Precursor Ratio-Driven Morphological Evolution of CVD-Grown MoS₂ Microstructures

Abstract

The morphology of CVD-grown molybdenum disulfide (MoS₂) is sensitive to the local precursor environment, which governs nucleation density, edge stability, and growth kinetics. In this work, we systematically investigate the effect of precursor ratio on the morphological evolution of MoS₂ microstructures synthesized by atmospheric-pressure chemical vapor deposition on SiO₂/Si substrates. By varying the relative amounts of MoO₃ and sulfur precursors, distinct growth regimes were obtained, ranging from compact hexagonal and quasi-circular domains to multilayer hexagonal structures, triangular domains, and dendritic morphologies. At higher MoO₃ loading, growth is dominated by dense nucleation, leading to isolated few-layer hexagonal domains. A moderate reduction in Mo precursor concentration promotes diffusion-assisted growth and secondary nucleation, resulting in multilayer hexagonal structures with aligned or slightly rotated stacked layers. Under sulfur-rich conditions, morphology evolves into triangular domains due to anisotropic edge stabilization, while further increase in sulfur concentration gives rise to branched dendritic structures through kinetically limited, diffusion-dominated growth. Cross-sectional FIB analysis reveals dense vertical stacking and lateral displacement of upper layers in the multilayer hexagonal domains. Raman and photoluminescence measurements confirm strong correlations between morphology, layer thickness, and optical response, with thinner regions exhibiting reduced Raman peak separation, enhanced photoluminescence intensity, and blue-shifted excitonic transitions. The results show that, under fixed APCVD reactor geometry, source positions, temperature profile, carrier-gas flow, and nominal growth duration, the nominal MoO₃ to sulfur source-loading ratio reproducibly correlates with the transition between compact hexagonal, multilayer hexagonal, triangular, and dendritic MoS₂ morphologies on amorphous SiO₂/Si substrates. Because vapor-phase Mo- and S-containing partial pressures were not directly measured, this ratio is treated as a nominal source-inventory descriptor rather than as a direct vapor-phase stoichiometric ratio. The observed trends are interpreted using a qualitative thermodynamic and kinetic framework based on precursor fluence, nucleation density, edge stability, and diffusion-limited growth.

1. Introduction

The controlled synthesis of two-dimensional transition metal dichalcogenides (TMDs), particularly molybdenum disulfide (MoS₂), has attracted significant attention due to their unique structural, electronic, and optoelectronic properties [1,2]. Among the various synthesis techniques, chemical vapor deposition (CVD) has emerged as a versatile and scalable method for the growth of MoS₂ with tunable morphology and layer thickness [3,4]. A critical parameter governing the CVD growth process is the precursor concentration, which directly influences the local chemical environment, supersaturation, and growth kinetics at the substrate surface. Singh et al. demonstrated that the local precursor concentration boundary layer plays a decisive role in controlling MoS₂ layer number and domain quality in CVD growth, highlighting that even small variations in effective precursor flux at the substrate surface produce significant changes in the resulting microstructure [5].

Manipulation of precursor concentrations enables control over the morphology and structural characteristics of MoS₂, thereby determining its physical and chemical properties [6,7]. Variations in the effective Mo-to-S precursor ratio can lead to the formation of diverse microstructures, including triangular [6,8], hexagonal [6,7], truncated triangular [9], and dendritic domains [10]. These morphological transitions are closely associated with changes in vapor-phase composition and the resulting anisotropic edge growth dynamics [11,12]. Mouloua et al. further showed that distinct MoS₂ morphologies are obtained as a direct function of substrate position within the CVD tube furnace, confirming that local precursor concentration gradients are a primary determinant of the final growth morphology [13].

Previous studies have demonstrated that the domain shape of MoS₂ is highly sensitive to the Mo-to-S ratio. Wang et al. reported that variations in precursor ratio (Mo to S > 2, ≈1:2, and <2) significantly alter the edge growth kinetics, leading to systematic transformations in crystal morphology [6]. Similarly, Zhu et al. showed that tuning the effective precursor ratio allows for control over surface morphology, crystalline quality, and photoluminescence properties of MoS₂ [7]. In sulfur-rich environments, large-area monolayer MoS₂ with well-defined triangular domains and sharp edges is typically favored due to enhanced lateral growth and anisotropic S-edge stabilization [14,15]. Van der Zande et al. further demonstrated that grain boundaries and domain morphology in CVD-grown MoS₂ are intimately linked to local growth kinetics and precursor availability during nucleation [4].

Furthermore, Özden et al. [9] highlighted the importance of vapor confinement and transport in determining the local Mo and S concentrations near the substrate, emphasizing that the effective Mo-to-S vapor ratio in the growth zone plays a decisive role in controlling domain size, shape, and uniformity. Under controlled sulfur conditions, increasing Mo concentration has been shown to induce a sequence of morphological transitions, where MoS₂ domains evolve from triangular to hexagonal, and then become truncated triangular, and eventually revert to triangular shapes due to changes in edge stability and growth kinetics [12]. Complementary theoretical insight was provided by Cao et al., who demonstrated that the equilibrium shape and edge configuration of monolayer MoS₂ are strongly controlled by the chemical potentials of Mo and S species [11]. Using density functional theory and Wulff construction analysis, they predicted a morphology evolution from dodecagonal or mixed-edge structures under Mo-rich conditions to hexagonal domains at intermediate chemical potential, and finally to triangular domains under S-rich conditions [11,14]. Moreover, previous experimental studies have shown that the morphology of MoS₂ domains is highly sensitive to the relative Mo and S supply [14]. Xu et al. reported that triangular, near-hexagonal, and truncated-triangular MoS₂ flakes can be synthesized on SiO₂/Si substrates by CVD and proposed that the final domain shape is determined by the relative growth rates of Mo-terminated and S-terminated edges under different sulfur environments [8]. Under strongly sulfur-rich and high-flux conditions, the growth front becomes susceptible to diffusion-limited instabilities, where preferential attachment of adatoms at protruding edge sites promotes branching and the formation of dendritic morphologies rather than compact polygonal domains [10,16].

