frequency (RF) tags, optical MEMS, conformal X-ray imagers, and electronic-eye cameras.^26–35 

Various types of MEMS devices with small size and low power consumption have been developed to satisfy the requirements of current applications. However, with regard to flexibility, a number of problems arise, such as how to achieve sufficient flexibility and seamless integration into a complete flexible system. The main difficulty here lies in how high performance can be retained in flexible devices compared with traditional rigid devices, while also realizing new capabilities.

Investigations of flexible and stretchable electronics were originally motivated by their use in flexible displays,^36–41 but more recently attention has focused on applications to biomedical devices that are able to intimately integrate with the human body and on devices that exploit curvilinear, ergonomic, or bio-inspired layouts.⁴² Various options have been considered to provide flexibility for MEMS devices in such applications. One approach involves the use of new electronic materials, such as nanomaterials and organic materials that are intrinsically flexible^43–47 or composite elastomer conductors that are stretchable.^16,48 However, organic materials pose challenges such as degradation in device performance and incompatibility with previous fabrication processes such as deposition, crystallization, and doping. On the other hand, recent studies^10,12,42,49 have demonstrated that with suitable material selection, mechanical design, and integration strategies, traditional inorganic-material-based structures can be made mechanically compliant, avoiding the limitations of incompatibility and performance degradation. In this way, the advantages of MEMS devices in terms of their small size, high performance, and low energy consumption can be preserved in a flexible format for new applications in biomedical and wearable electronics.^50–52 

Recent advances in materials science and manufacturing have provided opportunities for the development of flexible electronics with unique potential for use in bio-integrated applications. Kim⁵³ reviewed flexible and stretchable electronics that could be integrated with the brain, skin, and heart to study disease states or establish human–machine interfaces. Kim et al.⁵⁴ described the manipulation and use of semiconductor nanomaterials for various components of mechanical devices and the integration of these devices into flexible plastic foils and stretchable rubber sheets. Suitable materials and their processing for stretchable electronics in bio-integrated surgical tools and bio-inspired cameras were discussed by Kim et al.¹ Yu et al.⁵⁵ presented some inorganic semiconductor nanomaterials, together with corresponding design strategies, for use in flexible electronics for biomedical applications. The mechanical properties and electronic performance of inorganic materials for flexible electronics, together with appropriate design strategies, were outlined by Kim et al.⁴² Ma et al.⁵⁶ reviewed the underlying ideas, fabrication/assembly routes, and microstructure–property relationships of stretchable electronics. However, to date, there have been no reviews focusing on flexible MEMS devices.

The present paper describes a variety of flexible MEMS devices based on inorganic materials. Section II highlights the key design and fabrication strategies of MEMS devices aimed at realizing flexibility, including transfer printing, device integration, and various interconnections according to optimized mechanical configurations. Section III describes typical flexible MEMS devices and their corresponding applications. Generally, the performance of these flexible MEMS devices can reach levels comparable to those of conventional wafer-based technologies, which exceeds the range accessible to known organic-material alternatives. Section IV highlights the key requirements for the further development of flexible MEMS devices. As pointed out in the concluding Sec. V, flexible MEMS sensors have a promising future for application across a range of fields, particularly in wearable devices, health-monitoring systems, and biomedical engineering.

MEMS thin-film structures are mechanically bendable. However, the substrate, which only plays a supporting role, is too rigid to bend. One direct method for achieving flexibility is to replace the rigid substrate with a flexible one. Transfer printing is a technology that transfers high-performance inorganic materials grown on primary substrates to other heterogeneous substrates,^57,58 thereby solving the problem of incompatible substrates. The materials or structures that are transferred in this manner are generally thin, ranging from a few hundred nanometers to a few micrometers, and thus are flexible. The transfer operation is generally accomplished using additional transfer stamps, which are generally made of a silica-gel material with small Young’s modulus, such as polydimethylsiloxane (PDMS). Under van der Waals interactions, the stamp picks up the released devices from the donor wafer and then transfers them to the receiver substrate to achieve device flexibility [Fig. 1(a)].⁵⁹ 

