Ultra-low detection limit chemoresistive NO2 gas sensor using single transferred MoS2 flake: an advanced nanofabrication

In this work, a method of fabricating a NO2 nano-sensor working at room temperature with a low detectable concentration limit is proposed. A 2D-MoS2 flake is isolated by transferring a single MoS2 flake to SiO2/Si substrate, followed by applying an advanced e-beam lithography (EBL) to form a metal contact with Au/Cr electrodes. The resulting chemoresistive nano-sensor using a single MoS2 flake was applied to detect a very low concentration of NO2 at the part-per-billion (ppb) level. This result is obtained due to the ability to create microscopic nano-sized MoS2 gaps using e-beam lithography (300 nm–400 nm). Experimental results also show that the sensor can capture changes in concentration and send the information out extremely quickly. The response and recovery time of the sensor also reached the lowest point of 50 and 75 ms, outperforming other sensors with a similar concentration working range.


Introduction
The amount of poisonous gases emitted into the environment has signicantly increased due to industrial development as well as improvements in quality of life and transit accessibility, causing substantial harm to the ecosystem and the general public, making it necessary to monitor the air quality in metropolitan areas. 1,2 Among the aforementioned air contaminants, NO 2 is one of the most deadly as at concentrations over 1 ppm, it causes or aggravates respiratory disorders including emphysema and bronchitis as well as severe damage to lung tissues in humans. 3 Therefore, research on gas sensors is helpful for the NO 2 emission monitoring and controlling process and is a potential research direction where a large number of scientists are currently interested.
Two-dimensional (2D) materials have been applied in various research applications and industries because of their potential in chemical gas sensing, electronics, optoelectronics, optics, and energy generation. [4][5][6][7] In previous works, the resistive NO 2 gas sensor was fabricated using 2D materials (molybdenum disulde MoS 2 /graphene hybrid), which showed high sensitivity, good selectivity, and a low detection limit. 8 However, the concentration limit of that type of sensor is still relatively high and is not suitable for monitoring NO 2 in the surrounding environment. The nano-sensor deployed by the combination of two-dimensional (2D) materials of atomic-layer thickness and the advantages of nanofabrication is an urgent research direction because of excellent outcomes in the chemical gas sensing eld. However, the biggest challenge in this research direction is how we can make sensor devices at the nanoscale to quickly trap these changes in 2D materials, i.e., what is the optimal design and what are the proper materials. This has been the motivation for us to conduct experiments to fabricate a highly sensitive nanoscale NO 2 gas sensor using advanced materials. MoS 2 has recently been suggested as a promising material for NO 2 gas sensing 9-12 because, when combined with graphene to create composite or hybrid materials, it increases sensitivity due to synergistic effects and eliminates the issue of high recovery time of graphene-only NO 2 sensor. 13 Moreover, the challenge in developing chemoresistive sensors is the distance between electrodes, which is limited by e-beam lithography technology. The short distance between electrodes and the sensing area at truly nanoscale give such advantages as fast response and low detection limit. 14 In this work, we transferred a single MoS 2 ake to SiO 2 /Si and applied advanced e-beam lithography to make chemoresistive nanosensors to detect a very low concentration of NO 2 gas. transferred to SiO 2 /Si. To fabricate a chemoresistive NO 2 gas sensor, two steps of e-beam lithography were applied. In the rst step, the array of Au/Cr marker was fabricated on SiO 2 /Si using 100 nm thick PMMA as e-beam resist. Then, the single MoS 2 ake was transferred to SiO 2 /Si (with Au/Cr marker on top) using 3M Scotch tape. The intermediate SEM observation step is to nd the MoS 2 ake on SiO 2 /Si aer the transferring process and identify the MoS 2 ake position using Au/Cr marker. Aer identifying the MoS 2 ake location on the array Au/Cr marker with high precision, the second e-beam step was performed on PMMA (100 nm)/MoS 2 /SiO 2 /Si to make metal contact by the lioff process. Then, two Au/Cr electrodes were deposited on MoS 2 /SiO 2 /Si by e-beam evaporation to make the nal chemoresistive nano-sensor device. The illustration of chemoresistive NO 2 nanosensor fabrication process is shown in Fig. 1.