Despite these advances, a comprehensive understanding of how precursor concentration governs the full spectrum of MoS₂ morphological evolution—from compact hexagonal and multilayer stacked domains through triangular to highly branched dendritic structures—as a continuous function of the Mo to S ratio on amorphous SiO₂/Si substrates remains incomplete. Existing studies have typically examined one or two morphological regimes in isolation, without establishing a mechanistic framework that connects all transitions within a single experimental series under otherwise fixed growth conditions [5,6,9,13]. Furthermore, the cross-sectional microstructure of multilayer hexagonal domains, including the lateral displacement mechanism of successive stacked layers, has not been directly characterized by FIB analysis and correlated with growth kinetics. Similarly, the spatial evolution of Raman and photoluminescence signatures across structurally complex dendritic domains, and its interpretation in terms of thickness-dependent electron–phonon coupling and trion/exciton balance, has not previously been reported for this morphological regime. In this work, we systematically investigate the morphology of CVD-grown MoS₂ microstructures across five distinct nominal precursor-loading conditions and correlate the observed structural and optical properties with the resulting growth regimes. The term “nominal precursor-loading ratio” is used throughout to describe the initially loaded MoO₃ and sulfur source inventory under a fixed APCVD reactor geometry, source placement, temperature program, carrier-gas flow, and growth duration. It is not used as a direct measure of vapor-phase Mo:S partial-pressure ratio at the substrate. The morphology sequence is therefore interpreted as a reactor-specific correlation between nominal source inventory, precursor fluence, nucleation density, edge stability, and diffusion-limited growth, rather than as a universal morphology map based only on loaded precursor masses.

2. Materials and Methods

2.1. Growth Method

For the growth of MoS₂ thin films, vapor-phase reactions of MoO₃ and sulfur (S) precursors were employed. The process was performed in a 1.2 m long clear fused quartz tube (inner diameter 4.6 cm, outer diameter 5 cm, (Raesch Quarz GmbH, Ilmenau, Germany). Silicon substrates (1 × 1 cm) with 90 nm SiO₂ were initially cleaned sequentially in acetone, isopropanol, and deionized water for 10 min each, and then subsequently dried under nitrogen gas. The cleaned substrates were then positioned at the center of the furnace within the quartz tube.

A ceramic boat containing MoO₃ powder (99%, Sigma-Aldrich Chemie GmbH, Taufkirchen, Germany), was placed at the center of the quartz tube, 40 cm from the inlet. To systematically investigate the effect of precursor ratio on MoS₂ morphology, the MoO₃ mass varied between 10 mg and 500 mg. The sulfur source (99%, Sigma-Aldrich Chemie GmbH, Taufkirchen, Germany), with mass varied between 200 mg and 400 mg, was positioned 27 cm upstream from the inlet in a lower-temperature zone maintained at 240 °C throughout the growth process. The resulting nominal atomic Mo:S source-loading ratios are summarized in Table S1. These ratios were calculated from the initially loaded MoO₃ and sulfur masses and describe the source inventory before the growth process. They are not equivalent to directly measured vapor-phase Mo:S partial-pressure ratios in the reaction zone. To clarify the limiting vapor-pressure and source-depletion scales associated with these source inventories, equilibrium vapor-pressure ceilings and lower-bound depletion times for MoO₃ and sulfur are estimated in Supplementary Note S1 and Table S2.

Prior to growth, the furnace was purged with 500 sccm N₂ for one hour. The furnace temperature was then ramped to 850 °C at a linear rate of 28 °C min⁻¹ under a continuous N₂ flow of 1000 sccm. Upon reaching 850 °C, the peak temperature was maintained for growth durations ranging from 10 to 30 min. However, all morphological characterization results reported in this work were obtained at a fixed growth duration of 10 min, with the MoO₃-to-sulfur precursor mass ratio as the sole independently varied parameter across the five growth conditions investigated. Time-series SEM observations at fixed precursor conditions (MoO₃ = 10 mg; S = 300 mg) for growth durations of 10–30 min, presented in Figure S1, confirm that the fundamental morphological character is established at 10 min and does not change with increasing growth duration, directly validating the decoupling of growth time and precursor ratio effects. All temperatures were monitored using calibrated thermocouples. Following the growth period, the system was cooled naturally to ambient temperature under a continued N₂ flow of 500 sccm at atmospheric pressure. The complete CVD temperature profile and N₂ flow rate evolution throughout the process are shown in Figure S2 of the supporting information.

2.2. Characterization of Synthesized Vertical MoS₂ Nanosheet

The surface morphology and topographical details of the different MoS₂ morphologies were carried out using scanning electron microscopy (SEM; RAITH 150, Raith GmbH, Dortmund, Germany) and atomic force microscopy (AFM; Dimension V, Veeco Instruments GmbH, Aschheim, Germany). All cross-sectional and structural analyses were performed using a FIB Nanoanalytics Auriga 60 dual-beam system (Carl Zeiss AG, Oberkochen, Germany) to analyze the inner morphology of distinct structures. Before cutting, protective carbon and platinum layers were deposited sequentially via an electron beam and a gallium ion beam, respectively. Furthermore, we investigated the material properties with Raman spectroscopy and photoluminescence (PL). The Raman and PL spectra were obtained using a WiTec Alpha 300 confocal Raman microscope (WiTec GmbH, Ulm, Germany). X-ray diffraction (XRD) measurements were performed on a Siemens D5000 diffractometer (Siemens AG, Munich, Germany) using a copper anode and radiation with a wavelength of 0.15406 nm (Cu Kα1) in Bragg–Brentano mode.

