Ultrasensitive H2S gas sensors based on p-type WS2 hybrid materials

Owing to their higher intrinsic electrical conductivity and chemical stability with respect to their oxide counterparts, nanostructured metal sulfides are expected to revive materials for resistive chemical sensor applications. Herein, we explore the gas sensing behavior of WS2 nanowire-nanoflake hybrid materials and demonstrate their excellent sensitivity (0.043 ppm-1) as well as high selectivity towards H2S relative to CO, NH3, H2, and NO (with corresponding sensitivities of 0.002, 0.0074, 0.0002, and 0.0046 ppm-1, respectively). Gas response measurements, complemented with the results of X-ray photoelectron spectroscopy analysis and first-principles calculations based on density functional theory, suggest that the intrinsic electronic properties of pristine WS2 alone are not sufficient to explain the observed high sensitivity towards H2S. A major role in this behavior is also played by O doping in the S sites of the WS2 lattice. The results of the present study open up new avenues for the use of transition metal disulfide nanomaterials as effective alternatives to metal oxides in future applications for industrial process control, security, and health and environmental safety.


Introduction
Gases with different properties, origins, and concentrations are pervasive in our environment. Some of these gases are highly toxic and hazardous, while others are essential for life or indicators of health status. Accordingly, sensors for gas detection and monitoring are needed in various sectors such as environmental protection, industrial process monitoring and safety, amenity, energy saving, health, and food industries [1]. Metal oxide semiconductors stand out as the most common active sensing materials used in practical devices. Grain size reduction to the nanometer range and relatively easy sensitization to various analytes (via either lattice doping with anions and cations or surface decoration with metals and metal oxides) have emerged as key strategies to improve their detection properties [2,3]. The main drawbacks of current metal oxide-based devices are associated with their typically high operating temperatures (several hundreds of degrees) and limited use in environments containing even traces of sulfur and sulfides, because of the easy poisoning of the catalytically active surfaces involved in the adsorption and sensing of analytes [4]. In principle, other types of gas sensors whose operation is based on, e.g., gas ionization [5] or changes in the optical properties of waveguides [6], may represent alternative solutions; however, their typical detection limits are at most a few parts per million of analyte. For this reason, and to further improve the existing sensing performance of resistive-type gas sensors, significant efforts are being directed towards developing new classes of materials that are less susceptible to poisoning, such as carbon nanotubes [7], graphene [8], conducting polymers [9], and transition metal dichalcogenide (TMD) nanostructures [10], to replace metal oxides.
Among the new types of nanoscopic sensors being studied, layered transition metal dichalcogenide (MX 2 , M = Mo, W; X = S, Se) nanostructures have recently attracted significant interest. Often compared to graphene and other two-dimensional (2D) nanomaterials, their properties present distinct advantages for electronic, optical, and electrochemical sensors [11]. Recent studies, mostly focused on MoS 2 [12][13][14][15] and WS 2 [16,17], highlighted the gas sensing appli-cations associated with the large specific surface area of the TMD materials, particularly of those with a mono-or few-layered microstructure. Previous studies suggest that the working mechanism of TMD gas sensors involves charge transfer-based conductance modulation [11,13,18]. Since the electron density is dependent on the dimensions of the crystal lattice, it is reasonable to expect that the gas sensing properties are also influenced by the dimensions and microstructure of the sensing material. This motivated our very recent investigation of the gas sensing properties of nanowire-nanoflake hybrid WS 2 nanostructures [19].
Therefore, in this work, the gas sensing characteristics of WS 2 nanowire-nanoflake hybrid materials were investigated using a simple two-terminal Taguchi-type sensor arrangement, operating at moderate temperatures (30 and 200 °C ). The sensors were exposed to five different analytes (H 2 S, CO, NH 3 , H 2 , and NO buffered in air) and their response, sensitivity, and recovery properties were assessed. The goal was to obtain experimental evidence supporting previous gas sensitivity data predicted from modeling studies, and to compare the experimental measurements to more recent molecular dynamics simulation results. We show that the WS 2 nanowire-nanoflake hybrids have particularly high sensitivity towards H 2 S, allowing detection and quantification of analyte concentrations on the order of parts per billion, competing with other known materials such as Fe 2 O 3 nanochains and nanoparticles [20,21], CuO-SnO 2 [22], CuO nanoparticles [23] and nanosheets [24], mesoporous WO 3 [25], CeO 2 nanowires [26], and PbS quantum dots [27] (for a comprehensive list, see Table S1 in the Electronic Supplementary Material (ESM)).

