Scalable integration of periodically aligned 2D-MoS₂ nanoribbon array

Two-dimensional (2D) molybdenum disulfide (MoS₂), one of the most promising transition-metal dichalcogenides (TMDs),^1,2 has received exceptional attention due to excellent electrical, mechanical, and optical properties. Since the early report on exfoliated MoS₂,³ a variety of synthetic methods^4–6 have been developed, leading to profound achievements in material and device properties.^7–11 Recently, there is a growing interest in applying 2D-MoS₂ nanostructures for sensor applications. However, fabricating robust 2D-MoS₂ nanostructures remains challenging due to the vulnerability of 2D materials to existing lithographic patterning processes. Besides, the active area patterned by these conventional methods is insufficient to be practically employed. To date, several alternative nanofabrication techniques have been reported, which relies on the following four methods: (1) selective growth on the patterned templates,^12,13 (2) reactive ion etching using patterned masks,^14,15 (3) direct removal by scanning lasers or ions,^16,17 and (4) peeling, transfer, and filling methods by controlling a surface adhesion energy.^18–20 

In method (1), the prepared Mo nanowires were converted to MoS₂ nanowires through pyrolysis in H₂S ambient; however, these MoS₂ nanowires exhibited unstable electrical properties owing to incompletely connected granular morphology and remaining precursor residues. In method (2), MoS₂ nanoribbons (MNRs) were patterned using reactive ion etching. This process was not suitable for a low cost and large area fabrication because the nanoscale etching mask was prepared by electron beam lithography (EBL). Method (3) relies on a scanning laser or an ion beam to selectively remove an undesired area of MoS₂. The rapid spreading nature of the beams deteriorated the edge roughness, which hinders achieving densely aligned MoS₂ nanostructures. In method (4), MoS₂ peeling, transfer, and filling processes were implemented by controlling the adhesion energy between a stamp and a substrate. Nam et al.¹⁸ successfully transferred the prepared MoS₂ nanostructures on a stamp onto a plasma-treated silicon oxide (SiO₂) substrate. However, preparing the MoS₂ nanostructures/stamp was complicated, and the fabrication yield was impractical. In the meantime, Zhao et al.¹⁹ introduced an Au-assisted peeling method to create various MoS₂ nanostructures, but pressuring and peeling the tape by hand is not appropriate for reliably fabricating the nanoribbons with a high aspect ratio. Also, the patternable area was very small because a polymethyl methacrylate (PMMA) grating pattern on their MoS₂ substrates was prepared by EBL. Recently, Hung et al.²⁰ reported the combined fabrication of polydimethylsiloxane (PDMS) micro-molding and thermolysis, achieving the MNR width of 157 nm. This process meets to a low cost and scalable fabrication, but the high-temperature thermolysis at ∼1000 °C needed to convert the crystallinity from MoS_x to MoS₂ can severely limit the versatility of the substrate. Overall, most of the methods discussed earlier hardly demonstrated the fabrication of densely aligned MoS₂ nanostructures with a practically large patterning area. Thus, developing a more advanced nanofabrication process is promptly required to meet the demand for practical sensor platforms.

In this work, we report on a facile, scalable, and repeatable process to fabricate the periodically aligned MNR array with a practically large area of 2.25 cm². First, the aligned Au nanoribbons are transferred onto a PMMA-coated MoS₂/SiO₂ substrate via the direct metal transfer (DMT) technique. In the DMT process, once a stamp having nanoscale grating is manufactured, it reliably and reproducibly transfers the metallic patterns with the same dimensions without using any developing solvents by just applying pressure and temperature; thus, it is eco-friendly and economical.^21,22 Next, tetra-fluoro-methane (CF₄) plasma etching is performed to transfer the pattern of Au nanoribbons into 2D-MoS₂, which resulted in creating the MNR array with the mean width of 260 nm at 650 nm pitch. The superior characteristics of the MNR array are comprehensively studied via various spectroscopic analyses and application sensor devices.

