Mingxi
Chen
a,
Jianwei
Chai
a,
Jing
Wu
a,
Haofei
Zheng
ab,
Wen-Ya
Wu
a,
James
Lourembam
a,
Ming
Lin
a,
Jun-Young
Kim
a,
Jaewon
Kim
a,
Kah-Wee
Ang
ab,
Man-Fai
Ng
c,
Henry
Medina
*a,
Shi Wun
Tong
*a and
Dongzhi
Chi
*a
aInstitute of Materials Research and Engineering (IMRE), Agency for Science, Technology and Research (A*STAR), 2 Fusionopolis Way, Innovis #08-03, Singapore 138634, Republic of Singapore. E-mail: henry.medinasilva@imec.be; tongsw@imre.a-star.edu.sg; dz-chi@imre.a-star.edu.sg
bDepartment of Electrical and Computer Engineering, National University of Singapore, 4 Engineering Drive 3, Singapore 117583, Republic of Singapore
cInstitute of High Performance Computing (IHPC), Agency for Science, Technology and Research (A*STAR), 1 Fusionopolis Way, #16-16 Connexis, Singapore 138632, Republic of Singapore
First published on 5th October 2023
Atomically-thin monolayer WS2 is a promising channel material for next-generation Moore's nanoelectronics owing to its high theoretical room temperature electron mobility and immunity to short channel effect. The high photoluminescence (PL) quantum yield of the monolayer WS2 also makes it highly promising for future high-performance optoelectronics. However, the difficulty in strictly growing monolayer WS2, due to its non-self-limiting growth mechanism, may hinder its industrial development because of the uncontrollable growth kinetics in attaining the high uniformity in thickness and property on the wafer-scale. In this study, we report a scalable process to achieve a 4 inch wafer-scale fully-covered strictly monolayer WS2 by applying the in situ self-limited thinning of multilayer WS2 formed by sulfurization of WOx films. Through a pulsed supply of sulfur precursor vapor under a continuous H2 flow, the self-limited thinning process can effectively trim down the overgrown multilayer WS2 to the monolayer limit without damaging the remaining bottom WS2 monolayer. Density functional theory (DFT) calculations reveal that the self-limited thinning arises from the thermodynamic instability of the WS2 top layers as opposed to a stable bottom monolayer WS2 on sapphire above a vacuum sublimation temperature of WS2. The self-limited thinning approach overcomes the intrinsic limitation of conventional vapor-based growth methods in preventing the 2nd layer WS2 domain nucleation/growth. It also offers additional advantages, such as scalability, simplicity, and possibility for batch processing, thus opening up a new avenue to develop a manufacturing-viable growth technology for the preparation of a strictly-monolayer WS2 on the wafer-scale.
New conceptsIn this manuscript, a sublimation-assisted and self-limited thinning approach for wafer-scale transition metal dichalcogenides (TMDCs) is demonstrated for the first time. Tungsten dichalcogenide (WS2) – a promising TMDC-based channel material for next-generation Moore's nanoelectronics is selected as a typical example. Based on sulfurization with pulsed sulfur gas supply and constant H2/Ar gas flow, the thinning effect coming from the volatile reduced WS2−x phase by H2 under a temperature substantially above the sublimation temperature of WS2 can become self-limited. Importantly, the self-limited nature of the thinning process enables the damage-free and strictly monolayer TMDCs formation. This new thinning process simplifies the current challenges in preparing the high-quality monolayer TMDCs, where an inevitable partial oxidation and physical damages associated from conventional etching processes severely degrade the film stoichiometry and uniformity. Notably, the universality of the self-limited thinning effect is also manifested on a broader number of different TMDCs (e.g. MoS2) and TMDCs prepared under various growth approaches (e.g. PVD sputtered WS2 and MoS2) and doped TMDCs system (e.g. Co-doped WS2 prepared by sulfurization). The findings indicate that this thinning method is broadly applicable to pure and doped TMDCs systems, which will accelerate the research of TMDCs with high PL quantum yield and adjustable band gap for large-area optoelectronics and spin-electronics. |
