Single-Crystal Pt-Decorated WO3 Ultrathin Films: A Platform for Sub-ppm Hydrogen Sensing at Room Temperature.

Hydrogen-related technologies are rapidly developing, driven by the necessity of efficient and high-density energy storage. This poses new challenges to the detection of dangerous gases, in particular the realization of cheap, sensitive, and fast hydrogen sensors. Several materials are being studied for this application, but most present critical bottlenecks, such as high operational temperature, low sensitivity, slow response time, and/or complex fabrication procedures. Here, we demonstrate that WO3 in the form of single-crystal, ultrathin films with a Pt catalyst allows high-performance sensing of H2 gas at room temperature. Thanks to the high electrical resistance in the pristine state, this material is able to detect hydrogen concentrations down to 1 ppm near room temperature. Moreover, the high surface-to-volume ratio of WO3 ultrathin films determines fast sensor response and recovery, with characteristic times as low as 1 s when the concentration exceeds 100 ppm. By modeling the hydrogen (de)intercalation dynamics with a kinetic model, we extract the energy barriers of the relevant processes and relate the doping mechanism to the formation of oxygen vacancies. Our results reveal the potential of single-crystal WO3 ultrathin films toward the development of sub-ppm hydrogen detectors working at room temperature.

Introduction

Transition metal oxides are a class of materials whose properties can be tuned by several external parameters, such as heteroepitaxial strain,[1] electrostatic doping,[2,3] intercalation of other species,[4] or oxygen vacancies.[5−7] This versatility recently triggered a broad technological interest, which goes from their use in oxide-based fuel cells,[8,9] memristive systems,[10−12] neuromorphic architectures,[13,14] chemical actuators,[15] and gas sensors.[16] In this framework, oxide compounds are of particular interest for building high-performance hydrogen sensors, motivated by the growing demand for using H2 gas to transport and store energy and to reduce the safety risks imposed by this explosive gas.[17,18] Several oxide materials have been employed for resistive hydrogen sensing, with TiO2, SnO2, VO2, and WO3 being among the most prominent ones.[19−22] WO3, in particular, is regarded as a very promising candidate because incorporation of hydrogen changes its electrical resistance by several orders of magnitude.[23,24] As summarized in Table , hydrogen sensing has been demonstrated with various forms of WO3, such as amorphous layers and nanostructured materials.[25−30] These sensors, however, usually require a high operational temperature, resulting in high power consumption and poor long-term stability, and have a sensitivity limited to hydrogen concentrations of hundreds of ppm.

Table 1

Comparison of WO3-Based Technologies for H2 Sensing

technology	operational temp (°C)	lower detection limit (ppm)	response time τ (s)
electrodeposited 200 nm films[26]	25	50000	42 (at 25 °C, 50000 ppm)
sputtered 160 nm films[25]	30–300	1250	–
nanoplates[30]	25	1000	24 (at 25 °C, 1000 ppm)
nanowires[28]	30–250	100	300 (at 250 °C, 1000 ppm)
nanocomposites[29]	180–240	50	26 (at 240 °C, 200 ppm)
nanodots[27]	25–350	20	100 (at 150 °C, 1000 ppm)
single-crystal 9 nm ultrathin films (this work)	25–65	1	1 (at 65 °C, 100 ppm)

In this work, we show that single-crystal WO3 ultrathin films with a Pt catalyst can detect H2 concentrations down to 1 ppm, with a fast response time, while working at or nearby room temperature. We show that hydrogen intercalation in the WO3 crystal lattice determines changes of more than 5 orders of magnitude in the electrical resistivity of the material and that this fully reversible process is mediated by a small amount of platinum catalyst. Response time and lower detection limit can be tuned by temperature, which are reduced by an order of magnitude at a slightly elevated temperature of 65 °C. We study the hydrogen dynamics by modeling the intercalation and deintercalation processes with a series of coupled kinetic equations and extract the characteristic energy scales related to H2 dissociation, incorporation, and recombination phenomena. We also demonstrate that hydrogen doping is a powerful control parameter to modulate the physical state of WO3, which can be stabilized in kinetically arrested configurations just below room temperature. Our results show that single-crystal WO3 ultrathin films are strong candidates for next-generation hydrogen sensing devices, combining high sensitivity, fast response, and device scalability.

