WorksPhysics and Astronomy
Toward Phonon-Limited Transport in Two-Dimensional Electronics by Oxygen-Free Fabrication
Abstract
Future electronics require aggressive scaling of channel material thickness while maintaining device performance. Two-dimensional (2D) semiconductors are promising candidates, but despite over two decades of research, experimental performance still lags theoretical expectations. Here, we develop an oxygen-free approach to push the electrical transport of 2D field-effect transistors toward the theoretical phonon-limited intrinsic mobility. We achieve record carrier mobilities of 91 (132) cm2V-1s-1 for mono- (bi-) layer MoS2 transistors on SiO2 substrate. Statistics from over 60 devices confirm that oxygen-free fabrication enhances key figures of merit by more than an order of magnitude. While previous studies suggest that 2D transition metal dichalcogenides such as MoS2 and WS2 are stable in air, we show that short-term ambient exposure can degrade their device performance through irreversible oxygen chemisorption. This study emphasizes the criticality of avoiding oxygen exposure, offering guidance for device manufacturing for fundamental research and practical applications of 2D materials.
Silicon complementary metal-oxide-semiconductor (CMOS) technology has been pivotal in the progress of modern electronics. Advances in nanofabrication techniques have enabled continued scaling of CMOS transistor technology.[1] Meeting the growing demand for cheaper, faster, and more energy-efficient electronics will require the ultimate scaling of the semiconductor channel thickness toward the sub-nanometre regime. However, achieving this with any three-dimensional semiconductor is challenging due to the increased surface roughness scattering from dangling bonds that severely degrades carrier mobility.[2] 2D transition metal dichalcogenide (TMD) semiconductors with intrinsic atomic-scale thickness ( nm) could be a solution.[3, 4, 5, 6, 7, 8, 9] Their van der Waals (vdW) layered nature provides dangling-bond-free surfaces such that high carrier mobilities are maintained even in monolayer devices. This has positioned 2D TMDs on the technology roadmap of industry, including companies such as the Taiwan Semiconductor Manufacturing Corporation (TSMC), Samsung, Intel, and IMEC.1 However, despite promising early device prototypes, 2D TMD transistor performance falls short of their theoretical expectations.[11, 12, 13] Realizing their full potential will depend on developing high-quality material growth, controlled doping, dielectric integration, and contact engineering methods.[4, 7, 9, 14] Encouragingly, there has been considerable progress in these areas. Chemical vapor deposition growth can produce highly crystalline materials with fewer defects and impurities to rival exfoliated material quality.[15, 16, 17] The vdW interfaces formed between TMDs and In[18, 19], Bi[20], Pt,[21], and Sb[22, 23] three-dimensional (3D) metals, as well as Y-doping[24] of source-drain contact regions, have led to ultralow contact resistances. Surface[25], substitutional[26], and sub-stoichiometric oxide[27, 28] doping have increased sheet conductance and on-state currents. Dielectric and substrate engineering to quench impurity and phonon scattering have boosted TMD carrier mobilities.[29, 30, 31, 32, 33] These advances have narrowed the experiment-theory device performance gap and showed that progress depends on tailored solutions that consider the unique nature of 2D materials: their extreme geometry.
2D geometry with extreme surface-to-volume ratios heightens the sensitivity of 2D materials to their environment.[34, 35, 36, 37, 38] A monolayer TMD comprises a three-atom crystal lattice where the layer of transition metal atoms is sandwiched between two layers of chalcogens. The top layer of chalcogen atoms forms a surface that can interact with ambient molecular absorbates. There are four possible interaction scenarios for molecular adsorbates on TMD surfaces (illustrated in Figure 1a).[35] These are generally divided into weakly interacting physisorption and strongly interacting chemisorption. Physisorption of a molecular adsorbate occurs on a (i) sulfur atom or (ii) a sulfur vacancy (), while chemisorption may exhibit either (iii) non-dissociative or (iv) dissociative coupling to a . For dissociative chemisorption, the molecule dissociates into individual atoms which substitutionally replaces a site. This substitution causes a disruptive and often permanent stoichiometry change, e.g., chemical oxidation. Ambient oxidation is an undesired material degradation of TMDs. It is well documented, as changes in stoichiometry are easily detected by experimental techniques like X-ray photoemission spectroscopy (XPS) (Extended Data Figure 1) or Raman and photoluminescence spectroscopy.[38, 39, 40, 41, 42, 43] In contrast, physisorption (i, ii) is often considered a reversible process where the smaller interaction energies allow adsorbate removal through thermal annealing or vacuum cycling.[44, 45] Consequently, TMDs such as MoS₂ and WS₂ are often considered ambient stable as many studies have confirmed their short-term resistance against (iv) dissociative chemical oxidation. However, the impact and reversibility of non-dissociative chemisorption (iii) remain elusive. Although some theoretical studies[35] have hinted at its significance, this process remains poorly understood due to challenges in experimental validation, especially at the device level, where ambient exposure is unavoidable in standard cleanroom fabrication.

