Visible range photoresponse of vertically oriented on-chip MoS2 and WS2 thin films

The excellent electrical properties of transition metal dichalcogenide (TMD) 2D materials promise a competitive alternative to traditional semiconductor materials for applications in optoelectronics, chemical sensing, as well as in energy harvesting and conversion. As the typical synthesis methods of TMDs produce nanoparticles, such as single or multi-layered nanoflakes, subsequent strenuous integration steps are necessary to obtain devices. Direct synthesis of the material on substrates would simplify the process and provide the means for large-scale integration and production of practical devices. In our approach, we synthesize MoS2 and WS2 thin films with a simple sulfurization of the respective metal films deposited by sputtering on Si/SiO2 chips, and study their optoelectrical properties at wavelengths of 661 nm, 552 nm, and 401 nm using pulsed lasers. Both TMD thin films are found to show photoresponsivities of up to ∼5 × 10−6 A W−1 with corresponding quantum efficiencies of ∼10−5, which are unexpectedly moderate, and can be attributed to their columnar microstructure, in which the basal planes of the hexagonal lattices are perpendicular to the substrate, thus, limiting the electron transport in the films parallel to the plane of the substrate.

The research field of 2D layered materials has experienced a renewed interest alongside the high popularity of graphene. 1 Unlike graphene, transition metal dichalcogenide (TMD) films have a direct bandgap, which can be exploited in a number of applications including energy storage, 2 gas sensing, [3][4][5] photodetectors, 6 and catalysis. 7 The structure of these 2D materials consists of a transition metal layer (M) sandwiched between two chalcogen layers (X 2 ), which in multi-layer structures are bound together by van der Waals forces. 8,9 Depending on the coordination of the stacked layers, the crystal structure of the material varies: for example, a tetragonal structure in octahedral coordination (1T-phase) results in a metallic band structure in MoS 2 . On the other hand, a hexagonal structure in the trigonal prismatic coordination (2H-phase) forms a semiconducting structure. 1,4 The various combinations of transition metals and chalcogenides provide a great amount of different thin film compositions, of which MoS 2 and WS 2 are the most widely researched. The bandgap of the TMDs depends on the number of the layers, e.g., ∼1.8 to 1.9 and 1.8-2.1 eV for monolayer structures of MoS 2 and WS 2 , respectively; whereas, multilayer films approach the bulk values of ∼1.2 to 1.3 and ∼1.4 eV, respectively. 4,9,10 The bandgaps are in the visible light spectrum, 4 making them useful in photoconductive applications such as solar cells, 11 phototransistors, 10 and optical detectors. 8 In this work, the structure and photoconductivity of MoS 2 and WS 2 thin films that we synthesized by a simple sulfurization of their corresponding metal films are studied. Identical thin films have been proven to have excellent chemical sensing properties in our previous work; 4 thus, the purpose of the current study is to reveal further details of their properties and potential use in devices for optoelectronics.
The samples used for the photocurrent measurements were identical to those we used in our previous experiments for chemical gas sensing. 4 The MoS 2 and WS 2 thin films were made by sulfurizing their corresponding metal thin films at 800 ○ C for 1 h under N 2 flow in a 2 ′′ quartz tube reactor installed in a tube furnace (Thermo Scientific Thermolyne). The metal films were deposited on Si/SiO 2 chips by sputtering (Torr International PVD System).

ARTICLE scitation.org/journal/adv
For electrical probing, Ti/Pt electrodes (500 × 3000 μm 2 each with 500 μm gap) were deposited on the top of the WS 2 and MoS 2 films by sputtering through a laser patterned alumina shadow mask of 150 μm thickness. The photocurrent measurements were carried out using three different laser beams (Coherent ® High Performance OBIS TM Laser System, TEM00 mode, with λ = 661 nm, 552 nm, and 401 nm and 1/e 2 beam diameters of 9 mm, 7 mm, and 8 mm, respectively) modulated with a 1 kHz square wave trigger from a signal generator (33220A, Agilent Technologies, Inc.). A load resistor of 19.6 MΩ was connected in series with the sample and probed using an oscilloscope (Agilent InfiniiVision DSO-X 3024A with 10 MΩ probe, Agilent N2863B, Agilent Technologies, Inc.). A schematic of the measurement setup is provided in the supplementary material (Fig. S1). The measurements were averaged with 4096 samples and laser powers were modulated as follows: 1 mW, 2 mW, 5 mW, 10 mW, and 20 mW for all wavelengths; 50 mW for both 401 nm and 661 nm; and 100 mW for 661 nm. The incident beam power on the samples was calculated by integrating the Gaussian intensity profiles over the exposed area between electrodes. Six samples for both materials were measured for photocurrent. For the response time assessment, the oscilloscope data were fitted with exponential functions (both response and recovery curves at 20 mW power for all wavelengths). Three samples were measured for both MoS 2 and WS 2 to enable statistics. RC parameters for the samples with electrode patterns were measured using an LCR meter at 1 kHz (Agilent 4284A precision LCR meter).