3. Results and Discussion

3.1. Precursor Concentration-Dependent Morphological Evolution and Grain Growth in CVD-Grown MoS₂

Figure 1 shows the SEM images revealing a systematic evolution of MoS₂ morphology with varying MoO₃ and sulfur precursor concentrations. As shown in Figure 1a,b, at higher MoO₃ loadings (500 mg and 100 mg with 200 mg S), the surface is populated by compact hexagonal and quasi-circular multilayer domains. Moreover, as shown in Figure 1b, the hexagonal multilayer MoS₂ structures exhibit two distinct stacking morphologies. In some cases, the additional layers grow directly above the base layer while preserving the lateral orientation and crystallographic alignment, indicating a relatively ordered stacking process. In other cases, the upper layers are slightly rotated with respect to the base layer, producing a star-like morphology. However, such multilayer stacking is not evident in Figure 1a, where the higher MoO₃ concentration (500 mg) mainly results in scattered few-layer hexagonal and quasi-circular MoS₂ domains with Raman peak separations consistent with mono-to-few-layer thickness (Figure S3). The observed morphological evolution can be understood through the combined influence of local supersaturation, chemical potential, and edge energetics [17]. However, because the vapor-phase concentrations at the substrate were not measured directly, the following interpretation should be regarded as qualitative. The nominal precursor-loading ratio is used here as an experimentally controlled descriptor of the initial source inventory under fixed reactor conditions, rather than as a direct measure of reaction-zone partial pressures. To estimate the limiting source-zone pressure and depletion scales, equilibrium vapor-pressure ceilings were calculated from tabulated vapor-pressure data in Supplementary Note S1.

At the measured sulfur-source temperature of 235–240 °C, the total sulfur vapor-pressure ceiling is approximately 10.8–12.7 mbar, while at the MoO₃ source temperature of 850 °C, the MoO₃ vapor-pressure ceiling is approximately 25.9 mbar. A saturated-carrier-gas upper-bound model at 1000 sccm N₂ gives lower-bound depletion times of approximately 1.4–3.2 min for 200–400 mg sulfur and 0.06–3.0 min for 10–500 mg MoO₃. These are limiting estimates, not measured depletion times, because they assume complete equilibration of the carrier-gas stream with the source and neglect finite evaporation kinetics, powder-bed geometry, boundary-layer transport, condensation/re-evaporation, and reaction losses. Therefore, the observed morphology sequence is discussed in terms of nominal source inventory and possible precursor fluence under the fixed APCVD configuration, not in terms of directly measured vapor-phase stoichiometry. The nominal precursor loading ratio is used here as an experimentally controlled descriptor of the initial source inventory under fixed growth conditions, rather than as a direct measure of reaction-zone partial pressures.

$$S_{\mathit{eff}}\sim {\displaystyle \frac{p_{\mathit{Mo}}\cdot p_S^2}{K_g\left(T\right)}},$$

which governs the overall thermodynamic driving force for growth. Here, effective supersaturation is treated as a phenomenological parameter that depends on the local activities of Mo- and S-containing vapor species. These activities are governed not only by precursor loading, but also by source temperature, source depletion, carrier-gas transport, condensation, and surface reactions. Therefore, the model is used to rationalize the observed morphology sequence, not to extract absolute gas-phase concentrations. At high MoO₃ loading, the effective supersaturation is expected to be elevated, resulting in an increased nucleation rate and, consequently, a high nucleation density. Under these conditions, precursor species are rapidly consumed by the growing nuclei, limiting both lateral growth and vertical stacking. In addition, the effective diffusion length of adatoms is reduced due to frequent capture by nearby nuclei, suppressing mass transport toward existing domains. Consequently, growth remains nucleation-dominated, resulting in isolated few-layer hexagonal or quasi-circular structures. When the nominal MoO₃ loading is reduced, the observed morphology suggests a decrease in nucleation density and a larger relative contribution from surface diffusion, enabling adatoms to migrate toward and attach to existing MoS₂ domains. In this regime, the lateral edge growth rate may be described phenomenologically as

$$v_{edge}\propto \left(S_{eff}-1\right)exp\left(-{\displaystyle \frac{E_{edge}}{kT}}\right)$$

where $v_{edge}$ is the lateral edge growth velocity, $S_{eff}$ is the effective supersaturation defined above, $E_{edge}$ is the activation energy for adatom incorporation at the MoS₂ edge, $k$ is the Boltzmann constant, and $T$ is the absolute growth temperature. This expression remains sufficient for domain expansion while also facilitating secondary nucleation on top of stable base layers [18].

Beyond nucleation and diffusion kinetics, the compact polygonal shape of MoS₂ domains is influenced by the relative stability of different edge terminations, including Mo-terminated zigzag, S-terminated zigzag, sulfur-passivated zigzag, and armchair-type edges. Density functional-theory studies show that these edge stabilities are strongly dependent on the sulfur chemical potential and on edge reconstruction. For example, Cao et al. reported that Wulff-type equilibrium constructions predict a morphology evolution from mixed-edge or near-hexagonal shapes to triangular domains as the Mo/S chemical environment is varied [11]. More recent edge-reconstruction calculations by Li and Ding showed that the relative formation energies of reconstructed MoS₂ zigzag and armchair edges can differ by approximately 0.1–0.4 eV Å⁻¹ depending on sulfur chemical potential; in particular, the energy difference between stable Mo-zigzag and S-zigzag reconstructions can exceed 0.4 eV Å⁻¹ under sulfur-rich chemical potentials and remains on the order of ~0.2 eV Å⁻¹ near intermediate sulfur chemical potentials [19]. These literature values provide a numerical scale showing that moderate changes in chemical environment can substantially alter which edge terminations are retained during growth.

