Skip to main content
Advanced Science logoLink to Advanced Science
. 2025 May 28;12(31):e01989. doi: 10.1002/advs.202501989

VOx‐Based Non‐Volatile Radio‐Frequency Switches for Reconfigurable Filter

Dabin Seo 1, Dahyeon Kim 2, Jiyeon Ryu 1, Changwoo Pyo 1, Seungchan Lee 2, Tae‐Sik Yoon 1,3, Myungsoo Kim 1,2,
PMCID: PMC12376588  PMID: 40432607

Abstract

Vanadium oxide (VOx) based memristor is a promising candidate for next‐generation non‐volatile memory and radio‐frequency (RF) switches due to its compatibility with wafer‐level integration and high‐frequency operation. This work demonstrates high‐performance VOx memristor with gold and silver electrodes, achieving a higher cutoff frequency compared to previously reported VOx devices. The devices exhibit long retention, high endurance (≈103 cycles), and nanosecond switching speeds, enabling the fabrication of RF switches with a cutoff frequency of ≈4.5 THz, low insertion loss (< 0.46 dB), and high isolation (>20 dB) from 0.1 to 20 GHz with stable operation extended to frequency up to 67 GHz. Leveraging these switches, a reconfigurable X‐band bandpass filter whose is realized center frequency is tuned from 8.2 GHz in the OFF state to 7.6 GHz in the ON state, achieving a tunable range of ≈600 MHz. This demonstration paves the way for compact and versatile RF front‐ends with improved frequency agility in advanced communication systems.

Keywords: 5G/6G communication, non‐volatile memory, radio‐frequency switch, reconfigurable filter, vanadium oxide


Vanadium oxide‐based memristor devices achieve high performance for next‐generation non‐volatile memory and RF switches. These devices demonstrate a cutoff frequency of ≈4.5 THz, low insertion loss (<0.46 dB), and high isolation (>20 dB) from 0.1 to 20 GHz with stable operation extended to 67 GHz. Leveraging these switches, a reconfigurable X‐band bandpass filter is realized for advanced RF systems.

graphic file with name ADVS-12-e01989-g003.jpg

1. Introduction

The burgeoning growth of wireless communication technologies, spurred by advancements in 5G/6G, the Internet of Things, and autonomous systems, drives an unprecedented demand for reconfigurable radio‐frequency (RF) front‐end systems.[ 1 , 2 , 3 , 4 , 5 ] To dynamically adapt to diverse operating frequency, bandwidths, and communication standards, these systems require components that offer both flexibility and efficiency, with the RF switch central to their operation as a critical element enabling rapid and efficient signal routing between multiple circuit paths.[ 6 , 7 , 8 , 9 , 10 , 11 ] While traditionally dominated by semiconductor technologies like p‐type intrinsic n‐type (PIN) diodes and field‐effect transistors, conventional RF switches present limitations, including high insertion loss, degraded isolation at higher frequency, and most significantly, the requirement of continuous power consumption to maintain their operational state (static power consumption), even when no signal is being transmitted.[ 12 , 13 , 14 ] These limitations pose substantial challenges for battery‐powered devices and energy‐conscious applications.

To overcome these drawbacks, memristive devices such as resistive random‐access memory (ReRAM), conductive‐bridging RAM (CBRAM), and phase‐change memory (PCM) have recently gained considerable attention as promising candidates for implementing next‐generation RF switches.[ 15 , 16 , 17 ] Memristive devices exploit a nonvolatile resistive switching mechanism, whereby the resistance of the device is reversibly modulated between a high resistance state (HRS) and a low resistance state (LRS) in response to an applied voltage or current.[ 18 , 19 ] This switching is typically associated with the formation and rupture of conductive pathways, often referred to as filaments, driven by electrochemical redox reactions, the migration of mobile ions, or valence change processes that locally alter the material's conductivity.[ 20 , 21 ] This inherent non‐volatility eliminates static power consumption, offering a decisive advantage over traditional volatile switches. The potential of memristor for achieving low insertion loss and high isolation across a wide frequency range has been demonstrated in several recent studies, exploring materials such as TaOx, HfOx, TiOx.[ 22 , 23 , 24 ] Particularly, structural engineering approaches like multi‐material stacks further enhance their durability and pave the way for the development of highly reliable memristive RF switches. Despite these advancements, challenges persist regarding switching speed, endurance, and seamless integration with existing RF circuit architectures.

Among various memristor materials, vanadium oxide (VOx), a transition metal oxide, has attracted considerable attention due to its distinctive material properties. Vanadium's rich polymorphism, characterized by multiple oxidation states (e.g., V2+, V3+, V4+, V5+) and the formation of numerous stable and metastable phases, including the Magnéli phases (VnO2n−1), provides a versatile platform for tailoring electrical and structural properties for various electronic applications. In addition, mobile dopants such as interstitial boron within VOx exhibit high diffusivity, imparting intrinsic metallic properties to the material.[ 25 , 26 ] This behavior indicates that VOx is capable of accommodating metallic ions with high mobility, allowing them to migrate efficiently within its layers. These unique characteristics make VOx a promising material in memristor applications, where it has demonstrated desirable attributes such as rapid switching speeds, robust endurance, and scalability for nanoscale device integration.[ 27 , 28 , 29 , 30 , 31 ]

This study leverages the distinctive properties of VOx to develop high‐performance memristive devices employing gold (Au) and silver (Ag) electrodes, forming conductive filaments, achieving a significantly improved figure of merit compared to previously reported oxide‐based devices. Our devices represent exceptional characteristics, including extended retention (> 60 h), high endurance (≈103 cycles), and nanosecond‐scale switching speeds. Building upon these Ag/VOx/Au devices, we successfully fabricated RF switches that exhibit a cutoff frequency of ≈4.5 THz, low insertion loss (< 0.46 dB), and high isolation (> 20 dB) across a frequency range of 0.1 to 20 GHz, while maintaining stable operation across an extended frequency range up to 67 GHz. To demonstrate practical applicability, a reconfigurable X‐band bandpass filter was implemented using these switches, achieving a tunable center frequency (7.6–8.2 GHz). In the OFF state, the filter exhibits a 3‐dB bandwidth of ≈600 MHz with an insertion loss of 5.6 dB, while in the ON state, it achieves a 3‐dB bandwidth of ≈650 MHz with an insertion loss of 9.6 dB. The filter also maintains a good return loss (> 20.2 dB), along with additional advantages of fast switching speeds, and zero‐static power dissipation due to their non‐volatile nature. These findings highlight the substantial potential of VOx‐based memristor technology to enable the creation of compact, versatile, and frequency‐agile RF front‐ends for the next‐generation of advanced communication systems.

