Gas Sensors Based on Mechanically Exfoliated MoS2 Nanosheets for Room-Temperature NO2 Detection

The unique properties of MoS2 nanosheets make them a promising candidate for high-performance room temperature gas detection. Herein, few-layer MoS2 nanosheets (FLMN) prepared via mechanical exfoliation are coated on a substrate with interdigital electrodes for room-temperature NO2 detection. Interestingly, compared with other NO2 gas sensors based on MoS2, FLMN gas sensors exhibit high responsivity for room-temperature NO2 detection, and NO2 is easily desorbed from the sensor surface with an ultrafast recovery behavior, with recovery times around 2 s. The high responsivity is related to the fact that the adsorbed NO2 can affect the electron states within the entire material, which is attributed to the very small thickness of the MoS2 nanosheets. First-principles calculations were carried out based on the density functional theory (DFT) to verify that the ultrafast recovery behavior arises from the weak van der Waals binding between NO2 and the MoS2 surface. Our work suggests that FLMN prepared via mechanical exfoliation have a great potential for fabricating high-performance NO2 gas sensors.


Introduction
In recent years, transition metal dichalcogenides (TMDs) have attracted much interest due to their unique layered structure, electronic and energy storage properties, which can be exploited in numerous devices such as sensors, field-effect transistors and supercapacitors [1][2][3][4]. MoS 2 , as the frontrunner in the TMDs family, has also been extensively investigated as a potential gas-sensing material because of its tunable band gap, large surface-to-volume ratio, and various active sites [5][6][7]. Gas-sensing properties are closely related to the size of the gas-sensing materials. According to reports, all electrons inside the gas sensing channel can be affected by the adsorbed gas when the grain size is smaller than two times the Debye length (so-called grain-size control). The space-charge layer then penetrates into the whole sensing channels, and the response is drastically promoted [8,9]. Structurally, MoS 2 is a layered material in which S-Mo-S atoms are closely packed in a hexagonal arrangement, and each neighboring layer is connected by van der Waals forces [10]. Due to the strong intra-layer interactions and the relatively weak interactions between these layers, the synthesis of single or few-layer MoS 2 nanosheets (FLMN) becomes possible by mechanical exfoliation from bulk MoS 2 [11,12]. With this method which is the easiest and fastest way to produce the pristine, highly exhibits much higher electronic mobility due to lower Schottky barriers, which makes it more attractive for gas sensing [14][15][16]. Moreover, it is reported that the physical adsorption of gas molecules on MoS2 surface can overcome the shortcoming of difficult desorption from twodimensional materials [17], which is beneficial to improve the recovery characteristics of gas sensors. Hence, the FLMN prepared via mechanical exfoliation show great potential in fabricating highperformance NO2 gas sensors.
In this work, the FLMN prepared via mechanical exfoliation is transferred to an Al2O3 ceramic substrate with Ag-Pd interdigital electrodes. Exfoliated MoS2 nanosheets are interconnected among interdigital electrodes to form sensitive channels. The planar gas sensors based on FLMN show high responsivity to NO2 and ultrafast recovery behavior without heating unit. We hope that this work can provide a useful guideline for the application of two-dimensional (2D) MoS2 in high-performance gas sensors.

