1. Introduction
Over the last decade, transition metal dichalcogenides (TMDs) have attracted immense interest because of their outstanding electrical, optical and chemical properties [
1,
2,
3]. These TMDs are typically attributed as MX
2, where M is a transition metal atom (i.e., molybdenum, Mo) and X refers to a chalcogen atom (i.e., sulfur, S) to form a TMD (i.e., molybdenum disulfide, MoS
2). Monolayer MoS
2, with a direct band gap of 1.8 eV and a three atoms thick nanosheet, shows potential applications in the fields of electronics, optoelectronics and valleytronics due to optical transparency, high carrier mobility, no-dangling bonds and atomically thickness [
3,
4,
5]. In addition, conformal growth of MoS
2 nanosheets over pre-patterned substrates enables an added advantage for effective routes towards fabrication; the miniaturization of integrated circuits and structurally excellent TMD materials are essential for flexible optoelectronics such as LEDs and may lead to bandgap and exciton engineering through local strain generation due to excellent conformal growth and layer bendability [
6,
7]. The growth on the patterned substrate is also fundamental for the exploitation of the anisotropy by design in MoS
2 and related TMDs, thus enabling the engineering of the electronic band structure as a function of the local strain. Anisotropic effects are relevant in a broad range of applications. For instance, when we consider the resulting nanosheets as 2D rippled membranes, we are devising new physical concepts to enhance the optical, plasmonic, and catalytic performance of the pristine materials. Further, direct growth on engineered patterned substrates promotes strain engineering, with implications on relevant properties of TMDs such as the thermal/electronic transport or the exciton physics, which are dramatically affected by the anisotropy-dependent degree of strain [
8,
9]. On another front, direct growth of high crystalline MoS
2 on conductive substrates would open up a new pathway for easy integration in memory devices by serving as a bottom electrode and also providing an underlying conductive substrate [
10]. MoS
2 and other TMDs, such as TaS
2, are considered as diffusion barriers in ultra-scaled microelectronics (<5 nm technology node) directly in contact with TaN or Cu, which are commonly used as interconnects in back-end-of-line compatible processes [
11,
12]. In this regard, a significant demand to develop an efficient approach to direct growth of MoS
2 on metal or conductive substrates at a large scale is still demanding, although the high process temperature poses severe constraints in terms of substrate stability during growth. In contrast, multilayered MoS
2 with an indirect band gap of 1.3 eV, disordered structures with exposed edge defects, chemical stability and high surface area are a favorable ground for heterogeneous catalysis reaction with promising impact on hydrogen storage and fuel cells [
4,
13,
14,
15]. It is known that excellent catalytic activities of MoS
2 are greatly enhanced by using conductive layers as a growth substrate, thus providing cost-effective, high-performance catalysts in electrocatalysis over Pt [
13].
Before the practical applications, so far, several considerable efforts have been focused on the preparation of large-scale MoS
2. Approaches, such as physical and chemical exfoliation, chemical synthesis, atomic layer deposition, laser annealing, physical vapor deposition and chemical vapor deposition, have been reported [
16,
17,
18,
19]. Among the proposed methods, chemical and mechanical exfoliation, physical vapor deposition, and chemical vapor deposition schemes are the most used ones [
18,
20]. Specifically, chemical exfoliation or sonication are versatile methods for the low cost, scalable production of monolayer 2D materials [
16,
21]. Mechanical, or tape, exfoliation allows to obtain high crystal quality of monolayer MoS
2 but is beneficial for fundamental property studies only [
17]. Indeed, small size, nonuniform thickness and agglomeration in solution are drawbacks of this method. An explicit understanding of these variables is critical for precise control of MoS
2 morphology with large coverage during their growth. Such desired knowledge could further allow for the synthesis of other TMDs that consists of both vertical and horizontally grown layer structures [
22].
