Ultra-compact and high-performance suspended aluminum scandium nitride Lamb wave humidity sensor with a graphene oxide layer

Abstract

The utilization of Microelectromechanical Systems (MEMS) technology holds great significance for developing compact and high-performance humidity sensors in human healthcare, and the Internet of Things. However, several drawbacks of the current MEMS humidity sensors limit their applications, including their long response time, low sensitivity, relatively large sensing area, and incompatibility with a complementary metal–oxide-semiconductor (CMOS) process. To address these problems, a suspended aluminum scandium nitride (AlScN) Lamb wave humidity sensor utilizing a graphene oxide (GO) layer is firstly designed and fabricated. The theoretical and experimental results both show that the AlScN Lamb wave humidity sensor exhibits high sensing performance. The mass loading sensitivity of the sensor is one order higher than that of the normal surface acoustic wave (SAW) humidity sensor based on an aluminum nitride (AlN) film; thus the AlScN Lamb wave humidity sensor achieves high sensitivity (∼41.2 ppm per % RH) with only an 80 nm-thick GO film. In particular, the as-prepared suspended AlScN Lamb wave sensors are able to respond to the wide relative humidity (0–80% RH) change in 2 s, and the device size is ultra-compact (260 μm × 72 μm). Moreover, the sensor has an excellent linear response in the 0–80% RH range, great repeatability and long-term stability. Therefore, this work brings opportunities for the development of ultra-compact and high-performance humidity sensors.

---

Introduction

In the past few years, humidity detection has attracted great attention due to its widespread applications across various fields, including industrial production, agricultural monitoring, Internet of Things, and human healthcare.^1–5 Consequently, a variety of humidity sensors have been explored, such as resistance-type,^6,7 capacitance-type,^8–10 impedance-type,^11–13 voltage-type,^14,15 and surface acoustic wave (SAW)-type humidity sensors.^16–20 With the advent of self-powered technology, voltage-type humidity sensors based on gradient power generation and resistance-type humidity sensors based on piezoelectric devices and triboelectric nanogenerators (TENGs) have been proposed in recent years.^21–23 Generally, a promising candidate for humidity sensors requires high sensitivity, and fast response in a wide relative humidity (RH) range. The evolution of wireless devices has heightened the demand for next-generation humidity sensors. Microelectromechanical Systems (MEMS) technology based solutions are getting more and more attention due to their high-performance, low-cost, and compact-size.^24–28 In recent years, piezoelectric surface acoustic wave (SAW) technology with a humidity sensing layer has gained widespread adoption, owing to its simplified fabrication process.^29–31 However, the ongoing miniaturization trend in wireless devices necessitates the development of smaller sensors. Traditional SAW-based humidity sensors face challenges in this regard, as they rely on extensive interdigitated transducers (IDTs) and large reflector arrays.³² Moreover, the solid structure of SAW-based sensors mandates thicker sensing layers to achieve satisfactory detection sensitivity. Bulk lithium niobate (LiNbO₃)^18,29,33,34 and thin-film aluminum nitride (AlN)^16,30 are among the most commonly used piezoelectric materials in SAW-based humidity sensors. Thanks to the simple fabrication process, LiNbO₃-based devices have the advantage of low-cost. However, limitations arise from their high stiffness coefficients and incompatibility with complementary metal–oxide-semiconductor (CMOS) technology, posing challenges for achieving adequate detection sensitivity and seamless system integration. Conversely, AlN-based humidity sensors demonstrate excellent compatibility with the CMOS process and integration capabilities, although there is scope for further enhancement of sensing performance.³⁵

Lamb wave resonator (LWR), which utilizes Lamb waves propagating along the suspended piezoelectric thin film, has become a popular tool.^36–38 While LWR-based acoustic filters have been extensively studied in radio frequency (RF) communication applications,³⁹ their potential for humidity sensing remains largely untapped.⁴⁰ Unlike SAW resonators, LWRs provide smaller dimensions and higher phase velocities.⁴¹ Crucially, the suspended structure of LWRs capitalizes on lower stiffness coefficients, enhancing sensitivity to mass loading changes. As such, LWR-based sensors hold promise for next-generation humidity sensing applications. In addition to structural considerations, the choice of piezoelectric thin films plays an important role in achieving high sensitivity to mass loading. A commonly used piezoelectric material in such suspended structures is an AlN thin film.^41,42 Recently, aluminum scandium nitride (AlScN) has been widely investigated by researchers.^43–45 Scandium doping results in not only higher piezoelectric coefficients, but also lower stiffness coefficients.^46,47 The sensing ability of the humidity sensors also rely on the selection of sensing materials. Carbon nanomaterials,^48,49 polymers,⁵⁰ and semiconductor metal oxides⁵¹ are frequently used. Recently, graphene oxide (GO) has emerged as a promising candidate for humidity sensing due to its large surface-to-volume ratio and high hydrophilicity.⁵²

