Lithography-free, Pd-based bimorph cantilever switches for zero-standby-power chemo-mechanical H₂ detection

Abstract

Currently available hydrogen (H₂) detection methods are fundamentally inefficient because, although leaks are rare, continuous, power-consuming monitoring across large sensor networks is necessary owing to the flammable and explosive nature of H₂. A cost-effective fabrication strategy for chemo-mechanical switches that enables zero-standby-power H₂ detection is presented herein. The switches comprise a palladium (Pd)-based bimorph cantilever structure and are fabricated by tilted metal deposition on water-soluble electrospun nanofiber templates, without the need for conventional photolithography. This approach also permits the use of nonhazardous chemicals, thereby reducing the environmental burden and safety risks during fabrication. The Pd-based cantilevers function as chemo-mechanical H₂ switches; H₂ absorption in the Pd layer induces asymmetric expansion within the bimorph, bending the cantilever, thus closing the gap and establishing electrical contact. The chemo-mechanical H₂ switches are experimentally characterized for various fabrication parameters, and their H₂ detection performances are evaluated in terms of detection threshold, sensitivity, and transient response. The H₂ switches exhibit noise-level currents of ~2.2 pA in the absence of H₂, a low detection limit of 0.3% H₂, short response time of 37.2 s, and high sensitivity exceeding 1.35 × 10⁵.

Introduction

Hydrogen (H₂) is considered a clean alternative energy source and is used in many industrial sectors, including energy, chemical processing, and manufacturing^1–4. However, the colorless and odorless properties of H₂ make its leakage difficult to detect in industrial environments⁵. In addition, it is highly flammable and explosive at concentrations above 4% in air, implying that infrequent and small leaks can lead to catastrophic accidents^6,7. Therefore, the continuous and unattended monitoring of leakage is an essential prerequisite for ensuring safety in facilities that handle H₂.

H₂ sensing has been achieved using various transduction mechanisms, including electrochemical cells, thermal conductivity, catalytic combustion, optical detection, and chemo-resistive readout, which typically enables rapid electrical transduction through continuous resistance modulation^8–17. However, these conventional H₂ sensors operate in a continuously powered mode, drawing energy at all times regardless of whether a relevant event occurs. Despite significant progress in miniaturization and low-power operation, these sensors still continuously consume several microwatts of power, even in the absence of leakage events. This persistent power drain not only reduces energy efficiency but also increases manual maintenance requirements, such as battery replacements¹⁸. This issue becomes particularly critical in nonurban regions¹⁹, where both electrical infrastructure and manpower for maintenance are limited but H₂ is increasingly being considered as an alternative means of electricity generation to the conventional grid supply²⁰. Therefore, event-driven H₂ sensing capabilities, in which the sensor can remain inactive with near-zero standby-power consumption until awakened by H₂ leakage, are preferred for long-term operation under such constraints²¹.

Among the efforts to demonstrate such event-driven operations, chemo-mechanical approaches for H₂ detection have recently gained attention^22–27. These approaches typically rely on the large lattice expansion of palladium (Pd) upon H₂-selective absorption²⁸, which can intentionally induce mechanical deformation or displacement in thin films, wires, or 3D micro/nano structures with nanogaps. The expansion of Pd directly changes the mechanical contact, enabling chemo-resistive detection with high sensitivity where the contact resistance changes between cracks^26,27,29. However, such cracks are typically formed randomly under mechanical stress, making it difficult to precisely define the dimensional parameters and thus limiting reproducibility³⁰. Meanwhile, Pd-based chemo-mechanical switches offer an event-driven H₂ sensing approach, where conduction is turned on or off between nanogaps³¹. In particular, switch-type H₂ sensors exhibit a theoretically infinite on/off signal ratio once the H₂ concentration surpasses a certain threshold, and virtually no standby power is consumed until H₂ exposure occurs^30–36. This unique characteristic provides a promising method for highly sensitive and energy-efficient H₂ monitoring. Earlier approaches for switching operations also relied on randomly formed cracks or gaps induced by mechanical prestresses on Pd; however, the lack of control over their dimensions and locations often resulted in a nonuniform sensing performance in terms of sensitivity, repeatability, and reproducibility^30,33–36.

With the advent of microelectromechanical systems (MEMS) and nanoelectromechanical systems (NEMS) processes, the large-scale integration and improved reproducibility of Pd-based chemo-mechanical H₂ switches have been achieved^21,37,38. For example, Kiefer et al. reported large-array Pd-based chemo-mechanical switches with vertical nanogaps³⁷. The switches consist of a surface-micromachined Pd/titanium/polysilicon trimorph 3D cantilever and a fixed Pd electrode, offering complementary metal-oxide-semiconductor-compatible fabrication processes for chemo-mechanical H₂ switches. Nevertheless, their manufacturing processes are complex and costly, requiring multiple photolithography steps and hazardous chemicals, such as photoresists, developers, and hydrogen fluoride, to pattern and release 3D Pd-based micro/nano structures. This increases both the manufacturing complexity and environmental burden of Pd-based chemo-mechanical switches, contradicting the goal of employing hydrogen as a clean and sustainable energy source^39,40.