2. Results and Discussion

2.1. Device Schematics and Images with Material Characterization

VOx was employed as the active layer in both memristor and RF devices, exhibiting unique multiple valence properties while forming a uniform thin film within the device structure. The 3D schematic views of the device in Figure 1a,b highlight the crossbar array architecture and the metal‐insulator‐metal (MIM) configuration, respectively. Figure 1c presents a scanning electron microscopy (SEM) image of the fabricated RF switch, integrated with a ground‐signal‐ground (GSG) coplanar waveguide for RF signal transmission. Atomic force microscopy (AFM) analysis of the Ag/VOx/Au stack (Figure S1, Supporting Information) reveals a VOx active layer thickness of ≈36 nm and an average surface roughness (R a) of 0.27 nm, indicating a smooth and uniform film. The fabrication process flow of the VOx RF switch is detailed in Figure S2 (Supporting Information). Briefly, the process involves electron‐beam (E‐beam) evaporation of the bottom Au electrode onto the silicon oxide (SiO2) substrate, followed by sputtering and wet‐etching of the 20 µm × 20 µm VOx active layer. Finally, a 100 nm‐thick top Ag feedline electrode was patterned and deposited onto the VOx layer.

Figure 1.

Figure 1

Device schematics and images with material characterization. a) Schematic illustration of the VOx‐based memristor crossbar array. b) Zoomed‐in perspective view and side view cross‐sectional illustration of the Ag/VOx/Au structure on a SiO2 substrate, demonstrate the overall device layout and the detailed layered configuration. c) The representative top view SEM image of the signal transmission area, and the dashed circle marks the area where the VOx was deposited. d) XPS spectra of O 1s and V 2p region for the sputter‐deposited VOx layer, which is etched 10 nm from the surface. e) Cross‐sectional HR‐TEM micrograph of C/VOx/SiO2 (left), and magnified VOx layer in the red‐solid‐lined rectangles (right). f) SADP image of C/VOx/SiO2 layer.

X‐ray photoelectron spectroscopy (XPS) analysis was performed to investigate the composition and chemical bonding states of the deposited VOx layer. The O 1s spectrum revealed peaks for V─O bonds (OL, 528.7 eV) and oxygen vacancies (OV, 529.4 eV). V 2p1/2 and V 2p3/2 spectra were deconvoluted into V5+ and V4+ peaks, obtained at 522.4 and 520.8 eV for V 2p1/2, and 515.1 and 513.5 eV for V 2p3/2, respectively, with higher V4+ peak intensity. This indicates mixed phases of VO2 and V2O5, differing from stoichiometric V2O5 with dominant V5+ peaks.[ 32 , 33 , 34 , 35 , 36 ] According to the previous report by Ryu et al.,[ 29 ] rutherford backscattering spectroscopy analysis and auger electron spectroscopy of VOx thin films which deposited under identical conditions in this confirmed an oxygen‐to‐vanadium ratio below 2.5, consistent with oxygen‐deficient VOx layers. These oxygen vacancies facilitate Ag+ ion migration by lowering the activation energy and creating diffusion pathways confirmed by nudged elastic band (NEB) simulation and density functional theory (DFT) calculations, enabling fast resistive switching at low voltage.[ 37 , 38 ] More detailed XPS results, and their corresponding analysis are discussed in Note S1 (Supporting Information).

In addition to the chemical bonding states, the microstructure of VOx layer was examined using transmission electron microscopy (TEM) and selected area diffraction patterns (SADPs) as shown in Figure 1e,f. For these analyses, the VOx layer was deposited on a SiO2 substrate under the same conditions as those used for the Ag/VOx/Au device, except for the deposition time. An amorphous carbon (C) layer was coated on the VOx thin film prior to TEM sample preparation to prevent ion beam damage during focused ion beam milling. Figure 1e shows a cross‐sectional high‐resolution transmission electron microscopy (HR‐TEM) micrograph of C/VOx/SiO2 specimen, and Figure 1f shows its SADP. The HR‐TEM images confirm that the VOx layer exhibits a homogeneous amorphous microstructure with an almost negligible amount of partially crystalline phases (depicted as red rectangles in Figure 1f), which is consistent with the SADPs result. In the amorphous structure of the VOx layer, there are no preferential ion migration paths, such as grain boundaries in the polycrystalline layer. Therefore, it leads to reliable device operation as will be further discussed. The amorphous nature of VOx layer makes it have a flat surface, consistent with the AFM result in Figure S1 (Supporting Information).[ 39 , 40 , 41 ] This flat surface is advantageous to achieve stable device operation, particularly in filamentary‐type devices, as it minimizes the variation caused by the non‐uniform electric field in the device with rough metal‐dielectric interfaces.[ 42 , 43 ] The inherent flatness of VOx mitigates these issues, ensuring reduced variability and enhanced reliability in electrical properties, as detailed in the next section.

