Highly Crystalline CVD-grown Multilayer MoSe2 Thin Film Transistor for Fast Photodetector

Hexagonal molybdenum diselenide (MoSe2) multilayers were grown by chemical vapor deposition (CVD). A relatively high pressure (>760 Torr) was used during the CVD growth to achieve multilayers by creating multiple nuclei based on the two-dimensional crystal growth model. Our CVD-grown multilayer MoSe2 thin-film transistors (TFTs) show p-type-dominant ambipolar behaviors, which are attributed to the formation of Se vacancies generated at the decomposition temperature (650 °C) after the CVD growth for 10 min. Our MoSe2 TFT with a reasonably high field-effect mobility (10 cm2/V · s) exhibits a high photoresponsivity (93.7 A/W) and a fast photoresponse time (τrise ~ 0.4 s) under the illumination of light, which demonstrates the practical feasibility of multilayer MoSe2 TFTs for photodetector applications.

by Kwon et al. 9 to enhance the photoresponsivity of multilayer MoS 2 phototransistors. On the other hand, it is known that MoSe 2 can provide higher photoresponsivity compared to MoS 2 due to the quantum confinement effect during the bandgap transition 10 , which implies that using an advanced device structure may not be needed for MoSe 2 phototransistors to achieve high sensitivity. In addition, Choi et al. 6 reported that TMD multilayers have an advantage over the monolayers that photoresponse is achievable over a broad range of the electromagnetic spectrum from ultraviolet to near infrared. Therefore, multilayer MoSe 2 can be a strong contender for an active channel material of future phototransistors.
In this study, we present the chemical vapor deposition (CVD) growth of MoSe 2 multilayers at a relatively high pressure. The CVD methods reported up to date involve low-pressure deposition with slow nucleation rates [11][12][13][14][15] , resulting in triangular single layers terminated by either transition metal (e.g., Mo) or chalcogen atom (e.g., Se) for TMDs. In contrast, we demonstrate that indirect-bandgap MoSe 2 multilayers can be grown by using a high-pressure CVD method. Based on the two-dimensional nucleation theory, a relatively high pressure at a fixed temperature can induce a large nucleation rate before film growth occurs, and thus, the formation of multilayers is promoted. The microstructure of hexagonal MoSe 2 grains is examined using X-ray diffraction (XRD), high-resolution transmission electron microscopy (TEM) and Raman spectroscopy. Our multilayer MoSe 2 TFTs exhibit ambipolar behaviors with high photoresponsivity (93.7 A/W) and reasonably large field-effect mobility (~10 cm 2 /V · s). This highly crystalline and photo-responsive multilayer MoSe 2 is anticipated to be used in a myriad of potential applications for interactive electronics.

Results and Discussion
Figure 1(a) shows a schematic CVD process for the growth of hexagonal multilayer MoSe 2 . Unlike recent studies 16,17 , MoO 3 and Se are contained in the same alumina boat located at the upper stream of the CVD furnace. A SiO 2 /Si substrate is placed downstream, facing up on the alumina boat and is left to react with the source materials. The furnace temperature was ramped up to 800 °C and 750 °C for the sources and the substrate, respectively, at a rate of 15 °C/min in an evacuated ambient. The vaporized source molecules are efficiently transferred to the substrate surface by the temperature gradient with the aid of carrier gases such as Ar and H 2 . During the chemical reaction between MoO 3 and Se, H 2 gas acts as a catalyst 18 . After the CVD process, hexagonal MoSe 2 crystallites of various thickness and size were obtained. Figure 1(b) shows an optical image of an as-grown hexagonal MoSe 2 grain with a thickness of approximately 11 nm, measured by atomic force microscopy (AFM). In comparison, Fig. 1(c) shows the height profile of an as-grown MoSe 2 bilayer using a relatively low pressure (< 760 torr). While the thickness of a MoSe 2 monolayer lies generally between 0.6 nm and 1 nm [18][19][20] , hexagonal MoSe 2 multilayers obtained in the present work show an average thickness of approximately 12 nm with the size in the range of 4-9 μ m. The shape of film depends on the film thickness, which is attributed to the difference in the edge formation energy between Mo-edge or Se-edge termination. In single or few layered MoSe 2 films, the difference in the edge formation energy depending on Mo-edge (100) or Se-edge ( ) 100 termination results in triangular shaped grains, but since the effect of specific edge termination is cancelled out by alternating Mo-and S-edges in multilayer MoSe 2 films, forming a hexagonal shape is expected from the hexagonal crystal structure of MoSe 2 film 21 .
