Nucleation and growth mechanism for atomic layer deposition of Al₂O₃ on two-dimensional WS₂ monolayer

Due to their exceptional physical, electrical, and optical characteristics, two-dimensional (2D) transition metal dichalcogenides (TMDs) have generated a lot of attention in the scientific community.¹ Alternative designs for optoelectronic devices are being studied for upcoming technological nodes as Moore’s law approaches its scaling limits.^1,2 A new field of study in nanotechnology for 2D materials was established in 2004 by Geim and Novoselov’s discovery of a graphene monolayer.³ Transition metal dichalcogenides (TMDs), which are 2D semiconductors, have shown to be a great substitute for silicon-based transistors in the sub-5 nm range^2,4 owing to their wide range of conductivity measurements. The short-channel effect,⁵ which is caused by source-drain tunneling and the loss of gate electrostatic control over the channel, is what prevents silicon-based devices from being further advanced. Whenever the gate length is shortened, a serious on/off current deterioration occurs. Due to their superior semiconductor features, such as high mobility, a wider bandgap, and a lower in-plane dielectric constant, TMD-based transistors perform better than silicon-based transistors.^6–8 However, the thoroughly passivated surface of 2D materials provides the opportunity to create transistors with unparalleled levels of defect-free at the nanoscale range. A low subthreshold swing (70 mV/dec) and a high on/off current ratio (up to 10⁸) were two characteristics of the first MoS₂ transistor demonstration.⁹ Since TMD monolayer flakes are accessible by exfoliation processes, a wide range of TMDs is now being explored in semiconductors.^10,11 Researchers have also made incredible progress in establishing TMD manufacturing methods that are practical for commercial production, such as chemical vapor deposition (CVD). For effective gate control in cutting-edge technology, nanometer-thin high-k dielectric films must be deposited on an ultrathin TMD channel. However, there is currently little knowledge of the processes governing the formation and nucleation of high-k dielectric films during atomic layer deposition (ALD) on TMD materials made by CVD. The majority of reports^12,13 focus on the ALD process settings, the physical characteristics of the generated films, structure determination, and domain size alone. However, TMD CVD’s function has not yet been investigated. The majority of CVD-grown TMDs are polycrystalline, and the starting surface as well as the CVD growth conditions have a significant impact on the size, surface coverage, and in-plane orientation of the crystal grains.¹⁴ The subsequent ALD process of high-k dielectric films can be impacted by the TMD grain size, the existence of intragrain defects, surface coverage, and in-plane crystal grain orientation, respectively. Furthermore, selective decoration of the TMD surface structure by ALD can offer a thorough comprehension of the TMD layer structure and characteristics, which is necessary for adjusting the layers to specific applications. We, thus, examine atomic layer deposition (ALD) of Al₂O₃ on CVD-grown TMDs in this study. Self-terminating gas-solid chemical reactions are the foundation of ALD. The ALD deposition principle enables conformal deposition and nanoscale control of layer composition and thickness. ALD has been used successfully in several high-k dielectric applications, as per the reports.