Experimental Details
A typical mechanical exfoliation process is shown in Figure 1. The bulk MoS2 crystal and scotch tape were purchased from XFNANO Materials Tech Co. (Nanjing, China). First, a piece of scotch tape was adhered onto a bulk MoS2 crystal for about 5 s and then the scotch tape was removed carefully with as small angles as possible. Secondly, the scotch tape with the MoS2 flakes was folded and separated repeatedly many times to get thin MoS2 flakes, and a metallic luster can be clearly seen on the scotch tape surface. Then, the scotch tape with MoS2 thin flakes adhered tightly onto a clean Al2O3 ceramic substrate with Ag-Pd interdigital electrodes, and this adhesion state was maintained for 6 hours before the scotch tape was removed. Subsequently, the acetone was dripped onto the substrate to remove adhesive residue from the scotch tape. Finally, an FLMN gas sensor without the heating unit was obtained and used for further characterization. The morphologies of exfoliated MoS2 nanosheets were observed by scanning electron microscope (SEM, Hitachi SU5000, Tokyo, Japan), transmission electron microscope (TEM, JEM-2100, JEOL, Tokyo, Japan) and atomic force microscope (AFM, Cypher S, Asylum Research, Oxford Instruments, Abingdon, UK). Molecular structure of exfoliated MoS2 nanosheets was evaluated by Raman spectra (Renishaw inVia, Renishaw, Gloucestershire, UK). The gas sensor surface was observed through a biological microscope (Eclipse-E200, Nikon, Tokyo, Japan). All the electrical measurements were carried out on a CGS-8 intelligent gas sensing analysis system (Beijing Elite Tech Co., Ltd, Beijing, China) at room temperature of 25 °C. The laboratory humidity is 55% relative humidity (RH), and the volume of our test chamber is 1000 mL. The NO2 gas sources of different concentrations (5 ppm, 10 ppm, 20 ppm, 50 ppm, 100 ppm, and 200 ppm, 21%vol O2 with 79%vol N2 The morphologies of exfoliated MoS 2 nanosheets were observed by scanning electron microscope (SEM, Hitachi SU5000, Tokyo, Japan), transmission electron microscope (TEM, JEM-2100, JEOL, Tokyo, Japan) and atomic force microscope (AFM, Cypher S, Asylum Research, Oxford Instruments, Abingdon, UK). Molecular structure of exfoliated MoS 2 nanosheets was evaluated by Raman spectra (Renishaw inVia, Renishaw, Gloucestershire, UK). The gas sensor surface was observed through a biological microscope (Eclipse-E200, Nikon, Tokyo, Japan). All the electrical measurements were carried out on a CGS-8 intelligent gas sensing analysis system (Beijing Elite Tech Co., Ltd, Beijing, China) at room temperature of 25 • C. The laboratory humidity is 55% relative humidity (RH), and the volume of our test chamber is 1000 mL. The NO 2 gas sources of different concentrations (5 ppm, 10 ppm, 20 ppm, 50 ppm, 100 ppm, and 200 ppm, 21%vol O 2 with 79%vol N 2 as balanced gas) were bought from Dalian Special Gases co. LTD (Dalian, China), which had been calibrated by Fourier transform infrared spectrometer (spectrum 100, PerkinElmer, Waltham, MA, USA). The response of the gas sensors is defined as the ratio of the resistance of the sensors in tested gases (R g ) to that in the air (R 0 ). For oxidizing tested gases, that is response = R g /R 0 , while for the reducing tested gases, response = R 0 /R g [18][19][20]. The time taken by a sensor to reach 90% of the total resistance change was defined as the response/recovery times [21].