Over the last 5 and more years, there has been significant progress in the effort for growing large-area atomically thin 2D TMDCs, including WS2. Among the various growth methods developed, including chemical vapor deposition (CVD), metal organic chemical vapor deposition (MOCVD), physical vapor deposition (PVD), and molecular beam epitaxy (MBE),11–13 CVD and MOCVD are considered as the two most promising front-runners in terms of the quality of the grown TMDCs layer and the wafer-size for uniform growth. Both are able to grow monolayer TMDCs on the wafer-scale (up to 4 inch wafer for CVD and up to 8 inch wafer for MOCVD).14,15 However, both CVD and MOCVD also have their own shortcomings for adoption in scalable mass-production. For CVD, where metal oxide (e.g., WO3, MoO3etc.) powders are typically used as metal precursors, it is difficult to further scale up the wafer-size due to its so-called distance-dependent growth mechanism.16,17 In comparison, MOCVD offers a good control over the area uniformity, and therefore greatly improved scalability.18 However, an un-acceptably long growth time (more than 26 h) is required to achieve monolayer growth.19
Other than those limits associated with the individual growth methods, a more general issue with almost all TMDCs growth arises from the difficulty in controlling the layer number. Unlike graphene growth, which is governed by a catalytically driven self-limiting growth mechanism, the CVD/MOCVD/PVD/MBE growth of TMDCs is non-self-limiting growth.20,21 The 2nd layer nucleation/growth occurs after substantial growth of the 1st layer domains, typically at the centers of domains, due to competition between lateral vs. vertical growth, or at their grain boundaries after their coalescence.22–24 This poses a serious challenge for their consistent and robust usage in nanoelectronics since the presence of 2nd layer domains change the electronic structure in TMDCs (since the electronic structure of the bilayer TMCDs is significantly different from monolayer TMDCs), consequently resulting in variation of the device parameters and affecting circuit yields.25–27
Alternative approaches based on the post-growth treatments (e.g., low-power oxygen plasma,28,29 laser thinning method,30 XeF2 gas etching,31etc.) were also used to achieve monolayer TMDCs. Promising thinning effects had been demonstrated by trimming down the few-layer TMDCs designated areas to atomically-thick layers. However, there are also several limitations in these approaches. First, partial oxidation happens on the post-treated TMDCs due to the thinning process being performed under ambient conditions. Such ambient air oxidation would induce uncontrollable doping in TMDCs with O2, H2O or N2, and it is necessary to perform extra sulfurization on post-treated TMDCs to restore their stoichiometry. Second, the etching results are highly sensitive to the exposure time and power of the oxygen plasma, laser beam, and XeF2 etchant gas. More importantly, these post-etching methods are not self-limiting, and it is thus very challenging to avoid the occurrence of physical damages and complete etching on the thinnest TMDCs areas.26,31
Here, we present the use of an in situ self-limited thinning of few-layer WS2 formed by the sulfurization of WOx (W) for the growth of a wafer-scale fully-covered strictly-monolayer WS2. This approach plays with the thermodynamic instability of the WS2 top layers as opposed to the stable bottom monolayer WS2 on suitable substrates (with strong interface coupling), especially in its reduced state, i.e., WS2−x, above the vacuum sublimation temperature of WS2 as revealed in Density functional theory (DFT) simulations. However, we also found that the alternating switching between thinning and non-thinning via a pulsed supply of sulfur precursor vapor in a constant flow of H2 and Ar gas is essential for process robustness and reliability. Strictly-monolayer high quality WS2 films have been prepared successfully on 4 inch substrates using this method with excellent uniformity. The grown monolayer WS2 was further examined by optical microscopy, Raman microscopy, photoluminescence (PL), X-ray photoelectron spectroscopy (XPS) and transmission electron microscopy (TEM) to reveal its high quality. A number of FETs were fabricated on the synthesized monolayer WS2 to demonstrate its potential application in electronics and optoelectronics. Other than overcoming the major intrinsic limitation of conventional vapor-based growth methods in preventing the 2nd layer WS2 domain nucleation and growth, additional potential advantages of the self-limited thinning approach could also include: (1) scalability (growth area is determined by the original sputter-deposited WOx (or W) film coverage); (2) simplicity in process methodologies (the use of industry-standard sputter-deposition and conventional sulfurization); and (3) possibility for batch processing (several WOx(W)-deposited wafers can be processed simultaneously in a large growth reactor).