Experimental and modeling details
The WS 2 nanohybrids were synthesized according to the procedures detailed in our previous report [19]. In brief, WO 3 nanowires were first synthesized by hydrothermal recrystallization of Na 2 WO 4 ·2H 2 O in acidic environment at 180 °C under autogenic pressure, and then sulfurized at 800 °C for 10 min in sulfur vapor to convert them to WS 2 nanowire-nanoflake hybrids. The microstructure of the as-obtained materials was determined by field emission scanning electron Nano Res. microscopy (FE-SEM, Zeiss Ultra Plus) and transmission electron microscopy (TEM, JEOL JEM-2200FS). The chemical surface composition of the nanomaterials was assessed by a Kratos Axis Ultra X-ray photoelectron spectrometer (with a monochromatic Al Kα source operated at 150 W) equipped with a delay line detector and a charge neutralizer. The spectra were processed with the Kratos software.
Two-terminal Taguchi-type resistive gas sensors were fabricated on test chips with Pt/Ti (300 nm/45 nm) microelectrodes on a Si/SiO 2 substrate (a 300 nm-thick thermal oxide deposited on B-doped p + -Si). The electrode patterns were defined by optical lithography using the lift-off method. To prepare the sensor devices, a small amount (~ 20 mg) of the WS 2 nanowire/ nanoflake hybrid powder was dispersed in 10 mL acetone by ultrasonication, then deposited between the platinum electrodes by drop casting, and finally dried under ambient conditions for 1 day before the gas sensing measurements. The amount of drop-cast material was chosen in such a way to produce a reasonable conductive active layer, with typical resistance values between 3 and 100 MΩ at room temperature.
The gas sensing measurements were performed by testing the chips in a Linkam TMS 94 gas flow chamber. The sample resistance was measured at a 1 V bias, using a computer-controlled Hewlett-Packard 3458 A multimeter. The sensor devices were tested at 30 and 200 °C on air-buffered analytes (H 2 S, CO, NH 3 , H 2 , and NO) with nominal concentrations between 1 and 600 ppm, using a MKS Type 247 mass-flow controller. Some of the sensors were also tested on sub-ppm concentrations (between 10 and 1,000 ppb) of H 2 S at 200 °C .
In addition, we performed first-principles calculations based on density functional theory (DFT), using the generalized gradient approximation (GGA) exchange-correlation functional of Perdew, Burke, and Ernzerhof (PBE) [28], as implemented in the Quantum Espresso code [29]. The important role of van der Waals (vdW) interactions in the layered structure of WS 2 was accounted for by including the Grimme's semiempirical dispersion correction (D2), which describes the vdW interactions by a pairwise force field [30]. We used norm-conserving Goedecker-Hartwigsen-Hutter-Teter pseudopotentials [31], whereas the electronic wavefunctions were expanded in a plane-wave basis set, with energy cutoffs of 90 and 360 Ry for the wavefunctions and the charge density, respectively. A 10×10×1 Monkhorst-Pack [32] grid was used for k-point sampling within a single unit cell of bulk WS 2 . The total energy and the forces on each atom were converged to within 1 mRy/atom and 0.2 meV, respectively, thus allowing us to obtain a quantitative description of the sensing behavior of the 2D-based materials. Furthermore, 4×4 slabs were built to simulate the adsorption of an individual molecule of H 2 S on both single-layer (32 S and 16 W atoms) and bilayer (64 S and 32 W atoms) WS 2 , to take into account the possible intercalation of gas molecules between the adjacent layers. Due to the periodic boundary conditions, this geometry corresponds to 6.28 × 10 13 molecules/cm 2 in the monolayer. The distance between two neighboring gas molecules was thus larger than 12 Å. In order to avoid unphysical interactions between images along the non-periodic direction, the distance between monolayers was set to ~ 20 Å, with a dipole correction layer. Integrations over the Brillouin zone were carried out using a regular mesh of 3×3×1 k-points for the structural relaxation of the slab, and at the zone center for a single molecule. A denser mesh of 6×6×1 k-points was used for the density of states (DOS) and Bader charge transfer calculations [33]. The calculations were also extended to study the adsorption of H 2 , CO, NO, and NH 3 gases on the WS 2 monolayer; spin-polarized calculations were performed in the case of NO adsorption.