2.2. DC Switching Characteristics of Ag/VOx/Au Device

Direct current (DC) electrical characterization was performed on as‐fabricated Ag/VOx/Au crossbar devices, revealing their non‐volatile resistive switching behavior. The devices were SET from the HRS to the LRS by applying a voltage sweep, and the non‐volatile nature of the switching allowed the devices to retain their LRS state without continuous input power. Figure 2a presents the current–voltage (IV) characteristics of a 4 µm × 4 µm device, obtained by sweeping the voltage applied to the Ag top electrode from 0 to +3.0 V, back to 0 V, then to −1.1 V, and finally returning to 0 V, while grounding the Au bottom electrode. A compliance current of 1 and 3 mA was enforced during the SET and RESET processes to prevent permanent device breakdown. The device exhibits a stable ON/OFF ratio of ≈107, as demonstrated by the retention measurements in Figure 2b. Both the LRS and HRS were maintained for over 60 h at room temperature under a read voltage of +0.1 V, indicating outstanding stability and reliability, surpassing previously reported VOx‐based memristor devices.[ 28 , 44 , 45 ] Figure 2c further confirms the robustness of the device through read disturbance testing, where applying over 105 read pulses of +0.02 V with a 10 µs width resulted in negligible degradation of both resistance states. Figure S3 (Supporting Information) demonstrates the high endurance of the device, maintaining a stable ON/OFF ratio over 103 switching cycles. The distribution of SET voltage (V SET) and RESET voltage (V RESET) over these cycles is presented in Figure 2d, revealing median values of +2.55 and −0.9 V, respectively. These relatively higher switching voltages, compared to some previously reported VOx memristor, are attributed to the crystalline structure and the overlap area of our devices.[ 46 , 47 ] As the amorphous phase of VOx inherently contains a significant number of defects, and larger cells contain proportionally more defects than smaller ones, the activation energy for defect‐driven filament formation is reduced in larger cells.[ 48 , 49 ] This leads to lower switching voltages in larger devices, a trend we confirmed by fabricating a device with a 100 µm diameter using a shadow mask (Figure S4, Supporting Information). This larger device exhibited significantly reduced V SET and V RESET of +0.49 and −0.13 V, respectively, albeit with a reduced ON/OFF ratio of ≈100 times compared to the smaller crossbar device. This observation suggests the correlation between the overlap area, switching voltage, and ON/OFF ratio in the Ag/VOx/Au memristor. Figure 2e presents the rapid SET switching capability, with the device transitioning from the HRS to the LRS within a 100 ns pulse. Figure 2f reveals the corresponding power consumption and switching energy during this SET event, highlighting the energy efficiency of the switching process. A comprehensive comparison of the VOx device performance against previously reported devices is provided in Table S1 (Supporting Information). Notably, our device exhibits improved or comparable performance across key metrics, including a higher ON/OFF ratio, longer retention, and greater cycling endurance. Furthermore, Figure S5 (Supporting Information) demonstrates the tunability of the ON state resistance through compliance current control. Increasing the compliance current during the SET process results in a lower ON state resistance, indicating the ability of the device to withstand high currents while achieving a desirable low‐resistance state for RF switch applications.

Figure 2.

Figure 2

DC switching characteristics of VOx‐based memristor. a) Representative I–V curve of bipolar non‐volatile resistive switching effect in Ag/VOx/Au device with 4 µm × 4 µm overlap area and SET compliance current of 1 mA. Step 1: The voltage is swept from 0 to +2.4 V causes a sharp current increase to the compliance value, indicating the SET transition from HRS to LRS. Step 2: Decreasing the voltage from +3 to 0 V maintains the LRS, demonstrating non‐volatile characteristics. Step 3: The voltage decreases from 0 to −1 V. The sharp drop in current ≈−0.8 V shows the RESET process from LRS to HRS. Step 4: The voltage returns to 0 V. b) Retention performance of HRS and LRS states measured at room temperature (RT) with a +0.1 V of V Read. The device demonstrates an ON/OFF ratio of 107 with a compliance current of 3 mA. Retention time exceeds 60 h, indicating stable non‐volatile characteristics. c) Read disturbance characteristics of HRS and LRS under 105 read pulse stress, indicating minimal degradation and demonstrating the reliability of the resistive switching behavior. d) Distribution of the V SET and V RESET observed in the VOx memristor switch for ≈103 cycles. The median V SET and V RESET are +2.55 and −0.9 V, respectively. e) Nanosecond input voltage and output current pulse waveform characteristics of the VOx memristor during the SET processes. The VSET pulse amplitudes were set to ≈2.0 V with a pulse width of 100 ns. f) Consumed energy and power during the 100 ns SET pulse. The consumed energy is determined by integrating the power over time, yielding a measured SET energy of 3.1 nJ.

2.3. Switching and Conduction Mechanism of Ag/VOx/Au Memristor Devices

To elucidate the transport mechanism in the Ag/VOx/Au device, current fitting analyses and low‐temperature measurements were conducted, providing strong evidence for a filamentary conduction model involving the rapid migration of Ag+ ions within the VOx layer. In the HRS, as shown in Figure 3a, a linear relationship between ln(I/T 2) and V 1/2 is observed, indicative of Schottky emission. Further analysis of the IV characteristics at varying temperatures (Figure 3b,c) reveals non‐linear I–V curves in the HRS, with current increasing with temperature increases, which is a typical characteristic of insulators and semiconductors. The HRS data were best fitted by the Schottky emission model, demonstrating good agreement when plotting logarithmic J (A/cm2) versus E 1/2 (V/cm)1/2 across all temperatures. The Schottky barrier height was calculated using Equation (1):[ 50 ]

J=AT2expqϕBqE4πεrε0kTA=120mm0 (1)

where J is the current density, A* is the effective Richardson constant, m0 is the free electron mass, m* is the effective mass, q is the electron charge, T is the absolute temperature, E is the electric field across the dielectric, k is Boltzmann's constant, ε 0 is the permittivity of free space, and ε r is the relative dielectric constant. Using an effective thickness of ≈40 nm and assuming m */m 0 ≈ 1, the extracted barrier height (ϕ B) is ≈0.245 eV at 300 K, based on the relative dielectric constant of VOx (ε r = 14.6) measured from capacitance–voltage (CV) (Figure S6, Supporting Information). In the LRS, the double‐logarithmic IV plot (Figure 3d) exhibits a slope of ≈1, indicative of Ohmic conduction behavior resulting from the formation of conductive filaments. The decrease in LRS current with increasing temperature (Figure 3e) suggests metallic‐like conduction within the filament. To further characterize the nature of these filaments, the temperature coefficient of resistance in the LRS was determined using Equation (2):[ 51 ]

RT=R01+αTT0 (2)

where R 0 is the resistance at a reference temperature T 0, and α is the temperature coefficient of resistance. Linear fitting of the normalized LRS resistance (R T/R 0) as a function of temperature (Figure 3f) yields a temperature coefficient of resistance (α) of 4.2 × 10−3 K−1 for the Ag/VOx/Au device. This value closely matches the reported α of Ag (4.1 × 10−3 K−1), strongly suggesting that the conductive filaments are primarily composed of Ag, formed by the diffusion of Ag+ ions from the top electrode into the VOx layer.[ 52 ]

Figure 3.