Ambient pressure chemical vapor deposition (AP-CVD) is a facile, efficient, scalable method to grow large-scale monolayer MoS
2 aiming for the fabrication of integrated devices [
23,
24]. Conversely, CVD is also a flexible, cost-effective, and scalable process for growth optimization from flat horizontal to vertically oriented structured MoS
2 for energy storage applications. Despite atomically thin MoS
2 have been successfully grown on SiO
2/Si, also by our research team, [
24] there are still some limitations in extending this process to direct synthesis of large-area monolayers on different substrates such as pre-patterned and metal or conductive substrates. In this respect, computational studies, based on ab-initio methods, such as density functional theory, provided useful insights in predicting and clarifying the growth mechanism involved in CVD growth of TMDs and similar systems [
25,
26,
27]. Having the application target in mind, precise tuning of the growth orientations of horizontally and vertically aligned MoS
2 is critically important to benefit from their tailored materials properties and device functionalities in various fields. Furthermore, by modifying the configurational design inside the CVD reactor, the direct growth of large-scale MoS
2 on any arbitrary conductive substrate would open up an easy route for bottom contacts, thus facilitating MoS
2 integration in devices.
In this work, we investigated the growth behavior of mono to a few layers of MoS2 from molybdenum trioxide (MoO3) and sulfur (S) solid powders as a precursor in the AP-CVD process. We explicitly demonstrated the local changes of the S to MoO3 precursor positions in the growth zone inside the CVD reactor, which play a key factor in the changing of MoS2 nanosheets orientation. We successfully synthesized high-quality MoS2 flat monolayers and vertically aligned bulk MoS2 up to cm2 scale with good uniformity on different substrates such as SiO2/Si, pre-patterned SiO2 and TaN. In addition, we grew isolated single domains to continuous MoS2 conformally on the pre-patterned surface without any ruptures. The growth formation, crystallinity, extension of monolayer and vertically grown MoS2 layers were characterized by a series of techniques such as Raman spectroscopy, X-ray photoelectron spectroscopy (XPS), scanning electron microscopy (SEM) and photoluminescence (PL). We conclude with an outlook on the prospective scientific future of device developments based on MoS2 grown by optimized CVD methods.
3. Results and Discussion
Horizontally and vertically aligned MoS
2 were synthesized using a two-zone-furnace CVD apparatus on different substrates such as SiO
2/Si, pre-patterned and conductive substrates. Initial experiments were developed on flat SiO
2/Si substrates later extended the optimized growth approach to other substrates in terms of flat few layers MoS
2.
Figure 2a–f represents schematics and SEM images of horizontally and vertically aligned MoS
2 on the different substrates. The as-grown MoS
2 morphology on flat SiO
2/Si in the large region is shown by the SEM image as having a lateral size of more than 200 µm of continuous film, as shown in
Figure 2a. Here we observed that the triangular MoS
2 domains merge to yield a large-scale uniform single layer. The MoS
2 domains are large and tend to connect to each other despite the fact that their orientation on the surface is not controlled. Thus, the amount of grain boundaries and possibly other defects is reduced with respect to the case of relatively small domains randomly oriented in the space. As further evidence, we show large-scale lateral growth of monolayer MoS
2 up to a centimeter scale in
Figures S2 and S3 in Supplementary Materials. Using the same experimental conditions in CVD, we obtained a large-area continuous monolayer MoS
2 on a patterned SiO
2/Si substrate (
Figure 2b). The cross-sectional view of the monolayer MoS
2 is also characterized with transmission electron microscopy (TEM), where we observe that the MoS
2 layer conformally follows the trenches of the pre-patterned substrate as also detailed in the inset TEM cross-sectional image in
Figure 2b (see also
Figure S4 in Supplementary Materials). By optimizing the CVD growth conditions, we also synthesized highly crystalline few-layers MoS
2 on TaN. One of the critical parameters during the direct growth of MoS
2 on the TaN substrate is the growth temperature because of the lower TaN evaporation temperature with respect to the MoS
2 growth temperature. Although it is well documented that the chemical reaction from Ta(N) and S to form TaS
2 happens at around 900 °C at ambient pressure [
30], we could still note a minimal presence of TaS
2 phonon modes in Raman measurements, related to the imbalance between the formation of MoS
2 and TaS
2 in an atmosphere supersaturated in S, as in our case. Here, we controlled the MoO
3 temperature by lowering it to 650 °C to preserve TaN film.