In the present APCVD experiments, however, the reaction-zone Mo- and S-containing partial pressures were not measured directly. Therefore, we do not use Wulff construction to calculate sample-specific equilibrium shapes or absolute edge free energies. Instead, the Wulff-type argument is used only as a qualitative thermodynamic reference: lower-energy edge configurations are expected to be preferentially retained under near-equilibrium growth, whereas higher-energy edges tend to shrink or disappear. The experimentally observed transition from compact hexagonal or quasi-circular domains to triangular domains is therefore interpreted as being consistent with a change in the relative stability and growth velocity of MoS₂ edge terminations under different nominal source-loading conditions, but not as a quantitative determination of edge energies in our reactor. Minimization of ${\gamma }_{total}$ overall facet orientations predict the equilibrium crystal morphology for a given set of chemical potentials. Under near-stoichiometric (balanced) chemical potential conditions, the energies of Mo-terminated and S-terminated edges become comparable, promoting symmetric growth and resulting in hexagonal domains [11]. The resulting multilayer stacking behavior is governed by local growth kinetics: near-equilibrium incorporation leads to aligned multilayers, whereas kinetically limited attachment results in slight rotational misalignment and the formation of star-like structures [19,20]. Therefore, the transition from scattered single-layer domains at high precursor loading to well-defined multilayer hexagonal structures at reduced precursor concentration can be consistently explained by a shift from nucleation-dominated growth to surface diffusion-assisted lateral and vertical growth by secondary nucleation under near-equilibrium conditions.

Figure 1c,d illustrates the morphological transitions of the MoS₂ nuclei at reduced MoO₃ concentrations and increased sulfur supply. These conditions result in a clear morphological transition, evolving from well-defined triangular domains to dendritic structures. As shown in Figure 1c, d, decreasing the MoO₃ concentration to 20 mg and 10 mg (with S = 300 mg) promotes the formation of large triangular MoS₂ domains, indicating a shift toward a sulfur-rich growth environment. Under these conditions, the relative chemical potential of sulfur increases, altering the stability of different edge terminations. Specifically, S-terminated zigzag edges become energetically more favorable than Mo-terminated edges, leading to anisotropic growth and the formation of triangular domains.