In this study, we present a simple manufacturing process for chemically actuated mechanical switches that enable event-driven H₂ detection, further improving our earlier proof-of-concept demonstration⁴¹. For the chemo-mechanical H₂ switches, Pd/Cr bimorph cantilevers were formed by the tilted deposition of Cr and Pd layers onto aligned electrospun polymer nanofibers. This fabrication approach eliminates photolithography-dependent structuring while enabling geometrically defined nanoscale cantilevers, electrode pads, and gaps. Furthermore, the fabrication process was designed to use eco-friendly and safe materials, including polyethylene glycol (PEO) nanofibers^42,43, deionized (DI) water, and isopropyl alcohol (IPA)^44,45 as the sacrificial pattern, dissolving medium, and releasing solvent, respectively. Upon exposure to H₂, only the Pd layer in the Pd/Cr bimorph cantilever selectively absorbs H₂ and expanded. This asymmetric expansion bends the cantilever downward, and once the deflection exceeds a critical threshold, an electrical contact is formed with the underlying fixed electrode by closing the nanogap, generating a conductive electrical path. However, below this threshold, the switch remains electrically and mechanically open, thereby consuming no standby power. Such event-driven functionality allows the sensors to eliminate the persistent power drain of conventional Pd-based chemo-resistive H₂ sensors, which rely on uninterrupted electrical bias to continuously track resistance changes. The sensor also exhibited a near-zero noise-level current (~4 pA), detection limit of 0.3% H₂, and high on/off current ratio of >1.35 × 10⁵, while showing no response to changes in humidity and other gas exposures and only minor signal variations with temperature, enabling selective H₂ detection with significantly low power consumption.

Results and discussion

Figure 1a shows a cross-sectional schematic of the bimorph switches used for chemo-mechanical H₂ detection. The switches consisted of Pd/Cr bilayered cantilevers, fixed lower electrodes, and a designated nanogap, as shown in the inset of Fig. 1a. To achieve chemo-mechanical H₂ switches, we developed a cost-effective nanofabrication process using electrospinning and physical vapor deposition without any photolithography steps, as shown in Fig. 1b. First, we deposited directionally aligned PEO nanofibers on a thermally oxidized silicon substrate through a far-field electrospinning process with electrostatic field-assisted alignment, as shown in Fig. 1b–i^46,47. For electrospinning, a 10 wt% PEO solution dissolved in DI water was prepared and loaded into a 10 mL syringe. The solution was injected at a flow rate of 0.2 µL/min through a needle with an inner diameter of 159 µm, while a voltage of 4.98 kV was applied between the needle and substrate at a distance of 88 mm. As illustrated in Fig. S1a, the substrate chip was positioned between two grounded auxiliary electrodes for electrostatic field-assisted alignment. During electrospinning, the charged PEO nanofibers were electrostatically attracted to the grounded electrodes, leading to the aligned deposition of the nanofibers on the substrate (Fig. S1b). The deposited PEO nanofibers had an average diameter of 275.1 nm, as shown in Fig. S1c. To verify the fiber directionality and process uniformity enabled by the grounded electrodes design, we additionally performed an image-processing-based statistical analysis, as provided in Fig. S2. The nanofibers exhibited a diameter standard deviation of 22.7 nm and a mean absolute angular deviation of 1.11°, confirming that the electrospinning process is well controlled under the proposed configuration.

Fig. 1. Structure, fabrication process, and switching mechanism of the Pd/Cr bimorph chemo-mechanical H₂ switch.

a Schematic of the Pd/Cr bimorph chemo-mechanical H₂ switch composed of Pd/Cr bilayer cantilevers and a predefined nanogap. b Lithography-free fabrication flow: (i) directional electrospinning of aligned sacrificial PEO nanofibers on thermally oxidized Si using field-assisted alignment; (ii, iii) tilt angle e-beam deposition of Cr (20 nm) and Pd (50 nm) at θ = 15°, 30°, 45°, and 60° to form bilayer cantilevers; (iv) additional normal-incidence Pd deposition (30/40/50 nm) to extend the free end for reliable contact upon bending; (v) release by immersion and bath sonication in DI water (3 h) to dissolve the sacrificial PEO and open nanogaps; (vi) IPA displacement (3 h) to lower surface tension and suppress stiction, followed by oven drying. c Operating principle (no H₂): Pd remains in the α-phase; the cantilevers stay unbent and the nanogap is open (I = 0). d Operating principle under H₂: Pd transforms to β-PdH_x (~3.5% lattice expansion), inducing bimorph bending, contact formation with the lower electrode, and the onset of current (I = I_Signal)