Acoustic MEMS devices such as film bulk acoustic resonators (FBARs)^60–63 and piezoelectric micromachined ultrasonic transducers (PMUTs),^64–72 require the space provided by a cavity to allow free vibration of the resonant body. A misalignment in the cavity or inadequate boundary constraints will result in poor performance or even failure to work properly. To facilitate transfer printing with adequate cavity space for high-performance flexible acoustic MEMS devices, Sun et al.^73,74 proposed the FlexMEMS technique. The key part of this technique is the transfer printing of the MEMS structure onto a flexible substrate [Fig. 1(b)]⁵⁸ with a predefined cavity, and accurate alignment of the structure and cavity is essential.^75,76 Experimental implementation of the FlexMEMS technique has shown that the performance of the flexible device is comparable to that of its counterparts on rigid substrates. For different substrates, both flexible Lamb wave resonators (LWRs) on polyimide (PI) and polyethylene terephthalate (PET) substrates [Fig. 1(c)] have been realized.⁷⁴ FlexMEMS-enabled devices have demonstrated excellent mechanical flexibility and electrical performance during and after bending trials.

Special release techniques have also been used in lieu of stamps to fabricate flexible devices. For example, the thermal-release transfer-printing technique was used to directly pick up a prepared three-dimensional (3D) PMUT and transfer-print it onto a mica substrate, thereby avoiding the problems encountered when picking up using a thermal-release tape without additional complex treatment. The PMUT device was used for curved surface imaging because of its excellent flexibility and high transparency [Fig. 1(d)].⁷⁷ 

Transfer printing can be used to fabricate a flexible-device array without additional processing steps. In a batch-fabrication method, the production efficiency can be greatly enhanced, and the cost of fabrication can be accordingly reduced.^78,79 Flexible film bulk acoustic-wave filters consisting of several FBARs were fabricated using this method.⁷⁸ In another example, a high-aspect-ratio ultrathin strain sensor was integrated with a flexible printed-circuit substrate [Fig. 1(e)].⁷⁹ Jiang et al.^80,81 successfully demonstrated large-area transfer printing of 3D high-aspect-ratio MEMS structures using the difference in adhesion between the donor and receptor [Fig. 1(f)]. Transfer printing of a device array can also be done in a step-by-step and repetitive manner. Using a transfer machine with high resolution and multidimensional position control, micrometer-level transfer accuracy and a large throughput per hour can also be achieved.

Integrating MEMS devices and sensors with flexible substrates⁸² can provide new capabilities and greatly enhance functionality. Direct integration of devices and flexible substrates is another effective method for realizing flexibility and stretchability of a system, offering the advantage of process compatibility between MEMS and a complementary metal oxide semiconductor (CMOS) or a polymer substrate. With this approach, commercially available MEMS devices can be used directly.⁸³ For example, the components of a system can be realized by micromachining, microelectronic technology, bio-/chemotechnology, or MEMS technology. Each technology provides different devices with specific functionality. Through integration, it is possible to combine components and create a micro/nanosystem with multiple functions.⁸⁴ The most commonly used method is heterogeneous integration,⁸⁴ in which MEMS devices are directly bonded to a flexible substrate or flexible wires to achieve overall system flexibility.

Yamada et al.⁸³ proposed a pseudo-system-on-chip (pseudo-SoC) integration by incorporating an electrostatic MEMS and its driver, which was a CMOS-level system, for mobile electronics applications [Fig. 2(a)]. This integrated system offers the complementary advantages of a system-in-a-package and a system-on-a-chip.^85–87 Jentoft et al.⁸⁸ demonstrated a sensor array that could grasp sensing events and detect object shapes. The sensor array was formed by direct integration of commercial barometers and commercially compatible flexible printed-circuit boards [Fig. 2(b)]. In addition, IC-integrated MEMS devices are highly desirable owing to their simultaneous elimination of the bias board and interconnection cables, which simplifies implementation procedures and improves system reliability. To avoid the need for a high-temperature deposition process in the CMOS post-treatment process, low-temperature deposition processes such as sputtering, spin coating, and electroplating were employed by Xu et al.³² to fabricate an IC-integrated flexible shear-stress sensor skin [Fig. 2(c)]. Through-hole vias and multilayer electrical routings were used to realize flexibility of the system proposed by Hsu et al.⁸⁹ Xiao et al.⁹⁰ used the flip-chip-on-flex technique to attach sensor and actuator chips to the same flexible substrate within a small space [Fig. 2(d)]. Overall system-level flexibility can be achieved using the above methods.