Results and discussion
The hydrothermally synthesized WO 3 samples were composed of precisely oriented nanowires, as revealed by the scanning and transmission electron microscopy measurements ( Fig. 1(b)). Based on the high-resolution transmission electron microscopy (HRTEM) image in Fig. 1(c) and the selected area electron diffraction (SAED) pattern, the spacing of the lattice fringes was found to be 0.39 nm, which can be indexed as the (001) plane of hexagonal WO 3 . This indicates that the growth of the nanowires occurred along the [001] Nano Res.
direction. After sulfurization, the morphology and other characteristics of the as-formed WS 2 nanowire/ nanoflake structure ( Fig. 1(d)) were identical to those reported in our previous work. The hybrids consisted of elongated rod-shaped structures with partially peeled surfaces, thus forming nanoflakes ( Fig. 1(e)) that are an integral part of the original nanorod [19]. The length of the hybrid structures is around 1-3 μm. Their core has a typical diameter of 200-300 nm, whereas the size of the partially peeled flakes is ~ 200 nm in diameter and 10 nm in thickness (Fig. 1). The high-resolution image of a single-crystal 2D WS 2 nanoflake ( Fig. 1(f)) reveals crystal planes with a spacing of 0.27 nm, which can be indexed as the (100) plane of hexagonal WS 2 .
Resistive gas sensor measurements of the WS 2 hybrid materials dispersed on the test chips indicated p-type semiconducting behavior, as the resistance increased for reducing gases (H 2 S, CO, NH 3 , and H 2 ) and decreased for nitric oxide, which is an oxidant (Fig. 2). In simple terms, the sensing mechanism is based on the localization of positive charge in the lattice by the surface-adsorbed electron donor molecules (i.e., on the decrease in conductance caused by reducing chemicals) and vice versa (i.e., the increased conductance caused by electron acceptors on the surface, which induce p-type doping in WS 2 ). The results are consistent with sensing mechanisms recently proposed for TMDs [11,13,18]. The sensing performances of the present materials, however, are superior to those measured for WS 2 thin films [17] and comparable to the performance of MoS 2 and WS 2 flakes [10][11][12][13][14]. It is worth noting here that the parent WO 3 nanowires are n-type semiconductors that display an opposite sensing response to the same gas molecules [34].
The sensors show a very rapid response to all gases examined. The typical time constants are between 1 and 2 min at 200 °C , and slightly higher (~ 3-6 min) at lower temperatures. Since the gas response reflects the gas-solid interfacial equilibrium at the surface of WS 2 , the faster response at higher temperatures can be expected, as a result of the increased reaction rates. A similar behavior was also found for the sensor recovery. The faster gas desorption and setting of the air-WS 2 equilibrium at the interface are the main drivers of the rapid sensor recovery at higher temperatures.
The sensor response to the analytes shows a nearly linear dependence on the gas concentration, with a slight decay in sensitivity at higher concentrations. The sensitivity to H 2 S is particularly high (0.023 ppm -1 ) compared to the other analytes (Fig. 3), which prompted us to carry out further tests on the WS 2 -H 2 S system.  Exposing the sensor to sub-ppm H 2 S levels revealed that the detection limit is as low as ~ 20 ppb at 200 °C , with a corresponding sensitivity of 0.043 ppm -1 .