Results and Discussion

WO3 Material Preparation

We grow a 24 unit cell (uc) crystalline WO3 film on a SrTiO3 (001) substrate by pulsed laser deposition (growth details in the Experimental Section and ref (31)). The WO3 structural properties are investigated by X-ray diffraction (XRD) and presented in Figure a, where finite size oscillations indicate high crystalline quality and confirm the expected film thickness of 9 nm. The reciprocal space map in Figure b shows that the material is a single crystal, coherently oriented to the substrate lattice. Surface topography, investigated by atomic force microscopy (AFM) in Figure c, shows a step and terrace structure which mimics the underlying substrate morphology. The hydrogen reaction is made possible by the presence of a catalyst, whose function is to facilitate the splitting of H2 molecules from gas phase into atomic hydrogen, which is absorbed into the WO3 lattice. In this work, we use platinum, which has the advantage of a high catalytic efficiency together with a low reactivity with oxide materials.[32,33] A Pt layer having nominal thickness of 1 nm is deposited at room temperature by evaporation in vacuum (Figure d). Deposition of this low amount of material is expected to form sparse droplets of a few nanometers, and their catalytic activity is confirmed by the data discussed in the following section. Electrical measurements are performed on 50 μm × 50 μm patterned squares measured in van der Pauw configuration, as shown in the optical image of Figure e (details in the Experimental Section). The experimental setup is sketched in Figure f. We mount the sample in a vacuum chamber (base pressure p < 1 × 10–4 mbar), which is placed in a dark environment. Mixtures of H2 in Ar carrying gas, with concentration CH ranging from 20% down to 1 ppm, are prepared in a series of preparation chambers before exposing the sample. The system also allows to regulate the total pressure of the gas mixture from 1 bar down to 1 mbar.

Figure 1

Characterization and preparation of WO3 thin films. (a) θ–2θ X-ray diffraction scan showing the (001) and (002) peaks of the WO3 thin film with finite size oscillations around the sharp peaks of the SrTiO3 substrate. (b) Reciprocal space map around the (103) peak. (c) Atomic force microscopy of the WO3 surface. (d) Schematic of the WO3/SrTiO3 heterostructure after the evaporation of the Pt catalyst. (e) Optical image of the WO3 sample patterned in van der Pauw geometry (the image is taken before the removal of the resist for better contrast). (f) Schematic of the experimental setup comprising the sample anchored on a Peltier element (red), the vacuum chamber SC1, the four-probe configuration used for resistive readout, the series of preparation chambers (PC1, PC2), gas bottles, pressure sensors, and vacuum pumps used to prepare the gas mixture (details in the Experimental Section).

WO3 Response to Hydrogen

The high crystal quality of the WO3 material determines a high electrical resistance of the pristine device R = 400 MΩ at room temperature. This corresponds to a resistivity of 360 Ω cm, consistent with previous reports of high quality WO3 thin films with negligible amount of oxygen vacancies.[31] In Figure we show that upon exposure to a high hydrogen concentration (CH = 5%, 1 bar total pressure) at room temperature the resistance drops by 5 orders of magnitude, and after a few minutes it saturates to a constant value. When the sample chamber is evacuated, the resistance stays approximately constant, indicating that the doped state is stable over time. When the doped device is exposed to air, the electrical resistance increases over time, recovering more than 3 orders of magnitude in the first 10 min and reaching the initial state after several hours.

Figure 2

Room temperature electrical and optical changes with hydrogen gas. The WO3 thin film is exposed to different environmental conditions: 1 bar of air, 1 bar of hydrogen mixture, and vacuum. Both electrical resistance and sample color (optical images) change upon exposure to H2 and recover their initial state once in air (see also the video in the Supporting Information).

During the electrical measurement, we took a series of photographs of an unpatterned WO3 region which is presented in Figure at the corresponding time. The material color changes, switching to blue upon the introduction of the gas (doped state), and going back to the initial transparency when the resistance progressively recovers its original value (see also the video in the Supporting Information). This color change is a well-known effect of hydrogen doping of WO3, consistent with previous reports.[26,34,35] Both the resistance and optical data indicate that the measured changes in the sample properties are due to hydrogen intercalation and deintercalation into the WO3 lattice.[36] Moreover, the data show that hydrogen doping is a reversible process and that the deintercalation is facilitated by the presence of air.