Results and Discussion
In order to further verify the few-layer microstructure of MoS 2 nanosheets, SEM, TEM, Raman spectra and AFM images of MoS 2 nanosheets are shown in Figure 2. From Figure 2a, a sheet with a smooth surface is observed on the substrate, and the sheet surface shows the same pattern as the substrate, which looks transparent. This phenomenon, that the fuzzy pattern of the substrate reveals on the sheet surface, can be mainly attributed to the fact that the thickness of the sheet is thin enough to allow the electron beam to penetrate through it at the acceleration voltage of 30 kV, resulting in the reception of signals from the substrates. A high-resolution TEM (HRTEM) image is shown in Figure 2b, and a lattice fringe spacing of 0.27 nm corresponding to the crystal planes (100) of MoS 2 . The ordered lattice arrangement of MoS 2 indicates that MoS 2 has good crystallinity [22]. In Figure 2c, Raman spectra of bulk MoS 2 and exfoliated MoS 2 nanosheets are shown as measured using a 532 nm laser at room temperature of 25 • C. Two characteristic vibration modes can be observed in the spectrum of bulk MoS 2 , the in-plane E 1 2g mode at 381.8 cm −1 results from opposite vibration of two S atoms with respect to the Mo atom while the A 1g mode at 407.3 cm −1 is associated with the out-of-plane vibration of only S atoms in opposite directions [23,24]. Generally, the frequency difference (∆) of the two dominant modes can be used to estimate the number of MoS 2 layers [25]. as balanced gas) were bought from Dalian Special Gases co. LTD (Dalian, China), which had been calibrated by Fourier transform infrared spectrometer (spectrum 100, PerkinElmer, Waltham, MA, USA). The response of the gas sensors is defined as the ratio of the resistance of the sensors in tested gases (Rg) to that in the air (R0). For oxidizing tested gases, that is response = Rg/R0, while for the reducing tested gases, response = R0/Rg [18][19][20]. The time taken by a sensor to reach 90% of the total resistance change was defined as the response/recovery times [21].

Results and Discussion
In order to further verify the few-layer microstructure of MoS2 nanosheets, SEM, TEM, Raman spectra and AFM images of MoS2 nanosheets are shown in Figure 2. From Figure 2a, a sheet with a smooth surface is observed on the substrate, and the sheet surface shows the same pattern as the substrate, which looks transparent. This phenomenon, that the fuzzy pattern of the substrate reveals on the sheet surface, can be mainly attributed to the fact that the thickness of the sheet is thin enough to allow the electron beam to penetrate through it at the acceleration voltage of 30 kV, resulting in the reception of signals from the substrates. A high-resolution TEM (HRTEM) image is shown in Figure 2b, and a lattice fringe spacing of 0.27 nm corresponding to the crystal planes (100) of MoS2. The ordered lattice arrangement of MoS2 indicates that MoS2 has good crystallinity [22]. In Figure 2c, Raman spectra of bulk MoS2 and exfoliated MoS2 nanosheets are shown as measured using a 532 nm laser at room temperature of 25 °C. Two characteristic vibration modes can be observed in the spectrum of bulk MoS2, the in-plane E 1 2g mode at 381.8 cm −1 results from opposite vibration of two S atoms with respect to the Mo atom while the A1g mode at 407.3 cm −1 is associated with the out-ofplane vibration of only S atoms in opposite directions [23,24]. Generally, the frequency difference (Δ) of the two dominant modes can be used to estimate the number of MoS2 layers [25]. Compared to the bulk MoS2, the E 1 2g peak of MoS2 nanosheets shifts from 381.8 cm −1 to 383.4 cm -1 , whereas the A1g mode experiences almost zero shift. The value of Δ is 23.9 cm −1 for the exfoliated MoS2 nanosheets which consisted of about four-monolayer MoS2, agrees well with the reported results in the literature [23]. Figure 2d,e shows the AFM image and the corresponding quantitative Compared to the bulk MoS 2 , the E 1 2g peak of MoS 2 nanosheets shifts from 381.8 cm −1 to 383.4 cm −1 , whereas the A 1g mode experiences almost zero shift. The value of ∆ is 23.9 cm −1 for the exfoliated MoS 2 nanosheets which consisted of about four-monolayer MoS 2 , agrees well with the reported results in the literature [23]. With the thickness of a MoS 2 monolayer of about 0.65 nm [26,27], this suggests that the as-prepared FLMN are composed of 2-5 monolayers MoS 2 .