To rationalize the present experimental findings and understand how H 2 S interacts with WS 2 , we carried out ab initio calculations using a simplified model involving single-and bilayer pristine/undoped WS 2 . The calculations of the adsorption of H 2 , CO, NO,  Fig. 2). The data point labeled with an asterisk denotes the sensitivity (0.043 ppm -1 ) measured at 20 ppb H 2 S. and NH 3 were based on the work of Zhou et al. [18], who used first-principles simulations to determine the most favorable geometries of these molecules on monolayer WS 2 . The same approach was followed in this work to estimate the adsorption energy of H 2 S on WS 2 (see details in Table S2 in the ESM). The relaxed structure of H 2 S adsorbed on the monolayer is shown in Fig. 4 , and E T [X] are the total energies of the supercell with the adsorbed molecule, the clean WS 2 slab, and the adsorbed molecule, respectively, in their optimized configuration. The calculated energies show that the adsorption of NO (-509.3 meV) is the most favorable among the systems investigated, followed by H 2 S (-181 meV) and NH 3 (-171.7 meV). The adsorption energies of H 2 (-57.4 meV) and CO (-84.7 meV) are considerably smaller. The adsorption energies thus follow a similar qualitative trend as the experimental sensitivities, with the exception of NO (see Fig. 3(b)). This trend is also in agreement with previous calculations based on the local density approximation (LDA) [18]. The negative sign in the calculated adsorption energies indicates that all adsorption processes are thermodynamically favorable; moreover, higher absolute values denote higher binding with the slab. The present adsorption energies show some correlation with our experimental results on multilayered WS 2 , which revealed good sensitivity towards H 2 S (2.3 × 10 -2 ppm -1 ), ammonia (7.4 × 10 -3 ppm -1 ), and NO (4.6 × 10 -3 ppm -1 ), along with moderate sensitivity towards CO (2 × 10 -3 ppm -1 ) and H 2 (2.3 × 10 -4 ppm -1 ). Furthermore, the calculated E ads of H 2 S on the WS 2 bilayer shows a slight increase (-246.2 meV) with respect to that measured for the monolayer. This may suggest a qualitatively higher sensitivity towards H 2 S in the case of the nanoflakes, compared with the monolayer. We also explored the possible intercalation of gas molecules between the adjacent WS 2 monolayers; however, our simple computational model predicts that the adsorption process is not energetically favorable in this case, due to the small interplanar distance between atoms in the adjacent layers (~ 3.1 Å) and the resulting repulsion between the S atoms belonging to the slab and the molecule. Nevertheless, we cannot rule out this possibility in the experiments, because structural defects in the material can provide perfect gates for the intercalation of gas molecules.
Electronic structure calculations were then carried out to understand the effects of the gas molecules on the electronic properties of WS 2 . The band structure and density of states calculated for the H 2 S molecule adsorbed on a WS 2 monolayer, shown in Fig. 5 (the results for the other gas molecules are available in Figs. S6-S9 in the ESM), are consistent with those reported in Ref. [18]. As shown in Fig. 5(a), the H 2 S molecule has little influence on the band structure of WS 2 in the combined system, which remains almost the same as that of the pristine layer. The highest occupied molecular orbital (HOMO) of H 2 S (green line in Fig. 5(a)) is located on the 2p orbital of the S atom and lies ~ 0.4 eV below the top of the valence band, as evidenced by a peak in the DOS curve (green line in Fig. 5(b)). The lowest unoccupied molecular orbital (LUMO) of H 2 S lies approximately 2.9 eV above the top of the conduction band (not shown). Therefore, the H 2 S molecule does not provide donor or acceptor states within the band gap of the WS 2 layer. The same behavior is observed in the case of H 2 S adsorbed on bilayer WS 2 (Figs. 5(c) and 5(d)) and also for the other gases, with the exception of NO. In particular, our results show that only the NO molecule introduces three impurity levels within the band gap of WS 2 , behaving as an acceptor (Fig. S8 in the ESM), in good agreement with the experiments.
Moreover, a Bader analysis was performed to study the local charge transfer associated with the adsorption of the gas molecules on the pristine WS 2 monolayer (Table S2 in the ESM). The results for H 2 , CO, NH 3 , and NO are in good agreement with previous data [18]. Accordingly, no sizable charge transfer is observed for H 2 and CO, and only a very small amount of charge is transferred in the case of NH 3 and NO. The NH 3 molecule acts as a donor by transferring 0.017 electrons to the monolayer, whereas the NO molecule acts as an acceptor, capturing 0.008 electrons. Similar to H 2 and CO, H 2 S also exhibits very limited charge transfer over the pristine WS 2 slab, which indicates that the excellent sensitivity of the hybrid material towards the H 2 S molecules observed in the experiments may have a different origin than the direct charge transfer within the clean monolayer. The minor amount of charge transfer observed for the most stable configuration is also compatible with the orbital mixing charge transfer theory. As discussed earlier, the mechanism of charge transfer between an adsorbate and a surface is partially governed by the mixing of the molecular HOMO and LUMO with the orbitals of the monolayer [35]. In the most favorable configuration of H 2 S on pristine WS 2 , the H atoms (LUMO) are oriented towards the monolayer, accepting electronic density from it. On the other hand, the HOMO of the molecule, located on the S atom, provides a slightly lower charge density to the slab because it is placed further away, resulting in a very small net charge transfer in the full system. We investigated the effect on the charge transfer of a H 2 S molecule with the opposite orientation, i.e., with the H pointing away from the slab [35]. As expected, the direction of the charge transfer was inverted, and the gas molecule could be considered as a donor. This possibility cannot be ruled out for some of the adsorbates under the actual experimental conditions. However, in the case of the pristine WS 2 monolayer this configuration is clearly energetically unfavorable.