Subsequently, Cr films with a thickness of 20 nm and Pd films with a thickness of 50 nm were sequentially deposited on the aligned PEO nanofibers by tilted-angle deposition via electron-beam deposition to form cantilever structures, as shown in Fig. 1bii, iii, with tilt angles of 15°, 30°, 45°, and 60°. Additionally, Pd layers with thicknesses of 30, 40, and 50 nm were further deposited without tilting to enable contact between the cantilever and the fixed lower electrode when the cantilever was bent, as shown in Fig. 1biv. In the absence of this additional Pd deposition, the released cantilevers lacked the necessary length to establish contact between the cantilever and lower electrode. Consequently, even when bent downward by H₂ absorption, they failed to establish an electrical connection and thus could not operate as chemo-mechanical switches. The chips were then immersed in DI water and sonicated for 3 h to release the cantilevers by dissolving the sacrificial PEO nanofibers, as shown in Fig. 1bv. Immediately afterward, they were placed in an IPA bath for 3 h. This bath had a lower surface tension than that of DI water and served to replace the DI water, which prevented the stiction of the cantilever structures by reducing the surface tension, as shown in Fig. 1bvi⁴⁸. The chips were then dried in a convection oven at 60 °C to remove the residual solvent.

To demonstrate the cost-effective and ecofriendly fabrication of chemo-mechanical H₂ switches, we introduced a lithography-free process using aligned electrospun PEO nanofibers as sacrificial templates, followed by the tilted deposition of Cr/Pd layers. This approach eliminated the costly photolithography required in previous MEMS-based chemo-mechanical H₂ sensors while reliably defining nanoscale gaps with sufficient dimensional control for switch operation, compared to Pd crack-based H₂ sensors on polymer substrates. The process also relied only on processing materials, including PEO, DI water, and IPA, which are nonhazardous and easy to handle, minimizing risks to both the environment and the operators, compared to the hazardous, toxic, and volatile chemicals used in previous studies, such as photoresists, developers, and hydrogen fluoride^49,50.

Figure 1c, d schematically illustrate the operating principle of the chemo-mechanical H₂ switch. In the absence of H₂, the Pd layer remains in the α-phase and the cantilevers stay unbent, resulting in no current flow, as shown in Fig. 1c. Upon exposure to H₂, Pd transforms into the β-phase PdHₓ, which accompanies a lattice expansion of 3.5%³¹ (Fig. 1d). The associated volumetric strain induces the bending of the Pd/Cr bimorph cantilevers, leading to mechanical contact between the electrodes and the subsequent onset of current flow. This event-driven mechanism ensures zero standby power consumption in the off state while delivering a clear electrical signal only when H₂ is present. Moreover, to enhance reliability, we intentionally connected three cantilever switches in series within a single device, which mitigated the risk of false switching due to accidental gap closure or stiction.

Figure 2 presents the morphological characterization of the fabricated Pd/Cr bimorph cantilevers after removal of the sacrificial PEO nanofibers. We investigated the morphology of the device fabricated with an angular deposition angle of 45° and an additional Pd layer of 30 nm. Figure 2a shows an optical image of six sensors fabricated on an oxidized silicon substrate with dimensions of 9 × 9 mm². The actual footprint of each device is 2860 × 400 µm². A magnified scanning electron microscopy (SEM) image of the yellow-boxed region in Fig. 2a is shown in Fig. 2b, where the bimorph nanostructures are positioned within the three sky-blue boxes. In the top-view SEM image in Fig. 2c, cantilevers with nanoscale gaps are clearly observed, where the dimensions of the cantilever correspond to the diameter of the electrospun PEO nanofibers of 275.1 nm. In addition, sacrificial PEO nanofibers were continuously deposited across the chips. The cantilevers and adjacent lower electrodes were electrically isolated and measured optically and electrically using SEM and a multimeter, respectively. Cross-sectional SEM analysis further verified that the bimorph structure had a nanogap after release, as shown in Fig. 2d. The device was passivated with a Pt layer to suppress the electrostatic deformation under SEM observation. Figure 2e shows the energy-dispersive X-ray spectroscopy (EDS) mapping, which confirms the elemental composition of the structure and indicates distinct elemental boundaries within the Pd/Cr bilayer. Top-view EDS mapping is presented in Fig. S3 and confirms that bimorph nanostructures and gaps were formed, and PEO was successfully removed.

