Controllable synthesis and optoelectronic applications of wafer-scale MoS2 films

The chemical vapor deposition (CVD) method is widely used for synthesizing two-dimensional (2D) materials such as molybdenum disulfide (MoS2) because of the process’ simplicity, relatively low cost, compatibility with other process, and tendency to result in high-quality crystalline materials. However, the growth of films with a uniform large area of several square centimeters with control of the number of layers remains challenging. Here, a MoS2 synthesis technique that enables thickness and size control of wafer-scale films with high uniformity and continuity is proposed. This CVD technique is a powerful and simple method to control the layer number and size of MoS2 films without using additive chemicals or a complex process. The thickness of the MoS2 films can be controlled from one to four layers by adjusting the concentration of MoO3. MoS2 films with dimensions greater than 10 cm can be grown by manipulating the Ar/H2S ratio. In addition, a photodetector based on CVD-grown MoS2 is shown to exhibit a high current on–off ratio of 105 and gate-tunability. It also shows a high responsibility of 1.2 A W−1, external quantum efficiency of 345%, and a specific detectivity of 1.2 × 1011 Jones. The proposed CVD technique can provide a facile direction for the controllable synthesis of wafer-scale MoS2 films with diverse applications in future optoelectronic devices.


Introduction
Two-dimensional (2D) transition-metal dichalcogenides (TMDs) are representative layered materials that consist of a transition-metal layer composed of, for example, Mo or W, and two chalcogen layers composed of S, Se, or Te. Two-dimensional TMDs have attracted extensive attention because of their outstanding mechanical, optical, and electrical properties and strong potential for use in applications involving nanoelectronics and optoelectronics [1][2][3][4][5]. In particular, MoS 2 is among the most studied 2D TMD materials because of its unusual physical properties, which include high flexibility, optically controllable valley polarization, and unusual bandgap transition [6][7][8]. MoS 2 has been used in various optoelectronics devices, including photodetectors, light-emitting diodes, and solar cells, because of its high electrical mobility (∼200 cm 2 V −1 s −1 ) and excellent quantum luminescence efficiency [9][10][11]. Because of the strong potential and wide applicability of 2D TMDs, wafer-scale synthesis methods of these materials have been developed to enable their practical device applications. Synthesis techniques such as atomic layer deposition (ALD), metal-organic chemical vapor deposition (MOCVD), and CVD can fulfill this requirement. ALD is a suitable method for growing large-area 2D TMDs with excellent thickness control because of its self-limiting mechanism [12,13]. However, the crystal quality of the resultant nanosheets is relatively low. MOCVD has been used to produce wafer-scale films of 2D TMDs with high crystal quality because, unlike general CVD methods, it uses all-gas-phase precursors [14][15][16]. However, MOCVD has critical shortcomings, including long processing times (>20 h), high operating costs, complex equipment, and a limited number of compatible metal-organic precursors. By contrast, CVD involves a simple process and relatively low-cost setup and provides a high degree of freedom for manipulating the structure of 2D TMDs [17][18][19]. Because of these advantages, the CVD method is the most widely used method for growing 2D TMDs. However, the CVD process also has limitations, including lower crystal quality compared with that of exfoliation samples and low reproducibility of the growth. In particular, CVD suffers from poor control of the thickness and size of the grown films because the use of a solid-state precursor makes controlling the vapor pressure difficult. Although some strategies have been developed for growing multilayer 2D TMDs, the resultant films exhibited a limited size and area, on the scale of micrometers. The synthesis of uniform waferscale films of 2D TMDs with controlled thickness remains a challenge.