The relatively large flow rates of injected Ar and H 2 gases in this work are anticipated to increase the pressure in the chamber. Based on two-dimensional nucleation theory 22 , a relatively high pressure induces large nucleation rates, and thus multiple nuclei may form on already existing ones, resulting in multilayered structures. The rate of forming stable two dimensional nuclei is usually described by the following equation: where n s is the concentration of single adsorbed atoms on a closed packed surface, n* is the concentration of atoms in contact with a two-dimensional nucleus with a critical size, ν the vibration frequency, Δ G m the free energy of activation for a surface diffusion jump, and Δ G 2D the free energy of formation of a critical nucleus. The latter is again represented by the following relationship: where V m is the molar volume of the solid phase, σ s (T) is the free energy of step formation per unit length, h the height of the monolayer step, and Δ g the difference of molar free energy between the vapor and solid phases in the case of CVD deposition. The growth of MoSe 2 layers is attributed to the condensation of vaporized MoO 3 and Se radicals on the substrate, forming Mo x Se 2-x nuclei. As the pressure increases, the n s also increases, resulting in the enhancement of the nucleation. Another effect of higher pressure is the increase of Δ g, as more MoSe 2 molecules assemble on the substrate surface. A relatively large Δ g value results in decreased Δ G 2D barrier, thereby contributing to higher nucleation rate rather than growing from the existing nuclei. It is thus expected that the growth of multilayers is attributed to the formation of additional nuclei to the existing ones with a large nucleation rate. In order to confirm this theory, relatively thin few-layered MoSe 2 films were grown at a lower pressure. Figure 1(c) is an optical image of a 1.5 nm-thick MoSe 2 grain, which was synthesized at a reduced pressure using lower gas flow rates (Ar: 50 sccm, H 2 : 10 sccm). The pressure-dependent film thickness was also demonstrated previously for graphene and graphite layers: Graphite was obtained at relatively high pressures 23 , while lower pressures were used for the formation of graphene 24 . Figure 2(a) shows the XRD pattern of our MoSe 2 having a clear hexagonal monocrystalline structure 25 . The symmetry of 2H MoSe 2 belongs to the space group D 4 64 (P6 3 /mmc), which reveals characteristic peaks at 2θ = 13.72°, 27.62°, 41.88°, 56.59° and 56.97° corresponding to the (002), (004), (006), (110) and (008) diffractions for MoSe 2, respectively. The most intense (002) peak indicates the preferential growth of the MoSe 2 crystallites in the (002) direction. To assess the presence and quality of the MoSe 2 films, Raman spectroscopy with a laser wavelength of 514.5 nm was carried out. The as-grown MoSe 2 atomic layers in this study exhibit several signatures of MoSe 2 in the Raman shift ranging from 200 cm −1 to 360 cm −1 . Along with the out-of-plane A 1g mode ( Fig. 2(b)) the atomic vibrations corresponding to several less prominent modes including the E 2g 1 (in-plane) and B 2g 1 modes are in agreement with information provided in former reports available in the literature 10,26,27 . The latter modes have significantly lower intensities compared to the most intense out-of-plane A 1g peak located at 240.9 cm −1 . The typical Raman active modes, i.e. the broad & weak E 2g 1 peak located at 287.4 cm −1 and the B 2g 1 peak located at 350 cm −1 , are observed. The B 2g 1 mode is a shear mode corresponding to the vibration of two rigid layers against each other and appears at relatively low frequencies. The A 1g mode is an out-of-plane vibration involving only the chalcogen atoms (Se) while the E 2g 1 mode involves the in-plane displacement of the transition metal (Mo) and chalcogen atoms (Se). Figure 2(c) shows a plan view low magnification TEM image of an as-synthesized MoSe 2 flake about 3-4 μ m large. The inset in Fig. 2(c) represents a selected area electron diffraction (SAED) pattern taken from within the flake. The pattern reflects well the hexagonal monocrystalline structure along the (002) zone, confirming the XRD results. Figure 2(d) consists of a high resolution TEM image of the thin edge of a MoSe 2 flake, and the inset is a fast Fourier transform (FFT) pattern from the entire area of the figure. The above analyses clearly indicate the presence of a highly crystalline hexagonal MoSe 2 phase. Figure 3(a) shows a three-dimensional (3D) schematic of a MoSe 2 TFT device. The electrodes consist of Ti as an adhesion layer and Au, and the devices were annealed at 200 °C under atmospheric conditions with Ar and H 2 in order to reduce contact resistance and remove the photoresist remnants. Figure 3(b) shows a 3D topography AFM image, where the thickness of the MoSe 2 channel is approximately 20 nm. Figure 3(c) shows the typical drain current (I ds ) versus gate voltage (V gs ) characteristics of the TFT and the extracted mobility values are also plotted as a function of V gs . The field effect mobility (μ eff ) was calculated using the following relationship; μ eff = g m * L/(WC ox V ds ), where L is the channel length (~6.35 μ m), W is the channel width (~2.12 μ m), C ox is the capacitance of the gate insulator per unit area, and V ds the applied drain-source voltage (1 V). The maximum transconductance, g m , was extracted to be approximately 54 μ S. The devices exhibit an ambipolar behavior with a predominant p-type characteristic, with the highest μ eff being approximately 10 cm 2 /V · s and an ON/OFF current ratio of ~10 3 , while the electron mobility was extracted as 2.14 cm 2 /V · s, at the positive region. The output characteristics (I ds − V ds ) were measured in the negative V ds range (Fig. 3(d)), which also depict clear p-type behaviors.