Fig. 2. Structural characterization of the fabricated Pd/Cr bimorph H₂ switches.

a Optical image of a 9 × 9 mm² oxidized-Si chip containing six Pd/Cr bimorph H₂ switches. b SEM of the yellow-boxed region in (a); three bimorph nanostructure sites are indicated by the sky-blue dashed boxes. c Top-view SEM image showing a suspended Pd/Cr cantilever and the nanogap formed after removing the sacrificial PEO nanofiber; the cantilever and adjacent film are electrically isolated. d Cross-sectional SEM of a representative device with Pt passivation (tilted deposition angle θ = 45°, additional Pd_add = 30 nm) confirming the nanogap and suspended nanostructure. e EDS elemental map (C, O, Si, Cr, Pd, Pt) of the cross section, showing distinct elemental boundaries in the Pd/Cr bilayer. f–h Cross-sectional SEM images for devices fabricated at θ = 60°, 30°, and 15° (all with Pd_add = 30 nm), demonstrating robust nanogap formation across angles. i Measured gap size versus deposition angle, increasing from ~26 nm at 15° to ~160 nm at 60°, attributed to the shadowing of the nanofiber template during tilted deposition

To analyze the influence of the deposition angle on the gap geometry, cross-sectional SEM analysis was performed on the samples deposited at 60°, 45°, 30°, and 15°. For this analysis, the nanofiber-deposited substrates were physically diced prior to dissolution in a DI water bath to minimize structural damage to the cantilevers during dicing for cross-sectional exposure. The SEM images in Fig. 2f–h clearly show that cantilevers with defined nanogaps were formed at deposition angles of 60°, 30°, and 15°, respectively. We also prepared a sample without tilt angle, but a Pd/Cr layer with a continuous thin film was formed without structuring the cantilevers, and this system did not operate as a chemo-mechanical switch, consistent with a previous study⁵¹. Figure 2i plots the measured gap size as a function of the deposition angle, showing that the gap dimensions decreased from 160 to 26 nm as the angle decreased from 60° to 15°. This result originates from the shadowing of the nanofiber template during the angled deposition because the nanofibers partially block the deposition of Cr and Pd; thus, one side is coated, while the opposite side remains uncovered, which defines the gap size between the cantilever and lower electrode. Consequently, the deposition with tilt angle defines the nanogap size and the corresponding H₂ detection threshold.

Figure 3 shows the schematics and experimental results of the fabricated Pd/Cr bimorph switches under H₂ exposure. As shown in Fig. 3a–d, the electromechanical behavior can be divided into distinct operation states depending on the H₂ concentration, which is consistent with that observed in previous Pd-based switch-type H₂ sensors^37,38. Specifically, without H₂ exposure, the Pd layer remained unexpanded, and the cantilever maintained its initial configuration without bending. The nanogap remained open, and the switch was disconnected both electrically and mechanically, as illustrated in Fig. 3a. When exposed to H₂, selective absorption in the Pd layer caused asymmetric lattice expansion relative to the Cr layer, which leads to downward bending of the cantilever, as shown in Fig. 3b. In this state, the gap is not closed, and no measurable current flows through the switch. As the H₂ concentration increased and reached a threshold value, the cantilever deflection became sufficiently large to close the nanogap and form a mechanical contact with the lower electrode, as illustrated in Fig. 3c. At this transition, the switch turned on and the current abruptly increased. At H₂ concentrations above the threshold, additional expansion of the Pd layer increased the contact force, enlarged the effective contact area, and reduced the interfacial resistance, as shown in Fig. 3d. As a result, the current increased further compared to the threshold state, allowing the switch to not only detect H₂ leakage at a defined threshold concentration but also to distinguish higher concentrations by the magnitude of the current response. To verify full nanogap closure in the on-state, we fabricated a control device without cantilevers, consisting of a continuous Cr (20 nm)/Pd (90 nm) thin film patterned using the same geometry. As shown in Fig. S4, the continuous thin film exhibited higher currents both with and without 2% H₂ under the same bias voltage than the device with fully actuated bimorph cantilever switches. This difference originates from the absence of contact resistance and the shorter planar conduction path in the continuous film, whereas the switch inherently includes mechanically formed Pd-Pd contact resistance and a three-dimensional current path. These results confirm that the cantilevers fully close the nanogap upon actuation, and that the reduced on-state current is due to intrinsic contact and geometric effects rather than incomplete deflection.

Fig. 3. Chemo-mechanical operation and H₂ response of Pd/Cr bimorph switches.