A CVD method for the wafer-scale synthesis of a monolayer MoS 2 film using MoO 3 and H 2 S as precursors has been reported [20]. The CVD-grown monolayer MoS 2 showed high uniformity, good continuity, and high electrical mobility (∼1 cm 2 V −1 s −1 ). However, control of the layer number and size of MoS 2 films synthesized using precursors of MoO 3 and H 2 S and their optoelectronic devices applications is also challenging. In the present work, a controllable synthesis method for wafer-scale, homogeneous MoS 2 films is proposed. The relation between the precursor concentration and the size of the grown MoS 2 and that between the precursor concentration and the MoS 2 layer thickness were systematically investigated. The size of the obtained MoS 2 film can be controlled by manipulating the ratio between Ar and H 2 S gases. For instance, the size of the MoS 2 film can be expanded to 10 cm when a Ar/H 2 S ratio of 500 is used in the growth process. The growth length is linearly proportional to the Ar/H 2 S ratio. In addition, the thickness of the obtained MoS 2 film can be controlled from one layer (1L) to four layers (4L) by increasing the concentration of MoO 3 from 5 to 20 mg. The relation between number of layers and the concentration of MoO 3 exhibits a nearly linear proportion. The 1 × 1 cm 2 MoS 2 films (1L-4L) exhibit high uniformity and continuity in area. Their optical properties and thickness were characterized by Raman spectroscopy and photoluminescence (PL) spectroscopy. Moreover, a photodetector array based on a CVD-grown MoS 2 film with high uniformity was fabricated. It showed a high photoresponsivity of 1.2 A W −1 , an external quantum efficiency (EQE) of 354%, and a specific detectivity (D * ) of 1.2 × 10 11 Jones, which are comparable to, or even greater than, those of other devices based on CVD-grown MoS 2 . The photodetector also exhibits highly gate-tunable optoelectronic properties and a high electrical current on-off ratio of 10 5 . These results confirm that wafer-scale CVD-grown MoS 2 has strong potential for use in next-generation optoelectronic devices and systems.

Experimental details
Synthesis of MoS 2 films: Before the synthesis, a 300 nm SiO 2 /Si wafer was cleaned for 10 min in acetone and isopropyl alcohol (IPA), respectively, and then rinsed in deionized water (DIW) several times. MoO 3 powder (99.5%, Sigma-Aldrich) was used as a precursor for the growth of MoS 2 . As shown in figure S1, MoS 2 films were prepared in a custom-made quartz tube furnace under a low pressure of 10 −3 Torr. The MoO 3 powder in a quartz boat was placed at the center of furnace and a Si/SiO 2 substrate was placed downstream from MoO 3 powder. The furnace was heated to 600°C at a rate of 20°C min −1 . After H 2 S gas was injected into the quartz tube at 600°C, a MoS 2 film was directly grown for 30 min by sulfurization of MoO 3 . The sample was rapidly cooled down to room temperature after growth. In whole process, Ar gas used as a transporting material and for control of size with H 2 S gas. The density and supply duration of H 2 S and Ar gas were exactly adjusted by mass flow controller and CVD system.
Characterization of CVD-grown MoS 2 : Raman spectroscopy and PL spectroscopy (NTMDT AFM-Raman spectroscope) were carried out using a 532 nm wavelength laser to obtain PL and Raman spectra. To avoid thermal damage to the samples, the laser intensity was set at a value less than 1 mW. Before measurements, the instrument was calibrated using the Si peak at 520 cm −1 .
Fabrication of electrical devices based on as-grown MoS 2 : Electrical devices were fabricated using conventional photolithography. Metal electrodes of Cr/Au(5/50 nm) were deposited onto as-grown MoS 2 by a lift-off process using an e-beam evaporation system. Isolation of channel area of MoS 2 was performed by additional photolithography and plasma etching. The electrical properties of devices were evaluated using a semiconductor parameter analyzer (Agilent 4155C). The photoelectrical properties of the devices were investigated under illumination of a solid-state laser with a broad wavelength range from 400 to 700 nm.

