A Review of Flexible Acceleration Sensors Based on Piezoelectric Materials: Performance Characterization, Parametric Analysis, Frontier Technologies, and Applications

Abstract

Acceleration sensors are tools for detecting acceleration and serve purposes like fault monitoring and behavior recognition. It is extensively employed in a variety of industries, including aerospace, artificial intelligence, biology, and many more. Among these, one of the major research hotspots and challenges is the development of low-energy, self-powered, miniature, mass-produced sensors. Due to its capacity to perceive human behavior and identify errors, the flexible acceleration sensor offers a distinct advantage in the use of flexible and miniaturized sensing systems. This review analyzes the current state of piezoelectric flexible acceleration sensors’ applications in the areas of sensitive materials, processing technology, and device structure and briefly summarizes the fundamental properties of these sensors. Additionally, it ends with a prognosis for the future growth of flexible piezoelectric acceleration sensors.

1. Introduction

Sensing technology is an important support in the field of modern information technology, and together with communication and computers, it can collect, transmit, and process external measurement signals [1,2]. An acceleration sensor is a device that senses acceleration, which allows the acquisition of various vibration signals to further perform functions such as behavior recognition and fault monitoring [3,4]. It has important applications in aerospace, power and energy, power equipment, health monitoring, and other fields [5]. Driven by the new generation of information and communication network technology, the industry and Internet will develop in a broader, deeper, and higher direction [6,7]. The manufacturing industry will rapidly shift to digitalization, networking, and intelligence [8], and perception is a key source of information for this process. Flexible perception is an important technological support for achieving multi-dimensional spatial freedom [9]. Flexible sensing can change the traditional physical form of information devices and systems to organically integrate information with people, objects, and environments [10]. The flexibility of acceleration sensors is therefore also an important upgrade for functions such as behavior recognition and fault monitoring. Acceleration sensors can be divided into capacitance types, resistance types, optical fiber types, inductance types, and electric vortex types [11,12]. Among them, piezoelectric material is used as the sensitive material for piezoelectric acceleration sensors. Piezoelectric materials can realize the mutual conversion between vibration signals and electrical signals to directly generate electrical signals in response to vibration signals without external power supplies, which can realize a zero-power consumption system design [13,14] and has a unique advantage in the application of flexible and miniaturized sensing systems.

In this review, the most recent advances in the manufacturing technology, device structure, and sensing material research of flexible piezoelectric acceleration sensors are summarized. It is also introduced simultaneously along with the performance parameters. The potential future of the flexible piezoelectric acceleration sensor is briefly discussed in Section 6.

2. Development, Parameters, and Relationship between Parameters of Flexible Piezoelectric Acceleration Sensors

2.1. Development

Flexible piezoelectric acceleration sensors are devices with both flexible and acceleration-sensing functions, and their core sensitive materials are piezoelectric materials. Flexible piezoelectric acceleration sensors have broad application areas in behavior recognition, robotics, and the Internet of Things (IoT), as shown in Figure 1 [15,16,17,18,19,20,21,22,23,24,25,26,27].

When compared to a standard acceleration sensor and a single flexible acceleration sensor, the flexible piezoelectric acceleration sensor is superior; it can be considered a combination and further development of both. It can achieve self-power supply, high stability, and high toughness based on the benefits of piezoelectric materials and lacks the awkwardness and low sensitivity of conventional accelerometers. Figure 2 shows the comparison of the above sensors and the common structure of the flexible piezoelectric acceleration sensor [28,29,30,31,32,33,34,35].

2.2. Parameters

The knowledge of basic parameters is critical for the selection of pressure sensors as every application scenario has its unique requirements for sensor characteristics [36]. The performance of flexible piezoelectric accelerometers is evaluated by the following parameters.

2.2.1. Frequency

The natural frequency is the frequency that causes the sensor to exhibit the resonance phenomenon. The external vibration frequency forces the sensor to generate spontaneous vibration, and when the vibration amplitude reaches its maximum, it is resonant. Resonance will generate heat in the sensor heat, as well as structural damage [37], so it is necessary to use the sensor below this frequency. The natural frequency (resonance frequency) is given by

$$\mathrm{f}=\frac{\sqrt{\frac{\mathrm{k}}{\mathrm{m}}}}{2\mathsf{\pi }}$$

(1)

where m is the mass of the sensor and k is the structural stiffness.

2.2.2. Sensitivity

Sensitivity is the most basic and important parameter of an accelerometer. The sensitivity of a piezoelectric sensor is defined as the ratio of electrical output to mechanical input. Higher measurement accuracy can be obtained by increasing the sensitivity. For piezoelectric vibration sensors, charge sensitivity can better characterize the properties.

It is known that the acceleration value is proportional to the square of the signal frequency for the same displacement, so the amplitude of acceleration signals in different frequency ranges can vary considerably and may even exceed the measurement range of the accelerometer, causing distortion. As can be seen, the sensitivity of the accelerometer needs to be fully estimated when measuring high-frequency signals and low-frequency signals. As a general principle, the higher the sensitivity, the smaller the measuring range, and conversely, the smaller the sensitivity, the larger the measuring range.

2.2.3. Pre-Tightening Torque

The preload force is the force that is added in advance to enhance the reliability and tightness of the connection before the relevant structure is subjected to the working load, and also to prevent gaps and relative slippage between the connected devices after being subjected to the load. The magnitude of the preload force has an important influence on the degree of stiffness of the part, and therefore the preload force is also one of the main factors affecting the resonant frequency [38].

If the pre-tightening force is small, the stiffness of the sensor will decline, resulting in a lower resonant frequency [39], which ultimately increases the product’s upper limit of frequency response and frequency response error. As the pre-tightening force gradually increases, the product stiffness, resonant frequency, and the upper limit of frequency response will increase, but the error will decrease. Of course, the increase in pre-tightening force is limited, and when it approaches the bearing limit of the part structure and sensitive elements, it will lead to the structure’s deformation or even damage, causing reductions in stiffness and frequency response limit and an increase in error [40].