Figure 1 depicts the fabrication of the MNR array via the DMT process. First, a vapor phase self-assembled monolayer (SAM) with chemical formula CF₃(CF₂)₅(CH₂)₂SiCl₃ (tridecafluoro-1,1,2,2-tetrahydro-octyl-trichlorosilane) and 30 nm-thick Au were sequentially deposited on a SiO₂ stamp with grating patterns (see Fig. S1 of the supplementary material). The SAM can reduce the adhesion energy between the stamp and the Au layer, thereby enabling easy transfer of the Au layer onto the 100 nm-thick PMMA (2% diluted in anisole)-coated MoS₂/SiO₂ substrate (see Fig. S2 of the supplementary material for MoS₂ growth and transfer). The PMMA/MoS₂/SiO₂ substrate was placed below the stamp with Au/SAM layers. Then, the temperature was increased slowly to 130 °C while a pressure of 80 psi was simultaneously applied to ensure conformal contact between the stamp and the substrate [Fig. 1(a)]. After maintaining at 130 °C for 10 min, the temperature was naturally decreased to 80 °C, and the stamp was detached from the substrate, leaving the Au nanoribbons with 400 nm width at 650 nm pitch [Fig. 1(b)]. The Au layer on the grating (protruding) surface of the stamp was selectively transferred because the structural characteristic of the stamp prevents Au in the deep trenches from contacting the PMMA layer. A transferrable area could depend on a temperature of stamp detachment, transferred metal, and stamp dimension, as described in our previous work.²¹ Next, using the Au nanoribbons as an etching mask, the unprotected PMMA region was etched for 120 s by oxygen (O₂) plasma (power: 20 W, pressure: 20 mTorr, O₂: 50 SCCM), and then, the exposed 2D-MoS₂ was etched by CF₄ plasma (power: 20 W, pressure: 20 mTorr, CF₄: 50 SCCM) [Fig. 1(c)]. After immersing in acetone to lift-off the Au/PMMA layers [Fig. 1(d)], the periodically aligned MNR array with the area of 2.25 cm² was successfully achieved. For electrical measurement, the interdigitated (60 nm Au/5 nm Ti) contact electrode (IDE) with a channel length of 20 µm and a total width of 4000 µm was constructed perpendicular to the aligned MNR array [Fig. 1(e)].

Figure 2(a) shows the scanning electron microscopy (SEM) images of the periodically aligned MNR array at 650 nm pitch and a practically large area of 2.25 cm². The magnified SEM image [see Fig. 2(b) and the inset] clearly showed the well-aligned smooth edges of the MNR without folded or disconnected regions. This superior alignment and morphology were maintained after the Au IDE construction [see Fig. 2(c) and the inset]. Considering the pitch (650 nm) and the channel width (200 µm), the number of MNRs was ∼307. The mean width and the standard deviation of the MNR at 650 nm pitch were ∼265 nm and ∼40 nm, respectively [Fig. 2(d)]. Additionally, we fabricated a periodic 2D-MoS₂ nanomesh structure through the DMT process (see Fig. S3 of the supplementary material). Both MNR and nanomesh structures with the same dimension and patterning area were repeatedly fabricated by reusing the stamp, which is the fundamental difference from the existing plasma etching processes [method (2) using EBL]. To confirm the advantage of our method, we fabricated the reference MNR array on the same SiO₂ substrate by utilizing a conventional 325 nm laser interference lithography (LIL), which has been considered most useful for a rapid and large area nanopatterning.²³ However, the reference MNR was easily peeled off the substrate and entangled in many regions because a typical lift-off chemical solution containing 1-methyl-2-pyrrolidone (NMP) or di(propylene glycol) methyl ether to completely remove a photoresist etching mask weakened the adhesion between the MNR and the SiO₂ substrate (see Fig. S4 of the supplementary material).