From a thermodynamic perspective, this behavior can be understood as a deviation from near-stoichiometric conditions toward a sulfur-rich regime, where the balance between Mo and S chemical potentials favor asymmetric edge stabilization. As a result, the equilibrium crystal shape evolves from symmetric hexagons to anisotropic triangles, consistent with edge-energy minimization principles described by Wulff construction [21,22]. With a further increase in sulfur concentration to 400 mg (Figure 1e), the MoS₂ growth morphology transforms into dendritic or branched structures. In this regime, the growth process becomes increasingly dominated by kinetic limitations rather than thermodynamics. The high sulfur supply enhances the local growth, while the reduced availability of Mo and surface diffusion restricts uniform lateral growth. Consequently, adatoms preferentially attach to protruding regions of the crystal edges, where the local flux is higher. This leads to the amplification of edge instabilities and promotes branching, resulting in dendritic morphologies. Additionally, the dendritic structures are frequently observed to grow on top of an existing monolayer or few-layer MoS₂ base, indicating that once a continuous film is established, excess precursor species contribute to secondary growth under diffusion-limited conditions. To further elucidate the influence of precursor concentration on the growth behavior, the average grain size distribution was analyzed, as shown in Figure 1f. A significant variation in grain size, ranging from approximately 5 µm to 40 µm, is observed depending on the precursor conditions. Under higher nucleation density conditions, associated with increased supersaturation, smaller grain sizes dominate due to the simultaneous formation of multiple nuclei and limited lateral growth. In contrast, reduced nucleation density enables enhanced adatom diffusion and sustained lateral growth of existing domains, resulting in larger grain sizes. Moreover, the grain size distributions obtained under sulfur-rich conditions (MoO₃ = 20–10 mg and S = 300–400 mg) exhibit a polymodal character, including a reduced population of compact domains in the intermediate 20–25 µm size range (Figure 1f). The distribution can be rationalized by the coexistence of domains that nucleated at different times and therefore experienced different local growth histories during the same CVD run. Domains that nucleated early under locally high supersaturation were surrounded by a higher density of neighboring nuclei; their lateral expansion was therefore limited by precursor competition and reduced adatom diffusion length, leading predominantly to smaller domains in the 5–15 µm range. In contrast, domains that nucleated in locally less crowded regions or at later stages after partial precursor depletion had reduced the nucleation rate, experienced weaker inter-domain competition, and could continue lateral growth to larger sizes of approximately 30–40 µm. Under these conditions, relatively few domains are expected to remain in the intermediate 20–25 µm range because the growth environment favors either early nucleation with restricted lateral expansion or sustained growth of more isolated domains. The reduced 20–25 µm population reflects the spatial and temporal heterogeneity of nucleation, precursor depletion, and inter-domain competition in the APCVD process. The same argument also resolves the apparent coexistence of small and large domains on a single substrate: the substrate does not experience a perfectly uniform growth environment; instead, it contains local variations in precursor flux, nucleation density, and available diffusion area. Furthermore, under strongly sulfur-rich conditions, perturbations on the growth front experience enhanced local diffusion flux, amplifying edge instabilities rather than smoothing them. This promotes anisotropic dendritic branching rather than continued isotropic expansion [23,24], redirecting growth toward branched morphologies and further suppressing the population of intermediate-sized compact domains. To confirm that the observed morphological transitions are not accompanied by any structural phase transformation, X-ray diffraction (XRD) measurements were performed on samples representing the three principal growth regimes: hexagonal (MoO₃ = 100 mg; S = 200 mg), triangular (MoO₃ = 10 mg; S = 300 mg), and dendritic (MoO₃ = 10 mg; S = 400 mg), as shown in Figure S4. All three diffraction patterns are fully indexed to the 2H-MoS₂ phase [25,26]. The dominant (002) reflection at 2θ ≈ 14° is present in all samples with comparable relative intensity, confirming the layered hexagonal stacking sequence of the 2H polymorph across all precursor conditions. Higher-order basal plane reflections at (004), (006), and (008) are consistently resolved, indicating a well-ordered layered structure in each morphological regime [27,28]. However, the XRD measurements presented here serve solely to verify phase identity and exclude competing crystalline phases; no compositional or defect-density information is inferred from them. Critically, no diffraction peaks attributable to the metallic 1T or distorted 1T′ phases are detected in any sample, confirming the 2H polymorph across all growth conditions. It is acknowledged that bulk XRD cannot detect amorphous suboxide (MoO_3−X) or oxysulfide phases, so their local presence within the defective dendritic networks cannot be fully excluded by the present measurements. Nevertheless, several lines of evidence indicate that such phases do not primarily drive the dendritic morphology. Li et al., combining experiments with kinetic Monte Carlo simulations on amorphous SiO₂/Si substrates, showed that dendritic MoS₂ growth is governed by the Mo adatom concentration gradient under high S/Mo conditions, and that the dendritic character persists even when MoO residues are explicitly included as growth units—i.e., it is set by the Mo concentration gradient rather than by residue chemistry [29]. In agreement with this, spatially resolved structural studies of dendritic MoS₂ grown under comparable sulfur-rich conditions consistently report phase purity at the level of individual branches: atomic-resolution ADF-STEM of dendrites on SiO₂/Si revealed the same hexagonal MoS₂ lattice on both basal planes and branch edges, with Mo/S zigzag terminations [10], while multi-location SAED and HRTEM on individual branches confirmed single-phase MoS₂ of uniform crystallographic orientation, with MoO_3−X suboxides acting only as transient intermediates that fully convert to MoS₂ during growth [30]. Furthermore, according to the Mo-to-S phase diagram presented in Wu et al., MoS₃ is the dominant vapor-phase monomer under strong sulfur excess, implying fully sulfur growth species prior to attachment [31]. Since the dendritic morphology in the present work appears at the highest nominal sulfur loading (Mo:S ≈ 1:180), it is attributed to diffusion-limited kinetic growth instabilities under extreme sulfur excess rather than to oxysulfide-driven compositional effects, because the growth conditions are the dominating driver of the morphological transitions, and surface residues are just supporting factors [10,30,32].

3.2. Formation and Characterization of MoS₂ Hexagonal and Dendritic Microdomains

The multilayered hexagonal MoS₂ structures were investigated with SEM, AFM, and FIB to elucidate their morphological characteristics. The detailed morphology of the multilayer hexagonal structures is presented in Figure 2. These structures exhibit a distinct vertical stacking configuration, where additional layers are formed on top of an underlying MoS₂ base layer. In Figure 2a, a tilted SEM image provides clear visualization of this stacked architecture. The inset SEM image highlights that successive layers exhibit a shift with respect to the underlying layer while maintaining their crystallographic orientation. A three-dimensional representation of these structures is shown in Figure 2b, offering a comprehensive view of the multilayer arrangement. To gain further insight into the vertical growth mechanism, cross-sectional analysis was performed using FIB milling. Prior to sectioning, protective carbon and platinum layers were sequentially deposited using an electron beam to prevent surface damage. Subsequently, precise cross-sections were prepared using a gallium ion beam. The cross-sectional FIB images shown in Figure 2c,d reveal the internal morphology of the multilayer hexagonal MoS₂ structures grown on the SiO₂/Si substrate. The base layer appears as a relatively continuous film with a thickness of approximately 20–35 nm, while the subsequently stacked layers exhibit thicknesses in the range of 100–200 nm. A notable feature is the lateral displacement of the upper layers with respect to the base layer, with a sliding distance of ~40–100 nm. This lateral offset can be understood in the context of the growth mechanism discussed earlier. Under surface diffusion limited growth conditions, adatoms arriving on the surface preferentially attach on energetically favorable sites, such as step edges or existing terraces [33,34,35]. However, due to local variations in adatom flux, diffusion length, and attachment kinetics, the nucleation of upper layers does not necessarily occur exactly above the center of the underlying domain. Instead, slight asymmetries in mass transport to the edges lead to off-centered nucleation, which manifests as a lateral shift of the upper layers relative to the base layer. This behavior is further facilitated by the weak van der Waals interaction between layers, which allows lateral relaxation to minimize local strain and surface energy without disrupting the overall crystallographic orientation [19,20,33]. Furthermore, the FIB cross-sections do not show any voids or porous features between the stacked layers, indicating that the structures are densely packed. This suggests that the growth proceeds via continuous material incorporation rather than incomplete coalescence, consistent with a sufficiently high local precursor supply during vertical growth. However, in many cases, the upper layers exhibit a reduced lateral size compared to the base layer, forming progressively smaller hexagonal terraces. This morphological evolution can be attributed to a gradual decrease in effective adatom supply at higher layers due to limited vertical mass transport from the gas phase and shadowing effects, as well as Ehrlich–Schwoebel growth instabilities [23,24].

