Thermal-stimulated phase transition of vanadium dioxide enabling versatile transduction and smart applications

Vanadium dioxide (⁠ $VO 2$⁠), a prototypical transition metal oxide with a strongly correlated system, has garnered significant attention as a promising smart material, primarily due to its remarkable metal–insulator transition (MIT) with reversible structural phase transformation. This transition fundamentally governs electronic, optical, and thermodynamic behaviors, coupled with dramatic changing properties in electrical conductivity, optical transmittance and emissivity, thermal conductivity, and a strain transformation with volumetric expansion.¹ Besides, this unique MIT characteristic, occurring near room temperature at approximately 340 K, makes $VO 2$ suitable for practical deployment with minimal energy input for activation and enhances the ability to respond to subtle environmental changes. Then, coupled with a sub-nanosecond transition speed, $VO 2$ becomes a candidate for use as a sensitive element in many cutting-edge applications, ranging from smart windows and energy-efficient buildings to advanced optoelectronics and next-generation information processing systems.^2,3 Notably, $VO 2$ is distinct from other MIT materials, in which the degrees of freedom of charge, lattice, orbitals and spin are strongly coupled, allowing for the existence of multiple phase transition mechanisms for the MIT of $VO 2$⁠. In theoretical studies, there are three generally accepted mechanisms for the phase transition of $VO 2$⁠, namely, the Peierls phase transition, the Mott phase transition, and the Peierls–Mott phase transition, explaining the relationship between the complex electronic structure and the structural phase transition observed in $VO 2$ from different perspectives, but still in the exploration stage.^4–6 

Thermally stimulated transitions have long been a focal point of research because they are achievable through various methods such as radiant heating, electric heating, conduction heating, and laser heating and can even be naturally triggered by ambient temperature changes. These varied thermal-stimulated approaches offer researchers significant flexibility in manipulating material properties,⁷ as shown in Fig. 1, thereby rendering $VO 2$-based sensitive elements adaptable to a broad of practical applications, including smart energy-efficient windows, camouflage, thermal rectification, switches, resonators, actuators, and temperature-sensitive devices. To date, most research focuses on the thermochromic property of $VO 2$⁠,⁸ and efforts include doping $VO 2$ with other elements to lower the transition temperature, synthesizing nanostructured $VO 2$ to achieve more uniform and faster switching, and designing composite materials to augment functionality.⁹ Additionally, investigations extend to exploring the potential of $VO 2$ in adaptive camouflage systems, where the color or reflectance of $VO 2$ changes in response to temperature can be utilized in military applications and bio-inspired designs.^10,11 In the field of temperature-sensitive devices, especially in flow rate detection, its potential is increasingly evident while still in the early stages of exploration. These $VO 2$-based sensitive elements exhibit an exceedingly high-temperature coefficient of resistance (TCR) near the phase transition point, which is pivotal for conducting high-precision temperature monitoring.¹² Also, this temperature sensitivity enables the development of advanced sensing systems capable of accurately measuring fluid flow rates based on the thermal method (anemometric or calorimetric), thereby opening up new possibilities for applications with the high-sensitivity and ultra-low flow range flow detection demand. Although the inherent nonlinear response of the material may initially be seen as a limitation to its efficacy as a sensitive material, further research into the modeling of $VO 2$ can help to capture the complex thermal hysteresis behavior under various temperature conditions and address the issue of inconsistent input–output relationships, which provides a basis for designing more precise control strategies and compensation techniques,^13,14 paves the way for the development of more accurate and reliable $VO 2$-based devices and opens up new possibilities for the development of novel sensing applications. Therefore, these challenges must be addressed when applying $VO 2$ to practical thermal-induced applications: (1) precise control over the transition temperature, (2) fabricating high-quality, functional thin films with consistent properties, (3) nonlinear response behaviors and thermal hysteresis effects, (4) scalable and cost-effective synthesis routes, and (5) energy efficiency during operation.

This article presents a comprehensive overview of $VO 2$⁠, especially focusing on its role as a multifunctional transduction element in the development of various domains. First, we introduce the electronic and lattice structural transformations process within the distinctive MIT behavior and delve into the influence of these transitions on material properties, such as optical transparency, electrical resistivity, thermal conductivity, and mechanical strain energy. Furthermore, we systematically review the diverse applications of $VO 2$ across four major domains: optical, electrical, thermal, and mechanics. Compared to previous reviews,^15–17 we mainly focus on the topic confined in the most widely used thermal-stimulated MIT of $VO 2$⁠, and its latest developments and applications over the past five years, particularly emphasizing $VO 2$’s role as a transduction medium that enables convert information into measurable physical properties changes in optical, electrical, thermal, and mechanics, highlighting its related critical research, recent advancements and breakthroughs, and the potential directions for future investigations in each field. Lastly, we also proposed a perspective on the progression of thermal-stimulated systems based on $VO 2$⁠, providing insights into optimizing material synthesis, device design, and performance, specifically addressing the challenges posed by the intrinsic thermal hysteresis behavior and fabrication complexities, aimed to pave the way for the next generation of high-performance multifunctional devices that will revolutionize fields ranging from environmental monitoring and energy management to biomedical diagnostics and advanced materials science.

The MIT in $VO 2$ constitutes a reversible, first-order phase transformation occurring around a transition temperature (⁠ $T c$⁠). This transition comes with the structural phase transition (SPT) from the metallic rutile phase, denoted as $VO 2$(R), to the insulating monoclinic phase, denoted as $VO 2$(M).^18, Figure 2(a) illustrates the tetragonal rutile structure of $VO 2$(R), with tetravalent $V 4 +$ ions situated at the corners and body center of a rectangular lattice, forming a slightly asymmetric octahedral $VO 6$ unit where each $V 4 +$ ion is surrounded by six $O 2 −$ ions. Along the c-axis, V-V chains align linearly with a lattice distance of 2.85 Å.¹⁹ The energy band structure of $VO 2$(R) shown in Fig. 2(b) reveals that under high temperatures, some of the $d / /$ orbitals of V in $VO 2$(R) are partially filled at the lower energy levels, with the higher energy states overlapping with $π ∗$ orbitals. The Fermi level $E F$ resides between the conduction band $π ∗$ and the valence band $d / /$⁠, enabling some of the originally localized d-orbital electrons to participate in electrical conductivity, thus making metallic behavior upon $VO 2$(R). Compared to $VO 2$(R), $VO 2$(M) is characterized by a distorted octahedral arrangement around the vanadium atoms with a lower symmetry, as shown in Fig. 2(c). The V-V alignment along the c-axis takes on a zigzag pattern with lattice distances of 3.19 and 2.60 Å. Figure 2(d) presents the energy band structure of $VO 2$(M), with the dimerization of $V 4 +$ ions along the c-axis and the displacement of the octahedral $VO 6$ unit, antibonding $π ∗$ orbitals elevate, leading to the splitting of the $d / /$-band into two subbands. The $E F$ lies within the forbidden gap, thus imparting insulating properties to $VO 2$(M).^20,21 However, the MIT mechanisms in $VO 2$ are still controversial, and the above-discussed in this section is based on the current mainstream understanding of the MIT/SPT, combining the Peierls model related to structural distortions and the Mott model related to electronic structure to comprehend the entire transition process.

