Optical lithography technique for the fabrication of devices from mechanically exfoliated two-dimensional materials

Optical lithography technique has been applied to fabricate devices from atomically thin sheets, exfoliated mechanically from kish graphite, bulk MoS2 and WSe2. During the fabrication processes, the exfoliated graphene, few-layer MoS2 and WSe2 sheets have been patterned into specific shapes as required and metal contacts have been deposited on these two-dimensional sheets to make field effect devices with different structures. The key to the successful implementation of the technique is the appropriate alignment mark design, which can solve the problems of aligning photomasks to the random location, orientation and irregular shape exfoliated two-dimensional sheets on the substrates. Raman characterization performed on the patterned two-dimensional sheets after the fabrication processes shows that little defects have been introduced during fabrication. Field effect has been observed from I-V characteristics with the highly doped silicon substrate as the back gate. The extracted field effect hole and electron mobilities of graphene are ~1010cm2V-1s-1 and ~3550cm2V-1s-1 respectively; and the field effect carrier mobilities of MoS2 and WSe2 are ~0.06cm2V-1s-1 and ~0.03cm2V-1s-1, separately, which are comparable with experimental results of other reports. Display Omitted Optical lithography has been applied for the fabrication of 2D electronic devices.The exfoliated graphene, few-layer MoS2 and WSe2 sheets have been patterned.The key to the successful implementation is the appropriate alignment mark design.The final electronic devices show little defects.


Introduction
As Moore's Law drives the feature size of silicon transistors to a few nanometers, device operation will reach its limits. Therefore, in order to improve future circuit performance, alternative materials and their associated fabrication technologies need to be developed. In the pursuit of new materials for circuits beyond the silicon era, a great deal of attention has been paid to atomically thin materials such as graphene [1], two-dimensional (2D) hexagonal boron nitride (h-BN) [2] and transition metal dichalcogenides (TMDs) [3][4][5]. Each of these materials has its own particular physical properties that may play an important role in building high performance devices. For example, graphene is well known for its high carrier mobility [1]. The electron mobility of graphene from kish graphite or highly ordered pyrolytic graphite (HOPG), prepared by mechanical exfoliation, can reach as high as 15,000 cm 2 V −1 s −1 even at room temperature, outperforming that of silicon by ten times. However, pristine graphene does not have a bandgap, a property that limits its usefulness in certain applications. Some of the TMDs (such as MoS 2 and WSe 2 ) have attracted much attention for their high on/off current ratio and low off-state current, due to their intrinsic bandgap [3][4][5], which is complementary to graphene.
Although several large area 2D material synthesis methods, including sublimation of SiC [6] and chemical vapor deposition (CVD) [7][8][9][10][11], have been studied extensively, mechanical exfoliation method can produce 2D flakes with better crystal quality and purity [1,4]. Thus, from the fundamental research, device fabrication, and cost points of view, the mechanical exfoliation technique still prevails. However, other characteristics associated with the exfoliated 2D material sheets are the small size (typically in the micron range), the random distribution and orientation of the sheets on the substrate, as well as the irregular shape of the sheets. Such features of exfoliated sheets pose challenges to device fabrication, mainly in two aspects: patterning of the 2D materials and alignment of the metal contacts to the patterned 2D sheets.
Thus far, most researchers try to avoid patterning the exfoliated 2D materials and fabricate electronic devices on as-exfoliated pristine 2D materials, which make the dimension of the devices hard to control and limits the practical application of such kind of devices. In addition, electron beam lithography (EBL) is the dominant patterning technique in aligning metal contacts to exfoliated 2D sheets to make electronic devices. However, EBL has some drawbacks. For instance, EBL is a serial process and can take a long time to expose the metal contact openings. Moreover, the electron beam irradiation can induce defects in the 2D sheets [12][13][14], which may degrade the electrical properties of 2D materials [15,16].