^15–19 For instance, ALD has been used to create Si-based dielectric films effectively on FETs with a thickness of less than 2 nm.^12,20,21 Fast layer closure is necessary to create nanometer thin, hole-free dielectric films; hence, the nucleation mechanisms have been addressed.^21,22 When the substrate is not yet entirely covered by the ALD-grown layer, the substrate impacts the growth per cycle (GPC, the quantity of material deposited each reaction cycle) and the growth mode that controls how the material is distributed on the substrate. When the film closes and the GPC (steady growth per cycle) becomes constant, the impact of the original substrate surface fades. Since HfCl₄ quickly interacts with the –OH surface groups, the development of HfO₂ layers using HfCl₄/H₂O ALD on a completely hydroxylated SiO₂ surface illustrates rapid film closure.¹² Since HfCl₄ is not highly reactive with the H groups of the Si surface, the same mechanism results in sluggish layer closure for an H-terminated Si surface. Another well-known instance of this is the development of Al₂O₃ on H-terminated Si using trimethylaluminum (TMA)/H₂O, which at first develops an island morphology rather than a continuous Al₂O₃ film.²² As a result, a lot of reactive surface sites are needed for the film to close quickly. TMD materials suspended bond-free surface is an advantage in electrical devices. However, the absence of reactive surfaces makes it difficult to deposit dielectrics via ALD.²³ It has been shown that high dielectric material nucleation on exfoliated TMDs occurs extremely slowly during the ALD process, with the layers not closing until they reach a significant thickness (e.g., >10 nm).^24–29 On the inert surface of TMDs, attempts have been made to resolve the nucleation issue for ALD. These methods, unfortunately, face the risk of oxidizing the TMDs.³⁰ ALD may enable physisorption-based nucleation by lowering the temperature.²⁸ Purity is the concern and loses when the ALD temperature is dropped. In addition to this issue, it is unknown how ALD on TMD materials works.³¹ Typically, chemisorption mechanisms that occur during the successive, self-limiting surface reactions of ALD precursors are used to explain how ALD grows.³² However, in addition to chemisorption, extra atomistic interactions, such as physisorption and surface diffusion of adsorbed precursor molecules (species), which are weakly bound, may also contribute to the formation of ALD on 2D materials. Additionally, the creation of nanoparticles from the aggregated species can then disperse and agglomerate.^33–38 In this research, we study the nucleation, development, evolution, and ALD of a polycrystalline monolayer of WS₂ produced via metal-organic CVD. Through complementary physical characterization methods that measure the total quantity of deposited material, its surface coverage, and the Al₂O₃ particle size distribution, the initial nucleation, growth mode, and layer closure processes during ALD are examined (PSD). Assuming selective nucleation at WS₂ grain boundaries, crystal edges, and surface defects, we evaluate the ALD growth process using this model. A lateral growth mechanism is then applied to close the thin films. It is suggested that the theory of NP diffusion and collision effects be used to explain the impact of ALD temperature. Additionally, we investigated the impacts of the early TMA pulses as a functionalization strategy to comprehend their initial nucleation mechanism and their relevance to the ALD process.