Results
Figure 1(a) shows CVD-grown 1L MoS 2 on a SiO 2 /Si(300 nm/500 μm) substrate, where the MoS 2 was grown under different Ar/H 2 S gas ratios at a fixed temperature of 600°C and with a fixed MoO 3 concentration of 5 mg. The process was performed in a homemade CVD apparatus at a pressure of 10 −3 Torr. As the ratio of Ar/H 2 S was increased from 50 to 500, the growth length of MoS 2 increased from 2.4 to 10.3 cm. The growth length as a function of the Ar/H 2 S ratio is plotted in figure 1(b), where the sublinear Ar/H 2 S ratio dependency of the growth length shows a slope of ∼0.017. In the present work, the maximum Ar/H 2 S ratio in the CVD apparatus was limited to 500. If the experiments could be conducted at Ar/H 2 S ratios greater than 500, a MoS 2 film larger than 10 cm could likely be grown on SiO 2 . This simple CVD method does not require any additional complex processes. The results reveal that the film size of MoS 2 can be controlled by manipulating the Ar/H 2 S ratio.  Figure 1(c) shows a highly uniform and continuous MoS 2 film grown on the SiO 2 /Si. The Raman spectrum was obtained for qualitative analysis of the MoS 2 . The laser intensity was sustained at 1 mW to avoid thermal damage to the sample. Figure 1(d) shows a typical Raman spectrum of CVDgrown MoS 2 with a peak distance of 20.5 cm −1 between the A 1g mode at 405.3 cm −1 and the E 1 2g mode at 384.8 cm −1 , indicating that it is a perfect monolayer [21,22]. The PL spectrum of MoS 2 also indicates a typical monolayer, showing peaks at 1.89 and 2.03 eV, which correspond to the A 1 and B 1 direct excitonic modes, respectively [23]. The Raman mapping across an area 50 × 50 μm was performed and a related statistical histogram was plotted ( Figure S2). The histogram proves that the average peak distance between the A 1g mode and the E 1 2g mode of 20.5 cm −1 is the same result as single Raman data. The Raman mapping image and histogram show high uniformity and homogeneity of the MoS 2 sample. Figure 2(a) demonstrates that MoS 2 can be synthesized with a thickness from 1L [24][25][26] to 4L over an area of 2 × 2 cm 2 . Each of the CVD-grown MoS 2 films is highly uniform and continuous. The number of layers of MoS 2 was controlled by adjusting the concentration of MoO 3 while the other growth conditions were fixed (temperature: 600°C, Ar/H 2 S = 200). Compared with bare SiO 2 , the sample contrast increases with increasing MoS 2 thickness because of light interference between the sample and the SiO 2 substrate. To identify the number of layers of these samples, Raman spectroscopy was conducted ( figure 2(b)). According to a previous report, Raman spectroscopy is a powerful tool to confirm the MoS 2 thickness from 1L to 4L. The distances between the E 1 2g and A 1g peaks were calculated from the spectra of the 1L to 4L samples (figure 2(b)) (1L: 19.5 cm −1 , 2L: 21.6 cm −1 , 3L: 23.6 cm −1 , 4L: 24.3 cm −1 ). These results are in good agreement with those of a previous report [21,22]. To confirm thickness of these samples more precisely, AFM measurements were conducted. The height profile revealed that the thicknesses of the MoS 2 films were about 0.8, 1.5, 2.5, 3.5 nm ( Figure S3). The thickness of monolayer MoS 2 is about 0.7-0.8 nm. Considering a 0.2 nm roughness value of amorphous SiO 2 substrate, the number of layers of each sample was estimated to be 1, 2, 3, and 4 respectively. Also, AFM topography of MoS2 films with various thickness shows high continuity and uniformity. As the concentration of MoO 3 was increased from 5 to 20 mg, the number of layers of MoS 2 increased from 1L to 4L ( figure 2(c)). The relation between the thickness of MoS 2 and the concentration of MoO 