Undergoing the MIT process, $VO 2$ undergoes crystal and electronic band structure alterations that inherently lead to property changes, including optical, electrical, thermal, and mechanical. In the high-temperature R-phase, $VO 2$ exhibits a significantly decreased transmittance in the infrared region, concurrently with a highly reflective, due to the overlap between $E F$ and $d / /$ vanishes the bandgap and enhances the absorption of incident photons, as shown in Fig. 3(a).²² These remarkable temperature-dependent optical properties make $VO 2$ a promising smart material for thermochromic applications, especially in the near-infrared range for smart windows and in the mid-infrared range for smart radiative devices. Therein, the development trend in the thermochromic $VO 2$ films for smart windows is especially toward achieving a balance between high visible light transmission and solar regulatory efficiency through material modifications, as illustrated in Fig. 3(b).²³ Besides, thermal effects on $VO 2$ electrical properties are particularly pronounced, which can be seen in Fig. 3(c). The resistivity has an over three-order magnitude change from a low-temperature insulating state to a high-temperature metallic phase, attributed to the structural transition resulting in a rearrangement of electronic states and the occupancy of d-orbitals enabling electrons to move more freely. It is worth noting that $VO 2$ boasts an ultrahigh TCR value, around $5× 10 5$ ppm/K, compared to conventional thermal sensing materials, as demonstrated in Fig. 3(d), indicating that even a small change in temperature can lead to a dramatic shift in its electrical resistance, which making $VO 2$ highly valuable for temperature-responsive devices such as bolometers, flow sensor, and others.²⁴ Additionally, thermal properties of $VO 2$⁠, including heat capacity and thermal conductivity, also undergo significant variation with the external thermal stimuli. In Fig. 3(e), $VO 2$ exhibits a lower thermal conductivity in the metallic phase than expected, contravening the classical Wiedemann–Franz Law, indicating that the transport behavior of electrons in $VO 2$ is independent of heat energy transfer.²⁵ This unique characteristic opens avenues for utilizing $VO 2$ in thermoelectric devices and smart thermal management systems, enabling more efficient control of heat energy. Regarding the mechanical property of $VO 2$⁠, Fig. 3(f) depicts the thermodynamic stability regions for the various phases displayed by $VO 2$ (⁠ $M 1$⁠, $M 2$⁠, or R), which also indirectly reveals the mechanical strength under the corresponding conditions.²⁶ During the phase transition process, vanadium ion pairs undergo dimerization and tilt relative to the tetragonal c-axis, leading to an approximately 1% expansion along the c-axis and an overall volume expansion of 0.3%, as illustrated in the inset of Fig. 3(g).^27,28 This reversible expansion causes changes in the generation and distribution of internal stresses of $VO 2$⁠, leading to an increase in the strain energy density and approaching 1  $J / cm 3$⁠, nearly three times more than that of human muscles shown in Fig. 3(h).²⁹ By taking advantage of this superior mechanical response during phase transitions, $VO 2$ has great potential for designing deformable structures and integration into and microelectromechanical systems.

Recently, the thermochromic properties of $VO 2$⁠, marked by a dramatic change in the transmittance along with the MIT, have received much attention in the field of smart windows. This technology offers new possibilities for energy-efficient buildings and intelligent light regulation by virtue of its ability to autonomously regulate the amount of solar energy (⁠ $Δ$T $s o l$⁠) that enters the room while maintaining the visible light transmittance (⁠ $T l u m$⁠) with external temperature changes (⁠ $Δ$T), as shown in Fig. 4(a).^8,30 However, there are still several technical challenges in practical applications, such as higher intrinsic phase transition temperature (⁠ $T c$⁠), lower visible light transmittance and solar energy regulation efficiency, and lower color comfort (brownish-yellow).³¹ In response to the above challenges, researchers are continuously exploring new material synthesis pathways, microstructure modulation, elemental doping, and composite material design strategies, aiming to optimize the thermochromic properties of $VO 2$ and give the film more functionality to meet particular needs.³² 

To begin with, the elemental doping strategies are widely investigated to lower the $T c$ to near room temperature in the range of 293–308 K. Figure 4(b) shows the incorporation of dopants such as tungsten (W) or boron (B), both of which can reduce the T_c to ambient levels.³³ This phenomenon can be explained by an increase in the concentration of electrons occupying the d-orbitals through doping, leading to a narrowing of the bandgap, and less thermal energy is required for electron movement, thereby facilitating the reduction of the $T c$⁠. Also, the dopant ions will replace $V 4 +$ or $O 2 −$ in $VO 2$⁠, destroying the bonding of V-V chains in the monoclinic structure and then reducing the $T c$⁠. Besides, the co-doping method utilizing the synergistic interaction among different dopants, like Sn-W, W-Sr, and Fe-Mg, can be feasible to achieve overall performance (⁠ $T c$⁠, $T l u m$⁠, and $Δ T s o l$⁠) improvements in smart windows that might not be possible with mono-doping.³⁴ Notably, the strategic co-doping of hydrogen and oxygen vacancies, which is the tailored control over synthesis parameters, resolves the trade-off between lowering the $T c$ and maintaining the desired MIT performance.³⁵ However, the study of elemental doping in $VO 2$ thin films should be further investigated, as its outcomes are influenced not only by the choice of dopant ions but also by the adopted synthesis method and the specific properties of the ion precursors employed.³⁶ 

Furthermore, multilayer, porous artificial nanostructures and composite material film design represent another performance improvement strategy in smart windows.³⁷ For instance, introducing the anti-reflective (AR) layer with destructive interference, such as $Cr 2 O 3$ and $SiO 2$⁠, as the top and bottom layers in the $VO 2$ sandwich structure, can maximize the $T l u m$ and block over 95.8% of ultraviolet radiation.^38, Figure 4(c) shows an innovative $VO 2$-integrated movable AR structure containing the two-phase (liquid-gas) regulation effect of solvents, replacing the traditional solid-only AR layer, which can obtain both high $Δ$T $s o l$ (20.8%) and T_lum (42.5%).³⁹ In addition to adding a continuous film on the $VO 2$ surface, the modulation of the morphology of the film, like porous or grid structures, kirigami-inspired designs, can also result in a transmittance increase and reflective reduction, especially with additional potential when incorporating morphing mechanisms.⁴⁰ As presented in Fig. 4(d), a self-rolling $VO 2$ nanofilm structure has a synergistic interplay between shape deformation and light transmission, which has led to significant breakthroughs in enhancing solar modulation performance with ultrahigh $Δ$T $s o l$ (42.14%) and $T l u m$ (61.01%), achieving the highest values among all $VO 2$-based smart window systems, and enabling multi-level temperature modulation according to the season and climate automatically.⁴¹ 

Likewise, several investigations have been conducted on the composite film design to realize the color tunability of $VO 2$⁠, addressing limitations in visual appeal and the commercialization impact of smart window products brought by the inherent brownish-yellow color.⁴² Typically, the composite film is composed of $VO 2$ nanoparticles and various color-changing materials like organic compounds, hydro-condensates, and ionic liquids. Therein, the coordination complex formed by combining metal ion coordinating sites IL-Ni-Cl (Ionic liquid-nickel-chlorine) with $VO 2$ results in a composite film exhibiting exceptional thermochromic properties, which demonstrates a remarkable color shift, transitioning from a light brown at low temperatures to a deep green at elevated temperatures.⁴³  Figure 4(e) introduces a $VO 2$ composite film structure design of a periodic silver (Ag)-nanodisk array with the localized surface plasmon resonance effect, regulating color within the entire visible spectrum by adjusting the spatial periodicity of the array and the diameters of the embedded Ag nanodisks.⁴⁴ 

Although the research on $VO 2$-based thermochromic smart windows has made fruitful progress, some directions in the practical application of $VO 2$ thin films still need to be explored in depth. At first, developing efficient, cost-effective methods for producing large-area, uniformly thick $VO 2$ films with consistent performance necessitates breakthroughs in materials synthesis and deposition techniques. Then, research should concentrate on enhancing the structural stability and durability of the $VO 2$ films, including measures to prevent oxidation, mitigate phase transition fatigue, and increase resistance to environmental erosion. Finally, optimizing smart window design for indoor human comfort by modulating the skin-sensitive band in the infrared region. In summary, $VO 2$-based thermochromic smart windows are still necessary to continue exploring and innovating to realize their wide application in building energy saving and improving the quality of life.