Figure 3

Switching and conduction mechanism of Ag/VOx/Au device. a) ln (I/T 2) versus V 1/2 curves in the HRS, indicating Schottky emission conduction. b,e) Typical IV characteristics of VOx memristor measured at different temperatures, with (b) showing data for the HRS with a 6 µm × 6 µm overlap area and (e) showing data for the LRS with a 2 µm × 2 µm overlap area. c) Fitted data for HRS using the Schottky emission model, demonstrating the linear relationship with the value of R 2 ≈0.99 in the logarithmic J versus E 1/2 plot. d) log (I) versus log (V) curves in the LRS, indicating Ohmic conduction. f) Normalized temperature‐dependent resistance in the LRS with the slope representing the α.

Building on these findings, the conduction mechanism can be further understood by examining the interface properties. Considering that the work function of n‐type VOx typically ranges from 4.6 to 5.02 eV, the higher work function of the Au electrode (5.1 eV) is consistent with the formation of a Schottky barrier at the VOx/Au interface.[ 53 , 54 , 55 ] Conversely, the lower work function of the Ag electrode (4.32 eV) suggests the formation of an Ohmic contact at the Ag/VOx interface, aligning with the observed experimental IV characteristics.[ 56 ]

2.4. Radio‐Frequency Characterization of VOx RF Switches

The high‐frequency performance of the VOx RF switch was evaluated by measuring its scattering parameters (S‐parameters) in both the ON and OFF states using a vector network analyzer (VNA). Accurate measurements were ensured by performing an on‐wafer short‐open‐load‐through (SOLT) calibration to eliminate the influence of the test cables and probe station. The calibration accuracy was verified by measuring the calibrated S11 and S21 of reflection standards, as shown in Figures S7–S10 (Supporting Information). To extract the intrinsic S‐parameters of the VOx RF switch, de‐embedding procedures were performed using on‐chip test patterns to remove the contributions of probe‐pad and interconnect parasitics.

Figure 4a,b present the de‐embedded S21 results for Ag/VOx/Au devices with an overlap area of 4.3 µm × 0.5 µm and switching with a compliance current of 10 mA. The VOx switch demonstrates excellent RF performance, exhibiting an insertion loss of only 0.45 dB in the ON state and an isolation of 20 dB in the OFF state up to 20 GHz. An equivalent lumped element circuit model (Figure S11, Supporting Information) was used to extract the key performance parameters: ON state resistance (R ON), OFF state capacitance (C OFF), and cutoff frequency (F CO). From the de‐embedded S‐parameters, R ON and C OFF were determined to be 5 Ω and 7 fF, respectively, resulting in a calculated F CO of ≈4.5 THz based on the relation F CO = 1/2πR ON C OFF.[ 57 ] To assess the applicability of the switch at a higher frequency, measurements were extended up to 67 GHz (Figure S12, Supporting Information), demonstrating reliable performance in this regime and emphasizing its potential for integration into advanced mmWave technologies. A benchmark comparison of various RF switch technologies (Table S2, Supporting Information) reveals that the VOx switch offers comparable or superior performance to PCM and microelectromechanical system (MEMS) switches, with the added advantages of heater‐less ambient integration and low switching voltage. Notably, it exhibits a higher ON/OFF ratio and cutoff frequency among oxide‐based RF switches.

Figure 4.

Figure 4

Radio‐frequency characterization of VOx RF switches. Experimental S‐parameter data in both the a) ON state and b) OFF state Ag/VOx/Au structured RF switches with an overlap area of 4.3 µm × 0.5 µm with measurements up to 20 GHz. The VOx RF switch was measured with a current compliance of 10 mA. The values of C OFF (7 fF) and R ON (5 Ω) were extracted using a lumped element equivalent circuit model. Calculated from this, the cutoff frequency is 4.5 THz. The insertion loss at 20 GHz in the ON state was 0.455 dB, and the isolation at 20 GHz in the OFF state was 20.032 dB. The dashed lines are the simulated data from the equivalent circuit model. Power‐handling data of VOx‐based RF switches, with c) representing the ON state and d) the OFF state. The black line shows the output power on a dBm scale, while the red line shows the normalized S21 on a dBm scale as a function of input power. In both states, no 1 dB power compression occurred at input power above 10 dBm, the measurement limit of the VNA source.

High‐power handling capability is crucial for system‐level applications of RF switches. Therefore, the 1 dB compression point (P 1dB) was evaluated through power‐dependent loss measurements. Figure 4c,d show the power handling characteristics of a 2 µm × 2 µm VOx RF switch at 40 GHz for input power levels ranging from −20 to +10 dBm. In the ON state, the output power exhibits a linear relationship with the input power, with no observable output power compression (> 0.01 dB) up to an input power of +10 dBm. In the OFF state, the switch maintains high isolation, with output power exhibiting losses exceeding 35 dB and remaining linear for input powers up to +10 dBm.

2.5. Reconfigurable Bandpass Filter Using VOx RF Switch

The increasing demand for compact, low‐loss, and reconfigurable bandpass filters (BPFs) in microwave and RF communication systems has driven the exploration of novel switching technologies.[ 58 , 59 , 60 ] Reconfigurable BPFs offer dynamic frequency tuning through external control, leading to benefits such as improved signal‐to‐noise ratio and enhanced out‐of‐band performance.[ 61 ] While various approaches for frequency tuning have been proposed, including RF MEMS switches, PIN diodes, and ferroelectric capacitors, they often suffer from limitations in switching speed, power consumption, or integration complexity.[ 62 , 63 , 64 ]

This work presents a novel reconfigurable X‐band BPF employing VOx memristor devices integrated within a 50‐µm‐pitch GSG configuration. This design leverages the fast switching speed and non‐volatility of the VOx switch to achieve zero‐static power consumption, offering a distinct advantage over conventional approaches.[ 65 ] The filter, optimized using the Advanced Design System (ADS) simulator, was fabricated on a 430 µm‐thick sapphire substrate (εr = 11.6). Figure 5a,b show the schematic and optical image of the compact 10.8 mm × 0.9 mm filter, respectively. The total resonator length (l 1 + l 2 + l 3) is 5.62 mm, the coupling length (l 2) is 1 mm, and the gap between the feed lines and the resonators (g 0) is 0.015 mm. Detailed information on the fabrication process and structural dimensions is provided in Figure S13 (Supporting Information). The integrated VOx switches exhibit stable and low‐power operation, as demonstrated by the DC I–V curves in Figure S14 (Supporting Information), showing consistent characteristics over 100 cycles with both V SET and V RESET within ±1 V.