As mentioned earlier, the demonstration of the wafer-scale monolayer WS2 formation shown in Fig. 1c is based on the sublimation-assisted and self-limited thinning process of few-layer WS2. Indeed, an apparent color difference can be seen from the photographs of WS2 films grown in the midway (top) and completion (bottom) of the PSS process growth cycle (Fig. 1f) due to the variation of the bandgap in WS2 of different thicknesses. The large value of Δω (∼64 cm−1) and small PL emission suggested that the WS2 grown in the midway of the PSS growth cycle (the state after ramping up from 600 °C to 1000 °C in PSS process) was few-layer thick (measured on spot 5 in Fig. 1g and h). Comparatively, a monolayer WS2 would be formed if its growth cycle went through the entire PSS process, as evidenced by the value of Δω ∼ 61.5 cm−1 (spots 1–4 in Fig. 1g) and strong PL emission at 1.98 eV (spots 1–4 in Fig. 1h). In short, the self-limited thinning PSS process consisted of 5 steps: (1) multi-layer WS2 growth in the early low-temperature stage, (2) reduction of the top few-layer WS2, (3) sublimation into the volatile WS2−x species (thinning process), (4) sulfur diffusion to restore WS2−x back to stable WS2 on sapphire, and (5) a formation of stable and strictly monolayer WS2 (Fig. 1i–m). The detailed growth mechanism will be discussed with XPS, TEM and DFT results in the later sessions. Notably, the self-limited thinning PSS process can also be manifested on PVD-sputtered WS2 (Fig. S2, ESI†) and MoS2 (Fig. S3, ESI†), in which the few-layer regions are trimmed down into strictly-monolayer with full surface coverage. In addition, this PSS monolayer WS2 method can be extended to the large area uniform formation of doped W1−x(Mx)S2 monolayer, as revealed by our preliminary work with Co doping (Fig. S4, ESI†). These findings are an embodiment of the universality of the self-limited thinning effect in the PSS process.
The as-grown PSS-WS2 films were transferred onto a SiO2/Si substrate for further examination through a conventional van der Waals 2D layer transfer method.35Fig. 2a indicates the optical microscope image of a transferred monolayer WS2 film on the substrate of SiO2/Si, while Fig. 2b shows the mapping image of the PL peak intensity over the red dashed box region in Fig. 2a. The homogeneous color contrast shown from the mapping image suggests that the transferred WS2 film maintains a high degree of integrity with minimum defects. This is further confirmed by AFM measurement (Fig. 2c), where a surface roughness (Rq) as low as 0.212 nm is obtained without revealing any evidence of the presence of bilayer domains. The height profile along the blue-dotted line indicates that the thickness of WS2 is about 0.65 nm, agreeing well with the reported value (∼0.67 nm) for monolayer WS213 and thus unambiguously confirming that a strictly monolayer WS2 is indeed prepared by the PSS sulfurization and subsequent self-limited thinning process. TEM characterization was carried out to reveal the crystalline structure of the transferred monolayer WS2 (Fig. 2d and e). Excellent uniformity of the monolayer WS2 is evident from the TEM image (over 60 nm × 60 nm), where neither bilayer regions nor grain boundaries are present. A well-defined atomic arrangement is observable in the High-Resolution TEM image, indicating a high crystallinity of WS2. The measured d-spacing of the (1010) planes are about 0.27 nm, which coincides well with the reported value for 2H-WS2.36 To assess its single crystallinity, selected-area electron diffraction (SAED) was performed on four different locations (over μm-scale area) of the monolayer WS2 (Fig. 2f). As shown in Fig. 2f, each pattern shows one set of six hexagonal dots, indicating high crystal crystallinity, while all of the patterns present an almost identical lattice orientation, suggesting the presence of large grains in μm scale.