Operation states: a ambient air, initial nanogap is open; b sub-threshold H₂, slight bending and a narrowed gap; c at threshold, gap closure and electrical turn-on; d above threshold, increased contact force enlarging the effective contact area and lowering interfacial resistance. e–h Step-response measurements at V_bias = 0.2 mV with H₂ steps of 0.1, 0.3, 0.5, 1.0, 1.5, 2.0, and 4.0% (1000 s exposure/2000 s of purging in air). The baseline noise of ~4 pA indicates near-zero standby power. The devices with tilted deposition θ = 60°, 45°, 30°, and 15° and Pd_add= 30 nm show angle-dependent thresholds consistent with the initial gap size: 60°: no response across all concentrations; 45°: threshold of 1.0%; 30°: threshold of 0.5%; 15°: threshold of 0.3%. i Device fabricated at θ = 15° with Pd_add = 40 nm maintained a 0.3% threshold but yielded a larger response at the threshold due to a reduced initial gap. j Linear I–V characteristics of the θ = 15°, Pd_add = 30 nm device under 2% H₂, confirming an ohmic contact after gap closure

Figure 3e–h shows the chemo-mechanical responses of the fabricated Pd/Cr bimorph switches deposited at angles of 60°, 45°, 30°, and 15°, all with an additional Pd layer of 30 nm. The H₂ detection limit is defined as the lowest H₂ concentration that produces a measurable response with a signal-to-noise ratio (SNR) higher than 3⁵². The measurements were performed under varying H₂ concentrations of 0.1, 0.3, 0.5, 1.0, 1.5, 2.0 and 4.0% at a bias voltage of 0.2 mV, with each step consisting of 1000 s of exposure followed by 2000 s of recovery, as shown in Fig. S5. When the H₂ concentration was below the threshold or absent, the current remained at a noise level of approximately 4 pA, implying a near-zero standby power consumption. Once the concentration exceeded the threshold of each device, the current increased abruptly as the cantilever deflected sufficiently to close the nanogap and form contact with the lower electrode. We also experimentally confirmed that the threshold concentration depends on the deposition angle and the resulting nanogap size. The device with a 60° deposition angle exhibited no response across all H₂ concentrations owing to the relatively large initial nanogap, as displayed in Fig. 3e. At 45°, the device exhibited the highest threshold of 1% (Fig. 3f). At 30°, the threshold concentration decreased to 0.5% (Fig. 3g), consistent with the smaller nanogap requiring smaller Pd expansion for closure. At 15°, the lowest threshold of 0.3% H₂ was obtained, as shown in Fig. 3h, because the small nanogap allowed the cantilever to form contact with only a slight deflection at low H₂ concentrations. Figure 3i shows the results for the 15° device with an additional 40-nm-thick Pd layer. As shown in Fig. S6, a finer evaluation with 0.05% H₂ increments confirms that the detection limit of the 15° tilt angle device with an additional 40 nm Pd layer remains 0.3%, consistent with the device with 15° tilt angle and an additional 30 nm Pd layer, while the reduced initial gap enables a larger signal at the threshold H₂ concentration. With the 15° tilt angle and an additional 40 nm Pd layer device, we also examined the switching response versus bias voltage from 0.2 to 5 mV to assess contact reliability, as presented in Fig. S7. The switching behavior remained unchanged for bias voltages from 0.2 to 1 mV, whereas a 5 mV-induced irreversible contact modification and a drastic reduction of the on/off ratio. This degradation is attributed to localized current, Joule heating-based micro-welding, and metal transfer⁵³. Accordingly, all switching characterization in this work used a bias voltage of 0.2 mV to ensure reliable contact operation and to avoid bias-induced degradation.

Additionally, Fig. 3j shows the current versus voltage curve of the 15° device with an additional 30-nm-thick Pd layer under 2% H₂ exposure, and we experimentally confirmed that the chemo-mechanically induced contact interface exhibits ohmic behavior. Although an apparent off state leakage current of about 4 pA was observed under the sourcemeter readout, we further evaluated the baseline current using a system electrometer, as summarized in Fig. S8. The electrometer measurement shows a reduced and more stable baseline current of 2.2 pA, suggesting that the measured leakage is largely limited by the readout instrument. This result indicates that even lower standby power and a higher on-off ratio are potentially achievable by improving the measurement electronics.