Figure 2c shows the SEM image of the edge part taken at the interface of MoS
2 on TaN with exposed silicon on the bottom. The SEM image reveals the domain structure of large-area, horizontally grown monolayer MoS
2 without any vertically-aligned layers, as shown in
Figure 2a–c.
The vertically aligned MoS
2 growth was obtained by controlling the insertion of sulfur precursor during the temperature ramping-up stage because of substantial differences in the growth kinetics. In CVD growth, we control the sulfur flow that reaches the MoO
3 source by changing the position of the boat containing sulfur with respect to the heat zone furnace in such a way that the kinetics in the sulfur transport acts as the governing factor for the vertical growth orientation. We speculate that during the growth process, intensive compression between the highly dense bulk MoS
2 domains leads to the collision with other MoS
2 islands, thus causing vertical growth. In the vertical growth regime, the sulfur reaction with MoO
3 yields a concentration gradient normal to the substrate at the substrate surface. This event could have a role in promoting the out-of-plane vertical growth of MoS
2 due to the Mullins-Sekerka mechanism and significantly reducing the sulfur flow downstream during the growth [
22,
29].
Figure 2d–f show SEM images of the vertically aligned MoS
2 nanosheets grown on different substrates. A uniformly covered MoS
2 grown on flat SiO
2/Si substrate is clearly shown in
Figure 2d. Furthermore, we used the same growth conditions to yield vertical MoS
2 nanosheets on patterned substrates.
Figure 2e clearly shows the presence of triangular vertical MoS
2 domains after the growth. Bulk MoS
2 domains appear very bright compared to the few-layers MoS
2 domains, an indication of a large density of exposed domain edges. In addition, the cross-sectional SEM image in the inset of
Figure 2e demonstrates that the as-grown MoS
2 triangular domains are nearly perpendicular to the patterned substrate surface (see also the cross-section SEM image in
Figure S5 in Supplementary Materials). Such findings possibly claim a larger density of defects in the grown MoS
2 with respect to the flat case. In a simplified picture, in the vertical growth, MoS
2 domains are formed at a random orientation each other, possibly promoting a high number of defects at their edges, with the final result that a consistent density of defects would be formed. On the other hand, during the CVD growth on TaN, the critical drawback is to control the TaN evaporation temperature and the transition metal source temperature (i.e., molybdenum). Therefore, we precisely investigated different temperatures to preserve the underlying TaN film. Facilitating the same growth approach with the slightly reduced temperature down to 650 °C, we successfully synthesized MoS
2 on TaN, as shown in
Figure 2f, where a top view of vertical MoS
2/ TaN/SiO
2 interface is imaged by SEM. Surprisingly, we observed flakes appearing as bright or dark in the SEM image. This fact may be correlated with the competing formation of TaS
2 flakes along with MoS
2 as a collateral reaction between S and the TaN substrate according to the observed TaS
2-related feature in the Raman spectrum in Figure 4d.
To gain additional characterization to our MoS
2 growths, micro-Raman spectroscopy, photoluminescence (PL), and atomic force microscopy (AFM) (
Figure S6 in Supplementary Materials) were also employed to probe the structure, optical response and thickness uniformity of flat horizontal MoS
2 monolayers as shown in
Figure 3a–d.
Figure 3a shows the Raman spectra from MoS
2 nanosheets grown on SiO
2/Si flat substrate recorded at different positions. The measurements give E
12g (in-plane) and A
1g (out of plane) phonon modes located at 385.5 and 405.2 cm
−1, respectively. The wavelength difference of 19.7 cm
−1 confirms the growth of a MoS
2 monolayer, consistent with values reported in the literature [
31].