To investigate the nucleation and growth mechanisms of Al₂O₃ on a polycrystalline WS₂ monolayer, the ALD Al₂O₃/1.3 Ml WS₂/1.6 μm SiO₂/Si (100) samples were prepared. To stop the development of native oxides, the Si wafers were first cleaned using cyclic HF and then subjected to wet thermal oxidation.³³ Then, using W(CO)₆ and H₂S as precursors, a closed WS₂ monolayer was produced by MOCVD at 950 °C, although alternative material sources and surface choices were used and their relevant MOCVD process has been previously reported.^34,35 Al₂O₃ ALD was then created using a Polygon® 8200 platform and a PULSAR® 2000 reactor with an inert gas valve, and the ALD reactor is a 200 °C hot-wall cross-flow reactor type. The temperature was maintained at 18 °C for trimethylaluminum (TMA) and H₂O precursor bubbles. As a carrier gas, nitrogen (N₂) was employed.³⁵ The samples were shifted into a loading lock before Al₂O₃ ALD was formed to prevent background oxidants’ formation.³⁶ 

AFM is an effective method for characterizing surface morphology. WS₂ surface morphology, grain size distribution, and Al₂O₃ deposition by ALD were all studied using atomic force microscopy (AFM, ICON PT, Brooker, and UHV3500 SPM, RHK) in non-contact mode. By comparing the uncovered WS₂ area to the Al₂O₃ surface area, it can calculate the Al₂O₃ surface coverage and film deterioration as a function of ALD cycles. In this manner, a computer model was used to explore the nucleation and growth process of Al₂O₃ deposition on WS₂ surfaces. The Al₂O₃ surface on the substrate was also examined using a time-of-flight secondary ion mass spectrometer (TOF-SIMS). Based on the secondary ions’ depth of escape in a 2D growth mode, the deposited material covers the surface uniformly, and it also intensifies the substrate elements’ TOF-SIMS signal (s). A secondary ion’s escape depth typically ranges between one and two atomic layers.³⁷ I/I₀= exp (−t/λ), where t is the layer thickness and also the secondary ion’s escape depth, results in a sharp drop.³⁸ Since certain areas of the substrate are still exposed after a few cycles in non-2D development, the substrate signal deteriorates significantly and slowly.^39–41 As a result, the substrate signal decay is highly sensitive and may be utilized to track the end and the growth of a film effectively. To determine the vibrations of the WS₂ lattice and a plane in front of the Al₂O₃ ALD, Raman spectra at 523 nm were taken. The elemental abundance of W and S as well as the thickness of the WS₂ film are measured using Rutherford backscattering spectrometry (RBS) with a 1.52 MeV He⁺ ion beam.

For better interpretation, the information is classified into four sections. We describe the composition and morphology of the polycrystalline WS₂ single crystal layer before ALD in Sec. III A. We trace the growth progress of the Al₂O₃ ALD on the WS₂ single crystal layer up to 600 ALD cycles in Sec. III B. Using AFM, ERD, and TOF-SIMS to examine the surface coverage, morphology, and amount of material deposited (Al areal density), we assess the growth mode and explore the implications of the ALD process settings. We discuss the Al₂O₃ ALD growth mode on polycrystalline WS₂ in Sec. III C. In Sec. III D, we examine how WS₂ crystal domains change in orientation and 2D structure as they develop on single crystals of WS₂.

The chemical bonding, surface composition, hydrophilicity, crystal structure, and shape of the 2D material have an impact on the nucleation and growth behavior of ALD on the substrate. To comprehend the structure, crystallinity, and morphology of WS₂ on a SiO₂/Si substrate, we first measured and obtained AFM pictures, RBS data, and Raman spectra. According to Fig. 1(a), the layer is mostly made up of a closed polycrystalline single layer with a few triangular WS₂ crystal grains on top. The SiO₂ substrate is mostly parallel to the basal planes of the WS₂ crystals. According to AFM measurements made after 60 cycles, the closed monolayer’s grain size ranges between 10 and 200 nm with an average of 64 nm [Fig. 1(c)]. Triangular grains on the initially closed monolayer have a grain size distribution that ranges from 100 to 1100 nm, with an average of 460 nm [Fig. 1(d)]. A spiral structure with a limited number of layers is generated when a few triangular WS₂ crystals develop. The surface roughness (rms) of WS₂ is 0.4 nm, according to AFM studies.

The elemental amounts of W and S as well as the thickness of the WS₂ layer are determined using Rutherford backscattering spectrometry (RBS). W and S both have a primary content of 1.2 × 10¹⁵ and 2.6 × 10¹⁵± at/cm², respectively. The S/W ratio is 2:1, which is almost the stoichiometry predicted for WS₂. The connection that a monolayer of WS₂ consists of 1.0 × 10¹⁵ W at/cm² assuming the nominal density (7.02 g/cm³) and ML thickness (0.55 nm) of WS₂ allows the W content measured by RBS to be translated into monolayer content (ML). According to the RBS measurements, 1.3 ML of WS₂ was deposited. Inferring that the MOCVD deposition produces a closed monolayer with a few crystals at the top, we compute the percentage area of triangular crystals in the second layer based on the RBS result, and 30% of the second layer’s computed area is made of triangular crystals. The 1.3 ML WS₂ is thought to be composed of a closed WS₂ layer with some sizable triangular crystals in the second layer based on the strong agreement between the AFM and RBS measurements. In other words, as seen in Fig. 1(e), the basal plane of WS₂ crystals in the first (70%) and second (30%) monolayers, step edges, and grain boundaries characterize the WS₂ surface. According to Fig. 1(b), the in-plane (E_12g) and out-of-plane (A_1g) vibrational modes exhibit vibrations at 355.2 and 417.2 cm⁻¹ in the Raman spectrum, respectively.⁴² The overall details of the possible models conferred are as follows and their detailed information is included in the following section as well for better understanding. We tried to brief the details as Model 1, Model 2, and Model 3.