Note that generally the PMMA interlayer is effective for easy and fast lift-off in acetone, but its sidewall near the bottom can be damaged by penetrating plasma ions. Furthermore, CF₄ plasma ions can etch the SiO₂ substrate even at a low plasma power, which may lead to an unstable gating effect in a three-terminal application device. A dielectric substrate, which is more resistive and non-selective to the CF₄ plasma, may resolve this issue. Herein, however, we focused solely on minimizing the damage to SiO₂ substrates. Figures 3(a)–3(d) sequentially show the SEM images of MNR arrays fabricated by applying various etching times (15, 30, 45, and 60 s) of CF₄ plasma at 20 W. At 15 s of the etching [Fig. 3(a)], the unmasked MoS₂ region was partially removed. When it was extended to 30 s [Fig. 3(b)], the unmasked region mostly disappeared, but remaining regions were folded and agglomerated along the MNR edges. Therefore, we increased the etching duration to completely etch the unmasked MoS₂ region, and at 45 s [Fig. 3(c)], the MNR width became narrower than the width of the Au nanoribbon mask, which was due to the undercut effect of CF₄ plasma ions. When it exceeded 60 s [Fig. 3(d)], the width was substantially reduced, and uneven and disconnected regions were observed. The mean width of the MNR after 60 s of the etching time was measured to be 135 nm, and the smallest width of a single MoS₂ nanoribbon was ∼80 nm. Figure 3(e) displays the variation in the mean width of MNR (left y-axis) and the accordingly etched depth of the SiO₂ substrate (right y-axis) as the function of etching times. The mean width linearly decreased at an estimated reduction rate of 7.3 nm/s with the increasing etching time. Considering the edge roughness, 40 s of the etching time was applied to integrate the optimized MNR array with the mean width of 260 nm.

To confirm the material characteristics of the MNR array, various spectroscopic analyses including X-ray diffraction (XRD), Raman, Photoluminescence (PL), and X-ray photoelectron spectroscopy (XPS), and global-gated electrical transfer characteristic were complementarily performed. Figure 4(a) shows the XRD 2$θ$ scans, corresponding to (002), (100), (101), (006), and (105) lattice planes of hexagonal MoS₂ (JCPDS card no. 77-1716). No significant peak shifts or new miscellaneous phases were found, meaning that the intrinsic crystallinity was adequately maintained in the MNR array. The material structure was examined using Raman spectroscopy [Fig. 4(b)]. The pristine MoS₂ presented the inherent peaks at 382.4 cm⁻¹ for E_2g (in-plain mode) and 407.5 cm⁻¹ for A_1g (out-of-plane mode), consistent with a typical few layered MoS₂ spectra.²⁴ The peak positions barely shifted in the MNR array (382.4 cm⁻¹ for E_2g and 407.1 cm⁻¹ for A_1g), but the intensity ratio (A_1g/E_2g) was markedly reduced from 2.6 (for the pristine MoS₂) to 2.2 (for the MNR array). This finding means that the “in-plane mode” vibration can be preferentially activated in such a terrace-terminated layer characterized by the laterally reduced width.²⁵ The similar inter-peak distance between E_2g and A_1g can be an indicator of the residue-free MNR surface, which was also confirmed by the PL spectra [Fig. 4(c)]. The absolute PL intensity was slightly weakened in the MNR due to the reduced active area and the open edges,^17,26 but the two exciton peaks were clearly observed at 1.83 eV and 1.97 eV related to the K-point transition in the Brillouin zone.¹ By contrast, the reference MNR with residues or the photoresist-coated MoS₂ displayed broad peaks with fluctuating backgrounds, implying that application devices based on such platforms will produce unstable characteristics.

Chemical properties of the MNR with ribbon edges were elucidated using XPS analysis. The pristine MoS₂ showed the characteristic peaks at 228.72 eV for Mo 3d_5/2 and at 231.87 eV for Mo 3d_3/2 [Fig. 4(d)]. After the CF₄ plasma etching, both Mo 3d and S 2p peaks were barely shifted, while the peak intensities were considerably decreased. In the previous literature,^27,28 the decrease in S 2s peaks (S—Mo—S bonding) and/or the increase in the Mo⁶⁺ peak (at ∼236.5 eV), featured as the indicator of a damaged basal plane of layered 2D-MoS₂, were typically observed after the plasma etching. In the MNR, there was no significant appearance of those peaks since the top surface of the 2D-MoS₂ was not damaged by the CF₄ plasma, owing to the protection with the Au nanoribbon mask. Therefore, it is expected that the deconvolution and/or broadening of both Mo 3d and S 2p peaks toward higher binding energies were mainly due to the formation of oxidized states²⁹ along the MNR edges [Figs. 4(d) and 4(e), respectively]. In fact, in the XPS spectra, we found the increased atomic percentage of oxygen (from 24.3 to 33.5 at. %) in the as-made MNR. The transfer characteristics of the drain current-gate voltage (I_D − V_G) [Figs. 4(f) and 4(g)] using a back-gating were measured at a channel length of 20 µm. The pristine MoS₂ device exhibited typical n-type transistor characteristics. However, in the MNR array device, the drain current was suppressed at positive voltages and relatively increased when the voltage polarity was reversed. The presence of carbon, fluorine, and oxygen atoms within the MNR might reduce the density of electrons along the channel. It was also previously reported that fluorine-rich molecules can provide a p-doping effect due to its high electronegativity.^30,31

