Mini-review on VO₂-based Sensors Utilizing Metal-insulator Transition

3.1 Photodetectors

VO₂, with an energy bandgap of 0.7 eV, is a promising material for photodetectors owing to its ability to convert light from various wavelengths into electrical signals [7]. In addition, the MIT in VO₂ occurs at a sub-nanosecond timescale, enabling the development of high-performance ultrafast photodetectors. However, there are several challenges in applying VO₂ in photodetectors. It is attributed to the short lifetime of the photogenerated carriers and their insufficient concentration to trigger MIT [20,21]. To address these challenges, recent studies have reported the implementation of high-performance photodetectors employing heterojunctions and other functionalization approaches [10-12].

Hong et al. [10] studied the localized surface plasmon resonance (LSPR) and MIT induced by visible-near-infrared light in Ag-decorated VO₂ nanorod arrays (NRs) for applications in broadband photodetectors. As shown in Fig. 1 (a), V2O5 NRs with a thickness of 250 nm were deposited using an e-beam evaporator at a glancing angle of 85°. This deposition angle satisfied the requirement of exceeding 80°, as reported by Sun et al. [22], to achieve a uniform and well-ordered array of nanorods. Subsequently, VO₂ NRs with a diameter of 50 nm were fabricated via heat treatment in a reducing ambient. Further, Ag-decorated VO₂ NRs were prepared by depositing Ag top electrodes with a spacing of 100 μm via e-beam evaporation.

Fig. 1.

(a) Schematic of the fabrication of Ag-decorated VO2 nanorods (NRs). (b) Ratio of Rdark to Rphoto as a function of light intensity for the Ag-decorated VO2 NRs. Electronic band structure of Ag-decorated VO2 NRs (c) in the dark and (d) under illumination. Reprinted with permission from Ref. [10] Copyright (2019) American Chemical Society. (e) I–t curves of the Sb2Se3/VO2 heterostructure under different light intensities. (f) Wavelength dependent responsivities of the Sb2Se3/VO2 device. (g) Responsivities and current of the Sb2Se3/VO2 device as a function of light intensity. Reprinted with permission from Ref. [11] Copyright (2021) John Wiley and Sons.

To demonstrate that the electrostatic charging of the VO₂ surface can induce MIT behavior, finite-difference time-domain (FDTD) simulations were conducted under various incident light wavelengths. The electric field distribution at the interface between the Ag NPs and the VO₂ layer induced by LSPR was calculated, demonstrating a notable enhancement in the electric field at 663 nm (visible region) and 1000 nm (near-infrared region). The significant plasmon resonance frequency shift and enhanced electric field in specific regions observed in the Ag NPs support the occurrence of strong electron-electron coupling at the Ag NPs and VO₂ layer interface over a broad range of incident wavelengths.

As illustrated in Fig. 1 (b), the ratio of the device resistance in the dark (R_dark) to that under white-light illumination (R_photo) was plotted as a function of light intensity (L_light) at an applied voltage of 5 V. The Ag-decorated VO₂ NRs device exhibited a slight increase in the R_dark/R_photo ratio up to a light intensity of 50 μWcm^-2. Whereas, other devices exhibited negligible R_dark/R_photo ratios. Furthermore, as the light intensity increased up to 10 mWcm^-2, the R_dark/R_photo ratio of the Ag-decorated VO₂ NRs device gradually increased, whereas the other devices maintained a negligible R_dark/R_photo ratio below 1.3. Notably, at a white light intensity of 10 mWcm^-2, the R_dark/R_photo ratio of the Ag-decorated VO₂ NRs device reached 40.1, which was comparable to the value of 47.01 induced by thermally induced MIT. This finding demonstrates that the localized heating effect owing to the plasmonic NPs induces the MIT in the Ag-decorated VO₂ NRs device at light intensities exceeding 10 mWcm².