Beyond the transmittance characteristics, the capability of $VO 2$ in dynamically modulating its infrared emissivity (⁠ $ϵ$⁠) has been widely investigated and utilized as “smart” thermal emitters, especially in camouflage and passive radiator applications.⁴⁵ These thermal emitters can automatically respond to environmental stimulation (⁠ $Δ$T) and transfer reversibly between the high emissivity $ϵ H$ metallic state (⁠ $T> T c$⁠) and the low emissivity $ϵ L$ insulating state (⁠ $T< T c$⁠), enabling manipulation of its infrared radiation (IR) power (⁠ $Δ$P $r a d$⁠) within the infrared spectrum, bringing great prospects for military and aerospace applications.^45,46 However, some drawbacks have limitations on the practical use, and the central problem among them is low infrared emission tunability (⁠ $Δϵ$⁠, defined as the difference between $ϵ H$ and $ϵ L$⁠). The current research mainly focuses on structure design, such as optical cavities and metasurface, and composite with other materials to achieve the desired modulation in $Δϵ$⁠.

Constructing Fabry–Pérot (F-P) optical cavity structures has been demonstrated as a highly effective strategy for enhancing the $Δϵ$ of $VO 2$⁠.⁴⁷ For a typical F-P cavity, an infrared transparent layer is inserted between $VO 2$ and a highly reflective substrate (normally consisting of a metal material) to form a $VO 2$ transparent cavity interference filter, as the inset graphs depicted in Fig. 4(f). When $VO 2$ transforms to the metallic state at high temperatures, the $VO 2$-based F-P cavity structure forms a resonance, resulting in a very high IR absorption and enhancing the infrared emission, also shown in Fig. 4(f).⁴⁸ Notably, the material selection and thickness of the infrared transparent layer have greatly influenced the infrared emission. For instance, the F-P cavity consists of $VO 2/ SiO 2$/Au, achieving good performance with $Δϵ$ value of 0.49, but has high $ϵ L$ due to the absorption peaks near 9  $μ$m in $SiO 2$ being relatively close to the peak emissivity of the blackbody at 25  $°$C.⁴⁹ Then, selecting the hafnium oxide (⁠ $HfO 2$⁠), with an ultrahigh infrared transmittance within 5–12  $μ$m, instead of $SiO 2$ as an infrared transparent layer can efficiently reduce the $ϵ L$ and enhance $Δϵ$ value up to 0.55.⁵⁰ Besides, according to Kirchhoff’s law, the $Δϵ$ increases with a greater thickness of $HfO 2$ due to more infrared absorption.⁵¹ Also, a greater thickness of the $VO 2$ layer, from 120 to 360 nm, can efficiently improve the $ϵ H$ and $Δϵ$⁠, with both improvements of around 2.3 times.⁵² 

Metasurfaces, on the other hand, are artificially engineered subwavelength structures normally designed with an array of $VO 2$ patterns,⁵³ as shown in Fig. 4(g). This structural design with different feature sizes can interact with the metallic phase of $VO 2$ to form a localized surface plasmon resonance effect and then lead to an increment in $ϵ H$⁠. In the low temperature, it exhibits a slightly lower $ϵ L$ compared with that of the unstructured $VO 2$ film, determined by the reduction of the infrared absorption with a decrease in the covered feature size. Thus, the metasurface design of $VO 2$-based emitters has a 30% $Δϵ$ enhancement, providing an efficacy strategy in modulating thermal radiation characteristics.⁵⁴ In addition, the array geometry design of the metasurface affects the shift of the absorption peak position. Specifically, the decrease in gap size has a similar effect to increasing the length, both acting to large the coverage area and causing a dipolar coupling between each unit in the $VO 2$ array, then enhancing the plasma resonance effect and leading the absorption peaks to increase and shift to longer wavelengths, which allows for camouflage in more complex environments and facilitates the development of passive radiators.⁵⁵ 

To further develop infrared emissivity materials that exhibit stable performance when subject to external environment change, an adaptive control thermal emitter has been put forward, consisting of a three-layer structure composed of $VO 2$/graphene/carbon nanotubes, through current heating to reduce the IR intensity, enabling rapid switching for thermal invisibility, and boasting low power consumption and great reliability.⁵⁶ Then, a graded doped W $VO 2$ thin film structure has been devised for emissivity regulation without the need for external energy sources, as depicted in Fig. 4(h). By programming the doped material so that its $ϵ$ follows an inverse $T 4$ relationship with temperature, which effectively cancels out the inherent $T 4$ dependency, making the surface thermal radiation nearly independent of temperature variations. This allows the emissivity to be manipulated arbitrarily within a wide temperature range, which also can maintain a consistently stable infrared signature despite rapid temperature fluctuations, offering an advantage over traditional strategies that rely on feedback loops to adjust emissivity.⁵⁷ In addition, a $VO 2$ microdevice based on integrated Joule heating triggering has been developed to demonstrate shape transformation controllable capabilities through electronically programmed pulses, providing technical and theoretical support for the development of next-generation intelligent, adaptive infrared camouflage.⁵⁸ 

In short, the current research on $VO 2$ emissivity modulation for military camouflage and aerospace passive radiators is still in the primary stage and faces a few challenges. First, the optical properties of $VO 2$ changing accordingly in the visible, near-infrared, mid- and far-infrared, and terahertz bands should be considered simultaneously when regulating the emissivity, also combining the requirements of different applications. For example, smart radiator applications mainly require infrared variable emission, but adaptive camouflage technology involves the adjustment of optical properties across multiple spectral bands so that the target can real-time adjust in response to changes in the surrounding background, thereby realizing a multi-band camouflage effect. In addition, high solar absorptivity remains a significant barrier to effective passive radiator applications,^59,60 with the thickness, surface preparation, and fabrication techniques of the coating material all having a significant impact on this parameter, thus further research could be center on these directions to reduce the solar absorptivity in order to minimize the heat input while maintaining good tunability, contributing to the development of more effective smart radiator devices. Then, due to the limitation imposed by fabrication technique levels and the inherent complexity of structural materials, especially in the metasurface, research mainly showcases small-scale prototypes rather than realizing large-scale flexible manufacturing processes. Therefore, the popularization of $VO 2$ as a “smart” thermal emitter for practical applications remains for extensive investigation and development.

$VO 2$ has an ultrahigh TCR value around $T c$⁠, significantly reflecting the change in resistance (⁠ $Δ$R) corresponding to the small temperature variation of the material caused by thermal stimulation with the absorbed IR, thus making it an ideal thermal sensing material for IR intensity (⁠ $Δ$I) detection, especially as the bolometer basing the thermal detecting principle. The $VO 2$-based bolometer, as an uncooled detector, has an average specific detectivity of around $2× 10 7 cm Hz 1 / 2 W − 1$ and can be operated near room temperature without requiring refrigeration, resulting in significant reductions in size, weight, and energy consumption.⁶¹ Despite this, there are several limitations that need to be addressed, including the nonlinear resistance response with a harmful hysteresis behavior within the MIT process of $VO 2$⁠, which is the detecting material undesirable, and the inherent poor detecting performance of the bolometer compared with the photovoltaic infrared detectors, such as the detectivity and response time.