Evaluation methods
The morphology and nanofabrication of the MoS 2 /SiO 2 /Si sample were determined using high-resolution scanning electron microscopy (HR-SEM; SU 8010, Hitachi) and highresolution e-beam lithography (e-Raith 150). High-resolution transmission electron microscopy (HR-TEM; JEM-2100F, JEOL) was used to gauge the atomic structure and thickness of transferred MoS 2 . Raman spectroscopic measurements were conducted in WITec Raman imaging with a 532 nm laser source to identify MoS 2 quality before/aer the transferring process. The NO 2 gas sensor was installed inside a sealed chamber that was attached to NO 2 gas cylinders with various concentrations. The I-V characteristics and resistance of the sensor were recorded using a semiconductor probe station (Keithley SCS-4200). The detailed measurement setup for the NO 2 gas sensor was described in previous work. 8

Results and discussion
As-grown MoS 2 -graphene hybrid and a single MoS 2 ake on SiO 2 /Si aer the transfer process are depicted in SEM images in Fig. 2(a). It is obvious that the MoS 2 with 2D crystallite was grown on graphene and was triangle-shaped. The triangle-like MoS 2 akes are with average sizes of around 1-2 mm. The MoS 2 akes were grown in a large-scale area on graphene with uniform thickness as shown in AFM images in Fig. 2(b). In Fig. 2(c), the transferred MoS 2 seems to induce defects during the transfer process by breaking into smaller akes compared with the original MoS 2 ake as-deposited on graphene (see AFM image in Fig. 2(b)). In order to measure the thickness of MoS 2 akes on graphene, the AFM with tapping mode was carried out. Results in MoS 2 ake average thickness is around 5 nm, according to 10 layers of MoS 2 , consistent with the report in ref. 8. Fig. 2(d) shows the cross-sectional view of HR-TEM of transferred MoS 2 on SiO 2 /Si substrate, which indicates 8 layers of MoS 2 . Fig. 3 shows the Raman spectra of a few layers of MoS 2 growth on graphene and their Raman mapping. In Fig. 3(a), two Raman peaks were assigned to MoS 2 material (E g at ∼378.5 cm −1 and A 1g at ∼405.8 cm −1 ). 8 The E g modes are doubly degenerate in-plane vibrational modes and A 1g is an outof-plane vibration mode, which agrees well with previous observations in b-phase bulk MoS 2 . 15 In comparison with the Raman spectrum of transferred pure MoS 2 akes (transferring from MoS 2 /graphene/SiO 2 /Si sample) on SiO 2 /Si substrate, both peak frequencies of b-phase (E g and A 1g ) of MoS 2 -graphene shi to the higher wavenumber region. The Raman peaks of transferred MoS 2 akes on SiO 2 /Si, as shown in Fig. S1, † are at 370.9 and 401.2 cm −1 for E g and A 1g mode, respectively. The shi in Raman peaks of MoS 2 material is due to either a strain/stress free state aer transferring from graphene or the inuence of the thickness of a few-layers MoS 2 on Raman shi. 15 The G and 2D bands of graphene were assigned to two minor peaks at 1584.3 cm −1 and 2677.2 cm −1 , respectively. The in-plane vibration of sp 2 -carbon atoms was attributed to the G band, while defects in graphene were associated with the D band's breathing modes of six-atom rings. The D band, which is at around 1350 cm −1 indicating defects in graphene, was not present in the sample aer depositing MoS 2 as in Fig. 3(a). However, the 2D peaks indicating the number of layers of graphene or graphene quality appeared at 2677.2 cm −1 . The intensities of graphene-related peaks are lower than MoS 2 peaks because of either the thickness of the few-layer MoS 2 or the focus point of the laser from Raman. Fig. 3(b) shows an optical image of MoS 2 -graphene and the selected area for Raman mapping. Fig. 3(c and d) are images for Raman mapping for average peak intensity centered at 405.8 cm −1 and 378.5 cm −1 for A 1g and E g mode of MoS 2 material, respectively. The A 1g peak has clear contrast on the area of MoS 2 akes in comparison with the E g peak, which indicated the higher intensity of A 1g mode in Fig. 1 The illustration of the chemoresistive NO 2 nanosensor fabrication process. Fig. 3(e and f) are images of Raman mapping of G and 2D peaks related to graphene, respectively. In conversion, the mapping of peaks related to graphene has more uniformity in color, except at the graphene grain boundary, which gives higher intensity in Fig. 3(e and f) due to the thicker thickness of folded graphene at the boundary. These Raman mapping images conrm the MoS 2 as isolated akes on the base planar graphene layer. Fig. 4(a) shows the fabricated chemoresistive sensor using a single transferred MoS 2 ake. The Au/Cr metal contact on a single MoS 2 ake is with a distance of 300 nm between two electrodes. The fabricated sensor was tested several times during the day and on different days. Results showed a consistent response with the same testing conditions (NO 2 gas concentrations and bias voltage). Fig. 4(b-d) shows the sensor responses with a low concentration of NO 2 gas. The I-V curves were recorded with synthetic air and different NO 2 gas concentrations of 1, 10, and 100 ppb as shown in Fig. 4(b). The fabricated sensor showed a clearly different response with various NO 2 concentrations. Fig. 4(b) indicates the maximum  response of the fabricated sensor with 100 ppb NO 2 . I-V curves in Fig. 4(b) show that the sensor response increased dramatically in the forwarding region when exposed to NO 2 gas. The non-linear behavior in the I-V characteristic of Au/MoS 2 /Au sample is due to either space-charge limited current in 2D materials 16 or oxidized MoS 2 surface giving the nano-junction with Au electrode. With the same sample structure without MoS 2 akes, pure graphene sample (with two Au/Cr electrodes and a distance of 300 nm) shows linear ohmic contact and negligible change when exposed to 100 ppb NO 2 gas (see Fig. S2 in ESI †). Fig. 4(c) shows the transient response of the fabricated sensor with 5 V bias exposed under 3 repeat cycles of 1, 10, and 100 ppb. The sensor response shows good repeatability and the gas-induced-current increased by two orders of its value when exposed to NO 2 gas. The response/recovery time of the fabricated sensor is estimated at around 50/75 ms. Fig. 4(d) shows the transient response to 3 cycles of 100 ppb NO 2 at a lower bias voltage of 1 V. The current increased only 4-fold with 100 ppb NO 2 at 1 V bias, which is much smaller than with a 5 V bias. The NO 2 gas generated more current in Au-MoS 2 -Au at a higher bias voltage in comparison with a low bias voltage due to the inuence of the external electrical eld on electron-hole separation and charge collection at the electrode. 14,16 The performance of the fabricated NO 2 gas sensor can be compared with previously developed sensors (Table 1). [17][18][19][20][21] All of them are sensors that work at NO 2 concentrations at ppb, so all comparisons are fair. It is evident that the proposed sensor exhibited a low limit of detection and high sensitivity, among others. In addition, the response/recovery time of the proposed sensor also outperforms others by being more than 60 times faster. In Table 1, the sensor response (S) at xed bias voltage is dened as the ratio of current change when the sensor is exposed to NO 2 :

