WS₂ and MoS₂ thin film gas sensors with high response to NH₃ in air at low temperature

Continuous and reliable detection of different gases is essential in industrial process monitoring, vehicle emission control, in and outdoor air quality safety and environment protection [1]. In these applications, traditionally metal oxide semiconductors such as SnO₂, WO₃, CeO₂, Nb₂O₅ and ZnO as well as their metal or metal oxide decorated derivatives have been most commonly utilized as sensing materials [2–5]. While these sensors and their arrays offer excellent and reliable discrimination and even quantification of analytes, their operation is only possible at high temperatures that requires considerable power sourcing. Nowadays, with the spread of internet-of-things and dispersed networks that involve complex sensing systems with small but large number of devices, power consumption becomes a significant factor. Powering such devices is even more cumbersome in the case of remote and autonomous off-grid systems [6–8] running on batteries and/or energy scavenging units with limited power output [9]. Accordingly, new materials that would allow for low temperature operation could inevitably alleviate power related challenges and contribute to better and more robust sensor networks.

2D layered transition metal dichalcogenides (TMDs) have recently been found very attractive for a wide range of applications such as energy storage, photodetectors and switches, electrocatalysis, photocatalysis as well as for chemoresistive sensors [10–14]. TMDs (MX₂, where M is a transition metal and X is a chalcogen) have a layered structure, in which the coordination of metal atoms can be trigonal or octahedral resulting in either hexagonal or rhombohedral structure symmetry [15–17]. These materials have not only high specific surface area but as the covalent layers are hold together with weak van der Waals forces, and thus have large interplanar spacing, intercalation of small moieties in between the layers is also possible making such structures suitable for multitude interactions with the environment. Recent studies of WS₂ and MoS₂ as well as their hybrid structures (functionalized nanofibers [18, 19], quantum dots [20] and metal doped nanoflower structures [21]) in gas sensing applications indicate substantial response at low operating temperatures with particular selectivity to certain gases such as H₂S and NH₃ [18, 22–24]. The lowered operation temperature of transition metal chalcogenide-based sensors, in reference to the oxides of the corresponding metals, is due to their typically lower bandgap and better conductivity. Furthermore, similar to semiconducting metal oxide sensors, the gas selectivity of TMDs is in many instances is associated to the affinities of surfaces for adsorbing different analytes and related surface charging/polarization effects [18, 24]. In addition, it has been identified that apart from surface charging/polarization, reversible doping of the chalcogenide lattice with heteroatoms can significantly contribute to sensing [18].

The aim of the present work is to elaborate on the low temperature gas sensing properties of TMDs produced by sulfurizing sputter deposited tungsten and molybdenum films. As we show, the direct synthesis of WS₂ and MoS₂ thin films on Si chips is robust and offer a route which is up and down-scalable without necessitating any transfer steps of 2D materials [24]. Thin films of crystalline WS₂ grown on Si chips display distinct selectivity towards NH₃ with a sensitivity of 0.10 ± 0.02 ppm⁻¹, even at 30 °C. Interestingly, the surfaces remain practically irresponsive to other analytes (CO, H₂, and H₂S), which is unexpected in the light of our previous studies on WS₂ nanowire-nanoflake hybrids that showed unprecedented selectivity and sensitivity to H₂S.

Materials and sensor fabrication

Materials characterization

Electrical and gas sensing measurements

AFM images of large scans areas (10 × 10 μm²) revealed the sulfurized Mo films were uniform, whereas the sulfurized W layers had some small pinholes of 0.5–2 μm in diameter (not shown here). Apart from the surface defects in the sulfurized W layer, both products were found homogeneous with grain-like morphology. The original particle size of the metal films increased from 23 ± 4 nm (Mo) and 21 ± 4 nm (W) to 37 ± 9 nm and 36 ± 8 nm, respectively. The particle size distributions (inset panels) seem to follow a log-normal curve rather than Gaussian. The increased particle diameters are likely caused by the formation of the corresponding sulfides, having lower density than that of the corresponding metals (figures 1(a)–(d)).