The optical images of the gas sensor without and with FLMN are displayed in Figure 3a-c. The blank Al 2 O 3 ceramic substrate with Ag-Pd interdigital electrodes is clearly seen in Figure 3a, and the bare gap is quite clean without any materials. From Figure 3b,c, the FLMN with metallic luster are randomly dispersed on the Al 2 O 3 ceramic substrate with interdigital electrodes, and the interlaced MoS 2 nanosheets are bridged between the adjacent electrodes, which form the sensing channels on the gap. In order to support the issue that the sensing channels are composed of bridged MoS 2 nanosheets, gas sensors with and without FLMN were exposed to 5 and 200 ppm NO 2 at room temperature and their response and recovery behaviors are shown in Figure 3d. It can be seen that the resistance of the gas sensor coated with FLMN increases upon injection of NO 2 and that it returns to its original value after removing the NO 2 . The higher the concentration, the larger the resistance change. The resistance responses of the uncoated gas sensor to 5 and 200 ppm NO 2 are shown in the inset of Figure 3d. No matter whether the uncoated gas sensor is exposed to 5 ppm NO 2 , 200 ppm NO 2 or air, there is no change in the resistance, and its resistance is much higher than the resistance of the gas sensor coated with FLMN. These results prove that bridging MoS 2 nanosheets between adjacent electrodes form a conduction channel that reduces the resistance of the coated gas sensor below its resistance before coating. The observed NO 2 response patterns can, therefore, be ascribed to the MoS 2 nanosheets.  [26,27], this suggests that the asprepared FLMN are composed of 2-5 monolayers MoS2. The optical images of the gas sensor without and with FLMN are displayed in Figure 3a-c. The blank Al2O3 ceramic substrate with Ag-Pd interdigital electrodes is clearly seen in Figure 3a, and the bare gap is quite clean without any materials. From Figure 3b,c, the FLMN with metallic luster are randomly dispersed on the Al2O3 ceramic substrate with interdigital electrodes, and the interlaced MoS2 nanosheets are bridged between the adjacent electrodes, which form the sensing channels on the gap. In order to support the issue that the sensing channels are composed of bridged MoS2 nanosheets, gas sensors with and without FLMN were exposed to 5 and 200 ppm NO2 at room temperature and their response and recovery behaviors are shown in Figure 3d. It can be seen that the resistance of the gas sensor coated with FLMN increases upon injection of NO2 and that it returns to its original value after removing the NO2. The higher the concentration, the larger the resistance change. The resistance responses of the uncoated gas sensor to 5 and 200 ppm NO2 are shown in the inset of Figure 3d. No matter whether the uncoated gas sensor is exposed to 5 ppm NO2, 200 ppm NO2 or air, there is no change in the resistance, and its resistance is much higher than the resistance of the gas sensor coated with FLMN. These results prove that bridging MoS2 nanosheets between adjacent electrodes form a conduction channel that reduces the resistance of the coated gas sensor below its resistance before coating. The observed NO2 response patterns can, therefore, be ascribed to the MoS2 nanosheets. The FLMN gas sensor was repeatedly exposed to gas pulses with concentrations ranging from 5 to 200 ppm NO2, separated by periods of fresh air in between. The resulting transient responserecovery curve of the FLMN gas sensor is shown in Figure 4a. The resistance of the gas sensor gradually increases with the NO2 concentrations as it is exposed to NO2, implying that the N-type response behavior of MoS2 nanosheets is found in the detection of NO2. According to this figure, the response of the gas sensor continuously increases as the NO2 concentration is ramped up from 5 to The FLMN gas sensor was repeatedly exposed to gas pulses with concentrations ranging from 5 to 200 ppm NO 2 , separated by periods of fresh air in between. The resulting transient response-recovery curve of the FLMN gas sensor is shown in Figure 4a. The resistance of the gas sensor gradually increases with the NO 2 concentrations as it is exposed to NO 2 , implying that the N-type response behavior of MoS 2 nanosheets is found in the detection of NO 2 . According to this figure, the response of the gas sensor continuously increases as the NO 2  20, 50, 100 and 200 ppm NO 2 , respectively. After removing the NO 2 from the gas sensor, the resistance of the gas sensor can return completely each time with almost no drift. The recovery behavior is very fast and recovery time constants are as short as 2-4 s. As the MoS 2 gas sensors reported in the previous works often fail to recover [28][29][30], the degree of recovery is an important indicator of the quality of the gas sensor. Herein, the recovery characteristic of the MoS 2 gas sensor is investigated by calculating the recovery rate, defined as follows [31,32].