Figure 1. Molecular adsorption on monolayer TMDs. (a) Schematic of possible diatomic molecular adsorbate interaction scenarios with a monolayer TMD. The dotted outlines represent a sulfur vacancy . (b) Adsorption energy of , , and on the pristine surface and sulfur vacancy sites of a monolayer , and resultant (c) electron transfer from TMDs to molecular adsorbates. Electronic band structure and the projected density of states (PDOS) of with sulfur vacancies during

Extended Data Figure 1. X-ray photoemission spectroscopy of the W 4f and W 5p energy regions of pristine, freshly cleaved, and oxidized WS₂ samples. The data intensity (energy scale) is normalized (aligned) to the peak maximum intensity (position) for comparison purposes.
In this work, we study the impact of ambient molecular exposure on the device performance of 2D TMD transistors. Using a purpose-built facility, we can complete the entire fabrication of 2D devices in an inert argon environment. We found that minimizing oxygen exposure throughout the fabrication process results in more than an order of magnitude enhancement across key transistor figures of merit, such as carrier mobility, sheet conductance, and carrier density. This enables us to push our device carrier mobility toward the theoretical limit arising from intrinsic phonon scattering.[11, 12, 13, 46] Recognizing molecular oxygen as a critical factor limiting device performance sets a significant milestone. This marks a qualitative step in bridging the theoretical-experimental discrepancy and guides the manufacture of high-performance 2D devices for fundamental research and practical applications.
To understand the impact and likelihood of physisorption and non-dissociative chemisorption (hereafter referred to as chemisorption in the rest of the manuscript for simplicity), we first perform adsorption energy () and charge transfer () analysis using density functional theory (DFT) simulations (details in Methods, Extended Data Figure 2-Extended Data Figure 5, and SI Figs S1-S7). The simulations evaluate the interactions of common ambient molecules , , and with monolayer and . Using a supercell, we consider both pristine (no ) and defective (single ) monolayers. Such a supercell size corresponds to a defect density of , comparable to expected experimental values.[47]

Extended Data Figure 2. Molecular adsorption on negatively charged sites on the TMDs’ basal plane. (a) Single defect formation in a (4×4×1) supercell as a chalcogen vacancy (), resulting in a defect density of . Unlike the charged neutral state, with a single site, we find three possible, stable configurations for molecular adsorption based on their center of mass position relative to the site (defined as , , and ). These geometric configurations for adsorption and charge delocalization arise from the readjustment of TM-TM bonds relative to the site, as described by the Jahn-Teller distortion mechanism. Notably, if the distance between an molecule and the site is considerably larger ( Å), only physisorption occurs due to weak interactions (as shown in (b)). However, strong chemisorption is observed when occupies the site, leading to significant charge transfer and -doping. Two possible chemisorption configurations can occur depending on the lowest energy states: (c) vertical and (d) horizontal with respect to the 2D basal plane.

Extended Data Figure 5. Effect of oxygen molecular adsorption on monolayer WS₂. (a) Adsorption energy of O₂, N₂, and H₂O on the pristine surface and sulfur vacancy sites. (b) Estimated electron transfer from WS₂ to adsorbate molecules.
(d) physisorption and (g) chemisorption. (e) Molecular orbital diagram of the molecule showing the highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO). Schematic of the electron transfer process and local electronic structure during physisorption (f) and chemisorption (h) on defective . The embedded plot shows the charge density difference, where blue regions represent electron accumulation and red regions represent electron depletion. Electrons transfer from TMD to the LUMO of and accumulate on the molecule in the shape of orbitals.
We calculate the of , , and for (Figure 1b) and (Extended Data Figure 5a). In pristine TMDs, all studied molecules show weak physisorption with characteristic energies meV. In contrast, sulfur vacancies in TMDs support the chemisorption of to the site, while interactions with and remain limited to physisorption. Notably, chemisorption is significantly stronger than physisorption, with values almost an order of magnitude larger, eV. Consequently, this strong interaction between and , coupled with the large electronegativity of molecular oxygen, should result in substantial charge transfer at the chemisorption site. Indeed, our Bader charge analysis confirms this behavior (Figure 1c and Extended Data Figure 5b). chemisorption leads to considerably more charge transfer with for (Figure 1c) and for (Extended Data Figure 5b) compared to physisorption, where for , , and . As further validation, calculations performed using different-sized supercells and defect densities show similar trends for and (SI Figure S1 and Table-S1). The calculated band structure shows that on the surface of TMDs act as electron donors and induce localized defect states in the band gap. When molecules are physisorbed at the site, the lowest unoccupied molecular orbitals (LUMO) of are slightly below the defect levels, suggesting potential electron transfer from the TMD to the molecules (Figure 1d-Figure 1f). In contrast, the chemisorption of fills the sites with oxygen atoms, leading to the disappearance of the defect level. Electron transfer results in the complete occupation and delocalization of the LUMO (Figure 1g and Figure 1h). These simulation results suggest that exposure affects TMD device performance. Even for ambient stable TMDs with low defect densities, the large of chemisorption can lead to unwanted and irreversible -doping. Uncontrolled doping from ambient molecular adsorbates at the source/drain and channel regions leads to detrimental effects such as larger parasitic contact resistance, undesired threshold voltage shift, and lowered sheet conductance.
However, experimentally assessing the impact on devices from strongly chemisorbed is challenging. So far, studies have measured the effects on 2D TMD devices from weakly physisorbed molecules and shown that it is reversible through vacuum and thermal cycling.[44, 45] Evaluating the impact from potentially irreversible chemisorbed molecules will require minimal ambient exposure, even during material synthesis and device fabrication. Currently, 2D TMDs are commonly synthesized through mechanical exfoliation or chemical vapor deposition (CVD). Exfoliation is usually conducted in standard laboratory or cleanroom environments. It can also be performed relatively quickly in controlled environments, such as an inert argon-filled glovebox with minimal and levels. However, for CVD-grown TMDs, limiting ambient exposure is difficult due to the technical challenge of integrating large-growth furnace tubes with gloveboxes. More significant concerns are the materials processing and device fabrication steps. CVD films are typically wet transferred from growth to device substrates. Also, device fabrication will require cleanroom lithography and thin film deposition processes, which must now be in controlled inert conditions.
Therefore, to experimentally evaluate the impact of molecular chemisorption on TMD devices, we designed a solution for TMD device fabrication with minimal ambient exposure. Our approach is based on purpose-built gloveboxes integrated with nanolithography, thin film deposition, and solvent processing capabilities (Extended Data Figure 6), with typical and levels maintained at 1 ppm. For this study, we fabricated sets of and back-gated FETs on 290 nm , utilizing both exfoliated and CVD materials. All devices fabricated in our glovebox facility are hereafter referred to as ‘inert.’ We fabricated equivalent device sets in ambient conditions as a control, hereafter referred to as ‘ambient.’ In total, 38 and 24 devices were measured for statistical analysis of FET metrics, including threshold voltage, sheet conductance, mobility, hysteresis, and interfacial trap densities.