A number of studies have investigated the fabrication of flexible and stretchable devices by direct deposition of inorganic materials on a polymer substrate in a bottom–up process. For example, following the initial spinning of a polymer substrate on a silicon or glass substrate, active material was deposited on this polymer, and the sacrificial layer was finally released.^91,92 To fabricate a flexible magnetic sensor, Wu et al.⁹³ spun PI on a silicon substrate, onto which they deposited two layers of Ni electrodes by electron-beam evaporation and a 2-µm piezoelectric film of ZnO by direct current (DC) reaction magnetron sputtering [Fig. 3(a)]. The characteristics of this magnetic sensor were investigated both experimentally and theoretically. Sun et al.⁹⁴ deposited Mo (0.2 μm)/AlN(1.5 μm)/Mo(0.2 μm) stack layers onto an oxide layer and then used a lithographic process to pattern the device layer. After the PI layer had been coated, the AlN/PI module was released from the Si test wafer. This method avoided the problems that would have been caused by the high temperatures involved in direct deposition of an AlN film on the PI, such as stress-induced crimp. The AlN/PI film had a similar elastic modulus to polyvinylidene fluoride (PVDF), but greater fabrication-process compatibility [Fig. 3(b)]. Feng and Liu⁹⁵ fabricated microcavities on a flexible copper sheet by micromachining and then deposited onto this a lead zirconate titanate (PZT) slurry to produce a PZT micro-pillar array, which, after development, was released as a curved thin PZT layer. The transducer thus fabricated achieved a center frequency of 26 MHz and a −6 dB bandwidth of approximately 65%. Zhang et al.⁹⁶ used an HF/HCl escharotics system to pattern a 500 × 200 × 1 μm³ PZT cantilever beam. The backside of the wafer was dry-etched. Subsequently, each sensor formed a silicon island. The sensor array was made flexible by coating it with a polymer [Fig. 3(c)]. A flexible capacitive MUT (CMUT) was demonstrated on an ultrathin PI substrate that was spun on a silicon wafer from which it was then peeled off [Fig. 3(d)].⁹⁷ Zhou et al.⁹⁸ explored the possibility of directly integrating piezoelectric ZnO and AlN thin films on a spin-coated PI substrate to realize flexible acoustic-wave resonators. A variety of flexible MEMS devices have been fabricated using bottom-up processes.^99–101 In addition to the above-mentioned methods, flexibility can be realized by filling trenches between unit cells. In a demonstration by Caronti et al.,¹⁰² CMUT dies were embedded using a flexible backing process and coating layers, and the resulting array assembly could withstand a bending radius of less than 10 mm [Figs. 3(e) and 3(f)].

Coutts et al.¹⁰³ demonstrated an island–bridge structure by fabricating a rigid-flex Kapton (PI film) substrate using a wet-etching process. A MEMS-tunable conformal frequency-selective surface (FSS) was integrated onto this substrate [Fig. 4(a)]. The use of Kapton instead of silicon for the islands offered the advantage of easier processing. A method for fabricating CMUT arrays was proposed by Chong et al.,¹⁰⁴ who used a novel rivet structure to fasten isolated metal islands to a flexible polymer film. The rivet structure securing the metal islands was formed by nickel overplating, which solved the problem of achieving adhesion between the metal and polymer. A photoresist was patterned to form via holes and served as the polymer film [Fig. 4(b)]. Dann et al.¹⁰⁵ also used an island–bridge technique to realize a flexible device, with a 12-mm-diameter ring-annular array mounted on top of an array of islands that withstood 10 000 bending cycles at an angle of 60° [Fig. 4(c)]. These types of integration methods offer advantages for the fabrication of flexible device systems. Multilayer functional systems such as those described here can be crucially important for cost-sensitive applications.

In brittle inorganic materials, one important route to flexibility and stretchability involves the use of specialized structural layouts and mechanical designs.⁴² One of the most intuitive approaches is to adopt special electrical connections to transform in-plane stresses into out-of-plane electrode changes. The use of special electrode interconnections combined with inorganic materials based on elastomeric substrates has shown good strain tolerance. Because the strain and stress caused by bending and stretching are borne mainly by the flexible interconnections on the substrate, a wide range of tensile properties becomes possible.