The photo-induced MIT in the Ag-decorated VO₂ NRs can be explained by the simplified band structure of VO₂ near the Fermi level (Fig. 1 (c), (d)). In the dark, the Ag-decorated VO₂ NRs exhibited a monoclinic VO₂ (M1) phase. The formation of V–V dimers in VO₂ (M1) resulted in alternating short and long V–V distances, leading to the localization of d-orbital electrons. The long distance of 3.12 Å corresponded to the separation between zigzag chains of V–V pairs, which elevated the 3d_|| and 3d_||^* states above the Fermi energy (E_F). The short distance of 2.65 Å between V–V dimers splits the nonbonding 3d_|| state into the occupied 3d_|| (lower energy) state and the unoccupied 3d_||^* (high energy) state. This resulted in the insulating behavior of VO₂ (M1) with a bandgap of 0.7 eV.

When light was incident on the Ag-decorated VO₂ NRs, the strong electric field generated at the Ag NPs and VO₂ NRs interface owing to LSPR induced changes in the electronic states of the VO₂ NRs. If this electric field is sufficiently strong to induce the MIT, a phase transition to the metallic tetragonal VO₂ (R) phase occurs. In this phase, the parallel alignment of V atoms along the rutile c-axis allows all metallic V atoms to share dorbitals and electrons, thereby forming a conduction band composed of 3d_|| and 3d_||^* states. Consequently, the Fermi level is positioned within the overlapping region of the 3d_|| and 3d_||^* states. This theory, combined with earlier FDTD simulation results, demonstrates that the photo-induced MIT induced by LSPR can lead to a structural phase transition (SPT) in Ag-decorated VO₂ NRs across a broad range of incident wavelengths.

A heterojunction is well suited for enhancing performance metrics, such as responsivity and response time, by increasing the lifetime of charge carriers through the separation of electrons and holes. Consequently, recombination owing to the bandgap difference between the two materials is reduced. Xin et al. [11] reported a VO₂/Sb₂Se₃ heterojunction device that maximized the charge separation efficiency and enabled the development of a self-powered photodetector (Fig. 1 (e), inset). The type-II heterojunction formed between the VO₂ and Sb₂Se₃ facilitated efficient charge separation. When light is incident on a heterojunction, electrons in the valence band are excited to the conduction band, creating electron-hole pairs (EHPs). The built-in electric field within the heterojunction drives the separation of EHPs, with electrons migrating to Sb₂Se₃ and holes migrating to VO₂. This separation induces a higher electric potential on the VO₂ side, thereby generating a light-driven current that flows from VO₂ to Sb₂Se₃ in the external circuit. Consequently, a high-performance self-powered photodetector is realized.

Sb₂Se₃ exhibits high absorption in the visible and near-infrared regions, whereas VO₂, with its MIT characteristics, is responsive across a wide range of wavelengths. The VO₂/Sb₂Se₃ heterojunction device operates over a broad wavelength range of 365–1000 nm, with a particularly outstanding responsivity at 520 nm. Fig. 1 (e) shows the transient current curves of the device under different light intensities (1–50 mWcm^-2) at 520 nm. As the light intensity increased, the current magnitude continued to rise, and similarly, the responsivity increased, reaching its peak at 7 mWcm^-2 (Fig. 1 (e), (f) : 0.234 A/W). Beyond this intensity, the responsivity decreased with a further increase in the light intensity. In addition, the device exhibited a fast response speed, with rise and decay times of 200 and 360 μs, respectively, as a film photodetector. The ability to generate a high photocurrent of 3 μA without an external bias further underscores the significant potential of the VO₂/Sb₂Se₃ heterojunction device for optoelectronic applications.