Figure 4a summarizes the current as a function of the H₂ concentration for devices with different deposition angles, including variants with additional Pd layers, and Fig. 4b shows the current on/off ratio, which represents the sensitivity of the switches. The switches exhibited highly sensitive chemo-mechanical responses with an on/off current ratio exceeding 1.35 × 10⁵ because the current remains near zero in the open state and increases sharply upon gap closure. Beyond the threshold, the current increased further because of the enlarged contact area and reduced interfacial resistance, enabling the discrimination of H₂ concentrations above the threshold. However, at higher H₂ concentrations, the current eventually saturates once the effective contact area is fully engaged, and an additional contact force cannot further reduce the resistance⁵⁴. Figure 4c shows the detection thresholds of H₂ concentrations across the different devices. The threshold concentration increased with nanogap size, confirming that a larger gap requires greater Pd expansion and cantilever deflection for closure. Figure 4d, e show the H₂-exposure responses of the 15° device with an additional 30-nm-thick Pd layer under different relative humidities (RHs) and temperatures, respectively. The device was tested at its threshold concentration (0.3%) and at 2% H₂, where saturated responses were observed, under dry and 80% RH conditions and at 25 and 80 °C. Figure 4f shows the response changes with respect to the humidity and temperature. When the relative humidity increased from dry to 80% RH, the current decreased only marginally by 3.26% at 0.3% H₂ and 0.42% at 2% H₂. This occurred because water molecules in air do not affect the Pd lattice expansion underlying the chemo-mechanical actuation⁵⁵. By contrast, when the temperature increased from 25 to 80 °C, the response decreased by 12.16% at 0.3% H₂ and 11.76% at 2% H₂, resulting from the positive temperature coefficients of resistance in the Pd and Cr bimorph layers. Nevertheless, a detection limit of 0.3% was maintained because the thermal volume expansion of the bimorph was negligible compared with the hydrogen-induced volume expansion of Pd. Considering this environmental stability, the use of Pd and a mechanical switching mechanism for H₂ detection provided the sensors with H₂ selectivity capabilities with no detectable responses to other gases, as shown in Fig. S9. While the present sensors show short-term selectivity against representative interfering gases (Fig. S9), it is well known that Pd-based hydrogen sensors can suffer from surface contamination or poisoning under prolonged exposure to strongly adsorbing species such as CO, NO_X^56–58. Previous studies have reported that such effects can be mitigated by employing selective filtration or protective coating strategies, for example, polymer or silica membranes that preferentially allow H₂ permeation while blocking interfering gases^59,60. These approaches can be readily integrated with the present chemo-mechanical switching platform for long-term practical deployment. We compared the sensing characteristics of our chemo-mechanical switch with previously reported chemo-mechanical H₂ sensors to evaluate its relative performance. Figure S10 plots the sensor sensitivity at 2% H₂ as a function of the limit of detection for the reported chemo-mechanical sensors. Table S1 further summarizes the working principles, fabrication methods, dimensional controllability, detection ranges, and sensitivities of the reported devices. They indicate the improved sensitivity and lower limit of detection of our device compared with existing chemo-mechanical H₂ sensors.

Fig. 4. H₂ sensing characteristics and environmental robustness of the Pd/Cr bimorph chemo-mechanical switch.

H₂ sensing characteristics and environmental robustness of the Pd/Cr bimorph chemo-mechanical switch. a Current versus H₂ concentration for devices with different tilted-deposition angles and additional Pd thicknesses; V_bias = 0.2 mV. b Current on/off ratio extracted from (a), exceeding 1.35 × 10⁵ after gap closure. c Detection thresholds summarized for each device; the threshold increases with the initial nanogap size set by the deposition angle. d, e Environmental robustness of the θ = 15° device with Pd_add = 30 nm measured at its threshold concentration (0.3% H₂) and at 2% H₂ under controlled humidity (dry and 80% RH at 25 °C) and temperature (25 and 80 °C at RH = 0%). f Compiled response changes with respect to RH and temperature. The RH caused only marginal changes because water does not affect Pd lattice expansion, whereas increased temperature reduced the current via the positive temperature coefficient of resistance (TCR) of the metallic Pd/Cr; the 0.3% detection limit is maintained

To evaluate the repeatability of the fabricated H₂ switches, cyclic sensing tests were performed at a 2% H₂ concentration. Figure 5a–d shows the transient responses of the devices with an additional 30-nm-thick Pd layer and deposition angles of 60°, 30°, 45°, and 15°, each subjected to five repeated cycles of H₂ exposure for 1000 s followed by 2000 s of recovery. The switch with a deposition angle of 60°, which had the largest nanogap (160 nm), exhibited poor repeatability, as shown in Fig. 5a. After three cycles, no measurable current was detected even under the 2% H₂ exposure. By contrast, the switches fabricated at 45°, 30°, and 15° maintained a stable operation over all five cycles, confirming reliable contact switching under repeated H₂ exposure, as shown in Fig. 5b–d. Based on these results, we further developed the proposed fabrication process to reduce the gap by increasing the Pd deposition thickness from 30 to 40 and 50 nm in the 15° switch.

Fig. 5. Repeatability of Pd/Cr chemo-mechanical H₂.

a–d Transient current responses at V_bias= 0.2 mV for devices with Pd_add = 30 nm and tilted deposition θ = 60°, 45°, 30°, and 15°; five cycles of 2% H₂ steps (1000 s of H₂ exposure/2000 s of air purging). a Device fabricated at θ = 60° with the largest initial gap (∼160 nm) lost conduction after three cycles. b–d Devices at θ = 45°, 30°, and 15° maintained stable turn-on/turn-off performances over all cycles. The baseline noise is ∼4 pA, indicating near-zero standby power. e Device fabricated at θ = 15°, Pd_add = 40 nm is pre-closed after fabrication; initial H₂ exposure increased the current and Pd-induced stress then reopened the gap, after which stable switching was observed. f θ = 15°, Pd_add = 50 nm: permanent stiction keeps the contact closed and the switching capability is lost. g Average saturated current at 2% H₂ over the last three stable cycles: devices that fully close under H₂ show similar currents at Pd_add = 30 nm; increasing Pd_add increases the on-state current via an enlarged current pathway. h Average on/off ratio over the last three cycles: >10⁵ for stable switching devices; ∼1 for the pre-closed Pd_add = 50 nm case. i Response and recovery times extracted with the 1–1/e criterion (response: time to 63.2% of the steady-state value; recovery: decay to 36.8%) for Pd_add= 30 nm