2.2.4. Noise

Noise is the interference signal superimposed on the useful signal, which has a large interference with the analysis and utilization of the measured signal. The noise of the measurement system can be divided into two parts: internal noise and external noise. The signal interference outside the measurement system itself, such as the vibration of the external environment and the interference of the electromagnetic field, is the main source of the external noise. Additional shielding measures can be used to achieve the purpose of noise reduction for such signals.

Internal noise comes from the system itself, also known as inherent noise or local noise, and can only be reduced by designing reasonable parameters. Internal noise determines the resolution and limits the dynamic measurement range of the system. For example, low-frequency vibration signals are very weak and must be measured with a low-noise vibration sensor to ensure that the useful signal is not submerged by the background noise. The internal noise of piezoelectric accelerometers can be further divided into mechanical thermal noise, electrical thermal noise, and noise generated by charge amplifiers [39]. Mechanical thermal noise is related to the mass, spring coefficient, and mechanical resistance of the sensor inertial system. It can be reduced by increasing the mass and quality factors or reducing the resonant frequency. Electrical thermal noise is related to the loss factor of sensing materials, and studies have shown that materials with fewer impurities and defects are often used for low-noise devices [41]. In the study of sensors, 1/f noise and its thermal noise are mainly considered in the noise generated by charge amplifiers [42]. In addition, the sensor can sometimes be regarded as a capacitor and the presence of capacitors in charge amplifiers can also lead to the generation of KT/C noise [43], which cannot be ignored.

2.3. Relationship between Parameters

As shown in Figure 3a,b [44,45], the output voltage and sensitivity of the piezoelectric accelerometer are plotted against frequency, respectively. There is an inverse ratio between sensitivity and resonant frequency, and only at the expense of one parameter can the other be improved. If the sensor has a higher sensitivity, smaller changes in the measurement can be recognized. But a small interference in the measured object can lead to a significant change in the output of the sensor. When a small disturbance occurs, the signal is quickly amplified, which greatly affects the stability and detection range of the sensor. Therefore, by increasing the sensitivity, the accuracy of the detection can be improved, but this needs to be done within reasonable limits. Figure 3c [46] shows the sensitivity of the sensor measured at different preload torques. It can be seen that within a certain range of values, increasing the preload force appropriately improves the sensitivity of the sensor and its frequency response is smoother in the lower frequency bands, because if the preload force is appropriate, the contact stiffness and pre-strain of the element will improve, leading to an increase in the stress applied to the element, induced charge generation, and increased sensitivity. And once this threshold is exceeded, the sensitivity and frequency response performance of the system will decrease. This phenomenon may be due to the deformation of small bumps on the wafer surface beyond the elastic limit, plastic deformation, or cracking, which increases the contact area and reduces the contact stiffness, resulting in reduced sensitivity and poor stability of the frequency response [47]. In Figure 3d [48], the temperature gradually increases from 0 to 1000 °C and the frequency excitation increases from 200 to 600 Hz; the sensitivity of the sensor still fluctuates around a horizontal line, reflecting the good temperature stability of the sensor in a certain range. In Figure 3d, the temperature gradually increases from room temperature to 650 °C and the frequency excitation increases from 120 to 520 Hz; the sensitivity of the sensor still fluctuates around a horizontal line, reflecting the good temperature stability of the sensor in a certain range. Figure 3e [49] shows the output voltage noise spectral density of piezoelectric acceleration sensors. It can be seen that the main noise sources are the thermal noise of the bias resistor, and the input current and voltage noise of the operational amplifier at high frequencies [50]. Figure 3f [49] shows the output charge as a function of acceleration for the 600 Hz case. The figure shows that the charge sensitivity is 2.3 pC/g at both room temperature and 900 °C, with a deviation at 900 °C because more noise disturbances are introduced at this time compared to room temperature.

Table 1. Performance comparison of different types of piezoelectric acceleration sensors.

Types	Mold Area (mm × mm)	Frequency Range (kHz)	Sensitivity	References
Resistive type 1	5 × 5	13.3	0.545 mV/g	[51]
Resistive type 2	3.5 × 3.5	1.5	4 mV/g	[52]
Capacitive type	6.5 × 5	6	/	[53]
PZT	6 × 6	17.4	14 mV/g	[54]
AlN	6.8 × 6.8	1.1	400 mV/g	[55]
AlN_C500	2.3 × 2.3	12.5	355 mV/g	[50]
AlN_S200	2.3 × 2.3	42.8	176 mV/g	[50]
YCOB	/	0.6	2.4 pC/g	[48]
BTS	/	3	2.6 pC/g	[47]
N/A	/	6	3.5 pC/g	[56]

shows the performance comparison of different types of piezoelectric acceleration sensors. It should be pointed out that due to the different designs and use ranges of different acceleration sensors, their sensitivity units are also different. Obviously, the input of the acceleration sensor is acceleration, which is generally expressed by the gravitational acceleration g. If the output of the sensor is a voltage change, such as mV, then the unit of sensitivity is mV/g; if the output is a change in the amount of charge (e.g., pC), the sensitivity is measured in pC/g.