Contents lists available at ScienceDirect
Microelectronic Engineering In our work, we report the development of the fabrication of devices from exfoliated 2D sheets using optical lithography. The advantage of employing optical lithography lies in the low cost of the process, the speed of exposure and ease of patterning. The key to the successful implementation of the technique is the alignment mark design. The technique has been applied to fabricate electronic devices designed with a field effect transistor (FET) structure -field effect devices (FEDs)using pre-patterned graphene, few-layer MoS 2 and WSe 2 sheets as the channels, respectively. To our knowledge, the optical technique reported here is the first of its kind for devices fabricated from exfoliated atomically thin materials.  Fig. 1(a) has been employed to divide the substrates into 320 μm × 320 μm grids, so that desired exfoliated 2D sheets can be found under the optical microscope and traced via the labels on the grid lines. In the case of graphene, in order to solve the problem of aligning to the randomly located, oriented, and irregularly shaped exfoliated graphene, alignment marks consisting of four sets of squares (20 μm × 20 μm each) arranged in an octagonal shape around a small octagon in the middle have been designed, as shown in Fig.  1(b). Since any one pair of squares can be used for the alignment, four pairs of squares oriented in different directions can take into account of the random orientation of the exfoliated sheets. Fig. 1(c) and Fig. 1(d) show photomask design of the channel and the electrodes of graphene FED, respectively. Each design has its own alignment marks (crosses), to be aligned onto a pair of squares in the pattern of Fig. 1(b). For the MoS 2 and WSe 2 devices, a different structure has been fabricated with the photomasks in Fig. 1(f, g). Instead of octagonal-square alignment marks in Fig. 1(b), a pair of crosses alignment marks as in Fig. 1(f) has been used to align the pattern in Fig. 1(g) to that in Fig. 1(f). Fig. 1(e, h) shows how the different layers of photomasks are incorporated together.

Experiments
Firstly, 280 nm of silicon dioxide (SiO 2 ) has been grown on highly ndoped silicon wafer by thermal wet oxidation, which results in optimal color contrast between the 2D material and substrate [17,18]. The  silicon substrate serves as a back gate electrode for the FED, while the SiO 2 serves as the gate dielectric. The wafer then has been diced into 1 cm × 1 cm squares. Afterwards, grid lines ( Fig. 1(a)) have been defined on substrates by etching 20 nm deep trenches into the SiO 2 layer with reactive ion etching (RIE). Then, the substrates have been cleaned by sonication in acetone, isopropyl alcohol (IPA), de-ionized (DI) water sequentially and blown dry with N 2 gas. Thereafter, the SiO 2 surface has been treated with O 2 plasma to enhance the bonding between the 2D material and SiO 2 surface [19]. Then, 2D materials have been exfoliated mechanically from the kish graphite (Graphene Supermarket, Inc.), bulk  MoS 2 , and WSe 2 (2D Semiconductors, Inc.) with scotch tapes and transferred onto the substrates. To avoid leaving much tape residual, water soluble tape (3 M water soluble wave solder tape 5414) has been used. After the mechanical exfoliation, the substrates have been soaked in 60°C DI water to remove the tape residual. Once the graphene and few-layer MoS 2 , WSe 2 sheets have been identified by optical microscope according to their color contrast [18,20,21] (Fig. 2(a), Fig. 3(b), and Fig. 4(a, b)), the locations of the 2D sheets on substrates have been recorded by the grid labels they belong to.
Then, for graphene FED fabrication, negative photoresist AZ 2035 has been spun onto the substrate and the octagonal-square alignment marks ( Fig. 1(b)) have been defined in SiO 2 layer by optical lithography and RIE of SiO 2 with the center of the octagon pattern in photomask aligned to the graphene sheet with the help of Karl Suss MA/BA8 Mask Aligner, as shown in Fig. 2(b, c) and Fig. 3(c-e).
Then, photoresist has been removed with acetone. For MoS 2 and WSe 2 devices fabrication, it is not necessary to etch the octagonalsquare alignment marks into SiO 2 , since the CF 4 based plasma that is employed to pattern the MoS 2 and WSe 2 sheets also etches into the SiO 2 , thus leaving a pair of crosses as alignment marks in the SiO 2 (as shown in Fig. 4(c, d)).
Subsequently, the 2D channels of FEDs have been patterned and defined, as follows (shown in Fig. 3(f-h) and Fig. 4(c, d)). After a layer of positive photoresist SPR 350 has been spin coated on the substrates, for the graphene device, the photomask in Fig. 3(f) has been aligned to the substrate by aligning the crosses to a certain pair of etched octagonal-square alignment marks in SiO 2 . After resist exposure and developing, the etch mask with 2 μm width for the channel fabrication has been formed, as indicated in Fig. 3(h). For the MoS 2 and WSe 2 devices, the photomask with a different pattern (as shown in the insets of  Fig. 4(c, d)) has been used for photoresist exposure. Then, the graphene sheet has been patterned by O 2 plasma and CF 4 based plasma has been used to pattern MoS 2 and WSe 2 sheets. After removing the photoresist by acetone, the channels of the FEDs have been formed and a pair of crosses alignment marks has been transferred into SiO 2 for the MoS 2 and WSe 2 devices (Fig. 4(c, d)).