Figure 5.

Figure 5

The typical measured and simulated S‐parameters of the filter are presented for both switch states. a) Schematic diagram of the X‐band filter with key dimensions labeled and b) optical image of the fabricated filter. The dashed box indicates the VOx memristor switches with an overlap area of 5 µm × 30 µm. c) The typical measured and simulated S‐parameters of the filter in both switch states: i) Measured S11 and ii) S21 showing the tunable bandpass filter using VOx switches in ON and OFF states. iii) Comparison of measured and simulated S‐parameter in the OFF state, and iv) in the ON state, with the dashed lines representing simulation results obtained from ADS.

Figure 5c presents the measured and simulated S‐parameters of the filter in both the ON and OFF states, demonstrating its reconfigurable behavior. In the OFF state, the filter operates at a center frequency of 8.2 GHz (simulated: 8.15 GHz) with an insertion loss of 5.6 dB, a 3‐dB bandwidth of 600 MHz, and a return loss of 20.2 dB. In the ON state, the center frequency shifts to a lower value of 7.6 GHz (simulated: 7.6 GHz), with an insertion loss of 9.6 dB, a bandwidth of 650 MHz, and a return loss of 28.2 dB. The quality factor (Q) of resonators with VOx switches, estimated using Equation (3), is 112.3 in the OFF state and 89.4 in the ON state.[ 66 ]

Q=fcFBW (3)

f c is the center frequency and FBW is fractional bandwidth.

The measured center frequency closely match the ADS simulation results, validating the filter design. Further optimization of parameters such as substrate permittivity and metal conductivity can enhance accuracy and reliability (Note S2, Supporting Information). However, considering the complex interactions within the filter structure, more in‐depth theoretical studies are warranted to fully elucidate the underlying physics. Figure S15 (Supporting Information) illustrates the simulated surface current density distribution, highlighting the activation of the parasitic patch through the connection between the main patch and the parasitic patch, demonstrating the principle of frequency tuning via switch integration. Table S3 (Supporting Information) benchmarks the performance of this VOx‐based BPF against other reported X‐band filters, emphasizing its superior advantages of zero‐static power dissipation due to the non‐volatile nature of the VOx switches, coupled with fast switching speed.

3. Conclusion

In conclusion, this work has successfully demonstrated the development of nanoscale, non‐volatile RF switches based on VOx for next‐generation wireless communication systems. The fabricated Ag/VOx/Au switches exhibit robust non‐volatile bipolar resistive switching behavior, with a high ON/OFF ratio of ≈107 and switching voltages below 3 V. Furthermore, the VOx switches demonstrate rapid switching with pulse operation, consuming a maximum power of less than 18.5 mW and an energy below 3.2 nJ. Advanced frequency characterization techniques, including precise calibration and de‐embedding, reveal outstanding RF performance, including an exceptionally low insertion loss of 0.45 dB and isolation exceeding 20 dB across a wide frequency range. The practical application of these switches was demonstrated through their integration into a reconfigurable X‐band filter, achieving tunable frequency with fast switching speeds and, crucially, zero‐static power dissipation due to the non‐volatile nature of the VOx memristor. As shown in Figure 6 , our device demonstrates a 4 times improvement in retention and a 105 times increase in ON/OFF ratio compared to previously reported VOx‐based memory, achieved through contact metal engineering and device structure optimization. Furthermore, the device exhibits a 60% higher cutoff frequency compared to previously reported oxide‐based RF switches. To further enhance device performance, additional studies on the optimization and analysis of VOx film characteristics are required. These findings underscore the significant potential of our device as high‐performance, energy‐efficient, nanoscale switches for future reconfigurable RF front‐end systems in advanced wireless communication applications.

Figure 6.

Figure 6

a) Benchmarking the ON/OFF ratio and retention times of VOx memristor against other reported VOx‐based devices. The plotted data highlight the outstanding performance of this work, achieving a significantly higher ON/OFF ratio (≈107) and extended retention times (>105 s) compared to previously reported devices with different electrode configurations.[ 27 , 28 , 29 , 44 , 45 ] b) Cutoff frequency comparison of VOx RF switches with various non‐volatile RF switch technologies. The green region represents 2D‐based RF switches, the blue region corresponds to PCM‐based RF switches, and the red region denotes oxide‐based RF switches. The highlighted performance of this work demonstrates a VOx RF switch achieving a cutoff frequency of 4.5 THz, showing superior potential in oxide‐based RF switches.[ 23 , 24 , 67 , 68 , 69 , 70 ] c) Radar chart of representative non‐volatile RF switching materials. While exhibiting a superior cutoff frequency, 2D materials are limited by low cycling endurance; similarly, PCM's scalability is constrained by its heater requirement. Oxide materials, in contrast, demonstrate a well‐balanced performance across all key metrics.

4. Experimental Section

Device Fabrication and Characterization

The VOx memristor and RF switches were fabricated on a SiO2 substrate, with the memristor patterned in a crossbar array structure and the RF switches patterned in a 50‐µm‐pitch GSG configuration, both using photolithography. An E‐beam evaporator was used to deposit the ground pads and the bottom electrode consisting of 2‐nm‐thick Cr for the adhesion layer and 60‐nm‐thick Au. Then, a 40‐nm‐thick VOx layer was deposited onto the entire wafer by RF magnetron sputtering using a high‐purity VO2 target (V:O = 1:2) with a diameter of 2 inches. The deposition was conducted at room temperature, with a base pressure in the low 10−6 Torr range and a working pressure of 20 mTorr under an Ar atmosphere (20 sccm). The RF power was maintained at 150 W during the deposition. The VOx active layer was patterned by photolithography, followed by an etching process. To ensure a small area of the top electrode, it was patterned by electron beam lithography (EBL) and deposited by E‐beam evaporator of 2‐nm‐thick Cr for the adhesion layer and 100‐nm‐thick Ag. The structure of the device was analyzed via high‐resolution transmission electron microscopy (HR‐TEM; JEM‐2100F, JEOL). The compositions and chemical bonding states of the VOx layer were examined through X‐ray photoelectron spectroscopy (XPS; K‐alpha, ThermoFisher). XPS was performed using a monochromated Al‐Kα source ( = 1486.6 eV), with the system featuring a micro‐focused monochromator and a double‐focusing hemispherical analyzer with a 128‐channel detector. The experimental XPS spectra were deconvoluted using the CasaXPS software. Scanning electron microscopy images were collected on a (Cold FE‐SEM; S‐4800, Hitachi High‐Technologies) instrument with the beam energy at 5 kV.