3. Materials for Flexible Piezoelectric Acceleration Sensors

Flexible substrates, electrodes, functional components, and packaging materials make up the majority of flexible electronic devices [57]. Flexible piezoelectric acceleration sensors are consistent with it, and its performance and use are typically impacted by its piezoelectric material, substrate material, preparation circumstances, and preparation procedure. The core of the smart age is made up of piezoelectric materials, which perform both active (producing sound waves and moving objects) and passive (sensing) activities. Piezoelectric materials are frequently employed to capture outside signals and are a crucial component of the intelligent era [58]. The foundation substance can support the adaptability of sensor components and further safeguard the upkeep and regular operation of the sensor form. To make the piezoelectric sensor exhibit a high response output and good flexibility, it is best to choose piezoelectric materials with a high piezoelectric coefficient and flexibility, as well as ensure sensitivity and mechanical endurance. Sometimes, there are substrates between the samples to be measured and the sensors, which affect how the two interact. In this case, it is necessary to establish a connection between the substrate and the respective electric system, and comprehensively consider solving the practical problem. Hofmann et al. invented a textile-reinforced structure that can itself be used for sensing and integrated into a fiber-reinforced plastic composite to enable structural health detection. Apparently, the fiber-reinforced plastic composite is used as a substrate for insulation. In order to connect to the electric system, the researchers used copper tape with conductive adhesive to achieve electrical contact and, after solidification, exposed the tape by removing the hardening resin and finally soldered to the cable [59]. That is, in the case of the substrate as an insulating layer, the wiring can be completed on the substrate and connected to the power system. In fact, it is also possible to connect using wireless communication methods such as Bluetooth and WiFi. When the substrate is used as a protective layer, the power system of the substrate should try to ensure that the signals received by the substrate and sensors are not distorted and attenuated, with specific measures that include shielding, noise reduction, impedance matching, and so on. Microstrip antennas for sensor signal transmission are widely studied, which can realize wireless transmission, mask the sensor without affecting its ability to receive a signal proportional to the measurement signal, and also realize a low response of the device in different scattering angle cases through scattering offset mechanisms and stealth characteristics, which will be extremely beneficial to the stability of sensor signal transmission [60].

Some material parameters can conversely be measured by piezoelectric sensors, such as the measurement of sample texture, which involves the problem of inverse scattering of the sample texture [61]. The problem is to determine the basic material properties or surface characteristics of an object from the measurement or observation of scattered waves or signals. When an incident wave or signal interacts with a textured surface, scattering occurs and a scattering signal is obtained. Using a piezoelectric sensor, the acquired signal is collected, processed, and analyzed, and the sample texture can be analyzed. However, in the specific implementation, the determination of sample texture using piezoelectric sensors is still a great challenge because of the limited data, the complexity of the calculation, and the uniqueness of the results. Advanced mathematical and computational methods need to be developed to cope with it. Stoynov et al. studied the scattering problem of cracked piezoelectric planes, and with known far-field data, the stress intensity factor (SIF) at the crack edge was found [62]. Valagiannopoulos et al. derived the dielectric coefficient of anisotropic materials using the Taylor series expansion and appropriate polarization excitation after measuring the projection coefficient [63].

However, piezoelectric materials such as ZnO, PZT, BaTiO₃, and PVDF, which have been widely studied at present, have not achieved the function of having both flexibility and excellent piezoelectric properties. Although ZnO is a semiconductor, PZT, BaTiO₃, and other induced ceramics have high ferroelectricity, but they are poor in flexibility. In addition, PZT-based piezoelectric materials will be gradually abandoned because of their harm to the environment and biology. Similarly, PVDF and its copolymers are flexible, but their piezoelectric properties are too poor. Crystal structures, SEM images, and related devices of four piezoelectric materials, ZnO, PZT, BaTiO₃ and PVDF, were compared, as shown in Figure 4 [64,65,66,67,68,69,70,71,72,73,74,75].

To sum up, there is an urgent need to explore and design materials with high voltage and high flexibility to fabricate piezoelectric acceleration sensors [76]. It was found that piezoelectric polymers such as PVDF and PVDF-TrFE can be used for the preparation of current flexible piezoelectric acceleration sensors. The preparation methods are divided into two types: One is to directly prepare piezoelectric polymers into nanofibers by electrospinning and other technologies. For example, piezoelectric polymers can be processed into PVDF-TrFE nanofibers. Ghosh et al. realized the preparation of this material, finding good softness and air permeability of this material (see Figure 5a) [77]. Another example is that inorganic piezoelectric materials are compounded with flexible polymers as nano-fillers, and then after micro-processing, flexible piezoelectric sensors can be fabricated [78]. Ceramic/polymer composites were first proposed in 1978, in which electrical properties have been greatly improved and are expected to exhibit some new properties [79,80]. Inorganic materials in the material will be very sensitive to applied forces due to high voltage [81,82]. In addition, due to the high elasticity of the elastic matrix in the composite, they also have good shape preservation and flexibility. Common inorganic piezoelectric materials include LiNbO₃, PbTiO₃ nanowires, PbTiO₃ nanoparticles, etc. Elastic matrix is divided into non-piezoelectric organic materials and piezoelectric organic materials, and the former mainly include polydimethylsiloxane (PDMS), polyvinyl alcohol (PVA), and polystyrene (PS). Figure 5b–d show the composite process of piezoelectric ceramics and piezoelectric organic materials, among which Figure 5b shows the overall composite process [83] and Figure 5c [84] shows the change in material microstructure during the composite process, and Figure 5d [85] further shows the networking situation of piezoelectric ceramic particles in piezoelectric organic materials of PVDF. The composite of piezoelectric ceramics and non-piezoelectric organic materials is shown in Figure 5e–h [86]. Alluri et al. synthesized a flexible film of BaTiO₃ and PDMS. Finally, they found that the film had good flexibility and rolling ability. This is because PDMS is chemically stable, highly insulating, inexpensive, corrosion-resistant, resistant to high and low temperatures, has good dielectric properties, is highly elastic, and has a high recovery force. It is cross-linked into a film by a polydimethylsiloxane curing agent [87] and can be stretched to 300% of its original length [88].

4. Structure of Flexible Piezoelectric Acceleration Sensors

The traditional piezoelectric acceleration sensor structure is shown in Figure 6a [83]. According to the classification of the vibration mode and sensor structure, traditional acceleration sensors can be divided into compression-type (d33), shear-type (d15), and bending-type (d31) [89,90]. Compression-type piezoelectric acceleration sensors mainly consist of a fixation nut, a mass block, a piezoelectric sheet, and a base. The number of piezoelectric pieces can be increased or decreased, but it is necessary to pay attention to whether the series or parallel connection is used. Shear-type sensors are mainly composed of a fixed screw, a mass block, a piezoelectric piece, and a base [91]. Unlike compression sensors, which can use a single piezoelectric plate, shear sensors generally use pairs of piezoelectric plates. The structure of a flexible piezoelectric acceleration sensor is similar to that of a traditional piezoelectric acceleration sensor, but many rigid components are replaced by flexible materials to meet the need for flexibility.

