Piezoelectric composites for gas sensing: evolution of sensing and transduction designs

Abstract

The development of piezoelectric composite (PEC) gas sensors is progressing rapidly, driven by innovations that range from atomic-level material design to system-level integration. While prior reviews have largely concentrated on material properties or transduction mechanisms, this review uniquely connects fundamental sensing principles with the challenges of integrating these sensors into larger systems. We comprehensively analyze recent progress in PEC gas sensors, highlighting their evolution from rigid structures to multifunctional, miniaturized systems. Key advances include: engineering interfacial coupling between gas sensitivity and energy harvesting to significantly enhance chemisorption reaction rates and piezoelectric efficiency; developing multiscale computational models for the rational design of high-performance sensors; and implementing AI-driven signal processing and energy management systems that mitigate environmental interference and power constraints. This review systematically summarizes the working principles, materials selection, transducer design, novel devices enabled by synergistic interactions, and AI-assisted gas recognition systems for PEC gas sensors. Finally, it offers critical perspectives on both the challenges and opportunities facing this technology.

---

Introduction

Recent progress in piezoelectric composite (PEC) gas sensors arises largely from innovations spanning microscale material design to macroscale device integration.^1–6 Drawing inspiration from Yamazoe et al.'s⁷ “receptor-transducer-utility factor” theory (developed for oxide semiconductor sensors) and pioneering work on piezoelectric nanowire gas sensors,⁸ Su et al.⁹ established three core functional pillars for PEC gas sensors: (1) gas sensing capability, enhanced through optimized surface chemistry (e.g., maximizing specific surface area and catalytic activity); (2) transduction efficiency, improved via piezoelectric energy conversion (e.g., controlling grain size in ceramics/polymers and enhancing electromechanical coupling); and (3) material utilization, optimized by designing porous microstructures and efficient catalyst distribution. This framework integrates chemical-electrical interactions with piezoelectric mechanics, providing a unified design approach. Crucially, it simultaneously enables the sensor's dual function: detecting gases via chemisorption and converting mechanical energy into measurable electrical signals. Realizing this dual function effectively requires interfacial engineering^6,10–12 to synchronize chemical reactions at the energy-harvesting interfaces and convert gas information (type and concentration) into electrical outputs.

However, beyond the material level, research efforts remain fragmented across disciplines such as computational modeling, circuit design, and algorithm development. This fragmentation, evident in disconnected advances in areas like multifunctional composites,¹³ interface engineering,¹⁴ and bioinspired flexible systems,^15,16 obstructs the establishment of clear relationships between material design, manufacturing approaches, and overall sensor performance. Consequently, progress toward developing next-generation sensors is hindered.

To address this complexity, our review organizes these diverse research threads into a cohesive framework. We evaluate the evolution of PEC gas sensors—from early simple ceramic–polymer blends to sophisticated engineered microstructures—and analyze how innovations in fabrication overcome historical limitations like poor filler dispersion and inadequate interfacial control. We examine functionalized ceramic nanowires and organic–inorganic composites, considering not only their performance advantages but also critical trade-offs in durability and manufacturing complexity.

Moreover, we assess emerging paradigms such as AI-enhanced systems¹⁷ and multiscale simulation-guided design, focusing on their potential to improve sensor selectivity and reduce errors. Insights are drawn from a system-level perspective inspired by engineering principles.^16,18

By synthesizing this multidisciplinary knowledge, our work bridges material innovations with system-level engineering needs. Fig. 1 presents our framework, which tackles the fragmentation in PEC sensors. At its core lies the integration of the three pillars—material utilization, transduction efficiency, and sensing capability. Upstream support comes from multiple fabrication processes and simulation-guided design, while downstream support is provided by AI-enhanced signal processing and miniaturized backend circuits. This structure explicitly links atomic-scale material design with the performance of functional device systems.^13–18 The ultimate goal is to enable the development of robust PEC gas sensors that operate reliably in real-world settings beyond the laboratory, offering practical solutions for applications in smart cities, health monitoring and so on.

Fig. 1 A framework spanning the entire upstream-downstream in the PEC gas sensor field. Ranging from multiscale simulation-guided material fabrication processes to portable backend circuitry and application-layer machine learning algorithms.

1. Structural evolution in PEC-based gas sensors

Piezoelectric sensors offer inherent advantages for miniaturization. They directly convert mechanical stress into electrical energy, eliminating moving parts – unlike triboelectric systems that require movable components to generate displacement currents.^19,20 However, three challenges hinder PEC gas sensor development: poor understanding of autonomous chemical information loading mechanisms;^4,7,21,22 low energy conversion efficiency at micro/nano scales;^23–26 and difficulty controlling interfacial states precisely.^8,19,27–33

A 2013 breakthrough came when Z. L. Wang's group created the first self-powered H₂S sensor using ZnO nanowires, combining gas sensing with piezoelectric functions.²⁷ This pioneered piezoelectric ceramic nanowire gas sensors and proved material architecture's critical role. For these sensors, key performance factors—sensing, transduction, and utilization efficiency—are tightly linked.^{2,21,27,34–44} Hydrothermally grown ceramic nanorods (Fig. 2a) face limitations due to fixed shapes and simple 1–3 connectivity. They inefficiently harvest energy from multidirectional mechanical inputs,^{6,7,26,45–49} restricting detection range and conversion efficiency. Li et al. confirmed this computationally (Fig. 2b): nanowires perform best under vertical loads but show strong direction dependence under other forces.⁴⁵

Fig. 2 Advancements in PENG-based gas sensor and structural innovations. (a) Flexible piezoelectrically self-powered H₂S sensor. Reproduced with permission from ref. 8. Copyright 2014, IOPscience. (b) Phase-field and ML co-optimization. Reproduced with permission from ref. 45. Copyright 2022, Wiley-VCH. (c) Coaxial electrospun smart fiber. Reproduced with permission from ref. 60. Copyright 2023, Elsevier. (d) LiCl-PVDF flexible nanocomposite. Reproduced with permission from ref. 10. Copyright 2023, Elsevier.

To overcome directional constraints, researchers develop composites. Ceramic–polymer designs leverage sophisticated topologies,⁵⁰ while continuous ceramic frameworks reduce internal impedance.⁵¹ But composites require optimizing two aspects simultaneously: integrating gas-sensitive materials into matrices while controlling interface compatibility;^{11,19,20,26,49} and managing chemical-specific reactions at energy conversion interfaces—including heterojunction potential changes^52–54 (e.g., metal–organic framework, MOF/polymer charge transfer barriers⁵³) and hybridized orbital interactions^55–57 (e.g., NH₃-triggered d-p orbital coupling in core–shell nanofibers⁵⁷).

Recent work exposes new composite challenges. Su et al.⁹ found bulk polymer matrices isolate conductive polymer receptors from piezoelectric ceramic transducers. This restricts target molecule access to transducers and weakens electrostatic shielding, degrading information encoding. Polymer-ceramic stiffness mismatch also causes poor encapsulation, reducing stress transfer and electromechanical coupling.^20,57–59

These issues demand new structural approaches. Coupling gas-sensitive and piezoelectric materials in hierarchical architectures could boost material utilization efficiency, strengthen chemical information encoding via enhanced electrostatic shielding, and optimize energy conversion through improved stress transfer. Emerging techniques target interfaces—for example, Chen et al.'s coaxial spinning method (Fig. 2c) integrates receptor-transducer-utilization functions in single fibers, achieving outstanding NO₂ sensing.⁶⁰

For absolute signal fidelity needs, decoupled designs physically separate sensing and energy conversion units. Guan et al.'s humidity sensor (Fig. 2d)¹⁰ prevents chemical interference this way but sacrifices integration.⁴² Recent advances resolve this trade-off: novel designs integrate modules spatially while maintaining functional independence, enabling both interference prevention and miniaturization.^61–63

2. Evolution of piezoelectric composite fabrication

The structure and performance of piezoelectric composites (PECs) depend heavily on how they are made. Creating complex structures requires precise manufacturing methods. This chapter reviews the progression from early solution-based techniques to advanced hybrid processes, showing how structural precision has advanced with technological breakthroughs. We analyze seven key methods (Fig. 3a–g), highlighting their role in solving critical challenges: achieving uniform filler dispersion (1980s–2000s), optimizing interfaces (2010s), and enabling complex structural design (post-2020). Crucially, advanced architectures—like hierarchical or 3D-structured composites—only became possible because processing techniques advanced too. These innovations show how manufacturing capabilities drive better performance, proving that process innovation and material design are interdependent for next-generation piezoelectric composites.

Fig. 3 Evolution of fabrication processes for piezoelectric composite materials. (a) Flexible PVDF composite films were prepared via a solution casting method. Reproduced with permission from ref. 66. Copyright 2021, Springer Nature. (b) Schematic diagram of the hot-pressing process for constructing textured BT2/PVDF composites and the corres-ponding stress distribution in the final product. Reproduced with permission from ref. 72. Copyright 2018, Elsevier. (c) Schematic diagram of three-dimensional multilayer nanofiber membranes (NFMs) and cross-sectional SEM images of the NFMs. Reproduced with permission from ref. 74. Copyright 2020, Elsevier. (d) Schematic of the solution spin-coating process for double-layered BT/PVDF nanocomposite films. Reproduced with permission from ref. 23. Copyright 2018, Elsevier. (e) 3D-printed stretchable kirigami piezoelectric nanogenerator. Reproduced with permission from ref. 86. Copyright 2020, Elsevier. (f) Fully transparent and flexible charge-generating piezoelectric nanodevices with ZnO nanorods and SEM image. Reproduced with permission from ref. 34. Copyright 2009, Wiley-VCH. (g) Schematic diagram of the preparation of polydopamine-modified BT/P(VDF-TrFE) nanocomposite. Reproduced with permission from ref. 31. Copyright 2020, Elsevier.

2.1. Solution casting

Solution casting is one of the earliest methods for making composites.^64,65 It involves dissolving polymers in solvents, dispersing fillers by stirring, and evaporating the solvent to form films. While scalable, traditional solution casting often spreads fillers randomly. Recent improvements have increased its usefulness. Cheng et al.⁶⁶ showed this by making ultrathin polyvinylidene fluoride (PVDF)/MXene/AgNW films (Fig. 3a). They dispersed the materials sequentially in dimethylformamide (DMF) followed by post-treatment, creating a 3D conductive network. This resulted in exceptional electromagnetic interference (EMI) shielding (41.26 dB) and thermal conductivity (0.78 W m⁻¹ K⁻¹).

Despite its simplicity, conventional solution casting struggles with filler clumping at high concentrations. For instance, lead zirconate titanate (PZT) particles in nanocellulose composites aggregate, degrading dielectric properties,⁶⁷ and PVDF/ZnO films face trade-offs between porosity and piezoelectric response.⁶⁸ Modern approaches integrate advanced dispersion or post-processing. Techniques like dielectrophoresis (DEP) [e.g., Wang et al.⁶⁹] greatly improve filler arrangement. A high-frequency AC field drives piezoelectric particles in the polymer matrix to form aligned chains or columns, boosting piezoelectric response and electromechanical coupling. Solution casting has also been adapted to create flexible molecular crystal composites, like biodegradable biosensors.⁷⁰ These advances show how traditional methods are being redesigned to meet modern demands: controlling microstructure, maintaining scalability, and enabling multifunctionality.

2.2. Hot-pressing for textured composites

Hot-pressing uses high temperature and uniaxial pressure to densify materials, align structures, and bond interfaces.⁷¹ Its core functions are: (1) inducing plastic deformation to eliminate voids; (2) aligning anisotropic fillers using shear forces; (3) increasing crystallinity and phase purity through thermal activation.

In PECs, hot-pressing mechanically aligns fillers while optimizing the polymer matrix. Fu et al.⁷² demonstrated this with BaTi₂O₅ (BT2) nanorod/PVDF composites (Fig. 3b). Hot-pressing at 200 °C under 10 MPa aligned BT2 nanorods horizontally. The compression shear forces also boosted β-phase formation in PVDF (54% crystallinity after poling). This textured structure improved stress transfer and dielectric polarization, increasing the energy harvesting figure of merit (FOM = d₃₃ × g₃₃) by 300% and achieving a power density of 27.4 μW cm⁻³.

2.3. Electrospinning/electrospraying for 3D architectures

Electrospinning uses high-voltage fields to make ultrafine polymer fibers (50–500 nm diameter),^57,73–75 while electrospraying generates functional micro/nanoparticles.⁷⁶ Combining them builds precise 3D architectures. Li et al.⁷⁴ achieved this by alternating electrospun PVDF nanofibers (83.5% β-phase vs. ∼53% in pure PVDF) with electrosprayed CsPbBr₃@PVDF beads (Fig. 3c). The composite reached a piezoelectric constant d₃₃ of ∼38 pC N⁻¹ (vs. ∼25 pC N⁻¹ for pure PVDF) and a sensitivity of 10.3 V (vs. 2.1 V for pure PVDF). Similarly, 3D honeycombs scaffolds for biomedical use combine aligned polycaprolactone (PCL) fibers with bioactive glass nanoparticles.⁷⁷ Thermal systems using BN-modified fibers and AgNWs reached 8.38 W m⁻¹ K⁻¹ conductivity.⁷⁸

Achieving uniform deposition with stochastic dynamics remains a challenge. New solutions like automated multi-axial systems and machine learning for parameter optimization⁷⁹ are bridging lab innovation to industrial production. This highlights a trend: advanced manufacturing now focuses on integrating system-level functionality alongside structural complexity.

2.4. Spin-coating for layered interfaces

Spin-coating uses centrifugal force from high-speed substrate rotation to deposit uniform thin films.⁸⁰ The process involves: (1) dispensing the solution, (2) rotational spreading to form a wet film, and (3) solvent evaporation for solidification. Adjusting spin speed and solution viscosity controls film thickness (10 nm–10 μm).

Hu et al.²³ made a double-layered BaTiO₃ (BTO)/PVDF film using spin-coating. They overlaid a high-filler layer (20 vol% BTO) with a pure PVDF layer (Fig. 3d). This design improved mechanical flexibility (18.12% strain vs. 5.27% for single-layer films) and piezoelectric output (6.7 V) by reducing interface defects and increasing charge accumulation at the boundary. This shows spin-coating's ability to balance mechanical flexibility and functional performance through controlled layering.

Challenges like edge bead effects and uneven solvent evaporation are being tackled with roll-to-roll methods,⁸¹ plasma pretreatment,⁸² and AI-assisted optimization,⁸³ showing how classic techniques evolve through interdisciplinary innovation.

2.5. Additive manufacturing

Additive manufacturing (AM), or 3D printing, builds 3D objects layer-by-layer using digital model data.⁸⁴ It slices a computer-aided design (CAD) model into 2D cross-sections and sequentially deposits materials (polymers, metals, ceramics) to achieve complex, custom geometries.