For the drain (D) and source (S) contacts fabrication, the regions of contacts with a gap of 3 μm for the graphene device and 15 μm for MoS 2 and WSe 2 devices have been defined on negative photoresist AZ 2035 using photolithography with the help of alignment marks in SiO 2 (octagonal-squares for graphene and a pair of crosses for MoS 2 and WSe 2 ) formed in previous processes. Afterwards, Ti (10 nm)/Al (200 nm) metal stack has been deposited through E-beam evaporation system and lifted-off in acetone solvent, as shown in Fig. 3(k) and Fig. 4(e, f). After metal deposition, the devices have been annealed in forming gas (N 2 /H 2 ) at 300°C for 2 h to remove the photoresist residual [22]. Electrical characterization has been carried out using a Keithley 4200-SCS parameter analyzer and Micromanipulator manual probe station at room temperature in air. Raman measurements have been performed on the 2D channels in a confocal Raman spectrometer (inVia Renishaw) with a 100× magnification objective in air ambient environment. The wavelength of the laser is 514 nm. In order to reduce the sample heating effect, the laser power has been kept~200 μW. Fig. 5(a, b) shows the Raman spectrum of the graphene channel after the device fabrication. Raman peaks at 1350 cm − 1 , 1580 cm −1 and 2700 cm − 1 , identified as D band, G band and 2D band respectively, have been observed (Fig. 5(a)). A closer look at the 2D band reveals that the peak is symmetric and can be fitted very well with a single Lorentzian peak (Fig. 5(b)), indicating that the graphene channel is single layer graphene (SLG) [23]. The D band around 1350 cm − 1 is observed to be significantly lower than the G band, suggesting that little defects have been induced to the graphene via the fabrication processes. Fig. 5(c) depicts the Raman spectrum of the few-layer MoS 2 channel. The peaks at 382.7 cm − 1 and 407.6 cm − 1 are attributed to the in-plane mode (E 1 2g ) and out-of-plane mode (A 1g ), respectively [24]. The Raman spectrum of the few-layer WSe 2 channel is displayed in Fig. 5(d). The E 1 2g mode (247.8 cm − 1 ) and A 1g (257.5 cm − 1 ) mode [25] also have been observed. The weak peak at 307.1 cm − 1 arises from the interlayer interaction [26]. No other notable Raman peak belonging to other materials (except SiO 2 from the substrate) is found in the Raman spectrum of MoS 2 or WSe 2 channel within the 100-1000 cm − 1 range (not shown here). The good signal-to-noise (S/N) ratio in Fig. 5(c, d) demonstrates that little damage has been introduced to MoS 2 and WSe 2 sheets from the fabrication processes. Fig. 6(a-f) shows the transfer characteristics ( Fig. 6(a-c)) and output characteristics (Fig. 6(d-f)) of graphene, MoS 2 and WSe 2 based FEDs. Since graphene, MoS 2 and WSe 2 based devices own significantly different electrical properties, due to different electronic band structures of the 2D materials and the barrier heights between the metal contacts and 2D materials, various voltage sweeping ranges have been employed for different devices to exhibit the electrical properties of each device distinctly. The influence of the gate voltage V GS on the drain current I DS with fixed drain voltage V DS for graphene, MoS 2 and WSe 2 devices is depicted in Fig. 6(a-c). For graphene FED (Fig. 6(a)), the drain current I DS decreases as the gate voltage increases from − 25 V to 5 V, and starts to increase when the gate voltage surpasses 5 V, showing ambipolar response. The Dirac point, where the drain current is at its minimum and the carriers are depleted most, is seen to occur at around 5 V. When V GS is swept from 5 V to −25 V, a low on/ off current ratio of~1.4 is obtained, due to the absence of bandgap [27]. Compared with graphene, MoS 2 FED exhibits the notable n-type semiconducting behavior (the threshold voltage is about 30 V as shown in Fig. 6(b)), while WSe 2 FED shows p-type behavior with the threshold voltage of −20 V (Fig. 6(c)) , which is consistent with the results of Ref. [3] and Ref. [5]. In addition, the corresponding current on/off ratios of the MoS 2 and WSe 2 devices are~2 × 10 3 (V GS is swept from 0 V to 60 V) and~5 × 10 2 (V GS ranges from 0 V to − 70 V), respectively, which are much higher than that of graphene.