Herein, we demonstrate an AlScN-based Lamb wave humidity sensor with a GO film. Leveraging a suspended structure composed of an AlScN thin film and a GO sensing layer, this humidity sensor demonstrates enhanced sensitivity to mass loading changes. The lower stiffness constants of the AlScN thin film further contribute to its sensitivity. Additionally, half-circle-shaped reflectors are integrated into the sensor structure to enhance the quality factor (Q factor). These design features yield ultra-compact sensors (260 μm × 72 μm) with high sensitivity (∼41.2 ppm per % RH with an 80 nm GO film), excellent linearity across the 0–80% RH range, great long-term stability and repeatability, and rapid response (∼2 s). The proposed AlScN Lamb wave humidity sensor represents a promising avenue for future technological advancements in consumer electronics and healthcare instruments.

Results and discussion

Design of the AlScN Lamb wave humidity sensor

Here, an acoustic fundamental symmetric (S0) mode Lamb wave is excited and propagated in the suspended 0002-polar Al_0.84Sc_0.16N thin film. In our design, the LWR with 3-pairs of IDTs is supported by a silicon substrate with an air cavity enabling great acoustic energy confinement. Then, the GO layer is deposited on the LWR surface for humidity detection. Fig. 1(a) shows a schematic of the proposed LWR-based humidity sensors. The dimensional parameters of the LWRs are defined in Fig. 1(b), where L is the length of the LWRs, W defines the width of LWRs, Λ depicts the pitch of LWRs, and t_AlScN represents the thickness of the AlScN thin film. Fig. 1(c) shows the frequency tuning capability of LWRs, since working frequency is determined by the pitch of resonators. In our work, an LWR with a 24 μm pitch is utilized for humidity sensing, working at around 300 MHz.

Fig. 1 (a) Schematic of the Lamb wave humidity sensor. (b) Geometric parameters of the LWR with length L, width W, pitch and thickness t_AlScN. (c) Simulated admittance spectra of AlScN LWRs with varying pitch. (d) Simulated admittance response and Q factor of AlScN LWRs with/without half-circle-shaped reflectors; insets are the simulated displacement fields of AlScN LWRs with a perfectly matched layer (PML). Simulated relative frequency shift as a function of mass loading for (e) the Lamb wave humidity sensor with an S0 mode shape and (f) the SAW humidity sensor with a Rayleigh mode shape.

The Q factor is one of the essential parameters for Lamb wave humidity sensors. The energy dissipated to the substrate through the anchor (anchor loss), the energy loss from the interface between the piezoelectric layer and electrodes (interface loss), and other loss sources (other loss) are the main energy dissipation mechanisms. Researchers often assume that the anchor loss mainly contributes to the Q factor of LWRs. To this end, our design focuses on decreasing the displacement fields in the substrate to reduce the anchor loss. In general, the acoustic wave (energy) in the vibration region easily travels to the substrate through the anchor, resulting in large anchor loss. With the help of the designed half-circle-shaped reflectors, the main acoustic wave reflects back to the vibration region. To save time and computing resources, only half of the structure is built with symmetry planes. The simulating structure consists of the AlScN LWR with a perfectly matched layer (PML). PML-based 3D finite element analysis (FEA) using COMSOL® Multiphysics is utilized to calculate the admittance responses and displacement fields of LWRs. As shown in Fig. 1(d), the Q_r of AlScN LWRs with half-circle-shaped reflectors is twice that of normal AlScN LWRs, and is defined as,

where f_r is the resonant frequency, Δf_{3 dB} is the frequency difference between left and right 3 dB points. The insets also show large displacement fields in the substrate, which indicate massive energy leaks through the anchor. In contrast, more energy is confined in the vibration region for the LWR with reflectors.