3 shows linear proportionality, demonstrating that the number of layers of MoS 2 can be perfectly controlled by adjusting the concentration of MoO 3 . Thus, the proposed CVD technique is a simple and powerful method to control the thickness of large-scale and highly uniform MoS 2 films without the assistance of complex processes or chemical additives. Large-area CVD-grown MoS 2 films were prepared on 300 nm SiO 2 /Si substrates, and an electrical device based on the synthesized MoS 2 films was fabricated via conventional photolithography, with the deposition of Cr/Au(5/50 nm) electrodes by electron-beam evaporation (figures 3(a), (b)). After the fabrication, the device was annealed at 230°C for 4 h to improve the contact resistance and remove impurities. The electrical properties of the MoS 2 transistors were measured under dark conditions before the optoelectronic characterization ( figure 3(c)). The 1L-MoS 2 field-effect transistor (FET) exhibited typical n-type semiconducting properties ( figure 3(c)). When the gate voltage (V G ) was varied from 0 to 80 V while the optical power and drain voltage (V D ) were fixed at 2.15 nW and 1 V, respectively, the drain current under illumination (I lllumination ) was greater than that under dark conditions (I Dark ). Moreover, the generated photocurrent (I ph = I lllumination − I Dark ) could be enhanced by controlling the gate voltage. That is, the MoS 2 -based photoresistor showed highly gate-tunable optoelectronic properties, consistent with previous reports of devices based on exfoliated MoS 2 or CVD-grown MoS 2 .
To precisely evaluate the enhancement of the photocurrent generation efficiency by the gate voltage, photoresponsivity (R), which is the one of a critical parameters used to characterize the performance of a photodetector, was estimated. The photoresponsivity is defined as the ratio between the output photocurrent (I ph ) and the input optical power (P in ) in the active area of a device: R = I ph /P in . In figure 3(d), the photoresponsivity is ∼1 mA W −1 under a zero gate voltage. As the gate voltage increases to 80 V, the photoresponsivity increases dramatically. The maximum value of the photoresponsivity is ∼0.5 A W −1 at V G = 80 V. The photocurrent at V G = 80 V is 500 times greater than that at V G = 0 V (1 mA W −1 ). As mentioned above, a phototransistor composed of a mechanically exfoliated MoS 2 monolayer exhibits a maximum responsivity of 0.42 mA W −1 at V G = 0 V and 7.5 mA W −1 at V G = 40 V. This result is comparable to those previously reported for CVD-grown MoS 2 films [24][25][26]. Conversely, the photoresponsivity decreases when a negative gate voltage is applied ( figure 3(d)). According to previous studies, the V G -dependent photoresponsivity of a MoS 2 phototransistor is a consequence of n-type doping of MoS 2 [27,28]. After deposition of the Au metal contact, MoS 2 exhibits a quasi-equilibrium Fermi level, which can be moved by controlling the gate voltage. When a negative voltage is applied, the Fermi level shifts from the conductance band of MoS 2 toward its valence band, which indicates the formation of a larger barrier between the conductance band of MoS 2 and the Fermi level of Au metal. The photogenerated carriers are resistant to drift because of the large barrier, and the density of the photocurrent is therefore lower than that under the zero gate voltage. On the contrary, when a positive voltage is applied, the Fermi level moves from the valence band toward the conductance band, which results in the formation of a smaller barrier between the conductance band of MoS 2 and the Fermi level of Au metal. The photocurrent increases because of the smaller barrier.