3D printing offers unmatched design freedom for PECs. Dong's team⁸⁵ created stretchable silver/ceramic lattices for wearables. Zhou et al.⁸⁶ 3D-printed kirigami-inspired BTO/P(VDF-TrFE) nanogenerators (Fig. 3e) with 300% stretchability and 6 V output. Although ink stability and resolution pose challenges,⁸⁴ AM's ability to merge complex shapes with functional materials makes it key for next-generation flexible electronics. This shows how fabrication innovation expands application potential.

2.6. Hydrothermal growth

Hydrothermal growth synthesizes crystals in sealed high-temperature/pressure water solutions.⁸⁷ Temperature gradients dissolve precursors and drive recrystallization onto substrates. This provides precise control over morphology (e.g., nanorods), with low-energy processing, high crystallinity, and no need for sintering.

In piezoceramics, this method is popular for its low-temperature control and adjustable morphology. Choi et al.³⁴ made fully transparent nanogenerators by hydrothermally growing ZnO nanorod arrays on flexible indium tin oxide (ITO)/polyether sulfone (PES) substrates at 95 °C (Fig. 3f). The vertically aligned nanorods (1.5–2 μm long) worked with Pd/Au bump electrodes to collect current efficiently (10 μA cm⁻²). The device achieved 70% transparency (92% array transmittance), suitable for touchscreens.

Vertically aligned ZnO nanorods on flexible polymers boost energy output by combining semiconducting and piezoelectric effects. Hybrid systems (e.g., PVDF/ZnO) amplify performance further.⁸⁸ While ZnO dominates due to tunable shape and scalability,⁸⁹ new materials like Pb(Mg_1/3Nb_2/3)O₃–PbTiO₃ (PMN–PT)⁹⁰ drive research into flexible, high-response composites.⁹¹

2.7. Integration of multiple methods: hierarchical composites

Combining multiple techniques creates hierarchical structures. Guan et al.³¹ integrated electrospinning and ultrasonic adsorption to make PDA-modified BTO@P(VDF-TrFE) fibers (Fig. 3g). This gave 25.6 V output and survived 10 000 cycles. Similarly, porous PVDF/BTO foams made with salt leaching and two-step blending used space charge electrets to boost d₃₃ by 28.5%.⁹² Non-covalent strategies, like glycine-MoS₂/PVDF composites, enable self-poled alignment with 213.3 MPa strength.⁹³ These hybrid approaches overcome limitations like filler clumping and boost performance through synergy, although scale-up and uniformity remain challenging.

Future progress will likely combine computational modeling, automated processes, and sustainable materials to solve these problems. As fabrication technologies mature, their convergence with smart design will redefine possibilities in energy harvesting, biomedical devices, and more. Ultimately, this proves that the roadmap for PEC advancement lies not just in materials, but critically in the processes shaping them.

3. Evolution of piezoelectric composite gas sensor fabrication

Piezoelectric composite (PEC) gas sensor performance relies heavily on fabrication advances. Methods progressed from basic homogeneous films to sophisticated 3D architectures. Early techniques focused on uniform filler dispersion and core functions. Modern methods prioritize intricate designs and hybrid materials to boost multifunctional integration. Unlike standard PEC fabrication, gas sensors specifically integrate gas-sensitive materials (like metal oxides or conductive polymers) and require precise control over piezoelectric–chemoresistive interfaces. Advances in PEC processing—filler dispersion, interfacial engineering, and structural complexity—directly enable sensor improvements. Below, we analyze six key fabrication methods (Fig. 4a–f) that enhance sensitivity, selectivity, and self-powered operation.

Fig. 4 Fabrication process and characterization of piezoelectric composite-based gas sensors. (a) Freestanding doped polyaniline–polyvinyl alcohol composite film. (b) Schematic diagram of a sensing element composed of two crossing carbon fibers (CFs) functionalized with ZnO nanowires (NWs) and its corresponding SEM image. Reproduced with permission from ref. 95. Copyright 2017, Elsevier. (c) Chemical deposition manufacturing ZnO nanowire-based nanogenerator (NG) can function as a self-powered gas sensor. Reproduced with permission from ref. 27. Copyright 2013, IOPscience. (d) 3D printing techniques in gas sensor devices: Vat polymerization, Material extrusion, Material jetting. Reproduced with permission from ref. 27. Copyright 2013, IOPscience. (e) Schematic diagram of a core–shell structured smart textile along with its SEM and FTIR characterizations. Reproduced with permission from ref. 60. Copyright 2023, Elsevier. (f) Fabrication process of a porous composite sensor involving: Sacrificial template method for preparing porous PZT/PDMS films; in situ polymerization of PAN coatings; corona polarization. Reproduced with permission from ref. 9. Copyright 2024, Elsevier.

3.1. Solution casting

Building on traditional methods, Mahato et al.⁹⁴ made PANI/PVA membranes via low-temperature oxidative polymerization and solution casting (Fig. 4a). Doping PANI with camphorsulfonic acid (CSA) or L-aspartic acid (ASP) tuned conductivity (9.30 × 10⁻³ S cm⁻¹) and delivered 49% sensitivity to methanol at 200 ppm. While simple and mechanically stable, humidity remains an issue—a challenge later solved by hybrid methods.

3.2. Electrospinning-chemical deposition synergy

To improve environmental robustness, Calestani et al.⁹⁵ fused electrospun carbon fibers (CFs) with ZnO nanowires (NWs) using ionic layer adsorption and hydrothermal growth (Fig. 4b). The brush-like NW arrays (50 μm⁻²) enabled dual-mode operation: chemiresistive ethanol detection (46% response at 50 ppm) and piezoelectric strain monitoring. Self-heating CFs removed external heaters, achieving both toughness and versatility in structurally integrated sensors.

3.3. Chemical deposition nanowire as self-powered sensors

Xue et al.²⁷ created self-powered sensors using wet-chemical ZnO nanowire (NW) nanogenerators (Fig. 4c). Vertically aligned NWs (5 μm) on Ti foil showed H₂S-dependent output decay (0.45 V to 0.198 V at 1000 ppm), proving gas adsorption alters surface charges. Encapsulation tests confirmed free-carrier screening's role. This foundational work enabled later functionalized nanowire sensors, demonstrating how synthesis precision drives energy-autonomous detection.

3.4. Additive manufacturing for programmable designs

Zhou et al.⁹⁶ summarized 3D printing's role in designing gas sensors. Vat polymerization, material extrusion, and jetting built hierarchical structures like Menger sponges and Fe₂O₃–CuO junctions (Fig. 4d). Direct ink writing (DIW)-printed SnO₂ scaffolds improved acetylene sensitivity (150% at 100 ppm) by optimizing porosity. Integrated microfluidic channels and resonant cavities show how 3D printing breaks geometric constraints to enhance gas diffusion and signal accuracy.

3.5. Electrospinning for core–shell smart textiles

Chen et al.⁶⁰ bridged wearability and function using coaxial electrospinning (Fig. 4e). A PVDF/PZT core provided piezoelectricity; a PANI/PVP shell detected NO₂. Co-spinning at 19 kV created breathable textiles with dual pressure-gas sensing. The sensor detected 100 ppb NO₂ and showed 57 mV kPa⁻¹ pressure sensitivity—thanks to how PANI's conductivity shifts during gas adsorption. Electrospinning's versatility supports next-gen textiles, though throughput limits⁷³ require scaling solutions.

3.6. Integration of multiple methods for ternary ordered assemblies

Li et al.⁹ combined techniques to make a porous PZT/PDMS composite via sacrificial sugar templates, then polymerized polypyrrole (PPy) in situ (Fig. 4f). The ternary system merged PZT (energy conversion), PPy (NH₃ sensing), and a 40% porous polydimethylsiloxane (PDMS) matrix (better gas diffusion). Corona polarization (20 kV, 12 h) aligned PZT domains, achieving 4.29% ppm⁻¹ NH₃ sensitivity. Simulations confirmed porosity boosts piezoelectric polarization—but complex assembly and humidity remain challenges.⁶³

3.7. Emerging frontiers

Progress from simple films to ternary composites shows how fabrication advances reshape sensors. Early methods set functional baselines; modern hybrid strategies leverage 3D designs and material synergies to overcome past limits. Scalability (especially in electrospinning), environmental stability, and reproducibility now push research toward machine-learning optimization⁸² and sustainable materials. As additive manufacturing and bioinspired designs mature, they will further blur boundaries between structural precision and multifunctional performance, making piezoelectric gas sensors smarter and more capable.

4. Functionalized nanowire piezoelectric (FNP) gas sensors

FNP gas sensors convert mechanical energy (vibrations, pressure) into electrical power while detecting gases.⁹⁷ Zinc oxide (ZnO) nanowires generate electricity when bent or compressed, powering the sensor. Gas molecules alter this electrical signal by transferring charge⁴⁰ or reacting on the nanowire surface.²⁷ This eliminates the need for external power during sensing.⁹⁸

As the most established piezoelectric sensing approach, FNP sensors achieve high sensitivity and selectivity using noble metal dopants (Pd, Pt), coatings (ZnSnO₃), or heterojunctions (NiO/ZnO). However, current research prioritizes material enhancements like doping methods over unresolved real-world challenges. Critically, self-powering applies only to sensing—auxiliary components (wireless transmitters, data processors) still need external power, limiting system autonomy.

4.1. Pure ZnO

Pure ZnO nanomaterials use their inherent piezoelectric and gas-sensing properties in self-powered systems. For example, vertically aligned ZnO nanowires (NWs) on titanium foil create a piezoelectric nanogenerator (PENG) (Fig. 5a). Xue et al.²⁷ first used this design to detect H₂S (Fig. 5b). When compressed, the ZnO NWs generate a piezoelectric signal that both powers the sensor and carries the gas signal. H₂S molecules reduce the output voltage by blocking surface charges. The sensor achieved 13.1% sensitivity to 100 ppm H₂S, working consistently up to 1000 ppm.²⁷ This work established ZnO NWs as key for self-powered sensors. Flexible ZnO nanosheets⁹⁹ extend this capability to humidity sensing. By relying on humidity-dependent charge screening effects, these sensors work at room temperature across 10–95% relative humidity (RH), with 2.96% sensitivity at 10 ppm.

Fig. 5 Material structure evolution and working mechanisms of deposited ceramic nanowire gas sensors. (a) Pure ZnO nanowire arrays on Ti foil substrates first demonstrated piezoelectric-gas sensing coupling for self-powered H₂S detection; (b) Piezoelectric output serves dual roles (power supply and sensing signal), with gas adsorption directly modulating signal amplitude through free carrier density variation. Reproduced with permission from ref. 27. Copyright 2013, IOPscience. (c) Pd/ZnO nanoarrays feature Pd nanoparticles on ZnO nanowires with PDMS encapsulation; (d) Pd-catalyzed ethanol oxidation synergizes with piezoelectric effects under mechanical stress, dynamically tuning output via electron density modulation. Reproduced with permission from ref. 39. Copyright 2013, Royal Society of Chemistry. (e) Core–shell ZnSnO₃-coated ZnO nanowires enhance LPG sensing; (f) ZnSnO₃ provides Lewis acid sites and lowers activation energy, amplifying piezoelectric shielding through optimized LPG adsorption and charge transfer. Reproduced with permission from ref. 36. Copyright 2015, American Chemical Society. (g) NiO/ZnO heterojunctions fabricated via NiO deposition on ZnO nanowires; (h) H₂S reacts with NiO to form NiS, collapsing the p–n junction barrier and suppressing piezoelectric output via carrier density surge for specific detection. Reproduced with permission from ref. 102. Copyright 2016, Elsevier.

4.2. Noble metal-doped ZnO

Adding noble metals (Pd,³⁹ Pt,¹⁰⁰ Au,¹⁰¹ Cu³⁵) to ZnO nanostructures improves sensitivity and selectivity through catalytic activity. A Pd/ZnO nanoarray sensor has Pd nanoparticles on ZnO NWs grown on Ti foil (Fig. 5c). Under mechanical stress (Fig. 5d), the piezoelectric output changes because Pd speeds up ethanol oxidation, altering electron density at the ZnO surface. This design achieved a 25 s response time, 59 s recovery time, and detected 216 ppb ethanol at room temperature.³⁹ The combination of piezoelectricity and catalysis allows high-performance, self-powered sensing without external heating or UV light. Key performance metrics of various metal-doped FNP gas sensors are summarized in Table 1.

Table 1 Metal-doped FNP gas sensors

Material	Gas	Contribution	Enhancement strategy	Performance	Ref.
Cu–ZnO nanowires	H2S	Ultra-high room-temp H2S sensitivity	Cu catalytic activation	Sens.: 1045% (1000 ppm); Res./Rec.: 100/60 s	35
Au–ZnO nanowires	Ethanol	Linear ethanol response	Au-enhanced oxidation	Sens.: 72.1% (1200 ppm); linear: 200–1200 ppm	101
Pt–ZnO nanoarray	Ethanol	Selective ethanol detection	Pt catalytic acceleration	Sens.: 37.14% (1000 ppm); Sel. >5× (vs. methanol MeOH)	100
Pd/ZnO nanoarray	Ethanol	Sensing without external heating or UV activation	Pd catalytic effect and Schottky barrier formation	Sens.: 108% (800 ppm); output: 0.25 V; selectivity >5× (vs. MeOH, H2, etc.)	39

---

4.3. Sensitive material coating

Coating composite structures, like ZnSnO₃ on ZnO NWs, improves sensing through reactive interfaces (Fig. 5e). ZnSnO₃ offers Lewis acid sites that adsorb liquefied petroleum gas (LPG). Under mechanical stress (Fig. 5f), this composite shows 83.23% sensitivity to 8000 ppm LPG, far exceeding pure ZnO.³⁶ Low reaction activation energy in ZnSnO₃ and strong piezoelectric screening enable this room-temperature operation.

4.4. Heterojunction structures

Heterojunctions use interfacial energy barriers for selective gas detection.^{37,38,102–104} A NiO/ZnO NW sensor has NiO nanocones forming a p–n junction with ZnO (Fig. 5g). Exposing it to H₂S (Fig. 5h) causes NiO to form NiS, collapsing the junction into an ohmic contact. This increases electron density, reducing piezoelectric output by ∼85% at 800 ppm H₂S.¹⁰² It detects H₂S from 10–1000 ppm rapidly (<100 s), showing promise for real-time monitoring. Table 2 provides a comparison of performance characteristics and mechanisms for different heterojunction-based FNP gas sensors.