Here, R 0 and R g are the resistances of the gas sensor before and after exposure to the target gas, and R r is the stable resistances after putting the gas sensor back to air. As shown in the inset of Figure 4a, the FLMN gas sensor shows an outstanding recovery rate greater than 97%, which implies good recovery behavior for the FLMN gas sensor to detect NO 2 gas.  In order to evaluate the repeatability and reversibility, the FLMN gas sensor is continuously placed into and removed from NO2 of the same concentration, and Figure 5a shows the response and recovery curves for three cycles when the gas sensors alternately change between air and 100 ppm NO2. Generally, the target gas is difficult to desorb completely from gas sensor surface without any stimulation of external field such as thermal field [36], optical field [42], etc., resulting in the long recovery time of gas sensor and the large drifting baseline at room temperature [31,32]. The above phenomenon is mainly caused by the chemical adsorption formed on the surface of sensing materials which makes it difficult for gas molecules to desorption [43]. In our work, for each cycle, the response of the gas sensor is 29 (the resistance changes from about 3 MΩ to 88 MΩ in the case of NO2 adsorption) and the response/recovery times are 42/2 s as shown in Figure 5a. It is worth mentioning that the recovery curve can return quickly to the baseline for each time with almost no drift, namely, the gaseous NO2 can be completely desorbed from the gas sensor without any extra stimulus like optical or thermal sources. The results illustrate that our FLMN gas sensors are superior in terms of repeatability and reversibility compared to other gas sensors working at room temperature. Interestingly, and contrary to many reported works [20,32,[44][45][46], the recovery time of our sensors is far shorter than response time. The reason for this abnormal behavior is attributed to the fact that physical adsorption of gases is dominant on the mechanically exfoliated FLMN surface which is reported to have few defects [47]. While chemical adsorption mainly takes place on the other reported gas sensors due to the formation of defects on the surface of MoS2 synthesized by the wet chemical method [48,49]. In order to verify the physical adsorption behavior of NO2 on the MoS2 surface mentioned above, the parameters related to the adsorption configuration, such as the adsorption energy, the distance between NO2 and MoS2, the bond lengths of gas molecules, etc., are calculated based on density-functional theory (DFT). All DFT calculations were performed as implemented in the Vienna ab-initio simulation package (VASP) [50,51], and the exchange-correlation potential is In order to illustrate the contrast between our gas sensor and the state of the art, the performance of our gas sensor is compared to other reported MoS 2 gas sensor in Table 1. The gas sensor based on FLMN presents a very high response of 4.4 and a very fast recovery time of~2 s at 5 ppm NO 2 gas, which does not require any heating unit to realize the detection of NO 2 at room temperature. As far as we know, our gas sensor features the fastest recovery time and the highest response for the detection of NO 2 at room temperature. In order to illustrate the data reliability, error bars have been calculated by the standard deviation formula [40], and the response versus NO 2 concentration index fitting curve with error bars is shown in Figure 4b. It exhibits small deviations for the FLMN gas sensor, indicating the data are reliable in the whole concentration range from 5 to 200 ppm. Based on the least squares method [41], the fitting equation of the response Y and NO 2 concentration X can be represented as Y = 81.68 − 77.26 × e (−X/295.29) − 2.44 × e (−X/10. 48) , and the regression coefficient R 2 is 0.965 at the concentration range from 5 to 200 ppm. The response curve shows optimal linear dependence in range of 5 to 100 ppm and then sign of slight saturation behavior at the NO 2 concentration larger than 100 ppm. From the inset of Figure 4b, the FLMN gas sensor shows a high response to NO 2 at room temperature while only minimal responses toward other gases such as ammonia, formaldehyde, ethanol, acetone and methanol, which is of an excellent cross sensitivity toward NO 2 .