Although a precise description of the doping mechanism is still a matter of debate, several reports show that the incorporation of atomic hydrogen determines electron doping of WO3.[24,37,38] The Pt catalyst plays a primary role in the intercalation process, as without it the resistance of WO3 remains constant upon exposure to H2 (Figure S1). Increasing (decreasing) the amount of Pt catalyst, the material response to hydrogen can be made faster (slower), as discussed in Figure S2. In our experiments, the hydrogen is diluted in a carrier gas. We tested different inert gases (Ar, N2, and He) and observed similar responses, indicating they have a negligible contribution in the intercalation process (Figure S3). While still showing a good response, a slower intercalation rate is observed when the carrier gas is air. We attribute this difference to the presence of O2 molecules that can react with H atoms to form water.[36] The intercalation rate is also reduced when no carrier gas is used, even if the H2 partial pressure is the same. The rate of the deintercalation process is quite slow both in a vacuum and in an inert gas atmosphere, while it is significantly enhanced in the presence of O2 molecules (further details in Figure S4).

Hydrogen Sensing

The strong and fast response of single-crystal WO3 ultrathin films to H2 makes them appealing candidates to develop hydrogen gas sensors. For this purpose, we characterize in Figure a the resistive changes of WO3 for different values of CH. The measurements are performed with the following experimental procedure: (i) evacuation of the sample chamber (p < 1 × 10–4 mbar), (ii) exposure to the H2/Ar mixture at 1 bar total pressure for 1 min, (iii) chamber evacuation, and (iv) exposure to air at 1 bar for 10 min. This procedure is repeated for different hydrogen concentrations between 1 and 4000 ppm and for two sample temperatures, namely 25 and 65 °C. We note that steps (i) and (iii) are used to remove H2 and O2 from the gas chambers, respectively, thus guaranteeing high purity of the mixtures used in the subsequent steps (ii) and (iv). The resistance drop ΔR during step (ii) becomes progressively larger with increasing CH, and it is almost independent of CH at high hydrogen concentration, indicating a saturation regime. In the vacuum step (iii), a small increase of resistance is observed, and it is more pronounced at 65 °C. Upon exposure to air in step (iv), the resistance recovers toward the initial value. We note that due to the short duration of this step, the resistance recovery at the highest values of CH is not complete. However, as discussed below, this does not influence the sensor operation. Raising the temperature determines a faster response and higher sensitivity, with an overall similar sample behavior.

Figure 3

Resistive sensing of different hydrogen concentrations. (a) Resistance variation upon 1 min of exposure to different concentrations of H2 gas in Ar (1 bar total pressure). The experimental steps consist of (i) sample chamber evacuation (p < 1 × 10–4 mbar), (ii) exposure to the H2/Ar mixture for 1 min, (iii) chamber evacuation, and (iv) exposure to air for 10 min. The curves are measured at a sample temperature of 25 °C (solid blue) and 65 °C (dashed red). (b) Percentage resistive variation at 1 s after exposure to H2 and (c) corresponding response time. (d) Recovery time upon exposure to air. The dotted lines are linear fits to the solid markers, while the grayed-out data points are either below the sensor detection limit (low ppm) or in its saturation regime (high ppm).

We can quantitatively evaluate the performance of WO3 as an H2 sensor by considering the relative resistance change ΔR/R0 at 1 s after the introduction of H2 gas. This analysis is reported in Figure b, where we show that at intermediate concentrations (solid markers) the response depends on CH according to the power law ΔR/R0 ∼ (CH)α. We find α = 1.3 at 25 °C and α = 1.5 at 65 °C, indicating that a similar relationship regulates the trend at both temperatures. The onset of the saturated response of the WO3 device (shaded markers in Figure b) depends on its operational temperature. At 25 °C the response follows the power law in the range 10–1000 ppm, while at 65 °C this window is shifted to lower concentrations in the range 1–100 ppm. This indicates that by increasing the sensor temperature by a few tens of degrees, it is possible to lower the detection limit by 1 order of magnitude or, in other words, that the sensitivity range can be tuned by controlling the sample temperature.