Table 2 Heterojunction FNP gas sensors

Material	Gas	Contribution	Enhancement strategy	Performance	Ref.
CuO/ZnO PN-junction	H2S	PN → Ohmic transition for sensitivity boost	H2S-induced interfacial conversion	Sens.: 629.8% (800 ppm); DL: 50 ppm	38
p-Si/n-ZnO (SAM)	NO2	Visible-light-driven NO2 sensing	SAM-guided surface chemistry	DL: 750 ppb; Sens.: 23.5% (amine), 12.8% (thiol)	37
NiO/ZnO nanowire	H2S	PN → Ohmic transition for ultra-high H2S sensitivity	H2S-induced NiO → NiS conversion and interface engineering	Sens.: 536% (1000 ppm); Res. time: 15 s; Rec. time: 10 min (60 °C/sunlight)	102
α-Fe2O3/ZnO nanowires	Ethanol	Heterojunction-enhanced ethanol response	Charge separation at interface	Sens.: 706.8% (700 ppm); Res. <60 s	103
SnO2/ZnO nanoarray	H2	Synergistic H2 sensing	Band engineering	Sens.: 471.4% (800 ppm); DL: 100 ppm	104

---

4.5. Non-ZnO materials

Non-ZnO materials expand FNP sensing with unique chemical interactions and structures. Flexible CdS nanorods detect H₂S:⁸ sulfur vacancies on CdS enhance adsorption while maintaining mechanical durability. Sensitivity is 62.5% at 600 ppm H₂S, with stable performance over 1000 bends. For humidity, CeO₂/ZnO heterostructures use rare-earth oxide chemistry:¹⁰⁵ CeO₂ surface hydroxyl groups actively adsorb water. Sensitivity reaches 82.1% at 95% RH across 20–95% RH. This merges ZnO's piezoelectricity with CeO₂'s hygroscopicity for new self-powered environmental monitors.

4.6. Perspectives

Current research focuses heavily on material innovation. Yet challenges like standardizing signal amplification and reducing noise limit IoT compatibility. Future work must combine material design with low-power electronics – like efficient transmitters, transducers, or smart algorithms – to enable practical, large-scale sensor networks. Interdisciplinary efforts (materials, electronics, data analysis) are vital to move from lab prototypes to real applications, such as smart environmental monitors or wearable health devices.

5. Evolution of decoupled piezoelectric composite gas sensors

Decoupled piezoelectric composite (PEC) gas sensors separate power generation and sensing into distinct modules—physically and functionally. This reduces signal interference and allows flexible design.¹⁰⁶ Early decoupled designs minimized interference but faced three problems: bulky size, poor energy efficiency,⁴² and complex assembly. Recent advances aim to keep functional independence while making systems smaller and more integrated. This chapter traces three key stages: modular separation, co-substrate integration, and multimodal isolation, each solving earlier problems while keeping core advantages (Fig. 6a–h).

Fig. 6 Evolution of decoupled piezoelectric-gas sensing architectures. (a) A flexible self-powered NH₃ sensor based on Au-decorated MoSe₂ nanoflowers, driven by a MoS₂-based piezoelectric nanogenerator (PENG); (b) Solvothermal synthesis of MoSe₂ nanoflowers with surface-anchored Au nanoparticles (NPs). Reproduced with permission from ref. 61. Copyright 2019, Elsevier. (c) MXene/Co₃O₄ composite-based formaldehyde sensor; (d) Humidity interference correction via compensation model and SEM/TEM characterization for enhanced environmental adaptability. Reproduced with permission from ref. 42. Copyright 2021, Elsevier. (e) SnO₂/MXene nanocomposites enabling high-sensitivity NO₂ detection with ultralow detection limit; (f) Co-planar integration of gas-sensing and piezoelectric units on a shared substrate. Reproduced with permission from ref. 62. Copyright 2023, American Chemical Society. (g) Dual-modal sensor with independent PVDF/CNT (piezoelectric) and PAAS/CNT (humidity-sensing) modules to eliminate cross-talk; (h) SEM validation of gas-sensing/piezoelectric structural morphology and interfacial properties. Reproduced with permission from ref. 63. Copyright 2025, Elsevier.

From bulky modules to co-substrate and multimodal designs, these systems now better balance compactness with independence. Early modular designs prevented signal interference but needed complex wiring. Co-substrate systems shrank device size but created new interference issues. Latest multimodal designs use isolated signal patterns for multiple sensing functions without interference—and no signal degradation. Yet research still focuses too much on materials. Scalable manufacturing and electronics integration get little attention. For real-world use, future efforts must link nanomaterial advances with circuit design to create efficient components and standardized interfaces.

5.1. Early decoupled architectures

Early systems used physically separate piezoelectric and sensing modules connected with wires.^42,61 Zhang et al. (2019) powered an Au–MoSe₂ ammonia sensor with a MoS₂-based piezoelectric nanogenerator (PENG) (Fig. 6a and b), achieving 18 s response and 16 s recovery times.⁶¹ Later, their 2021 work⁴² used ZnO/MXene nanowires (Fig. 6c) to drive an MXene/Co₃O₄ formaldehyde sensor, with algorithms cutting humidity interference (Fig. 6d). These proved decoupling works but had to sacrifice size and energy efficiency. Unresolved issue: manufacturing flaws—like unstable MXene oxidation or uneven MoS₂ layers—still block large-scale production.^61,73

5.2. Co-substrate integration strategies

To reduce bulkiness of decoupled PEC gas sensors, co-substrate designs merged modules on shared flexible substrates. Gasso and Mahajan⁶² integrated a MXene/SnO₂ nanofiber sensor (Fig. 6e) on a single substrate (Fig. 6f), achieving a 0.03 ppb NO₂ detection limit and stable performance under humidity variations. Although this approach improved portability, it introduced cross-modal interference between modules, highlighting the need for better isolation methods in compact systems.

5.3. Multimodal integration with signal isolation

The latest designs leverage isolated signal types for multifunctionality while maintaining decoupling. Yang et al.⁶³ created a dual-modal sensor using physically separate resistive (humidity) and voltage (pressure) signal pathways (Fig. 6g). The PAAS/CNT humidity film gave a 1100% resistance shift (23–98% RH) with minimal hysteresis. Simultaneously, the polyacrylate sodium (PAAS)/carbon nanotube (CNT) piezoelectric film delivered 0.11 V kPa⁻¹ sensitivity (3–115 kPa, Fig. 6h). This design prevents cross-talk without algorithm compensation, establishing a new approach for multimodal integration. The breakthrough shows how signal isolation enables multifunctionality while preserving decoupled architectures’ essential benefits.

5.4. Emerging frontiers

Despite progress, a gap persists between lab prototypes and practical devices. Research still emphasizes optimizing materials, but unreliable nanomaterial synthesis and poor interfacial durability prevent consistent reproduction. Furthermore, practical integration demands dedicated circuits for managing signals, energy, and interference – components complex and costly to develop.^107–109

Future efforts must bridge materials science and electronics engineering. Success means creating adaptive compensation modules, low-power transducers, and standardized manufacturing methods. Only through this combined approach can decoupled sensor architectures move beyond theory and deliver reliable solutions for industrial safety and environmental monitoring.

6. Integration of piezoelectric driving module and gas sensor

Piezoelectric energy harvesting and gas sensing can now merge into a single composite material—a major advance over separate designs. This requires careful coordination of how electrical and mechanical forces interact within materials. Unlike decoupled PEC gas sensors with distinct modules, these coupled systems combine energy capture and sensing in one material. This creates inherently compact gas-sensitive piezoelectric devices. Unlike rigid nanowire sensors (FNP), coupled PEC sensors mix flexible polymers with inorganic ceramics. Adding organic polymers complicates design (e.g., polarizing materials), but delivers crucial flexibility FNP systems lack.

Progress—from simple blends to sophisticated layered structures—has boosted performance through precisely designed material interactions. Yet challenges remain in balancing electrical sensitivity, chemical stability, and durability. Future progress needs computational modeling plus experiments to optimize these systems.

6.1. Microscale integration: material coupling

Coupled PEC sensors work by tightly integrating piezoelectric and gas-sensitive materials microscopically. Early attempts used material mixtures for simplicity. Lin et al.¹¹⁰ blended tetrapod ZnO (T-ZnO) with polyvinylidene fluoride (PVDF) for an oxygen-sensing mask (Fig. 7a). T-ZnO did double duty: (1) sensing oxygen by changing electron density when gas adsorbed, and (2) generating piezoelectric power when bent (Fig. 7b). PVDF bound the structure and enhanced piezoelectric output. Oxygen adsorption lowered T-ZnO's free electrons, weakening charge screening and amplifying piezoelectric voltage (0.1–0.3 V from breathing). This enabled battery-free lung monitoring. However, crude interfaces limited sensitivity and reliability.

Fig. 7 Integration strategy of self-powered gas sensor and piezoelectric driving module. Random structural: (a) T-ZnO nanostructure/PVDF composite sensor, (b) synergy between T-ZnO gas-sensitive response and PVDF piezoelectric output, achieving self-driven sensing signals through interfacial charge transfer. Reproduced with permission from ref. 110. Copyright 2022, IOPscience. Intra-material microstructure: (c) Sm-PMN-PT/PEI composite system: high-response piezoelectric ceramic filler (Sm-PMN-PT) embedded in hydrophilic PEI matrix; (d) physical images of composite structure with SEM/TEM interface analysis. Reproduced with permission from ref. 11. Copyright 2023, Royal Society of Chemistry. Layered mesostructure: (e) PZT/PPy porous composite sensor: PZT nanofillers provide piezoelectricity, PPy coating enables ammonia sensitivity, porous structure optimizes gas diffusion and piezoelectric conversion efficiency; (f) SEM analysis. Reproduced with permission from ref. 9. Copyright 2024, Elsevier. Device macrostructure: (g) PVDF/PANI corrugated sensor; (h) Fabrication process. Reproduced with permission from ref. 111. Copyright 2018, Springer Nature.

To control interfaces better, Su et al.¹¹ made humidity-sensitive textiles by electrospinning hydrophilic polyetherimide (PEI) with samarium-doped PMN-PT (Sm-PMN-PT) piezoceramic (Fig. 7c). Sm-PMN-PT harvested energy while PEI's amine groups (–NH₂) grabbed water molecules. Humidity changes altered PEI's dielectric properties, shielding the piezoelectric field. This achieved self-powered sensing (0.9%/RH sensitivity; 20 s response). The porous fiber network survived >7500 bends and allowed moisture penetration (Fig. 7d).

6.2. Mesoscale integration: hierarchical and porous architectures

Researchers designed porous and core–shell structures to optimize performance. Hierarchical porosity became key for fast charge and gas transport. Li et al.⁹ made a porous PZT/PDMS/PPy composite for ammonia. PZT generated piezoelectric power inside PDMS, while polypyrrole (PPy) coated pores as the gas sensor (Fig. 7e). Ammonia adsorption lowered PPy's conductivity, reducing shielding of PZT's piezoelectric field. With 20–40% porosity, sensitivity reached 4.29% ppm⁻¹, balancing gas access and energy conversion (Fig. 7f).

Core–shell fibers gave finer control. Chen et al.⁶⁰ used coaxial electrospinning to make fibers with a PVDF/PZT piezoelectric core and a PANI/PVP sensing shell (Fig. 4e). Pressure activated the core. NO₂ adsorption raised PANI's dielectric constant, shielding the piezoelectric potential. Optimized structure (8 μm core/16 μm shell) achieved 0.32 ppm⁻¹ sensitivity and 100 ppb limit of detection (LOD).

6.3. Device-scale integration: macroscopic heterostructures

At device scale, smart geometry enables real-world use. Fu et al.¹¹¹ built a foldable bellows breath sensor with polyaniline (PANI) electrodes on a PVDF film (Fig. 7g). Breathing airflow bent the PVDF bellows, generating power. PANI's resistance shifted with gases like ethanol or NOx. Five electrodes with unique dopants detected multiple gases simultaneously. PANI's resistance controlled PVDF's power output, giving humidity-resistant operation (40–100% RH) at 57 mV kPa⁻¹ sensitivity (Fig. 7h).

6.4. Emerging frontiers

Barriers remain for practical use. Getting electromechanical parts and gas-sensitive reactions to work perfectly demands precision at all scales. Testing combinations wastes time, like fine-tuning Sm-PMN-PT/PEI humidity responses¹¹ or PANI/PVDF electrical interplay.¹¹¹ Without predictive models for changing conditions (mechanical stress, humidity shifts), reliability suffers. To overcome these barriers, computational materials science is essential. Multiscale simulations can virtually prototype layered designs. Linking atomic to device scales should fast-track better PEC gas sensors, discussed next.

7. Multiscale simulation for rational design of gas sensors

Multiscale simulations—spanning quantum to continuum scales—are essential for probing multiphysics coupling and guiding high-performance sensor design. Although seamless integration across all scales remains challenging, hierarchical modeling provides a practical solution: distinct physical and mathematical frameworks operate independently at each scale.

For example, quantum mechanics (e.g., DFT) and molecular dynamics (MD) resolve atomic adsorption dynamics and interfacial charge transfer. These results then parameterize mesoscale phase-field models, whose outputs (e.g., effective dielectric constants) define boundary conditions for macroscopic finite element method (FEM) analyses. This approximation-driven workflow balances accuracy with computational efficiency, avoiding the prohibitive cost of fully unified frameworks. This chapter highlights advances in modeling adsorption dynamics, interfacial effects, microstructure, and device optimization (Fig. 8).

Fig. 8 Multiscale simulation techniques spanning micro-to-macro levels. Density functional theory (DFT): (a) first-principles DFT calculations on Au-MoSe₂ composites for NH₃ adsorption analysis via charge density/DOS; (b) DFT study of band structure changes in Au–MoSe₂ composites. Reproduced with permission from ref. 61. Copyright 2019, Elsevier. (c) DFT evaluation of COCl₂ adsorption energy and electronic properties on AlNNT. Reproduced with permission from ref. 112. Copyright 2016, Springer Netherlands. Molecular dynamics (MD): (d) MD tracking of COCl₂ adsorption/diffusion dynamics on AlNNT; (e) MD analysis of Ti₃C₂T_x–OH and fluoropolymer matrix interactions. Reproduced with permission from ref. 26. Copyright 2022, Springer Nature. Phase-field simulation (PFS): (f) phase-field simulation of water adsorption and proton hopping in fabrics. Reproduced with permission from ref. 11. Copyright 2023, Royal Society of Chemistry; (g) phase-field modeling of gas sensing under NH₃ gradients. Reproduced with permission from ref. 9. Copyright 2024, Elsevier. Finite element method (FEM): (h) FEM analysis of shielding effects in piezoelectric transduction; (i) MATLAB-based FEM simulation of M–S–M structures. Reproduced with permission from ref. 43. Copyright 2015, Elsevier; (j) electric field shielding model for core–shell sensors. Reproduced with permission from ref. 60. Copyright 2023, Elsevier.

7.1. Microscopic scales: quantum mechanics to molecular dynamics

Quantum mechanical calculations and molecular dynamics simulations combine to analyze atomic charge transfer and molecular interface dynamics. Density functional theory (DFT) resolves electronic structure changes and energy evolution during adsorption, while MD reveals dynamic behaviors and interfacial mechanisms. Together, they quantitatively predict binding energy, charge transfer efficiency, and gas adsorption selectivity for sensing materials.^61,112

For instance, DFT-derived binding energies inform MD simulations of gas diffusion,^113,114 which feed into phase-field polarization models.^9–11 Here, MD simulates representative volumes (e.g., hundreds of molecular chains) to calculate the time- and spatial-average polarization vector in specific sub-domains. This averaged polarization (P) is assigned directly to corresponding materials phases in the phase-field grid.²⁶ However, this simplification can obscure nonlinear polarization details.