It is worth investigating the characteristics of p-type MoSe 2 TFTs, since most recent studies on MoSe 2 TFTs exhibited n-type behaviors 25,28 . Here, it is hypothesized that the band structure of hexagonal MoSe 2 multilayers is influenced by a large density of trap sites created by an annealing process during the CVD growth. A recent study demonstrated that irradiation with MeV α particles or thermal annealing at sub-decomposition temperature (~600 °C) creates anion vacancies in TMD materials such as MoS 2 , MoSe 2 , and WSe 2 29 . Similarly, we can expect that Se vacancies may be created in the present MoSe 2 film when the decomposition temperature (~650 °C) is reached after the CVD growth for 10 min. Such defects may also significantly affect the electrical behavior through Fermi level pinning, making the theoretical prediction of the Schottky barrier height (Φ Bn = Φ m − χ, where Φ m and χ are the metal work function and the semiconductor's electron affinity, respectively) ineffective 30,31 . In order to extract the sub-gap states, a temperature-dependent analysis was performed 32,33 . Figure 4(a) shows the I ds − V gs characteristics at different temperatures between T = 300 and 400 K. Figure 4 , where I ds,0 is a prefactor and E a is the activation energy 32 . The variation of E a at different gate voltages is shown Fig. 4(c), from which the density of sub-gap states can be obtained by where q is the elementary charge. Figure 4(d) shows a large density of sub-gap states in the band gap near E V + 0.35 eV (E V being the valence band maximum), which is 0.07-0.2 eV below the midgap energy (E m ), since the band gap of bulk MoSe 2 is reported to be 0.84-1.1 eV 10,34,35 . Therefore, at the Ti/Au metal-MoSe 2 junction, Fermi level pinning 36 caused by the gap states is expected to occur in such a way that the Schottky barrier height for holes becomes smaller (Φ Bp ≈ E g /2 − 0.14 eV) than that for electrons (Φ Bn ≈ E g /2 + 0.14 eV), resulting in p-type-dominant ambipolar behavior as shown in Fig. 3(c).
In order to investigate the optoelectronic properties of MoSe 2 multilayers, the photoresponse of the TFTs were examined. Figure 5(a) shows the transfer characteristics under illumination with a 638-nm laser as a function of V gs at various incident power densities (from 20 to 2560 mW/cm 2 ). The drain bias (V ds ) was fixed at 1 V. The photoresponsivity is defined as R = I ph /(P inc S), where R is the photoresponsivity, I ph (= I total − I dark ) is the photo-induced photocurrent, S is the channel area of the device and P inc (W/ cm 2 ) = P tot /A laser (A laser is the area of laser spot) is the incident power density. The calculated photoresponsivity values with respect to the incident power density are shown in Fig. 5(b). The maximum photoresponsivity of 93.7 A/W was achieved at V gs = − 65 V with the lowest incident power density (20 mW/cm 2 ). Notably, this is the highest value reported up to date concerning CVD-grown MoSe 2 TFTs, and the photoresponsivity is also comparable to that of mechanically exfoliated MoSe 2 TFT [37][38][39][40] . Recently, Tongay et al. 29 reported that the trap sites related to the anion vacancies, whose energy levels are located between the conduction band (CB) and the valence band (VB), can enhance the photoresponsive properties of TMD. Our temperature-dependent measurements (Fig. 4) also exhibit a large density of sub-gap states in as-grown multilayer MoSe 2 , which can be the origin of the large photoresponsivity in our multilayer MoSe 2 TFTs, by providing excess photo-induced hole carriers upon exposure to light. To evaluate the devices for potential application as photodetectors, the photoswitching behavior was also examined. Figure 5(c) shows the time-resolved drain current when the incident laser is switched on and off with a power density of 2560 mW/cm 2 and a time period of 20 s, while the gate voltage is fixed at 20 V. As shown in Fig. 5(d), the photoswitching behavior consists of a relatively short rising time (τ rise ~ 0.4 s) and a short decay time (τ decay ~ 0.2 s) that form a nearly-ideal rectangular pulse. Such characteristics indicate that MoSe 2 devices are promising for photodetector applications.