To evaluate the response time of WO3, we consider a simplified model describing the time-dependent resistance change as ΔR/R0 ∼ e–, where the response time τ is the fitting parameter. We find a lower τ for higher hydrogen concentration and sample temperature in step (ii), reaching τ = 1 s for CH = 100 ppm at 65 °C (Figure c). We note that independently of the value of τ and in the whole range of explored parameters, the presence of hydrogen is always detectable within the 1 min time frame used in the experiments. Figure d shows the recovery time of doped WO3 after exposing the device to air in step (iv). Again, lower τ is measured at higher temperature and higher initial hydrogen concentration in WO3. Finally, we note that the sensor response is independent of the initial resistance value R0 (Figure S5), meaning that in the explored range of parameters WO3 works as a differential sensor, with a response depending only on CH and not on the doping state of the material. In addition, the sensor preserves its functionality in a low-pressure H2 atmosphere, demonstrating its operability also as possible detector for pure H2 gas (Figure S6). These characteristics make WO3 an interesting material for H2 sensing that can be used in a large variety of environments.

Kinetic Model

To obtain an insight into the mechanism regulating the doping of WO3 thin films, we study the intercalation and deintercalation kinetics in a controlled atmosphere. For this purpose, we consider the minimal model of two-step reactions schematically illustrated in Figure a, where each step is described by a rate constant ka, and its reverse by k–a. We describe the intercalation and deintercalation processes by the balance equationswhere with H(Pt) we indicate a hydrogen atom on a catalyst site, with H(surf.) anywhere on WO3 or Pt surface, and with H(bulk) intercalated in the WO3 lattice. Our model describes the intercalation and deintercalation considering different initial conditions. The intercalation experiment starts with undoped WO3 and an infinite reservoir of H2 in gas phase, allowing to consider the dissociative adsorption (k1,i) of H2 on Pt as an irreversible process. The subsequent hydrogen exchange between WO3 surface and WO3 bulk is an equilibrium reaction, regulated by rate constants k2,i and k–2,i. The deintercalation experiment is modeled considering an initial condition where WO3 is in a state of saturated doping and in a vacuum atmosphere. Hydrogen atoms in the crystal lattice migrate toward and from the surface (rate constants k–2,d and k2,d) where they can recombine to form H2. Differently from the intercalation process, in this step the hydrogen recombines in a spontaneous reaction that does not require a catalyst. H2 molecules can thus form anywhere on WO3 surface. Subsequently, hydrogen molecules desorb and are pumped away in an irreversible process (k–1,d).

Figure 4

Kinetics of hydrogen intercalation and deintercalation. (a) Schematics of the processes involved in H2 exchange with WO3. The rate constants associated with each process are indicated on the respective reaction arrow. (b) Experimental curves of intercalation in low-pressure hydrogen mixture (CH = 20%, 5 mbar total pressure). (c) Deintercalation curves (vacuum atmosphere) at different temperatures. The conductance is divided by its final (σ∞) and initial (σ0) value for intercalation and deintercalation curves, respectively. The dashed lines are fits obtained from numerical solution of eqs and 1b. (d–g) Rate constants extracted from the fits. The solid lines are Arrhenius fits used to extract the activation energy of the reaction steps.

Hydrogen intercalation is experimentally studied starting from an undoped, low-conductance state. In Figure b we show intercalation curves performed in a low-pressure (5 mbar) gas mixture with CH = 20% at different sample temperatures. Higher temperatures determine faster intercalation, with the conductance reaching a saturation value after a few tens of minutes. The deintercalation process in Figure c is studied starting with WO3 in a high-conductance state. At the beginning of the measurement, WO3 is exposed to a vacuum atmosphere, and the conductance decreases over time, with a faster rate at higher temperature. We note that in vacuum the deintercalation process is much slower than in air, so that the curves do not reach full saturation in the duration of the experiment (2 h), in agreement with our interpretation of the role played by oxygen in accelerating the desorption process (Figure S4).