7.1.1. Quantum mechanics for adsorption and charge transfer. DFT uses electron density—not multi-electron wavefunctions—to study electronic structures, simplifying calculations while maintaining accuracy.¹¹⁵ In gas sensing, DFT analyzes atomic-scale interactions like charge transfer and orbital hybridization during adsorption. For Au–MoSe₂ composites (Fig. 8a and b), DFT revealed enhanced NH₃ adsorption via Au-induced charge transfer, evidenced by Mo-4d/Au-4f orbital hybridization (−0.18 to −0.05 eV).⁶¹ Similarly, COCl₂ adsorption on AlN nanotubes (AlNNTs, Fig. 8c) reduced the bandgap (4.407 to 3.008 eV) via weak Al–O van der Waals interactions.¹¹²

7.1.2. Molecular scale dynamics and interface engineering. MD simulates atomic/molecular motion to capture adsorption kinetics and interfacial interactions.^116,117 In Ti₃C₂T_x MXene/PVDF composites (Fig. 8e), hydrogen bonding between –OH groups and PVDF chains boosted β-phase alignment, increasing vertical polarization by 20% (371.42 D vs. 308.13 D).²⁶ For CNT@BNNT heterostructures, MD quantified piezoelectric frequency tuning (∼50 GHz at ±10 V nm⁻¹) and identified thermomechanical damping as the primary energy dissipation mechanism.³³ MD also confirmed selective COCl₂ adsorption on AlNNTs (Fig. 8d; radial distribution peak at ∼1 nm) over CO₂/NO₂.¹¹² Combined DFT/MD linked COCl₂ selectivity to atomic charge transfer and mesoscale diffusion.

7.2. Mesoscopic scale: phase-field methods from polarization domains to functional units

Phase-field methods, grounded in thermodynamics and kinetics, simulate mesoscale morphology evolution.^118–122 They use conserved field variables (e.g., concentration) and non-conserved order parameters (e.g., polarization) to model microstructural changes without explicit interface tracking. Conserved variables obey mass conservation; non-conserved parameters quantify structural ordering, enabling predictive modeling under thermodynamic driving forces.¹¹⁹

7.2.1. Domain-scale and meso-structure. For ferroelectric polycrystals, phase-field simulations resolve polarization domain evolution and domain-wall dynamics under external stimuli.^119–122 In humidity-sensitive textiles (Fig. 8f), adsorbed water layers (relative permittivity ε = 79) reduced piezoelectric potential by 40% at high humidity, per phase-field dielectric shielding models.¹¹

For composites (e.g., fiber-reinforced matrices, porous architectures), phase-field modeling quantifies effective property evolution during phase transitions—like dielectric degradation or mechanical anisotropy.^9,19,26 These analyses bridge atomistic polarization mechanisms (e.g., dipole reorientation) with macroscopic performance, closing a gap left by traditional mesoscale-blind methods.^29,45 In gas sensors, phase-field models optimized porosity (40%) in ternary composites (Fig. 8g) to balance stress concentration and piezoelectric coefficients.

7.2.2. Numerical implementation. While phase-field models define the physical framework, their numerical solution employs tailored algorithms. Although finite element method (FEM) is widely used for multiphysics coupling, Fourier spectral perturbation iterative methods are often preferred in phase-field simulations due to their computational efficiency in handling periodic boundary conditions and gradient energy terms.^45,60,123 This highlights the complementary roles of physical modeling (phase-field) and numerical discretization (FEM, spectral methods) in multiscale simulations.

7.3. Macroscopic scale: multiphysics FEM for sensor response optimization

FEM couples’ structural mechanics, fluid dynamics, and acoustics to simulate complex physics and optimize macroscopic sensor performance.¹²⁴ In piezoelectric-chemical models, FEM links geometry to performance. Li et al.⁹ showed that NH₃-triggered dielectric reduction in polypyrrole coatings enhances field penetration (Fig. 8h). For ZnO Schottky junctions, equivalent circuit modeling⁴³ (Fig. 8i) proved compressive strain lowers Schottky barrier height (ΔΦ) by 0.15 eV, boosting current response by 5359% in vacuum or gases (H₂, NO₂). Chen et al.'s 60 core–shell sensor model (Fig. 8j) demonstrated that NO₂ adsorption increases shell dielectric constant, slashing piezoelectric output by 60%. Critically, Li et al.'s⁹ NH₃ adsorption data and Chen et al.'s⁶⁰ NO₂ adsorption metrics for FEM relied on mesoscale phase-field simulations—highlighting multiscale collaboration.

7.4. Future perspectives

Future efforts should focus on optimizing parameter transfer connecting multiple scale models (e.g., machine learning-assisted surrogate models^125,126) and correcting digital twin models based on experimental benchmarks. By embracing scale-specific approximations, computational tools will accelerate sensor design, bridging atomic insights and macroscopic functionality without overburdening computational resources. The applicable scenarios and inherent limitations of the key simulation methods discussed in this chapter are systematically compared in Table 3, providing a concise reference for sensor design.

Table 3 Applicable scenarios and limitations of simulation methods

Method	Scale	Applications	Key limitations
DFT	Atomic	– Gas adsorption energy	– Small systems
		– Electronic structure (band/DOS)	– Static approximations
		– Charge transfer	– Neglects thermal/dynamic effects
MD	Molecular	– Adsorption kinetics	– Short timescales
		– Interfacial interactions	– Force field dependency
		– Piezoelectric tuning	– Missing quantum effects
PFS	Mesoscopic (μm)	– Polarization domain dynamics	– Parameter transfer errors
		– Microstructure optimization	– Simplified nonlinearity
		– Property prediction	– Requires order parameters
FEM	Macroscopic (Device)	– Multiphysics coupling	– Loses mesoscale details
		– Device structural optimization	– Mesh sensitivity
		– Field distribution	– Complex geometry setup

---

8. Circuit design strategies for portable PEC gas sensors

Piezoelectric composite (PEC) gas sensors differ fundamentally from piezoelectric acoustic wave sensors. Their gas-sensitive signals overlay piezoelectric outputs from mechanical stimuli,^{9,12,20,49,60,78} complicating signal separation compared to resonant frequency-based acoustic devices. Unlike micro-electro-mechanical system (MEMS)-based acoustic wave sensors,^3,127–132 PEC gas sensors lack standardized manufacturing, causing inconsistent back-end circuit designs.

Development faces knowledge asymmetry: material researchers lack circuit expertise, while engineers struggle with PECs’ variable outputs like nonlinear impedance¹³³ and charge density shifts.¹³⁴ Closing this gap requires cross-disciplinary collaboration and standardized signal conditioning.

Current testing relies on lab instruments (Fig. 9a), but portable/wearable systems demand miniaturized circuits (Fig. 9b). Lab equipment (electrometers,⁷³ oscilloscopes²⁴) offers ultrahigh impedance (>200 TΩ), ultralow noise (<1 fA), and wide dynamic ranges – features hard to replicate portably due to power/size/cost limits.^11,109 This chapter examines portable circuit compromises and solutions.

Fig. 9 Compromises from laboratory-grade instrumentation to portable back-end circuits. (a) Device performance testing using an oscilloscope and electrometer, highlighting advantages of instrument-based measurements. Reproduced with permission from ref. 73. Copyright 2021, Springer Singapore; from ref. 24. Copyright 2025, Elsevier. (b) Key challenges in portable circuit design and corresponding solutions. (c) The bent structure design realizes milliwatt-level power output. And circuit design focus on rectification/storage, and power management, achieving efficient energy conversion and stable output through optimized architecture. Reproduced with permission from ref. 25. Copyright 2015, Elsevier. (d) A high-efficiency energy management circuit for impact-type piezoelectric energy harvesters, combining self-powered synchronized switch harvesting on inductor and adaptive impedance matching to significantly improve energy harvesting efficiency and output power. Reproduced with permission from ref. 108. Copyright 2022, Wiley-VCH.

8.1. Rectification and energy harvesting

Full-wave rectifiers¹³⁵ achieve 81% efficiency (vs. 40% for half-wave¹³⁶). Bennet's doubler circuits¹³⁷ passively double voltages using capacitors/diodes, amplifying voltage 10× and energy storage 100×. Active switches (mechanical triggers;¹³⁸ silicon-controlled rectifiers, SCR;¹³⁹ metal oxide semiconductor field effect transistor, MOSFET¹⁴⁰) synchronize with motion, boosting output currents 45× under matched impedance. These methods overcome piezoelectric outputs’ intermittency and low current. Synchronized switch harvesting on inductors (SSHI) by Ammar et al.¹⁴¹ hits 92% efficiency via inductor-piezoelectric resonance, ideal for low-frequency (<10 Hz) human motion.

8.2. Structural design for self-powered efficiency

Recent advances in portable circuit design bridge the performance gap with lab equipment. For example, curved piezoelectric generators with dual-arc polyimide substrates (Fig. 9c, right) focus mechanical stress on PVDF layers by shifting the neutral plane, boosting voltage fivefold.²⁵ Integrated circuits combine parallel PVDF units to amplify current output and use full-wave rectification with bq25504 power management chips (Fig. 9c, left). This delivers stable 3.3 V direct current (DC) in 10 seconds, achieving 22% mechanical-to-electrical conversion efficiency. Bio-inspired origami-flexible hinges¹⁴² keep 98% output stability under 120° bending—ideal for harvesting deformation energy in wearables.

8.3. Low-power readout circuits

New low-power readout circuits precisely interface emerging sensors. Puyol et al. designed an ultra-low-power resistive circuit for gas sensor matrices,¹⁴³ built in 180 nm complementary metal-oxide-semiconductor (CMOS). It covers a wide dynamic range (1 kΩ–33 MΩ) with 57 dB minimum SNR, critical for detecting ppb-level NH₃ and NO₂. Its relaxation oscillator architecture minimizes energy consumption, drawing ≤194 μA and using just 1.21 nJ per conversion. Key features include configurable bias voltages (50 mV–1 V) and hierarchical current mirrors that cut mismatch and flicker noise. The design works with diverse sensors (conductive polymers, graphene) and suits system on chip (SoC)-integrated internet of things (IoT) multi-sensor arrays. Subthreshold asynchronous analog-to-digital converters (ADCs) like Ippili et al.¹⁴⁴ cut static power to 72 nW by activating only during transitions, fitting intermittent gas monitoring. Time-domain signal processing using resistance-frequency converters and ring oscillators (Liotta et al.,¹⁴⁵ Park et al.¹⁴⁶) achieves 0.8% linearity across 0.5–100 kΩ while avoiding conventional amplifiers’ 1/f noise (Flicker Noise).

8.4. Energy management circuit design

Energy management circuits for impact harvesters (Fig. 9d) achieve 90.4% efficiency in low wind via self-powered synchronized switch harvesting on inductor (SP-SSHI) rectifiers and adaptive buck–boost converters,¹⁰⁸ extending field sensor runtime. SP-SSHI (Fig. 9d, bottom left) uses symmetric MOSFET topology instead of diodes. It recovers charge via inductor-capacitor (LC) oscillation, nearly doubling piezoelectric voltage with self-powered logic. For dynamic impedance matching (Fig. 9d, top right), an adaptive load network adjusts impedance (1–100 kΩ) during vibration cycles using Buck–Boost converters and relaxation oscillators. This delivers 294.2 μW output at 40 kΩ—6.1% higher than conventional approaches. Perovskite-PENG hybrids integrate dual-input bq25570 energy management chips with priority power multiplexing, achieving 87% system efficiency outdoors.¹⁴²

8.5. Dynamic compensation for precision

Dynamic compensation techniques enable precision in portable sensors.¹⁴⁷ Chopping and auto-zeroing cut 1/f noise and input offsets,¹⁴⁸ reducing temperature sensor errors below 2 μV. The main error source involves saturation current variability and bias current deviations inherent in low-cost CMOS. Σ–Δ modulation with dynamic element matching (DEM)¹⁴⁹ gives Hall sensors 24-bit resolution, enabling sub-microtesla detection with minimal calibration. These methods trade bandwidth for precision, suiting low-speed applications like environmental monitoring. Thermal Σ–Δ modulators in smart wind sensors use the sensor's thermal capacitance as a loop filter,¹⁵⁰ maintaining ±0.1 °C accuracy across temperatures. Self-sensing MEMS resonators dynamically tune drive frequencies through in situ Q-factor monitoring, stabilizing gas sensors at ±0.1 ppm °C⁻¹.¹⁵¹

8.6. Perspectives

Despite innovations in rectification topologies, adaptive power management, and dynamic error cancellation enabling milliwatt-level harvesting and precision sensing in wearables/IoT, these remain underutilized in PEC gas sensor backend circuits. Recent advances show promise: Li et al.⁹ engineered porous architectures enabling adaptive impedance matching, while Su et al.¹¹ demonstrated humidity-responsive dielectric-tunable textile sensors. Future progress requires sophisticated circuit architectures combined with AI-driven signal extraction to bridge lab-portable performance gaps while maintaining sub-milliwatt operation.

9. Multi-disciplinary applications

PEC gas sensors show significant potential across industries despite being in development (Fig. 10). Research prototypes enable innovative solutions in safety, healthcare, environment, and agriculture through self-powering, flexibility, and hybrid compatibility. Emerging trends like energy autonomy, multimodal sensing, and biocompatible designs drive progress toward predictive monitoring systems.

Fig. 10 The application potential of piezoelectric gas sensors across multi-disciplinary. (a) Industrial safety: monitoring mining environments through body motion-driven piezoelectric nanogenerators (PENGs) to detect H₂S, CH₄, ethanol, and humidity in real time. Reproduced with permission from ref. 152. Copyright 2018, Elsevier. (b) Medical diagnosis: biocompatible amino acid-based piezoelectric materials capable of in vivo implantation and degradation. Reproduced with permission from ref. 154. Copyright 2021, American Association for the Advancement of Science. (c) Smart infrastructure: light-driven self-powered integrated nano sensor systems for indoor gas detection, featuring high-performance three-dimensional Pd/SnO₂ sensor arrays and wireless smart gas sensing networks. Reproduced with permission from ref. 157. Copyright 2021, American Chemical Society. (d) Precision agricultural: gas sensor in electronic nose (E-nose) applications to support agricultural cycles. Reproduced with permission from ref. 159. Copyright 2022, Elsevier. (e) Environmental monitoring: ammonia sensor capable of visual alarms for food storage or contamination detection scenarios. Reproduced with permission from ref. 218. Copyright 2020, Springer Nature.