Figure 5e, f show the transient responses of the 15° switches with 40- and 50-nm-thick Pd layers, respectively, under five repeated cycles of 2% H₂ exposure. The devices were confirmed to be in a pre-closed-gap state immediately after fabrication, as indicated by their current response under initial H₂ exposure. In typical continuous Pd films, H₂ absorption reduces the current because of the higher resistivity of PdH_x. By contrast, our devices initially showed an increase in the current upon H₂ exposure, indicating that the observed response originated from Pd expansion and chemo-mechanical resistive behavior rather than from thin-film conduction. These results indicate that the cantilevers and corresponding gaps were formed, but the stress generated during drying induced the closure of the gaps in both devices. Specifically, for the 15° switch with a 40-nm-thick Pd layer, the switches initially exhibited a closed gap state in the absence of H₂, as shown in Fig. 5e. The current initially increased upon H₂ exposure and then decreased sharply to the noise level after H₂ was turned off, confirming that the Pd-induced mechanical stress was sufficient to separate the pre-closed contact interface. This device subsequently exhibited fast and flat transient responses with stable repeatability. However, the 50-nm-thick Pd-layer devices exhibited permanent stiction, where the gap was closed immediately after fabrication, as shown in Fig. 5f. Even under subsequent H₂ exposure, the stress generated by the Pd expansion did not reopen the contact, indicating irreversible adhesion of the interface. We further evaluated device-to-device consistency across independent fabrication batches for devices with 15° tilt angle and a 30 nm additional Pd layer. We prepared five samples in total (n = 5), one device from each of five independent batches and tested their switching responses under five repeated exposures to 2% H₂, as shown in Fig. S11. Across the five devices, the on-switching current at 2% H₂ exhibited an average value of 0.445 ± 0.026 μA. The corresponding on/off current ratio showed a mean value of ~1.11 × 10⁵ for all batches, with the relative variation of 5.8%, while maintaining current on/off ratio on the order of 10⁵. All devices exhibited stable and repeatable switching responses under repeated exposures to 2% H₂, with comparable switching current and on/off ratio distributions. Using the same device configuration (15° tilt angle with an additional 30 nm Pd layer), we additionally assessed long-term switching stability, as shown in Fig. S12. The device was subjected to 50 consecutive H₂ exposure cycles at both 0.3% H₂ and 2% H₂, using a 300 s H₂ exposure followed by a 600 s recovery in air for each cycle. At the threshold concentration of 0.3% H₂, the device showed considerable response in all 50 cycles, although the on-state current exhibited cycle-to-cycle deviation due to marginal contact formation, as presented in Fig. S12a. At 2% H₂, the device maintained stable and repeatable switching operation throughout the repeated actuation (Fig. S12b). These results confirm robust long-term switching stability under both low- and high-concentration conditions.

The average saturated currents over the last three stable cycles at 2% H₂ are shown in Fig. 5g. When the additional Pd layer thickness was fixed at 30 nm, the devices that fully closed under 2% H₂ (45°, 30°, and 15°) exhibited similar current levels during H₂ exposure. With an increase in the Pd layer thickness, the fully closed state yielded a higher current, which was attributed to an enlarged current path. Figure 5h shows the average on/off current ratio over the final three cycles. Unlike the other sensors that exhibited stable switching with on/off ratios exceeding 10⁵, the 15° sensor with a 50-nm-thick Pd layer, which lacked an initial gap, exhibited a very low on/off ratio of ~1. Figure 5i shows the response and recovery times extracted from the three cycles for the four samples, which showed stable and repeatable switching responses at 2% H₂. Both parameters were defined using the 1–1/e time constant, where the response time corresponds to the duration required for the output to reach 63.2% of its steady-state value and the recovery time corresponds to the time for the output to decay to 36.8% of that value. This definition, which is commonly adopted in general mechanical systems, was also applied to our devices to ensure consistency^61,62. With an additional Pd layer of 30 nm, the switches exhibited response times of 231.00, 102.60, and 97.73 s at deposition angles of 45°, 30°, and 15°, respectively, while the corresponding recovery times were 35.46, 71.62, and 87.63 s. For the devices with a 40-nm-thick Pd layer at a deposition angle of 15°, the response and recovery times were 37.46 and 93.2 s, respectively. In summary, smaller gaps yielded faster responses because a shorter expansion distance was required for contact, reaching 37.46 s for the 15° device with a 40-nm-thick Pd layer. Conversely, larger initial gaps enabled faster opening of the contact during recovery, resulting in shorter recovery times. This behavior is attributed to the stronger restoring force of the cantilever caused by a larger deflection required for contact, which corresponds to the gap, resulting in faster reopening of the contact interface. The overall results indicate that the sensing characteristics depend strongly on the structural parameters, including the deposition angle (15–60°) and the additional Pd layer thickness (30–50 nm). Among the tested structural parameters, the 15° switch with a 40-nm-thick Pd layer exhibited the desired sensing characteristics, including fast response, low limit of detection (0.3%), high sensitivity, and stable repeatability. As a result, the 15° device with a 40-nm Pd can be employed as H₂ switches with improved sensing performance owing to their smaller effective gaps.