Figure 3b shows the Raman measurements on the MoS
2 grown on the pre-patterned substrate. The measurement gives the same E
12g and A
1g phonon modes with a frequency difference of ~20 cm
−1, which validates the growth of a monolayer MoS
2. As for TaN, we acquired the Raman spectrum of the substrate before and after the MoS
2 growth, as shown in
Figure 3c. The Raman spectrum from bare TaN (red line in
Figure 3c) confirms the crystallinity of TaN with a primary first-order acoustic mode (A) centred near 200 cm
−1 [
32,
33]. The occurrence of slight variations in the spectra region from 115 cm
−1 to 230 cm
−1 in TaN is typically interpreted as stoichiometry modifications, either excess presence of Ta or excess N [
32,
33]. After MoS
2 growth (black line in
Figure 3c), Raman spectrum evidence the presence of TaN acoustic modes at low Raman shifts together with the presence of MoS
2 E
12g and A
1g main phonon modes located at 381.7 cm
−1 and 404.3 cm
−1 with a wavelength difference of 22.6 cm
−1 corresponding to a thickness of 3 layered-MoS
2 nanosheets.
Figure 3d represents the PL spectra obtained on as-grown MoS
2 on flat SiO
2/Si (red), patterned SiO
2/Si (green) and TaN (black) substrate. The PL spectrum shows a high intense PL response peaked around the optical bandgap of 1.831 eV on monolayer MoS
2 grown on flat SiO
2/Si, which accounts for the direct gap transition. At room temperature, the pristine monolayer MoS
2 shows a high-quality strong PL peak associated with the band-to-band optical transition at the K point, as reported in previous studies [
31,
34]. The PL spectrum of MoS
2 on the patterned substrate is located at 1.843 eV with a slightly low intensity compared to the one on flat SiO
2/Si; however, these changes are possibly due to the local strain and minimal thickness variations at the trenches of the acquired region. Furthermore, we obtained the PL spectrum of as-grown three-layered MoS
2 on TaN recording an optical bandgap of 1.839 eV, still measurable though appearing as a low-intensity peak when plotted together with the PL response from the other two substrates. The intensity reduction is possibly related to the different number of layers, being the PL peak intensity highly enhanced in the monolayer limit. When reduced at the single layer, the band gap in MoS
2 shifts from indirect to direct [
34]. The direct optical transitions happen between the conduction band minimum and the two-valence band maxima at the K point of the Brillouin zone, which splits due to the spin–orbit coupling. Here the acquired experimental bandgap values show strong emission claiming for a good quality monolayer MoS
2, with reduced defect density, both on flat and pre-patterned SiO
2/Si substrates, with peak values for A exciton (1.831 eV and 1.843 eV, respectively) well in agreement with values reported in literature [
31].
With a similar analytical approach as used for the flat MoS
2 nanosheets, we considered micro-Raman spectroscopy to analyse the vertically aligned MoS
2 growths.
Figure 4a–d shows the Raman spectra of as-grown vertically aligned MoS
2 flakes on flat SiO
2/Si, pre-patterned SiO
2/Si and TaN substrate, respectively. As shown in
Figure 4a, the Raman spectrum obtained from MoS
2 grown on pre-patterned SiO
2/Si substrate (red line) clearly show the main two MoS
2 phonon modes E
12g and A
1g located at 382.9 cm
−1 and 409.1 cm
−1, respectively. As a further assessment, a dedicated MoS
2 growth on flat SiO
2/Si substrate targeting similar MoS
2 thickness as in the vertical growth case was performed, and its Raman spectrum is shown in
Figure 4a (black line) for comparison. The frequency difference of the E
12g and A
1g Raman modes gives 26 cm
−1, as evidenced in
Figure 4b, where the phonon mode region corresponding to the dashed box drawn in
Figure 4a is plotted. Such value is compatible with the presence of more than 6 MoS
2 layers and an overall thickness of >4.2 nm, considering the thickness of a MoS
2 monolayer equal to 0.7 nm, as known from the literature [
35,
36]. Raman spectra of MoS
2 on TaN shows MoS
2 phonon modes at 381.6 cm
−1 for in-plane E
12g and 409.8 cm
−1 for out-of-plane A
1g with a frequency difference of 28.2 cm
−1, which corresponds to an eight layer-thick MoS
2 on average. In addition, a broad range (100–250 cm
−1) of Raman peaks (highlighted by the dashed box) is evident, confirming the presence of preserved TaN after MoS
2 growth, as shown in
Figure 4c.