Zoom In Zoom Out Reset image size

Raman spectra of the films displayed in figures 1(e) and (f) indicate the materials are crystalline MoS₂ and WS₂, respectively [25, 26]. Two strong peaks are observed for MoS₂ at 378 and 405 cm⁻¹ that correspond to the in-plane (E¹_2g) and out-of-plane (A_1g) vibration modes [27]. According to the difference of Raman shifts and also to the intensity ratio of these two peaks, MoS₂ is multi-layered [28]. Likewise, the WS₂ film has two intensive peaks at 349 and 415 cm⁻¹ that can be assigned to in-plane (E¹_2g) and out-of-plane (A_1g) vibration modes, evidencing that the structure is multi-layered [29, 30]. The presence of the longitudinal acoustic mode vibrations (LA) and their higher-order harmonics (2LA and 4LA) are also visible in the spectra. The low intensity of the 2LA peaks confirms the chalcogenides are multi-layered [26]. Note, that in each spectrum, the relatively intensive peak at around 520 cm⁻¹ is due to the Si substrate.

Grazing incidence x-ray diffraction patterns of the original and sulfurized metal films are displayed in figures 1(g) and (h). The metal films pattern fits well the reference patterns PDF 42-1120 and PDF 01-1204 for Mo and W, respectively. In addition, the metallic tungsten film shows some W₃O which is indexed according to PDF 02-1138. The pattern of MoS₂ shows two strong reflections at 33.2° and 59.5°, which may be indexed as the (100) plane and superposed reflections from the (110) and (008) planes of the hexagonal form of the crystal (PDF 75-1539). However, the strong reflections of the (002) and (102) planes at 14.1° and 36.0° of layered structure that should be also present in the pattern are missing from our diffractogram suggesting the lack of long-range order in stacking of the layers in the material. Another explanation might be a specific crystal orientation on the substrate with basal planes being perpendicular to the surface. However, this latter option is unlikely, since the AFM images show spherical particles on the surfaces that do not resemble oriented crystals. On the other hand, in the case of the WS₂ film, the pattern can be assigned to the planes of the hexagonal lattice (PDF 02-0131). The reflection at 14.1° corresponds to the (002) plane, the broad peak at 33.1° is due to the superposed intensities of the (100) and (101), whereas the broad reflection centered at 59.0° can be assigned to the unresolved (110), (008) and (112) planes [29]. It is also worth noting here, that the broad reflection at around 20.0° is due to amorphous SiO₂ of the substrate.

The composition of the films was analyzed with EDS (table 1). The aforementioned pinhole defects seen in WS₂ films comprise only silicon and oxygen confirming that the defects are indeed voids in the film. Since the thickness of TMD films is ∼20 nm, both Si and O of the Si/SiO₂ substrate are visible in the energy-dispersive x-ray spectra even at locations without pinholes. The ratio of sulfur and metal is very close to 2:1 for both TMD films, which implies the films are at least nearly stoichiometric. In the case of WS₂ there is a notable amount of C contamination present, which was not observed in MoS₂ films.