9.1. Industrial safety: hazard detection

He et al.'s flexible e-skin¹⁵² uses coated ZnO nanowires (Pd/CuO/TiO₂) to detect ethanol (59.82% response), H₂S (79.27%), CH₄ (−87.50%), and humidity (87.76%) in mines. Harvesting energy from arm movements (Fig. 10a), it enables battery-free monitoring to prevent explosions. Ma et al.'s non-contact sensors¹⁵³ detect leaks up to 1 m away (57 pC kPa⁻¹ sensitivity), ideal for petrochemical plants. Future work needs universal IoT protocols and machine learning for extreme conditions (>80% humidity, pH <5).

9.2. Biocompatible disease management

γ-Glycine/PVA biofilms¹⁵⁴ achieve wafer-scale self-assembly and controlled biodegradability, with a high d₃₃ (5.3 pC N⁻¹) and voltage output (>150 mV) for implantable muscle/respiratory monitoring in rodents (Fig. 10b). Applications include smart wound dressings and absorbable cardiac sensors that eliminate secondary surgeries. Wearable MXene/MoS₂ sensors¹⁵⁵ detect diabetes-related acetone (5–30 ppm) via resistance changes. MoS₂'s piezoelectricity enhances sensitivity: mechanical bending generates polarization charges that lower the MXene/MoS₂ interface energy barrier, accelerating charge transfer. Dual-layer PLLA films¹⁵⁶ correlate with spirometers (0.99 R²) via self-shielded designs that suppress electromagnetic interference, enabling ambulatory metabolic tracking. Key challenges for clinical translation: improve selectivity, integrate functional coatings with piezoelectric, model gas-piezoelectric interactions in physiological systems, and validate biodegradable implants.

9.3. Smart infrastructure

A self-sustaining indoor gas monitoring network (SINGOR) network¹⁵⁷ combines Pd-decorated 3D SnO₂ sensors with triboelectric-solar hybrid harvesting (4.3 μW per module), achieving 1 ppb toluene/acetone detection. Principal component analysis-support vector machine (PCA-SVM) algorithms enable humidity-robust gas identification, while bluetooth low energy (BLE) connectivity facilitates real-time air quality mapping in smart homes (Fig. 10c).

Sweat-permeable 3D PVDF piezoelectric nanoyarn fabrics¹⁵⁸ offer ultrahigh tensile strength (313.3 MPa) and rapid unidirectional liquid transport (4s sweat permeation). Integrated with wireless STM32 systems, they monitor body movements and environment as self-powered switches for smart beds and safety alarms.

9.4. Precision agriculture: from soil to storage

E-nose systems¹⁵⁹ combine piezoelectric and MOS sensors for soil VOC analysis and crop health monitoring (98% pest/disease accuracy). Applications include grain storage contamination alerts via ethylene/ethanol detection and AI-driven vertical farm control (Fig. 10d). Flexible piezoelectrics endow plant wearables with transpiration monitoring capability. An MOF-5/thymol blue-coated micro quartz sensor¹⁶⁰ detects green leaf volatiles (GLVs) by integrating frequency shifts (mass sensing) and colorimetric signals (pH reaction). It achieves 62.5 ppb to 250 ppm detection of 1-hexanol – a key GLV biomarker – enabling field-deployable early pest detection.

9.5. Environmental monitoring

Punetha et al.¹⁶¹ developed a self-powered wearable ammonia gas sensor using an optimized rGO/WO₃/PVDF nanocomposite. By adjusting material ratios and heat treatment processes, they created a device that combines piezoelectric energy harvesting with real-time gas detection (Fig. 10e). This sensor shows excellent performance with high sensitivity (R_g/R_a = 4.72 at 50 ppm ammonia), fast response (32 s) and recovery (78 s), and visual alerts, making it suitable for portable safety monitoring in food storage or pollution detection scenarios in research settings.

Although direct applications of PEC gas sensors in environmental monitoring are still limited, recent progress in adjacent offers valuable references. For instance, a hybrid sensing system by Shen et al.⁵⁵ using triboelectric and chemoresistive mechanisms can harvest energy from engine vibrations to detect NO₂. This system compensates for humidity/temperature variations and provides visual alerts when thresholds are exceeded, demonstrating potential practical solutions for vehicle emission monitoring despite using triboelectric energy generation.

Innovative approaches using piezoelectric effect enhances gas-sensitive response also show promise. Researchers developed environmental sensors based on 2D piezoelectric crystals (MoS₂/WS₂ layered structures),¹⁵⁵ which automatically adjust their sensing parameters to compensate for temperature changes (±0.1 ppm °C⁻¹) during air quality measurements. This strain-responsive technology, originally designed for atmospheric sampling, could potentially enhance signal stability in future PEC-based monitoring systems.

10. Future perspectives of piezoelectric gas sensors

Advances in multi-sensing integration, flexible electronics, AI, and self-powering systems will propel piezoelectric gas sensors toward broader industrial use. Fig. 11 illustrates key innovations addressing current limitations while enabling new functions—such as hybrid sensor arrays and self-sustaining IoT networks—which enhance selectivity, stability, and deployment scalability. Material breakthroughs combined with intelligent designs are shifting piezoelectric composites from labs to real-world applications in precision healthcare and resilient infrastructure. This section highlights five critical frontiers for next-generation sensors.

Fig. 11 Future development trends. (a) and (b) Multi-gas sensors: (a) wireless SAW microsensor with integrated temp/CO₂/NO₂ sensors. Reproduced with permission from ref. 163. Copyright 2011, Elsevier (b) semiconductor metal oxide (SMO) gas sensors with uniform sensor arrays fabricated via glancing angle deposition (GLAD) technique. Reproduced with permission from ref. 164. Copyright 2022, American Chemical Society. (c) and (d) Flexible electronics: (c) MoS₂-based flexible integrated circuit. Reproduced with permission from ref. 169. Copyright 2024, Springer Nature. (d) Laser lift-off (LLO) method for transferring flexible PZT thin films and working mechanism of planar-electrode PENGs. Reproduced with permission from ref. 168. Copyright 2016, Springer Nature. (e) and (f) Integration with AI and ML: (e) edge intelligence system for coal mine gas monitoring; (f) the NSGA-II-optimized single-hidden layer neural network in the system. Reproduced with permission from ref. 175. Copyright 2020, Elsevier. (g) Self-supply systems: self-powered flexible multifunctional monitoring system integrating TENG and PMUT, enabling synchronous detection of indoor environmental parameters and self-sustained energy supply. Reproduced with permission from ref. 204. Copyright 2019, Elsevier. IoT applications: (h) framework of self-powered integrated wireless electronic nodes (SIWEN) using perovskite/polymer PNGs (P-PNGs) as dual-functional power sources and sensors, capable of harvesting energy from diverse environments. Reproduced with permission from ref. 165. Copyright 2020, Royal Society of Chemistry (i) wireless self-powered high-performance nanostructured gas sensor networks for smart home applications. Reproduced with permission from ref. 157. Copyright 2021, American Chemical Society.

10.1. Integration of multi-sensing and arraying capabilities

Research pursues two complementary strategies. First: integrating arrays of diverse gas sensors. Piezoelectric ceramic (PEC) sensors show promise for miniaturization but face material compatibility hurdles. Surface acoustic wave (SAW) sensors enable multi-parameter detection (e.g., CO₂, NO₂, temperature) via on-chip designs (Fig. 11a),¹⁶² while metal-oxide semiconductor (MOS) sensors use wafer-scale glancing angle deposition (GLAD) to create uniform arrays (<10% batch variation, Fig. 11b).¹⁶³ In contrast, PEC integration lags due to incompatible fabrication processes.

Second: single-sensor multi-parameter detection, limited by unresolved signal decoupling and scalable manufacturing issues. Materials like MoS₂ hybrids¹⁶⁴ and MXene/SnO₂ composites⁶² enable lab-scale multi-sensing but lack standardized real-world designs. Early integrated sensors prioritized function over size;⁶¹ later systems like PZT/PANI core–shell structures⁶⁰ and porous composites¹⁶⁵ improved miniaturization but introduced signal noise. Zhong et al.¹⁶⁶ recently unified mechanical, acoustic, and thermal sensing on one platform—offering a framework for future multi-gas PEC sensors.

10.2. Fully flexible wearable systems: materials to circuits

For wearable comfort, all components—sensors to circuits—must be flexible. Rigid parts create pressure points, causing discomfort during prolonged use. This demands co-optimizing materials and circuitry beyond sensor innovation alone.

Current flexible sensing materials (e.g., layered porous composites for harsh environments,¹⁶⁶ conformal polymers for health monitoring¹⁶⁷) exhibit exceptional durability and multi-functionality. Yet connections to rigid processors create stiff zones, limiting practicality. A MoS₂-based flexible circuit¹⁶⁸ marks progress: it integrates 112 thin-film transistors on bendable substrates, maintaining function under 1.23% strain and full voltage tolerance (Fig. 11c).

However, most flexible research targets non-gas applications (e.g., energy harvesting). Gas detection needs stretchable piezoelectric films without cracks (e.g., via laser lift-off³²) and encapsulation layers balancing gas entry with chemical resistance. Combining insights from high-temperature composites,¹⁶⁶ self-powered designs (e.g., PZT nanogenerators, Fig. 11d¹⁶⁷), and deformable circuits¹⁶⁸ could yield systems for real-time gas monitoring in dynamic settings—from industrial gear to personal biosensors.

10.3. AI-driven edge intelligence

AI integration is transforming gas monitoring via localized data processing and adaptive learning. While edge intelligence (AI + edge computing) advances fields like industrial automation¹⁶⁹ and healthcare,¹⁷⁰ piezoelectric systems need co-development of materials, algorithms, and flexible hardware. Core–shell piezoelectric sensors^57,60 deliver clearer signals for AI detection. Yet current microcontroller-based implementations (e.g., 8-bit quantized neural nets¹⁶³) mostly serve MOS sensors (95% VOC accuracy in <10 ms; Fig. 11b). Bio-inspired spiking neural networks (SNNs) could bridge this gap—their event-driven processing matches piezoelectric responses in milliseconds, potentially halving computation.¹⁷¹ This synergy requires joint sensor-algorithm development.

Cross-domain methods tackle environmental noise and data scarcity: time-frequency analysis (STFT, MFCC) from acoustic sensing¹⁷² boosts gas recognition in noisy conditions. Reinforcement learning¹⁷² cuts IoT power 40% via adaptive sampling. Generative adversarial networks (GANs)¹⁷³ synthesize realistic gas profiles for rare-detection training.

A coal mine monitoring prototype¹⁷⁴ demonstrates integration (Fig. 11e). Piezoelectric arrays paired with TinyML hardware use an optimized neural network (SRWNN) to predict gas concentrations (Fig. 11f), responding 38% faster than centralized systems (>90% accuracy). Distributed training analyzes multi-sensor data (gas, CO, airflow) for adaptive hazard detection.

Gas detection algorithms benefit from acoustic sensor advances: ultrasonic catalysis + CNN temperature compensation improved H₂ sensing by Yang et al.¹⁷⁵ BOSCH achieved industrial multi-gas detection.¹⁷⁶ ANN-optimized AlN cantilevers diagnosed plant diseases via VOC profiling by Li et al.,¹⁷⁷ scalable via federated learning.¹⁷⁸ Field-deployable QCM e-noses could gain specificity using lightweight ML models (e.g., Kuchmenko et al.,¹⁷⁹ Huang & Wu,¹⁸⁰ van den Broek et al.,¹⁸¹ Anyfantis & Bilonas¹⁸²). Gaggiotti et al.'s hpDNA-peptide sensor fusion advanced phytochemical analysis (90% ANN accuracy),¹⁸³ complementing terpene detection work (Zhang et al.,¹⁸⁴ Mirzaei et al.¹⁸⁵).

While CNNs/ANNs handle dynamic signals¹⁸⁶ and enable edge deployment,^163,172,187 data scarcity,¹⁸⁸ sensor aging,¹⁸⁹ and poor model interpretability persist. Solutions include hybrid architectures (attention-CNNs,¹⁷⁶ GNNs^190,191), bioinspired materials,^192,193 federated learning,^194–196 neuromorphic designs (SNNs^197,198), and explainable AI tools (e.g., SHapley Additive exPlanations, SHAP¹⁹⁹). These bridge lab research and real-world uses in environmental protection.^200,201 Future systems may merge sensing and computation via memristive gas sensors,²⁰² creating adaptive platforms where materials and edge AI collaborate.

10.4. Fully energy-autonomous systems

Traditional self-powered sensors harvest mechanical energy only for sensing—not for signal processing, requiring external power. New approaches decouple energy harvesting from detection:

A TENG-PMUT hybrid²⁰³ exemplifies this: the triboelectric nanogenerator (TENG) scavenges energy (7.5 mW output), while a separate piezoelectric ultrasonic transducer (PMUT) handles gas sensing (Fig. 11g). Integrated storage and low-power circuits enable true autonomy, treating energy purely as power. This overcomes piezoelectric nanogenerators’ (PENGs) low output vs. TENGs or flexible thermoelectric generators (FTEGs).

Cross-domain models offer templates: TENG humidity sensors²⁰⁴ run battery-free (1.26 kHz/%RH sensitivity; 50 cm wireless). FTEG wearables²⁰⁵ harvest heat (1600 μW at ΔT = 2.3 °C) for motion tracking. Adapting these to gas sensors requires pairing efficient harvesters with circuits tuned to material responses.

Industrial deployment hurdles remain: atomic-layer coatings²⁰⁶ boost durability but complicate integration. Intermittent TENG output (mW-scale) struggles to power protocols like long range wide area network (LoRaWAN),^207,208 needing hybrid solar-thermal-mechanical harvesters.²⁰⁹

Future progress hinges on co-designing materials, energy systems, and algorithms. Embedding TENG layers in piezoelectric composites or 3D-printed structures could merge functions. Bio-inspired event-driven schemes^210,211—transmitting data only when gas thresholds breach—slash energy use for sporadic monitoring.

Though no fully autonomous PEC gas sensors exist yet, the path is clear: separate energy harvesting from sensing, blend hybrid energy sources, and adopt adaptive algorithms. Addressing robustness and scalable integration will enable maintenance-free networks.

10.5. IoT-enabled smart networks

Piezoelectric gas sensors in IoT networks promise smarter environmental and safety monitoring but require further development. Traditional sensors’ reliance on external power or high temperatures hinders miniaturization. Emerging self-powered solutions scavenge vibrations or airflow instead.