Bandpass Filter Fabrication

The 1.0 cm × 1.0 cm area of a 430 µm‐thick sapphire substrate was dried by a nitrogen gun after rinsing with acetone, and isopropyl alcohol (IPA). The 360 nm‐thick Au electrode was deposited on the sapphire substrate and patterned by a photolithography system. To minimize the undercutting of the etchant into the metal, inductively coupled plasma‐reactive ion etching (ICP‐RIE) was performed for 120 s using Cl2 gas, after which etching was performed using Au etchant and Cr etchant. The 300 µm × 400 µm area VOx active layer was deposited by the RF magnetron sputtering with a shadow mask. The top electrode, consisting of a 2‐nm‐thick Cr adhesion layer and a 100‐nm‐thick Ag layer, was patterned by EBL to form a 5 µm × 30 µm overlap area with the VOx/Au stack, followed by deposition using an E‐beam evaporator.

DC Measurements

The electrical characteristics of the device including resistive switching, endurance properties, voltage pulse measurements, and read pulse disturbance measurements were examined using a Keysight B1500A semiconductor parameter analyzer with a B1530 waveform generator/fast measurement unit pulsing module under ambient conditions. For the voltage pulse measurements, a rectangular pulse with a rising time of 100 ns, a pulse width of 100 ns at an amplitude of 2 V, and a falling time of 100 ns was applied to the device. The low‐temperature measurements were conducted using a Keithley 4200‐SCS semiconductor parameter analyzer from 100 to 300 K under vacuum conditions with the chamber pressure maintained at 10−6 torr. In the measurement process, the voltage was applied to the top Ag electrode while the bottom Au electrode was grounded.

RF Measurements

The Formfactor Infinity GSG probes and the Keysight P5024A‐402 VNA was used for on‐wafer measurements from 10 MHz to 20 GHz, while the Keysight N5247A VNA was employed for measurements extending up to 67 GHz to evaluate higher frequency operation. The RF power was set to a nominal level of −20 dBm for small‐signal measurements. Before the actual device measurement, a Formfactor 138–356 Impedance Standard Substrate and WinCal software (FormFactor Inc.) were used to perform the SOLT calibration. Due to the non‐volatile resistive switching of the VOx switch device, a forward and reverse DC bias was employed to turn the switching devices ON and OFF. S‐parameters were measured for each device in both the ON and OFF states, and parameters were extracted from the de‐embedded S‐parameters to determine the ON and OFF state equivalent lumped circuit component models. A de‐embedding process utilizing an open‐through configuration was employed to eliminate the pad and interconnect resistances. The power handling capability was evaluated in both the ON and OFF states by measuring the S‐parameters as the devices were incrementally subjected to RF power levels ranging from −20 to +10 dBm.

Conflict of Interest

The authors declare no conflict of interest.

Author Contributions

D.S. and D.K. contributed equally to this work. D.S., D.K. and J.R. performed all the material characterizations, device fabrications, DC/RF measurements, and electromagnetic simulations. C.P. and S.L. performed data curation and formal analysis. D.S. and M.K. examined all of the electrical data and prepared the manuscript, with help from the other collaborators. M.K. coordinated and supervised the research.

Supporting information

Supporting Information

ADVS-12-e01989-s001.docx (6.5MB, docx)

Acknowledgements

This research was supported by the National Research Foundation (NRF) funded by the Korean government (MSIT) (RS‐2024‐00337000, IITP‐2025‐RS‐2022‐00156361) and Korea Institute for Advancement of Technology (KIAT) grant funded by the Korea Government (MOTIE) (P0023703). This study contains the results obtained by using the equipment of UNIST Central Research Facilities(UCRF).

Seo D., Kim D., Ryu J., Pyo C., Lee S., Yoon T.‐S., Kim M., VOx‐Based Non‐Volatile Radio‐Frequency Switches for Reconfigurable Filter. Adv. Sci. 2025, 12, e01989. 10.1002/advs.202501989