Before the actual sensor measurements, the electrical conductivity of each film was assessed with I–V measurements. Both MoS₂ and WS₂ films were sufficiently conductive (with resistances of several tens of megaohms and hundreds of megaohms, respectively) to perform sensor measurements on those. The I–V curves were found linear for MoS₂, whereas WS₂ showed slight nonlinearity (figures 2(a) and (d), respectively), suggesting a barrier between the semiconducting film and the Ti/Pt electrodes [31]. Although the primary direct contact with the chalcogenide films was meant to be Ti, some Pt also deposited on the films around the perimeter of the Ti under-metallization due to the imperfect contact between the shadow mask and the substrate. Considering the typical bandgap, electron affinity and the work function values of bulk MoS₂ (E_g ∼ 1.3 eV, χ ∼ 4.0 eV and φ ∼ 5.25 eV)[32, 33] and WS₂ (1.4, 4.0 and 5.1 eV) [34] as well as the work function values of Ti (φ_Ti ∼ 4.3 eV) and Pt (φ_Pt ∼ 5.3 eV), we find that the contact between Ti and the chalcogenide films is Schottky-type (figures 2(b) and (e)), whereas it is quasi-ohmic with Pt (figures 2(c) and (f)). Furthermore, because of the better alignment of the Fermi level of Pt with the valence band of MoS₂ in reference to WS₂, the ohmic character is also expected to be better, which explains the good linearity of the I–V slopes and higher conductivity for the MoS₂ based devices and slight nonlinearity measured for the WS₂ based ones.

Zoom In Zoom Out Reset image size

In general, WS₂ based devices were electrically less conductive but had considerably higher response to the analytes, although with higher device-to-device variation (average of six devices for each gas) in reference to MoS₂ sensors (figures 3 and 4(c)). The films show a small response to H₂ (0.00005 ± 0.00003 ppm⁻¹ and 0.00016 ± 0.00017 ppm⁻¹ for MoS₂ and WS₂, respectively). This is comparable to previous results with WS₂ nanowire/nanoflake hybrid structures [18]. For H₂S and CO the responses were either nonexistent or inconsistent. Regardless of the variance of the response to other analytes, for NH₃ the devices display a consistent and very good response (0.0026 ± 0.0022 ppm⁻¹ and 0.0586 ± 0.0305 ppm⁻¹ for MoS₂ and WS₂, respectively).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Since the response to NH₃ with both sensors was very distinct at 1 ppm gas concentration, additional measurements were performed in the sub-ppm NH₃ concentrations. Significant responses to 200 ppb of the analyte could be measured with average sensitivities of 0.0315 ± 0.00282 ppm⁻¹ and 0.104 ± 0.0243 ppm⁻¹, respectively (figure 4). According to the signal to the base noise ratio, the limit of detection (LOD) is under 100 ppb for both materials. The response time of the MoS₂ based sensors to the analytes were found to be rather short for NH₃ (5–6 min) and slightly longer for other gases (10–30 min). The sensor recovery time constants were long (30–60 min) and varied greatly among the gases. On the other hand, sensors made of the WS₂ films showed a consistent and quick (5–6 min) response and recovery for each gas, which is excellent considering that the measurements were carried out at near room temperature.

According to the reducing nature of all the analyte gases and the response of the sensors, it can be concluded that both WS₂ and MoS₂ exhibit p-type semiconducting behavior, and the resistance of the sensors increase with the concentration of the analytes. This is usually explained by surface adsorption and subsequent charge transfer based mechanisms; the reducing gases donate electrons (or localize holes) in the materials, thus reducing the concentration of charge carriers, and consequently reduce the conductivity of the material [18, 20]. A number of other studies show selectivity to NH₃ for both WS₂ [22, 23, 31, 35] and MoS₂ [20, 36, 37]. In the case of MoS₂, the LOD of 100 ppb is similar to other results (300 ppb measured and 50 ppb predicted) [24]. In the case of WS₂, the reported LODs are about a decade worse (∼1 ppm)[38, 23] than our data. (In table S2, we list a more detailed comparison of the results with other reported values for MoS₂ and WS₂ based devices and commercial NH₃ sensors.)