Model 1: 2D growth, the ideal ALD growth model, the self-limiting aspect of ALD leads to excellent step coverage and conformal deposition on high aspect ratio structures. Some surface areas will react before other surface areas because of different precursor gas fluxes.

Model 2: Nanoribbon growth, which is the 3D growth model based on the sample surface, in our cases, the major nucleation sites are on the grain boundaries, but all the grain boundaries are highly sensitive and can nuclei on all nucleation site on every cycle, but the noninteractive region will deposit more by the lateral growth.

Model 3: Hemi-cone growth— we found out that even if the grain boundaries are causing selective growth, which leads to the third model, we need to provide the nucleation density value and the growth rate by each cycle number, leading to a slower film closure due to the better quality of the 2D material film surfaces (WS₂).

In this section, we discussed Al₂O₃ ALD process development, evolution, and morphology on WS₂. To examine the nucleation and growth processes of Al₂O₃ on WS₂, we utilized a 1.3 ML polycrystalline Al₂O₃ sample and a 1.6 μm SiO₂/Si (100) substrate. After purification with cyclic HF, the influence of native oxide is avoided by thermal oxidation. Trimethylaluminum (TMA) and oxidant H₂O were employed as precursors, bubbling at 18 °C to produce WS₂ films at 950 and 150 °C through the MOCVD process and nitrogen (N₂) was used as a carrier gas. We pumped the chamber up to a 10⁻¹ Torr base pressure (P_base). The AFM images following Al₂O₃ ALD for various ALD cycles are shown in Figs. 2(a)–2(f). The Al content as estimated by ERD is coupled with the AFM data, Figs. 4(a) and 4(b), from nucleation through linear development, and the evolution of Al₂O₃ and ALD on WS₂ is investigated. The growth is impeded during the first 400 cycles of Al₂O₃ ALD on WS₂, which is a substantially longer period than Al₂O₃ ALD on H-terminated Si and synthetic MoS₂ produced via sulfurization.^28,38,43 The initial nucleation, the transition state, and the linear growth stage are the three phases that contribute to the general growth mechanism of ALD. WS₂ step edges, grain boundaries, and a few point defects are the only places where Al₂O₃ islands are produced during the first nucleation process, leaving the basal plane mostly uncovered. The nucleation density was calculated based on the surface morphology of the Al₂O₃/WS₂ surface and the number of ALD cycles of the AFM film, as shown in Fig. 4(e). The nucleation density increases nearly linearly throughout the first 200 ALD cycles, suggesting that a rising number of Al₂O₃ nuclei are formed during this period. The initial WS₂ monolayers-enhanced nucleation at the grain boundaries is the primary cause of the growing nucleation density. This increase is influenced by the flushing time after the water pulse, Fig. 4(c), and therefore, it could indicate a slow change in surface composition. The S-H groups at the grain boundaries may slowly oxidize during H₂O pulse and rinsing, which increases the TMA adsorption rate at the grain boundaries, allowing new nuclei to be formed during the first 200 ALD cycles.