Owing to the MIT characteristics of VO₂ and its excellent photoresponsivity, it can be utilized in a wide range of applications, including ReRAM, smart windows, and nanoantennas. Zhou et al. [12] leveraged these properties to develop a threshold switching selector for photodetectors in the NIR (near-infrared) region and for optical neural network applications. Fig. 2 (a) and (b) depict the current-voltage characteristics of the VO₂ and Au@VO₂ devices at room temperature under varying light power densities from a laser (808 nm). As the power density increases, the photocurrent increases by more than two orders of magnitude, resembling the temperature-dependent I-V characteristics. This behavior is attributed to the strong IR absorption of VO₂ films, which promotes temperature elevation, thereby inducing MIT in the VO₂ films.

Fig. 2.

(a) I–V characteristics of (a) bare VO2 and (b) Au@VO2 photodetectors at various power densities. (c) Rise and fall times of VO2 and Au@VO2 photodetectors as a function of power densities. (d) Responsivity and detectivity of VO2 and Au@VO2 photodetectors as a function of power densities. (e) Schematic VO2 photo-selector model for the design of an optical convolution engine and (f) convolutional neural network. Reprinted with permission from Ref. [12] Copyright (2021) John Wiley and Sons.

Notably, the photocurrent of the Au@VO₂ sample was saturates at 8.59 W/cm² illumination. The resistance of the VO₂ and Au@VO₂ samples at 8.59 W/cm² is comparable to their resistance under dark conditions at 70°C. Further, it is observed that bare VO₂ saturates at a higher power density of 12.96 W/cm² (Fig. 2 (c)). For the bare VO₂ device, it took 446 ms to reach a stable ON state and 662 ms to switch to the OFF state. In contrast, the Au@VO₂ device required only 138 ms to reach the ON state and 156 ms to switch the OFF at 8.59 W/cm². This is attributed to n-type carrier injection by the Au nanoislands owing to LSPR-induced electron injection during the overheating process, which enhances not only the response speed but also the responsivity and detectivity of the photodetector (Fig. 2 (d)). The bare VO₂ device achieves a maximum responsivity of 261.1 mA/W and a detectivity of 1.14 × 10¹¹ Jones at 12.96 W/cm², whereas the Au@VO₂ device reaches a maximum responsivity of 502.1 mA/W and a detectivity of 1.83 × 10¹¹ Jones at 8.59 W/cm². Both devices exhibited their highest responsivity and detectivity immediately following the phase transition. However, as the power density increased further, the increase in photocurrent became negligible, and both responsivity and detectivity decreased. This indicates that the IR response is dependent on the phase transition and suggests that electron injection due to LSPR from the Au nanoislands enhances the IR response.

To explore the potential of a neural network selector, Zhou et al. [12] designed an optical convolution engine based on a VO₂ device with exceptional IR response. When the cross point was irradiated with a 4 W/cm 2 laser (808 nm), it can directly switch the VO₂ selector to its low-resistance state (LRS). As shown in Fig. 2 (e), although the top electrode (TE) caused a shading effect on the underlying VO₂ film, the light absorption in the adjacent regions of the VO₂ film, combined with Joule heating, generated sufficient thermal effects to switch the VO₂ films at the electrode edges to the LRS. The significant difference between the LRS and high-resistance state (HRS), along with their integrated two-terminal structure, forms the basis for implementing an optical convolution engine. In this convolutional neural network, the current values of the VO₂ device in the LRS and HRS can be used to generate binary outputs (“1” and “0”), as illustrated in Fig. 2 (e) and (f), demonstrating the computational process.

3.2 Gas sensor

While the interest in hydrogen as a renewable energy source has increased, reliable hydrogen detection has become important because of its flammability and explosiveness. It is well known that VO₂ detects hydrogen well and that hydrogen adsorption can trigger the MIT in VO₂ [23,24]. Therefore, several studies have focused on VO₂-based high-performance gas sensors utilizing MIT [13,14,25].