The feature of humidity sensing is numerically simulated utilizing 2D FEA with COMSOL® Multiphysics. For the resonant-type humidity sensor, the mass loading effect of the devices leads to a frequency shift, which increases as a function of humidity. To this end, mass loading is added to the structure in the simulation. As shown in Fig. 1(e) and (f), the characteristics of the frequency shift on both Lamb wave-based and SAW-based sensors are measured, and the insets show the mode shapes for the S0 mode Lamb wave and Rayleigh mode (one type of SAW). For the AlScN Lamb wave-based sensor, the mass loading is set from 0 to 0.07 g m⁻², and then the relative frequency shift is changed from 0 to 0.74%. Compared with the traditional AlN SAW-based sensor, it shows about 10 times higher sensitivity of mass loading contributed by the smaller stiffness constant of Sc-doping and smaller coefficients of mass sensitivity by the suspended structure. Additionally, simulation results show a linear relationship between the frequency shift and mass loading, which agrees with the previously proposed equation:⁵³

where f₀ is the fundamental resonant frequency, f is the resonant frequency under certain mass loading, c_m is the coefficient of mass sensitivity of materials, and is the change of mass in per unit area.

Fabrication of AlScN Lamb wave humidity sensors

Fig. 2(a)–(i) illustrate the fabrication process flow for the AlScN Lamb wave humidity sensor. The 10 nm/100 nm Ti/Pt layers are first deposited using electron beam evaporation and lift-off as bottom electrodes on a 4-inch silicon (100) wafer (Fig. 2(a)). Notably, since high-quality deposition of the AlScN thin film requires a relatively smooth and flat substrate, the AZ® 5214 reversal process is utilized for negative sidewalls. Then, a 500 nm AlScN film is deposited using a DC-pulsed magnetron co-sputtering system with 4-inch Al and Sc targets with an EVATEC CLUSTERLINE® 200 MSQ multi-source system at 350 °C (Fig. 2(b)). Different concentrations of Al/Sc can be obtained by varying the power ratio between Al and Sc targets. In this work, 1000 W/300 W power on Al/Sc targets is set for the Al_0.84Sc_0.16N film. Additionally, the film quality of the sputtered AlScN film is characterized. The scanning electron microscopy (SEM) images of the AlScN film at the center, middle, and also edge are captured (Fig. 2(l)). Abnormal orientation grains (AOGs) can be found at the middle and edge of the wafer. The growth of AOGs is contributed by the intensity of the ion bombardment. Thus, AOGs are more likely distributed near the boundary of the wafer. The existence of AOGs deteriorates the crystalline quality, since the AlScN film with AOGs generally has a lower piezoelectric constant. The full width at half maximum (FWHM) value in the X-ray diffraction (XRD) rocking curve is another essential parameter to determine the AlScN film quality. The FWHM achieves 1.5°/1.7°/1.8° at the wafer center/middle/edge, indicating better crystalline quality at the center (Fig. 2(m)). Fig. 2(n) shows the atomic force microscopy (AFM) amplitude image of the AlScN film, which illustrates a smooth film surface. Next, as shown in Fig. 2(c), a 2 μm SiO₂ hard mask is deposited using plasma enhanced chemical vapor deposition (PECVD) and patterned using reactive ion etching (RIE). According to our previous works, an optimized Cl-based inductively coupled plasma (ICP) etching recipe is required for AlScN film vertical etching.⁵⁴ Thus, the Al_0.84Sc_0.16N film is fully etched using a recipe of Cl₂/BCl₃/N₂ = 25/30/20 sccm, 350 W RF power (Fig. 2(d)). Fig. 2(j) shows a tilted SEM view of the etching result, which indicates a clean and smooth etching surface with great uniformity. According to Fig. 2(k), a vertical etching profile (∼77°) is achieved, which results in better efficiency of acoustic reflection at the boundary. After that, the remaining hard mask is removed through RIE, the same as the process in hard mask patterning (Fig. 2(e)). Then, a 200 nm Al layer is deposited by magnetron sputtering and patterned (Fig. 2(f)). To achieve a free-standing structure only connected by anchors, the silicon layer beneath is removed through rapid dry etching with XeF₂ gas (Fig. 2(g)). The SEM image of the AlScN Lamb wave humidity sensor before GO drop-casting is shown in Fig. 2(o). The energy dispersive spectroscopy (EDS) results in Fig. 2(p) verify the presence of Al and Sc in the area of Fig. 2(o), while the distribution of both elements is uniform. The admittance response and fitting of the AlScN Lamb wave humidity sensor without a GO layer are measured, as presented in Fig. 2(q). The modified Butterworth–Van Dyke (mBVD) model is utilized to extract the equivalent electric parameters,⁵⁵ which are listed in the figure. R_m, L_m, and C_m are the motional resistance, inductance, and capacitance, respectively; R_s is the loaded resistance in series; R₀ and C₀ are the static resistance and capacitance, respectively. The sensor works at 299.6 MHz, achieving a k_t² of 4.2% and a Q_r of 1248.3.where k_t² is the electromechanical coupling coefficient, f_r is the resonant frequency, and f_a is the anti-resonant frequency.