Wang et al. [92] presented the design and testing of a uniaxial piezoelectric acceleration sensor made from cellulose paper and zinc oxide nanowires (ZnO-NWs). The device is shown in Figure 6b. As only the paper needs to be cut and then the growth of ZnO nanoparticles is carried out in a hydrothermal method, it does not require expensive and complex instrumentation. Figure 6c illustrates the experimental apparatus for this paper-based accelerometer with a sensitivity of 16.3 mV/g at a natural frequency of 84.75 Hz for this accelerometer weighing 61 mg. Zhang et al. [93] developed a 1D–3D fully piezoelectric nanocomposite based on perovskite BaTiO₃ and PVDF-TrFE and applied it in a material generator (hNCG) device (see Figure 6d). The output of such a flexible hNCG can be as high as 14 v and 4 μA, which is significantly higher than current devices made using piezoelectric ceramic films. The finite element analysis showed that the superiority of the piezoelectric composite is due to the piezoelectric synergy of the two materials and the effective pressure transfer ability of the piezoelectric composite.

Respiratory health surveillance with very low detection limits is extremely challenging. Given this, Yuan et al. [94] from Xidian University designed a flexible bionic sensor with a detection limit as low as 0.0005 N. The specific structure of the sensor is shown in Figure 6e, which has the advantages of short response time and high response repeatability. In addition, reasonable design and optimization of the back-end circuit can make it monitor various breathing states, including apnea, cough, and deep breathing, which is extremely expandable and a breakthrough in the wearable field.

Zhang et al. [95] designed a self-powered high-sensitivity acceleration sensor based on the triboelectric nanogenerator, as shown in Figure 6f. The sensor is composed of an LMMD and an nn-PVDF film, where the LMMD has high surface tension, high mass density, high elasticity, and mechanical robustness. The measurement shows that the acceleration sensor detection ranges from 0 to 60 m/s². It also exhibits excellent stability, having potential applications in equipment vibration measurement monitoring and troubleshooting.

Figure 6g shows the fabrication process of a flexible piezoelectric sensor made from a group III nitride film, designed by Chen et al. [96]. Group III nitrides exhibit environmentally friendly, chemically stable, biocompatible, resistant, and stable input and output characteristics.

In Figure 6h, a PVDF sensor patch [97] was designed to detect heartbeat and respiratory signals. The sensor patch is composed of three layers: the PDMS covering module, the PVDF polymer film, and the Mylar layer from top to bottom. The upper and lower surfaces of the PVDF sensor film are, respectively, equipped with two printed electrodes for signal conduction and grounding, which are connected to a shielded cable to prevent electrical noise. The Mylar layer provides insulation between the printed electrodes and the chest cavity. It was found that compared with the flat PVDF film structure, the curved structure can increase the detection signal by 151%.

5. New Technologies of Flexible Piezoelectric Acceleration Sensor

5.1. Three-Dimensional Printing Technology

5.1.1. Introduction to 3D Printing Technology

The future application areas of sensors will develop in several directions, such as real-time monitoring, advanced diagnostics, and interconnected feedback, which will undoubtedly place higher demands on the sensitivity, resolution, strength, and hardness of sensors. The emergence and development of 3D printing technology obviously provide great possibilities for this.

Three-dimensional printing technology, also known as additive manufacturing, enables the production of precisely shaped, smooth-surfaced parts, where the printing process can be controlled automatically, started and stopped at any time, or incorporated into complementary manufacturing processes or embedded in sub-components manufactured with conventional processes. The production process for 3D-printed sensors can thus be seamlessly achieved by embedding the sensor into a printed structure or by printing the entire sensor from the ground up [98]. Sensors with complicated geometry, complex functionality, and ease of assembly are becoming possible thanks to advancements in multi-process and hybrid 3D printing technologies [99,100].

5.1.2. Three-Dimensional Printing Technology Classification and Process Flow

The whole process of 3D printing technology is shown in Figure 6 [101]: First, a model of the object to be printed is created using software such as Solidworks. Once the 3D model is created, it is then converted into an STL file, which stores all the information about the details of the model. The 3D printer obtains a series of 2D cross-sectional layers by splitting in the Z-direction. Finally, the desired object is printed layer by layer. Taking the printed flexible sensor and “glove” as examples, Figure 7(bi) shows the CAD design of the flexible sensor. Figure 7(bii) and Figure 7(biii) show the un-flexed and flexed states of the sensor, respectively. Figure 7(ci) and Figure 7(cii), respectively, show the CAD design of “gloves” and the finished product after printing. Figure 7(ciii) and Figure 7(civ), respectively, show the printed “gloves” in two states: un-flexed and flexed [102]. The piezoelectric nanocomposites were printed using 3D printing technology and the device used was a custom PuSL fabrication system, as shown in Figure 7d. As seen in Figure 7e–h, the printed piezoelectric nanocomposites have a fine surface structure and excellent surface finish [103]. Bodkhe et al. reasonably adjusted the nozzle size, pressure, and solution concentration to obtain 1D, 2D, 2.5D, and 3D shapes of PVDF, as shown in Figure 7i–l [104].

Three-dimensional printing technologies can be subdivided into seven main categories according to specific manufacturing techniques, and their detailed technical processes and characteristics are as follows:

• Fused Deposition Modeling (FDM)

The process works by melting and extruding a thermoplastic filament from a nozzle, which is mainly from thermoplastic materials such as polyamide (PA), polylactic acid (PLA), and polycarbonate (PC). The melted material is deposited on a production platform, where it is cooled and solidified to form a single layer. This process is repeated layer by layer to form a 3D model. FDM has the characteristics of being low cost and open source, but its print resolution and print speed are often low. Crump first proposed FDM [105], its device is shown in Figure 8a [106].