Khan et al.'s vibration-powered node¹⁶⁵ monitors infrastructure (10 mW cm⁻²; Fig. 11h). Song et al.'s solar-assisted platform¹⁵⁷ combines gas sensors (4.3 μW per module) and AI for air quality (Fig. 11i). Flexible AlN harvesters power pollution sensors via exhaust gas flow.²¹² PVDF-based sensors dual-function as gas detectors and energy harvesters (99% accuracy) in logistics.²¹³

Challenges include sensitivity drift in humidity/temperature shifts²¹⁴ and limited selectivity. Memristor sensors²¹⁵ improve this (room-temperature operation; 0.34 mW). Future gains need better hybrid materials (e.g., metal-coated designs^101,114), adaptive AI analysis,¹⁵⁷ and multi-source harvester designs.²¹² Wearables (45.6 mW²¹⁶) and road harvesters (92 mW²¹⁷) prove scalability. PEC research should prioritize durability and practical IoT integration for smart-city deployment.

11. Conclusion

Piezoelectric composite gas sensors have evolved from concept to application through synergistic advances in materials and engineering. Innovations in structure—such as layered nanowires and decoupled components—now resolve earlier trade-offs between energy efficiency and signal accuracy. Manufacturing techniques overcome past limitations (agglomeration, interface defects), enabling precision-engineered 3D composites with controlled porosity and tailored coatings. Multiscale simulations have further shifted design from trial-and-error to physics-driven optimization.

Despite progress, critical challenges remain: computational models lack complete physical mechanisms, causing simulations to diverge from experiments. Self-powered systems still cannot generate sufficient energy. Humidity and temperature fluctuations undermine field reliability. Standardized protocols for data processing and IoT connectivity are still missing, hindering scalability.

Future solutions require tightly integrating material properties with electronics—for example, adaptive power management and AI-enhanced signal analysis. Shared AI systems could address sustainability and data scarcity challenges.

For piezoelectric sensors to transition from lab devices to flexible real-world tools, cross-domain collaboration is essential. By unifying these advancements, they could transform critical sectors—from medical monitoring to climate resilience—establishing a key role in next-generation IoT networks.

Abbreviations

PEC	Piezoelectric composite
TENG	Triboelectric nanogenerator
MOF	Metal–organic framework
DEP	Dielectrophoresis
EMI	Electromagnetic interference
DMF	Dimethylformamide
BT2	BaTi2O5
FOM	Figure of merit
PCL	Polycaprolactone
BN	Boron nitride
CsPbBr3	Cesium lead bromide
BTO	BaTiO3
P(VDF-TrFE)	Poly(vinylidene fluoride-trifluoroethylene)
CAD	Computer-aided design
ITO	Indium tin oxide
PES	Polyethersulfone
PDA	Polydopamine
PANI	Polyaniline
PVA	Polyvinyl alcohol
CSA	Camphorsulfonic acid
ASP	L-Aspartic acid
CF	Carbon fiber
DIW	Direct ink writing
PZT	Lead zirconate titanate
PVDF	Polyvinylidene fluoride
PVP	Polyvinylpyrrolidone
PDMS	Polydimethylsiloxane
PPy	Polypyrrole
SAM	Self-assembled monolayer
FNP	Functionalized nanowire piezoelectric
PENG	Piezoelectric nanogenerator
RH	Relative humidity
LPG	Liquefied petroleum gas
MXene	Transition metal carbide/nitride
Co3O4	Cobalt oxide
LOD	Limit of detection
CNT	Carbon nanotube
PAAS	Polyacrylate sodium
T-ZnO	Tetrapod ZnO
Sm-PMN-PT	Samarium-doped Pb(Mg1/3Nb2/3)O3-PbTiO3
PEI	Polyetherimide
DFT	Density functional theory
MD	Molecular dynamics
FEM	Finite element method
SCR	Silicon-controlled rectifier
MOSFET	Metal-oxide-semiconductor field-effect transistor
DC	Direct current
CMOS	Complementary metal-oxide-semiconductor
ADC	Analog-to-digital converter
SNR	Signal-to-noise ratio
IoT	Internet of things
1/f noise	Flicker noise
SSHI	Synchronized switch harvesting on inductor
LC	Inductor-capacitor
DEM	Dynamic element matching
SINGOR	Self-sustaining indoor gas monitoring
PCA-SVM	Principal component analysis-support vector machine
BLE	Bluetooth low energy
GLV	Green leaf volatiles
rGO	Reduced graphene oxide
SAW	Surface acoustic wave
GLAD	Glancing angle deposition
SHAP	Shapley additive explanations
CNN	Convolutional neural network
ANN	Artificial neural network
GAN	Generative adversarial network
SNN	Spiking neural network
TinyML	Tiny machine learning
GNN	Graph neural network
LoRaWAN	Long range wide area network
PMUT	Piezoelectric micromachined ultrasonic transducer
FTEG	Flexible thermoelectric generator

Conflicts of interest

The authors declare no competing financial interest.

Data availability

No primary research results, software or code have been included and no new data were generated or analysed as part of this review.

Acknowledgements

This work was funded by the National Science Fund for Distinguished Young Scholars (Grant No. 62225106), the National Natural Science Foundation of China (Grant No. U24A20229, 62074027), Sichuan Innovation Research Group Project (Grant No. 2025NSFTD0008), Sichuan Province Hong Kong, Macao and Taiwan Science and Technology Innovation Cooperation Project (Grant No. 2024YFHZ0367) and the Natural Science Foundation of Sichuan Province of China (Grant No. 2024NSFSC0462).