Fig. 2 Schematic illustration of the AlScN Lamb wave humidity sensor. (a) Bottom electrode patterning and lift-off. (b) AlScN thin film depositing by magnetron co-sputtering. (c) SiO₂ hard mask depositing and patterning. (d) AlScN thin film patterning by Cl-based ICP etching. (e) SiO₂ hard mask stripping by RIE etching. (f) Top electrode patterning. (g) Structure releasing by XeF₂ dry etching. (h) GO solution drop-casting on the sensor surface. (i) GO layer drying at room temperature. SEM pictures of AlScN etching results, (j) tilted view, and (k) cross-section view. (l) SEM top view of the (0002) polar oriented AlScN surface with different locations on the wafer. (m) XRD rocking curve of the AlScN film. (n) AFM amplitude image of the AlScN film. (o) SEM image of the AlScN humidity sensor without a GO layer. (p) EDS mapping analysis of the AlScN humidity sensor surface. (q) Measured admittance response and MBVD fitting of the AlScN humidity sensor before GO drop-casting.

To prepare the GO sensing layer, a GO solution with an original concentration of 4 mg mL⁻¹ is mixed with deionized (DI) water with a volume ratio of 1 : 20, and stripped by ultrasonic treatment for 10 min. Then, 2.5 μL of the prepared GO solution is drop-cast onto the surface of the Lamb wave sensor as a humidity sensing layer utilizing a micro-syringe (Fig. 2(h)). After the sensing layer is drop-cast, the sensor is maintained at room temperature until it is dried (Fig. 2(i)).

The GO layer is first measured by AFM in the tapping mode using a Park NX20. The amplitude images of four positions on the GO film are shown in Fig. 3(a). The captured amplitude images show great uniformity of the GO film. The distribution of C and O elements in the GO layer drop-cast on the AlScN film can be observed from the EDS map (Fig. 3(b)). The oxygen-containing molecules are widely distributed on the GO layer. Thus, a great hydrophilic property is expected for the GO layer. Then, the thickness of the GO film is measured using a surface profiler, and the average thickness of the GO film is approximately 80 nm (Fig. 3(c)). In addition, the inset picture also indicates the great smoothness of the GO film on AlScN. Fourier transform infrared spectroscopy (FTIR) is also performed to characterize the functional groups of the GO layer (Fig. 3(d)). Three main peaks at 3370, 1725, and 1625 cm⁻¹ are observed in the spectrum. The broad peak at 3370 cm⁻¹ with high intensity is correlated with the vibration of the O–H molecular bond, while the peaks at 1725 and 1625 cm⁻¹ correspond to the vibration of the C=O bond and skeletal ring.⁵⁶ The inset shows the comparisons of the intensity peaks at 3370 cm⁻¹ of the sensor under 0 and 65% RH conditions. The larger peak for the sensor at 65% RH illustrates the larger adsorption of H₂O molecules. The SEM image of the AlScN humidity sensor covered by the GO layer shows a clear and smooth structure, which can be seen in Fig. 3(e). The optical microscopy image showing the top view of the fabricated AlScN Lamb wave humidity sensor is presented in Fig. 3(f). The surrounding of the device is also suspended due to the isotropic etching of XeF₂ gas. The half circle shaped reflectors at the end of the anchors are defined for acoustic energy confinement.

Fig. 3 Characterization of the AlScN Lamb wave humidity sensor. (a) AFM amplitude images of the GO layer at different positions. (b) EDS mapping analysis of the GO layer on the AlScN film. (c) Thickness profile of the GO film; the inset shows the optical image of the GO film on AlScN. (d) FTIR spectrum of GO; the inset shows the comparison of the peak intensity at around 3370 cm⁻¹ at 0% and 65% RH levels. (e) SEM image of the AlScN Lamb wave humidity sensor. (f) Optical image of the AlScN Lamb wave humidity sensor.

Performance characterization of the AlScN Lamb wave humidity sensor

The device is tested to characterize the sensing performance. The sensing performance characterization setup is shown in Fig. 4(a) and (b). During the test, the device is fixed in a metallic chamber. Test data are recorded and processed using a network analyzer with the MATLAB program. By varying the flow rate of dry and wet N₂ gas, the humidity inside the chamber can be adjusted from 0 to 80% RH. The real-time humidity and temperature in the chamber are calibrated using a hygrothermometer connected to the outlet of the chamber.