MoS 2 akes.
where I a is the current of the sensor in the presence of synthesized air only, and I g is the current in the presence of NO 2 at given concentrations. The response time is dened as the time required for the sensor to reach 90% of the current change (DI) when the sensor is exposed to a given concentration of NO 2 . The recovery time is dened as the time needed to recover 90% of the initial baseline aer NO 2 is turned off. The sensor response of pure graphene with NO 2 is shown in Fig. S2(b and c). † Obviously, the Au-MoS 2 -Au showed signicantly higher responsivity in comparison with pure graphene (for responsivity of pure graphene, see Fig. S1 in ESI †). The sensor made of Au-MoS 2 -Au has increased responsivity from 4 to 100 with 100 ppb NO 2 at 1 V and 5 V bias, respectively. Meanwhile, the sensor response of pure graphene is not signicant, which is less than 1% in Fig. S2(b and c), † at 1 V and 5 V bias. The selectivity of the fabricated sensor towards NO 2 , in comparison with common toxic gases/moisture such as H 2 , H 2 O, CO, and CO 2 , is shown in Fig. S2(d). † In Fig. S2(d), † examined gases/moisture were measured at 1000 ppb, and the negative sensor response indicates that the current was reduced with H 2 , CO, and CO 2 gas. Results showed good selectivity of the fabricated sensor towards NO 2 gas, which is promising for practical applications.

Conclusions
In this work, MoS 2 was successfully deposited on graphene by thermal CVD. The single MoS 2 ake was then transferred successfully to SiO 2 /Si substrate as isolated akes. Advanced ebeam lithography was applied to fabricate the nanosensor with a single MoS 2 ake only. The created nanosensor could identify low NO 2 concentrations between 1 and 100 ppb at room temperature, showing a fast response/recovery of 50/75 ms. Thanks to the recent advantages of nanofabrication in combination with the development of 2D materials, the chemical gas sensor at truly nanoscale showed its advantages with low detection limit and fast response and is promising for detecting NO 2 at the atomic level in practical applications. In future works, the performance of the fabricated sensor will be considered over a long-term period.

Conflicts of interest
There are no conicts to declare.