Strelcov et al. [13] introduced a gas sensor based on the MIT of VO₂ nanowires. The sharp phase transition characteristics of VO₂ render it highly sensitive to small changes in the surrounding gas environment. To facilitate operation at room temperature, they induced phase transitions in the VO₂ NWs through Joule heating by applying a current directly to the NWs. This approach eliminated the need for external heating devices because the current itself regulated the temperature within the NWs, resulting in minimal heat loss and enhanced sensor sensitivity. Moreover, the simplicity of the device design (Fig. 3 (a)) reduced response times and improved efficiency. Fig. 3 (b) illustrates the dependence of the sensor signal on the gas concentration.

Fig. 3.

(a) Schematic of the suspended VO2 nanowire (NW) gas sensor. The voltage response of self-heated VO2 NWs to gas pulses as a function of pressure for four different gases. (c) Transition power versus gas pressure graph for self-heated VO2 nanowires. Reprinted with permission from Ref. [13] Copyright (2009) American Chemical Society. (d) SEM image of Pd-decorated VO2 NWs. (e) Current changes upon exposing Pd-decorated VO2 NWs to hydrogen. (f) Recovery rate of Pd-decorated VO2 NWs. Reprinted with permission from Ref. [14] Copyright (2009) American Chemical Society.

The sensor response, ΔV, follows a square-root dependence with the increase in concentration. In contrast, the transition power, PMIT, exhibits a linear dependence on gas pressure (Fig. 3 (c)). The operation of this mesothermistor is related to the suspended state of the NWs, as shown in Fig. 3 (a). A simple heat balance evaluation assumes that the temperature of the self-heated suspended NWs decreases linearly from the highest temperature T at the center to the electrical lead temperature T₀ (T₀: contact temperature). The Joule power generated in the NWs is dissipated through heat transfer to the ambient gas, heat conduction to the metal contacts, and negligible radiative losses. The suspended VO₂ NW sensor operates over a wide pressure range and exhibits excellent sensitivity to both inert and reactive gases. Thus, a novel type of nanoscale gas sensor was proposed, utilizing the pressuredependent MIT in single-crystal suspended VO₂ NWs as the sensor signal. The sensor operated over a wide pressure range and offered excellent sensitivity to both inert and reactive gases. The performance of this sensor could be further enhanced by reducing the diameter of the NWs, thereby increasing their length.

Baik et al. [14] proposed a Pd-decorated VO₂ nanowire device capable of highly sensitive hydrogen detection based on the MIT utilizing Pd nanoparticles (NPs) (Fig. 3 (d)). Pd NPs are considered promising as efficient catalysts for reactions with hydrogen because of their exceptional abilities to absorb and release hydrogen. A Pd-decorated VO₂ nanowire device demonstrated high sensitivity specifically to hydrogen, suggesting its potential for selective hydrogen sensing applications.

As shown in Fig. 3 (e), when hydrogen was introduced into the Pd-decorated VO₂ nanowire biased at 10 V, the current gradually increased. Depending on the H₂ exposure time, at ambient temperatures within the range of 45–55°C, the current increased by approximately 1000-fold, indicating a transition from the insulating to the metallic state, occurring approximately 10 minutes after hydrogen exposure. At 60°C, this transition occurred within approximately 5 min. In contrast, when a hydrogen flow was introduced to the bare VO₂ NWs, the change in current was less than 10%. This significant difference was attributed to the catalytic dissociation of diatomic hydrogen on the Pd NPs, which generated a substantial number of hydrogen atoms on the nanowire surface. These hydrogen atoms then “spilled over” and diffuse into the vanadium oxide, triggering MIT in the VO₂ nanowires. Following the termination of hydrogen exposure after 60 min, the recovery of the nanowire conductance to its initial value was exceedingly slow. Even after 24 h in a vacuum environment (<10^-5 Torr, room temperature), minimal change in the conductance value was observed. When air was introduced into the chamber, the device current began to decrease immediately, likely owing to the catalytic oxidation of hydrogen by Pd, resulting in the formation of water. However, the return of the current to its initial value required more than 4 h at 50°C, 30 min at 80°C, and 5 min at 100°C (Fig. 3 (f)).