Fig. 4 Static performance characterization of the AlScN Lamb wave humidity sensor. (a) Schematic diagram of the test setup. (b) Picture of the test setup in the lab. (c) S11 response of the sensor as a function of RH. (d) Stability test of sensors. (e) Frequency shift of the humidity sensor with/without the GO layer versus RH curve. (f) Working frequency of the sensor versus RH curve. (g) S11 amplitude of the sensor versus RH curve. (h) S11 amplitude of the sensor at 298 MHz.

We investigated the S11 responses of the fabricated AlScN Lamb wave sensor from 0 to 80% RH increments of 10% RH, as illustrated in Fig. 4(c). As simulated in the previous section, due to the mass loading effect, the resonant frequency of the sensor decreases with increasing humidity levels. At the same time, the damping loss also increases with larger loading mass. Thus, the S11 amplitude decreases with higher RH. Fig. 4(d) shows the frequency responses of the sensor for about ten days, indicating that the sensor has good long-term stability. Fig. 4(e) depicts the frequency shift feature of sensors with and without a GO layer. The ability of humidity sensing is totally contributed by the GO layer. The sensitivity of the sensor is generally quantified as,

where ΔRH is the change in relative humidity. Fig. 4(f) illustrates the resonant frequency and fitting curve in the 0–80% RH range. The sensor achieves a linear response between the frequency and RH, and the sensitivity is about 41.2 ppm per % RH with a slope of −0.01245 MHz per % RH. Among the most previously reported SAW-based sensors with a GO sensing layer, due to the low mass sensitivity of the structure, an over 100 nm-thick GO layer is utilized for higher sensitivity. However, it is more likely to observe the nonlinearity behavior in such sensors, because the water molecules usually penetrate into the interlayer of the GO film at a high RH level (>70% RH). According to these findings, our fabricated sensor uses only an 80 nm-thick GO layer for trade-off between sensitivity and linearity. Then, the relationship between the S11 amplitude and RH of the sensor is also demonstrated in Fig. 4(g). More data can be found in Fig. S1 and S2 (ESI†). The sensor also yields a linear response between amplitude and RH, with a slope of 0.01254 dB per % RH. Additionally, the linearity of the sensor can also be found for the S11 amplitude shift at a certain frequency. As shown in Fig. 4(h), the linear response of the S11 amplitude at 298 MHz and RH is demonstrated.

The humidity hysteresis of the humidity sensor is defined as the maximum difference of measured signals during the adsorption and desorption processes.^57,58 Thus, the humidity hysteresis characteristic of the humidity sensor is tested by increasing the RH level from 0 to 80%, and then decreasing from 80% to 0. As shown in Fig. 5(a), the humidity hysteresis of the AlScN Lamb wave humidity sensor is about 15% RH, which is relatively large. The repeatability characteristic of the sensor is measured by repeating the adsorption and desorption processes (RH varying between 0 and 80%). The excellent sensing repeatability of the sensor is demonstrated in four cycling tests, as shown in Fig. 5(b). Next, the response and recovery speed of the sensor are investigated, when the RH rapidly changes from 0 to 80%. As illustrated in Fig. 5(c) and (d), the response time of the sensor is about 2 s, while the recovery time of the sensor is about 20 s. Fig. 5(e) and (f) show the mechanism of GO used for humidity sensing. For the GO layer, water molecules are easily absorbed at the active sites (–COOH, –OH, and –CO) through hydrogen bonds, which causes the mass loading on humidity sensors. When the RH level is high, most of the active sites are occupied by H₂O. Then, with the further increase of the RH level, the second H₂O layer is adsorbed by hydrogen bonds, and H₂O molecules even go into the interlayer of GO, and as a result, causes larger mass loading. In contrast, during the desorption process, the water molecules detach from the GO layer by breaking hydrogen bonds, which decreases the mass loading. The ultra-short response time is benefited by the thin GO film for rapid adsorption of H₂O, which is much faster than those of previously reported SAW-based humidity sensors, as listed in Table 1. In addition, as the first proposed AlScN Lamb wave humidity sensor, compared with other SAW humidity sensors, our sensor has higher sensitivity with a thin sensing layer, and much better linearity. Additionally, the applications of the AlScN Lamb wave humidity sensors for human respiratory monitoring are conducted and displayed in Fig. S3 (ESI†). The better detection of respiratory behavior will be a focus in our future work.

Fig. 5 Dynamic performance characterization of the AlScN Lamb wave humidity sensor. (a) Hysteresis characteristic of the humidity sensor. (b) Repeatability test of the humidity sensor as the RH is repeatedly varied between 0 and 80% RH. (c) The response and (d) recovery processes of the sensor. Sensing mechanisms of (e) the adsorption process and (f) desorption process.

