The demonstrated relationship between MoS2 and the surface morphology of the substrate reveals that the regular performance may be affected by the presence of metal oxide dielectrics.16 To realize better surface morphology, we synthesized MoS2 films on both SiO2 and Si3N4 surfaces. Optical microscopy was employed initially to observe as-synthesized and converted MoS2 films (shown in Fig. 2a). Figure 2(b) illustrates the AFM images of annealed (converted) MoS2 films fabricated on both SiO2 (upper panel) and Si3N4 (lower panel) surfaces at annealing temperatures ranging from 600°C-1000°C. At different annealing temperatures, SiO2 surfaces show the appearance of numerous grains (shown in Fig. 2c) with a height between 2 and 8.5 nm. Intriguingly, these grains are nearly absent on Si3N4 surfaces up to an annealing temperature of 800°C as illustrated in Fig. 2(d). This indicates superior uniformity and crystallinity in MoS2 films compared to SiO2 surfaces. During the annealing process, Si or oxygen may migrate from SiO2 to the MoS2 interface with lower sulfur content, leading to grain formation. This migration process will be discussed in a later section focusing on the interface analysis of MoS2 films grown on both surfaces.
At annealing temperatures exceeding 800°C (SiO2) and 900°C (Si3N4), we observe larger grains with significant boundaries on both surfaces. We suggest that these grains may correspond to either Mo or MoOx, depending on the specific surface. The AFM and X-ray reflectivity (XRR) techniques were employed to estimate the roughness and thickness of the MoS2 films, respectively. High-quality MoS2 film formation typically begins above 600°C during annealing,30 thus we skipped AFM analysis for the film formed at 500°C. Figure 2(e) and (f) illustrate the roughness (Ra) and thickness as a function of annealing temperatures. The Ra values of the MoS2 films on Si3N4 surfaces are found to be below 0.5 nm with a thickness of ~ 4 nm (between 600°C and 800°C), indicating a smoother surface compared to films formed on SiO2 surfaces using a similar technique, as reported previously.31 As the annealing temperatures rise from 800°C, the Ra values for both surfaces show an increase. This can be attributed to less sulfur content in MoS2 films at high annealing temperatures, leaving only metal or metal oxide on the surface. The film thickness on SiO2 surfaces varies as the height of remaining structures fluctuates with increasing annealing temperatures. The thickness, however, depends on the spin coating and concentration of the precursor solution.
Due to the improved surface morphology observed in our synthesized MoS2 films on Si3N4 compared to the SiO2 surface, we intend to concentrate on analyzing Si3N4 surfaces for the subsequent stages of our analysis. Raman spectroscopy was next employed to examine the converted MoS2 films. Figure 3 displays Raman intensity (a.u) plotted against Raman shift (cm − 1) at different annealing temperatures. A mechanically exfoliated MoS2 crystal is used as a reference to assess the distinctive Raman peaks of films produced at annealing temperatures ranging from 500°C to 1000°C. Two characteristic Raman active modes emerge: the out-of-plane vibration of S atoms (A1g), exhibiting a Raman shift around ∼406 cm− 1, and the in-plane vibration involving Mo and S atoms (E2g) at approximately ∼382 cm− 1 during annealing temperatures ranging from 700°C to 800°C. This closely corresponds to our reference MoS2 crystal(A1g = 407 cm− 1 and E2g = 383 cm− 1) and aligns with the previous report.32,33 The characteristics A1g and E2g peaks with a difference of 24 cm− 1, suggests the formation of semiconducting 2H- multilayer MoS2 films34 within the similar temperature range. This matches well with our estimated thickness as determined by XRR.
On the other hand, the emergence of A1g and E2g peaks at lower annealing temperatures (< 700°C) shows a minimal presence or reduced intensity, suggesting the absence of MoS2 film formation. Additionally, we also assume the absence of MoS2 film on both surfaces at higher annealing temperatures, starting from 900°C.