Figure 5(a) shows a device system that employs this strategy. Thin wavy metal lines directly deposited and patterned on the PI provide interconnections between the separately fabricated and assembled cells. By initially coating PI onto Cu sheets and then baking them on a hotplate, five separate pieces of Cu sheets were then patterned in an island–bridge-structured geometry using pulsed-laser ablation. PDMS served as a temporary substrate, onto which the PI/Cu sheet was laminated for laser ablation. The five layers were aligned to the previous layer of electrodes after assembly of the transducer arrays. For relatively small or negligible strain in the system, it could be designed to withstand different stretchability and twisting actions using a related strategy.¹⁰⁶ In a similar manner, high-performance composite material and flexible interconnectors were used to form a conformal ultrasonic device that could monitor a central blood-pressure waveform [Fig. 5(b)].¹⁰⁷ Figure 5(c) shows a soft microfluidic assembly of sensors, circuits, and radios for the skin fabricated using this strategy.¹⁰⁸ 

The devices described in the preceding paragraph were based an “S” shaped geometry. In another approach, a serpentine geometry has been adopted. Figure 5(d) shows an example of the use of serpentine-shaped interconnectors to provide device flexibility. The manufacturing process was compatible with the MEMS technology, and the serpentine interconnection carried most of the strain during the stretching process to ensure normal operation of the whole system.¹⁰⁹ In addition to the stretchable system shown Fig. 5(d), other flexible devices were also fabricated using serpentine structures.¹¹⁰ In addition to the above examples, there have been a number of other studies of interconnection structure, for example, mechanical structure and optimization,^{107,111–122} aimed at realizing flexibility and stretchability of inorganic devices such as MEMS.

In addition to the methods described so far, origami and kirigami structures^123–125 have been extensively explored to provide flexibility and stretchability of devices. Figure 6(a) shows a miniature 50-mm³ folded inertial measurement unit (IMU) fabricated using a MEMS process. First, wafer-level and high-aspect-ratio single-axis sensors were fabricated and then interconnected using flexible hinges, forming a 3D configuration similar to a silicon origami. The process for a double-sided folded IMU was initiated using deep reactive-ion etching (DRIE) of 500-µm-deep blind via holes patterned on the handle side of a silicon-on-insulator (SOI) wafer. Then, flexible hinges and interconnects were fabricated. Finally, the sensor was defined through top-side processing and wafer etching using SiO₂ as a hard mask [Figs. 6(a) and 6(b)].^126,127 Figure 6(c) shows an origami-structured flexible electronic substrate whose faces are parallel to the target of the attachment surface. The parallel faces are important from an engineering point of view in that they make direct contact with the surfaces of the target of attachment and mounted electronic elements such as sensors or light emitters. These characteristics allow the realization of flexible devices that can detect a shear force or flow velocity parallel to the object surfaces.¹²⁸ Using a kirigami structure, Yamamoto et al.¹²⁹ designed an all-printed planar-type multifunctional health-monitoring patch based on an in-plane PET film integrated with an acceleration sensor for motion detection and with skin temperature and electrocardiogram sensors [Fig. 6(d)].

Controlled buckling of semiconductor nanoribbons is another general approach for obtaining stretchable electronic devices. First, an elastomeric substrate is mechanically prestretched. Then, nanoribbons are defined on it. Release induces compressive forces on the ribbons, which leads to “wavy” deformation (Fig. 7). The buckling amplitudes can be controlled using the prestrain level. The tensile properties of the device are positively correlated with the amplitude of fluctuation.¹³⁰ There have been a number of theoretical analyses of this strategy and investigations of the appropriate materials and their responses to applied strain.^42,131–133 More sophisticated systems that contain integrated circuits, MEMS devices, and microfluidics could be built on the basis of this approach. Since this strategy has been summarized elsewhere,¹³⁰ it will not be considered further in this paper.