References

1. J. Yun, M. Cho, K. Lee, M. Kang and I. Park, Sens. Actuators, B, 2022, 372, 132612.
2. L. Zhu and W. Zeng, Sens. Actuators, A, 2017, 267, 242–261.
3. S. S. Ba Hashwan, M. H. Md Khir, I. M. Nawi, M. R. Ahmad, M. Hanif, F. Zahoor, Y. Al-Douri, A. S. Algamili, U. I. Bature, S. S. Alabsi, M. O. B. Sabbea and M. Junaid, Discover Nano, 2023, 18, 25.
4. J. Dai, O. Ogbeide, N. Macadam, Q. Sun, W. Yu, Y. Li, B.-L. Su, T. Hasan, X. Huang and W. Huang, Chem. Soc. Rev., 2020, 49, 1756–1789.
5. A. Nasri, M. Pétrissans, V. Fierro and A. Celzard, Mater. Sci. Semicond. Process., 2021, 128, 105744.
6. A. T. Le, M. Ahmadipour and S.-Y. Pung, J. Alloys Compd., 2020, 844, 156172.
7. N. Yamazoe, G. Sakai and K. Shimanoe, Catal. Surv. Asia, 2003, 7, 63–75.
8. P. Wang, P. Deng, Y. Nie, Y. Zhao, Y. Zhang, L. Xing and X. Xue, Nanotechnology, 2014, 25, 075501.
9. Y. Li, W. Li, Z. Jin, X. Luo, G. Xie, H. Tai, Y. Jiang, Y. Yang and Y. Su, Nano Energy, 2024, 122, 109291.
10. H. Guan, R. Yang, W. Li, Y. Tao, C. Chen, H. Tai, Y. Su, Y. Wang, Y. Jiang and W. Li, Sens. Actuators, B, 2023, 377, 132996.
11. Y. Su, Y. Liu, W. Li, X. Xiao, C. Chen, H. Lu, Z. Yuan, H. Tai, Y. Jiang, J. Zou, G. Xie and J. Chen, Mater. Horiz., 2023, 10, 842–851.
12. C. Chen, M. Jiang, X. Luo, H. Tai, Y. Jiang, M. Yang, G. Xie and Y. Su, Sens. Actuators, B, 2022, 370, 132441.
13. Y. Zhou, X. Zhao, J. Xu, Y. Fang, G. Chen, Y. Song, S. Li and J. Chen, Nat. Mater., 2021, 20, 1670–1676.
14. X. Zhao, Y. Zhou, J. Xu, G. Chen, Y. Fang, T. Tat, X. Xiao, Y. Song, S. Li and J. Chen, Nat. Commun., 2021, 12, 6755.
15. X. Zhao, Y. Zhou, Y. Song, J. Xu, J. Li, T. Tat, G. Chen, S. Li and J. Chen, Nat. Mater., 2024, 23, 703–710.
16. X. Zhao, Y. Zhou, W. Kwak, A. Li, S. Wang, M. Dallenger, S. Chen, Y. Zhang, A. Lium and J. Chen, Nat. Commun., 2024, 15, 8492.
17. X. Zhao, Y. Zhou, A. Li, J. Xu, S. Karjagi, E. Hahm, L. Rulloda, J. Li, J. Hollister, P. Kavehpour and J. Chen, Nat. Electron., 2024, 7, 924–932.
18. J. Yin, S. Wang, T. Tat and J. Chen, Nat. Rev. Bioeng., 2024, 2, 541–558.
19. Y. Su, G. Xie, H. Tai, S. Li, B. Yang, S. Wang, Q. Zhang, H. Du, H. Zhang, X. Du and Y. Jiang, Nano Energy, 2018, 47, 316–324.
20. Y. Su, J. Wang, B. Wang, T. Yang, B. Yang, G. Xie, Y. Zhou, S. Zhang, H. Tai, Z. Cai, G. Chen, Y. Jiang, L.-Q. Chen and J. Chen, ACS Nano, 2020, 14, 6067–6075.
21. B. Yu, Y. Fu, P. Wang, Y. Zhao, L. Xing and X. Xue, Phys. Chem. Chem. Phys., 2015, 17, 10856–10860.
22. H. Karami, B. Thurn, N. K. De Boer, J. Ramos, J. A. Covington, J. Lozano, T. Liu, W. Zhang, S. Su and M. Ueland, Nat Hazards, 2025, 121, 31–60.
23. P. Hu, L. Yan, C. Zhao, Y. Zhang and J. Niu, Compos. Sci. Technol., 2018, 168, 327–335.
24. J. Zhang, C. Li, C. Lv, H. Yu, W. Ma, M. Lei, X. Chen, M. Liu and X. Zhang, Mater. Today Phys., 2025, 50, 101606.
25. W.-S. Jung, M.-J. Lee, M.-G. Kang, H. G. Moon, S.-J. Yoon, S.-H. Baek and C.-Y. Kang, Nano Energy, 2015, 13, 174–181.
26. Y. Su, W. Li, X. Cheng, Y. Zhou, S. Yang, X. Zhang, C. Chen, T. Yang, H. Pan, G. Xie, G. Chen, X. Zhao, X. Xiao, B. Li, H. Tai, Y. Jiang, L.-Q. Chen, F. Li and J. Chen, Nat. Commun., 2022, 13, 4867.
27. X. Xue, Y. Nie, B. He, L. Xing, Y. Zhang and Z. L. Wang, Nanotechnology, 2013, 24, 225501.
28. Y. Hu, L. Lin, Y. Zhang and Z. L. Wang, Adv. Mater., 2012, 24, 110–114.
29. Y. Su, M. Yao, G. Xie, H. Pan, H. Yuan, M. Yang, H. Tai, X. Du and Y. Jiang, Appl. Phys. Lett., 2019, 115, 073504.
30. Y. Su, G. Xie, S. Wang, H. Tai, Q. Zhang, H. Du, H. Zhang, X. Du and Y. Jiang, Sens. Actuators, B, 2017, 251, 144–152.
31. X. Guan, B. Xu and J. Gong, Nano Energy, 2020, 70, 104516.
32. K. Park, J. H. Son, G. Hwang, C. K. Jeong, J. Ryu, M. Koo, I. Choi, S. H. Lee, M. Byun, Z. L. Wang and K. J. Lee, Adv. Mater., 2014, 26, 2514–2520.
33. K. You, C. Li, D. Zhou and K. Bi, 2022 IEEE International Conference on Industrial Technology (ICIT), IEEE, Shanghai, China, 2022, pp. 1–5.
34. M. Choi, D. Choi, M. Jin, I. Kim, S. Kim, J. Choi, S. Y. Lee, J. M. Kim and S. Kim, Adv. Mater., 2009, 21, 2185–2189.
35. Y. Fu, Y. Zhao, P. Wang, L. Xing and X. Xue, Phys. Chem. Chem. Phys., 2015, 17, 2121–2126.
36. Y. Fu, Y. Nie, Y. Zhao, P. Wang, L. Xing, Y. Zhang and X. Xue, ACS Appl. Mater. Interfaces, 2015, 7, 10482–10490.
37. M. W. G. Hoffmann, L. Mayrhofer, O. Casals, L. Caccamo, F. Hernandez-Ramirez, G. Lilienkamp, W. Daum, M. Moseler, A. Waag, H. Shen and J. D. Prades, Adv. Mater., 2014, 26, 8017–8022.
38. Y. Nie, P. Deng, Y. Zhao, P. Wang, L. Xing, Y. Zhang and X. Xue, Nanotechnology, 2014, 25, 265501.
39. Y. Lin, P. Deng, Y. Nie, Y. Hu, L. Xing, Y. Zhang and X. Xue, Nanoscale, 2014, 6, 4604–4610.
40. Z. Wen, L. Zhu, Z. Zhang and Z. Ye, Sens. Actuators, B, 2015, 208, 112–121.
41. W. Zang, Y. Nie, D. Zhu, P. Deng, L. Xing and X. Xue, J. Phys. Chem. C, 2014, 118, 9209–9216.
42. D. Zhang, Q. Mi, D. Wang and T. Li, Sens. Actuators, B, 2021, 339, 129923.
43. R. Zhou, G. Hu, R. Yu, C. Pan and Z. L. Wang, Nano Energy, 2015, 12, 588–596.
44. D. Zhu, T. Hu, Y. Zhao, W. Zang, L. Xing and X. Xue, Sens. Actuators, B, 2015, 213, 382–389.
45. W. Li, T. Yang, C. Liu, Y. Huang, C. Chen, H. Pan, G. Xie, H. Tai, Y. Jiang, Y. Wu, Z. Kang, L. Chen, Y. Su and Z. Hong, Adv. Sci., 2022, 9, 2105550.
46. Y.-F. Lin, J. Song, Y. Ding, S.-Y. Lu and Z. L. Wang, Appl. Phys. Lett., 2008, 92, 022105.
47. A. R. Prasad and A. Joseph, RSC Adv., 2017, 7, 20960–20968.
48. H. Mahdavi, N. Mazinani and A. A. Heidari, Polym. Int., 2020, 69, 1187–1195.
49. Y. Su, G. Xie, T. Xie, H. Zhang, Z. Ye, Q. Jing, H. Tai, X. Du and Y. Jiang, J. Phys. D: Appl. Phys., 2016, 49, 215601.
50. G. Zhang, P. Zhao, X. Zhang, K. Han, T. Zhao, Y. Zhang, C. K. Jeong, S. Jiang, S. Zhang and Q. Wang, Energy Environ. Sci., 2018, 11, 2046–2056.
51. R. Sahore, B. L. Armstrong, X. Tang, C. Liu, K. Owensby, S. Kalnaus and X. C. Chen, Adv. Energy Mater., 2023, 13, 2203663.
52. S. Wen, S. Chen, W. Gao, Z. Zheng, J. Bao, C. Cui, S. Liu, H. Gao and S. Yu, Adv. Mater., 2023, 35, 2211175.
53. H. Roh, D. Kim, Y. Cho, Y. Jo, J. A. Del Alamo, H. J. Kulik, M. Dincă and A. Gumyusenge, Adv. Mater., 2024, 36, 2312382.
54. F. Ren, Z. Liang, W. Zhao, W. Zuo, M. Lin, Y. Wu, X. Yang, Z. Gong and Y. Yang, Energy Environ. Sci., 2023, 16, 2579–2590.
55. Q. Shen, X. Xie, M. Peng, N. Sun, H. Shao, H. Zheng, Z. Wen and X. Sun, Adv. Funct. Mater., 2018, 28, 1703420.
56. S. Li and L. Zhang, npj Comput. Mater., 2024, 10, 138.
57. D. Liu, Q. Shi, S. Jin, Y. Shao and J. Huang, InfoMat, 2019, 1, 525–532.
58. S. Yu, F. Qin and G. Wang, J. Mater. Chem. C, 2016, 4, 1504–1510.
59. W. Jiang, K. Hu, N. Lv and Z. Lyu, Sens. Actuators, A, 2021, 332, 113203.
60. C. Chen, G. Xie, J. Dai, W. Li, Y. Cai, J. Li, Q. Zhang, H. Tai, Y. Jiang and Y. Su, Nano Energy, 2023, 116, 108788.
61. D. Zhang, Z. Yang, P. Li, M. Pang and Q. Xue, Nano Energy, 2019, 65, 103974.
62. S. Gasso and A. Mahajan, ACS Appl. Nano Mater., 2023, 6, 6678–6692.
63. L. Yang, Q. Liu, M. Li, Y. Liu, X. Li, Q. Liu, T. Zhu, Y. Lu, X. Liu and D. Wang, Sens. Actuators, B, 2025, 423, 136778.
64. S.-Y. Bae, I.-Y. Jeon, J. Yang, N. Park, H. S. Shin, S. Park, R. S. Ruoff, L. Dai and J.-B. Baek, ACS Nano, 2011, 5, 4974–4980.
65. E. Fontananova, J. C. Jansen, A. Cristiano, E. Curcio and E. Drioli, Desalination, 2006, 192, 190–197.
66. H. Cheng, Y. Pan, Q. Chen, R. Che, G. Zheng, C. Liu, C. Shen and X. Liu, Adv. Compos. Hybrid Mater., 2021, 4, 505–513.
67. M. Latif, Y. Jiang and J. Kim, Mater. Today Adv., 2024, 21, 100478.
68. J. T. Kloster, M. J. Danley, V. K. Lai and P. Zhao, BME Front., 2023, 4, 0009.
69. Z. Wang, J. Cheng, R. Hu, X. Yuan, Z. Yu, X. Xu, F. Wang, J. Dong, R. Gong, S. Dong and H. Wang, J. Mater. Chem. A, 2021, 9, 26767–26776.
70. H.-Y. Zhang, Y.-Y. Tang, Z.-X. Gu, P. Wang, X.-G. Chen, H.-P. Lv, P.-F. Li, Q. Jiang, N. Gu, S. Ren and R.-G. Xiong, Science, 2024, 383, 1492–1498.
71. C. Liu, W. Wu, J. Liang and H. Zhou, Opt. Express, 2024, 32, 42793–42800.
72. J. Fu, Y. Hou, X. Gao, M. Zheng and M. Zhu, Nano Energy, 2018, 52, 391–401.
73. D. Wang, D. Zhang, P. Li, Z. Yang, Q. Mi and L. Yu, Nano-Micro Lett., 2021, 13, 57.
74. Y. Li, M. Xu, Y. Xia, J. Wu, X. Sun, S. Wang, G. Hu and C. Xiong, Chem. Eng. J., 2020, 388, 124205.
75. A. Greiner and J. H. Wendorff, Angew. Chem., Int. Ed., 2007, 46, 5670–5703.
76. N. Bock, T. R. Dargaville and M. A. Woodruff, Prog. Polym. Sci., 2012, 37, 1510–1551.
77. Y. Chen, M. Shafiq, M. Liu, Y. Morsi and X. Mo, Bioact. Mater., 2020, 5, 963–979.
78. K. Zhao, S. Wei, M. Cao, M. Wang, P. Li, H. Li, X. Zhang, Y. Zhang and Y. Chen, Compos. Sci. Technol., 2024, 245, 110323.
79. D. Ji, Y. Lin, X. Guo, B. Ramasubramanian, R. Wang, N. Radacsi, R. Jose, X. Qin and S. Ramakrishna, Nat. Rev. Methods Primers, 2024, 4, 1.
80. H. Min, N. Wang, N. Chen, Y. Tong, Y. Wang, J. Wang, J. Liu, S. Wang, X. Wu, P. Yang, H. Shi, C. Zhuo, Q. Chen, J. Li, D. Zhang, X. Lu, C. Zhu, Q. Peng, L. Zhu, J. Chang, W. Huang and J. Wang, Nat. Nanotechnol., 2024, 19, 632–637.
81. X. Fu, Y. Cao, X. Song, R. Sun, H. Jiang, P. Du, D. Peng and L. Luo, Adv. Funct. Mater., 2025, 35, 2412507.
82. D. O. Baumann, F. Laufer, J. Roger, R. Singh, M. Gholipoor and U. W. Paetzold, ACS Appl. Mater. Interfaces, 2024, 16, 54007–54016.
83. J. Zhang, J. Wu, V. M. Le Corre, J. A. Hauch, Y. Zhao and C. J. Brabec, InfoMat, 2025, e70005.
84. K. V. Wong and A. Hernandez, ISRN Mech. Eng., 2012, 1–10.
85. Z. Wang, X. Yuan, J. Yang, Y. Huan, X. Gao, Z. Li, H. Wang and S. Dong, Nano Energy, 2020, 73, 104737.
86. X. Zhou, K. Parida, O. Halevi, Y. Liu, J. Xiong, S. Magdassi and P. S. Lee, Nano Energy, 2020, 72, 104676.
87. J. H. Kim, D. Andeen and F. F. Lange, Adv. Mater., 2006, 18, 2453–2457.
88. S. Abubakar, S. T. Tan, J. Y. C. Liew, Z. A. Talib, R. Sivasubramanian, C. A. Vaithilingam, S. S. Indira, W.-C. Oh, R. Siburian, S. Sagadevan and S. Paiman, Nanomaterials, 2023, 13, 1025.
89. M. A. Mustaffa, F. Arith, N. S. Noorasid, M. S. I. M. Zin, K. S. Leong, F. A. Ali, A. N. M. Mustafa and M. M. Ismail, Micromachines, 2022, 13, 2200.
90. F. Li, M. J. Cabral, B. Xu, Z. Cheng, E. C. Dickey, J. M. LeBeau, J. Wang, J. Luo, S. Taylor, W. Hackenberger, L. Bellaiche, Z. Xu, L.-Q. Chen, T. R. Shrout and S. Zhang, Science, 2019, 364, 264–268.
91. L. Gu, J. Liu, N. Cui, Q. Xu, T. Du, L. Zhang, Z. Wang, C. Long and Y. Qin, Nat. Commun., 2020, 11, 1030.
92. H. Jiang, L. Song, Z.-X. Huang, M. Liu, Y. Zhao, S. Zhang, J. Guo, Y. Li, Q. Wang and J.-P. Qu, Nano Energy, 2022, 104, 107921.
93. X. Huang, Y. Wang and X. Zhang, Nano Energy, 2022, 99, 107379.
94. M. Mahato and B. Adhikari, Synth. Met., 2016, 220, 410–420.
95. D. Calestani, M. Villani, M. Culiolo, D. Delmonte, N. Coppedè and A. Zappettini, Sens. Actuators, B, 2017, 245, 166–170.
96. S. Zhou, Y. Zhao, Y. Xun, Z. Wei, Y. Yang, W. Yan and J. Ding, Chem. Rev., 2024, 124, 3608–3643.
97. S. Xu, B. J. Hansen and Z. L. Wang, Nat. Commun., 2010, 1, 93.
98. Z. Wen, Q. Shen and X. Sun, Nano-Micro Lett., 2017, 9, 45.
99. E. Modaresinezhad and S. Darbari, Sens. Actuators, B, 2016, 237, 358–366.
100. Y. Zhao, X. Lai, P. Deng, Y. Nie, Y. Zhang, L. Xing and X. Xue, Nanotechnology, 2014, 25, 115502.
101. L. Xing, Y. Hu, P. Wang, Y. Zhao, Y. Nie, P. Deng and X. Xue, Appl. Phys. Lett., 2014, 104, 013109.
102. Z. Qu, Y. Fu, B. Yu, P. Deng, L. Xing and X. Xue, Sens. Actuators, B, 2016, 222, 78–86.
103. D. Zhu, Y. Fu, W. Zang, Y. Zhao, L. Xing and X. Xue, Mater. Lett., 2016, 166, 288–291.
104. Y. Fu, W. Zang, P. Wang, L. Xing, X. Xue and Y. Zhang, Nano Energy, 2014, 8, 34–43.
105. D. Zhu, Y. Fu, W. Zang, Y. Zhao, L. Xing and X. Xue, Sens. Actuators, B, 2014, 205, 12–19.
106. Y.-X. Zhen, G. Wang, Y.-F. Li and Y. Yu, Front. Chem., 2025, 12, 1532018.
107. J. Toledo, V. Ruiz-Díez, P. Schwarz, H. Seidel and J. L. Sánchez-Rojas, in Smart Sensors, Actuators, and MEMS VIII, ed. L. Fonseca, M. Prunnila and E. Peiner, Barcelona, Spain, SPIE, 2017, vol. 10246, p. 148.
108. L. Yang, T. Wei and N. Chen, Energy Technol., 2022, 10, 2200434.
109. Z. Li, B. Cui, B. Wang, F. Shen, H. Wu, Y. Gong, Y. Peng, H. Guo and D. Zhang, Sci. China: Technol. Sci., 2025, 68, 1310101.
110. Y. Lin, Z. Long, S. Liang, T. Zhong and L. Xing, J. Phys. D: Appl. Phys., 2022, 55, 224001.
111. Y. Fu, H. He, T. Zhao, Y. Dai, W. Han, J. Ma, L. Xing, Y. Zhang and X. Xue, Nano-Micro Lett., 2018, 10, 76.
112. M. Shahabi and H. Raissi, J. Inclusion Phenom. Macrocyclic Chem., 2016, 86, 305–322.
113. Y. Wang, W. Wang, Z. Zhang and P. Li, Appl. Surf. Sci., 2022, 571, 151355.