X-ray diffraction (XRD) was employed to analyze the crystalline phase and orientation of MoS2 films. In Fig. 4(a), XRD patterns are depicted for MoS2 crystals and MoS2 films on the Si3N4 surface, after annealing from 500°C to 1000°C. The exfoliated MoS2 crystal displays diffraction peaks at 2θ𝝌/𝞅 angles of 32.7°, 47.5°, and 58.4°, corresponding to (1 1 0 0), (2 2 0), and (1 2 1 0) orientations. Converted MoS2 films show similar peaks at temperatures between 700°C to 800°C, maintaining identical orientations. The (1 1 0 0) and (1 2 1 0) orientations are attributed to MoS2 films with a hexagonal lattice structure, while (2 2 0) corresponds to Si, consistent with previous XRD results.35 At temperatures above 800°C, additional patterns appear, along with an extra peak at 40.3° (1 1 0), indicating MoS2 or Mo and a combination of these (Supplementary Figure S1), attributed to large grains confirmed by AFM topology (Fig. 2b). However, MoS2 films on SiO2 surfaces show a similar orientation (Fig. 4b), with an expanded peak around 20.4°, likely assigned to the SiO2 surface during annealing.36,37
In addition, we determined the Full Width at Half Maximum (FWHM) based on the area of XRD patterns of MoS2(1 1 0 0), MoS2(1 2 1 0), and Si(2 2 0) orientations taking MoS2 crystals as a reference and the plots are illustrated in Fig. 4(c). The MoS2 films on Si3N4 show a decreasing FWHM with higher annealing temperatures, similar trends to those observed in the fabrication of ALD-assisted MoO3 film.38 This suggests an improvement in the crystal quality of the films. Notably, this improvement remains consistent for Mo and Si respectively. We then used Fusion360 to illustrate the crystal structure and lattice spacings (Fig. 4d) of our MoS2 film based on the crystallographic orientations obtained from the XRD patterns. In Fig. 4(e), a top view of the MoS2 crystal shows lattice spacings of 0.27nm and 0.16nm, corresponding well to the MoS2(1 1 0 0) and MoS2(1 2 1 0) planes. These results are in good agreement with a previous study by Heeseung Yang et. al.31
The above results and discussion highlight that MoS2 exhibits superior surface morphology in terms of uniformity, roughness, and crystallinity on Si3N4 compared to SiO2, as confirmed through AFM, Raman spectroscopy, and XRD analysis. These improvements are attributed to specific annealing temperatures in the range of 700°C to 800°C.
The annealed MoS2 thin film was analyzed by X-ray photoelectron spectroscopy (XPS) to accurately determine its stoichiometry, detect any residual precursor elements, and evaluate the oxidation state induced by either moisture or the annealing process. Figure 5 shows the core level XPS analysis of the Mo 3d, S 2p, and O 1s for both surfaces. Gaussian-Lorentzian fitting was used to identify individual components. On the SiO2 surface, Mo 3d peaks appear at 227.6 and 230.8 eV, corresponding to Mo 3d5/2 (cyan-blue line) and Mo 3d3/2 (green line) doublets, respectively, as shown in Fig. 5a. In addition, a peak at 225.5 eV attributed to sulfur atoms, denoted as 2s (orange line), is evident. In contrast, XPS analysis of films on the Si3N4 surface shows Mo 3d peaks at 228.1 and 231.4 eV, with the S peak at 225.9 eV. We attribute these binding energy peaks to the Mo-S bond in MoS2 films.39 The discrepancy in binding energies between Si3N4 and SiO2 surfaces may be due to surface oxidation of MoS2 or migration of Si or oxygen from SiO2 into the film. In order to accurately fit the Mo3d5/2 and Mo3d3/2 peaks, additional sets of low-intensity doublets were required to achieve a satisfactory fit. A peak corresponding to MoO2 (blue line) at a binding energy of 228.3 eV for Mo3d5/2, together with other peaks overlapping with Mo3d3/2 at 232.1 and 233.3 eV, representing MoO2 (dark cyan line) and MoO3 (wine line) respectively. These oxides can be attributed to oxygen migration during annealing or to the uneven surface of SiO2. Similarly, in addition to MoO2 (Mo 3d3/2), metal oxides at the binding energies of 229.2 eV and 232.8 eV are observed on the Si3N4 surface (Fig. 5d), possibly due to exposure to air or water, although the surface is initially free of oxide. These binding energies are in close agreement with those reported in the literature.40,39
XPS analysis of MoS2 films on Si3N4 reveals an oxidation-free surface compared to SiO2, and our focus then shifts to the analysis of trace elements on the Si3N4 surface (Fig. 5e). The observed binding energies at 161.7 and 162.8 eV, in agreement with previous report,41 correspond to the S 2p3/2 and S 2p1/2 peaks of sulfur. A weaker peak at 163.9 eV indicates subsurface sulfur rather than residual surface contamination.42,43 Similar sulfur peaks are observed in MoS2 films on SiO2 (Fig. 5b). Further investigation of the O1s core level is shown in Figs. 5c and 5f for both surfaces. The detection of oxygen as bound Mo-O at 529.35 eV suggests its possible origin from MoO3,39 which is clearly distinct from Si-O at 531.4 eV. Interestingly, these oxide species are not present in the film grown on Si3N4.