Based on the thermoplastic properties of PVDF and taking advantage of the combination of FDM and the corona poling process (IPC), Kim et al. achieved the low-cost, low-complexity, and flexible fabrication of PVDF piezoelectric sensors [107]. In the same year, they also found that FDM could significantly improve the uniform dispersion of BT nanoparticles in PVDF matrix, enhancing the piezoelectric properties. As a result, they designed a Multiwall carbon nanotube (MWCNT)/BT/PVDF nanocomposite film, whose tests showed excellent mechanical properties and durability [108].

2.

Direct Ink Writing (DIW)

Direct ink writing, as shown in Figure 8b [109], uses nozzles to extrude materials directly onto the manufacturing platform, where the material is deposited in a viscous liquid state to maintain its shape after deposition. Direct inkjet technology can be used in many materials, such as ceramics, plastics, food, hydrogels, living cells, and so on. The size of the nozzle, the viscosity and density of the material, the scanning speed, the jet speed, and other parameters can all be adjusted to obtain an optimal deposition object. Post-processing may require hardening the manufactured object and improving its mechanical properties through sintering, heating, UV curing, and drying.

DIW technology is suitable for multi-material, free-form, complex geometry printing and is often used for the one-step printing of force sensors. Using DIW to print composites of PDMS and barium titanate (BTO), the resulting samples are high-density, have a high-voltage electrical coefficient, and have tunable mechanical properties; however, the prints have a low resolution and are subject to complex post-processing processes such as high-temperature curing and photopolymerization [110].

3.

Photocuring (SLA, DLP)

Photocuring is curing layer by layer by UV light to form a three-dimensional structure on the platform. Light curing technology can be divided into two types according to the light source and the way it is processed: stereolithography equipment (SLA) and digital light processing (DLP) [111]. The former uses a moving laser to cure the photosensitive resin directly, as shown in Figure 8c [112], while the latter uses a laser or UV lamp as a light source to cure the exposed parts by means of a special pattern. SLA and DLP can produce high-precision structures with complex internal characteristics but have one disadvantage: only a single material can be used.

Biosensors require extremely complex information to be processed in a small mechanical building block, but conventional fabrication methods often struggle to achieve extremely strong mechanical coupling within the sensor, which makes it difficult to apply biosensors for high-precision measurements on tiny objects. Tiller et al. [113] proposed the use of DLP to achieve this requirement. First, barium titanate nanopowders were mixed with multi-walled carbon nanotubes (MWCNTs) into a PEGDA-based monomer resin, and the MWCNTs were oriented and aligned by an electric field to make piezoelectric devices.

4.

Lamination (LOM)

Figure 8d [114] shows the experimental apparatus designed for the LOM. The additive manufacturing technology for LOM was first developed by Helisys in 1988 [115]. It first uses a laser or knife to cut the sheet material, and after cutting off a layer, a new layer is firmly adhered up by a roller that compacts and heats/glues the sheets together. This process is repeated and the unwanted parts are later removed to obtain a complete 3D structure [116].

5.

Selective Laser Sintering and Selective Laser Melting (SLS and SLM, respectively)

Both SLS and SLM are based on powders, including plastics, metals, ceramics, and waxes, which can be printed with sufficient strength and density to meet aerospace or military requirements [117]. The device for SLA is illustrated in Figure 8e [118].

Song et al. [119] showed that SLS can be used to print PVDF/graphene nanocomposites with hollow axial-shaped arrays resulting in parts with higher conversion efficiency, higher piezoelectric performance with an optimum open-circuit voltage of 16.97 V, a short-circuit current of 274 nA, and the ability to light up to 10 LED bulbs simultaneously.

6.

Photopolymer Jetting (Ployjet)

The photopolymer injection technique was first proposed by Gothait [120]. The printing material chosen was a photosensitive resin, which was squeezed out of a nozzle and deposited onto a printing platform, and then it was cured by UV light, as shown in Figure 8f [110]. This method also requires layer-by-layer manufacturing. The advantage of optical polymer injection is that it can print a variety of materials and colors at the same time, and it has a high resolution, which is suitable for printing small and delicate objects. The disadvantage is that the parts produced by this method are weaker.

Ployjet enables printed, highly integrated sensors and actuators that are constructed from a variety of photosensitive polymers [121]. The ployjet-based 3D printing technology can also be used to prepare multifunctional, highly integrated, and efficient microfluidic devices. Based on previous research, Nigtingale et al. developed a microfluidic-chip-based wearable bioassay system for monitoring biomolecular levels using liquid chromatography technology [122].

Figure 8. Schematic diagram of the equipment of 3D printing. (a) Fused deposition modeling (FDM). Reprinted with permission from Ref. [106]. Copyright 2018, MDPI AG. (b) Direct ink writing (DIW). Reprinted with permission from Ref. [109]. Copyright 2020, Elsevier. (c) Stereolithography (SLA). Reprinted with permission from Ref. [112]. Copyright 2014, American Chemical Society. (d) Lamination (LOM). Reprinted with permission from Ref. [114]. Copyright 2016, John Wiley and Sons. (e) Selective laser sintering (SLS). Reprinted with permission from Ref. [118]. Copyright 2018, American Chemical Society. (f) Photopolymer jetting (Ployjet). Reprinted with permission from Ref. [109]. Copyright 2017, MDPI AG. (g) Binder jetting (3DP). Reprinted with permission from Ref. [123]. Copyright 2018, MDPI AG.

Figure 8. Schematic diagram of the equipment of 3D printing. (a) Fused deposition modeling (FDM). Reprinted with permission from Ref. [106]. Copyright 2018, MDPI AG. (b) Direct ink writing (DIW). Reprinted with permission from Ref. [109]. Copyright 2020, Elsevier. (c) Stereolithography (SLA). Reprinted with permission from Ref. [112]. Copyright 2014, American Chemical Society. (d) Lamination (LOM). Reprinted with permission from Ref. [114]. Copyright 2016, John Wiley and Sons. (e) Selective laser sintering (SLS). Reprinted with permission from Ref. [118]. Copyright 2018, American Chemical Society. (f) Photopolymer jetting (Ployjet). Reprinted with permission from Ref. [109]. Copyright 2017, MDPI AG. (g) Binder jetting (3DP). Reprinted with permission from Ref. [123]. Copyright 2018, MDPI AG.