In order to evaluate the repeatability and reversibility, the FLMN gas sensor is continuously placed into and removed from NO 2 of the same concentration, and Figure 5a shows the response and recovery curves for three cycles when the gas sensors alternately change between air and 100 ppm NO 2 . Generally, the target gas is difficult to desorb completely from gas sensor surface without any stimulation of external field such as thermal field [36], optical field [42], etc., resulting in the long recovery time of gas sensor and the large drifting baseline at room temperature [31,32]. The above phenomenon is mainly caused by the chemical adsorption formed on the surface of sensing materials which makes it difficult for gas molecules to desorption [43]. In our work, for each cycle, the response of the gas sensor is 29 (the resistance changes from about 3 MΩ to 88 MΩ in the case of NO 2 adsorption) and the response/recovery times are 42/2 s as shown in Figure 5a. It is worth mentioning that the recovery curve can return quickly to the baseline for each time with almost no drift, namely, the gaseous NO 2 can be completely desorbed from the gas sensor without any extra stimulus like optical or thermal sources. The results illustrate that our FLMN gas sensors are superior in terms of repeatability and reversibility compared to other gas sensors working at room temperature. Interestingly, and contrary to many reported works [20,32,[44][45][46], the recovery time of our sensors is far shorter than response time. The reason for this abnormal behavior is attributed to the fact that physical adsorption of gases is dominant on the mechanically exfoliated FLMN surface which is reported to have few defects [47]. While chemical adsorption mainly takes place on the other reported gas sensors due to the formation of defects on the surface of MoS 2 synthesized by the wet chemical method [48,49]. In order to verify the physical adsorption behavior of NO 2 on the MoS 2 surface mentioned above, the parameters related to the adsorption configuration, such as the adsorption energy, the distance between NO 2 and MoS 2 , the bond lengths of gas molecules, etc., are calculated based on density-functional theory (DFT). All DFT calculations were performed as implemented in the Vienna ab-initio simulation package (VASP) [50,51], and the exchange-correlation potential is treated with the Perdew-Burke-Eznerh of generalized-gradient approximation (PBE-GGA) [52,53]. The projector augmented wave (PAW) method is used to describe the electron-ion interaction [54]. For the structural relaxations and energy calculations, we employ the D2 method of Grimme (DFT-D2), which includes van der Waals (vdW) interactions [52,55]. All calculations are performed with a 3 × 3 × 1 supercell of MoS 2 containing 27 atoms, and the cut off energy for plane-wave expansion is 400 eV. The Brillouin zone is sampled with a grid of 9 × 9 × 1 conducted by the Monkhorst-Pack special k-point scheme [56]. For geometry optimization, all the internal coordinates are fully relaxed until the Hellmann-Feynman forces are less than 0.005 eV/Å. The three adsorption configurations of NO 2 molecules on MoS 2 surface, including the NO 2 adsorbed on the hollow, Mo-top and S-top sites of MoS 2 , are shown in Figure 5b-d, and Table 2 gives the adsorption parameters for three adsorption configurations including total energy (E tot ), adsorption energy (E ad ), the distance of adsorbed NO 2 to the MoS 2 monolayer (d zN-S ) and the bond length of gas molecules (l N-O ). The adsorption energy of gas molecules on MoS 2 surface is calculated as E ad = E tot − E MoS 2 − E Gas , where E tot is the total energy of MoS 2 with a molecule