Extended Data Figure 6. Our purpose-built glovebox fabrication facility. This customized setup allows complete fabrication of 2D devices in an argon environment with typical O₂ and H₂O levels maintained below < 5 ppm, usually < 1 ppm. Capabilities include nanolithography, wet chemistry, 2D material transfer, metal deposition, and atomic layer deposition.
Our theoretical results suggest that chemisorption leads to substantial charge transfer doping. We can determine the degree of charge doping by comparing the threshold voltages of ambient and inert devices. Room temperature transfer curves of representative exfoliated monolayer and inert (blue) and ambient (red) devices are shown in Figures 2a and 2b, where we observe typical -type semiconducting behavior. Consistent with our theoretical expectations, ambient devices exhibit substantial -doping due to electron transfer from TMDs to the molecular adsorbates. This -doping is characterized by significant shifts in toward positive gate voltages with V () and 35 V () corresponding to an electron doping of \times 10^{12}$ () and 3.3 (). Figure 2c (MoS2) and 2d (WS2) show statistics summarizing distributions. We observe significant doping for where mean V (exfoliated) and 65 (CVD) corresponding to an electron -doping of 6.5 (exfoliated) and 4.7 \times 10^{12}$ (CVD). For , we find more modest doping with mean V (exfoliated) and 3 V (CVD) corresponding to an electron doping of 1.3 \times 10^{12}$ (exfoliated) and 2.2 \times 10^{11}$ (CVD).

Figure 2.
Figure 2 (PDF p. 6). Charge doping from oxygen. Transfer characteristics of a monolayer (a) MoS and (b) WS back-gate field-effect transistor at a drain voltage V. The device schematic is shown as an inset. The device channel lengths are 1 m. The dotted lines represent the linear extrapolation to obtain threshold voltages (). Boxplots present statistical distributions of the extracted threshold voltages for (c) MoS and (d) WS devices. Significant negative shifts in threshold voltages are observed for inert devices compared to their ambient counterpart, which indicate reduced O-driven -doping.
The observed difference in doping levels between MoS and WS is surprising. Our DFT calculations suggest negligible differences in , and defect densities are also expected to be lower in MoS crystals with better growth quality, which should present fewer sites for O chemisorption. A possible explanation is that more defective WS crystals tend to undergo dissociative oxidation from defect pairing.[35, 41] The kinetics for dissociative oxidation likely differ from chemisorption and may be material-dependent. Furthermore, differences in overall stoichiometry, purity, and crystallinity are expected between MoS and WS crystals, suggesting more complex O interaction mechanisms. However, the difference in doping levels observed between CVD and exfoliated materials is expected. Exfoliated devices represent the lowest defect density and ambient exposure levels where 2D flakes are prepared inside the gloveboxes. For commercially purchased CVD TMDs, the post-growth packaging and shipment process is beyond our control. CVD devices thus represent higher defect densities and a ‘medium’ ambient exposure level, translating to more significant -doping.
Minimizing ambient exposure to prevent unwanted -doping enables our devices to reach high carrier densities up to cm (MoS) and cm (WS). For context, our MoS carrier densities approach degenerate doping levels[48] and are comparable with intentional surface transfer or solid-state doping techniques (- cm).[25, 27, 48] As a result, inert devices can attain more than an order of magnitude larger on-state current and current on/off ratios than their ambient counterparts (Figure 2a, b). Our inert monolayer MoS FET reached A/m (SI Figure S8), exceeding even ultra-short channel ( nm) FETs developed by IMEC (250 A/m).[49] Current on/off ratios for inert devices () exceed ambient devices (-) by around an order of magnitude. Such significant -doping was previously only achieved by intentional doping techniques. However, such doping introduces foreign impurities that lead to impurity-related Coulomb scattering or band structure changes impacting electrical transport, lowering carrier mobilities and sheet conductance. Our DFT calculations confirm that the chemisorption of O molecules can lead to band structure changes that increase effective mass and lower carrier mobility (Extended Data Figure 9). In contrast, such effects are not expected for our approach, which minimizes ambient adsorbates to preserve the intrinsic cleanliness of 2D TMDs.