Although the presence of thermal hysteresis and high sheet resistance in $VO 2$⁠, which can affect the readings and increase the noise figure, $VO 2$-based bolometers still have been widely investigated. To overcome this, recent studies have explored doping techniques to modify the $VO 2$ film morphology and can mitigate or even eliminate the thermal hysteresis effect.^62,63 As shown in Fig. 5(a), the TCR value of $VO 2$ film can be improved by W-doped from 2–3%/K to around 8%/K, and further, $W x V 1 − x O 2$ nanowire (NW) can reach a TCR value of over 10%/K and a low sheet resistance of $10 − 3Ω$/cm.⁶⁴ In addition, combining $VO 2$ with other materials that exhibit high light absorption capabilities is another strategy being explored to enhance its performance. A $VO 2$/carbon nano coils (CNCs) composite material has been developed and exhibits an ultrahigh TCR value of 35%/K, then be applied to fabricate the infrared microbolometers with a responsivity of $3.3× 10 5$ V/W, low response time (⁠ $∼$2 ms), and a non-hysteresis response working in the MIT region.⁶⁵ However, this composite material’s TCR is greatly influenced by the crystallinities of $VO 2$ and stresses on the CNC substrates. Thus, further research is needed to manufacture uniform, controllable, high-performance sensing materials.

The structural design of the $VO 2$-based bolometer, such as the microbridge, antennas, microtubular, and other types, can further improve the IR absorption efficiency and enhance the detection performance. A nanorod dimer-based metal nanoantenna, incorporating a nanogap feature filled with $VO 2$⁠, has been proposed as a high-performance bolometer. Compared with the conventional optical cavity, this structure can enhance the local electric field within the gap and generate localized areas of great heating performance, thus improving the sensitivity of the $VO 2$ resistance to long-wavelength infrared radiation. By adjusting the nanorod length and the nanogap width, this nanorod dimer nanoantenna can achieve tuning of the response to a specific wavelength with an ultrafast operating speed (⁠ $∼$picosecond), which is suitable for multi-spectral range detection.⁶⁶ Besides, as illustrated in Fig. 5(b), the $VO 2$ nanofilm has been rolled up into a microtube configuration, which provides a wide detecting angle range of $150 °$⁠, effectively minimizing heat conduction through a substrate that attains superior thermal insulation and also serving as an effective light trap to augment infrared absorption capabilities. Also, this tubular bolometer can achieve a noise-equivalent temperature difference of 64.5 mK and a much higher specific detectivity of $2× 10 8$ cm  $Hz 1 / 2 W − 1$ compared with the unreleased and released nanofilm structure. Notably, it also has been verified that this novel tubular $VO 2$-based bolometer demonstrates competitive performance advantages among typical bolometers, as demonstrated in Fig. 5(c), hopefully surpassing current photovoltaic infrared detectors through further optimization, opening up a new development direction for the long infrared region thermal detection technology.⁶⁷ 

In summary, the $VO 2$-based uncooled infrared detection technology demonstrates great potential for the development of infrared sensing, with its performance already approaching that of cooled infrared photon detectors. However, to fully harness this potential, several key aspects must be considered. The primary issues are the compatibility of manufacturing technology and the achievement of miniaturization and integration, which necessitates breakthroughs in material synthesis and processing. Another vital issue is the requirement for an extended operating temperature range, ensuring dependable detection under extreme conditions. In addition, the inherent hysteresis in $VO 2$ will lead to different temperature readings. Although doping techniques have been used to mitigate or even eliminate the impact of this hysteresis, they introduce additional complexities in the fabrication process. Therefore, developing accurate hysteresis modeling methods based on the underlying phenomena and mechanisms is a potential alternative approach, which can help to capture the resistance-temperature relationship and compensate hysteresis effect, also it has shown feasibility in other $VO 2$-based applications. Thus, further research focusing on nano-engineered structures and integration with emerging technologies could lead to a new generation of high-performance sensing solutions.

$VO 2$⁠, a typical Mott material, undergoes a reversible switch between a high-resistance state (HRS) and a low-resistance state (LRS), causing several orders of magnitude changes in resistance (⁠ $Δ$R) and realizing the modulation of their electric charge history (⁠ $Δ$q) with an ultrahigh switching speed within the sub-nanosecond and a large ON/OFF ratio, which makes it an ideal candidate for building high-performance, low-power, fast-response memristors.⁶⁸ The $VO 2$-based memristor, which operates via a thermal-triggered threshold switching (TS) mechanism, primarily exhibits volatile resistance changes driven by MIT. This volatile characteristic makes it suitable for application as a selector in resistive random access memory (RRAM) and neuromorphic physiological signal processing systems.⁶⁹ However, recent studies have shown that through specific modifications, such as high-temperature treatment or special doping techniques, $VO 2$ can be engineered to exhibit a non-volatile characteristic. This non-volatile behavior can achieve stable learning and memory operations and is highly promising for advanced non-volatile memory devices.⁷⁰ These two distinct properties of $VO 2$-based memristors provide flexible and efficient solutions for different information processing and storage applications. Despite these attractive properties, the $VO 2$-based memristor still faces challenges in practical applications, such as reproducibility of the phase transition, thermal stability, and CMOS process compatibility. Thus, current research focuses on material modification, device structure design, and thermal management strategies to reduce energy consumption and effectively integrate it into practical electronic systems.

To enhance the controllability of thermal activation, researchers have worked on the nanoscale device structural design and heating mechanisms. Figure 5(d) schematics a single device composed of thin layers of $VO 2$ integrated with a carbon nanotube (CNT) acting as a localized nano-heater, which can achieve sub-nanosecond periodic spike pulse in the absence of an applied capacitor, exhibiting fast discharge characteristics similar to those of biological neurons.⁷¹ Compared to $VO 2$-only devices, the introduction of the CNT heater enables precise control of heat transport to optimize thermal energy utilization and greatly affect the dynamic behavior of the device, such as an increase in frequency, a decrease in pulse energy, and a reduction in transient time scales by nearly three orders of magnitude. This $VO 2$-CNT composite structure also suggests a new method to achieve order-of-magnitude tuning of energy efficiency in electronic devices without the complex lithographic technique, which is important for the construction of efficient and scalable neuromorphic computing systems and other advanced electronic devices. Besides, the TS-based memristor with a Ti/Au/ $VO 2$/Ti/Au structure, as a selector working based on the Joule-heating induced MIT, realizes a range of benefits, including great stability observed in nearly 120 times cycles, threshold voltage tuneability with the temperature variation, and high selectivity (⁠ $∼$188), which is suitable for in-memory computing and conducive to the development of and future intelligent sensors.⁷² In practical applications, it is noteworthy that the performance of the $VO 2$-based devices is significantly influenced by the ambient environmental temperature. Then, the interplay between Joule heating, heat dissipation, and the external temperature on the $VO 2$ volatile memristor with the resistive switching mechanism shown in Fig. 5(e) has been thoroughly investigated, which can help to forecast the oscillator frequency with external temperature variation and highlights potential cross-talk issues effect, such as self-induced noise caused by local heating or cross-talk interference between adjacent $VO 2$ oscillators, providing technical guidance for the development of more efficient and adaptable neuromorphic systems.⁷³ 

Based on the above structural innovations and in-depth study of the thermal effect mechanism, further application exploration of $VO 2$-based memristors is showing a broad prospect, especially in the simulation of excitation and inhibition of biological neurons. Figure 5(f) demonstrates a thermoelectric logic circuit based on the $VO 2$ thin-film memristor, allowing the transmission of information over short distances utilizing heat as a carrier and over longer distances via electrical signals, akin to how biological neurons employ neurotransmitter molecules for local diffusion and axonal long-range electrical conduction.⁷⁴ Moreover, by investigating neuron models constructed using $VO 2$ active memristors, it has been possible to effectively embody the intricate dynamics and inherent randomness characteristic of biological neurons.⁷⁵ These found demonstrated that $VO 2$ memristor-based neurons are not only capable of realizing basic integrative functions, but can also simulate a richer array of biological neuron activities, such as threshold variability, refractory periods, and others, which signifies the great potential for building high-performance, scalable, and biologically plausible neuromorphic computing systems. However, these previous studies have shown that neuron models have primarily been limited to processing single-modal physical signals, which restricts their applicability in complex multimodal signal environments. Thus, a bio-inspired hybrid tactile-temperature synapse neuron has been developed in Fig. 5(g), consisting of a piezoresistive sensor serially connected to a $VO 2$ volatile memristor.⁷⁶ It leverages the voltage divider effect between the piezoresistive sensor and the $VO 2$ memristor, as well as the intrinsic thermal sensitivity resulting from the IMT, enabling sensing and fusion of the tactile and temperature signals, which are capable of recognizing complex multimodal patterns. This breakthrough holds significant implications for the advancement of future intelligent sensing systems, electronic skins, and human-machine interaction technologies, as it emulates the human sensory system’s ability to integrate diverse sensory information into a coherent perceptual experience. This dependency underscores the importance of understanding and controlling the thermal effects on the material’s electrical properties, especially since $VO 2$ exhibits a reversible IMT at a specific temperature point.