Among the most powerful MEMS devices that can be constructed using the approaches described above are micromachined ultrasonic transducers.^134–144 Figure 8(a)^,144 shows a flexible PMUT (FPMUT) array integrated on PDMS for brain ultrasound stimulation. Thinned bulk PZT, of thickness 55 μm, was bonded to a silicon wafer. Then, a flexible ultrasonic transducer array was fabricated by integrating it with a flexible substrate, as described in Sec. II. The bending radius was less than 5 mm, thus allowing the device to be in contra-shape contact with a mouse brain. Good ferroelectric properties could be achieved using PZT thinning. By measuring the output sound pressure, the sound intensity was found to reach 44 mW/cm² at 80 V, which meets the requirements for ultrasonic neural stimulation. Researchers from KU Leuven designed and fabricated a bendable SOI-based PMUT array [Fig. 8(b)].^145,146 PMUT array islands were connected to one another by silicon springs using a MEMS procedure without any additional processing. The PMUT performance was not compromised by the fabrication on silicon islands. Each bendable array contained six islands with dimensions 3 × 3 mm², including 3 × 3 PMUT cells, in which each PMUT had a diameter of 410 µm and a thickness of 6 µm, with a 1-µm PZT layer as piezoelectric material. The PMUT array could withstand a bending radius of 1 mm. The PMUT elements in the array provided a displacement response of 2.1 µm/V and a Q factor of 104 at 426 kHz. Another novel PMUT was fabricated in which pre-etched PZT blocks were introduced into holes in a PI substrate [Fig. 8(c)].^147,148 Good conformal contact with skin surfaces and a suitable resonant frequency made this device suitable for heart imaging. The resonant frequency remained unchanged when the device was wrapped around a 1-cm diameter cylinder. This flexibly packed ultrasonic transducer also demonstrated good waterproof performance after hundreds of ultrasonic electric tests in water. Figure 8(d)^,149 shows a conformable and stretchable PMUT interconnector that was combined with a flexible substrate. The sandwiched structure of the bulk PZT array provided good ultrasonic penetration when tested on different pork tissues. The device had a resonant frequency of 321.15 kHz and a maximum sound pressure level (SPL) of 180.19 dB at a depth of 5 cm in water. The device was considered very promising for adjuvant treatment of bone injury. Figure 8(e) shows a typical example of an FPMUT fabricated using transfer printing of PMUT from a silicon donor wafer to a PI substrate, in a technique developed at Tianjin University.¹⁵⁰ The measured resonant frequency, effective coupling coefficient, and displacement sensitivity of this device were 2.58 MHz, 1.42%, and 30 nm/V, respectively.

Another important flexible MUT is the capacitive micromachined ultrasonic transducer (CMUT).^151–153 Figure 9(a)^,154 summarizes the progress made in research on these devices by the group at the National Kaohsiung University of Applied Sciences, culminating in the fabrication of a transparent flexible CMUT using roll-lamination technology. The fabrication temperature of this device was below 100 °C, which is much lower than that for a conventional silicon-based CMUT and reduces the stress unevenness caused by high temperature. The high visible-light transmittance (>80%) of this transducer and its ability to operate on 40-mm-radius curved surfaces make it of critical importance for use in curved displays. Indium tin oxide–PET (ITO–PET) served as substrate, with SU-8 photoresist forming a sidewall, and vibrating membranes and a silver nanowire serving as a transparent electrode. As shown in Fig. 9(b), Zhuang et al.¹⁵⁵ fabricated flexible CMUT arrays for side-looking intravascular ultrasound (IVUS) imaging in which, to achieve flexibility, deep trenches in a silicon substrate were refilled by PDMS. The trench width varied between 6 µm and 20 µm, and the trench depth was 150 µm. The transducer array could be attached to a 450-µm-diameter tip, which illustrated the potential for catheter applications (this is less than the catheter size). The 250 × 250 µm² transducer elements had a resonant frequency of 5.0–4.3 MHz and a capacitance of 2.29–2.67 pF. It was experimentally verified that the DC bias voltage ranged from 70 V to 100 V. Using a similar technique, Chen et al.¹⁵⁶ fabricated a curved row–column-structured CMUT array. The final flexible device exhibited a 4.5-MHz centre frequency with a −3 dB fractional bandwidth of 82%.

Surface acoustic-wave resonators,¹⁵⁷ bulk acoustic-wave resonators,^93,158–162 electrostatic actuators,¹⁶³ and MEMS LWRs^58,74 have been demonstrated. Jiang et al.⁷⁵ fabricated a 2.6-GHz air-gap-type thin-film piezoelectric MEMS resonator on a flexible PET film using the FlexMEMS technique. At a quality factor Q of 946, the performance was comparable to that of FBAR on a silicon substrate while being bent to 1 cm. As shown in Fig. 10(a),¹⁶⁴ after transfer of an ultraflexible FBAR from silicon onto a PI thin film, another 4.6-µm-thick PI thin film was installed as an encapsulation layer at the mechanical neutral plane. The flexible resonator exhibited a Q of 1108, an effective coupling coefficient of 5%, and a minimum bending radius of 0.5 mm.