Extended Data Figure 9. The electron’s effective masses in defective monolayers (a) MoS and (b) WS, with various interaction configurations of O molecules. The dotted line presents a visual guide for the change in effective mass compared to the pristine case.
To investigate whether minimizing ambient adsorbates translates into superior electrical transport, we extract the carrier mobility and sheet conductance from two-terminal room-temperature transport measurements of 38 MoS (Figure 3a) and 24 WS (Figure 3b) few-layer and monolayer devices (details in Methods). Clear improvements are observed for inert (blue) compared to ambient (red) devices made from both exfoliated (diamond markers) and CVD (circle markers) materials. For exfoliated (CVD) MoS, we find that the mean mobility increases by a factor of 3 (3) to 86.516.7 cm/V s (25.75.8 cm/V s), while the mean sheet conductance increases by a factor of 9 (6) to 138.825.5 S/ (16.53.6 S/). Similar improvements are observed for exfoliated (CVD) WS, with mean mobility increasing by a factor of 10 (26) to 35.91.8 cm/V s (18.51.4 cm/V s), and the mean sheet conductance increases by a factor of 26 (10) to 23.61.9 (7.10.8 ) (details are in Extended Figure 10). At room (low) temperature, mobilities are typically phonon- (impurity-) scattering limited. Our measurements at 80 K show that mobility increases by a factor of 4 (3) to 470.80.2 (106.30.3 ) for exfoliated MoS (WS) devices. We measure similarly high mobility improvement by a factor of 3 (3) for monolayer CVD MoS (WS) devices. In contrast, the exfoliated ambient MoS device shows only a modest 19% improvement, suggesting that higher impurity levels from ambient exposure limit the low-temperature carrier mobility. All analyzed devices exhibit linear output behavior at both room and low temperatures, ruling out poor contacts as the cause of the different mobility trends (Extended Data Figure 11) between ambient and inert devices. Transfer length method measurements corroborate our device contact quality. We find that minimizing ambient exposure is also beneficial for contact quality. Inert devices show lower contact resistance = 102 k compared to ambient devices where = 1550 k, (Extended Data Figure 12). This lower can be understood by the decreased -doping of contact regions which decreases Schottky barrier heights.[66] The performance advantage of inert devices also extends to other figures of merit where we find consistent improvements in gate hysteresis and interface trap densities (Extended Data Figure 13).

Extended Data Figure 11. Extended Data Figure 11: Electrical data of inert MoS FETs. Output characteristics of a monolayer device measured at (a) 80 K and (b) 300 K. (c) Temperature-dependent transfer characteristics of a monolayer device showing mobility improvement with decreasing temperature due to quenching of phonon scattering. (d) Mobility as a function of temperature () showing phonon-limited behavior at higher temperatures.

Extended Data Figure 13. (a) Boxplot showing the distribution of threshold voltage hysteresis determined for inert and ambient devices at room temperature. The back gate voltage is swept from -30 to +70 V (forward sweep) and then back to -30 V (backward sweep) with a constant V. The hysteresis is then calculated from the difference of the threshold voltages determined from the forward and backward sweeps. (c) Boxplot showing the distribution of the interface trap density of inert and ambient devices calculated from the subthreshold slopes.

Figure 16 (PDF p. 7). Distribution of field-effect mobility () vs sheet conductance () for (a) MoS and (b) WS field-effect transistors (FETs) fabricated in inert (blue) and ambient (red) conditions. Mean mobility values show 2 orders of magnitude increase for inert devices compared to ambient devices. Similarly, sheet conductance is enhanced by 3 orders for MoS and 2 orders for WS. (c) When the temperature is decreased from 300 K to 80 K, significant enhancement is observed for inert devices compared to ambient devices, showing higher impurity-limited mobility for inert devices. Benchmarking FET performance against the existing literature for (d) exfoliated MoS and (e) CVD MoS. The red dashed line represents the theoretical limit of the room-temperature, phonon-limited mobility in 2D MoS. We benchmark the sheet conductance and back-gated field-effect mobility of FETs with comparable geometries on SiO. The data used for refs are provided as a table in the Supporting Information.[18, 20, 25, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60] (f) We also benchmark mobility as a function of FET channel thickness. We compare our devices with other state-of-the-art 2D FETs, traditional ultra-thin body (UTB) silicon PMOS and NMOS, III-V 2DEG MOSFETs, and IRDS roadmap targets.[2, 20, 22, 61, 62, 63, 64, 65]
Next, we benchmarked our device performance with other devices of similar architecture as reported in the literature. We summarize sheet conductance and mobility values of exfoliated 2D MoS (Figure 3d), CVD monolayer (Figure 3e) MoS, and WS (Extended Data Figure 14) devices. Our exfoliated mono- and bi-layer MoS FETs demonstrate record room-temperature mobility of and respectively, more than three times higher than previous reports of comparable device geometries. Likewise, we measure a record high mobility of for WS FET. The comparatively modest improvement for WS is likely due to crystal quality limitations such that transport is still defect-limited, whereas MoS growth processes are more mature. Notably, our MoS mobility marks a significant advance toward the theoretical expectations for phonon-limited transport in 2D TMDs (indicated by the horizontal red dashed line in Figure 3d). We have found that minimizing ambient exposure to molecular oxygen leads to superior devices. Notably, our fabrication strategy is also generally compatible with most reported contact, dielectric, and interface engineering techniques that can further improve 2D TMD device performance.