7.

Binder Jetting (3DP)

This method is similar to the aforementioned method, except that this method does not require any support structure, as the powder can support itself. Powder materials are often ceramics, gypsum, sugar, etc. Adhesive jetting is mainly used in the biological field, and it can print a wide range of materials, but the strength and surface roughness of the object are not so good. Figure 8g [123] shows the device used for 3DP.

The classification, materials used, advantages, and disadvantages of 3D printing technologies are summarized in

Table 2. Three-dimensional printing technology’s classification, commonly used materials, advantages, and disadvantages.

Categories	Materials	Advantages	Disadvantages
FDM	All-plastic materials (PA, PLA, PC, etc.)	Low cost Open-source code	Slow printing speed Low resolution
DIW	Ceramics, plastics, hydrogels, and even living cells	Fast printing speed High durability Suitable for repairing components	Subsequent processing required Low mechanical properties Poor surface finish
SLA, DLP	Resins, polymers, plastics	High speed High resolution Complex and high-precision structure	Single material Easy to damage the device Support structure required
LOM	Papers, plastics, metals	Low Cost High speed Color printing	Post-processing required Low resolution Design limitations
SLS, SLM	Plastics, metals, ceramics, waxes	High strength and stiffness Suitable for aerospace and military No support mechanism required	Slow printing speed Size limitation High power usage
Ployjet	Photosensitive resins	High resolution, suitable for small and delicate devices	Low strength
3DP	Ceramics, plasters, sugars	Can print simultaneously Color printing No support mechanism required	Poor surface strength and roughness Requires post treatment Low mechanical properties

:

5.1.3. Influence of 3D Printing Parameters on Sensors

The benefits of 3D-printed piezoelectric acceleration sensors are evident to all, but for this sensor to have high performance, it often requires good tuning of the parameters associated with 3D printing, which include extrusion rate, nozzle diameter, nozzle temperature, and bed temperature. The extrusion rate is positively correlated with the pressure and shear rate at the narrowing of the nozzle, and its variation may cause the print trace spacing to change and overlap [124]. Smaller nozzle diameters can damage the structure of materials such as fibers or cause nozzle clogging, resulting in reduced sensing capabilities [125], while using larger nozzles may reduce the print resolution [126]. Nozzle temperature and bed temperature also need to be strictly controlled, because above the appropriate temperature range, the material viscosity will be reduced, the print traces will be deformed, and even the material will be degraded [127]; below the appropriate temperature range, the material will not adhere to the surface of the printing bed. Kim et al., while using the 3D printing process to fabricate PVDF films, found that too high a bed temperature reduces the piezoelectric properties of PVDF films and decreases the sensitivity [108].

5.1.4. Application of 3D Printing Sensor

It is known that high-resolution physical sensors or microfluidic devices can be used for particle detection [128] and 4 μm single-cell capture [129]. However, the traditional manufacturing method is not only complicated, time-consuming, and lacks scalability and extensibility, but also has high manufacturing costs to achieve a high-resolution structure. With the rapid development of 3D printing technology, the characteristics of short manufacturing time and low manufacturing cost can obviously make up for these defects. The cost of a sensor manufactured by conventional methods is USD 215, while the cost of such a sensor printed by 3D printing (stereolithography) is limited to USD 200 [130]. In addition, microfluidic devices with complex structures, such as multiple PDMS layers, can be reproduced simply and quickly by 3D printing technology, without the need for soft lithography manufacturing requiring complicated processes such as optical rotation resistance, coating, and UV exposure, saving a lot of cost and labor.

An even greater feature of 3D-printed sensors is that they can achieve high sensitivity, such as omnidirectional radiation pattern sensing, which can be obtained by 3D printing with a larger response surface area, to achieve 2.25 times sensitivity [131,132,133,134,135]. Three-dimensionally printing sensors can also achieve high flexibility. It makes it easy to change the geometry of the sensor so that it can be paired with electrodes as an electrical sensor, or with smartphones as an optical sensor to match measurements with other commercial products of known and measurable size.

A flexible sensor typically consists of a sensing element, a flexible substrate, flexible electrodes, packaging, and wiring. The application of 3D printing to sensors varies depending on the components printed and often includes (1) custom molds with various surface structures to obtain sensor substrates, electrodes, and sensing elements; (2) the printing of flexible bases or major parts of sensors that are then used to compose sensors with good flexibility and ductility; (3) the manufacture of sensing elements with specially tailored geometries and microstructures; (4) the suitability for printing flexible electrodes. In some sensors, flexible electrodes are required to take advantage of their conductive and stretchable properties, which can lead to better impedance stability [136]; (5) complete printing of the entire sensor [120].

Figure 9a shows the design of a flexible tactile sensor proposed by Guo et al. [137]. The design develops a non-sintered ink with electrical conductivity and adjustable viscosity. Using a multi-material, multi-scale, and multi-functional 3D printing method, the designed sensor shows good linear current–voltage characteristics (see Figure 9b for the detailed process flow). Figure 9c [138] shows a piezoelectric pressure sensor fabricated by Fuh et al. The sensor substrate was fabricated by an FDM process and has a wavy shape. A layer of copper foil was placed on the substrate, and the piezoelectric composite PVDF was deposited on top of it by electrospinning. Finally, the sensor was encapsulated using PDMS film. Figure 9d [139] shows a functional acceleration sensor for piezoelectric signal readout. Bernasconi et al. combined 3D printing and inkjet material deposition intelligently to create a PVDF-TrFE piezoelectric layer and corresponding silver electrode. At the same time, stereolithographic techniques were selected to fabricate the structural components of the accelerometer. The results show that the proposed hybrid additive manufacturing technique is very promising for the mesoscale manufacture of electromechanical sensors.