Table 1 Comparison of the characteristics of acoustic-type humidity sensors

Ref.	Sensing type	Piezoelectric material	Sensing material	Thickness of sensing material (nm)	Sensitivity (ppm per % RH)	Response and recovery time (s)
This work	Lamb wave	AlScN (500 nm)	GO	∼80	∼41.2	2, 20 (0–80% RH)
32	Lamb wave	AlN (500 nm)	GO	∼160	∼16.6	9, 11 (20–80% RH)
10	SAW	AlN (1000 nm)	GO	∼90/∼210	∼33.2/∼111.7	9, 6/10, 9 (15–80% RH)
24	SAW	ZnO	GO	∼100/∼400	∼25.7/∼89.4	22, 5/NA (10–80% RH)
13	SAW	Quartz (bulk)	GO	∼70	∼2.5	NA
20	SAW	Quartz (bulk)	Polymer	NA	∼0.9	10, 10 (10–97% RH)
12	SAW	LiNbO3 (bulk)	MoS2/GO	NA	∼331.6	6.6, 3.5 (10–95% RH)
33	SAW	LiNbO3 (bulk)	GO	∼70	∼10	80, 72 (0–97% RH)
29	SAW	LiNbO3 (bulk)	Polyaniline	NA	∼16.8	NA
34	SAW	LiNbO3 (bulk)	CeO2/PVP	NA	∼19.1	16, 16 (11–95% RH)

---

Conclusions

In summary, we have proposed a novel AlScN Lamb wave humidity sensor, which utilizes a suspended structure and AlScN thin film to improve the sensing performance. The theoretical and experimental results show that the AlScN Lamb wave humidity sensor has varieties of advantages: (1) the mass loading sensitivity is one order higher than that of the traditional AlN SAW-based humidity sensor; (2) the sensor achieves an ultra-compact device size (260 μm × 72 μm); (3) the proposed sensor exhibits high sensing sensitivity (∼41.2 ppm per % RH) in a wide range (0–80% RH) with a thin sensing layer (80 nm); (4) the response time is only 2 s; and (5) moreover, the AlScN Lamb wave humidity sensor has excellent linearity, and great long-term stability and repeatability. In view of the many advantages of the AlScN Lamb wave humidity sensors, this study is expected to provide useful guides for the development of ultra-compact and high-performance humidity sensors.

Materials and methods

Materials

The N₂ gas used in the experiments was stored in a steel cylinder purchased from Singapore. The graphene oxide water dispersion (0.4 wt% concentration) used in the experiments was purchased from Graphenea Co., Ltd, Singapore. The N95 mask used for the insertion of the sensor was purchased from Winner Medical Co., Ltd, China.

Characterization and test methods

The field emission scanning electron microscope (FE-SEM) Hitachi Regulus 8230 from Hitachi High-Tech Corporation was used to investigate the humidity sensor structure. The humidity sensor was connected to a printed circuit board layer utilizing wire bonding. The frequency signals were collected using a network analyzer (R&S® ZNB20). The relative humidity level in the device chamber was measured using a hygrothermometer (Hygro-Thermometer AZ8721).

Author contributions

Zhifang Luo designed and performed the experiments, and drafted the manuscript. Dongxiao Li, Xianhao Le, and Tianyiyi He contributed to the testing setup. Shuai Shao contributed to fabrication. Qiaoya Lv and Zhaojun Liu contributed to characterization. Chengkuo Lee and Tao Wu supervised the project.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work was supported in part by the Lingang Laboratory under Grant LG-QS-202202-05, in part by the Natural Science Foundation of Shanghai under Grant 23ZR1442400, in part by Jiangsu Provincial Key Research and Development program (BE2023048), and in part by the Double First-Class Initiative Fund of ShanghaiTech University. The authors appreciate the device fabrication support from the ShanghaiTech Quantum Device Laboratory (SQDL).