Extended Data Figure 14. Benchmarking FET performance against the existing literature for exfoliated and CVD WS. Solid (hollow) stars represent exfoliated (CVD) FETs. [70, 71, 72, 73, 74]
These measurements and benchmarking highlight the significant impact of adsorption at the device level. Next, we investigate the reversibility of adsorption processes. Based on our DFT calculated (Figure 1b), weakly interacting physisorption should be reversible through simple annealing or vacuum cycling treatments. However, the large of chemisorption poses a significant barrier toward molecular desorption and will require considerable thermal energies. While 2D TMDs can remain structurally intact up to 500 °C,[67] devices may not tolerate similarly high temperatures. In integrated circuit manufacturing, back-end-of-line (BEOL) compatibility is an essential consideration with a temperature limit of ~400 °C.[5] Thermal effects such as strain and atomic diffusion at the contacts, dielectrics, and interfaces can degrade device performance and reliability at even lower temperatures below 300 °C.[22] With these considerations, we perform ab initio molecular dynamics (AIMD) simulation to understand the desorption process at elevated temperatures. We simulate molecules physisorbed and chemisorbed to TMD surfaces with double (Figure 4a, Figure 4b). At a temperature of 500 K, 75% (0%) of physisorbed (chemisorbed) molecules can desorb from the TMD surface. Importantly, our simulations (Figure 4c) suggest that chemisorbed molecules are more likely to undergo dissociative oxidation than desorb from the TMD surface. This is consistent with our previous work, where XPS measurements confirmed the tendency for dissociative oxidation in defective 2D WS with double .[41] Next, we experimentally assess the reversibility of device performance after thermal annealing at 200 °C for 2h. Figure 4d shows the box plots of sheet conductance measured for inert, ambient, and annealed devices. Inert devices show the highest mean sheet conductance of µS/□ while ambient devices show the lowest mean sheet conductance of µS/□. After thermal annealing, the mean sheet conductance of ambient devices improves by ~50% to µS/□ but remains far below that of inert devices. These measurements indicate that while thermal annealing can remove weakly physisorbed molecules to boost device conductance, chemisorbed molecules cannot be removed, limiting device performance.