Jiang et al.⁷⁸ constructed a flexible radio-frequency (RF) filter with a central frequency of 2.4 GHz based on film bulk acoustic-wave resonators [Fig. 10(b)]. The active material for each FBAR, which was a sandwiched structure with thin-film electrodes, was aluminum nitride thin film. The filter had a peak insertion loss of −1.14 dB, a −3 dB bandwidth of 107 MHz and a temperature coefficient of frequency of −27 ppm/°C. At the test state of 2.5-mm bending radius, the filter showed no significant performance degradation. This filter shows promise as an important component of future flexible wireless-communication systems.

LWRs^73,74 have also been demonstrated on a flexible substrate (PET) [Fig. 10(c)].⁵⁸ Using a PDMS stamp, three different LWR topologies were simultaneously transferred to a flexible substrate with air cavities using the FlexMEMS technique. The resonators exhibited similar performance to their counterparts on a silicon substrate and demonstrated excellent mechanical stability and flexibility under deformation. They had a quality factor of 1294 and a coupling coefficient of 2.2%. Jin et al.¹⁶⁵ deposited ZnO nanocrystals on a cheap and bendable plastic film [Fig. 10(d)]. The flexible surface acoustic wave (SAW) devices thus fabricated exhibited two wave modes: Rayleigh and Lamb waves with resonant frequencies of 198.1 MHz and 447.0 MHz, respectively. Particle concentration using SAWs was achieved with these devices.

Conventional force and strain sensors can be classified mainly as piezoresistive, capacitive, piezoelectric, and resonant types.^166–168 Flexible tactile and strain sensors have great potential for application in health monitoring and medicine, where they are required to make conformal contact with curved or nonconformal surfaces.¹⁶⁹ Shi et al.¹⁷⁰ constructed a flexible pressure sensor based on electroplated-Ni film, in which mechanical cracking is induced by stress concentration when the sensor is subjected to pressure, and this cracking in turn causes changes to the sensor resistance. This pressure sensor has a small response time, which is promising for wearable-device application such as pulse detection. Figure 11(a) shows a MEMS–CMOS integrated tactile sensor on a flexible and stretchable bus covering a social robot body.¹²⁰ In this configuration, the CMOS substrate was flip-bonded to a low-temperature co-fired ceramic (LTCC) substrate, which simplified the fabrication process. A flexible and stretchable wire was fabricated using metal etching and PI laser cutting. These tactile sensors could send coded digital signals under different forces.

A differential piezoresistive wind-speed sensor has been fabricated by mounting two piezoresistors with the same size on both surfaces of a flexible substrate [Fig. 11(b)].¹⁷¹ The substrate bends under the action of wind, resulting in changes in electrical resistance that depend on the wind speed. Experiments have shown that the sensor can measure wind speeds of up to approximately 20 m/s. Motivated by the need to measure the force and pressure on nonplanar surfaces for structural health monitoring, MEMS force sensors embedded in flexible PI substrates have been constructed, with an average piezoresistive gauge factor of 1.75 [Fig. 11(c)].¹⁷² The sensors have an average sensitivity of 1.25 V/N, with values ranging from 0.266 to 2.248 V/N. By using transfer printing, a fragile strain sensor has been integrated onto a flexible printed-circuit substrate. The output voltage of the manufactured sensor exhibited a linear dependence on strain.⁷⁹ High-performance wearable strain sensors based on a graphene/polymer composite film have been demonstrated for human-motion detection.¹⁷³ Other pressure sensors^100,174 [Fig. 11(d)] and tactile sensors¹⁷⁵ have also been explored for applications in health care,^176–179 in intracortical probe insertion mechanics,¹⁸⁰ and as shock sensors.⁹¹ 

With the use of metallic nanoparticles and nanowires, carbon nanotubes (CNTs), or graphene-based materials,¹⁸¹ the challenges posed by the fabrication of flexible high-stretchability strain sensors can be partially overcome. Channel nanocrack-based strain sensors (nCBSSs) and network nCBSSs have shown great potential.¹⁸² Zhou et al.¹⁸³ demonstrated strain sensors based on fragmented single-walled carbon nanotube (SWCNT) paper embedded in PDMS that retained sensitivity even at very high strain levels [Fig. 11(e)]. The sensing layers remained unaffected when a novel one-step laser encapsulation (OLE) method was used. Subsequently, nanocracking strain sensors were developed to detect motions around eyes, where oil and sweat are generated. Figure 11(f) ¹⁸⁴ presents a schematic of nanocracking strain sensor fabrication and encapsulation. Nanocracking strain sensors also show great potential as audio analog-to-digital converters for acoustic signature recognition and for other large motions of the human body.