5.2. MEMS Technology

5.2.1. Introduction to MEMS Technology

Prior to measurement, some sensors that need contact measurements are glued to the host structure. Obviously, this is less ideal for smaller structures due to the size of the sensor itself as well as the adhesive’s excessive thickness, which could alter the substrate’s behavior. Sensors should scale with the size of the host structure and should be optimally placed. The optimal sensor placement is a function related to the geometry of the structure and the vibration signal to be retrieved, which also requires the sensor to be much smaller than the host structure [140]. MEMS technology clearly meets these requirements.

MEMS technology gradually developed in the 1990s as an emerging technology in the field of micromechanics, which uses integrated circuit technology processes and micromachining methods to integrate electromechanical sensitive components and processing circuits based on various physical effects on a chip, successfully realizing the combination of micromachining and microelectronics technology [141]. Rather than improving the performance of these systems while reducing size and power, as traditional engineering efforts have done, the MEMS field has worked to fundamentally change the size, performance, and cost [142,143]. With the development of MEMS technology, the integration of the devices and the size of the chips continue to improve. It offers the following advantages: smaller device size, better electrical and mechanical properties, the possibility of mass production, integration in many areas, etc. [144,145]. These advantages are of great significance for the miniaturization, integration, and large-scale production of sensors. MEMS technology is also now deeply integrated with acceleration sensors. It is known that MEMS technology combined with optical inspection technology can realize the measurement of sub-angstrom-scale sensitivity and nanometer-scale acceleration, and the performance is close to the Brownian noise limit of mechanical structures. This sensitivity cannot be achieved by capacitive or piezoresistive methods, and increasing the mass of the acceleration sensor will be sensed at lower frequencies [146].

5.2.2. Process of MEMS Sensors

MEMS processes can be divided into the following categories: additive processes, subtractive processes, patterning, material property changes, and mechanical steps. Common additive processes include metal evaporation, metal sputtering, chemical vapor deposition of organics, chemical vapor deposition of inorganics, plasma-assisted chemical vapor deposition, thermal oxidation, and electroplating [147]. Subtractive processes include plasma etching, reactive ion etching, deep reactive ion etching, silicon wet chemical etching, and thin film wet chemical etching. Patterning mainly refers to photolithography processes. Material property changes include ion implantation, diffusion doping, and thermal annealing. Mechanical steps are polishing, wafer bonding, wafer slicing, lead bonding, and chip packaging [148].

Some commercial MEMS sensors are designed to be general-purpose with large bandwidths, and such sensors have large intrinsic frequencies; however, this can lead to poor sensitivity. To avoid this problem, dedicated MEMS sensors can be designed for different applications and different frequencies. For example, in civil structure health monitoring, low-seismic-level, low-frequency, and high-sensitivity acceleration signals are required, then MEMS accelerometers with low intrinsic frequency and high background noise can be designed to achieve the purpose [149]. Packaging technology, one of the key technologies for MEMS acceleration sensors, also has a critical impact on the performance of the overall device. Because of the influence of different package shells, die adhesives, and chip materials, thermal stresses are generated by temperature changes during packaging [150]. Li et al. found through theoretical analysis and simulation tests that the die adhesive dosage is inversely proportional to the sensitivity during packaging, and the binder thickness is inversely proportional to the residual stress after packaging. Uniform binder dosage and thickness with a suitable potting adhesive can ensure the consistency and improvement in device sensitivity [151].

Taking the piezoelectric MEMS sensor as an example, we introduce the flow of the additive process, subtractive process, and integrated process in detail. Firstly, these three processes are compared according to the conditions of device size, preparation speed, sensitivity, etc. (see Figure 10).

• Addition process

Additive processes refer to the deposition of piezoelectric films on a silicon substrate with a suitable insulating layer, followed by micromachining of the surface or silicon block. The process starts with the sequential deposition of a passive layer, an active layer, and an electrode layer. Different thicknesses of SiO₂, Si₃N₄, or polycrystalline silicon are first deposited on a substrate material such as Si to form the passive and sacrificial layers required for the sensor. In Figure 11a, the electrode films are deposited using sputtering and vacuum evaporation techniques, and the piezoelectric films are prepared using a combination of sputtering, chemical solution deposition, and evaporation techniques. To achieve the detection of charge and voltage in sensitive elements, the piezoelectric film is generally sandwiched between two electrode layers, or an interlocking electrode pattern is placed on top of the piezoelectric film. In the former case, the bottom electrode layer is used as a seed layer on which a textured piezoelectric film is deposited [152].

2.

Subtractive process

The subtractive process is the direct micromachining of single or polycrystalline piezoelectric materials, piezoelectric ceramics, etc., followed by appropriate plating [153] (see Figure 11b). It allows for the integration of sensors and actuators. It is worth mentioning that the etching of piezoelectric materials remains difficult for subtractive processes because of the stable nature of their oxides, which do not easily undergo the selective chemical reactions required for patterning. It is also the case that piezoelectric sensors implemented using batch micromachining processes lack a reliable and fast etching process. As a result, there are always limitations to such sensors [154].

3.

Integration process

Integration processes integrate micromechanical structures in silicon onto large piezoelectric substrates through bonding techniques, such as integrating precision micro-machined silicon structures onto PZT substrates. This has the advantage of being able to use a large number of Si micromachining techniques without the need to develop extensive micromachining capabilities for PZT. However, the success of this technique is predicated on the availability of low temperatures (preferably <200 °C) as well as precisely aligned bonding techniques. Andrea [154] et al. designed a MEMS actuator for PZT using an integrated process (see Figure 11c). This actuator is integrated with an embedded varistor, allowing the amplitude to be precisely controlled, and the stability and white noise to be effectively improved.

4.