Figure 4. Reversibility of O chemisorption. Ab initio molecular dynamics (AIMD) simulation of (a) physisorbed and (b) chemisorbed O at 500 K. The simulations indicate that of physisorbed O molecules desorb at 500 K, while no chemisorbed O molecules desorb. (c) Gibbs free energy profile of the initial state (IS), transition state (TS), and final state (FS) of O dissociation and desorption on MoS with adjacent double sulfur vacancies. The calculations show that desorption becomes energetically unfavorable once the O molecules are chemisorbed. Instead, the dissociation of O molecules is the more energy-favorable pathway. (d) Sheet conductance of MoS FETs inert and ambient devices (calculated from Extended Figure 15). The ambient devices were subsequently annealed at 473 K in a 10% H/Ar forming gas environment for two hours. Annealing partially recovers the conductance attributed to the desorption of physisorbed O. However, chemisorbed O molecules restrict its complete recovery.
We have identified the detrimental effect of oxygen on 2D TMD devices that were previously thought to be ambient stable. Adopting an oxygen-free fabrication approach, we push our device performance toward the theoretical phonon-limited intrinsic mobility. Our work highlights that atomically thin materials require radical strategies beyond conventional nanofabrication. This provides guidance for the community, which should adopt practices to limit oxygen exposure to 2D materials. Our study advances the understanding of 2D TMDs that can likely be expanded to other atomically thin vdW systems. Thousands of 2D materials have been theoretically predicted since the first isolation of graphene,[68] but experimental research has been largely restricted to a handful of 2D materials that are more ambient stable. Our oxygen-free fabrication approach should accelerate the fundamental understanding of 2D materials and their translation for practical applications.
Methods
Materials
and bulk crystals obtained from 6Carbon Technology and HQ Graphene were used for exfoliation. Commercial monolayers of and grown by chemical vapor deposition (CVD) on sapphire substrates from 6Carbon Technology were employed for CVD-based devices. A 300 nm -coated doped Si wafer served as the device substrate and gate dielectric in back-gated field-effect transistor (FET) device architectures.
Sample preparation
All fabrication steps for inert devices were carried out inside a glovebox with and levels maintained typically below 1 ppm. Initially, fresh substrates were cleaned under purging, followed by baking at 100 °C for 10 minutes. Exfoliation of 2D layers was performed using "blue-Nitto" low-adhesive tape (@Nitto Denko Corp.). First, thicker layers were cleaved from parent crystals and exfoliated 5-6 times within the same tape to thin down the layers and cover the entire tape surface. Subsequently, the 2D layers were lightly pressed onto the cleaned substrates immediately after removal from the hot plate. Finally, the samples were inspected under an optical microscope to identify the desired thickness of 2D layers.
CVD-grown monolayers were transferred from sapphire to the /Si substrate using a polymer-assisted wet-transfer method. The CVD TMD films were delaminated from their growth substrates under ambient conditions and transferred inside the inert glovebox after vacuum cycling in the antechamber. We conducted hotplate annealing before the final film transfer to the device substrate for fabrication. These combined annealing and vacuum treatments aim to eliminate physisorbed adsorbates.
Nanolithography
Contact patterning was conducted in an inert glovebox using the thermal scanning probe lithography (-SPL) system (Nanofrazor Scholar, Heidelberg Instruments). A bilayer -SPL resist consisting of 30/110 nm PPA/PMMA-MA was spin-coated, followed by baking at 140 °C for 90 s and 110 °C for 120 s, respectively. Details for the patterning parameters can be found in our previous study.[69] After thermal patterning, the exposed PMMA-MA layer was etched using 20 °C ethanol for 15 seconds. The In/Au metals (5/35 nm) were deposited by thermal evaporation at ~ bar with a substrate temperature of 0 °C.
To rule out the effect of the nanolithography method used in device fabrication on the observed improvement, we also define the exact width of metal contact by electron-beam lithography (EBL). Samples were spin-coated with PMMA A5 electron resist at 4000 rpm for 90 seconds, followed by hard baking at 180 °C for 2 minutes before removal from the glovebox. The resist-coated 2D layers were subsequently transferred to the EBL system. Next, the contacts were written by EBL, and In/Au metal (5/30 nm) was deposited via thermal evaporation. All the intermediate processes, such as sample development in MIBK:IPA (1:3), metal lift-off using 1165 solvent (Microposit), and further resist coating for the next step of lithography, were performed inside gloveboxes using interconnected sample-transfer lines without exposing the 2D materials to air at any stage.
Finally, the electrical connections and bond pads were defined by EBL, followed by 5/55 nm Cr/Au metal evaporation.
Ambient Environment Fabrication
All process steps and parameters, from exfoliation to resist coating, development, and liftoff, were explicitly performed in an air-atmosphere environment outside the glovebox.
Electrical Characterizations
The electrical measurements were carried out using a cryostat probe station under a base pressure of mbar. The system was cooled down to 80 K using liquid nitrogen. Keithley 2450 source meters and a Lakeshore temperature controller were utilized for the measurements.
Computational details
A monolayer TMD was considered with a single chalcogen vacancy as a defect site, and the influence of different defect densities was investigated. The amount of charge modulation in TMDs due to molecular adsorption is ascribed by . The orientation of molecules determines the actual and values. For example, side-on O-O configurations exhibiting stronger adsorption ( eV) and higher electron transfer ( e) compared to the end-on arrangements ( eV and e). The equilibrium band structure and atomic projected density of states (APDOS) are illustrated in ESI. Details about the DFT calculations and ab initio molecular dynamics (AIMD) simulations can be found in the Supporting Information.
Acknowledgments
We acknowledge the funding support from the Agency for Science, Technology and Research (ASTAR) (#21709, C230917006, and C230917007). C.S.L. acknowledges the support from ASTAR under its MTC IRG grant No. M23M6c0103 and MTC YIRG grant no. M21K3c0124. K.E.J.G. acknowledges support from a Singapore National Research Foundation Grant (CRP21-2018-0001). D.H. acknowledges funding support from A*STAR Project C222812022 and MTC YIRG M22K3c0105. Y.S.A. is supported by the Singapore Ministry of Education Academic Research Fund Tier 2 (Award No. MOE-T2EP50221-0019). Y.S.A. and S.W. are supported by the SUTD-ZJU IDEA Thematic Research Grant under the award number SUTD-ZJU (TR) 202203.
Contributions
S.M. and C.S.L. designed the experiment and carried out the transport measurements and result analysis. S.W. and Y.S.A. performed DFT and AIMD simulations. S.M., D.V., T.T.D, R.L. and Y.W.T contributed to the device fabrication and measurements. Y.T., S.M. and C.S.L. contributed to the python coding. F.B. and T.M. performed XPS studies. I.A.V., D.H., A.M., J.W.J, S.D. and K.E.J.G contributed to result interpretation and discussion. K.E.J.G., Y.S.A. and C.S.L. supervised the experiments and calculations. C.S.L. conceived the study. C.S.L. and S.M. wrote the manuscript with input from all authors.
Competing Interests
The authors declare no competing interests.
Extended Data

Extended Data Figure 3. Charge density difference after molecule adsorption on (a) pristine MoS, and (b-d) defective MoS sites. The energetically favorable arrangements for (b) physisorption and (c & d) chemisorption are shown here. Band structure and atomic projected density of states (APDOS) for (e) single chalcogen vacancy () in MoS basal plane, and MoS with adsorbed molecules, such as (f) physiosorbed , and (g-h) chemisorbed .