114. H. Demir, J. A. Greathouse, C. L. Staiger, J. J. Perry Iv, M. D. Allendorf and D. S. Sholl, J. Mater. Chem. A, 2015, 3, 23539–23548.
115. M. Orio, D. A. Pantazis and F. Neese, Photosynth. Res., 2009, 102, 443–453.
116. M. Karplus and G. A. Petsko, Nature, 1990, 347, 631–639.
117. B. Leimkuhler and C. Matthews, Molecular Dynamics: With Deterministic and Stochastic Numerical Methods, Springer International Publishing, Cham, 2015, vol. 39.
118. J.-J. Wang, B. Wang and L.-Q. Chen, Annu. Rev. Mater. Res., 2019, 49, 127–152.
119. L.-Q. Chen, Annu. Rev. Mater. Res., 2002, 32, 113–140.
120. F. Li, D. Lin, Z. Chen, Z. Cheng, J. Wang, C. Li, Z. Xu, Q. Huang, X. Liao, L.-Q. Chen, T. R. Shrout and S. Zhang, Nat. Mater., 2018, 17, 349–354.
121. R. J. Zeches, M. D. Rossell, J. X. Zhang, A. J. Hatt, Q. He, C.-H. Yang, A. Kumar, C. H. Wang, A. Melville, C. Adamo, G. Sheng, Y.-H. Chu, J. F. Ihlefeld, R. Erni, C. Ederer, V. Gopalan, L. Q. Chen, D. G. Schlom, N. A. Spaldin, L. W. Martin and R. Ramesh, Science, 2009, 326, 977–980.
122. Q. Li, L. Chen, M. R. Gadinski, S. Zhang, G. Zhang, H. U. Li, E. Iagodkine, A. Haque, L.-Q. Chen, T. N. Jackson and Q. Wang, Nature, 2015, 523, 576–579.
123. F. Risch, P. Koutsogiannis, Y. Tikhonov, A. G. Razumnaya, C. Magén, J. A. Pardo, I. Lukyanchuk and I. Stolichnov, ACS Nano, 2025, 19, 6993–7004.
124. M. Venkateswar Reddy, B. Hemasunder, S. V. Ramana, P. Ramesh Babu, P. Thejasree and J. Joseph, Mater. Today: Proc., 2023, S2214785323037227.
125. K. T. Butler, D. W. Davies, H. Cartwright, O. Isayev and A. Walsh, Nature, 2018, 559, 547–555.
126. F. Noé, A. Tkatchenko, K.-R. Müller and C. Clementi, Annu. Rev. Phys. Chem., 2020, 71, 361–390.
127. M. I. A. Asri, Md. N. Hasan, M. R. A. Fuaad, Y. Md Yunos and M. S. M. Ali, IEEE Sens. J., 2021, 21, 18381–18397.
128. R. Littrell and K. Grosh, J. Microelectromech. Syst., 2012, 21, 406–413.
129. C. Ge and H. Chen, J. Bionic Eng., 2024, 21, 2717–2729.
130. X. Le, Q. Shi, P. Vachon, E. J. Ng and C. Lee, J. Micromech. Microeng., 2022, 32, 014005.
131. H.-Y. Yeh and G.-H. Feng, in 2023 IEEE 36th International Conference on Micro Electro Mechanical Systems (MEMS), IEEE, Munich, Germany, 2023, pp. 239–242.
132. Y. Zhang, T. Jin, Y. Deng, Z. Zhao, R. Wang, Q. He, J. Luo, J. Li, K. Du, T. Wu, C. Yan, H. Zhang, X. Lu, C. Huang and H. Gao, Microsyst. Nanoeng., 2024, 10, 128.
133. D. Soleymanzadeh, H. Ghafarirad and M. Zareinejad, Measurement, 2019, 147, 106839.
134. N. Wang, J. Van Turnhout, R. Daniels, C. Wu, J. Huo, R. Gerhard, G. Sotzing and Y. Cao, ACS Appl. Mater. Interfaces, 2022, 14, 42705–42712.
135. J. H. Lee, K.-W. Kim, J. K. Kim and U. Jeong, ACS Appl. Electron. Mater., 2020, 2, 2550–2557.
136. H. Zhang, D. Galayko and P. Basset, Nano Energy, 2022, 103, 107816.
137. A. Ghaffarinejad, Y. Lu, R. Hinchet, D. Galayko, J. Y. Hasani, S. W. Kim and P. Basset, Electron. Lett., 2018, 54, 378–379.
138. G. Cheng, Z.-H. Lin, L. Lin, Z. Du and Z. L. Wang, ACS Nano, 2013, 7, 7383–7391.
139. T. Zhang, L. Xie, J. Li, Z. Huang, H. Lei, Y. Liu, Z. Wen, Y. Xie and X. Sun, Nanomaterials, 2021, 11, 2711.
140. Y. H. Liufu, D. Dai and Z. Liu, Nano Energy, 2023, 106, 108042.
141. M. Ben Ammar, S. Sahnoun, A. Fakhfakh, C. Viehweger and O. Kanoun, Sensors, 2023, 23, 1830.
142. A. A. Khan, A. Mathur, L. Yin, M. Almadhoun, J. Yin, M. H. Bagheri, M. F. A. Fattah, A. Rajabi-Abhari, N. Yan, B. Zhao, V. Maheshwari and D. Ban, Nat. Commun., 2024, 15, 9511.
143. R. Puyol, S. Petre, Y. Danlee, T. Walewyns, L. A. Francis and D. Flandre, IEEE Sens. J., 2022, 22, 9521–9533.
144. S. Ippili, V. Jella, A. M. Thomas and S.-G. Yoon, Nanoenergy Adv., 2021, 1, 3–31.
145. A. Liotta, G. Frattini, P. Giannelli, E. Bonizzoni and P. Malcovati, in 2022 17th Conference on PhD Research in Microelectronics and Electronics (PRIME), IEEE, Villasimius, SU, Italy, 2022, pp. 245–248.
146. Y.-J. Park, B. Jang, S.-M. Park, H.-C. Ryu, S. J. Oh, S.-Y. Kim, Y. Pu, S.-S. Yoo, K. C. Hwang, Y. Yang, M. Lee and K.-Y. Lee, IEEE Trans. Power Electron., 2018, 33, 3141–3156.
147. K. Makinwa and M. Pertijs, ESSCIRC 2007-33rd European Solid-State Circuits Conference, 2007, pp. 76–82.
148. M. A. P. Pertijs, G. C. M. Meijer and J. H. Huijsing, IEEE Sens. J., 2004, 4, 294–300.
149. T. Denison, K. Consoer, A. Kelly, A. Hachenburg and W. Santa, in 2007 IEEE International Solid-State Circuits Conference. Digest of Technical Papers, IEEE, San Francisco, CA, USA, 2007, pp. 162–594.
150. K. A. A. Makinwa and J. H. Huijsing, Sens. Actuators, A, 2002, 97–98, 15–20.
151. K. Wang, X. Guo, C. Han, L. Liu, Z. Wang, L. Thomsen, P. Chen, Z. Shao, X. Wang, F. Xie, G. Liu, L. Wang and S. Liu, ACS Nano, 2025, 19, 3818–3829.
152. H. He, C. Dong, Y. Fu, W. Han, T. Zhao, L. Xing and X. Xue, Sens. Actuators, B, 2018, 267, 392–402.
153. X. Ma, Q. Hu, L. Zhou, X. Xiang, Y. Qin, K. Zhang, P. He, Y. Dai, W. Niu and X. Zhang, Nano Res., 2024, 17, 7643–7657.
154. F. Yang, J. Li, Y. Long, Z. Zhang, L. Wang, J. Sui, Y. Dong, Y. Wang, R. Taylor, D. Ni, W. Cai, P. Wang, T. Hacker and X. Wang, Science, 2021, 373, 337–342.
155. S. Siraj, G. Bansal, B. Hasita, S. Srungaram, S. K. S. F. J. Rybicki, S. Sonkusale and P. Sahatiya, ACS Appl. Nano Mater., 2024, 7, 11350–11361.
156. L. Jin, Z. Liu, M. Altintas, Y. Zheng, Z. Liu, S. Yao, Y. Fan and Y. Li, ACS Sens., 2022, 7, 2281–2292.
157. Z. Song, W. Ye, Z. Chen, Z. Chen, M. Li, W. Tang, C. Wang, Z. Wan, S. Poddar, X. Wen, X. Pan, Y. Lin, Q. Zhou and Z. Fan, ACS Nano, 2021, 15, 7659–7667.
158. W. Fan, R. Lei, H. Dou, Z. Wu, L. Lu, S. Wang, X. Liu, W. Chen, M. Rezakazemi, T. M. Aminabhavi, Y. Li and S. Ge, Nat. Commun., 2024, 15, 3509.
159. T. Seesaard, N. Goel, M. Kumar and C. Wongchoosuk, Comput. Electron. Agric., 2022, 193, 106673.
160. L. R. Jaishi, W. Ding, R. A. Kittelson, F. Tsow and X. Xian, ACS Sens., 2024, 9, 6553–6562.
161. D. Punetha, A. Kumar, S. K. Pandey and S. Chakrabarti, in Organic and Hybrid Sensors and Bioelectronics XVI, ed. I. Kymissis, E. J. List-Kratochvil and S. Inal, SPIE, San Diego, United States, 2023, p. 28.
162. C. Lim, W. Wang, S. Yang and K. Lee, Sens. Actuators, B, 2011, 154, 9–16.
163. M. Kang, I. Cho, J. Park, J. Jeong, K. Lee, B. Lee, D. Del Orbe Henriquez, K. Yoon and I. Park, ACS Sens., 2022, 7, 430–440.
164. R. K. Sonker, B. C. Yadav, A. Sharma, M. Tomar and V. Gupta, RSC Adv., 2016, 6, 56149–56158.
165. A. A. Khan, M. M. Rana, G. Huang, N. Mei, R. Saritas, B. Wen, S. Zhang, P. Voss, E.-A. Rahman, Z. Leonenko, S. Islam and D. Ban, J. Mater. Chem. A, 2020, 8, 13619–13629.
166. Y. Zhong, Y. Wang, L. Ma, P. He, J. Qin, J. Gao and J. Li, ACS Appl. Mater. Interfaces, 2025, 17, 11154–11163.
167. W. Gao, S. Emaminejad, H. Y. Y. Nyein, S. Challa, K. Chen, A. Peck, H. M. Fahad, H. Ota, H. Shiraki, D. Kiriya, D.-H. Lien, G. A. Brooks, R. W. Davis and A. Javey, Nature, 2016, 529, 509–514.
168. Y. Peng, C. Cui, L. Li, Y. Wang, Q. Wang, J. Tian, Z. Huang, B. Huang, Y. Zhang, X. Li, J. Tang, Y. Chu, W. Yang, D. Shi, L. Du, N. Li and G. Zhang, Nat. Commun., 2024, 15, 10833.
169. W. Dai, H. Nishi, V. Vyatkin, V. Huang, Y. Shi and X. Guan, IEEE Ind. Electron. Mag., 2019, 13, 48–56.
170. S. U. Amin and M. S. Hossain, IEEE Access, 2021, 9, 45–59.
171. D. Huo, J. Zhang, X. Dai, P. Zhang, S. Zhang, X. Yang, J. Wang, M. Liu, X. Sun and H. Chen, Sensors, 2023, 23, 2433.
172. Y. H. Jung, S. K. Hong, H. S. Wang, J. H. Han, T. X. Pham, H. Park, J. Kim, S. Kang, C. D. Yoo and K. J. Lee, Adv. Mater., 2020, 32, 1904020.
173. N. Yoon, W. Jung and H. Kim, Chemosensors, 2024, 12, 131.
174. Y. Zhang, H. Guo, Z. Lu, L. Zhan and P. C. K. Hung, Eng. Appl. Artif. Intell., 2020, 92, 103643.
175. Y. Yang, S. Lin and J. Hu, Sens. Actuators, B, 2023, 385, 133721.
176. L. Klibanov and P. Boldt, Preliminary analysis of Bosch BME688 4-in-1 environmental sensor with AI, 2021.
177. D. Li, B. Zhu, K. Pang, Q. Zhang, M. Qu, W. Liu, Y. Fu and J. Xie, ACS Sens., 2022, 7, 1555–1563.
178. A. Anyfantis, A. Silis and S. Blionas, Meas.: Sens., 2021, 16, 100049.
179. T. A. Kuchmenko, R. U. Umarkhanov, A. A. Shuba, E. S. Dorovskaya and A. E. Chernitskiy, IOP Conf. Ser.: Earth Environ. Sci., 2021, 640, 072028.
180. J. Huang and J. Wu, Chemosensors, 2020, 8, 73.
181. J. Van Den Broek, A. T. Güntner and S. E. Pratsinis, ACS Sens., 2018, 3, 677–683.
182. A. Anyfantis and S. Blionas, in 2020 IEEE International Instrumentation and Measurement Technology Conference (I2MTC), IEEE, Dubrovnik, Croatia, 2020, pp. 1–6.
183. S. Gaggiotti, S. Palmieri, F. Della Pelle, M. Sergi, A. Cichelli, M. Mascini and D. Compagnone, Sens. Actuators, B, 2020, 308, 127697.
184. C. Zhang, P. Chen and W. Hu, Chem. Soc. Rev., 2015, 44, 2087–2107.
185. A. Mirzaei, J.-H. Lee, S. M. Majhi, M. Weber, M. Bechelany, H. W. Kim and S. S. Kim, J. Appl. Phys., 2019, 126, 241102.
186. Q. Zhou, Q. Ding, Z. Geng, C. Hu, L. Yang, Z. Kan, B. Dong, M. Won, H. Song, L. Xu and J. S. Kim, Nano-Micro Lett., 2025, 17, 50.
187. H. Fang, S. Ma, J. Wang, L. Zhao, F. Nie, X. Ma, W. Lü, S. Yan and L. Zheng, Adv. Funct. Mater., 2024, 34, 2409045.
188. M. Ueland, S. Harris and S. L. Forbes, Forensic Sci. Int., 2021, 323, 110781.
189. C. Wang, L. Yin, L. Zhang, D. Xiang and R. Gao, Sensors, 2010, 10, 2088–2106.
190. M. Dai, M. F. Demirel, X. Liu, Y. Liang and J.-M. Hu, Comput. Mater. Sci., 2023, 230, 112461.
191. M. Dai, M. F. Demirel, Y. Liang and J.-M. Hu, npj Comput. Mater., 2021, 7, 103.
192. T. Wen, Q. Nie, L. Han, Z. Gong, D. Li, Q. Ma, Z. Wang, W. He, L. Wen and H. Peng, Food Chem., 2022, 387, 132905.
193. L. Uzun and A. P. F. Turner, Biosens. Bioelectron., 2016, 76, 131–144.
194. K. Cheng, T. Fan, Y. Jin, Y. Liu, T. Chen, D. Papadopoulos and Q. Yang, IEEE Intell. Syst., 2021, 36, 87–98.
195. I. Kholod, E. Yanaki, D. Fomichev, E. Shalugin, E. Novikova, E. Filippov and M. Nordlund, Sensors, 2020, 21, 167.
196. L. Witt, M. Heyer, K. Toyoda, W. Samek and D. Li, IEEE Internet Things J., 2023, 10, 3642–3663.
197. J. K. Eshraghian, M. Ward, E. O. Neftci, X. Wang, G. Lenz, G. Dwivedi, M. Bennamoun, D. S. Jeong and W. D. Lu, Proc. IEEE, 2023, 111, 1016–1054.
198. S. Ghosh-Dastidar and H. Adeli, Int. J. Neural Syst., 2009, 19, 295–308.
199. L. Antwarg, R. M. Miller, B. Shapira and L. Rokach, Expert Syst. Appl., 2021, 186, 115736.
200. X. Li, W. Sun, W. Fu, H. Lv, X. Zu, Y. Guo, D. Gibson and Y.-Q. Fu, J. Mater. Chem. A, 2023, 11, 9216–9238.
201. W. Xuan, M. He, N. Meng, X. He, W. Wang, J. Chen, T. Shi, T. Hasan, Z. Xu, Y. Xu and J. K. Luo, Sci. Rep., 2014, 4, 7206.
202. S. Y. Chun, Y. G. Song, J. E. Kim, J. U. Kwon, K. Soh, J. Y. Kwon, C. Kang and J. H. Yoon, Adv. Mater., 2023, 35, 2302219.
203. C. Sun, Q. Shi, D. Hasan, M. S. Yazici, M. Zhu, Y. Ma, B. Dong, Y. Liu and C. Lee, Nano Energy, 2019, 58, 612–623.
204. L. Xu, W. Xuan, J. Chen, C. Zhang, Y. Tang, X. Huang, W. Li, H. Jin, S. Dong, W. Yin, Y. Fu and J. Luo, Nano Energy, 2021, 83, 105814.
205. J. Yuan, Y. Zhang, C. Wei and R. Zhu, Adv. Sci., 2023, 10, 2303114.
206. B. Liu, T. Bi, Y. Fu, K. Kudara, S. Imanishi, K. Liu, B. Dai, J. Zhu and H. Kawarada, IEEE Trans. Electron Devices, 2022, 69, 949–955.
207. F. Adelantado, X. Vilajosana, P. Tuset-Peiro, B. Martinez, J. Melia-Segui and T. Watteyne, IEEE Commun. Mag., 2017, 55, 34–40.
208. M. A. M. Almuhaya, W. A. Jabbar, N. Sulaiman and S. Abdulmalek, Electronics, 2022, 11, 164.
209. K. Kaviyarasu, G. T. Mola, S. O. Oseni, K. Kanimozhi, C. M. Magdalane, J. Kennedy and M. Maaza, J. Mater. Sci.: Mater. Electron., 2019, 30, 147–158.
210. K. Zhao, G. Gu, Y. Zhang, B. Zhang, F. Yang, L. Zhao, M. Zheng, G. Cheng and Z. Du, Nano Energy, 2018, 53, 898–905.
211. S. H. Khan, A. Banerjee, K. H. Kim, J. Salvant, R. Looper, C. H. Mastrangelo and H. Kim, in 2020 IEEE 33rd International Conference on Micro Electro Mechanical Systems (MEMS), IEEE, Vancouver, BC, Canada, 2020, pp. 733–736.
212. F. Madaro, I. Mehdipour, A. Caricato, F. Guido, F. Rizzi, A. P. Carlucci and M. De Vittorio, Energies, 2020, 13, 4169.
213. C.-H. Chuang, D.-H. Lee, W.-J. Chang, W.-C. Weng, M. O. Shaikh and C.-L. Huang, IEEE Sens. J., 2017, 17, 2498–2506.
214. A. Kumar and N. Joshi, Nanobatteries and Nanogenerators, Elsevier, 2021, pp. 463–489.
215. M. Chae, D. Lee and H.-D. Kim, Micromachines, 2023, 15, 77.
216. J. Granstrom, J. Feenstra, H. A. Sodano and K. Farinholt, Smart Mater. Struct., 2007, 16, 1810–1820.
217. C. Li, F. Yang, P. Liu, C. Fu, Q. Liu, H. Zhao and P. Lin, Sensors, 2021, 21, 7708.
218. D. Punetha, M. Kar and S. K. Pandey, Sci. Rep., 2020, 10, 2151.

---