In general, the photoresponse mechanism of graphene is related to photothermoelectric (PTE) and bolometric effects. However, unlike graphene-base devices, optoelectronic devices of TMDs such as MoS 2 are more dependent on photoconductive (PC) and photogating (PG) effects than on photothermal and bolometric effects [29,30]. The PC effect results in photocurrents by creating electron-hole pairs when photons with an energy greater than the bandgap energy of monolayer MoS 2 (∼1.89 eV) are absorbed in MoS 2 devices. The PG effect is a special case of the photocurrent effect. The localized energy states at defects and impurities can trap the charge carriers generated under illumination. These trapped carriers can form local gates that severely affect carrier conductance. The photocurrent generated by the PG effect is proportional to g m ΔV th (g m = dI D /dV G , which is the transconductance; ΔV th is the change in the threshold voltage) [31]. That is, the PG effect can be described as the difference in V th between dark and illumination conditions. In figure 4(a), a clear change of ∼9 V is observed for the threshold voltage under illumination (V th,dark : 42.2 V, V th,illumination : 51.2 V), which indicates that a phototransistor based on 1L-MoS 2 exhibits strong dependency on the PG effect. The results demonstrate that phototransistors based on exfoliated MoS 2 exhibit a strong photovoltaic effect because of defects and impurities. Moreover, the structural defects and impurities in CVD-grown MoS 2 are assumed to be more abundant than those in intrinsic exfoliated MoS 2 crystals because of the relatively lower crystallinity and greater number of impurities generated by the complex synthesis process. For these reasons, the threshold voltage shifts under incident light until the localized trap states are fully occupied ( figure 3(c)).
To verify the optical power dependency of the photocurrent, the photocurrent is plotted as a function of the optical power in figure 4(a). The experiments were carried out using a laser with a 550 nm excitation wavelength and at a fixed V G of 50 V. To avoid thermal damage, the optical power should be less than 1 mW. When the maximum optical power in the MoS 2 channel area (2000 μm 2 ) was increased from 0.78 to 114.59 nW, the photocurrent increased gradually ( figure 4(a)). The photocurrent under a fixed V D of 1 V was extracted and plotted in figure 4(b). The sublinear relationship between photocurrents and optical powers can be described as I ph ∝ P α , where I ph is the photocurrent, P is the optical power, and α represents a constant. An α value of 0.78 was calculated through the power-function fitting method. According to previous reports, the sublinear power law I ph ∝ P 0.78 can be caused by the defects in MoS 2 and by absorbed molecules and impurities at the interface of MoS 2 and the SiO 2 substrate. Figure 4(c) shows the optical spectral response of a MoS 2 phototransistor at different excitation wavelengths in the range of visible light (from 400 to 700 nm) under V G = 80 V and V D = 1 V. As previously mentioned, the photoresponsivity was calculated by the equation R = I ph /P in . The EQE is a critical parameter to evaluate the performance of optoelectronic devices. The EQE is the ratio between the number of electron-hole pairs that participate in the photocurrent and the number of incident photons. It can be expressed as EQE = (I ph /e)/[P in λ/(h.c.)], where e is the elementary charge (1.6 × 10 −19 C), h is Planck's constant, λ is the excitation wavelength, and c is the speed of incident light. Both the photoresponsivity and EQE tend to decrease with increasing illumination wavelength, consistent with the previously discussed results. In particular, the responsivity and EQE increase to 1.2 A W −1 and 354% at 400 nm, respectively. The bandgap of 1L-MoS 2 is 1.89 eV, which corresponds to a wavelength of ∼660 nm. An incident photon with an energy greater than the bandgap energy of MoS 2 can theoretically be absorbed into MoS 2 phototransistors. According to a previous study, the absorbance of a MoS 2 monolayer decreases dramatically as the illumination wavelength increases from 300 to 1200 nm [32]. The remarkable increases of R and EQE at short wavelengths are attributed to a greater absorbance of photons. In addition, photons with a higher energy (shorter wavelength) can create more photocurrent because of a greater probability of overcoming the localized trapping by defects. Here, we note that most of the EQE values are greater than 100%, except that at 650 nm. The high EQE values are caused by the generation of multiple electron-hole pairs from a single absorbed photon. The carriers in the MoS 2 device can recirculate numerous times before recombination, resulting in greater efficiency.