Despite having a smaller length per area, the density of nuclei generated at the grain boundaries is noticeably higher than the density of nuclei generated at the crystal edges in the first cycle (roughly estimated from the intersection of the linear fit in Fig. 4(e). ∼2.1 × 10¹⁰/cm² nuclei in the first cycle (∼0.0078 × 10¹⁰/cm² nuclei per cycle) (there are a smaller number of larger triangular crystals in the second layer compared to a larger number of smaller crystals in the first layer). The AFM images indicate this distinction as well. Due to the closer spacing between the Al₂O₃ nuclei at the crystal borders, continuous nanoribbons form more quickly. The increased nucleation density is consistent with the surface morphology and suggests that there are more surface sites, such as S-H groups, at the crystal edges. There may be a large number of S-H groups on the crystal edges of the triangular WS₂ crystals in the partial second monolayer, which are easily accessible for ligand exchange reactions with TMA and/or H₂O. On the other hand, the first monolayers of WS₂ grains could be partly linked, for instance, by S-bridges. This leads to a lower number, lower accessibility, and lower reactivity of the reactive sites, which is associated with a lower nucleation density and slower coalescence.⁴² Finally, the nucleation density reaches a plateau at 3.6 × 10¹⁰/cm² after 200 cycles [Fig. 4(e)]. This may indicate that an equilibrium is reached between adsorption, in which new nuclei are formed, and coalescence of islands, in which their number is reduced.

As a result of conformal Al₂O₃ deposition in both the vertical and horizontal orientations, the nanoribbons eventually coalesce into a nanoscale membrane during the ALD process. In reality, the GPC of Al₂O₃ ALD is not far from the lateral and vertical growth rates of Al₂O₃ nanoribbons developing at the step edges, Fig. 4(c). As a result, between 200 and 600 cycles, the surface coverage gradually rises, and this is the layered closure’s limiting factor. After around 600 ALD cycles, linear growth is finally seen, indicating a multilayer closure with 1.65 × 10¹⁷ Al atoms/cm². In comparison to other substrates, Al₂O₃ ALD at 300 °C requires a substantially higher Al content to accomplish linear growth, such as 1.6 × 10¹⁶ atoms/cm² or 50 ALD cycles on H-terminated Si and 7.0 × 10¹⁶ atoms/cm² or 230 ALD cycles on polycrystalline MoS₂ with grain sizes between 10 and 50 nm.²⁸ After achieving the linear growth condition, the growth cycle (GPC) is 3.2 × 10¹⁴ Al atoms/cm². In conclusion, the nucleation and growth mechanism are significantly different from the typical 3D key mechanisms proposed in the literature for ALD⁴⁴ because of the heterogeneity of the surface and the different adsorption and nucleation kinetics at the grain boundaries and step edges of the WS₂ surface. We infer by discussing the outcomes of the ALD processing conditions, such as a lower deposition achieved by reducing the N₂ push flow, which is supported by AFM pictures in Fig. S1.⁴⁶ The simulation results for the Hemi-cone model and the 60-cycle model demonstrate strong nucleation, and the push flow has a modest growth rate of less than 150 °C (0.02 nm/cycle). A better film closure is indicated by the observation of points even after the simulation line for 2D development [Figs. 4(a) and 4(b)]. In these analyses, the H₂O solution’s rinse duration was adjusted to 10 s to prevent negative impacts on the reaction surface group ([Figs. 4(c) and 4(a)]. The test results of 0–200 cycles are shown, which is at 150 °C temperature, for 80, 100, 150, and 200 cycles. It is clear from the simulation that the 2D growth condition is the same as what we observed from imec clean (Al₂O₃ grows on SiO₂, and the test conditions are the same as the growth of WS₂/SiO₂/Si). In contrast, imec clean can simulate 2D growth conditions, such as all growth cycles growing 3.54 × 10¹⁵ × growth rate × cycle number. At low cycle numbers, WS₂/SiO₂/Si can be observed and have low adsorption on the WS₂ surface, which results in particles forming just at step heights, vacancies, and W–S–H bonds. As a result of this, surface irregularities are formed at the initial stage (nonconformal surfaces) because the nucleation density per cycle, growth rate along with grain boundary, as well as geometrical interaction between nanoparticles are accumulated into an F-factor, which is calculated in the second layer closure step: Sum of the growth number per cycle (lateral direction) and proposed hemi-cone model is provided and mentioned in Fig. 3 given below.