In addition to the flexible MEMS devices described above, a number of other flexible MEMS devices have been presented. These include RF MEMS switches.^185–187 Researchers from Nanyang Technological University transferred RF MEMS switches onto flexible printed-circuit boards using thermal compressive bonding, mechanical grinding, and wet removal of silicon [Fig. 12(a)].¹⁸⁵ These switches demonstrated low insertion loss (≤0.15 dB at 20 GHz) and high> isolation (∼21 dB at 20 GHz), which illustrates the potential for integration of RF MEMS components with other RF devices on organic substrates.^185,187

In addition to RF MEMS switches, electrostatic actuators on a flexible substrate were demonstrated by Kim et al.,¹⁸⁸ who presented a MEMS system on a flexible substrate interfaced with integrated driver circuitry [Fig. 12(b)]. A suspended diaphragm, made of parylene, served as a passivation layer to generate an acoustic pressure wave perpendicular to the substrate. A single actuator had an SPL of 27 dB at 25 kHz, and a 1 × 2 array emitted up to 34.6 dB SPL at 1-cm distance. A 4 × 4 flexible micro-actuator array with a resolution of 2 mm was presented by Streque et al.¹⁸⁹ [Fig. 12(c)]. Each actuator was made of a PDMS elastomeric membrane, which was magnetically actuated by coil–magnet interaction. This technique has potential for application in tactile displays and offers great advantages in terms of comfort in use.

Microcantilevers could be used as flexible stethoscope transducers.⁹⁶ A PZT thin film was deposited and patterned on silicon nitride, which served as the supporting layer, using the sol–gel technique. Polymer was coated on both sides of a 500 × 200 × 1 µm³ PZT cantilever beam to ensure flexibility of the sensor array. A PDMS-based electrostatically actuated MEMS cantilever beam was presented by Singh and Patrikar.¹⁹⁰ 

Using surface micromachining, three-axis accelerometers were realized on Si and flexible PI substrates [Fig. 12(d)].¹⁰¹ Testing revealed high sensitivities ranging from 22 fF/g to 27 fF/g in the z axis and from 12 fF/g to 18 fF/g in the lateral axes. Wafer-level packaged MEMS accelerometers between two PI layers were fabricated by Mahmood and Çelik-Butler.¹⁹¹ The devices could be bent up to a 1.0-cm radius and were compatible with robotic or prosthetic fingertips. In addition to these devices, flexible MEMS accelerometers for failure assessment in aerospace systems⁹⁹ and packaged flexible and bendable micro-accelerometers¹⁸⁷ have been demonstrated.

In the area of power generation, the availability of MEMS has promoted the development of flexible triboelectric nanogenerators (TENGs) that convert mechanical energy into electricity.¹⁹² A TENG based on contact electrification coupled with electrostatic induction between two media for energy conversion from dynamic stimuli was reported by Fan et al.¹⁹³ A wearable and washable textile-TENG was demonstrated using a synergetic triboelectric trapping layer of black phosphorus protected by cellulose-derived hydrophobic nanoparticles [Fig. 12(e)].¹⁹⁴ To widen the range of applications, attempts have been made to improve the performance of TENGs in various ways, such as modification of surface materials, introduction of micro/nanostructures, and fabrication of novel device structures.

The following criteria should be satisfied by a flexible MEMS device and its fabrication method:

1. To provide intimate contact with a complex surface, the device should be as bendable as possible and should preferably also have a certain level of stretchability.

2. The device should have stable and reliable performance under bending or stretching conditions.

3. Its fabrication process should preferably be compatible with the MEMS process to reduce cost, achieve high yield, and maintain high production throughput.

The transfer-printing technique transfers a MEMS structure from a donor wafer to a target polymer substrate. Thin-film MEMS structures thus produced can achieve high bending capability. In addition, transfer printing does not have strict requirements on the thermal budget during the manufacturing process, and therefore is compatible with the MEMS process. The use of van der Waals forces and a high-precision alignment machine allows low-cost production. However, a large-area flexible array must satisfy stringent requirements in terms of the alignment precision of the transfer machine. The disadvantage is that transfer printing suffers from difficulties in satisfying robust boundary constraints. In the case of bending and stretching, device performance is significantly affected by the strength of the boundary constraints, and inappropriate constraints can lead to device failure.