The other important characteristic of a photodetector is detectivity (D * ). The detectivity as a function of the illumination wavelength is plotted for the same optical power and gate and drain voltages in figure 4(b). If the shot noise from dark current dominates the total noise of a photodetector, then the detectivity can be defined as D * = RA 1/2 /(2eI dark ) 1/2 , where R is the photoresponsivity, A is the effective area, e is the elementary charge, and I dark is the dark current. Figure 4(d) does not show remarkable changes in the detectivity with increasing illumination wavelength. The maximum value of the detectivity is ∼1.22 × 10 11 Jones at a wavelength of 400 nm, which is comparable to the detectivity of photodetectors of TMDs such as ReS 2 , WSe 2 , and MoTe 2 , as well as to the detectivity of phototransistors based on MoS 2 [33][34][35].
Time-resolved experiments were conducted to investigate the photoswitching characteristics and reproducibility of 1L-MoS 2 phototransistors. As shown in figure 5(a), the drain current as a function of time was measured under repeated on-off photoswitching at a fixed V D of 1 V and a constant optical power at a wavelength of 550 nm. Figure 5(a) shows a highly sustainable and stable photoresponse during on-off photoswitching tests. In figure 5(b), the photocurrent was recorded in a higher temporal resolution to calculate the response rate. The rise time of the MoS 2 phototransistor is ∼0.7 s, and the decay time is ∼2 s. This rise time is longer than that in an exfoliated MoS 2 phototransistor (50 ms). According to previous reports, the response time of the photodetector is mainly affected by defect states and adsorbents [36][37][38][39][40]. TMDs such as MoS 2 and WSe 2 synthesized at high temperature show high density of defect states because of chalcogen vacancies. The defect states of vacancies near the Fermi level become additional trapping centers of photogenerated carriers, and yield long response time with low photocurrent [39,40]. Also, the absorbent such as environmental oxygen and water molecules can give critical effects on response time of 2D materials due to high surface to volume ratio. The absorbed oxygen molecules on the surface defects act another trapping center, and result in larger response time. Especially, the oxygen captures the photogenerated hole under illumination, and it hinder recombination of trapped holes and photogenerated electrons [38,41,42]. Because of the environmental absorbents, relatively many photodetectors based on TMDs shows tendency of larger decay time compared to rise time [43]. However, the responsivity and EQE were 70 times greater than those of the exfoliated MoS 2 , as discussed previously. In general, the more abundant localized trap states generated by defects and impurities because of the relatively low crystallinity of CVD-grown MoS 2 can enhance the photoresponsivity and the photoefficiency by the photogating effect but also result in a slower response. Consequently, a MoS 2 phototransistor exhibits a relatively high responsivity and a low response rate.

Conclusion
In the present work, the controllable synthesis of wafer-scale MoS 2 films was achieved via the CVD method using precursors of MoO 3 and H 2 S. The number of layers and the size of MoS 2 films were demonstrated to be controlled through manipulation of the concentration of MoO 3 and the Ar/H 2 S ratio, respectively. The CVDgrown MoS 2 film exhibited high uniformity and continuity. These results show that the proposed CVD technique is a simple and powerful method to control the size and thickness of not only 2D MoS 2 films but also films of other wafer-scale 2D materials. In addition, an optoelectronic device based on the as-grown MoS 2 exhibited gate tunability, a high responsivity of 1.2 A W -1 , an EQE of 345%, and a specific detectivity of 1.2 × 10 11 Jones. The CVD method described here can help expand the range of feasible applications of 2D materials, including MoS 2 , in practical nanoelectronic and optoelectronic devices.

Acknowledgments
This study was supported by the Basic Science Research Program (2021R1F1A1051987, 2018R1A6A3A11047867) through the National Research Foundation of Korea (NRF).

Data availability statement
The data generated and/or analysed during the current study are not publicly available for legal/ethical reasons but are available from the corresponding author on reasonable request.