Conclusion

We developed chemo-mechanical H₂ switches based on Pd/Cr bimorph cantilevers fabricated via a lithography-free process using electrospun nanofibers and tilted deposition. This process eliminated the need for photolithography, enabling the simple fabrication of nanoscale cantilevered switches. The Pd/Cr bimorph cantilevers exhibited event-driven switching operations, remaining electrically open with zero standby power in the absence of H₂ and closing upon H₂-induced Pd expansion. The threshold concentration was determined by the deposition angle and the resulting nanogap size, with smaller gaps enabling lower threshold values. The fabricated switches demonstrated a detection limit of 0.3% H₂, stable cycling performance, and high on/off current ratios exceeding 1.35 × 10⁵. These results confirm that the proposed fabrication strategy enables simple, ecofriendly, and reliable chemo-mechanical H₂ switches with zero standby power consumption.

Materials and methods

Directional electrospinning process

PEO (average molecular weight of 300,000; Sigma-Aldrich) powder was dissolved in DI water and magnetically stirred at 150 rpm for 12 h to prepare a 10% PEO solution. A schematic of the electrospinning process is shown in Fig. S1. The PEO solution was loaded into a 10 mL syringe, and a metal hub needle (30 gauge, Hamilton) was mounted at the syringe outlet. The syringe was then placed on a syringe pump (KDS Legato 210, KD Scientific) and electrospun at a flow rate of 0.2 µL/min. For directional electrospinning, two grounded electrodes were spaced 25 mm apart, and the sample was positioned between them. The distance between the needle tip and grounded electrodes was 88 mm, and a voltage of 4.98 kV was applied between the needle and grounded electrodes using a high-voltage generator (Korea Switching). Electrospinning was performed for 2 s.

Fabrication of switches for H₂ detection

The switches were fabricated on silicon substrates with a 300 nm wet-oxidized SiO₂ layer and saw-diced into 9 × 9 mm² chips. After directional electrospinning, a stainless-steel shadow mask was placed on each sample, and Cr (20 nm) and Pd (50 nm) were sequentially deposited via tilted e-beam evaporation (KVE-4000, Korea Vacuum Tech). After returning to normal incidence (0° tilt), an additional Pd layer was deposited by e-beam evaporation. Following the removal of the shadow mask, the samples were immersed in DI water, sonicated in a bath sonicator (Elmasonic P, Elma Ultrasonic) for 3 h, and immersed in IPA for 3 h. The chips were subsequently dried in a convection oven (OF-02GW, Jeiotech) at 60 °C for 6 h.

Morphological characterization of the Pd/Cr bimorph nanostructures

Surface morphology and elemental maps were acquired using SEM-EDS (IT-500, Jeol). For cross-sectional EDS, a 200 nm Pt passivation layer was deposited using sputtering (Sputter coater, Cressington).

Evaluation of H₂ sensing performance

A schematic of the air-diluted H₂ gas-sensing setup is depicted in Fig. S5. The total flow was maintained at 500 sccm throughout both the H₂ exposure and recovery phase, with exposure and recovery durations fixed at 1000 and 2000 s, respectively. During the recovery phase, air purging was performed while keeping the total flow rate constant. The mixed gas was passed through a quartz tube to expose the sample, and the switching current was recorded with a source meter (2614B, Keithley) under a 0.2 mV bias. Temperature-dependent responses were assessed via a quartz tube housed in a temperature-controlled furnace. The baseline current measurements at a lower noise level were performed using an electrometer (Keithley 6514).

Author contributions

D.K. and E.J. designed, fabricated, and tested the switches and wrote the manuscript. Prof. J.K. guided the device design, fabrication, and experiments, revised the manuscript, and reviewed it prior to submission. All authors contributed to the work and approved the submitted version.

Data availability

All data supporting the findings of this study are contained in the article and the Supplementary Information. Additional relevant data are available from the corresponding author upon request.

Conflict of interest

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41378-026-01269-2.