Rigid devices can be integrated directly into flexible substrates. Although the minimum bending radius is limited by the size of the rigid component, and the conformal contact with a complex surface is poorer, the device performance usually remains good. Simultaneous integration of different devices is also possible for a wide range of applications. By contrast, when MEMS structures are directly constructed on a flexible substrate, the quality of the films and MEMS structures on the substrate is limited because of process incompatibility, and therefore the device performance is degraded.

In the strategy of shaped electrical connection, specially shaped electrodes with functional layers are connected to a device, with system deformation being provided by a patterned electrode. This strategy facilitates integration of devices in 3D space. The resulting systems can bear large deformation and provide conformal contact with complex surfaces. However, the fabrication process is complex, and therefore achieving high process yield and throughput is difficult.

The crucial challenges for realizing flexible and stretchable MEMS devices are as follows:

1. MEMS structures must be able to move or be coupled with the external environment during operation. Therefore, appropriate boundary-constraint conditions should be met. For example, cavities in a flexible substrate and seamless bonding between the MEMS structure and substrate are very important. In particular, in the case of a large-area device array, alignment between structures and cavities seriously affects the device performance. A large misalignment can lead to serious degradation or even device failure.

2. In a flexible MEMS array, cross talk among elements cannot be neglected. For example, in imaging applications, there is significant coupling among the ultrasonic transducers in the array, whereas this does not usually need to be considered in conventional silicon-based PMUT arrays. Strong coupling significantly affects key resolution specifications.^195,196 MUTs based on a polymer substrate typically have a much lower output pressure than those based on a silicon substrate. For operation at a given resonance, a polymer membrane must be much smaller than a silicon membrane, because of its low flexural rigidity.

3. MEMS devices cannot operate properly without suitable and reliable packaging. MEMS devices contain movable and fragile structures, and therefore their packaging is different from that of conventional integrated circuits.¹⁹⁷ Wafer-level capping and hermetic/vacuum sealing are often required in MEMS devices. Reliable packaging of MEMS devices while maintaining mechanical flexibility is challenging. In addition, connection of MEMS devices to external circuits is often necessary. For flexible MEMS devices, achieving connections with low electrical loss and high mechanical robustness is a challenging task.

The work summarized in this paper has provided convincing proof that high-performance flexible and stretchable MEMS devices can be constructed using appropriate strategies. Fabrication processes include transfer printing, shaped electrical connection, and heterogeneous integration using diverse classes of inorganic material-based MEMS structures and flexible substrates (plastics, elastomers, and others). The flexible MEMS devices reviewed here include, among others, MUTs, pressure sensors, strain sensors, and resonators. Various disciplines, ranging from mechanical engineering, electrical engineering, and chemistry to materials science, have been involved in efforts to realize flexible MEMS sensors and systems.

From an engineering perspective, future flexible systems will be greatly influenced by the availability of flexible and stretchable MEMS devices. In biomedical applications, the low mechanical modulus of flexible MEMS devices will allow them to match shape with biological tissues. Moreover, system integration of flexible MEMS devices will contribute to future consumer electronics and the rapid development of the Internet of Things. In our view, the fundamental challenges posed by the development of flexible systems, and their wide scope of application, together with the considerable social implications of their use, provide strong motivation for continued and expanded research efforts in this field.⁴² 

This work was supported by the National Natural Science Foundation of China (No. 62001322), the Tianjin Municipal Science and Technology Project (No. 20JCQNJC011200), the National Key Research and Development Program of China (No. 2020YFB2008801), and the Nanchang Institute for Microtechnology of Tianjin University.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

 Xiaopeng Yang received the M.S. degree from Taiyuan University of Technology in 2018, majoring in Mechanical Engineering. He is currently pursuing a Ph.D. degree at Tianjin University. His research focuses mainly on flexible MEMS for wearable devices.

 Menglun Zhang received his B.S. and Ph.D. degrees from Tianjin University, China. Since 2016, he has been an assistant professor at the School of Precision Instruments and Optoelectronics Engineering, and at the State Key Laboratory of Precision Measuring Technology and Instruments, Tianjin University. He was a visiting scholar in the Bioelectronics Group at the University of Cambridge. His research focuses on piezoelectric MEMS sensors and actuators.