The evaluation of the bandgaps using diffuse reflectance spectroscopy failed on the original on-chip devices due to the strong superposed interference caused by the multiple parallel flat interfaces (i.e., Si-SiO 2 , SiO 2 -TMD, and TMD-air). To circumvent the interference patterns of plane-parallel interfaces, we roughened the surface of a quartz wafer (0.5 mm thickness) using 800-1200 grit sandpapers and, then, synthesized MoS 2 and WS 2 similar to those on the Si/SiO 2 chips. The as-made samples were then analyzed by optical transmittance spectroscopy (Cary 500, Varian, Inc.) at a window of 350-800 nm. Analyses of material structure and composition were carried out on cross-sectioned specimens. The thin lamellae of both types of films were cut out from the samples using focused ion beam (FIB, FEI Helios DualBeam) and mounted on transmission electron microscopy (TEM) grids for imaging, electron diffraction, and elemental analyses (JEOL JEM-2200FS, 200 kV FEG). The elemental composition of the films grown on the roughened quartz substrates was measured by energy-dispersive x-ray spectroscopy (Oxford Instruments X-Max N 80, Zeiss ULTRA Plus FESEM).
TEM analysis of the FIB cross-sectioned sulfurized metal films reveals that both types of materials have a well-oriented and crystallized layered structure (Figs. 1 and S2). The spacings of the vertical fringes are 6.2 ± 0.1 Å and 6.3 ± 0.1 Å (as calculated from electron diffraction patterns) for the sulfurized Mo and W, which are very close to the spacing of (002) planes of bulk hexagonal MoS 2 and WS 2 , respectively. 12,13 The slightly larger values may arise either from excess S between the sheets (or some other defects) or from the imperfect stacking between the sheets due to bending.
The orientation of the layers is mainly vertical, although on the surface of WS 2, we observe some horizontally oriented layers as well. This is in excellent agreement with the XRD patterns of WS 2 and MoS 2 we have reported previously, which were practically lacking the otherwise strong (002) and (102) reflections seen for polycrystalline or horizontally aligned TMDs. 4 The average thicknesses of the films are ∼90 and ∼110 nm for MoS 2 and WS 2 , which are significantly higher than the nominal thickness of the starting metal films (∼20 nm). However, these results are reasonable considering the increase in molar volumes of the metals upon sulfurization (approximately 3.4-fold for Mo → MoS 2 and 3.5-fold for W → WS 2 ).
The growth of vertically oriented MoS 2 /WS 2 flakes via sulfurization of Mo/W films has been reported in the literature. 12,13 Such films have been demonstrated to work well for applications in catalysis or gas sensing, presumably due to the high density of exposed reactive edges. 4,[12][13][14] In particular, it was shown that the vertical growth is preferred over horizontal growth when the thickness of the precursor metal film is more than 3 nm. 13 Because of the lateral confinement on the substrate, the volumetric expansion of the crystal takes place in the out-of-plane direction. Since grains grow preferentially in the fastest growth direction, 15 which is along the highly reactive edges of the planes in 2D crystals, 16 the resulting product has its basal planes in the normal direction of the substrate. It is worth pointing out here that the substrate is amorphous SiO 2 , thus, surface driven epitaxy can be ruled out here. On the other hand, it is not clear where the growth is seeded (surface, metal/substrate interface, or grain boundaries), and, thus, the exact kinetic pathway still remains unknown. We note that even if the flakes are predominantly vertically oriented, the surface does not consist of exposed edges only, as we also can find closed loops of the layered structure as well as lateral flakes on the top of the film (Fig. S2).