absorbed, E MoS 2 and E Gas are the energies of the pristine MoS 2 single layer and isolated gas molecule. From Table 2, the total adsorption energies of the three adsorption configurations are almost the same, indicating that three adsorption configurations are all possible for the adsorption of NO 2 on MoS 2 surface. the MoS2 monolayer (dzN-S) and the bond length of gas molecules (lN-O). The adsorption energy of gas molecules on MoS2 surface is calculated as Ead = Etot − EMoS2 − EGas, where Etot is the total energy of MoS2 with a molecule absorbed, EMoS2 and EGas are the energies of the pristine MoS2 single layer and isolated gas molecule. From Table 2, the total adsorption energies of the three adsorption configurations are almost the same, indicating that three adsorption configurations are all possible for the adsorption of NO2 on MoS2 surface.  The Ead of NO2 in the three adsorption configurations is very small, which the maximum Ead (NO2 adsorbed on hollow) is only 0.05 eV, and the distances of adsorbed NO2 to the MoS2 surface are large in all three adsorption configurations (dzN-S are all greater than 3 Å). The small Ead and the large dzN-S indicate that no chemical bond is formed between NO2 and MoS2 on MoS2 surface. Moreover, no matter at which adsorption site NO2 is adsorbed, the lN-O of NO2 gas molecules is approximately 1.218 Å, which lN-O is almost unchanged compared with lN-O (l~1.213 Å) of the free NO2 gas molecule. This further illustrates that the NO2 gas is not chemically adsorbed on the MoS2 surface, because the bond length will be greatly affected by the gas chemically adsorbed on the surface of the material [57,58]. Therefore, from the calculation results of Ead, dzN-S and lN-O, it is concluded that the NO2 gas is adsorbed on the MoS2 surface by a weak van der Waals interaction, i.e. the physical adsorption is the main factor for the adsorption of NO2 on MoS2 surface.
The response of gas sensing materials against target gas is mainly dependent on the electronic interaction between gases and materials, which occurs mainly on the surface of materials, i.e. it is a  The E ad of NO 2 in the three adsorption configurations is very small, which the maximum E ad (NO 2 adsorbed on hollow) is only 0.05 eV, and the distances of adsorbed NO 2 to the MoS 2 surface are large in all three adsorption configurations (d zN-S are all greater than 3 Å). The small E ad and the large d zN-S indicate that no chemical bond is formed between NO 2 and MoS 2 on MoS 2 surface. Moreover, no matter at which adsorption site NO 2 is adsorbed, the l N-O of NO 2 gas molecules is approximately 1.218 Å, which l N-O is almost unchanged compared with l N-O (l~1.213 Å) of the free NO 2 gas molecule. This further illustrates that the NO 2 gas is not chemically adsorbed on the MoS 2 surface, because the bond length will be greatly affected by the gas chemically adsorbed on the surface of the material [57,58]. Therefore, from the calculation results of E ad , d zN-S and l N-O , it is concluded that the NO 2 gas is adsorbed on the MoS 2 surface by a weak van der Waals interaction, i.e. the physical adsorption is the main factor for the adsorption of NO 2 on MoS 2 surface.
The response of gas sensing materials against target gas is mainly dependent on the electronic interaction between gases and materials, which occurs mainly on the surface of materials, i.e., it is a surface-controlled process [59]. As only a certain thickness of material surface can interact with the gas, only a certain depth of electron depletion layer can be formed with an order of 2-100 nm when the gas adsorbs on the surface of materials, which is usually called the Debye length (L) [8,9,60].