To evaluate the practicality as sensor platforms, photo-detection and gas sensing based on the MNR array were characterized. Figure 5(a) shows five cycles of light on-off behaviors of both pristine and MNR devices under the illumination of a white light emitting diode (LED, THORLABS) with a power density of 0.25 mW/cm² at 1 V. The photo-response of the MNR device was sufficiently repeatable and stable compared to the reference MNR device (Fig. S4 of the supplementary material) with residues where no photo-responses were found. The photo-response (up to 90% from a base resistance) and recovery (down to 90% from a saturated resistance) times increased slightly from 0.9 s and 0.8 s (for the pristine MoS₂) to 1.4 s and 1.3 s (for the MNR array). Figure 5(b) presents the photo-responsivities at various wavelengths such as 405, 617, and 850 nm, whose corresponding power densities were approximately 3.5, 11.9, and 13.5 mW/cm², respectively. The photo-responsivity was defined as the ratio of the generated photocurrent (I_photo) to a corresponding incident light power (P). The highest photo-responsivity values were 0.52 mA/W for pristine MoS₂ and 530 µA/W for MNR array under 405 nm illumination, which are acceptable results if considering a planar-type device structure with such a long channel length of 20 µm.

The gas sensing for triethylamine (TEA) and acetone gases was also characterized (see Fig. S5 of the supplementary material). Figure 5(c) depicts three cycles of gas on-off behavior upon the exposure to various gas concentrations diluted with high purity air. The injection of TEA gas increased the current of both devices because the TEA adsorption provided electrons to a 2D-MoS₂.³² The adsorption of acetone gas also induced the current increase. There are two possible scenarios for the current increase: releasing free electrons from a reaction between adsorbed oxygen and acetone^33,34 and band-bending at the metal-MoS₂ interface due to polarity change.^35,36 The gas responses at various concentrations are plotted in Fig. 5(d). At 250 ppm, the pristine MoS₂ device showed the gas responses of 143.8% for TEA and 7.8% for acetone, while the MNR device exhibited the reduced gas responses with 14.7% and 2.3%, respectively. We investigated the gas recovery time (for 60% decrease from a top current) of both devices at 500 ppm. Interestingly, the recovery time for acetone gas was markedly reduced to 77.7 s in the MNR device (from 114.5 s of the pristine MoS₂ device). The MNR sensor revealed a relatively small reduction in acetone gas response (7.8% → 2.3%) and faster recovery sensing behavior (114.5 s → 77.7 s) compared to those of the pristine MoS₂ sensor. The low gas response is ascribed to the reduced active area of MNR, on which less gas molecules can be absorbed. By contrast, this phenomenon is beneficial for speedy recovery behavior because the fewer absorbed gas molecules can desorb faster.²³ The minimum concentration was about 30 ppm for TEA and 100 ppm for acetone, at which gas responses were 37.3% and 5.5% for the pristine MoS₂ and 2.4% and 1.7% for the MNR array, respectively.

In conclusion, we demonstrated the scalable nanofabrication of the periodically aligned MNR array with the practically large patterning area of 2.25 cm² using the facile DMT process. The mean width of the MNR array was scalable from 463 nm to 135 nm, and the optimized width was measured to be 260 nm at 650 nm pitch. The residue-free MNR array exhibited the practical and robust material characteristics as futuristic sensor platforms. Our DMT based-nanofabrication technique proved to be very favorable for controlling the nanoribbon width, and pitch, thus probably expecting to employ it for scalable patterning of other 2D materials.

See supplementary material for detailed experimental description and data.

This work was supported by the Pioneer Research Center Program (No. NRF-2016M3C1A3908893) and by the Basic Science Research Program (No. NRF-2016R1A2B4006395) through the National Research Foundation of Korea (NRF) funded by the Ministry of Education. The research was partially supported by the GIST Research Institute (GRI) project through a grant provided by GIST in 2018.