In order to identify any change in the surface chemistry of the materials during the gas sensing experiments that may affect their sensing response, we carried out X-ray photoelectron spectroscopy (XPS) measurements of the WS 2 hybrid nanostructures. Three samples were studied (Table 1): the original WS 2 hybrids (sample #1), the hybrid materials exposed to air at 200 °C for 1 h (sample #2), and those subsequently exposed also to H 2 S at 200 °C for 1 h, in 1 ppm air (sample #3). The S 2p 3/2 and O 1s peaks at around 162.3 and 530.7 eV (associated with WS 2 and WO 3 , respectively) indicate the partial substitution of surface sulfur with oxygen when the WS 2 hybrid is heated in air. In addition, a new peak simultaneously appearing at 168.9 eV suggests the additional formation of sulfates, e.g., by the oxidation of sulfide and the reduction of O 2 to WO 3 . On the other hand, when the sample is exposed again to H 2 S (1 ppm in air), the sulfur concentration of the surface recovers the original value, despite the very low concentration of analyte. According to a recent report, for 2D WS 2 nanoparticles the replacement of S with O occurs even at room temperature, resulting in the formation of some amorphous/crystalline WO 3 phase in the nanoflakes below/above 250 °C [36]. However, the very efficient and fast recovery of the WS 2 structure even after a short treatment in 1 ppm H 2 S is unexpected and has important consequences for the sensing of H 2 S in the presence of O 2 (e.g., in ambient air). As mentioned is an n-type semiconductor, whereas WS 2 displays p-type semiconducting behavior. Accordingly, S doping in the WO 3 lattice or O doping in WS 2 are expected to cause significant changes in the electronic band structure and consequently alter the carrier localization upon adsorption of analyte gas molecules on the surface. Therefore, the H 2 S sensing process may be interpreted as a sequence of reversible adsorption/desorption and redox reactions, in which oxygen and sulfur compete for the anionic sites in the WS 2 lattice, thus modulating its band structure and electrical behavior, as schematically illustrated in Fig. 6.
In summary, our results indicate that the WS 2 nanohybrid possesses all key features (similar to other one-dimensional (1D) or 2D TMDs) to be applied as an effective gas sensor material, including a semiconducting nature and electrical transport properties adjustable via gas adsorption or (as observed in this study) lattice doping. Nevertheless, what renders the nanostructured hybrid probably superior to its 1D or 2D counterparts is the combination of the beneficial properties of nanowires and nanosheets. In particular, long 1D nanomaterials are preferred when electrically conductive networks need to be created between macroscopic electrodes. The long nanowires can establish percolation channels via a much lower number of interparticle connections than small zerodimensional (0D) nanoparticles or 2D sheets [34,37,38]. This is particularly important, since during the electrical measurements the interfacial transport might dominate and suppress the effects of gas adsorption on the nanoparticles. On the other hand, small and thin 2D nanosheets have the fascinating ability to adsorb a significant fraction of surface-bound analytes not only on the 2D crystal facets but also on their edges [39], which may contribute to enhance the sensitivity of the 2D crystals, relative to other nanoparticles lacking edge adsorption sites.

Conclusions
The WS 2 nanowire-nanoflake hybrid materials investigated in this study were found to possess electrical properties highly sensitive to H 2 S, along with moderate sensitivity to other analytes such as H 2 , CO, NH 3 , and NO in an air buffer. First-principles calculations based on the DFT-GGA approach with the PBE functional show that the intrinsic electronic properties of pristine WS 2 cannot fully explain its observed high sensitivity towards H 2 S, whose origin might also be attributed to defects and/or doping in the WS 2 lattice. To test this possibility, we carried out XPS measurements on pristine and modified WS 2 nanowire-nanoflake hybrids and found that O 2 present in the environment during the sensor measurements induce a partial (and reversible) substitution of S by O at the anionic sites of the lattice, which may explain the high sensitivity and selectivity of the WS 2 nanowire-nanoflake heterostructures towards H 2 S. Nano Res.