The electrical conductance of VO₂ NWs changes over two distinct timescales. The first is a rapid process occurring within 1–5 min, depending on the ambient temperature, and the second is a slower process, wherein the conductance reaches a steady state approximately 18 h after hydrogen exposure under a 1 V bias. However, when a Pd coverage corresponding to a mass thickness of 1 nm was applied, the reaction time was reduced from minutes to subseconds. This acceleration was attributed to the increased number of atomic hydrogen sources on the nanowire surface, which increased the rate of atomic H incorporation into the VO₂ surface. The incorporation of atomic H into VO₂ altered the electron-phonon interactions or contributes electrons, thereby increasing the electron density of VO₂. These changes reduced the MIT temperature of VO₂ by approximately 10°C, which is a crucial mechanism that significantly enhances the sensitivity of the hydrogen detection sensor. Researchers have proposed the possibility of electron-phonon interactions and changes in the electronic structure owing to hydrogen doping, stemming from the interaction between Pd NPs and VO₂ NWs.

Li et al. [25] reported a novel hydrogen sensor based on a Pd/a-WO₃/VO₂ structure. Fig. 4 (a) shows a schematic of the a-WO₃/VO₂ hydrogen-sensor device. Following the deposition of the VO₂ epitaxial layer on an aluminum oxide substrate using molecular beam epitaxy (MBE) technology, amorphous WO₃ and ultrathin Pd layers were grown using magnetron sputtering and e-beam evaporation, respectively.

Fig. 4.

(a) Schematic of a hydrogen sensor with a heterostructure of Pd decorated a-WO3/VO2 thin film. (b) Sensitivity of Pd decorated a-WO3/VO2 heterostructure as a function of temperature. (c) Temperature dependent response and recovery time of a Pd decorated a-WO3/VO2 gas sensor. Schematic for the operating mechanism of the Pd decorated a-WO3/VO2 heterostructure in (d) pristine state, (e) response state, and (f) recovery state. Reprinted with permission from Ref. [25] Copyright (2022) John Wiley and Sons.

To investigate the effect of the VO₂ phase transition on the performance of the sensor, the sensitivity was measured following its exposure to hydrogen gas (5 ppm) at 30°C and 80°C. The sensitivities in the insulating state (30°C) and the metallic state (80°C) were 258% and 18,000%, respectively, indicating a sharp increase in sensitivity following the phase transition (Fig. 4 (b)). In addition, the response and recovery times were significantly reduced to 680 and 30 s at 30°C, and 1910 and 45 s at 80°C, respectively (Fig. 4 (c)). These results suggest that the phase transition properties of VO₂ have a substantial impact on the performance of the sensor.

The entire sensing process of the Pd/a-WO₃/VO₂ sensor involves the following steps (Fig. 4 (d–f)):

In the pristine state (Fig. 4 (d)), oxygen adsorbed on the surface is reduced to O₂^- ions owing to the catalytic effect of Pd NPs and interfacial electron transfer within the a-WO₃/VO₂ layer. When the sensor is exposed to H₂ gas, in the response state (Fig. 4 (e)), Pd NPs catalytically dissociate hydrogen into H⁺ ions, which are then drawn into the WO₃ layer via an electron–proton co-doping process at the interface. This results in the metallization of the H-doped WO₃ layer. This process decreases the resistance of the a-WO₃ layer, and the sensing conductance signal is synchronously detected. In the third stage (Fig. 4 (f)), synthetic air is introduced to restore the device to its initial state. Here, the Pd NPs catalyze the dissociation of oxygen molecules into O^2- ions, which react with H atoms released from the a-WO₃ layer to form H₂O, subsequently evaporating from the surface owing to external heating and the intrinsic exothermic reaction. The excess electrons generated in the reaction are transferred back to the Pd NPs, where they reform O₂^- ions. During this process, the phase transition from VO₂ (M) to VO₂ (R) facilitates the transfer of more number of electrons from VO₂ to the WO₃ layer, facilitating the conductive H_xWO₃ formation and promoting dehydrogenation during the recovery process. The authors demonstrated significant performance enhancements, including detection limits, response time, and sensitivity, through interfacial charge transfer, depending on the electronic phase of VO₂. These results provide valuable insight into the design of high-performance gas sensors utilizing the MIT of VO₂.