Sensor fabrication on the silicon substrate

The basic flow of sensor fabrication on a silicon substrate is shown in Figure 11d [155], where spin-on glass (SOG) is first spin-coated onto silicon and dried at a certain temperature to cross-link the SOG. Next, an aluminum layer is evaporated to form an ohmic contact and pattern, and this is followed by alumina sputtering in an argon atmosphere (see Figure 11(d2)). The ZnO films are patterned using a mixture of a certain volume ratio of phosphoric acid, acetic acid, and aqueous solution. The second SOG layer is then spin-coated and patterned (see Figure 11(d3,d4)). The top electrode is then evaporated and patterned (see Figure 11(d5)). An SOG is spin-coated to encapsulate the sensor and is patterned, then the wafer is cut (see Figure 11(d6,d7)). Finally, it is processed using bulk micromachining or by cutting into strips (see Figure 11(d8)).

5.

Sensor fabrication on steel substrates

The manufacturing of sensors on steel substrates is more difficult compared to silicon [148], specifically related to two factors. One is that the suspension design of the steel substrate requires the steel sheet to be limited to a certain thickness and is therefore highly susceptible to damage during fabrication. After the vacuum chuck was invented to solve this problem, its use caused permanent deformation of the steel sheet, a deformation that, in turn, led to a change in the film thickness on the substrate. In response, the steel can be glued to the silicon shank with water and then the two wafers are heated to 100 °C, evaporating the water and releasing the steel substrate. The second is that steel has a large coefficient of thermal expansion. It is an order of magnitude greater than most micromachined compatible materials. This mismatch can lead to significant residual stresses and cause adhesion problems after film deposition.

For these two reasons, the temperature should be kept low during the manufacture of steel substrate sensors. First, the SOG is spin-coated onto the steel substrate and baked to flatten the steel substrate and prevent oxidation on the front side of the substrate. After the SOG layer has been applied to the steel substrate (Figure 11(e1)), the aluminum layer is evaporated and patterned to treat the bottom electrode (Figure 11(e2)). A second SOG layer is then coated; the benefit of using this additional SOG layer is its use as a buffer layer between the piezoelectric material and the aluminum layer, reducing residual stress gradients and the better adhesion of the piezoelectric material to the aluminum; it also allows the aluminum layer to be planarized for better deposition of zinc oxide. Continuing with the deposition and patterning of the ZnO layer on top of the second SOG layer, the SOG layer sandwiched between the ZnO and the underlying electrode is dry-etched in the SF plasma (Figure 11(e2)), a self-aligning process as the ZnO layer is used as a mask. The rest of the process is shown in Figure 11(e3–e7) and is the same as the process used in the Si substrate. After the sensor has been fabricated, the steel sheet is etched, as in Figure 11(e8), and finally tested for assembly.

Figure 11. Process implementation of MEMS piezoelectric acceleration sensors: additive process, subtractive process, integrated process, sensor manufacturing on silicon substrates, and sensor fabrication on steel substrates. Image for “integrated process”: Reprinted with permission from Ref. [154]. Copyright 2021, Elsevier.

Figure 11. Process implementation of MEMS piezoelectric acceleration sensors: additive process, subtractive process, integrated process, sensor manufacturing on silicon substrates, and sensor fabrication on steel substrates. Image for “integrated process”: Reprinted with permission from Ref. [154]. Copyright 2021, Elsevier.

5.2.3. Application of MEMS Flexible Piezoelectric Accelerometer

Li et al. [156] proposed an LC wireless passive flexible acceleration sensor to eliminate the difficulty of measuring acceleration on the surface of curved structures. The principle of this accelerometer lies in varying the distance between the accelerometer and the antenna, using RF wireless transmission to measure the acceleration signal, and extending the lifetime by preventing the wire from falling off during vibration. It consists of a flexible polyimide (PI) substrate and a planar spiral inductor coil, and the processes for manufacturing such sensors are shown in Figure 12a. The measurements show that the acceleration sensor has a measurement range of 20 to 100 m/s² and the acceleration and output voltage are almost linear with a maximum error of less than 0.037%. An ultrathin bendable flexible silicon-based piezoelectric sensor [157] was realized by wet anisotropic etching as a post-processing step, and the cutting of the film was achieved by a low-cost pre-etching cutting technique (see Figure 12b). Due to the optimized polarization procedure, this sensor has better piezoelectric properties and facilitates low-cost fabrication, as shown in Figure 12c. Figure 12d,e [158] show a flexible acceleration sensor composed of PVDF-TrFE and nanoclay composites that can be used for vibration measurement. The components of the sensor also include parallel stripe electrodes with impedance-matching circuits and an arc spring polyimide (PI) substrate. The test results in Figure 12f show that the design has high sensitivity, and the reasonable choice of the nanoclay composition can further promote the sensitivity. It is worth mentioning that this sensor also has good temperature stability and noise resistance.

6. Prospect and Outlook

In summary, the research progress of flexible piezoelectric acceleration sensors and their recent research results are introduced from the perspectives of properties, materials, structures, and processes. As a combination of flexible electronics, acceleration sensors, and other hot fields, flexible piezoelectric acceleration sensors make up for the shortcomings of traditional acceleration sensors, such as poor impact resistance and a complex manufacturing process. At the same time, it can achieve high flexibility, high sensitivity, self-power, and high reliability, especially suitable for acceleration measurement on some bending structures. Therefore, it has important applications in biomedicine, aerospace, wearable devices, the Internet of Things, and other fields.

Although the flexible piezoelectric acceleration sensor has made great progress and development in the past few years and has great potential and advantages in various fields at present, there are still some problems to be solved or improved in practical applications. Designing and fabricating high-performance piezoelectric sensors involve material innovation, structural innovation, material optimization, and principal innovation. However, most of the preparation processes are carried out under laboratory conditions, and due to the uncertainty of complexity, repeatability, and stability in practical applications, it is difficult to achieve the large-scale production of piezoelectric acceleration sensors under low-cost conditions. Three-dimensional printing technology and MEMS technology can be further improved to achieve the purpose of reducing costs to achieve mass production, which is a major aspect worth exploring. In addition, piezoelectric acceleration sensors with low power consumption meet the requirements of small size and low power consumption for future flexible microelectronic systems. Therefore, the development of integration technologies for flexible piezoelectric acceleration sensors and other electronic devices to achieve complete and complex functions is another hot research topic.