The MoS2 films have an oxidation-free surface on Si3N4, as verified by XPS analysis. However, it is suggested that during the annealing process, MoS2 layers are initially formed on the SiO2 surface, resulting in the diffusion of oxygen from SiO2 to MoS2. This ultimately leads to the formation of metal oxides at the interface between MoS2 and SiO2. The migration model is then briefly discussed in the following sections.
Elemental migration to MoS2, caused by substrates or surface adsorbates, can affect electronic devices44 when used as a channel. However, the precise migration mechanism between 2D MoS2 and dielectric surfaces remains incompletely understood. Figure 6(a) illustrates the migration model from SiO2 to MoS2 during annealing above 600°C. We propose that elevated annealing temperatures cause sulfur loss, leading to the migration of silicon or oxygen from SiO2 to MoS2 films. Initially, the migration tends to accumulate at the edges of the MoS2, particularly in regions of lower sulfur content. This process leads to the formation of an S-O bond. While the 2D layered MoS2 surface provides a dangling bond-free and inert interface, sulfur vacancies are expected to be the dominant defects with energy formation in the range of 1.3-1.5eV.45,46 A study of MoS2 film formation by sputtering shows initial Mo adlayer formation at the MoS2/SiO2 interface, with subsequent annealing leading to Mo layer diffusion into SiO2, resulting in the formation of molybdenum silicon oxide (MoOxSiy) layers.47 A similar layer may potentially be formed in our MoS2 films on the SiO2 surface.
To confirm the migration, thick MoS2 films on a SiO2 surface were subjected to core-level XPS analysis. Figure 6(b) shows the atomic concentration variations with Ar sputtering time, derived from the peak areas of Mo 3d, S 2p, Si 2p, N 1s, O 1s, and C 1s. The concentration of Mo and S decreases with sputtering time, while Si and O show an opposite trend after 180s. A slight shift of the Si 2p peaks (shown in Fig. 6c) to higher binding energies (102.6eV to 102.8eV) at 190s, corresponding to the charge effect associated with either SiO2.48,49 At the same time, the Mo 3d peaks (see supplementary information Figure S2 ) shift to lower binding energy, causing the disappearance of the S peak, suggesting the effect of Ar sputtering on MoS2 films. This shift is attributed to either the charging effect or changes in the chemical states from the film surface to the interface.50 We also performed thermal desorption spectroscopy (TDS) on MoS2 films deposited on the SiO2 surface using a quadrupole mass spectrometer (QMS). The TDS spectra (Fig. 6d) for mass-to-charge ratios of m/z = 32 and 64 show a sharp peak at a desorption temperature below 400°C, attributed to the desorption of organic matter. Since QMS selectively filters substances based on mass-to-charge ratio, both S and O2 can be assigned to m/z = 32, which persists up to about 730°C. In addition to direct S desorption, we also observed the presence of the sulfur oxidation product, namely SO2, expressed as m/z = 64 in Fig. 6(d), at the same desorption temperature as m/z = 32. It should be emphasized that m/z = 64 is referred to as SO2 rather than S2 and is in agreement with previous TDS investigations.51
The release of sulfur in the form of SO2 suggests oxygen transfer from SiO2 to Mo-S bonds in low sulfur MoS2 films, rather than from any adsorbates. This implies that migration occurs predominantly in the region of lower Mo or S content, leading to Mo-O and S-O bond formation at higher annealing temperatures, consistent with our proposed migration model. Notably, the organic matter in the films originates from the precursor solution itself, as evidenced by the absence of monolayer MoS2 degradation after environmental exposure.52 L.M. Farigliano et al. have theoretically documented similar desorption studies.53 The formation of SO2 appears to be related to two discrete intermediates: OSOMo and Mo-OSO structures. The former, formed by oxygen migration into Mo-S, undergoes desorption of SO2 with an energy barrier of 1.49 eV. The latter is formed by oxygen insertion bound to adjacent Mo atoms and undergoes desorption of SO and SO2 with energy barriers of 0.41 eV and 0.78 eV, respectively.