Results and discussion
With the diffusive transport model proposed by Kim et al. [28], both the hole and electron mobilities of the graphene channel with the dimension of 3 μm in length and 2 μm in width have been extracted to be ultrahigh (around 1010 cm 2 V −1 s −1 and 3550 cm 2 V −1 s −1 respectively), which correspond well to the reported values of unsuspended graphene [29]. Moreover, the current does not go to zero at the Dirac point, indicating the presence of high intrinsic carrier density in the graphene. The extracted intrinsic carrier density is about 1 × 10 12 cm − 2 . This means that at zero gate voltage, the Fermi energy is below the Dirac point, and the energy-wavevector (E-k) relationship may no longer be linear, which could explain the asymmetry of the I DS -V GS curve observed and the lower hole mobility compared with the deduced electron mobility. For MoS 2 and WSe 2 FEDs, the mobilities have been extracted using the equation μ = [dI DS /dV GS ] × [L/(WC i V DS )] [30], where L = 15 μm is the channel length, W = 2 μm is the channel width, and C i = 1.23 × 10 −4 F/m 2 is the capacitance between the channel and back gate per unit area (ε 0 ε r /d; ε r = 3.9; d = 280 nm). The mobilities of MoS 2 and WSe 2 FED with Ti metal contacts extracted from two point measurements are~0.06 cm 2 V −1 s −1 and~0.03 cm 2 V −1 s −1 , in agreement with the previous reports [3,31,32]. As seen, a mild hysteresis is observed when V GS is swept in reverse directions. The hysteresis can be attributed to charge-trapping effects at the MoS 2 or WSe 2 /SiO 2 interface and/or absorbed molecules (moisture and O 2 ) on the MoS 2 or WSe 2 surface [33,34]. Fig. 6(d-f) shows the dependence of drain current on the drain voltage (I DS -V DS ) with different gate voltages V GS for the graphene, MoS 2 and WSe 2 based FEDs. The I DS -V DS curves of graphene FED (Fig. 6(d)) have been observed to be linear when V DS is swept from − 0.1 V-0.1 V, suggesting that the graphene device behaves as a resistor. However, for MoS 2 or WSe 2 FEDs, I DS -V DS plots ( Fig. 6(e, f)) shows significantly different behavior. When the drain bias V DS of MoS 2 FED is in the range of 0-7 V, the drain current I DS is relatively small and stays almost constant for various gate voltages V GS , as shown in Fig. 6(e). When the drain bias V DS surpasses a certain value and the gate voltage V GS is beyond the threshold voltage, the drain current I DS starts to increase remarkably, which is consistent with the results reported in Ref. [34,35]. The nonlinearity of the I DS -V DS curve could be caused by the Schottky barriers formed between the 2D MoS 2 and Ti layer [36]. On the other hand, for the WSe 2 FED (Fig. 6(f)), the drain current I DS increases at low negative drain voltage and saturates to a certain extent at higher negative drain bias, behaving as a p-type field effect transistor (FET). The characteristic of current saturation in WSe 2 FED plays an important role in the future applications of organic light-emitting diode (OLED) displays [37].

Conclusions
Optical lithography technique has been applied for the fabrication of electronic devices from randomly distributed and oriented mechanically exfoliated 2D materials (graphene, few-layer MoS 2 and WSe 2 ) with random shape. During the development of fabrication processes, patterning of the 2D material channels and the alignment of metal contacts to the patterned 2D channels to make field effect devices have been achieved. The appropriate alignment mark design and implementation are enabling factors in these processes. According to Raman analysis, the 2D materials in the final devices show little defects. From the electrical measurements, the fabricated devices exhibit field effect characteristics, with the extracted mobilities of the 2D channels corresponding to previous reports. The added advantage of the optical technique is that it is fast, cheap and can be applied to the fabrication of devices from any atomically thin exfoliated material.