As mentioned before, the $VO 2$-based memristor has been implemented as the core component of neuromorphic systems, such as artificial synapses and artificial neuron functions, making it highly potential for applications in neuromorphic computing systems. Nonetheless, the path toward commercial viability of $VO 2$-based memristor technology is still fraught with hurdles. A key challenge is its performance stability over multiple read/write cycles, including drift in the resistor state and changes in the switching threshold. Besides, although $VO 2$ memristors theoretically enable low-energy operation, in practice, especially when triggering phase transitions, they tend to require large current densities, leading to localized overheating and high energy consumption. Last, the integration of $VO 2$ memristors into existing semiconductor technologies, especially in high-density integrated circuits, faces challenges of compatibility, interconnect technology, and thermal management.

$VO 2$ is an ideal material for developing a high-performance sensor due to its ultrahigh TCR around the $T c$ and rapid switching speed, especially in temperature and fluid sensing. This $VO 2$-based sensor, as a resistive sensor, allows the conversion of small temperature changes (⁠ $Δ$T) triggered by external factors, including the temperature (⁠ $Δ T d$⁠) or the flow rate (⁠ $Δ Q f l o w$⁠) to be measured, into electrical output signals (⁠ $Δ$R), making them potentially suitable for a wide range of applications, including human health and environmental monitoring systems.⁷⁷ Current research focuses on improving the temperature resolution of sensors, especially for skin temperature detection, which requires better than 0.1 K, and enhancing the stability of the sensors under different environmental conditions.⁷⁸ Furthermore, exploring flexible fabrication methods and advancing toward multi-parametric sensing capabilities are crucial steps in fully harnessing the potential of this exceptional material for the next generation of high-performance sensors.

Here, a study highlights the successful development of a printed $VO 2$ temperature sensor embedded within a smart wristband designed for remote skin temperature monitoring.⁷⁹ By means of tungsten doping, the TCR has a nearly doubled improvement, and the $T c$ was adjusted from 67 to 31  $°$C to match the human skin temperature range. Besides, considering that ambient humidity has a direct effect on $T c$⁠, shifting it toward a higher value, and the IMT phenomenon is even suppressed in arid conditions,⁸⁰ an encapsulation fluoropolymer layer with the capability against humidity has been coated on the surface of the printed $VO 2$ sensing layer to ensure that the sensor can still accurately detect subtle temperature variations on the skin, even under conditions of up to 90% relative humidity. This sensor can adapt to a moist skin surface environment from perspiration and achieve a high sensitivity of 2.78%/K with a 0.1 K resolution in the measuring range of 30–40  $°$C, which validates its suitability for point-of-care monitoring applications requiring continuous, real-time, and accurate temperature detection. In addition, a high-performance flexible fabric temperature sensor based on $VO 2$/PEDOT:PSS composites, as showcased in Fig. 5(h), was investigated and prepared by a low-temperature spraying method down to 75  $°$C.⁸¹ Also, choosing parylene as the encapsulation maintains good mechanical adaptability and stability and, most importantly, mitigates the effect of environmental humidity. This temperature sensor, through optimizing the ratio of composite materials, obtains a higher sensitivity of 2.712%/K with a resolution of 0.085 K compared to the untreated $VO 2$ sensor. Moreover, the outstanding breathability characteristic of flexible substrates utilized in these sensors ensures user comfort and compatibility with long-term skin contact, opening up new opportunities for smart wearable technology in healthcare, particularly respiratory monitoring.

Leveraging the high-sensitivity temperature sensing capability, the $VO 2$-based sensor can effectively capture the temperature fluctuations induced by fluid motion and realize accurate flow rate measurements, mirroring the functionality of a flow sensor. As shown in Fig. 5(i), a fast-response flexible flow sensor based on $VO 2$ thin films is presented and capable of real-time respiratory monitoring by the positive relationship between the electrical response and the breathing rate, particularly suited for clinical applications such as monitoring sleep apnea and other related breathing disorders.⁸² Additionally, this miniaturized sensor consisting of a PDMS/ $VO 2$/polyimide multilayer structure can be mounted on the human wrist to achieve a dual-parameters sensing capability for pulse and body surface temperature, which is attributed to the different frequency characteristics of the temperature and vibration responses, providing a portable and practical novel tool for health monitoring.⁸³ Furthermore, $VO 2$-based flow sensors also hold promise for extending into micro total analysis systems (⁠ $μ$TAS), where they can facilitate high-sensitivity detection in ultra-low flow rate regimes. Figure 5(j) depicts a $VO 2$-based thermal flow sensor with an anemometer configuration suited for detecting gas flow rates in the microliter per minute range.⁸⁴ Notably, $VO 2$ exhibits an obvious hysteresis effect in the resistance-temperature characteristic, represented by a major loop and multiple minor loops inside. As the temperature fluctuates non-monotonically and falls in the minor loop, the TCR undergoes a significant decrease, resulting in a reduction of sensitivity and narrowing the dynamic range. Thus, to effectively leverage the ultrahigh TCR property to enhance sensor performance, the working principle of this anemometer sensor is studied in-depth, and the optimal operating temperature corresponding to the maximum TCR value can be realized by controlling the heating current. The resistance change in the $VO 2$-based anemometric flow sensor is 59.1% response to the flow range of 0–37.8  $μ L s − 1$⁠, which is nearly 15 times higher than that in the conventional Pt-based sensor, proving the excellent performance and their potential for high-sensitivity flow detection applications. In order to improve the sensitivity of the anemometric micro-flow sensor, a calorimetric micro-flow sensor based on a dual-heater structure is proposed, as illustrated in Fig. 5(k).²⁴ This configuration consists of two thermistor-heater pairs located upstream and downstream, which can be independently controlled, ensuring the initial temperatures of thermistors maintain at the maximum TCR temperatures in the cooling and heating curves, respectively. The results show that the theoretical maximum sensitivity of this dual-heater calorimetric flow sensor is about 1.34 V/(⁠ $μ L min − 1$⁠) in the flow range of 0–0.2  $μ L min − 1$⁠, which is 80.5 times higher than that of the anemometric microfluidic sensor, demonstrating a breakthrough of utilizing the phase change material with a nonlinear response to achieving ultrahigh sensitivity in low flow rate detection.

Although some research results have been achieved in both theoretical and experimental aspects, a number of difficulties remain that need to be solved in future research. Therein, the presence of hysteresis behavior in $VO 2$ is an undesirable characteristic for sensing materials, and its intricate nonlinear responses still remain incompletely understood. Current research only considers the operating situation in which the temperature is located on the major hysteresis curve, without a reliable model describing the accurate characterization of the resistance of the $VO 2$ at arbitrary operating temperatures, thus absent theoretical basis for further optimizing the $VO 2$ micro flow sensor. Besides, in device preparation, there exists a pressing requirement to advance toward miniaturization and multifunctional integration of devices, as well as improving their sensitivity and long-term stability of sensors in a variety of environmental conditions.