3.3 Strain sensor

VO₂ is an exceptionally advantageous material for mimicking human body functions or detecting dynamic changes owing to its high sensitivity, rapid response, and metal-insulator transition (MIT) characteristics [26]. In addition, repeated strain can induce phase transitions between the M1 and M2 phases within VO₂ [15]. These properties render VO₂ a promising candidate for the fabrication of high-performance strain sensors.

Hu et al. [15] introduced a flexible strain sensor that induced MIT through an external strain in a VO₂ nanobeam structure. As shown in Fig. 5 (a), tensile and compressive strains were applied incrementally in steps of 0.05%, and the corresponding changes in the I–V characteristics were observed. In contrast to piezotronic-based strain sensors, the I–V curve of the VO₂ nanobeam device exhibited downward and upward shifts under tension and compression, respectively. This distinct response to different types of strain was attributed to the M1-M2 phase transition within the VO₂ nanobeam.

Fig. 5.

(a) I–V characteristics of VO2 nanobeam under different tensile and compressive strains. (b) Resistance hysteresis of the VO2 nanobeam under external strain and strain release. (c) Resistance changes of VO2 nanobeam owing to strain in (b). (d) Gauge factor as a function of strain derived from (b). (e) Response of VO2 nanobeam to strain switching. Reprinted with permission from Ref. [15] Copyright (2010) John Wiley and Sons.

As shown in Fig. 5 (b), the resistance changes owing to the strain increased under tension and the decreased under compression. To verify the stability of the device, the original state, tension, release, compression, and release were sequentially scanned thrice. The VO₂ nanobeam exhibited a stable and reproducible performance under repeated tensile and compressive strains. In addition, the variation in the resistance with strain was measured to confirm the operating principle of the device (Fig. 5 (c)). As the compression strain increased, the resistance changes decreased, which was attributed to the stepwise reduction of the M2 phase induced by the pre-applied tensile strain in the localized regions of the VO₂ nanobeam. As the tensile strain increased, a sufficient amount of the original M1 phase transformed into the M2 phase, which resulted in a relatively stable resistance change.

Most semiconductor-based strain mechanisms involve changes in the band structure of the semiconductor material under strain, typically leading to linear resistance changes. However, in contrast to conventional devices, a VO₂ nanobeam device does not exhibit a hysteresis effect (Fig. 5 (b)) and does not perfectly match the linear relationship with the applied stress, as shown in Fig. 5 (c). Therefore, the dominant cause of the resistance change in the VO₂ nanobeam device under strain is likely the phase transition rather than the piezoresistance effect.

The performance of the strain sensor is characterized by its gauge factor, with the VO₂ nanobeam sensor exhibiting a maximum gauge factor of 347 (Fig. 5 (d)). In contrast to other types of sensors, the VO₂ nanobeam sensor achieves its maximum gauge factor at the lowest applied stress. Although the gauge factor decreases with increasing stress, it remains higher than that of conventional metal strain gauges and doped Si strain sensors. In addition, as demonstrated in Fig. 5 (e), the response time under a 1 V bias shows a rapid response to strain switching. Because there is no significant change in the current at a 1 V bias across the original, compression, and tensile states of the VO₂ nanobeam, the Joule self-heating effect can be excluded. The relatively balanced performance in both the compression and tensile states suggests that the VO₂ nanobeam strain sensor is well-suited for applications in low mechanical strain ranges.