The kinetic equations associated with the reactions in eqs and 1b (see also the Supporting Information) are solved numerically to fit the experimental data (dashed lines in Figures b and 4c). Our model shows good agreement with the measurements and allows us to extract the rate constants of the different processes (Figures d to 4g). We find that k1,i is at all temperatures 1 order of magnitude larger than k–1,d, signaling a faster rate of H2 splitting compared to its recombination. This difference supports our choice to consider the dissociative adsorption of H2 an irreversible process during the intercalation experiment (i.e., k–1,i ≪ k1,i ). In the deintercalation experiment, instead, k−1,d becomes important because the vacuum atmosphere hinders the H2 splitting reaction by carrying away the H2 molecules as soon as they form. We fit the temperature dependence of the rate constants with an Arrhenius model, finding activation energies of 0.13 eV for k1,i and 0.12 eV for k–1,d. These results can be compared with ab initio computations that find activation energies in the range 0.06–0.42 eV for the dissociative chemisorption on Pt surfaces, with lower values expected for smaller particles.[39,40] Similarly, recombinative desorption at oxide surfaces entails activation energies in the range 0.05–0.9 eV depending on the hydrogen binding energy at the adsorption site.[41]

The surface-to-bulk diffusion rate constants k2,i/d show a weak temperature dependence (Figure f). The bulk-to-surface rates k–2,i/d in Figure g, instead, show a linear trend which allows to extract activation energies of 0.43 and 0.24 eV in the intercalation and deintercalation experiments, respectively. The higher activation energy required during intercalation stems from the fact that hydrogen splitting can only happen on the catalyst particles, so that the hydrogen atoms have to overcome two energy barriers: the diffusion on WO3 surface and then the diffusion in WO3 bulk. During deintercalation, instead, k–2,d is only related to the hydrogen diffusion in WO3 bulk because recombination can happen anywhere on the WO3 surface. Previous experimental and density functional theory calculations reported energy barriers between 0.26 and 0.37 eV for hydrogen migration from the Pt sites to WO3 surface,[42,43] which are compatible with our experimental results.

Electrical Transport

The previous analysis showed that both intercalation and deintercalation are thermally activated processes, so that their rate can be controlled changing the sample temperature. Furthermore, we showed that the room temperature resistivity of WO3 films can be continuously tuned by H2 intercalation. As a result, hydrogen kinetics in the WO3 lattice can be arrested by cooling down the material. This allows to get an insight into the electronic doping mechanism by measuring the transport activation energy of the free carriers induced by hydrogen doping. In Figure a, we show a series of resistance versus temperature curves measured on a single WO3 device under different doping conditions. Before each measurement, we expose the sample to hydrogen gas at room temperature in order to obtain the desired sheet resistance ρ300 K, after which we evacuate the sample chamber and start a cooling ramp. All curves show a semiconducting trend, where higher doping produces a shallower temperature dependence. We note that the resistance versus temperature curves are similar to what previously reported for WO3 doped with oxygen vacancies,[31,44] suggesting the occurrence of a similar mechanism of extrinsic doping. This interpretation is supported by recent experiments on WO3 nanowires which showed that H2 intercalation is accompanied by the formation of interstitial H2O molecules along with oxygen vacancies.[36,45]

Figure 5

Low-temperature transport at different levels of H2 doping. (a) Resistance versus temperature curves of a single WO3 device. The dashed lines are fits with the Arrhenius model for activated transport. (b) Activation energy as a function of room temperature resistivity ρ300 K. The solid line is a linear regression to the data points.

We consider an Arrhenius mechanism to describe the temperature dependence of the electrical transport σ = 1/ρ ∼ e–, where Ea is the activation energy. The resulting fits in the high-temperature regime (above 150 K, dashed lines) show good agreement with the experimental data and allow to extract the trend of Ea as a function of ρ300 K reported in Figure b. The data show the logarithmic dependence Ea = A log ρ300 K + B, with A = 19 meV and B = 90 meV. We show in the Supporting Information that the activation energy can be expressed aswhere γ is related to the extrinsic donor charge, ΔS° is the reaction entropy, and CH is the hydrogen concentration inside the WO3 lattice. Equation provides a direct relation between Ea and ρ300 K through the coefficient kBT300 K ∼ 26 meV. However, ρ300 K is also a function of CH, and by comparing eq with the experimental fit, we find that the resistivity can be expressed as ρ300 K ∼ (CH)−γ/0.3. We show in the Supporting Information that γ = 1/3 if the electron donors are oxygen vacancies. Substituting this value, we obtain an almost linear relationship between the conductance and hydrogen concentration in the material σ300 K = 1/ρ300 K ∼ CH, in agreement with the assumption we used to treat the experimental intercalation and deintercalation curves with the kinetic model. The linear dependence between hydrogen concentration in WO3 and electrical conductivity constitutes a powerful transduction scheme for the electrical sensing of H2 gas.