$VO 2$ is one of the candidate materials for the construction of thermal management and energy conversion devices, such as thermal diodes, due to its unique thermal property in MIT, possesses a relatively low thermal conductivity (k) in the low-temperature insulating phase, and transitions into the high-temperature metallic phase accompanied by dramatic increases in k.⁸⁵ The $VO 2$-based thermal diode properly utilizes this characteristic to regulate and control the direction of heat flow, akin to the electronic diode, facilitating efficient heat conduction in unidirectional while significantly reducing it in the opposite direction, which can be widely used in various applications such as thermal logic circuits, solar energy harvesting systems, and smart building refrigeration systems.⁸⁶ However, achieving an ultrahigh rectification ratio $η$⁠, which quantifies the disparity in heat conduction efficiency between bidirectional in a thermal diode, presents significant challenges and is cost-intensive due to the requirement of a temperature difference greater than 100 K to attain this effect.⁸⁷ Considering the $η$ mainly depends on the difference between the k of the two terminal materials and the temperature gradient. Therefore, research focuses on two core aspects, material selection and structural design, aiming at optimizing the heat flow path to enhance the performance of thermal diodes.

The conventional conductive thermal diode with a planar structure, as illustrated in Fig. 6(a), typically consists of two materials with different thermal conductivities, generally with phase change material (PCM) and non-PCM.⁸⁸ When the thermal diode is configured in the forward direction, the PCM terminal with high temperature exhibits high k and allows for freer heat transfer to the cold non-PCM terminal. Conversely, under the backward configuration with the temperature gradient flipped, $VO 2$ shifts to its low k state, offering greater resistance to heat flow, thus exhibiting an asymmetric heat transfer behavior. There has been research investigating the effect of different phase change materials, like $VO 2$ and nitinol, on the performance of thermal diodes. The $VO 2$-based thermal diode demonstrates a $η$ of 19.7%, along with a temperature differential of 69.5 K between its two terminals. This performance surpasses that of a nitinol-based thermal diode, illustrating that incorporating a PCM with attributes such as higher k, minimal thermal hysteresis, and rapid phase transition capabilities greatly bolsters the thermal rectification abilities of such devices.⁸⁹ To further enhance the rectification performance of thermal diodes, a strategy involving the use of dual PCMs, for instance, combining $VO 2$ with polyethylene, is proposed instead of conventional designs that pair a PCM with a non-PCM. This dual PCM approach with two opposite temperature dependence materials demonstrates strong thermal rectification with an ultrahigh $η$ of approximately 60%, which is nearly three times of the single PCM diode.^90,91 In addition, research has also delved into the impact of geometry on thermal diode performance. As depicted in Fig. 6(b), the thermal diodes based on spherical and cylindrical configurations show the $η$ of 20.8% and 20.7% with a PCM/non-PCM composition and also exhibit the $η$ of 63.5% and 63.2% within the dual-PCMs structure.⁹² Compared with the planar structure, the design of these diode structural geometries significantly governs the temperature distribution and heat flow dynamics inside the component while having a marginal influence on the optimal rectification coefficient.

In summary, while the selection and optimization of PCMs play a central role, the geometric design of thermal diodes is another important aspect being explored to push the boundaries of thermal rectification performance and expand the practical applications of these devices. Notably, the majority of research focuses on the steady-state heat transfer process within thermal diodes and related devices, future directions in thermal diode technology are expected to delve deeper into transient and dynamic behaviors, exploring novel materials and configurations that can enhance the performance of these devices under varying temperature conditions and time scales.

Thermal regulator and switch, as depicted in the middle and right of Fig. 6(c), respectively, are the other two elements related to intelligent thermal management, the former focuses on the dynamic temperature control (⁠ $Δ$T) by self-feedback regulation of heat flux (⁠ $Δ$Q) with variable thermal conductance (⁠ $Δ$k), while the latter respond to temperature thresholds by executing binary actions.⁹³ Incorporating $VO 2$ in these switchable thermal components has garnered significant interest for the potential of enhancing the ability to manage heat flow with its rapid switching speed and thermal conductivity change triggered by temperature variations. To assess the effectiveness of the $VO 2$-based switchable components in thermal regulation, a key performance indicator known as the on/off switching ratio (r, expressed as $r= Q o n/ Q o f f$⁠) has been proposed.⁹⁴ The higher the ratio, the greater the contrast between its heat transfer capabilities in the high-conductivity “on” state (⁠ $Q o n$⁠) and low-conductivity “off” state (⁠ $Q o f f$⁠), indicating a superior thermal regulation performance. Considering that $VO 2$ thermal switches may encounter challenges in achieving comparable switching ratios (r < 2) to solid–liquid PCMs, as less pronounced thermal conductivity changes during solid-state transitions, efforts are focused on optimizing structural designs and material modifications to enhance performance, especially in terms of switching ratios.

One of the most effective strategies for boasting the greatest on/off switching ratio is precisely regulation of the physical contact state with well-designed nanostructures. As shown in Fig. 6(d), an all-solid-state thermal switch device featuring a nanogap design has been developed, employing the phase transition of the $VO 2$ thin film layer to precisely control the interfacial pressure and contact surface area between polycrystalline silicon layers, then achieve the opening and closing state of nanogap structure.⁹⁵ This unique design facilitates a more dramatic alteration in the thermal conductivity across a wide range of temperature conditions, as depicted in Fig. 6(e), thus obtaining a significantly higher on/off ratio of nearly 2.75 and a 670% enhancement compared to that with the stand-alone $VO 2$ thin film thermal switch. Nevertheless, it should be noted that the performance of the thermal expansion switchable device is normally influenced by large thermal hysteresis. Therefore, recent research is mainly based on planar structures and further optimization of the thermal properties of $VO 2$⁠. As shown in Fig. 6(f), the single-crystalline $VO 2$ nanobeams structure is pivotal for precise control and measurement of thermal properties.²⁵ This design eliminates the influence of extrinsic grain boundaries and stress effects, ensuring that heat flows along the same pathway. Besides, the introduction of dopants like tungsten (W) in $VO 2$ can efficiently manipulate the thermal conductivity, as illustrated in Fig. 6(g). With the W-doping fraction x increasing in $W x V 1 − x O 2$⁠, the phase transition temperature toward a lower temperature and the thermal conductivity exhibits a sudden increase throughout the IMT process due to electrons in the metallic phase becoming better heat conductors, which endows these materials with great potential for use as thermal switches. Notably, the result of undoped $VO 2$ has demonstrated an anomalous electron thermal conductivity in the metallic phase, which is much lower than predicted by the Wiedemann–Franz law.

Overall, $VO 2$-based switchable thermal devices theoretically offer great potential for applications in the fields of thermal management, energy regulation, and adaptive insulation materials due to their unique thermal properties within the MIT process. However, despite the promising prospects, the practical applications of these devices are still relatively limited, and there remain several controversies in thermal conductivity modulation explained by the electron transport mechanism, which needs to be overcome by further scientific research.