In Fig. 4(a), the 0–200 cycle numbers cannot accurately fit the experimental results. This calculation formula will be restricted by the topography of the grain boundary surface if F = 10. Since WS₂ has a limited surface morphology, Fig. 2(a), AFM results show its morphology. This stage covers from the initial stage to the transition stage, so as shown in Fig. 2(e), after all, grain boundaries are covered by Al₂O₃; then, F-factor and nucleation density will no longer be added, only horizontal growth and as shown in Fig. 4(e), nuclei do not increase after 200 cycle number, as in Fig. 4(d). After 200 cycles, the slope of areal density for several ALD cycles gradually becomes a constant value. According to different test conditions, this growth model can determine which one has a faster growth rate. At the same time, it also verifies the WS₂ surface. Under this test condition, a nucleation site is still needed for surface adsorption and growth, Fig. 4(b), and the growth results under different test conditions (low push flow) are obtained. As shown in the AFM results in Fig. S3,⁴⁶ it can be seen that the growth still starts from the grain boundary, but the film closure occurs earlier, and the growth from the horizontal direction and the nucleation density has increased significantly, indicating that the horizontal growth rate is much larger than the vertical growth rate, which is the main cause of the faster film closure under the low push flow test condition.

The ALD develops in various directions on the surface depending on the growth mode, and WS₂ is increased by 60 ALD cycles with some sizable triangular crystals. This is complicated and has to be described appropriately. Instead, the complete layer closure was examined individually by watching the TOF-SIMS W⁺ signal decay as a function of Al content and examining the rise in AFM surface coverage as a function of A1 content, Figs. 5(a) and 5(b). Figure 5(a) depicts an Al₂O₃ growth film made from 600 cycles, with a total atom content of 1.65 × 10¹⁷/cm², a thickness of 46 nm, and a WS₂ surface (97% covered). The crystal base surface needs to be shaped into a sizable, inert triangle form. Al₂O₃ in WS₂ experiences a slower membrane closure process, which is consistent with the development in surface resistivity shown on the growth curve, as shown by the less severe decay in Fig. 5(b), compared to the perfect 2D growth. Because of the huge grain size and subsequent stabilization and deception of the signals at TOF-SIMS, which mostly impacts the W⁺ signals, the layers are closed by 400–600 ALD processes. When the normalized W⁺ signal declines from 10 to 3, it indicates that the layer has completely closed.

The surface coverage and morphological progression from Al₂O₃ to WS₂, Fig. 1(a), can be used to define the growth mode. Early nucleation and growth are driven by thermal ALD self-terminating chemical processes. The borders of the phases and the grain boundaries are where the majority of the nuclei are deposited. After 40 cycles, Fig. 4(e) displays the ALD reaction outcomes, including the nucleation of crystals along the device’s borders. As the ALD reaction proceeds, the grain boundaries increasingly widen until the Al₂O₃ film completely encases the reaction sites at all grain boundaries. The majority of the S–H surface groups at grain boundaries and step edges are covered by transition growth, which causes the nanoparticles to connect and turn into the growth model of the nanoribbons. The nanoribbons were then stretched both horizontally and vertically. Rapid nucleation happens at the step edge because of the poor coverage of the phase edge, yet this growth mode results in a slowly closing membrane. The WS₂ film oxidizes quickly when exposed to the environment because sulfur is sensitive to air.

