Effects of Phase Selection on Gas-Sensing Performance of MoS2 and WS2 Substrates.

Two-dimensional transition metal disulfides such as MoS2 and WS2 exhibit multiple phases. Altering their phase makes it possible to change their chemical and physical properties significantly. Although several phase-induced modification mechanisms have been reported, their effects on the gas-sensing performance of these substrates remain unknown. Here, the effects of phase selection on the gas-sensing characteristics of 1T' and 2H monolayer MoS2 and WS2 were explored using a density functional theory-based first-principles approach. The theoretical computations took into account the binding energy, band structure, theoretical recovery time, density of states, electron difference density, and total electron density. The results showed that there is a significant change in the density of states near the Fermi level as well as greater charge transfer between the gas in question and the substrate when the gas is adsorbed onto 1T' MoS2 and WS2. Thus, phase selection is important for improving the gas-sensing performance of monolayer MoS2 and WS2. This study provides theoretical evidence for increasing the sensing performance of polymorph films of these materials.

Introduction

Gas sensors are used in many fields and applications, including measuring environmental pollution, testing for fuel combustion during industrial production, and detecting the concentration of toxic gases at leakage sites.[1] Various types of gas sensors have been developed.[2] Among them, the ones with a two-dimensional (2D) planar structure exhibit large surface-to-volume ratios and thus greatly reduced volumes and improved portability. More importantly, because the entire surface of monolayer materials is sensitive to external factors, these materials contain a greater number of active sites.[3] Given these advantages, 2D materials have considerable potential for use in gas-sensing applications.[4−6] Among the 2D nanomaterials being explored, graphene is highly promising for use in gas sensors owing to its desirable chemical and physical properties.[7] However, the fact that graphene does not exhibit a bandgap has limited the application range and pushed the research focus to other graphene-like 2D materials such as transition metal disulfides (TMDs).[8−10] Compared with graphene, TMDs not only possess tunable electronic characteristics[5] but also exhibit a high signal response to external analytes.[11] Furthermore, gas sensors based on TMDs also have the advantages of an ultrafast response, low electron noise, low power consumption, and ultralow detection limit.[12]

As typical TMDs, MoS2 and WS2 are used preferentially as gas-sensing materials because of their high stability and excellent electrical properties.[13,14] To enhance their sensitivity as gas-sensing materials, it is common to modulate their surface characteristics through a chemical treatment (element doping or surface functionalization), by introducing defects, or via strain engineering.[15−18] However, chemical treatments and the introduction of defects can lead to lattice distortion within monolayer materials.[19] Furthermore, with respect to strain engineering, though theoretical calculations have been performed to simulate cases with a uniform strain, it is difficult to achieve an equivalent uniform strain experimentally.[20] Fortunately, transition metals exhibit different coordinations allowing both MoS2 and WS2 to exhibit polymorphs (such as 1T, 1T′, and 2H phases).[21] Significant changes can be achieved in the chemical and physical properties of these materials by altering their phase. These phase-change-induced variations in their properties have been exploited in several applications, such as the catalysis of the hydrogen evolution reaction and improving the reversible capacity of lithium-ion batteries.[22,23] However, the effects of such phase changes on the gas-sensing performance of these materials remain unknown.

Among the polymorphs of monolayer MoS2 and WS2, the 2H phase is the most common one. This phase exhibits a semiconductor-like band structure with a wide bandgap. In contrast, the 1T phase has a typical metal-crystal-like structure. By the dimerization distortion of the transition metal atoms, the metastable metallic 1T phase can be transformed into the 1T′ phase under certain conditions.[24] The 1T′ phase has a narrow bandgap and exhibits semimetallic properties. Moreover, 1T′ MoS2 and WS2 have been identified as quantum spin Hall insulators (QSHIs).[25] In QSHIs, the backscattering phenomenon is suppressed by their time-reversal symmetry when the surface-state electrons encounter impurities.[26] Thus, topological insulators have the advantages of high carrier mobility and low power consumption.[21] With the development of material fabrication technologies, 1T′ TMDs are becoming more and more accessible. Many previous experimental studies have shown simple ways to prepare 1T′-phase MoS2 or WS2.[27−29] These facts make 1T′-phase materials promising candidates for use in large-scale low-dissipation electronic devices.[25]

Herein, keeping both the stability and sensitivity in mind, the effects of phase changes on the gas-sensing performance of 2D materials were evaluated. Using first-principles calculations based on the density functional theory (DFT), the changes in the electronic and energy properties of monolayer MoS2 and WS2 upon interactions with the molecules of common toxic gases (NO2 and NH3) were studied. Detailed theoretical computations that considered the binding energy, theoretical recovery time, band structure, density of states (DOS), total electron density, and electron difference density (EDD) were performed. The results showed that the 1T′-phase materials exhibit higher sensitivity to gas molecules than the 2H-phase ones. This confirmed the non-negligible role of selecting the appropriate phase in improving the gas-sensing performance of devices based on these materials. Our work provides new insights and should help improve the sensing performance of gas-sensitive substrates.

Results and Discussion

After geometrical optimization, the lattice constants were set at a = b = 3.17 Å and a = b = 3.15 Å for 2H monolayer MoS2 and WS2, respectively. Furthermore, the lattice constants were a = 6.53 Å and b = 3.19 Å and a = 6.57 Å and b = 3.23 Å for 1T′-phase monolayer MoS2 and WS2, respectively. These were in good agreement with the previously reported values.[30,31] The adsorption sites are shown in Figure S1. The structure with the lowest negative binding energy was considered to exhibit the most stable adsorption and was thus used for the subsequent investigations of gas adsorption. All the binding energy results are listed in Tables S1–S4. After that, the thermal stability was explored by molecular dynamics (MD) simulations (as presented in Figure S2). After 500 steps (2.5 ps) at 400 K, there was no significant structural or energy change for all these four materials. Thus, the structures are stable thermodynamically, and the thermal stability of these materials is adequate for gas-sensing applications.

Ideally, gas sensors should exhibit high stability, sensitivity, and desorption speed. The thermodynamic stability of the entire system during the process of gas adsorption was considered first. Figure shows the top views of the adsorption models after geometry optimization, along with the bond length, adsorption height, and binding energy. After the geometry optimization process, no new bonds emerged at the interface for all the eight conditions investigated (Figure a–h). The corresponding side views are shown in Figure S3. The N–O and N–H bond lengths for the free, gas-phase NO2 and NH3 molecules were 1.21 and 1.02 Å, respectively. The variations in the bond length or adsorption height after the geometry optimization are shown in Figure i,j, and Table S5; a positive sign means an increase in the bond length or adsorption height, while a negative sign means a decrease. It can be seen from the results that the gases in the 1T′ MoS2 and WS2 systems exhibited more significant changes in their bond lengths and adsorption heights, indicating stronger interactions between the gas molecules and the monolayer 1T′ structure in question. The binding energies of the adsorption systems are shown in Figure k and Table S5. It can be seen that the gas-1T′ systems had lower negative binding energies than those of the 2H ones. A low negative binding energy indicates a strong affinity for analytes, and hence a strong signal response.[11,32,33] Therefore, the 1T′-phase materials show a high signal response while adsorbing gas molecules. In addition, a lower negative binding energy indicates a relatively more stable adsorption system.[34,35] Thus, the gas-1T′ adsorption systems were also more stable than their 2H counterparts.

Figure 1

Top views of (a) NO2–1T′ MoS2, (b) NO2–2H MoS2, (c) NH3–1T′ MoS2, (d) NH3–2H MoS2, (e) NO2–1T′ WS2, (f) NO2–2H WS2, (g) NH3–1T′ WS2, and (h) NH3–2H WS2 systems. (i) Bond lengths of gas molecules in adsorption systems and those of isolated gas molecules. (j) Adsorption heights of gas molecules in adsorption systems and initial height (k) binding energies of adsorption systems.

Next, to examine the sensitivity of the investigated materials, band structure analysis was performed. The Brillouin zone paths of the band structures are shown in Figure S4, while Figure a,d shows the band structures of the isolated 1T′- and 2H-phase materials. The bandgaps of 1T′- and 2H-phase monolayer MoS2 were 0.017 and 1.816 eV, respectively, while that of 2H monolayer WS2 was 1.974 eV and that of 1T′ monolayer WS2 was approximately 0.005 eV. These bandgap values were in good agreement with that reported previously[36−40] (Table S6). According to the results shown in Figure a,d, the bandgaps of the pristine 1T′-phase materials were much smaller than those of the 2H ones, indicating that the 1T′ structures exhibited higher conductivities. Besides, although the wider bandgaps of 2H materials result in a stronger signal response, they also make the system susceptible to high noise.[11] The band structures of the adsorption systems after their geometrical optimization are shown in Figure b,c,e,f. After gas adsorption, the impurity energy levels introduced by the gas molecules change the conductivity of the entire system, thus allowing for the ready detection of the gas molecules. On comparing Figure b,c, it can be seen that the bandgaps of the systems with the 1T′-phase materials were significantly broader as compared with those of their pristine counterparts. This was indicative of changes in the conductivity,[25] which could be detected. In addition, as can be seen from Figure b,c, the energy levels at the band extremes near the Fermi level became smoother, especially in the case of NO2 adsorption. Linghu et al. have suggested that a smoother change in the energy levels near the Fermi level after gas adsorption is indicative of high sensitivity during the detection of gas molecules.[41] Based on this fact, it can be inferred that 1T′-phase materials have better gas sensitivity than 2H-phase materials.

Figure 2

Band structures of (a) isolated 1T′ MoS2 and WS2, (b) NO2-1T′ system, (c) NH3-1T′ system, (d) isolated 2H MoS2 and WS2, (e) NO2-2H system, and (f) NH3-2H system. Insets in (a) are magnified views of encircled parts.

In addition to their band structures, the sensitivities of these gas-sensitive materials were also analyzed based on their DOS values. To better understand the electrical properties of the gas molecules, which were affected by the substrates after the adsorption process, the partial density of state (PDOS) values of NO2 and NH3 were determined, as shown in Figure . The black and red lines represent the PDOS of the N and O atoms, respectively, in free NO2 molecules. Furthermore, the filled areas represent the PDOS of the N and O atoms of NO2 molecules adsorbed onto the surfaces of the substrate materials. Next, Op and Np represent the p orbitals of the O and N atoms, respectively, and Hs represents the s orbital of the H atom. The free, gas-phase NO2 molecule exhibits an asymmetric spin-up and spin-down PDOS owing to the presence of unpaired electrons, while the NH3 molecule shows symmetric spin-up and spin-down PDOS as all its electrons are paired.[42] However, in the case of the NO2 molecules in the adsorption systems, symmetric activated states emerged after the molecules had been adsorbed onto the substrates. These activated states not only induced new peaks or resulted in peak splitting but also promoted the mixing of the orbitals of the gas and the substrate surfaces. In Figure a, the drops in the Op and Np peaks of 1T′ MoS2 at the Fermi level are greater than those of the peaks of its 2H counterpart. Figure b shows that the changes in the case of the group with 1T′ WS2 were more obvious. In particular, for the regions above the Fermi level, new peaks related to Np and Op emerged. Similarly, distinct changes can be seen in Figure c,d, which correspond to the adsorption of NH3 on 1T′ MoS2 and WS2. When NH3 molecules adsorbed on the surface of 2H-phase materials, shown in Figure c,d, the PDOS results were similar to those for the free gas molecules. The changes in the PDOS after adsorption were related to the behavior of the electrons, which was affected by the surfaces of the substrates.[43] Both NO2 and NH3 showed greater electron redistributions after they were adsorbed onto the surfaces of the 1T′-phase substrates, indicating that the 1T′-phase materials underwent stronger interactions with the gas molecules.

Figure 3

PDOS of (a) NO2 adsorbed on MoS2, (b) NO2 adsorbed on WS2, (c) NH3 adsorbed on MoS2, and (d) NH3 adsorbed on WS2. Lines represent PDOS of atoms in free gas molecules. Filled areas represent PDOS of atoms in gas molecules adsorbed onto surfaces of substrate materials.

The degree of overlapping of the PDOS peaks of the gases and the substrate materials was examined to further understand the sensitivity of the substrates to the gas molecules. As shown in Figure a, the overlap regions marked in gray ranged from −4.4 to −2.6 and 0.5 to 1.5 eV in the case of the adsorption of NO2 on the surface of 1T′ MoS2. For the NO2–2H MoS2 system, the overlap region only ranged from −3.5 to −2.2 eV. As shown in Figure b, when NH3 molecules were adsorbed on monolayer 1T′ MoS2, the overlap region extended from −6.3 to −5.5 eV and −1.2 to −0.1 eV. Furthermore, the overlap region extended from −5.7 to −4.8 and −0.5 to 0 eV for the NH3–2H MoS2 system. Therefore, the degree of overlap of the PDOS peaks around the Fermi level was greater when either gas was adsorbed on 1T′ MoS2. In other words, the gas molecules underwent stronger interactions with the surface of 1T′ MoS2.[36,44] Similar conclusions could be drawn in the case of WS2. With respect to the adsorption of NO2 molecules onto monolayer WS2, the gray areas ranged from −4.5 to −2.7 eV, with the range being 0.5–1.9 eV for 1T′ WS2 and −3.5 to −2.2 eV for 2H WS2, as shown in Figure c. Furthermore, as can be seen from Figure , when NH3 molecules were adsorbed onto monolayer WS2, the gray area extended from −6.4 to −5.5 eV. Thus, the range was −1.1 to −0.1 eV for 1T′ WS2 and −5.7 to −4.7 and −0.5 to 0 eV for 2H WS2. Hence, the degree of overlap around the Fermi level was greater when 1T′ WS2 was used as the sensing substrate. The overlapping of the PDOS peaks around the Fermi level is reflective of the strength of the interactions between the gas in question and the sensing surface.[45] This overlapping promotes charge transfer, and thus enhances the adsorption interactions between the gas and the material surface.[34,46,47] Therefore, the extent of charge transfer was greater, and the interactions were stronger between the gases and the 1T′-phase materials.

Figure 4

PDOS of (a) NO2–MoS2 system, (b) NH3–MoS2 system, (c) NO2–WS2 system, and (d) NH3–WS2 system.

Next, the total densities of states (TDOS) of the entire adsorption systems were determined to analyze further the interactions between the gas molecules and the sensing substrates (Figure a,b). Here, we only considered NO2. No obvious change in the TDOS profiles was observed when NO2 molecules were adsorbed onto the surface of 2H MoS2 or WS2. However, there was a left shift in the TDOS at approximately 0.58 and 0.49 eV in the cases of 2H MoS2 and WS2, respectively. This suggests that the extent of charge transfer during the adsorption process was not significant.[33,48,49] In contrast, changes in the TDOS profiles were observed in the cases of the 1T′ MoS2 and WS2 systems after the adsorption of NO2 molecules. A slight increase in the TDOS was observed at the Fermi level (see the magnified image inset in Figure a,b). This suggested that there was significant charge transfer between the gas in question and the surfaces of the 1T′-phase materials.[50] Therefore, the 1T′-phase materials were more sensitive to NO2 than the 2H materials.

Figure 5

TDOS for (a) NO2–MoS2 and (b) NO2–WS2 systems.

For visualizing charge transfer between the gases and the substrates, EDD analysis was performed, as shown in Figure . The pink regions represent the areas with electron accumulation, while the blue regions represent the areas with electron depletion. In the case of the adsorption of NO2 onto the surfaces of MoS2 and WS2, the isosurface value was set as 0.009e/Å3, as shown in Figure a,b,e,f. In the case of the adsorption of NH3 molecules on the surfaces of MoS2 and WS2, the isosurface value was set as 0.0025e/Å3, as shown in Figure c,d,g,h. The charge was transferred away from the substrate surface when NO2 molecules were adsorbed on it. On the other hand, when NH3 molecules were adsorbed on the substrate, the charge was transferred toward the substrate surface and accumulated there. This indicates that electron redistribution on the surfaces of the 1T′-phase substrates was more pronounced than that on the surfaces of the 2H substrates. Greater charge transfer occurred between the gas and the substrate in the cases of 1T′ MoS2 and WS2. The total electron density results are shown in Figure S5. The results further confirmed the occurrence of electron redistribution. Because a greater number of electron density overlaps exist at the interfaces of the gas-1T′ phase material systems, the potential for charge transfer between the gases and the 1T′-phase materials was greater. Hence, the 1T′-phase materials exhibited stronger adsorption with respect to the target gases.

Figure 6

EDD of (a) NO2–1T′ MoS2 system, (b) NO2–1T′ WS2 system, (c) NH3–1T′ MoS2 system, (d) NH3–1T′ WS2 system, (e) NO2–2H MoS2 system, (f) NO2–2H WS2 system, (g) NH3–2H MoS2 system, and (h) NH3–2H WS2 system.

Finally, the recovery times were considered. A short recovery time indicates a high desorption rate, which, in turn, is conducive for sustainable applications.[31,51] The theoretical recovery time (τ) is defined aswhere ν0 is the attempt frequency (s–1), Ebind is the binding energy (eV), KB is the Boltzmann constant (8.62 × 10–5 eV–1 K), and T is the temperature (K).[52] The calculated theoretical recovery times are listed in Table S7. It was found that gas-1T′ phase material systems exhibited reasonable recovery times[48,53,54] (the longest was 13.115 μs). Thus, all the investigated adsorption systems with 1T′ phase materials have moderate desorption rates.

Conclusions

The gas-sensing performances of monolayer 1T′- and 2H-phase MoS2 and WS2 with respect to NO2 and NH3 were analyzed theoretically. Based on the calculation results, it can be concluded that the bond lengths and adsorption heights of NO2 and NH3 molecules change more significantly after the molecules are adsorbed onto 1T′-phase materials than is the case for 2H-phase materials. The fact that the binding energy in the former case was negative also indicated that adsorption systems involving 1T′-phase materials are more stable. Besides, 1T′ MoS2 and WS2 have higher conductivities, and their energy levels near the Fermi level become smoother after gas adsorption. With respect to the entire adsorption systems, the phenomenon of electron redistribution was more pronounced after the adsorption of the gases on the 1T′-phase materials.

Moreover, the charge transfer between the gases and the 1T′-phase materials was greater, suggesting that the adsorption interactions were stronger. Owing to the mechanisms described above, the sensitivities of the 1T′-phase materials to NO2 and NH3 were superior to those of the 2H-phase ones. This superiority was more obvious during the detection of NO2. On the other hand, the theoretically determined recovery times indicated that 1T′-phase materials show moderate desorption rates. Thus, it can be concluded that 1T′-phase materials are promising for use in sensing applications, as they exhibit both high sensitivity and rapid desorption. Hence, the selection of the appropriate phase of the substrate is an important factor for enhancing the sensing performance of gas sensors. More relevant studies are expected to further explore the effect of phase selection on gas-sensing properties. For theoretical ones, phase selection could be combined with other treatment methods (such as doping) to further enhance gas-sensing properties; for experimental studies, gas sensors using 1T′-phase materials are highly expected.

Computational Details

The electronic structure calculations were performed based on the Perdew–Burke–Ernzerhof generalized gradient for the exchange–correlation energy, the double numerical plus polarization basis set, and dispersion correction.[55] All the binding energy values were obtained using Grimme’s empirical dispersion correction (DFT-D).[56,57] The DFT semicore pseudopod core treatment replaced the core electrons with single productive potential. This introduced some degree of relativistic correction within the core.[58,59] A vacuum layer of 13 Å was set to eliminate the interlayer interactions. The Brillouin Zone was set as a 4 × 4×1 Monkhorst–Pack k-point for the geometrical optimizations. The energy convergence precision was 1.0 × 10–5 hartree, the maximum atomic force was 0.002 hartree/Å, and the maximum displacement was 0.005 Å. The direct inversion in an iterative subspace size was set as 6, in order to accelerate the self-consistent field convergence.

The 4 × 4 × 1 supercell of monolayer WS2 consisted of 16 W atoms and 32 S atoms. All sets of the WS2 supercell were the same as those of the MoS2 supercell. The thermal stability of these materials was simulated by performing MD simulations at 400 K using the simple Nosé–Hoover heat bath scheme. The time step was 5.0 fs, and the total simulation time was 2.5 ps. The initial adsorption heights of the gas molecules were uniformly set as 3 Å in the cases of the 1T′- and 2H-phase systems. The equation for the binding energy iswhere Esystem, EMS2, and Egas are the total energies of the adsorption system, the isolated monolayer MoS2 or WS2, and a free gas-phase NO2 or NH3 molecule, respectively, for the same cell size. A negative binding energy value meant that an adsorbing force existed between the gas molecule and the material in question. Furthermore, a lower negative binding energy indicated stronger adsorption within the system.