With the formation of monolayer WS2 by the self-limited PSS process being confirmed by the experimental results presented above, X-ray photoelectron spectroscopy (XPS) was carried out to analyze the phase evolution in the original WOx film during/after the CSS and PSS process to gain insight of the monolayer formation mechanism (Fig. 3a). It was found that the sulfurization of the WOx film at 600 °C during the initial stage of the CSS and PSS process (for 10 min, see Fig. 1b) resulted in a substantial, but incomplete conversion of WOx to WS2. This is evident from the observation of three W 4f core-level peaks (i.e., two dominant W 4f7/2 (∼32.8 eV) and W 4f5/2 (∼35 eV)) and a faint W 5p3/2 (∼37.9 eV) (peaks) corresponding to WS2, together with two weak W 4f peaks (W 4f7/2 (∼36.1 eV) and W 4f5/2 (∼38.2 eV)) from WO3. Cross-section TEM of the 600 °C-sulfurized film indeed confirms the partial formation of few-layer WS2 (2–4 layers) on top, while a substantial amount of residual WOx remains beneath the WS2 formed (Fig. 3c–e). It is seen that the WOx phase remains even after further sulfurization at a much higher temperature of 1000 °C under a constant DES supply, i.e., CSS process (Fig. 3a), presumably due to the effective inhibition of S penetration by the top WS2 multilayer.
The multilayer nature of CSS-WS2 is revealed by its much larger A1g/E2g ratio of ∼0.6, which agrees with reported results,37 and significantly weaker PL intensity as compared to PSS-WS2 (Fig. 4a and b), which may also lead to a higher roughness of the CSS-WS2 surface (Fig. S5, ESI†). An apparent discrepancy in film color can also be observed from the photographs of the CSS-WS2 and PSS-WS2 films (Fig. 4c and d) due to the variation of the bandgap in WS2 of different thicknesses. In contrast, the WO3 phase is completely absent in the PSS processed sample. Only the three W 4f7/2 (∼32.8 eV), (W 4f5/2 (∼35 eV)), and W 5p3/2 (∼37.9 eV) peaks corresponding to WS2 phase are observed, in agreement with the confirmation of the formation of the monolayer WS2 in the PSS sample by the experimental data presented in previous sections.
As opposed to the multilayer WS2 in the CSS-WS2 sample, the formation of monolayer PSS-WS2 certainly indicates that effective thinning of the multilayer WS2 occurs only during the DES turn-off periods. This is presumably due to the etching and/or enhanced sublimation of the top WS2 layers by H2. In understanding this thinning process, we note that there are several previous reports relevant to the thermal and chemical stability of WS2, which, together with our own experimental observations, may help to reveal the underlying mechanism. First, it was previously reported that a post-growth thermal treatment under H2 + Ar mixed gas is not effective in etching CVD-grown WS2 at temperatures up to 900 °C.24 This is in agreement with our experimental finding that monolayer WS2 formation by PSS is possible only for temperatures above 950 °C. As shown in Fig. 4e and f, there is a significant increase in the PL intensity of PSS-WS2 without the presence of bilayer domains upon preparing the film above 950 °C. This temperature-dependent optical/structural transition is highlighted into two colored zones showing that the multilayer WS2 and strictly-monolayer WS2 were prepared at low temperature range (850–925 °C) and high temperature range (950–1025 °C), respectively. This finding validates that a high enough temperature (larger than the vacuum sublimation temperature ∼870 °C of WS2)38 is crucial for providing the thermodynamic driving force to trim down the overgrown WS2 top layers from the bottom monolayer WS2/SAP. Second, a study of the thermal stability of WS2 and MoS2 in vacuum set the sublimation and decomposition temperatures for bulk WS2 around 870 °C and 1040 °C, respectively, as observed from the compact WS2 beginning to have a detectable weight loss rate at 870 °C and sulfur peaks appearing at 1040 °C.38 Third, Wu et al. demonstrated the vapor-solid growth of high optical quality MoS2 monolayers by the sublimation of the MoS2 powder source at 900 °C.39 Meanwhile, a vacuum sublimation temperature of ∼930 °C was determined for bulk MoS2, a temperature higher than the vacuum sublimation temperature ∼870 °C of bulk WS2.38