When the crystallite size (D) is much larger than 2L, grain-boundary contacts display higher resistance and govern the electric gas sensitivity of the chain (grain-boundary control) [59]. When D decreases to come closer to 2L, the necks become the most resistant, controlling the gas sensitivity (neck control) [61]. Finally, when D is smaller than 2L, each constituent grain is fully depleted of conduction electrons as a whole. In this situation, the resistance of grains dominates the whole resistance of the chain and the gas sensitivity, in this case, is controlled by grains themselves (grain control) [9]. The reported results mentioned above illustrate that the gas sensing properties are closely related to the size of gas sensing materials. Figure 6 shows the schematic diagram of the gas sensing mechanism and equivalent circuit, and the exfoliated FLMNs are bridged randomly between the adjacent electrodes, which form the sensing channels on the gap. In our work, the gas sensor based on FLMN exhibits a higher response than other reported MoS 2 gas sensors which have been shown in Table 1, and their high responsivity can be attributed to the following two reasons. First, the thickness of MoS 2 is about 1.5-3.2 nm in Figure 6b, which belongs to the type of grain control, so the electrons of all over the sensing channels (including the bridging contact and the FLMN itself) can be affected by the adsorbed NO 2 , and the space-charge layer then penetrates into the whole sensing channels, which leads to a sharp decrease in conductivity and great improvement of the response [8,9]. As shown in the equivalent circuit diagram, all the electronic transport paths through the FLMN are controlled by NO 2 gas, so the total resistance of FLMN gas sensor can change greatly when cycling it between air and NO 2 . As we have shown in the schematic of Figure 6b, the main reason why FLMN have a large response is that the NO 2 gas controls all the conduction channels of carriers in the material, and the thickness of MoS 2 is an important factor determining its gas sensing properties. Secondly, the gas molecules are generally confined to adsorb on the active sites of materials in the case of chemical adsorption, which limits the number of gas molecules adsorbed on the surface of the material [62,63]. By means of physical adsorption and a small amount of chemical adsorption, NO 2 gas molecules can easily diffuse to the whole surface of the material, leading to the increase of the adsorption quantity of gas molecules on the materials surface [63]. When gas molecules are physically adsorbed the material surface, the electrostatic attraction between the material and gas causes the transformation of electrons from the surface of the material to form a dipole moment [47,64,65]. As the gas concentration increases, the continuous transfer of electrons will increase the resistance of materials, and the increase of dipole moment further promotes physical adsorption of gas molecules. Therefore, the resistance of material increases with the increase in gas concentration. In addition, there may be a few chemisorptions of NO 2 gas molecules on the mechanically exfoliated FLMN due to the existence of few defects on the surface of MoS 2 [66]. Here, chemisorption of NO 2 gas molecules on FLMN does not affect the fast desorption behavior, and the main reason is that Brunauer-Emmett-Teller (BET) water layers formed on FLMN surface may dissolve the chemisorbed NO 2 which can promote the rapid desorption of chemisorbed NO 2 [67].

Conclusions
In summary, an FLMN gas sensor via a facile way (mechanical exfoliation) was demonstrated to have excellent performance, which enables overcoming the limitations of 2D TMDs gas sensors such as low response and poor recovery. Through the comparison with the state of the art, the performances, including high response against NO2 at room temperature and the quick and complete recovery behaviors (the recovery time of 2 s, the recovery rate greater than 97%), are confirmed in this work. Based on density-functional theory (DFT), the calculation shows that the excellent performances at room temperature are mainly attributed to the physical adsorption of NO2 on FLMN surface and size effect from extremely thin thickness of FLMN. Thus, an FLMN gas sensor via mechanical exfoliation can resolve the low NO2-sensing performance issues in terms of response and recovery, and potentially open up a new avenue for gas sensing applications.
Author Contributions: W.L. designed the experiments, analyzed the data and wrote the paper. Y.Z. revised the paper. X.L., J.C. and X.G. performed the theoretical analysis. X.X. gave advice on the experiments. J.P. and X.Z. provided characterization instruments.

Conclusions
In summary, an FLMN gas sensor via a facile way (mechanical exfoliation) was demonstrated to have excellent performance, which enables overcoming the limitations of 2D TMDs gas sensors such as low response and poor recovery. Through the comparison with the state of the art, the performances, including high response against NO 2 at room temperature and the quick and complete recovery behaviors (the recovery time of 2 s, the recovery rate greater than 97%), are confirmed in this work. Based on density-functional theory (DFT), the calculation shows that the excellent performances at room temperature are mainly attributed to the physical adsorption of NO 2 on FLMN surface and size effect from extremely thin thickness of FLMN. Thus, an FLMN gas sensor via mechanical exfoliation can resolve the low NO 2 -sensing performance issues in terms of response and recovery, and potentially open up a new avenue for gas sensing applications.