In contrast to the multilayer hexagonal structures, the dendritic-shaped MoS₂ domains exhibit highly anisotropic and branched morphologies, as shown in Figure 3. These structures are predominantly observed on top of an underlying monolayer or few-layer MoS₂ film, where the monolayer regions appear as darker orange contrast, while the substrate is lighter. The dendritic-like structures (highlighted in green) are observed to form on top of the underlying monolayer MoS₂, indicating that the pre-existing layer acts as a preferential nucleation template. The early stage of growth is marked by a well-defined triangular nucleus (outlined by red dotted lines), which reflects the anisotropic edge energetics of MoS₂ under sulfur-rich conditions. This triangular core serves as a seed, from which growth proceeds outward. As growth continues, kinetic effects dominate, promoting preferential extension along specific directions and resulting in the characteristic branched dendritic morphology [10,34]. This suggests that growth initially follows a thermodynamically favored triangular configuration, consistent with anisotropic edge stability. However, as growth proceeds, the morphology deviates from equilibrium shapes due to changes in local growth conditions causing morphological instabilities [24]. The transition from compact triangular domains to dendritic structures can be attributed to a shift from thermodynamically controlled growth to kinetically dominated growth [10,34,35]. Under relatively balanced growth conditions, domain evolution is governed by minimization of edge free energy, resulting in well-defined triangular flakes. In contrast, under sulfur-rich and high precursor flux conditions, the attachment rate of adatoms exceeds their surface diffusion capability. As a result, atoms are incorporated at the first available sites rather than redistributing along energetically favorable edges. This leads to diffusion-limited growth, where material accumulates preferentially at protrusions or tips of the structure due to diffusion instabilities [23,36]. These regions experience enhanced local flux (a classical tip-effect instability), promoting accelerated growth along specific directions and giving rise to branched, fractal-like dendritic morphologies. The observed arms extending from the triangular core therefore reflect instability-driven growth rather than equilibrium shape evolution. Furthermore, the underlying SiO₂/Si substrate plays a crucial role in this process. Unlike crystalline substrates, the amorphous nature of SiO₂ provides a spatially non-uniform energy landscape for adatom diffusion. This lack of directional constraint allows growth to proceed along locally favorable diffusion pathways, further enhancing branching and morphological irregularity [10,34]. Nevertheless, the dendritic morphology arises from diffusion-limited kinetic growth instabilities governed by the Mo concentration gradient under extreme sulfur excess. Oxysulfide or suboxide phases are therefore not required for dendrite formation [10,29,32].

Furthermore, the internal structural continuity of the dendritic arms was investigated by performing cross-sectional FIB analysis along a dendritic structure: the arm apex, the mid-arm region, and the core/base region. Figure 4a shows the SEM image of the dendritic structure with the three FIB cut positions indicated. At all three positions (Figure 4b–d), the MoS₂ film appears as a continuous, dense layer with no internal voids, gaps, or porosity, and the protective C/Pt capping layer is conformally deposited directly on top of the MoS₂ surface in every cross-section, confirming the absence of subsurface discontinuities. A clear thickness gradient is observed along the arm: the apex measures approximately 9.5 nm (Figure 4b), the mid-arm region increases progressively from 12 nm to 30 nm (Figure 4c), and the core/base region reaches 27–38 nm, including a locally thicker multilayered patch (Figure 4d). This progressive thickening from tip to base is consistent with the proposed growth mechanism, in which the thinner arm tips extend rapidly via diffusion-limited, instability-driven growth, while the core and base regions accumulate additional material through continued secondary nucleation and vertical stacking over the full growth duration. For reference, optical and SEM images of the two distinct dendritic structures used for Raman/PL characterization and FIB cross-sectional analysis are provided in Figure S5, confirming that the measurements were performed on structurally distinct but nominally equivalent domains grown under identical conditions.

The insets in Figure 5f provide an optical image of the scanned region of the dendritic MoS₂ structure. Figure 5a,c, presents Raman intensity mapping images of the E¹_2g and A_1g peaks obtained using a 488 nm laser line at room temperature. The positions where the Raman spectra were acquired are labeled P1–P6 in Figure 5a. To further elucidate the layer thickness variations, peak frequency mappings of the E¹_2g and A_1g modes were extracted from the recorded Raman mappings and shown in Figure 5b,d. In these mappings, the brightness or darkness of the colors represents the shift in the peak frequencies. In the E¹_2g peak position Raman mapping (Figure 5b), the intensity exhibits variations corresponding to different thicknesses or layers of MoS₂. The uniform color contrast in each region indicates distinct thicknesses, with brighter colors indicating thinner MoS₂ layers. This observation is consistent with the understanding that as MoS₂ thickness decreases, the E¹_2g peak experiences a blue shift, which is represented by the brighter colors in the mapping. Similarly, Figure 5d shows the Raman peak frequency mapping of the A_1g mode, where darker colors signify a redshift of the A_1g peak. In contrast to the E¹_2g peak mapping, the A_1g peak frequency mapping shows that darker colors correspond to thinner areas, while lighter colors correspond to thicker layers of MoS₂. This relationship reflects the behavior of the A_1g peak, which experiences a red shift as the thickness of MoS₂ increases. Additionally, Figure 5e reveals that the wavenumber difference between the characteristic peaks is increasing for the bulk layers, indicated by brighter colors, compared to the thinner layers, represented by darker shades. This trend suggests that the spectral separation between the peaks decreases as the thickness of the MoS₂ layers decreases. This occurs if the measurement position is moved toward the tip of the dendritic MoS₂ structure. Indeed, based on the Raman peak separation values (Δω ≈ 22.4 cm⁻¹ at P5–P6 vs. ~25.1 cm⁻¹ at P1), the dendritic structures exhibit a pronounced thickness gradient: the outermost base periphery at positions P5–P6 corresponds to a few-layer MoS₂ region, while the central core and arm interior (P1–P4) represent a thick 3D multilayer crystal, consistent with the cross-sectional FIB-measured thicknesses of 9.5–38 nm.