Conclusions

To conclude, single-crystal Pt-decorated WO3 ultrathin films have proven extremely sensitive and fast probes for solid state hydrogen gas sensors. Thanks to the high crystal quality and high surface-to-volume ratio, our devices showed a 5 orders of magnitude change in electrical resistivity upon exposure to H2 gas at room temperature. The device response can be tuned by controlling the temperature, enabling detection of a hydrogen concentration down to 10 ppm at room temperature and down to 1 ppm at 65 °C, where we measured short response times as low as 1 s for concentrations above 100 ppm. We successfully modeled the intercalation and deintercalation of hydrogen in the WO3 lattice with a kinetic of thermally activated processes. Our analysis allowed to extract the activation energies of the different reactions, which are in good agreement with previous reports based on ab initio calculations. These results show that heteroepitaxial WO3 films hold great potential for high-performance hydrogen gas sensors, providing high sensitivity, fast response, high dynamic range, and a simple device architecture.

Experimental Section

Sample Growth

WO3 thin films were grown by pulsed laser deposition on commercially available 5 mm × 5 mm SrTiO3 (001) substrates, with TiO2 surface termination. The laser ablation was performed using a KrF excimer laser (Coherent COMPexPro 205, λ = 248 nm) with a 1 Hz repetition rate and 1 J cm–2 fluence. The target-to-substrate distance was fixed at 55 mm. The WO3 films were deposited from a WO3 sintered target at 500 °C substrate temperature and 8 × 10–2 mbar oxygen pressure. Film thickness was monitored in situ during growth by intensity oscillations of reflection high-energy electron diffraction (RHEED). At the end of the growth, the films were cooled down to room temperature in the same oxygen pressure used during deposition. Further details on the growth and sample characterization are reported in ref (31).

Van Der Pauw Geometry Fabrication

Standard e-beam lithography with PMMA resist was used to pattern WO3 heteroepitaxial films. The metal contacts were deposited by evaporation of 5 nm Ti and 45 nm Au at room temperature, followed by lift-off. Ar ion etching was used prior to metal deposition to etch 5 nm of WO3 in order to provide edge contacts to the film in an interdigitated pattern, which ensure a low contact resistance. The WO3 mesa was defined by a subsequent lithographical and etching step which fully removes the WO3 material from the regions surrounding the van der Pauw geometry. The etching was performed using a beam current of 2 mA and an energy of 500 eV, corresponding to an Ar ion flux of 0.25 mA cm–2. These parameters produce etched regions which are insulating at all temperatures.

Gas Mixture Preparation

To regulate the H2 concentration with high precision, we used the series of vacuum chambers described in Figure f: the preparation chambers PC1, PC2, and sample chamber SC1 having volume ratios 1:9:1. Before introduction of the gases, the chambers were evacuated to p < 0.1 μ bar. The hydrogen gas was sourced from a bottle of 20% H2/Ar. This gas blend was first introduced and measured by a pressure sensor in PC1 and then expanded to the volume PC1 + PC2, where pure Ar gas was introduced in order to reach the desired total pressure. To begin the experiment, the gas mixture was finally allowed to enter SC1. For example, to achieve the lowest hydrogen concentration used in the experiment (1 ppm in 1 bar total pressure), PC1 is first filled up with 55 μbar of H2 blend. We note that this value can be precisely evaluated as it is well above both the lower detection limit (1 μbar) of the pressure sensor used for the measurement and the chamber base pressure (0.1 μbar). The gas mixture is then introduced in the volume PC1 + PC2, where it is topped up with Ar to reach the total pressure of 1.1 bar. Finally, the sample is exposed to the gas mixture at 1 bar in the total volume PC1 + PC2 + SC1.

Electrical Measurements

The resistive measurements were performed with a Keithley 6430 sourcemeter using a four-probe configuration and a current bias of 100 nA.

Air Conditions

Air from the environment (relative humidity 50%) was used in the sensing measurements.

Numerical Solution of Kinetic Equations

The kinetic model in the Supporting Information was solved using the NDSolve function of Wolfram Mathematica. The resulting fit curves yield a coefficient R2 ∼ 0.99, indicating that our model is a good approximation of the experimental reactions, although higher-order corrections might be present.