In conclusion, we presented highly sensitive phototransistors based on CVD-grown hexagonal MoSe 2 multilayers. A relatively high pressure (> 760 Torr) in the CVD chamber, originating from a large flow rate of injected Ar and H 2 gases, stimulates the formation of multiple nuclei, resulting in multilayered MoSe 2 nanosheets. The decomposition temperature of MoSe 2 (~650 °C) was reached after the CVD growth for 10 min, which is supposed to induce Se vacancies in MoSe 2 . Such defects are believed to be the origin of the observed ambipolar conduction in our multilayer MoSe 2 TFTs, through the Fermi level pinning at the metal-MoSe 2 semiconductor interface. Moreover, the Se vacancies can enhance the optical properties of MoSe 2 devices, which were manifested by the highest photoresponsivity (93.7 A/W) reported to date and a fast response time (τ rise ~ 0.4 s) to the incident light. The results presented in this work will open up a new route for the fabrication of interactive electronics incorporating active-matrix displays and photosensing devices.

Methods
Synthesis of hexagonal MoSe 2 particles. MoSe 2 films were synthesized inside a CVD furnace composed of two zones; a 2-inch diameter horizontal tube furnace with a 2-inch diameter quartz tube. MoO 3 powders (Sigma Aldrich 99.5% purity) were placed on the right side of an alumina boat. Se powders (Sigma Aldrich 99.5% purity) were placed on the left side of the same boat. A 3 × 3 cm Si wafer with 300 nm SiO 2 grown on it was cleaned using acetone and isopropyl alcohol. The substrate was placed on top of the boat that is facing up. This configuration positions the sources in the hot zone and the substrate in the cold zone. The hot and cold zones were heated up to 800 °C and 750 °C, respectively, at 15 °C/min. The reaction chamber was constantly filled with 120 sccm Ar and 20 sccm H 2 gas for 20 min to allow the reaction to take place. After MoSe 2 growth for 20 min, the chamber was slowly cooled down to 650 °C for 10 min to generate Se vacancies and quenched to room temperature. Synthesis of triangular bilayer MoSe 2 particles. The substrate and sources were put on the same place of the multilayer growth method. The critical differences between the two growth methods were injection gas quantity and the chamber pressure. The chamber was created a vacuum state using a pump. After purging, the gases, 50 sccm Ar and 10 sccm H 2 , were constantly injected in the reaction chamber during the growth process. The reaction chamber pressure was kept around the atmospheric pressure. The other processes were same as the multilayer growth method.
Device fabrication. For the source and drain electrode, Ti (20 nm) was first deposited as an adhesion layer and Au was then grown (300 nm) using e-beam evaporation at room temperature. The electrodes were patterned by photolithography, resulting in a channel length of 6.35 μ m. The as-fabricated device was annealed at 200 °C under atmospheric conditions for 2 h while being exposed to 100 sccm Ar and 10 sccm H 2 gas to eliminate the photoresist residue and to reduce the contact resistance.
Characterization. Optical images of the hexagonal MoSe 2 TFT were taken using an optical microscope (BX51M, Olympus Co., JAPAN) with white light (100 W halogen lamp, U-LH100-3) in bright field imaging mode and a 50× objective lens. The TEM images and diffraction patterns were obtained using a transmission electron microscope (FEI Tecnai TM F20) operated at an acceleration voltage of 200 kV. For TEM sample preparation, the sample was cut to a 3 mm disk and the backside of the sample was hand-polished and dimpled down to about 5-10 μ m at the center of the sample. Then, the sample was ion-milled from the backsides at a 4.5° angle and at 4.5 kV using a Gatan PIPS TM until the small hole at the center of the sample was made. The topography of the MoSe 2 phototransistor was measured using an AFM (XE7 Atomic Force Microscope, Park Systems, South Korea) under non-contact mode with a 0.2 Hz scan rate. The electrical characteristics of the phototransistor were measured using a parameter analyzer (Keithley 4200 SCS) at room temperature. The photoresponsive properties of the MoSe 2 phtotransistor were evaluated using an illumination system composed of a Nikon Ti-e microscope with an Acton SP2300 spectroscope and a Zolix TLS3900x-500 tunable light source.