Furthermore, Figure 5f presents a statistical distribution of these differences observed from various points across the sample. This distribution presents the consistency and variability of the wavenumber differences between characteristic peaks dependent on the layer thickness throughout the sample. Figure 5g shows the spectral analysis obtained from six different points on the dendritic structure. Within the spectrum, the E¹₂g peak appears between 378 cm⁻¹ and 380 cm⁻¹, while the A₁g peak is between 400 cm⁻¹ and 403 cm⁻¹, representing the fundamental vibrational modes of MoS₂ [37,38]. In addition to these characteristic modes, the analysis reveals other signals associated with MoS₂. Lower-frequency combination modes including E¹₂g-LA at ~148 cm⁻¹ and A₁g-LA at ~180 cm⁻¹ are also observed [38,39,40,41]. The double longitudinal acoustic phonon mode (2LA) is observed at ~453 cm⁻¹. Additional higher-frequency combination modes are also present, including (E¹₂g + LA) at ~598 cm⁻¹, (E¹₂g + A₁g) at ~780 cm⁻¹, and (2A₁g) at ~820 cm⁻¹ [38,39,40,41]. These shear modes (SMs) involve the in-plane motion of metal and chalcogen atoms, while the layer breathing mode (LBM) entails the out-of-plane motion of metal and chalcogen atoms. Notably, while the SMs and LBMs are absent in single layers of MoS₂, they exhibit characteristic blue and red shifts, respectively, as the layer number increases [39,40,42].

Figure 5h illustrates the dependency of the deviation of characteristic peaks of MoS₂ on the layer numbers. Analysis reveals that as the layer number decreases, the E¹_2g and A_1g peaks exhibit slight shifts in their frequencies: the E¹_2g peak undergoes a blue shift, while the A_1g peak experiences redshift. The fitted E¹_2g, A_1g, and 2LA(M) peaks of MoS₂ and their respective peak position values respective to each position are shown in Figure 5i. In Figure 5j, the variation in these characteristic peaks across six different positions is depicted along with their wavenumber difference. Notably, from positions P1 to P4, the E¹_2g and A_1g peak positions show minimal change. However, at positions P5 and P6, the E¹₂g peak experiences a slight stiffening (~2 cm⁻¹, from 378.2 to 378.4 cm⁻¹), while the A₁g mode notably softens (~2.4 cm⁻¹, from 403.3 to 400.9 cm⁻¹). The stiffening and softening of the peaks decrease the wavenumber difference from Δω ~25.1 to ~22.4 cm⁻¹, indicating a change in the characteristic nature of the MoS₂ thickness from multilayered to few-layered formation [38,43]. The frequency variation in the E¹_2g and A_1g modes with layer thickness can be explained using the van der Waals force model [44,45,46]. With an increase in layer number, the inter-layer van der Waals force in MoS₂ strengthens, suppressing atom vibrations and resulting in higher force constants [46]. Consequently, the E¹_2g peak undergoes a redshift due to the shorter intra-layer S–Mo–S distance in thinner MoS₂ layers. Furthermore, the anomalous redshift of E¹_2g with increasing layer thickness can be attributed to structural changes or an increase in long-range Coulombic interlayer interactions, particularly involving Mo atoms, as the E¹_2g mode involves in-plane Mo atom vibrations. In contrast, the A_1g mode is less affected by these Coulomb potential screening effects, as it primarily involves out-of-plane S atom vibrations [43,46,47]. The FWHM widths of these Raman peaks, depicted in Figure 5k, further support these observations. The FWHM of the E¹₂g peak varies between 8.1 and 9.2 cm⁻¹ across all positions, with no clear monotonic trend, while the A₁g FWHM ranges from 9.4 cm⁻¹ at P1 to 10.9 cm⁻¹ at P2, before decreasing to 9.4–9.8 cm⁻¹ at P5–P6, consistent with broader linewidths in the thicker multilayer interior and narrower linewidths in the thinner base regions.

To decouple the contributions of strain, charge doping, and electron–phonon coupling to the observed anti-correlated mode shifts, the I(A₁g)/I(E¹₂g) integrated intensity ratio was extracted for all positions, P1–P6 and plotted alongside Δω in Figure S6. The ratio shows a general decreasing trend from 1.37 at P1 to 1.17 at P6, consistent with a progressive reduction in electron–phonon coupling strength toward the thinner base regions, where reduced interlayer van der Waals coupling weakens dielectric screening and lowers the local carrier density available for A₁g phonon renormalization [38,43,46]. All ratio values remain above unity across all positions, which is consistent with relatively low sulfur vacancy concentrations at the probed positions—significant vacancy-induced excess electron doping would selectively suppress A₁g intensity and drive the ratio below unity [38,43]. However, this indirect proxy cannot substitute for direct quantitative compositional analysis, and local non-stoichiometry—particularly at the edges and branching points of the dendritic structure where kinetically driven adatom incorporation may be less selective—cannot be excluded based on optical methods alone. The anti-correlated shifts at P4–P5, where Δω decreases from ~23.8 to ~22.4 cm⁻¹ while I(A₁g)/I(E¹₂g) simultaneously decreases, are therefore self-consistently explained: both trends reflect the same thickness-driven reduction in interlayer coupling and carrier density at the dendritic periphery [38,43,46]. The concurrent blue shift of the E¹₂g mode at P5–P6 relative to P1–P4 is attributed to reduced interlayer Coulomb interactions and weakened van der Waals coupling in the thinner peripheral regions. It is noted that the absolute positions of both modes are downshifted relative to pristine monolayer reference values, consistent with the presence of tensile strain or structural defects contributing to phonon softening across the entire dendritic structure [46,47,48,49].