Considering a high-performance micro-actuator demands high displacement and large force output with a rapid response, $VO 2$ emerges as a highly promising alternative driving material for the development of micro-actuators, owing to its unique thermally driven structural phase transition with a picosecond-scale response time, coupled with the high elastic modulus (E, $∼$140 GPa) and large strain (⁠ $ϵ$⁠, $∼$1%).⁹⁶ By controlling the temperature variation (⁠ $Δ$T), the $ϵ$ value can be finely tuned, thus realizing effective control of the deflection in $VO 2$-based thermally activated actuators.⁹⁷ However, the $VO 2$-based actuators remain the trade-off between a swift response time and large deformation, as well as the balance of structural simplicity with the need for multidirectional driven capabilities. Besides, $VO 2$-based actuators also inherently face challenges such as hysteresis effects and the presence of undesirable initial curvature, which has hindered widespread application.^98,99 Therefore, current research primarily focuses on leveraging the design and fabrication of single-crystalline or polycrystalline structures, microstructures, and integrated microsystems to achieve efficient thermal actuation.

In the typical actuator with polycrystalline arrangements, the control over the deposition orientation is employed to engender asymmetric stress distribution within $VO 2$-based actuators, thereby triggering controllable mechanical deformations. For example, the $VO 2$ exhibits a preferred orientation along the (011) plane when grown on carbon nanotube (CNT) films, and the c-axis of $VO 2$ sits in the film plane, which leads $VO 2$/CNT actuators to shrink to the $VO 2$ side. This $VO 2$/CNT bimorph actuator, driven by the thermally triggered MIT, can reach a displacement-to-length ratio of nearly 0.4 and has a response time of less than 10 ms.¹⁰⁰ To realize a bidirectional and large amplitude actuation, a high-performance micro-actuator composed of the $VO 2/ Al 2 O 3$/CNT eccentric coaxial nanofiber with nano-scale diameter has been developed, as shown in Fig. 7(a).¹⁰¹ By leveraging the synergistic effect of the eccentric coaxial structure, the thermal expansion mismatch between the layers, and the structural change during the MIT, this actuator can achieve both positive and negative displacements with the corresponding displacement-to-length ratio of 0.8 and 0.7, reaching a shorter response time of 2.5 ms, corresponding to a frequency of approximately 400 Hz. Notably, $VO 2$-based polycrystalline actuators generally face a serious problem in that the actuation of the single grain occurs along the c-axis direction, yet adjacent grains with different orientations tend to reduce the deformation of the overall structure of the $VO 2$ film. Thus, a $VO 2$ nanowire array/CNT bimorph actuator with highly anisotropic behavior has been developed, as illustrated in Fig. 7(b), enabling maximize the desired deformation in the longitudinal direction (aligned with the c-axis).¹⁰² This design has demonstrated actuation performance comparable to that of single-crystalline actuators, boasting a giant displacement-to-length ratio of 0.83. Besides, it has a volumetric work density of 2.64  $J cm − 3$⁠, and a much lower response time of over 65 ms, mainly owing to the large size and high heat capacity of centimeter-scale actuators.¹⁰³ 

Considering the polycrystalline actuator typically features a complex fabrication process and especially exhibits a significant initial curvature, as illustrated in Fig. 7(c), attributable to residual stresses accrued during the fabrication procedure, research has veered toward the development of the single-crystalline actuator with a simpler structure and superior stability. This actuator comprises one-dimensional $VO 2$ crystals with a radial gradient MIT and exploits asymmetry strain induced by the differential phase distribution to accomplish self-bending. Also, it can realize an ultrahigh response speed with kHz and a displacement-to-length ratio of approximately 1.¹⁰⁴ In previous research, axial doping has been proven as an efficient method for creating MIT domain structures in $VO 2$ nanobeams, but it might bring a great structural distortion and also be challenging to modulate along the special direction in the narrow space.^64,105 Therefore, a study for the stoichiometry engineering strategy for the controlled manipulation of $VO 2$ phases was conducted, and the ordered and controllable phase transition paths were successfully realized. Based on this approach, a family of high-performance single crystalline $VO 2$ actuators, including $M 1$-R, T-M $2$⁠, and M $2$-R configurations, has been prepared and achieved an ultrahigh volumetric work density (around 19.3  $J cm − 3$⁠) among $VO 2$-based actuators and response speed with 5 kHz.¹⁰⁶ In addition, in-plane movement actuators have also been extensively researched due to their high efficiency and compact design. Figure 7(d) presents a $VO 2$-based in-plane actuator completely based on the MIT-driven principle with a chevron configuration to amplify the lattice expansion. This actuator retains the ultra-fast response characteristics (⁠ $∼$2 kHz) of a single-crystalline actuator and a bidirectional actuation, which makes it well-suited for biomedical devices.¹⁰⁷ 

In the future, high-performance $VO 2$-based actuator requirements can be met by exploring new design structures and materials synthesis technology to increase the displacement output and work density. Also, optimizing the manufacturing process and reducing costs while maintaining or improving performance makes $VO 2$-based actuators more commercially available for production and application. The last but most important is maintaining stable performance under various environments, such as different temperatures, humidity, and atmospheric pressure conditions, and enhancing their adaptability in harsh environments.

The $VO 2$ integrated micro-electro-mechanical system (MEMS) resonator represents an emerging tunable device that skillfully blends the unique MIT property of $VO 2$ material with the precision manufacturing process of MEMS. It leverages the temperature-dependent phase transition of $VO 2$ accompanied by drastic changes in $ϵ$ to dynamically regulate resonant frequency (⁠ $Δf$⁠), exhibiting a wide range of potential applications in communications, tunable filters, and modulators.¹⁰⁸ Until now, research on $VO 2$-based MEMS resonators is still in the early stages, and mainly efforts to expand the range of frequency tuning and improve the energy utilization efficiency in the tuning process through micromechanical structural design, such as the cantilever, micro-bridge, and membrane types. Also, several active tuning methodologies (e.g., Joule heating and substrate heating) have been applied to control the structural phase transition of $VO 2$ and investigate the effect on frequency tuning behavior to achieve greater control and predictability in the tuning process.¹⁰⁹ 

As shown in Fig. 7(e), the classic mechanical resonators with cantilever and bridge-type designs are presented, and they exhibit distinct mechanisms for implementing frequency selectivity due to the inherent structural disparities. Specifically, the cantilever resonator with a free end primarily relies on geometric bending deformation to modulate its resonant frequency (5.1–5.25 kHz, 2.9% increase in resonant frequency), whereas the bridge-type resonator with both fixed ends exploits stress effects within the structure more dominantly for frequency tuning purposes (237.5–257.5 kHz, 8.4% increase in resonant frequency).¹¹⁰ It found that bridge-based resonators tend to display a larger frequency tuning range and more efficient energy utilization during the frequency tuning process, serving as a foundation for designing tunable bandpass filters with adjustable bandwidth characteristics. Additionally, different bending configurations of bridge structures, such as downward curvature, upward curvature, and bell-shaped deflections, inherently induced by the different interface stress in the fabrication process, manifest distinctive frequency tunability with resonant frequencies of 234.5, 246.6, and 266.7 kHz. Thus, dynamic tuning of the resonator frequency can be adjusted by strategically manipulating the microstructure.¹¹¹ To makeup for the nonlinear response behavior in the above $VO 2$-based out-of-plane resonators, a planar bridge-shaped comb drive MEMS resonator with unique stress distribution and tuning dynamics has been tentatively proposed and depicted in Fig. 7(f). Here, the $VO 2$ coating experiences structural phase transition as temperature increases, engendering tensile stress that can offset the initial compressive stress-driven decrease in resonant frequency and lead to an increase of approximately 2% in the resonant frequency,¹¹² which provides a promising strategy for mitigating nonlinearities and enhancing tunability in $VO 2$-based MEMS resonators.