References

1. Y. Lu, G. Yang, Y. Shen, H. Yang and K. Xu, Nano-Micro Lett., 2022, 14, 150.
2. Y. Liang, Q. Ding, H. Wang, Z. Wu, J. Li, Z. Li, K. Tao, X. Gui and J. Wu, Nano-Micro Lett., 2022, 14, 183.
3. Z. Zhang, T. Lu, D. Yang, S. Lu, R. Cai and W. Tan, Nano Lett., 2021, 21, 9030–9037.
4. A. Beniwal, P. Ganguly, A. K. Aliyana, G. Khandelwal and R. Dahiya, Sens. Actuators, B, 2023, 374, 132731.
5. T. He, F. Wen, Y. Yang, X. Le, W. Liu and C. Lee, Anal. Chem., 2023, 95, 490–514.
6. P. Guo, B. Tian, J. Liang, X. Yang, G. Tang, Q. Li, Q. Liu, K. Zheng, X. Chen and W. Wu, Adv. Mater., 2023, 2304420.
7. D.-T. Phan and G.-S. Chung, Sens. Actuators, B, 2015, 220, 1050–1055.
8. A. Tripathy, S. Pramanik, A. Manna, S. Bhuyan, N. F. Azrin Shah, Z. Radzi and N. A. Abu Osman, Sensors, 2016, 16, 1135.
9. N. Li, X.-D. Chen, X.-P. Chen, X. Ding and X. Zhao, IEEE Electron Device Lett., 2017, 38, 806–809.
10. Z. Duan, J. Li, Z. Yuan, Y. Jiang and H. Tai, Sens. Actuators, B, 2023, 394, 134445.
11. Z. Duan, Q. Zhao, S. Wang, Q. Huang, Z. Yuan, Y. Zhang, Y. Jiang and H. Tai, Sens. Actuators, B, 2020, 317, 128204.
12. J. Dai, X. Guan, H. Zhao, S. Liu, T. Fei and T. Zhang, Sens. Actuators, B, 2020, 316, 128159.
13. H. Zhao, T. Zhang, R. Qi, J. Dai, S. Liu, T. Fei and G. Lu, Sens. Actuators, B, 2018, 255, 2878–2885.
14. L. Li, Z. Chen, M. Hao, S. Wang, F. Sun, Z. Zhao and T. Zhang, Nano Lett., 2019, 19, 5544–5552.
15. D. Shen, M. Xiao, G. Zou, L. Liu, W. W. Duley and Y. N. Zhou, Adv. Mater., 2018, 30, 1705925.
16. X. Le, Y. Liu, L. Peng, J. Pang, Z. Xu, C. Gao and J. Xie, Microsyst. Nanoeng., 2019, 5, 36.
17. Y. Su, C. Li, M. Li, H. Li, S. Xu, L. Qian and B. Yang, Sens. Actuators, B, 2020, 308, 127693.
18. X. Li, Q. Tan, L. Qin, L. Zhang, X. Liang and X. Yan, Sens. Actuators, A, 2022, 341, 113573.
19. S. M. Balashov, O. V. Balachova, A. V. U. Braga, A. P. Filho and S. Moshkalev, Sens. Actuators, B, 2015, 217, 88–91.
20. Y. Li, P. Li, M. Yang, S. Lei, Y. Chen and X. Guo, Sens. Actuators, B, 2010, 145, 516–520.
21. M. Zhang, Z. Duan, B. Zhang, Z. Yuan, Q. Zhao, Y. Jiang and H. Tai, Nano Energy, 2023, 115, 108745.
22. Y. Jiang, Z. Duan, Z. Fan, P. Yao, Z. Yuan, Y. Jiang, Y. Cao and H. Tai, Sens. Actuators, B, 2023, 380, 133396.
23. Q. Zhao, Y. Jiang, L. Yuan, Z. Yuan, B. Zhang, B. Liu, M. Zhang, Q. Huang, Z. Duan and H. Tai, Sens. Actuators, B, 2023, 392, 134082.
24. X. Le, Q. Shi, P. Vachon, E. J. Ng and C. Lee, J. Micromech. Microeng., 2021, 32, 014005.
25. Z. Luo, A. Zhang, W. Huang, S. Shao, Y. Liu, T. Wu and Y. Zou, IEEE J. Sel. Top. Quantum Electron., 2023, 1–18.
26. K. Roy, J. E.-Y. Lee and C. Lee, Microsyst. Nanoeng., 2023, 9, 95.
27. X. Le, Q. Shi, Z. Sun, J. Xie and C. Lee, Adv. Sci., 2022, 9, 2201056.
28. A. Haroun, X. Le, S. Gao, B. Dong, T. He, Z. Zhang, F. Wen, S. Xu and C. Lee, Nano Express, 2021, 2, 022005.
29. T.-T. Wu, Y.-Y. Chen and T.-H. Chou, J. Phys. D: Appl. Phys., 2008, 41, 085101.
30. H.-S. Hong and G.-S. Chung, Sens. Actuators, B, 2010, 150, 681–685.