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

References

  • 1. Kim D., Yang S. J., Wainstein N., Skrzypczak S., Ducournau G., Pallecchi E., Happy H., Yalon E., Kim M., Akinwande D., Nat. Rev. Electr. Eng. 2024, 1, 10. [Google Scholar]
  • 2. Wainstein N., Adam G., Yalon E., Kvatinsky S., Proc. IEEE 2021, 109, 77. [Google Scholar]
  • 3. Lu Y., Zheng X., J. Ind. Inf. Integr. 2020, 19, 100158. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4. Elayan H., Amin O., Shihada B., Shubair R. M., Alouini M. S., IEEE Open J. Commun. Society, IEEE, Piscataway, NJ: 2020, pp. 1‐32. [Google Scholar]
  • 5. Feng M., Shen S. C., Caruth D. C., Huang J. J., in Proc. IEEE, IEEE, Piscataway, NJ: 2004, pp. 354‐375. [Google Scholar]
  • 6. Kumar R., Microsyst. Technol. 2021, 27, 2525. [Google Scholar]
  • 7. Bhatia V., Kaur S., Sharma K., Rattan P., Jagota V., Kemal M. A., Wirel. Commun. Mob. Comput. 2021, 8, 2021513. [Google Scholar]
  • 8. Ma L. Y., Soin N., Daut M. H. M., Hatta S. F. W. M., IEEE Access 2019, 7, 107506. [Google Scholar]
  • 9. Singh T., Mansour R. R., IEEE Trans. Microw. Theory Tech. 2021, 69, 4697. [Google Scholar]
  • 10. Lavadiya S. P., Sorathiya V., Kanzariya S., Chavda B., Naweed A., Faragallah O. S., Eid M. M. A., Rashed A. N. Z., Int. J. Commun. Syst. 2022, 35, 5141. [Google Scholar]
  • 11. Moon J. S., Seo H. C., Son K. A., Lee K., Zehnder D., Tai H., in IEEE MTT‐S Int. Microwave Symposium Digest, IEEE, Philadelphia, PA: 2018, pp. 836‐838. [Google Scholar]
  • 12. Lee J. L., Zych D., Reese E., Drury D. M., IEEE Trans. Microw. Theory Tech. 1995, 43, 250. [Google Scholar]
  • 13. Gopinath A., Rankin J. B., IEEE Trans. Electron Devices. 1985, 32, 1272. [Google Scholar]
  • 14. Li Q., Zhang Y. P., Yeo K. S., Lim W. M., IEEE Trans. Microw. Theory Tech. 2008, 56, 339. [Google Scholar]
  • 15. Singh T., Mansour R. R., IEEE Trans. Microw. Theory Tech. 2019, 67, 3237. [Google Scholar]
  • 16. Wang M., Rais‐Zadeh M., IEEE MTT‐S Int. Microw. Symp. Dig. 2014, 1. [Google Scholar]
  • 17. Pi S., Ghadiri‐Sadrabadi M., Bardin J. C., Xia Q., Nat. Commun. 2015, 6, 7519. [DOI] [PubMed] [Google Scholar]
  • 18. Chua L. O., IEEE Transactions on Circuit Theory 1971, 18, 507. [Google Scholar]
  • 19. Strukov D. B., Snider G. S., Stewart D. R., Williams R. S., Nature 2008, 453, 80. [DOI] [PubMed] [Google Scholar]
  • 20. Yang J. J., Pickett M. D., Li X., Ohlberg D. A. A., Stewart D. R., Williams R. S., Nat. Nanotechnol. 2008, 3, 429. [DOI] [PubMed] [Google Scholar]
  • 21. Xiao Y., Jiang B., Zhang Z., Ke S., Jin Y., Wen X., Ye C., Sci. Technol. Adv. Mater. 2023, 24, 2162323. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22. Lee M. J., Lee C. B., Lee D., Lee S. R., Chang M., Hur J. H., Kim Y. B., Kim C. J., Seo D. H., Seo S., Chung U. I., Yoo I. K., Kim K., Nat. Mater. 2011, 10, 625. [DOI] [PubMed] [Google Scholar]
  • 23. Chen S. C., Yang Y. T., Tseng Y. C., Chiou K. D., Huang P. W., Chih J. H., Liu H. Y., Chou T. T., Jhang Y. Y., Chen C. W., Kuan C. H., Ho E. M., Chien C. H., Kuo C. N., Cheng Y. T., Lien D. H., ACS Nano 2024, 1, 704. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24. Ruxandra B. E. A., Proc. Innov. Sustainability Conf. 2017, 41. [Google Scholar]
  • 25. Sellers D. G., Braham E. J., Villarreal R., Zhang B., Parija A., Brown T. D., Alivio T. E. G., Clarke H., De Jesus L. R., Zuin L., Prendergast D., Qian X., Arroyave R., Shamberger P. J., Banerjee S., J. Am. Chem. Soc. 2020, 142, 15513. [DOI] [PubMed] [Google Scholar]
  • 26. Alivio T. E. G., Sellers D. G., Asayesh‐Ardakani H., Braham E. J., Horrocks G. A., Pelcher K. E., Villareal R., Zuin L., Shamberger P. J., Arróyave R., Shahbazian‐Yassar R., Banerjee S., Chem. Mater. 2017, 29, 5401. [Google Scholar]
  • 27. Li W., Tuo Y., Mi W., Wang D., Wang M., Zhou L., Zhao J., Vacuum 2023, 209, 111794. [Google Scholar]
  • 28. Nirantar S., Mayes E., Rahman M. A., Ahmed T., Taha M., Bhaskaran M., Walia S., Sriram S., Adv. Electroni. Mater. 2019, 5, 1900605. [Google Scholar]
  • 29. Ryu J., Park K., Sahu D. P., Yoon T. S., ACS Appl. Mater. Interfaces 2024, 16, 26450. [DOI] [PubMed] [Google Scholar]
  • 30. Wu C., Feng F., Xie Y., Chem. Soc. Rev. 2013, 42, 5157. [DOI] [PubMed] [Google Scholar]
  • 31. Hu P., Hu P., Vu T. D., Li M., Wang S., Ke Y., Zeng X., Mai L., Long Y., Chem. Rev. 2023, 123, 4353. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32. Rakshit A., Mukherjee M., Chakraborty S., Mater. Sci. Semicond. Process. 2018, 88, 127. [Google Scholar]
  • 33. Plugaru R., Mihalache I., Romaniţan C., Comanescu F., Vulpe S., Craciun G., Plugaru N., Djourelov N., Sensors 2023, 23, 1759. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34. Faceira B., Nayak S. S., Teulé‐Gay L., Labrugère‐Sarroste C., Huang H. Y., Shao Y. C., Dong C. L., Rougier A., ACS Appl. Energy Mater. 2024, 7, 9882. [Google Scholar]
  • 35. Şakar B. C., Yıldırım F., Aydoğan Ş., J. Phys. D: Appl. Phys. 2024, 57, 305108. [Google Scholar]