As the sensing behavior of the thin films presented in the manuscript is similar to or better than those reported for other MX₂ based devices regarding NH₃, we may assume that the sensing mechanism is also similar, i.e. associated with surface adsorption and consequent partial charge transfer. Calculated (local-density approximation, LDA) surface adsorption energies of H₂, NH₃ and CO on MoS₂ with corresponding values of −82 meV, −250 meV and −128 meV, respectively, also seem to support the high response to NH₃ on MoS₂ films [20, 39]. Similarly, on WS₂, the calculated energies [18] are –57.4 meV (H₂), –171.7 meV (NH₃). and −84.7 meV (CO) indicating that the response to NH₃ is expected to be the highest. Here, we have to point out another effect regarding H₂S. Namely, considering the high adsorption energy of H₂S (−181 meV) on WS₂ nanosheets, we shall also expect to see good response to H₂S. However, this is not the case with our thin films. The reason is that competitive lattice doping with O and S is responsible for the detection of H₂S as we have concluded in our earlier work [18] based on XPS analysis of WS₂ nanowire-nanoflake hybrid structures, in which nanosheets of ∼5 nm thickness were sticking out from the surface of the nanowires. The structure is different now, since both WS₂ and MoS₂ films are comprising nanoparticles of 20–30 nm in their diameter with limited diffusion of O and S in the lattices, thus despite the large adsorption energy of H₂S, the sensor response is insignificant to this analyte.

To analyze the cross-sensitivity of the sensors, we performed two sets of measurements for two samples of each material (figure 5). In the first set, we applied gas pulses with concentrations of 1 ppm, 200 ppm, 10 ppm and 5 ppm for NH₃, H₂, CO and H₂S, respectively. In the other set, the respective concentrations were 10, 100, 5 and 5 ppm. The sensitivity to NH₃ remained high for all samples throughout the measurements (the first three pulses in each plot), which proves the repeatability and reproducibility of the devices. During the long 4th pulse of NH₃ (starting at 300 min) we introduced also H₂, then CO and H₂S for short intervals (30 min each starting at 330 min, 390 min and 460 min, respectively) to see the influence of the secondary analytes. The MoS₂ film was affected only by H₂ but only when its concentration was high relative to the concentration of NH₃ (figure 5(a)). The influence of secondary analytes on the response to NH₃ introduced at similar concentrations was insignificant (figure 5(c)). However, in the case of the WS₂ based device, we observe a decreased response to low concentration NH₃, when introducing the other gases at high concentrations (figure 5(b)). The effect is lower, when the analyte concentrations were adjusted to be similar (figure 5(d)). The results measured with WS₂ sensors indicate, that the secondary gas pulses clan flush away the surface adsorbed NH₃ thus influencing the sensor performance.

Zoom In Zoom Out Reset image size

Furthermore, it is worth pointing out that all sensors were operating in a rather reliable and repeatable manner throughout a period of more than nine month testing in various experimental conditions, which indicates that the thin film WS₂ and MoS₂ structures are very robust and practical.

A method to produce MoS₂ and WS₂ thin films by simply annealing sputtered Mo and W films in S vapor is proposed, and the gas response of the as-synthesized films are examined in the classical resistive sensor arrangement. Both materials showed excellent selectivity for NH₃ with very high corresponding sensitivity values (0.03 ± 0.0 ppm⁻¹ and 0.10 ± 0.02 ppm⁻¹, respectively) at 30 °C. Despite the low operation temperature, the WS₂ based sensors responded to the stimuli and then recovered after each gas pulse quite fast (5–6 min). The results obtained with our robust TMD films compete with other exfoliated or CVD grown single or few-layer chalcogenides and outperforms metal oxides. Since the process steps involved apply standard techniques, up and down-scalable, and the films produced are easy to process further (e.g. coating with other materials) the results presented in this paper may pave the road for sensor applications of TMD thin films.

Support received from the Väisälä Foundation, Riitta and Jorma J. Takanen Foundation, Academy of Finland (298409), University of Oulu (Project: Entity) and European Union Interreg Nord – Lapin liitto (Project: Transparent, conducting and flexible films for electrodes) are acknowledged. Kai Metsäkoivu (Center of Microscopy and Nanotechnology, University of Oulu) is acknowledged for the metal films deposition.

The authors declare no conflicts interest.