31. W. Xuan, X. He, J. Chen, W. Wang, X. Wang, Y. Xu, Z. Xu, Y. Q. Fu and J. K. Luo, Nanoscale, 2015, 7, 7430–7436.
32. J. G. Rodriguez-Madrid, G. F. Iriarte, J. Pedros, O. A. Williams, D. Brink and F. Calle, IEEE Electron Device Lett., 2012, 33, 495–497.
33. F. Wang, J. Jian, X. Geng, G. Gou, W. Cui, J. Cui, Y. Qiao, J. Fu, Y. Yang and T.-L. Ren, IEEE Sens. J., 2020, 20, 9733–9739.
34. Y. Liu, H. Huang, L. Wang, D. Cai, B. Liu, D. Wang, Q. Li and T. Wang, Sens. Actuators, B, 2016, 223, 730–737.
35. M. Hara, J. Kuypers, T. Abe and M. Esashi, Sens. Actuators, A, 2005, 117, 211–216.
36. Z. Luo, S. Shao, K. Liu, Y. Lu, A. Mazzalai, C. Tosi and T. Wu, Appl. Phys. Lett., 2022, 120, 173508.
37. S. Shao, Z. Luo and T. Wu, IEEE Electron Device Lett., 2021, 42, 1378–1381.
38. C.-M. Lin, Y.-J. Lai, J.-C. Hsu, Y.-Y. Chen, D. G. Senesky and A. P. Pisano, Appl. Phys. Lett., 2011, 99, 143501.
39. A. Gao, K. Liu, J. Liang and T. Wu, Microsyst. Nanoeng., 2020, 6, 74.
40. J. Pang, X. Le, Z. Xu, C. Gao and J. Xie, J. Micromech. Microeng., 2018, 28, 105016.
41. J. Zou, C.-M. Lin, A. Gao and A. P. Pisano, J. Microelectromech. Syst., 2018, 27, 973–984.
42. X. Le, L. Peng, J. Pang, Z. Xu, C. Gao and J. Xie, Sens. Actuators, B, 2019, 283, 198–206.
43. J. Wang, Y. Zheng and A. Ansari, Phys. Status Solidi RRL, 2021, 15, 2100034.
44. X. Liu, D. Wang, K.-H. Kim, K. Katti, J. Zheng, P. Musavigharavi, J. Miao, E. A. Stach, R. H. Olsson and D. Jariwala, Nano Lett., 2021, 21, 3753–3761.
45. S. Shao, Z. Luo, Y. Lu, A. Mazzalai, C. Tosi and T. Wu, J. Microelectromech. Syst., 2022, 1–10.
46. M. Akiyama, T. Kamohara, K. Kano, A. Teshigahara, Y. Takeuchi and N. Kawahara, Adv. Mater., 2008, 21, 593–596.
47. D. F. Urban, O. Ambacher and C. Elsässer, Phys. Rev. B, 2021, 103, 115204.
48. J. Qin, X. Yang, C. Shen, Y. Chang, Y. Deng, Z. Zhang, H. Liu, C. Lv, Y. Li, C. Zhang, L. Dong and C. Shan, Nano Energy, 2022, 101, 107549.
49. P. Zhu, Y. Kuang, Y. Wei, F. Li, H. Ou, F. Jiang and G. Chen, Chem. Eng. J., 2021, 404, 127105.
50. T.-G. Kang, J.-K. Park, G.-H. Yun, H. H. Choi, H.-J. Lee and J.-G. Yook, Sens. Actuators, B, 2019, 282, 145–151.
51. R. Nitta, H.-E. Lin, Y. Kubota, T. Kishi, T. Yano and N. Matsushita, Appl. Surf. Sci., 2022, 572, 151352.
52. F. Barroso-Bujans, S. Cerveny, A. Alegría and J. Colmenero, Carbon, 2010, 48, 3277–3286.
53. J. D. N. Cheeke, N. Tashtoush and N. Eddy, Proc. IEEE Ultrasonics Symposium, vol. 1, pp. 449–452.
54. Z. Luo, S. Shao and T. Wu, Microelectron. Eng., 2021, 111530.
55. J. D. Larson, P. D. Bradley, S. Wartenberg and R. C. Ruby, Proc. IEEE Ultrasonics Symposium, vol. 1, pp. 863–868.
56. S. Bose, T. Kuila, A. K. Mishra, N. H. Kim and J. H. Lee, J. Mater. Chem., 2012, 22, 9696–9703.
57. Z. Duan, Y. Jiang and H. Tai, J. Mater. Chem. C, 2021, 9, 14963–14980.
58. Z. Duan, Q. Zhao, S. Wang, Z. Yuan, Y. Zhang, X. Li, Y. Wu, Y. Jiang and H. Tai, Sens. Actuators, B, 2020, 305, 127534.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3nr05684h

---