The nucleation and growth mechanism for Al₂O₃ ALD on the WS₂ monolayer is complicated, as evidenced by the experimental AFM and ERD data outlined in Sec. III C. Although data show differing adsorption and nucleation kinetics at the grain boundaries and step edges, nucleation mostly occurs at these locations. The nucleation and growth process differ significantly from the typical 3D key mechanisms for ALD that have been suggested in the literature.⁴³ A more complex nucleation model was created. The extra material contains further information and relationships as in Fig. S2.⁴⁶ The growth curves may be statistically evaluated to comprehend the existing model shown in Fig. 6. The computation will be made more straightforward by assuming that the grain boundaries where growth begins are square at the surface. The number of grains per unit area may be computed using the formula 1/$l$², where l (nm) is the grain size. The total square side length (grain boundary length) is therefore 2/$l$ per unit area. Because it is believed to be significantly less than the grain size (∼1 vs 20 nm on average), the width of the grain border is disregarded.^44,45 In agreement with our AFM observations, the lateral and vertical growth rates are assumed to be equal to the GPC of Al₂O₃ ALD, given as Δr. The radius of the Al₂O₃ nanoribbon after n cycles is r = n ⋅ Δr deposited nanoribbons are modeled as half-cylinders, and the volume after n cycles can be expressed as follows:

The total Al content (atom density) deposited on WS₂ can be obtained as follows:

where ρ is the Al atomic density of Al₂O₃ (0.55 nm) calculated based on the density of ALD-Al₂O₃ (ρAl₂O₃ = 7.85 g/cm³). In Figs. 6(a) and 6(b), the dashed lines represent the estimated curves (b). The radius of the Al₂O₃ nanoribbon and, therefore, the total surface area of Al₂O₃ grow as the growth in the transition zone advances, which accounts for the rise in GPC. The GPC is even greater before the layer closes than in the linear growth model. Currently, the Al₂O₃ nanoribbon’s predicted surface area is more than the surface area of the ideal scenario, which explains why the GPC is higher. Once the WS₂ is completely covered by Al₂O₃ and linear growth is achieved, the appropriate fit of the nonlinear parts of the growth curve shows that the proposed growth model agrees with the experimental results.

This work provides details and discussion of Al₂O₃ ALD process development, evolution, and morphology on WS₂ including its nucleation and growth processes of Al₂O₃ on WS₂. Large WS₂ crystals have outstanding optical and electrical qualities, as well as great crystal quality, equivalent to samples that have been manually exfoliated and improved their electrical performance with monolayer SiO₂/Si substrates for WS₂ film transistor arrays. We demonstrated that atomic layer deposition of SiO₂/Si on WS₂ TMDs can act as a foundation for optoelectronics, valleytronics, and the next generation of electronic devices. Silicon substrates are subjected to surface modification techniques to reduce their reactivity with ALD reactants and manage silicon oxidation. The creation of such pretreatments will allow for the continuous growth of ALD layers on the substrate of the device that are of varied thicknesses. The growth study findings indicating that the WS₂ basal plane has a modest level of ALD-related inherent reactivity support the idea that functionalizing the TMD surface is necessary to improve the nucleation of ALD. Future studies in this area may benefit from these findings.

This work was financially supported by the “Center for the Semiconductor Technology Research” from The Featured Areas Research Center Program within the framework of the Higher Education Sprout Project by the Ministry of Education (MOE) in Taiwan. Also, it is supported in part by the Ministry of Science and Technology, Taiwan, under Grant Nos. MOST 110-2634-F-009-027 and 110-2622-8-009-018-SB and by the National Chung-Shan Institute of Science and Technology, Taiwan, under No. NCSIST-403-V309(110).

The authors have no conflicts to disclose.

Tsu-Ting Lee: Data curation (equal); Formal analysis (equal). Kashi Chiranjeevulu: Writing – original draft (equal); Writing – review & editing (equal). Sireesha Pedaballi: Software (equal); Writing – review & editing (equal). Daire cott: Formal analysis (equal); Resources (equal); Supervision (equal). Annelies Delabie: Conceptualization (equal); Resources (equal); Supervision (equal). Chang-Fu Dee: Conceptualization (equal); Resources (equal); Software (equal); Supervision (equal). Edward Yi Chang: Funding acquisition (equal); Project administration (equal); Resources (equal); Supervision (equal).

The data that support the findings of this study are available within the article and its supplementary material.