Additional desorbed species are then examined as shown in Figure S3. No detectable sulfur (S) or sulfur-related species were identified in the TDS spectrum at both lower and higher desorption temperatures. A broadened shoulder is evident at mass-to-charge ratios (m/z) of 17 and 18 in the temperature range of 100 to 200°C, indicating the desorption of H2O, although a definitive judgment proves difficult. A sharp peak of higher intensity is seen at all m/z ratios between 300 and 350°C, indicating the desorption of organic matter. In addition, smaller peaks of lower intensity at 440°C for m/z 31 and above 550°C present challenges in making conclusive judgments.
To illustrate the migration process on the Si3N4 surface, we conducted a depth profile XPS analysis of MoS2 films, as depicted in Fig. 7. The concentration of oxygen remains constant at below 5% ( above 60% for SiO2 surface at a sputtering time of 360 s) throughout the sputtering time (Fig. 7a). Similar decreasing trends to the SiO2 surface are observed for Mo 3d and S 2p. However, the Si content increased after 210 s of sputtering, and the N content increased after 270 s. In Fig. 7(b), the XPS spectra show a complex evolution in the line shape of the Si 2p peaks with sputtering time. The binding energies ranging from 100.9 to 101.8 eV correspond to the Si-N bond of Si3N4 (depicted by the black line). Nevertheless, a shift from a higher binding energy (101.9 eV) to a lower one (100.3 eV) indicates a Si-rich Si3N4 surface (depicted by the red line) due to the charge effect,54 though this is not evident in the MoS2 interface. We propose that this shift may lead to the formation of Si-N and Si-Si bonds on the Si3N4 surface, resulting in the enrichment of Si on the Si3N4 surface denoted as SixNy(x > 3,y < 4) in Fig. 7(b).55 The Si-enriched surface of Si3N4 may result from the annealing process of MoS2 films or the unexpected adjustment in the gas ratio during the formation of Si3N4 using LPCVD.
The migration process of Si3N4 to the MoS2 film encounters hurdles, which are mainly attributed to the increased oxidation resistance observed at higher temperatures.56 According to XPS results on Si3N4 produced by LPCVD on Si, surface migration of Si3N4 is only possible to Si that has undergone an oxygen-assisted plasma treatment. This treatment tends to lead to partial substitution of N by oxygen, resulting in the formation of Si-O bonds.57 However, it is important to note the possible presence of oxygen adsorbed from the environment (Fig. 5d), which manifests as MoO2, an unstable oxide. The elimination of this oxide requires the maintenance of a high-quality vacuum or inert environment during the preparation and preservation of MoS2 films. Apart from this consideration, it is possible to achieve an oxide-free surface for MoS2 films on the Si3N4 surface within the annealing temperature range of 700 to 800°C. The existence of oxygen (O2) or moisture (H2O) on the surface or interface of 2D MoS2 can deteriorate the field effect mobility and increase hysteresis when utilized as a channel in FETs. Additionally, oxygen impurities originating from the dielectric surface during the annealing process in the formation of MoS2 films can create a complex layer, potentially disrupting the usual electrical performance.58,59 Hence, achieving oxide-free semiconducting films is feasible by selecting an oxide-free dielectric, such as silicon nitride (Si3N4), as demonstrated in our study.
This achievement is significant as nitride-based dielectrics generally exhibit superior interfaces compared to oxide (SiO2) dielectrics. Such improvement can significantly enhance the performance of gate dielectrics, allowing a reduction in the interfacial defect density near the substrate,60 making them suitable for applications on a large scale to miniaturized electronics.