  • 36. Silversmit G., Depla D., Poelman H., Marin G. B., De Gryse R., Surf. Sci. 2006, 600, 3512. [Google Scholar]
  • 37. Patel K., Cottom J., Bosman M., Kenyon A. J., Shluger A. L., Microelectron. Reliab. 2019, 98, 144. [Google Scholar]
  • 38. Rasheed U., Imran M., Hussain F., Mumtaz U., Tighezza A. M., Khalil R. M. A., Ehsan M. F., J. Phys. Chem. Solids 2024, 193, 112214. [Google Scholar]
  • 39. Lau W. S., Zhang J., Wan X., Luo J. K., Xu Y., Wong H., AIP Adv. 2014, 4, 027120. [Google Scholar]
  • 40. Lee J. D., Shim B. C., Uh H. S., Park B. G., IEEE Electron Device Lett. 1999, 20, 215. [Google Scholar]
  • 41. Huang T. W., Lee H. Y., Hsieh Y. W., Lee C. H., J. Cryst. Growth. 2002, 237–239, 492. [Google Scholar]
  • 42. Raghavan N., Bosman M., Frey D. D., Pey K. L., Microelectron. Reliab. 2014, 54, 2266. [Google Scholar]
  • 43. Reuben J., Biglari M., Fey D., IEEE Trans. Nanotechnol. 2020, 19, 508. [Google Scholar]
  • 44. Li G., Wei J., Wang H., Xiong R., Wang D., Zhu Y., Liu Y., Zou Z., Xu J., Ma H., Ceram. Int. 2021, 47, 27479. [Google Scholar]
  • 45. Hota M. K., Nagaraju D. H., Hedhili M. N., Alshareef H. N., Appl. Phys. Lett. 2015, 107, 163106. [Google Scholar]
  • 46. Wang D., Zhang C., Han C., Qian L., Huang X., J. Alloys Compd. 2022, 921, 166226. [Google Scholar]
  • 47. Wong F. J., Sriram T. S., Smith B. R., Ramanathan S., Solid‐State Electron. 2013, 87, 21. [Google Scholar]
  • 48. Kim H. J., Park T. H., Yoon K. J., Seong W. M., Jeon J. W., Kwon Y. J., Kim Y., Kwon D. E., Kim G. S., Ha T. J., Kim S. G., Yoon J. H., Hwang C. S., Adv. Funct. Mater. 2019, 29, 1806278. [Google Scholar]
  • 49. Lee B., Fabrication and Characterization of Nanoscale Resistance Change Memory, Stanford University, Stanford, CA: 2012. [Google Scholar]
  • 50. Chiu F. C., Adv. Mate. Sci. Eng. 2014, 2014, 578168. [Google Scholar]
  • 51. Lee T. S., Lee N. J., Abbas H., Lee H. H., Yoon T.‐S., Kang C. J., ACS Appl. Electron. Mater. 2020, 2, 1154. [Google Scholar]
  • 52. Bid A., Bora A., Raychaudhuri A. K., Phys. Rev. B Condens. Matter Mater. Phys. 2006, 74, 035426. [Google Scholar]
  • 53. Zhao R., Gao Y., Guo Z., Su Y., Wang X., ACS Appl. Mater. Interfaces 2017, 9, 1885. [DOI] [PubMed] [Google Scholar]
  • 54. Lamarche R. M., Gasonoo A., Hoff A., Chernikov R., Welch G. C., Trudel S., Chem. Mater. 2023, 35, 2353. [Google Scholar]
  • 55. Ofuonye B., Lee J., Yan M., Sun C., Zuo J. M., Adesida I., Semicond. Sci. Technol. 2014, 29, 095005. [Google Scholar]
  • 56. Stössel M., Staudigel J., Steuber F., Simmerer J., Winnacker A., Appl. Phys. A: Mater. Sci. Process. 1999, 68, 387. [Google Scholar]
  • 57. Yang S. J., Dahan M. M., Levit O., Makal F., Peterson P., Alikpala J., Nibhanupudi S. T., Luth C. J., Banerjee S. K., Kim M., Roessler A., Yalon E., Akinwande D., Nano Lett. 2023, 23, 1152. [DOI] [PubMed] [Google Scholar]
  • 58. Jadidi I., Honarvar M. A., Khajeh‐Khalili F., Prog. Electromagn. Res. C 2019, 91, 241. [Google Scholar]
  • 59. Yang Q., Jiao Y. C., Zhang Z., IEEE Trans. Microw. Theory Tech. 2018, 66, 1926. [Google Scholar]
  • 60. Fan M., Song K., Fan Y., IEEE Trans. Circuits Syst. II Express Briefs. 2021, 68, 1758. [Google Scholar]
  • 61. Krishna N. V, Padmasine K. G., Mater. Sci. Semicond. Process. 2023, 154, 107181. [Google Scholar]
  • 62. Zhang N., Mei L., Wang C., Deng Z., Yang J., Guo Q., IEEE Trans. Electron Devices 2017, 64, 3377. [Google Scholar]
  • 63. Vryonides P., Nikolaou S., Kim S., Tentzeris M. M., Int. J. Microw. Wirel. Technol. 2015, 7, 655. [Google Scholar]
  • 64. Papapolymerou J., Lugo C., Zhiyong Z., Xiaoyan W., Hunt A., in IEEE MTT‐S Int. Microwave Symposium Digest, IEEE, San Francisco, CA: 2006, pp. 556‐559. [Google Scholar]
  • 65. Wang M., Lin F., Rais‐Zadeh M., in SiRF 2016 –2016 IEEE 16th Topical Meeting on Silicon Monolithic Integrated Circuits in RF Systems, IEEE, Austin, TX, 2016, pp. 38‐41. [Google Scholar]
  • 66. Basit A., Khattak M. I., Nebhen J., Jan A., Ahmad G., PLoS One 2021, 16, 0258386. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 67. Shim Y., Hummel G., Rais‐Zadeh M., in Proceedings of the IEEE Int. Conf. Micro Electro Mechanical Systems (MEMS), IEEE, Taipei, Taiwan: 2013, pp. 237‐240. [Google Scholar]
  • 68. Leon A., Reig B., Perret E., Podevin F., Saint‐Patrice D., Puyal V., Lugo‐Alvarez J., Ferrari P., IEEE Trans. Microwave Theory Tech. 2020, 68, 60. [Google Scholar]
  • 69. Kim M., Pallecchi E., Ge R., Wu X., Ducournau G., Lee J. C., Happy H., Akinwande D., Nat. Electron. 2020, 3, 479. [Google Scholar]
  • 70. Kim M., Ge R., Wu X., Lan X., Tice J., Lee J. C., Akinwande D., Nat. Commun. 2018, 9, 2524. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supporting Information

ADVS-12-e01989-s001.docx (6.5MB, docx)

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.


Articles from Advanced Science are provided here courtesy of Wiley

RESOURCES