Based on the experimental observations summarized above and knowing the critical role of H2 in the PSS process (note: the formation of WS2 monolayer by PSS failed in the absence of H2 gas; see Fig. S6, ESI†), we explain the self-limited thinning of multilayer WS2 by PSS for the formation of a stable monolayer WS2 as follows. (1) At a temperature substantially higher than the vacuum sublimation temperature of ∼870 °C of WS2, there exists a certain thermodynamic driving force for multilayer WS2 to sublimate, with a drastically accelerated sublimation for highly volatile reduced WS2−x (Fig. 1j). (2) At a temperature (e.g., 1000 °C) much higher than ∼870 °C and in the presence of H2, the top few-layer WS2 (formed after the initial 600 °C sulfurization) would undergo a reduction process via H2(g) + WS2(s) → H2S(g) + WS2−x(s), leading to a drastically enhanced sublimation in the volatile WS2−x and thus layer-by-layer thinning, during DES turn-off periods (Fig. 1k). (3) This thinning process would be inhibited or drastically minimized in the presence of the DES supply (i.e., during DES turn-on period), due to the massive annihilation of S vacancies (created either during previous DES turn-off period or current DES turn-on period) by the readily available S atoms and, consequently, the restoration of the volatile WS2−x back to more stable WS2. During this DES turn-on period, further conversion of underlying remaining WOx, if there is any, into WS2 also takes place due to enhanced S diffusion through the top WS2 multilayer, the thickness of which was trimmed down during the previous turn-off period. (4) Once the multilayer WS2 finally reduces to a monolayer limit after several DES turn-on/turn-off cycles, the sublimation would stop effectively, even for the volatile WS2−x phase, most likely due to the stronger interface coupling of the bottom monolayer WS2 with sapphire substrate, as compared to that for WS2–WS2 interface. It should be noted that a more negative binding energy (Eb) of −2.23 eV is obtained for the monolayer WS2–sapphire interface as compared to Eb = −2.01 eV for the WS2–WS2 interface by DFT calculations (Fig. 3b), which means a significantly increased sublimation temperature for the monolayer WS2 on the sapphire surface. The effective suppression of sublimation in monolayer WS2 is also evident from the observation of the persistent existence of the PSS-WS2 monolayer after high temperature annealing under the steady flow of the H2 + Ar mixed gases without a DES supply (Fig. S7, ESI†). (5) The final pulse DES supply in a complete PSS process cycle would restore the WS2−x phase to a stable WS2 monolayer via a reaction of WS2−x(s) + xS(g) → WS2(s), thus realizing the self-limited thinning of multilayer WS2 for monolayer WS2 formation (Fig. 1l and m). (6) On the other hand, the failure in the formation of monolayer WS2 by the CSS process is due to the overwhelming restoration process (WS2−x(g) + xS(g) → WS2(s)) against the thinning process involving (1) reduction (H2(g) + WS2(s) → H2S(g) + WS2−x(s)) and (2) subsequent sublimation (WS2−x(s) → WS2−y(s) + volatile WS2−x(g)) under an environment with a constant sulfur supply. Here, we would like to highlight that the self-limited thinning process can also be employed on other suitable substrates, as demonstrated by the formation of the stable monolayer WS2 on SiO2/Si substrates (Fig. S8, ESI†). However, it is found that the formation of the monolayer MoS2 fails on SiO2/Si (i.e., complete stripping-off of MoS2), as shown in Fig. S9 (ESI†). In contrast, a continuous monolayer MoS2 is successfully prepared on sapphire substrates under the same PSS processing condition (Fig. S3, ESI†), clearly indicating the crucial effect of substrate on the PSS process and its extendibility. The instability of monolayer MoS2 on SiO2/Si is most likely due to the weaker binding energy of the MoS2–SiO2 interface than that for the MoS2–sapphire interface.
To investigate the electrical properties of the monolayer WS2 film grown via the self-limited thinning PSS process, bottom-gated FET devices were fabricated, with a channel length L = 25 μm and a channel width W = 200 μm, respectively. Fig. S10 (ESI†) shows the schematic of the device structure. Details of the device fabrication steps are described in the Methods. The fabricated WS2 FET devices exhibited typical n-channel FET characteristics, as shown in Fig. 5, similar to previous reports for WS2 FETs either with exfoliated or CVD-grown monolayer WS2 as a channel.13,40,41 The WS2 FET mobility was calculated using the expression,
(1) |
Our values compare rather favorably to those reported in the literature (μeff = 0.16–1.97 cm2 V−1 s−1) for the exfoliated single crystalline monolayer WS2 FETs fabricated with similar source and drain metal contacts,9,42,43 implying the good quality of our monolayer WS2 film. It must be pointed out that, while the intrinsic mobility is a material-related parameter that is determined by intrinsic material properties such as crystallinity, crystal-defects/impurities, and grain sizes etc., μeff is also profoundly influenced by extrinsic effects, such as the contact resistance, gate-dielectric layer used, and integrity of the interface between the channel and gate-dielectric layer. For atomically-thin WS2 FETs, a μeff as high as 214 cm2 V−1 s−1 at 300 K was reported on single layer WS2 sandwiched between CVD-grown h-BN films.44 A μeff up to ∼115 cm2 V−1 s−1 at 300 K was also achieved for the exfoliated few-layer WS2 through monolayer-h-BN/Cr/Au tunneling barrier contact. In contrast, a low mobility of ∼1.44 cm2 V−1 s−1 at 300 K was extracted on the same few-layer WS2 film with direct Indium contact.42 Therefore, our reported carrier mobilities here should be taken as a lower limit of its intrinsic carrier mobility as there is still potential for improvement through various approaches, e.g., optimization of ALD oxide growth,45 choices of dielectric material (HfO2 instead of Al2O3),46 improvements in interface quality,44,47 while the contact resistance can also be further reduced.42,48 Nonetheless, our transport measurements indicate that a semiconducting WS2 monolayer film has been successfully grown uniformly by self-limited process on a 4 inch sapphire wafer.
(2 × 2) Al2O3 and (3 × 3) WS2 were employed in building the model of WS2 on Al2O3. The lattice mismatch was smaller than 0.7% with around 9.3 Å thick Al2O3 slab. In the optimization stage, we have frozen the bottom four layers (total nine layers). The optimized lattices of the sapphire substrate (bulk Al2O3) were a = 4.785 Å and c = 13.053 Å, while the bulk WS2 (2H) were a = 3.169 Å and c = 12.423 Å. The oxygen-terminated slab models were constructed based on the bulk structures.
Considering the interaction along the interface will affect the thermal stability significantly, the theoretical binding energy (Eb) at the WS2–WS2 interface and WS2–sapphire interface (Fig. 3b) was calculated by eqn (2):
Eb = Esubstrate+xWS2 − (Esubstrate+(x−1)WS2 + EWS2) | (2) |
To characterize the electrical performance, the monolayer WS2 was transferred onto the fresh target substrate by a conventional PMMA method.35 In short, PMMA was spun onto the as-grown WS2/sapphire with 3000 rpm for 60 s, and then heated at 150 °C for 2 min. Next, the PMMA-coated WS2 was separated from sapphire by dipping into 1 M KOH solution slowly. It was then picked up with a PDMS sheet, followed by rinsing in DI water several times to remove the residuals. Subsequently, the new PDMS stamp (DGL-45 × 45-0065-X4, GelPak) was then used to transfer the resultant film onto the target substrate and dried inside a fume hood overnight. Next, the PDMS was carefully detached from PMMA/WS2 at 150 °C, followed by a rinsing process to remove PMMA in the acetone and IPA baths. Finally, the sample was annealed at 300 °C for 30 min under Ar atmosphere.
The WS2 devices were fabricated based on the classical bottom-grid-top-electrode structure. The Ti/Au (10 nm/40 nm) electrodes were patterned by shadow mask under electron beam evaporation. The channel width and length were determined by Field Emission Scanning Electron Microscopy (FESEM, JEOL JSM7600F). Electrical measurements were carried out in air condition by a semiconductor device analyzer (B1500A, Agilent) equipped with probe stations (Probing Solutions, Inc.).
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3nh00358b |
This journal is © The Royal Society of Chemistry 2024 |