Extended Data Figure 4. Charge density difference after molecule adsorption on defective WS sites. The energetically favorable arrangements for (b) physisorption and (c & d) chemisorption are shown here. Band structure and atomic projected density of states (APDOS) for (e) single chalcogen vacancy () in WS basal plane, and WS with adsorbed molecules, such as (f) physiosorbed , and (g-h) chemisorbed .

Extended Data Figure 7. Dual sweep transfer characteristics of inert and ambient (a-b) bilayer MoS and (c-d) WS FETs.

Extended Data Figure 8. Electrical characteristics of a typical inert monolayer MoS FET. The transfer curves (a) and output curves (b) are presented for various drain and gate biases, respectively.

Extended Data Figure 10. Transfer characteristics at 300 K and 80 K for inert FETs. (a) exfoliated MoS, (b) CVD MoS, (c) exfoliated WS and (d) CVD WS FETs. Mobility is calculated for each case, and the summarized plot is presented in Figure 3c of the main text.
Extended Data Figure 12: (a) Transfer characteristics comparing inert and ambient few-layer MoS transistors of different channel lengths in a transfer length method (TLM) geometry measured at V. Extraction of TLM contact resistance () for (b) inert and (c) ambient devices at a carrier density of . Inert devices show improvement in the TLM mobility and lower contact resistance .

Figure 12.

Figure 15. Transfer characteristics and threshold voltages of inert and ambient devices. Ambient devices are subsequently annealed in a 10% Ar/H forming gas at 200 °C for 2 h. While annealing can partially restore the current and threshold voltage, complete reversal is not possible.
References
- [1]Moore, G. E. Cramming more components onto integrated circuits. Proc. IEEE 86, 82–85 (1998) 10.1109/JPROC.1998.658762.DOI
- [2]Uchida, K. et al. Experimental study on carrier transport mechanism in ultrathin-body SOI nand p-MOSFETs with SOI thickness less than 5 nm. in Digest. International Electron Devices Meeting, 47–50 (IEEE, 2002). doi:10.1109/IEDM.2002.1175776 10.1109/IEDM.2002.1175776.DOI
- [3]Zhu, K. et al. The development of integrated circuits based on two-dimensional materials. Nat. Electron. 4, 775–785 (2021) 10.1038/s41928-021-00672-z.DOI
- [4]Kim, K. S. et al. The future of two-dimensional semiconductors beyond Moore’s law. Nat. Nanotechnol. (2024) doi:10.1038/s41565-024-01695-1 10.1038/s41565-024-01695-1.DOI
- [5]Das, S. et al. Transistors based on two-dimensional materials for future integrated circuits. Nat. Electron. 4, 786–799 (2021) 10.1038/s41928-021-00670-1.DOI
- [6]Cao, W. et al. The future transistors. Nature 620, 501–515 (2023) 10.1038/s41586-023-06145-x.DOI
- [7]Liu, Y. et al. Promises and prospects of two-dimensional transistors. Nature 591, 43–53 (2021) 10.1038/s41586-021-03339-z.DOI
- [8]Jayachandran, D. et al. Three-dimensional integration of two-dimensional field-effect transistors. Nature 625, 276–281 (2024) 10.1038/s41586-023-06860-5.DOI
- [9]Wu, P., Zhang, T., Zhu, J., Palacios, T. & Kong, J. 2D materials for logic device scaling. Nat. Mater. 23, 23–25 (2024) 10.1038/s41563-023-01715-w.DOI
- [11]Sohier, T., Campi, D., Marzari, N. & Gibertini, M. Mobility of two-dimensional materials from first principles in an accurate and automated framework. Phys. Rev. Mater. 2, 1–21 (2018) 10.1103/PhysRevMaterials.2.114010.DOI
- [12]Zhang, C., Wang, R., Mishra, H. & Liu, Y. Two-Dimensional Semiconductors with High Intrinsic Carrier Mobility at Room Temperature. Phys. Rev. Lett. 130, 87001 (2023) 10.1103/PhysRevLett.130.087001.DOI
- [13]Zhang, C. & Liu, Y. Phonon-limited transport of two-dimensional semiconductors: Quadrupole scattering and free carrier screening. Phys. Rev. B 106, 1–9 (2022) 10.1103/PhysRevB.106.115423.DOI
- [14]Wang, Y., Sarkar, S., Yan, H. & Chhowalla, M. Critical challenges in the development of electronics based on two-dimensional transition metal dichalcogenides. Nat. Electron. (2024) doi:10.1038/s41928-024-01210-3 10.1038/s41928-024-01210-3.DOI
- [15]Amontree, J. et al. Reproducible graphene synthesis by oxygen-free chemical vapour deposition. Nature 630, 636–642 (2024) 10.1038/s41586-024-07454-5.DOI
- [16]Zhang, T., Wang, J., Wu, P., Lu, A.-Y. & Kong, J. Vapour-phase deposition of two-dimensional layered chalcogenides. Nat. Rev. Mater. 8, 799–821 (2023) 10.1038/s41578-023-00609-2.DOI
- [17]Zhu, J. et al. Low-thermal-budget synthesis of monolayer molybdenum disulfide for silicon back-end-of-line integration on a 200 mm platform. Nat. Nanotechnol. 18, 456–463 (2023) 10.1038/s41565-023-01375-6.DOI
- [18]Wang, Y. et al. Van der Waals contacts between three-dimensional metals and two-dimensional semiconductors. Nature 568, 70–74 (2019) 10.1038/s41586-019-1052-3.DOI
- [19]Lau, C. S. et al. Quantum Transport in Two-Dimensional WS₂ with High-Efficiency Carrier Injection through Indium Alloy Contacts. ACS Nano 14, 13700–13708 (2020) 10.1021/acsnano.0c05915.DOI
- [20]Shen, P.-C. C. et al. Ultralow contact resistance between semimetal and monolayer semiconductors. Nature 593, 211–217 (2021) 10.1038/s41586-021-03472-9.DOI
- [21]Wang, Y. et al. P-type electrical contacts for 2D transition-metal dichalcogenides. Nature 610, 61–66 (2022) 10.1038/s41586-022-05134-w.DOI
- [22]Li, W. et al. Approaching the quantum limit in two-dimensional semiconductor contacts. Nature 613, 274–279 (2023) 10.1038/s41586-022-05431-4.DOI
- [23]Su, T. et al. Semimetal contacts to monolayer semiconductor: weak metallization as an effective mechanism to Schottky barrier lowering. J. Phys. D: Appl. Phys. 56, 234001 (2023) 10.1088/1361-6463/acc53f.DOI
- [24]Jiang, J. et al. Yttrium-doping-induced metallization of molybdenum disulfide for ohmic contacts in two-dimensional transistors. Nat. Electron. (2024) doi:10.1038/s41928-024-01176-2 10.1038/s41928-024-01176-2.DOI
- [25]Kiriya, D., Tosun, M., Zhao, P., Kang, J. S. & Javey, A. Air-Stable Surface Charge Transfer Doping of MoS₂ by Benzyl Viologen. J. Am. Chem. Soc. 136, 7853–7856 (2014) 10.1021/ja5033327.DOI
- [26]Gao, H. et al. Tuning Electrical Conductance of MoS₂ Monolayers through Substitutional Doping. Nano Lett. 20, 4095–4101 (2020) 10.1021/acs.nanolett.9b05247.DOI
- [27]McClellan, C. J., Yalon, E., Smithe, K. K. H., Suryavanshi, S. V. & Pop, E. High Current Density in Monolayer MoS₂ Doped by AlOₓ. ACS Nano 15, 1587–1596 (2021) 10.1021/acsnano.0c09078.DOI
- [28]Choi, M. S. et al. High carrier mobility in graphene doped using a monolayer of tungsten oxyselenide. Nat. Electron. 4, 731–739 (2021) 10.1038/s41928-021-00657-y.DOI
- [29]Ma, N. & Jena, D. Charge scattering and mobility in atomically thin semiconductors. Phys. Rev. X 4, 1–9 (2014) 10.1103/PhysRevX.4.011043.DOI
- [30]Lau, C. S. et al. Gate-Defined Quantum Confinement in CVD 2D WS₂. Adv. Mater. 34, 2103907 (2022) 10.1002/adma.202103907.DOI
- [31]Zhang, Y. et al. Liquid-Metal-Printed Ultrathin Oxides for Atomically Smooth 2D Material Heterostructures. ACS Nano 17, 7929–7939 (2023) 10.1021/acsnano.3c02128.DOI
- [32]Lau, C. S. et al. Dielectrics for Two-Dimensional Transition-Metal Dichalcogenide Applications. ACS Nano 17, 9870–9905 (2023) 10.1021/acsnano.3c03455.DOI
- [33]Knobloch, T. et al. Improving stability in two-dimensional transistors with amorphous gate oxides by Fermi-level tuning. Nat. Electron. 5, 356–366 (2022) 10.1038/s41928-022-00768-0.DOI
- [34]Yang, S., Jiang, C. & Wei, S. Gas sensing in 2D materials. Appl. Phys. Rev. 4, 21304 (2017) 10.1063/1.4983310.DOI
- [35]Liu, H., Han, N. & Zhao, J. Atomistic insight into the oxidation of monolayer transition metal dichalcogenides: From structures to electronic properties. RSC Adv. 5, 17572–17581 (2015) 10.1039/c4ra17320a.DOI
- [36]Hu, Z. et al. Two-dimensional transition metal dichalcogenides: Interface and defect engineering. Chem. Soc. Rev. 47, 3100–3128 (2018) 10.1039/c8cs00024g.DOI
- [37]Chhowalla, M. et al. The chemistry of two-dimensional layered transition metal dichalcogenide nanosheets. Nat. