Highly Sensitive NO2 Detection by TVS-Grown Multilayer MoS2 Films.

Two-dimensional layered materials have been investigated for sensor applications over the last decade due to their very high specific surface area and excellent electrical characteristics. Although grain boundaries are inevitably present in polycrystalline-layered materials used for real applications, few studies have investigated their effects on sensing properties. In this study, we demonstrate the growth of two distinct MoS2 films that differ in grain size by means of chemical vapor deposition (CVD) and thermal vapor sulfurization (TVS) methods. Transistor-based sensors are fabricated using these films, and their NO2 sensing properties are evaluated. The adsorption behavior of NO2 on MoS2 is considered in terms of the Langmuir isotherm, and the experimental results can be well fitted by the equation. The CVD-grown film exhibits electrical properties 1-2 orders of magnitude superior to those of the TVS-grown one, which is attributed to the large grain size of the CVD-grown film. In contrast, the sensitivity to NO2 is unexpectedly found to be higher in the TVS-grown film and is of the same order of a previously reported record value. Transmission electron microscopy observations suggest that the TVS-grown film consists of multiple rotationally oriented grains that are connected by mirror twin grain boundaries. Theoretical calculation results reveal that the adsorption of NO2 on the grain boundary that we modeled is equal to that on the ideal basal plane surface of MoS2. In addition, the porous structure in the TVS-grown film may also contribute to enhancing the sensor response to NO2. This study suggests that a highly sensitive MoS2 sensor can also be fabricated by using a polycrystalline film with small grain size, which can possibly be applied to other two-dimensional materials.

Introduction

Two-dimensional materials have attracted attention over the years from researchers in various fields owing to their unique properties. Among them, metal chalcogenides are expected to be good candidates for electronic devices since they exhibit a wide variety of electronic and optical properties in accordance with their elemental composition and crystal structure.[1] Molybdenum disulfide (MoS2) has also been widely studied for its great potential in a field effect transistor (FET).[2,3] In addition to the high on/off current ratio and moderate electron mobility, the FET shows electric responses upon exposure to specific gas species, which has been the motivation for studying its potential as a material in sensor applications.[4−7] The MoS2 films used in highly sensitive nitrogen oxide (NO2) sensors are usually prepared by the exfoliation of bulk crystal or chemical vapor deposition (CVD).[8−11] However, these methods have difficulties in terms of thickness control and film uniformity that significantly affect the device fabrication yield. On the other hand, thermal vapor sulfurization (TVS), in which MoS2 is grown by heating a metallic Mo film in a sulfur vapor environment, has advantages in terms of the above-mentioned problems because the thickness and uniformity of the resulting MoS2 film are basically determined in accordance with the initial geometry of the pre-deposited Mo film.[12,13] Therefore, the TVS-grown film is likely to be compatible with the mass production of MoS2 devices. However, it typically shows relatively poor electric properties due to its small grain size. Specifically, it has a considerably denser grain boundary (GB) than the CVD-grown film. Although such films that possess polycrystalline nature seem to lack the potential to be used in a low-ppb (parts per billion) NO2 detectable sensor, few studies have examined the sensing performance of TVS-grown films.[14]

In this study, we fabricate an FET-based sensor using the TVS-grown MoS2 film and evaluate its electric properties and sensor response to NO2 gas for comparing its sensing ability with that of the CVD-grown film. Although most of the previous research has focused on the NO2 adsorption behavior only on the ideal basal plane surface of MoS2 to explain the sensor response, the effect of grain boundaries in MoS2 films is still unrevealed. In this regard, it is important to investigate the sensing ability of polycrystalline MoS2 containing a large amount of GB that would inevitably be present in the film while fabricating a sensor for practical use. To evaluate the net sensor response to low-ppb NO2, the drain current change when exposed to NO2 is measured in a nitrogen (N2) atmosphere for excluding the effect of oxygen and humidity on the sensing properties. In addition, all electrical measurements are conducted in a stainless steel chamber under dark conditions to avoid extrinsic effects induced by light illumination that can cause photocurrent generation[15,16] and desorption of adsorbed gas molecules on the MoS2 surface.[17,18] The TVS-grown film is found to have lower electrical properties than the CVD-grown film. Contrary to our expectation, however, the TVS-grown film shows higher response to NO2 than the CVD-grown film. The sensor exhibits clear response to NO2 with concentrations as low as 7 ppb and more than 90% change in drain current without any contribution of the extrinsic effect, whose responsivity is of the same order as a previously reported record value achieved with the support of light illumination.[19,20] To verify this interesting finding, we then demonstrate transmission electron microscopy (TEM) observations to investigate the grain structure of the TVS-grown film. The film is suggested to be polycrystalline, although it consists of well-aligned grains with mirror twin GBs. Theoretical calculations are performed to evaluate the adsorption behavior of NO2 on the MoS2 surface with and without the GB. The results reveal that the adsorption energy on the GB is estimated to be comparable to that on the ideal MoS2 surface. Furthermore, the porous structure found in the film possibly contributes to enhancing the NO2 adsorption properties. These results suggest that the highly crystalline film with a large grain size is not essential to improve the sensing performance of a MoS2-based sensor. This is rather favorable to sensor fabrication considering that the TVS method can synthesize a wafer-scale MoS2 film with uniform thickness at a relatively low cost. Our findings can help to clarify the mechanisms underlying NO2 sensing by MoS2 films and expand the strategies for developing an ultrasensitive chemical sensor utilizing two-dimensional materials.

Methods

MoS2 Growth

MoS2 growth was performed in a quartz tube furnace where three heating zones can be controlled independently. Al2O3(0001) substrates (sapphire C-face wafer) with mirror-polished surfaces were used for growth substrates. Prior to the growth, three cycles of Ar gas flushing followed by evacuation were carried out to reduce the concentration of residual oxygen in the furnace.

CVD growth was demonstrated under atmospheric pressure with 500 sccm (standard cubic centimeters per minute) of Ar used as a carrier gas. A continuous film of single-layer MoS2 was successfully synthesized on the sapphire substrate. MoO3 powder (99.95%) and high-purity elemental sulfur (99.9999%) were used as precursors. The growth setup is schematically illustrated in Figure a. The sulfur source is placed in the furnace upstream from MoO3 and the substrate where the temperature is controlled at around 140–180 °C during growth. The substrate is located downstream next to a MoO3-contained quartz boat. The growth zone temperature was 800 °C, at which the evaporated MoO3 reacts with the sulfur vapor, resulting in the formation of a MoS2 film on the surface of the substrate.

Figure 1

(a) Schematic illustration of the CVD growth furnace setup. The substrate and MoO3 powder are placed in a heating zone, and sulfur is placed upstream from the substrate. (b) Optical microscopy image of the MoS2 film grown on the substrate. (c) Dark-field TEM image of a CVD-grown MoS2 film. The inset shows the corresponding SAED pattern. Scale bars shown in (b,c) are 50 μm and 500 nm, respectively. (d,e) Raman and PL spectra of the CVD-grown MoS2 film taken on the growth substrate. The inset in (e) is a magnified image of Raman peaks.

TVS growth was conducted in the same furnace used for CVD. The Mo film (1.0 nm thick) was deposited on sapphire substrates at room temperature. The Mo source was evaporated by electron bombardment (EB) in a high-vacuum chamber. The elemental sulfur was used for Mo sulfurization. The Mo sample and the sulfur placed in a quartz boat were introduced into the furnace. The sample was first annealed at 300 °C for 30 min in a diluted H2/Ar mixture at atmospheric pressure to reduce native oxides on the Mo film. Then, the sample temperature was elevated up to 1000 °C with an Ar flow of 500 sccm. Meanwhile, the temperature of sulfur, which was placed upstream from the Mo sample, was also raised to 140 °C. After annealing the sample for 30 min, the growth temperature was maintained at 1000 °C, and the furnace was then cooled down under constant Ar flow.

Device Fabrication and Characterization Methods

After the growth, the MoS2 film was transferred onto a Si substrate with a 90 nm-thick thermal SiO2 layer by the conventional polymethyl methacrylate (PMMA) transfer method for back-gated FET fabrication.[21] Standard photolithography followed by oxygen plasma reactive ion etching was used to pattern the MoS2 film. The Ti/Au (10/50 nm) source and drain electrodes were deposited by EB evaporation. The typical channel length of the MoS2 FET is around several micrometers. All electrical measurements were performed using a semiconductor parameter analyzer (Agilent 4156C). Unless otherwise noted, two-terminal I–V (current–voltage) curves were measured for the MoS2 FETs in a N2 atmosphere at room temperature after thermal annealing at 200 °C under vacuum to evaluate the electric properties of the native MoS2 film. A sensing test was demonstrated using a home-built apparatus that is described in detail elsewhere.[22] Diluted NO2 was used as a test gas for carrying out measurements using the calibration gas generating equipment (GASTEC PD-1B). The sensing test was performed by monitoring the drain current of the FET-based sensor while NO2 was supplied. The back-gate bias was set to the value where the drain current was in the subthreshold region. The samples were annealed under vacuum conditions before every measurement to refresh the sensor. All measurements are conducted at room temperature under dark conditions to avoid light illumination effects on MoS2.

Photoluminescence (PL) and Raman spectroscopies (Horiba Jobin Yvon LabRAM HR-800) were performed at room temperature to confirm MoS2 formation with an excitation laser wavelength of 488 nm. X-ray photoelectron spectroscopy (XPS) was performed at room temperature using a Kratos AXIS-HSi system equipped with monochromatized Al Kα radiation source (photon energy 1486.6 eV). TEM and scanning TEM (STEM) images were taken at an acceleration voltage of 80 or 200 kV with an FEI Tecnai Osiris. The TEM specimens were prepared using a holey carbon gold grid by means of the above-mentioned transfer method.

Theoretical Calculations

Electronic structure calculations were performed using the first-principles calculation code, PHASE.[23] The exchange–correlation potential was treated with generalized gradient approximation (GGA) using the Perdew–Burke–Ernzerhof functional.[24] To describe the long-range interaction, the vdW-D2 method was adopted to include the van der Waals interactions.[25] Valence electrons are generated in Mo 4p64d55s1 and S 3s23p4 configurations. The cutoff energies were set on 25 Ryd for plane-wave expansion and 230 Ryd for integration of charge density in real space. The Bader charge analysis was carried out to evaluate the charge transfer from the MoS2 surface to a NO2 molecule.

Results and Discussion

CVD-Grown MoS2 Film

The schematic of the experimental setup for CVD growth is shown in Figure a. The MoS2 film is synthesized in a quartz tube furnace where MoO3 powder and elemental sulfur crystal are used as source materials (see Methods for details). Figure b shows an optical microscopy image of the MoS2 film grown on the sapphire substrate. MoS2 forms triangular islands, and their coverage increases to the right in the image, merging to form a continuous film. Previous studies have also reported that the CVD method has difficulty in precisely controlling the uniformity of growth coverage and the thickness of the MoS2 film.[26,27] The film in the full-coverage region is characterized by TEM analysis to evaluate the grain size of the CVD-grown MoS2 since it will be utilized for fabricating sensor devices as described later. A dark-field TEM image and the corresponding selected area electron diffraction (SAED) pattern in the inset are shown in Figure c. The typical grain size is found to range from 3 to 5 μm from the TEM observations where a single grain appears as a bright region in the image. Raman and PL spectroscopies are performed to analyze the layer number and crystallinity of the CVD-grown film. The Raman spectrum shows two prominent peaks at 387 and 406 cm–1, which correspond to the E2g1 and A1g modes of MoS2.[28] As shown in Figure S1a, no obvious Raman features related to structural defects in MoS2 are seen in the low-wavenumber region of the spectrum.[29] The full widths at half-maximum (FWHMs) are extracted by fitting the E2g1 and A1g peaks and are estimated to be 2.47 and 3.43 cm–1 (see Figure S1), respectively, indicative of the crystallinity of MoS2 being higher than in previous reports.[30,31] Furthermore, the layer number of the film is identified as a single layer from a distance of 19.1 cm–1 between the two peaks.[32,33] The film also shows a strong PL peak at 655 nm whose intensity is considerably higher than those of Raman peaks as shown in the inset of Figure e, which is due to the direct optical band gap inherent in single-layer MoS2, supporting the result of Raman spectroscopy.

After transferring the CVD-grown film onto a Si/SiO2 substrate, back-gated FETs were fabricated in a standard photolithography process. Note that the typical MoS2 channel size of the FET is around several micrometers in length, which is almost the same as the grain size. Before electrical measurement, the samples were first annealed under vacuum to eliminate the effect of surface adsorbates such as H2O and O2 molecules on the electrical properties of the FETs. Unless otherwise noted, all measurements were carried out in a N2 atmosphere under dark conditions at room temperature in a two-probe configuration. The FETs typically exhibit an n-type semiconducting behavior as can be seen in Figure b.[34] The electron mobility of the FET reached 14.9 cm2 V–1 s–1 with a current on/off ratio of more than 107, which compares favorably with the reported values.[35,36] Then, the sensing properties of the FET-based sensor are tested by measuring the change in drain current value when exposed to various concentrations of NO2. Figure c shows the normalized drain current (Id/I0) of the sensor with time for 7, 22, and 73 ppb NO2 exposure. The NO2 exposure leads to a rapid decrease in drain current in which the slope becomes steeper as the concentration increases. Responses to NO2 and other gas species were also measured to evaluate the selectivity as shown in Figure S2, which is key to the gas sensor. The threshold voltage shift in the positive direction is clearly observed in the I–V curves obtained before and after NO2 exposure as can be seen in Figure b, which explains the above-mentioned drain current change and corresponds to hole doping by NO2. The sensor response (ΔId/Id0) after 1 min exposure to NO2 is plotted as a function of concentration in Figure d. If it is assumed that the response depends on the NO2 coverage on the MoS2 surface, the sensing behavior relative to the NO2 concentration (CNO2) is explained by the Langmuir isotherm equation. The obtained response values are fairly fitted by the equation curve. The plot of CNO/(ΔId/Id0) on the concentration in the inset of Figure d also agrees with linear fitting, which is another expression of Langmuir isotherm, further illustrating the stability and repeatability of the sensor. This result also indicates either that the coverage of the adsorbed NO2 molecules cannot exceed one monolayer on the MoS2 surface or that even if the molecules form the second and subsequent adlayers on the surface, they do not contribute to the hole doping to MoS2. This suggests that the doping is caused by charge transfer from NO2 that directly adsorbs on MoS2 as will be discussed later. Here, NO2 adsorption on the MoS2 surface can be described as the following reaction equation: Sfree + NO2 ⇄ SNO (SNO and Sfree denote the site on the MoS2 surface with and without an adsorbed NO2 molecule, respectively). Assuming that the NO2 gas is constantly supplied in the sensing chamber, the above equation can be written as Sfree ⇄ SNO. Based on the first-order reversible reaction model, the drain current during the sensor response to NO2 gas follows I = Ieq + (Iini – Ieq) exp(−kt), in which Iini and Ieq are initial and equilibrium values of the drain current, respectively, k is the rate constant, and t is the elapsed time.[37] The fitted curves match the measured data well as shown in the inset of Figure c, and the rate constants are determined to be 0.0035 for 7 ppb, 0.0072 for 22 ppb, and 0.0165 for 73 ppb. The sensor shows a clear response to NO2 with a concentration as low as 7 ppb and exhibits more than 90% change in the drain current. Since the signal-to-noise (S/N) ratio is estimated to be ∼35, a 1 ppb or lower concentration of NO2 could be detected. However, the sensor after the measurement could barely recover its original property at room temperature even when exposed to N2, consistent with sensors reported elsewhere.[8−10] This implies that NO2 molecules are strongly adsorbed on the MoS2 surface. In this regard, photo-induced effects by light illumination on the sensor have been suggested to improve the recovery speed.[38,39]

Figure 2

(a) Schematic illustration of an FET-based MoS2 sensor (Vd: drain voltage, Vg: gate voltage). (b) Drain current (Id) as a function of Vg of a CVD-grown MoS2 FET with a channel length of 4 μm. The Vd was 1 V. The inset shows the Id – Vg curve displayed with the logarithmic axis. (c) Normalized Id of the CVD-grown MoS2 sensor when exposed to 7, 22, and 73 ppb of NO2. The inset shows the corresponding fitted curves. (d) Sensor response ((ΔId/Id0) after 1 min exposure to NO2 plotted as a function of the concentration with the fitted curve using the Langmuir isotherm equation. The inset shows the corresponding linear fitting of the measured data.

TVS-Grown MoS2 Film

Next, we evaluate the sensor utilizing the TVS-grown film. The MoS2 film is synthesized by sulfurizing the thin Mo film at an elevated temperature. After Mo film deposition, the sample was introduced into the quartz tube furnace. The growth process is schematically illustrated in Figure a. The Mo film reacting with vaporized sulfur at an elevated temperature gives rise to the MoS2 film where the number of layers varies accordingly depending on the thickness of the pre-deposited Mo film. An optical microscopy image of MoS2 transferred onto a pre-patterned SiO2/Si substrate shows a continuous film with uniform color contrast over the millimeter-scale range. The Raman spectrum of the as-grown MoS2 film is shown in Figure c. Two prominent peaks of MoS2 are clearly observed. The separation between the peaks is estimated to be 24.1 cm–1, indicating multilayer MoS2 that consists of four to five layers.[32,33] The peak located at 286 cm–1 as shown in Figure S3a appears when the film is a multilayer MoS2.[40] The LA(M) peak at ∼227 cm–1 is not visible in the spectrum. This is reported to be most prominent among the peaks that are induced by defects in MoS2, whose intensity is proportional to the density of the defects. Although the faint defect-induced peaks (150–200 cm–1) are visible,[29] their intensity is fairly low compared to the two prominent E2g1 and A1g peaks. As will be described later, these would be attributed to a certain number of structural defects present at twin GBs in the TVS-grown film. The stacking order of the film is found to be 2H phase, which is deduced from the Raman feature where the vibration modes of the 1T phase are barely seen.[41] Since the FWHM of the E2g1 mode is known to be relatively insensitive to the number of layers,[42] it could be used as an index for comparing the crystalline quality. The FWHM values of E2g1 and A1g of our film are found to be 3.0 and 3.2, respectively (Figure S3), which are narrower than previously reported values taken from TVS-grown films.[43−46] Furthermore, single-layer MoS2 films grown by CVD have exhibited the values in the range of 3.2–3.8 for E2g1 and 3.7–6.8 cm–1 for A1g,[47−49] suggesting that the crystalline quality of our TVS-grown film is comparable to that of the CVD-grown films in the previous reports. The chemical composition of the MoS2 film was characterized by XPS (see Figure S4). The elemental ratio S/Mo of the TVS-grown film was found to be ∼2.1, which is almost equal to the value of the CVD-grown film (data not shown). This is indicative of the fact that the amount of sulfur vacancies in the films is negligible, which is consistent with the results of Raman measurement that showed high crystallinity of the TVS-grown film. Therefore, the contribution of the vacancies to NO2 sensing is considered to be fairly small.

Figure 3

(a) Schematic illustration of the growth procedure and furnace setup for the TVS method. The substrate is placed in a heating zone, and sulfur is placed upstream from the substrate. (b) Optical microscopy image of the TVS-grown MoS2 film after being transferred onto a Si/SiO2 substrate. The scale bar is 300 μm. (c) Raman spectrum of the TVS-grown MoS2 film taken on the growth substrate. (d) Id as a function of Vg of a TVS-grown MoS2 FET with a channel length of 4.5 μm. The Vd was 1 V. (e) Normalized Id of the TVS-grown MoS2 sensor when exposed to 7 ppb of NO2. The inset shows the corresponding fitted curves.

The MoS2 FETs were fabricated on a SiO2/Si substrate to characterize their electronic properties. First, room-temperature I–V curves are measured for all devices fabricated to evaluate the electrical properties of the TVS-grown film. All the FETs exhibit an n-type semiconducting behavior, as observed in our CVD-grown film. The I–V curves measured in an air atmosphere (not shown) exhibited electron mobilities ranging from 0.1 to 3.0 cm2 V–1 s–1 at maximum and a current on/off ratio of 105–106, which are larger than the previously reported values of the TVS-grown film.[50,51] Note that their values can be further improved by measuring the devices in an inert atmosphere where MoS2 can avoid unintentional counter-doping induced by the absorbed H2O and O2 molecules.[52−54] The MoS2 film synthesized from the Mo film with thickness thinner than 1 nm exhibits higher electric resistance (not shown), which may be due to the insufficient surface coverage and/or the small grain size of MoS2. Mo film evaporation from the substrate surface at the growth temperature leads to reduced MoS2 film thickness. Also, in contrast to layer-by-layer growth for CVD,[55] TVS gives rise to multilayer MoS2 grains before the lateral growth completes the formation of the continuous layer, which leads to a distribution in the number of layers of the resultant film.[13]Figure d shows a typical I–V curve of the sensor measured in a N2 atmosphere after vacuum annealing. The electron mobility and the on/off ratio of the FET were 2.2 cm2 V–1 s–1 and 107, respectively. Then, we examine the gas sensing property of the TVS-grown MoS2 sensor. The sensor is found to detect NO2 with concentrations as low as 7 ppb and exhibited resistivity change by an order of magnitude, as shown in Figure e. The repeatability of the sensor was confirmed by obtaining I–V curves before and after the recovery. As can be seen from Figure S5, the I–V characteristics of the sensor after thermal annealing are completely returned to that of the initial state. The curve fitting of the sensor response as shown in the inset of Figure e was found to be deviated from the experimental data. As will be mentioned later, this deviation may be due to the presence of GBs or edges on the TVS-grown film where the adsorption property could differ from that on an ideal surface. However, the rate constant derived from the fitting curve reached 0.0069, which is almost twice the value of the CVD-grown film, which is indicative of having higher sensitivity than the CVD-grown film. Therefore, NO2 with a concentration of several hundreds of ppt (parts per trillion) or lower would probably be detectable. This is comparable in NO2 sensitivity to recent reports where the detection limit reached 100 ppt. It is noteworthy that our result is obtained at room temperature under dark conditions, while the above record value was achieved with the assistance of the extrinsic effects caused by red light illumination[19] or forming MoS2/ZnO nanocomposites in addition to near-infrared light illumination.[20] The sensor performances of the MoS2-based sensors for NO2 detection reported in previous studies are listed in Table S1. This is unexpected for the TVS-grown film, which usually exhibits inferior electronic properties to the exfoliated and CVD-grown ones. The typical size of the MoS2 grain of the TVS-grown film is in the range of several tens of nanometers,[12,46] which is 2 orders of magnitude smaller than that of the CVD-grown film. Considering the channel length of the FET, a significant amount of GB is probably present on the surface of the MoS2 sensor. Although the adsorption properties of gas molecules could be affected by GBs where the atomic structure is distorted, to the best of our knowledge, there have been no studies addressing this.

The TEM observation was then performed to characterize the grain structure in the TVS-grown film. The TEM sample was prepared by the conventional PMMA transfer process as described in the Methods section. Figure a shows the TEM image of the sample where the bright area corresponds to the MoS2 film transferred on the grid. A folded part observed at the film edge clearly shows that the film consists of four to five layers as can be seen in Figure b, which is consistent with the results of Raman spectroscopy. The 2H phase of MoS2 is confirmed again by observing the honeycomb lattice structure as shown in Figure S6, which is characteristic of the 2H stacking.[56] The SAED patterns obtained at the corresponding spots marked by circles exhibit a single set of hexagonal spots originated from a crystal structure of MoS2, indicating that the film grew laterally on the substrate and contains no rotational grain in the spot area (spot size: 1 μm). Furthermore, all the SAED patterns are found to be oriented in the same direction as displayed by the dashed lines. Note that the SAED patterns obtained at the areas located 0.5 mm or more apart on the same sample are oriented in the same angle as shown in Figure S7. These results suggest that all MoS2 layers in the film were epitaxially grown on the substrate in wafer scale, whereas the TVS growth of MoS2 generally results in a polycrystalline film with rotational grains even on a crystalline substrate.[57] However, it remains unclear whether the film consists of a single grain or multiple grains with a mirror twin GB because both can produce a single set of hexagonal spots in their electron diffraction patterns. Here, the hexagonal spots can be divided into two families, namely, [1̅100] and [11̅00], since the MoS2 crystal possesses a three-fold rotational symmetry. They are known to have a slight difference in their spot intensity that could be used to distinguish the crystal orientation of twinned grains (rotated 60 or 180° with respect to each other).[58] However, this method can no longer be used for multilayered 2H-MoS2 because the two layers in the stacking unit cell are inversely oriented, which is expected to break the three-fold symmetry and consequently produce hexagonal spots with an equivalent intensity. In fact, little difference was found in the spot intensity in our observation. Therefore, while the crystalline nature of the film is difficult to determine from the experimental results, we suppose that the film consists of twinned grains for two reasons: (1) MoS2 has been found to have an epitaxial relationship with the sapphire surface, giving rise to two preferential orientations related by 60° rotation.[27] Considering the theoretical prediction reported that MoS2 aligned along its preferential orientations grows by almost the same amount on the sapphire substrate,[59] this should be the case with our TVS-grown film. (2) There have been reports that the electrical properties of MoS2 are much less affected by the mirror twin GBs than by the tilt GBs.[60,61] The electrical properties of our film were poorer than those of the CVD-grown film but still better than those of the previously reported TVS-grown films. This could be explained if our TVS-grown film is assumed to be composed of the twinned grains. Although we could not directly observe the atomically resolved image of twin GBs by TEM, a certain number of line defects were found in the lattice images where the MoS2 crystal structure is distorted (Figure S8). This seems to be due to the presence of twin GBs in the topmost MoS2 layer.

Figure 4

(a) TEM image of a TVS-grown MoS2 film transferred onto a holey carbon TEM grid. (b) TEM image taken at the edge of a folded film. (c) SAED patterns taken at the corresponding areas marked in (a). Scale bars shown in (a,b) are 2 μm and 5 nm, respectively.

We propose the structure models of MoS2 with and without the twin GB and perform first-principles calculations to explore the NO2 adsorption behavior on the surface. In addition, the charge transfer between MoS2 and the adsorbed NO2 molecule is compared for both models. Details of the calculation are described in the Methods section. Figure a,b shows the most stable adsorption configuration of NO2 on an ideal MoS2 surface. The configuration of this model is in good agreement with previous theoretical results.[62] The adsorption energy (Ea) for this model is determined to be 0.202 eV. All the models that we considered are summarized with the corresponding adsorption energies in Figure S9. The models in which NO2 oriented toward MoS2 with the O atom pointing down are commonly found to be stable for each adsorption site. To investigate the electronic properties of the most stable configuration, the total density of states (DOS) and the electronic band structure are analyzed. As shown in Figure c,d, an unoccupied state located 0.5 eV above the valence band maximum (VBM) is introduced by adsorbed NO2,[62,63] as can be seen from the comparison with that of ideal MoS2 before NO2 adsorption (Figure S10). Then, Bader analysis is performed to estimate the charge-transfer value. The adsorbed NO2 is found to accept 0.11e from MoS2, indicating hole doping induced by NO2, and well explains the experimental results.

Figure 5

(a) Top and (b) side views of the most favorable configuration for NO2 on ideal MoS2. (c,d) The corresponding DOS and the band structure for the model shown in (a,b). (e) Model of MoS2 with the twin GB (between dashed lines). Triangles are shown as a guide to see the orientation of MoS2 grains. (f) Top and (g) side views of the most favorable configuration for NO2 on the twin GB shown in (e). (h) The corresponding DOS for the model shown in (f,g). The red (blue) lines in (c,d,h) correspond to the up-spin (down-spin) states, and the dashed line denotes the Fermi level.

Next, we constructed the model of MoS2 with a twin GB as shown in Figure e by referring to the atomic structure that is commonly observed in previous TEM observations.[58,61] Two grains rotated 60° from one another are connected at the boundary at 20° off of the zigzag direction of MoS2. The boundary is formed from periodic 4- and 8-membered rings whose periodicity varies in accordance with the angle relative to the zigzag direction. The most favorable configuration of NO2 on the twin GB is shown in Figure f,g. The others considered for the calculation are summarized in Figure S11 for comparison. It is found that Ea strongly depends on the adsorption site on the GB, which sharply contrasts to the ideal surface. This implies that the sensing properties are strongly affected by the presence of the GB depending on its atomic configuration. The DOS shows that localized states are created around 0 and 1 eV within the band gap of ideal MoS2. They are attributed to the structural defects in the twin GB, causing an upward shift of the Fermi level due to the n-doping effect of the periodic 8-4-4 ring structure.[58] On the other hand, the adsorbed NO2 state is introduced at ∼0.5 eV above the VBM as is the case of the ideal surface, whereas it is obscured by the presence of the defect states. The most stable adsorption site gives the Ea of 0.217 eV, which is slightly larger than the value for the case of the ideal surface. It derives a charge-transfer value of 0.11e, which is comparable to that on the ideal surface. Therefore, it turned out that an equal amount of hole doping can occur even on a polycrystalline MoS2 consisting of multiple twin grains. A large MoS2 grain is not always necessary for the sensor to be able to detect NO2 with high sensitivity, which supports the validity of our experimental results.

Another possible reason for higher sensitivity in the TVS-grown film than in the CVD-grown one is the high adsorption of NO2 onto the edge sites of the film. The TVS-grown film exhibits a characteristic feature of many nanometer-sized pores that appear as dark in the STEM annular dark-field image (Figure S12). Enhancement of NO2 adsorption properties along the edge sites has been reported to explain the superior response in a vertically aligned MoS2 film compared with the horizontally aligned one.[64] A considerable amount of edge sites present in our porous film could also play a significant role in detecting NO2 adsorption, resulting in enhancing the sensor performance. The gas adsorption behavior on the tilt GB between rotational grains, which is more typical for a polycrystalline MoS2 film, will also need to be further studied to clarify the effect of crystallinity on gas sensing properties from the sensor application point of view.

Conclusions

We demonstrated the comparative study of the NO2 sensing capability of MoS2 synthesized by CVD and TVS methods. The Langmuir isotherm model well described the NO2 adsorption behavior on the MoS2 surface that we observed. Although the TVS-grown film has electrical properties inferior to those of the CVD-grown one, the responsivity of the sensor is found to be higher for the former rather than for the latter. From TEM observations, it was supposed that the TVS-grown film is polycrystalline consisting of rotationally oriented twin grains. Theoretical calculation revealed that the NO2 adsorption on the twin GB that we assumed was estimated to be comparable to that on the ideal basal plane surface. In addition, the edge sites, which are present in the porous structure of the TVS-grown film, could possibly play a crucial role in enhancing the adsorption property, leading to a rise in sensor response to NO2. The findings of this study provide new insights into developing chemical sensors using MoS2 and other two-dimensional materials.