Figure 4d shows the enlarged Raman region limited by the dashed square in
Figure 4c to evidence the presence of a weak Raman peak around 280 cm
−1 assigned to the E
2g in-plane vibrational mode of 2H-TaS
2, which is an indication of the reaction between the interface of TaN and excess sulfur during the CVD growth process. The evidence of 2H-TaS
2 is unexpected based on thermochemistry considerations because the chemical reaction from Ta and S to form TaS
2 happens at around 900 °C at ambient pressure. Additional factors, such as a supersaturated sulfur atmosphere, out-of-equilibrium conditions, chemistry kinetics or catalytic effects, should be further considered to fully understand the formation of TaS
2. As noted above, the weak peaks on the broad region at low Raman shifts are assigned to TaN when stoichiometry modifications occur, either excess of Ta or excess N [
32,
33].
XPS analysis was performed to investigate the elemental composition and chemical bonding of the as-grown MoS
2.
Figure 5a represents the XPS spectrum of the Mo(3d) and S(2s) spectral region for the CVD grown MoS
2 monolayers. The spectral region contains the Mo 3d core-level line with Mo 3d 5/2 at 230.6 eV and Mo 3d 3/2 at 233.7 eV peaks, and the S 2s core-level line peaked at 227.8 eV, all pointing out to Mo-S bonding. The Mo peak positions are indicative of a MoS
2 arranged in a majority trigonal prismatic 2H-phase, in agreement with Raman spectroscopy. These results are consistent with previous works on peak positions for MoS
2 crystals. This observation constitutes the spectroscopic proof of the presence of Mo and S elements in a MoS
2 compound.
Figure 5b shows the XPS spectra recorded on vertical and flat aligned MoS
2 growth (in blue and magenta, respectively) on TaN substrate together with the bare TaN substrate (red). The spectral windows correspond to Ta-4d, Mo-3d, and S-2s core-level lines. The XPS investigation reveals the presence of Mo, S and TaN in the MoS
2 on TaN grown case. The TaN bare substrate (red line) XPS spectrum shows peaks at binding energies around 230.4 eV and 242.5 eV, which are distinctive of the 4d 5/2 and 4d 3/2 states of partially oxidized TaN, such as TaNO
x [
37]. Such analysis underlies that the pristine TaN substrate is partially oxidized at the surface, possibly due to exposure to the environment. This could be the evidence for the persistence of the TaN layer after the MoS
2 growth. XPS of MoS
2 on TaN shows prominent peaks at binding energies of around 229.4 eV and 232.6 eV, which are assigned to the doublet of Mo 3d 5/2 and Mo 3d 3/2, respectively. In addition, the sulfur peak (S-2s), located at 226.7 eV, is seen in both flat and vertical MoS
2. Interestingly, in the case of flat MoS
2 on TaN we do not observe any TaN peak (magenta line) around binding energy 242.5 eV of TaN(O
x), indicating that MoS
2 nanosheets are fully covering the TaN film. However, in the vertically-grown MoS
2 case (blue line), we clearly found an intense peak of TaNO
x at a binding energy of 242.7 eV (4d 3/2) and at 230.6 eV (4d 5/2), where an increase in the valley minimum between Mo related peaks is visible. As reasonable, in the case of vertically aligned MoS
2, a partial exposure of TaN substrate persists through the MoS
2 domains due to the different growth modes, which compromises the full surface coverage. By comparing the TaN (red) and MoS
2 on TaN (blue and magenta), it is also confirmed that TaN keeps preserved during the MoS
2 CVD growth process conditions. Thus, the XPS analysis provides further experimental support to show the formation of flat and vertically aligned MoS
2 on top of TaN substrate.