Optical absorption spectra (including scattering losses) calculated from transmission measurements show the excitonic peaks of both MoS 2 (at around 627 nm and 676 nm) and WS 2 (at around 528 nm and 626 nm) in Fig. 2. The peak positions are similar to those reported for multi-layered structures [17][18][19] in agreement with TEM imaging of the structures. While both films seem to absorb in the measured spectral range (350-800 nm), because of the strong light scattering from the rough interfaces, it is not possible to fit the absorption edge of the spectra and to determine the exact indirect band gap values of the films.
Since both films absorb in the entire visible spectrum, we applied three pulsed lasers emitting at 401 nm, 552 nm, and 661 nm to assess the corresponding photoconduction. The dependence of photocurrents on the laser powers follow well the simple allometric power function (y = ax b ) as shown in Figs. 3(a) and 3(b) with exponents between 0.24 and 0.42 (Table I)  those reported in the literature for 2D materials and nanostructures, such as graphene (∼0.25), 20 WS 2 nanowire/flake hybrid structure (0.19-0.23), 21 single layer WS 2 (0.4), 22 and MoS 2 thin films (0.67). 23 Such values imply complex, bimolecular process of carrier generation and recombination often observed in materials with high defect densities. 21,22 To be able to assess the photosensitivity of our devices, we calculate the incident laser powers on the devices (P dev. ) by integrating the Gaussian laser beam intensities according to the sensor area between the electrodes. The obtained photosensitivities are quite moderate having the highest values of 5.5 μA/W and 5.  reasonable, since the number of absorbed photons is also decreasing at a given laser power. The corresponding quantum efficiencies are also very moderate (∼10 −5 ). Comparing the measured photosensitivity values to other MoS 2 and WS 2 materials and devices (Table S2), we find a large variation of the reported data. However, it is clear that very large sensitivities were obtained only for small single crystals in a configuration having the channel parallel to the basal plane of the material. 8 Therefore, the low photosensitivity of our macroscopic films is reasonable, as it is caused by the columnar vertical alignment of the basal planes on the surface of the substrate. In a simplified picture, such a vertical alignment implies that a significant fraction of the photogenerated carriers shall inject (tunnel) across the adjacent layers, which is less probable than the seamless transport within the planes of the layers. In addition, stacking faults and other defects, which are certainly present in our films having composed of columnar crystals with a number of grain boundaries, provide a large density of recombination centers for the photogenerated carriers, and inherently result in a fractional power dependence of the photocurrent. 24 Despite the moderate photosensitivity, the rapid transients in the response to laser pulses indicate fast photogeneration and decay with corresponding time constants between 30 μs and 200 μs as displayed in Fig. 4. Considering the impedance and capacitance values of the sensors measured at 1 kHz (∼20 MΩ and 50 pF for MoS 2 and ∼100 MΩ and 20 pF for WS 2 ), the corresponding RC time constants of the measurement setup is ∼1 ms. This parasitic parallel capacitance induced by the electrodes via the silicon substrate effects as a low-pass filter, attenuating the generated signal. The time constant is an order of magnitude higher than the ones extracted from the transient measurements and suggests that photocarrier generation and relaxation is eventually faster than measured. Moreover, the measured time constants are very competitive compared to the other devices in the literature (Table S2), which have significantly better responsitivity. This implies that the performance of our device is, indeed, restricted by the vertical alignment of the structure with regard to electrodes.
In conclusion, sulfurized thin films of Mo and W form hexagonal layered MoS 2 and WS 2 bulk films with interlayer stacking perpendicular to the substrate. Both types of films show photoconduction; however, the photosensitivity and quantum efficiency of such planar on-chip devices is moderate, which can be attributed to tunneling barriers and, thus, limited the transport of photogenerated carriers across the vertical stacks of 2D layers. Despite the limited performance of the devices, our work highlights a simple approach for producing integrated TMD based devices such as photodetectors and switches in the future. Furthermore, with a proper optimization of the thickness of TMD films and with the use of vertical device geometry by applying, e.g., Pt bottom and semitransparent ITO, SWCNTs, or graphene top electrodes, we envision a significant improvement in both photosensitivity and quantum efficiency. The method might be also suitable for growing multilayered thin film heterojunction structures of TMDs by using corresponding evaporated metal thin film stacks as starting materials. In addition, we foresee applications in large area flexible electronics as well using detached TMD films from the substrate with subsequent transfer and lamination onto polymers having electrodes.
See the supplementary material for photoresponse measurement setup schematics, TEM micrographs, electron diffraction patterns, EDX composition information, and a performance comparison table of MoS 2 and WS 2 nanomaterials and devices.