It is noteworthy that different thermal activation methods [Fig. 7(g)], such as Joule heating and substrate heating, exert distinct impacts on the frequency-tuning behavior of $VO 2$-based resonators by influencing the stress distribution and structural response. In Joule heating, the local heating effect causes the $VO 2$ to undergo a structural phase transition and release a large amount of strain energy, accumulating compressive stress along the longitudinal direction of the bridge structure, thus causing a monotonic increase in the resonant frequency (⁠ $∼$11.9%). By contrast, the resonator experiences uniform heating overall under substrate heating, altering the boundary conditions and internal stress state within the bridge structure, which causes a non-monotonic changing trend in the resonant frequency (10.6% increase and 7.8% decrease).¹¹³ Thus, selecting proper thermal activation methods is pivotal to developing high-stability and tunable $VO 2$-based MEMS resonators adaptable to various applications.

$VO 2$-based MEMS resonators have exhibited remarkable tunability owing to the high volume working density and MIT characteristic of $VO 2$⁠, which facilitate a wide range of changes in resonance frequency and enable dynamic and reversible control by temperature field trigger. These findings lay the foundation for advancing $VO 2$-based MEMS resonator technology development and present a developing direction for designing next-generation adaptable resonators and filters for communications, sensing, and signal processing applications. However, achieving a more controlled and predictable frequency modulation while addressing challenges related to fabrication complexity, Q-factor enhancement, and minimizing hysteresis remains a focal point for advancing the practical implementation of these devices.

In this review, we focus on the thermally triggered metal–insulator transition of $VO 2$ with reversible structural phase transformation and present the changes in the crystal structure and electronic band configuration of $VO 2$ during this process, as well as the accompanying changes in a range of properties, including optical, electrical, thermal, and mechanical properties. Following this, we summarize a number of corresponding important applications based on the above characteristics, covering smart windows, camouflage techniques, infrared detectors, memristors, various sensors, thermal regulation systems, actuators, and resonators in microelectromechanical systems. For each application, we outline the specific mechanism of $VO 2$ action and the outstanding characteristics that are leveraged, while also provide a comprehensive review of the progress that has been made in the field, pointing out the remaining challenges and looking ahead to possible breakthroughs for future research. Although $VO 2$ has demonstrated its core value in these applications and has made compelling advances with great potential for future development, there are a number of pressing challenges that need to be faced in order to realize its full potential for utilized in the cutting edge of thermally triggered sensing systems, which specifically manifested in the following aspects:

1. The harmonization and deeper comprehension of the MIT mechanisms in $VO 2$⁠. Regarding the physical mechanism of the MIT process, literature reports are still controversial, and the viewpoints contain three main types, i.e., the Mott transition with the electron-electron correlation-driven mechanism, the Peierls transition driven by lattice structural dynamics, and a combined mechanism in which electron correlation and lattice structure together drive the $VO 2$ phase transition. The mechanism can be further investigated by combining new research tools, such as ultrafast microscopy techniques and improved computational methods, to gain profound insights into its fundamental workings. Recent research on the ultrafast phase transition processes of $VO 2$ induced by optical and electrical stimuli has made significant progress.^114,115 These findings have revealed two stages in the structural phase transition: first, an ultrafast structural disruption occurring on a femtosecond timescale, during which atoms rapidly move to positions similar to those in the metallic phase, though the unit cell still retains its monoclinic structure; followed by a lattice expansion process on a picosecond timescale, where the unit cell changes from the monoclinic structure to the final tetragonal structure. The results of these transient phase dynamics provide strong evidence for the debate over whether a monoclinic metallic phase exists. From this discovery, the understanding of the $VO 2$ phase transition mechanism will help to improve the modulation of the MIT behavior and $VO 2$ properties, optimizing the performance of the material and extending its application range.

2. The undesirable nonlinear response with hysteresis behavior in sensing systems. This refers to a loop-like pattern feature observed in several properties of $VO 2$⁠, such as refractive indices, resistance, thermal conductivity, and strain, which would affect the response time and accuracy of the $VO 2$-based device, especially in fast temperature-controlled applications where a sensitive and consistent response is required. For instance, the thermal flow sensor with resistance-temperature hysteresis behavior would be complex to predict and control, leading to potential inaccuracies in the measurements. Thus, it necessitates further research to better understand and mitigate its effects and that could be approached from two methods: the thin film material fabrication technology and the hysteresis phenomenon modeling.

   • The hysteresis behavior is often influenced by factors such as impurities, defects, grain boundaries, and film thickness in the $VO 2$⁠, which could potentially minimize the effect by refining the synthesis and deposition techniques. For example, using advanced epitaxial growth methods to produce single-crystal $VO 2$ films with minimal defects or carefully controlling the annealing conditions to optimize grain size and orientation can lead to reduced hysteresis phenomenon. Additionally, doping $VO 2$ with other elements can efficiently adjust its phase transition behavior and also eliminate the hysteresis behavior.

   • Developing accurate models to describe and predict the hysteresis behavior in $VO 2$ is crucial for designing and controlling sensing systems effectively. This model can be established based on the phenomenon and mechanism, the former is based on experimental data and can be incorporated into the control algorithms of sensing systems to compensate for the hysteresis effect in real-time operations, and the latter requires a deeper investigation of the fundamental mechanisms behind the hysteresis behavior, which can help reveal the microscopic processes driving the hysteresis and guide the development of more sophisticated analytical models.

3. The large-scale manufacturing techniques and multifunction integration. The existing $VO 2$ thin film fabrication techniques often require sophisticated equipment and are highly dependent on the process parameters, also the large-scale synthesis of $VO 2$ films with controllable shape, crystal orientation, and morphology remains a considerable challenge. Consequently, there is a pressing need for the development of alternative methodologies that can offer enhanced scalability, cost-effectiveness, and ease of implementation without compromising the desired material properties. One promising approach is the integration of $VO 2$ with advanced printing techniques, such as inkjet printing, which has already shown great potential in depositing functional materials with high resolution and pattern flexibility, making it an attractive option for fabricating $VO 2$-based devices. Notably, there are some challenges, such as developing stable ink formulation, optimizing printing parameters, and establishing efficient post-processing techniques, which should be addressed to further enhance the compatibility and performance of $VO 2$ in inkjet printing applications, ultimately paving the way for a new generation of $VO 2$-based devices and coatings with improved functionality and accessibility. Besides, multifunction integration of $VO 2$ should ensure that the individual functions synergistically interact without compromising the performance of any one aspect. This requires not only advances in material processing but also in device structure design that can effectively harness these properties concurrently. This includes the design of heterostructures, where $VO 2$ is combined with other materials to create synergetic effects, or the use of flexible substrates for the creation of conformable devices that can intimately interface with biological tissues for health monitoring purposes.

Overall, the cooperation of materials science, advanced fabrication technology, and computational modeling is catalyzing innovations that could see $VO 2$-based technologies play a key role in energy-saving buildings, next-generation electronics, and wearable health monitoring systems. With continued interdisciplinary efforts and targeted research, $VO 2$ promises to develop as a high-performance, adaptable material that will address global challenges in energy consumption, information processing, and more.

This work was supported in part by the National Natural Science Foundation of China under Grant Nos. 62103369 and U21A20519.

The authors have no conflicts to disclose.

Yushan Zhou: Conceptualization (equal); Formal analysis (equal); Investigation (equal); Validation (equal); Visualization (equal); Writing – original draft (equal). Shanqian Su: Formal analysis (supporting); Investigation (supporting); Validation (supporting); Visualization (supporting). Ziying Zhu: Investigation (supporting); Validation (supporting). Dibo Hou: Funding acquisition (supporting); Project administration (supporting); Writing – review & editing (supporting). H. Zhang: Funding acquisition (supporting); Project administration (supporting); Supervision (supporting); Writing – review & editing (supporting). Yunqi Cao: Conceptualization (equal); Funding acquisition (equal); Project administration (equal); Supervision (equal); Validation (equal); Writing – review & editing (equal).

Data sharing is not applicable to this article as no new data were created or analyzed in this study.