The room-temperature photoluminescence (PL) response of the dendritic MoS₂ structures (Figure 6) provides further insight into their thickness-dependent optical properties and local electronic environment. As shown in Figure 6a, the PL spectra acquired from positions P1–P6 exhibit two prominent features corresponding to the A and B excitonic transitions. These peaks originate from direct bandgap transitions at the K-point, where the splitting of the valence band due to spin–orbit coupling gives rise to the A (lower energy) and B (higher energy) excitons [50,51].

As shown in Figure 6, a clear spatial variation in PL intensity is observed from the mapping, where thinner regions exhibit significantly higher emission intensity, consistent with a direct bandgap transition, while thicker (bulk-like) regions appear darker due to the indirect bandgap transition characteristic of multilayer MoS₂. The abrupt blue shift in PL emission observed at positions P5 and P6 relative to P1–P4 directly reflects the thinner layer character of the dendritic base regions, where the direct bandgap transition dominates, as confirmed by the significantly enhanced PL intensity at these positions (Figure 6a,b). In thicker multilayer regions (P1–P4), stronger interlayer dielectric screening maintains a higher effective carrier density, shifting the spectral weight toward negatively charged trion (A⁻) emission, which appears ~30–40 meV below the neutral A-exciton (X⁰) energy and produces an apparent red shift in the observed PL peak position (Figure 6c) [52,53]. At the thinner base regions (P5–P6), reduced interlayer screening increases the exciton binding energy and lowers the local carrier density, visibly shifting the spectral weight from trion back toward neutral exciton emission and producing the observed blue shift [52,53,54]. This is further supported by the narrowing of both A and B exciton FWHM values at P5–P6 (Figure 6d), indicating relatively reduced disorder and narrower linewidths in the thinner peripheral regions compared to the thicker multilayer interior [7,10,35]. The sharp increase in B-exciton energy to ~1.987 eV at P5, combined with the A–B energy separation of ~0.13–0.158 eV, remains within the established range for MoS₂ and confirms that the spectral changes arise from thickness-driven trion/exciton balance rather than structural degradation [39,52,53].

Moreover, the relatively broader FWHM linewidths in the thicker multilayer regions (P1–P4) further reflect increased disorder, lattice distortion, and enhanced exciton–phonon scattering due to higher defect density, consistent with inhomogeneous broadening in CVD-grown multilayer MoS₂ [7,10,35].

4. Conclusions

In this work, the morphology of CVD-grown MoS₂ microstructures was investigated as a function of nominal MoO₃ and sulfur source loading under otherwise fixed APCVD conditions. Since reaction-zone partial pressures were not directly measured, the reported Mo ratios should be understood as nominal precursor loading ratios rather than absolute vapor-phase ratios. At relatively high MoO₃ loading, growth was dominated by a high nucleation density, resulting in compact hexagonal and quasi-circular single- or few-layer domains. A slight reduction in Mo precursor concentration promoted diffusion-assisted growth and secondary nucleation on pre-existing domains, leading to multilayer hexagonal structures with aligned or slightly rotated stacked layers. Cross-sectional FIB analysis revealed dense vertical stacking without void formation, while lateral offsets between successive layers indicated off-centered nucleation and interlayer relaxation during growth. Further reduction in MoO₃ concentration combined with increased sulfur supply shifted the system toward sulfur-rich conditions, producing large triangular domains through anisotropic edge stabilization. At the highest nominal sulfur loading, morphology transitioned into branched dendritic structures. This trend is consistent with a shift from compact edge-stabilized growth toward kinetically limited, diffusion-influenced aggregation, although quantitative extraction of supersaturation, precursor flux, or edge kinetic parameters were not performed in this study. Raman and photoluminescence analyses confirmed a strong correlation between morphology, thickness, and optical properties. The layer-dependent variation in the E¹_2g and A_1g Raman modes, together with the evolution of their peak separation, revealed the coexistence of multilayer and few-layer regions within the dendritic structures. Raman and photoluminescence analyses confirmed a strong correlation between morphology, layer thickness, and optical properties. The dendritic structures are 3D multilayer crystals exhibiting a pronounced thickness gradient from the thick multilayer core and arm interior—consistent with the FIB-measured cross-sectional thicknesses of 9.5–38 nm—to thinner few-layer regions exclusively at the outermost base periphery. The layer-dependent variation in the E¹₂g and A₁g Raman modes and the evolution of their peak separation (Δω decreasing from ~25 cm⁻¹ at the multilayer interior to ~22 cm⁻¹ at the peripheral positions P5–P6) confirm this thickness gradient quantitatively. Photoluminescence measurements further showed enhanced emission intensity and blue-shifted excitonic transitions at the thin peripheral positions P5–P6, consistent with reduced interlayer coupling and a thickness-driven shift in trion/exciton spectral balance toward neutral exciton emission, while the thicker interior regions exhibit reduced PL intensity and red-shifted excitonic transitions consistent with stronger interlayer dielectric screening and increased non-radiative recombination.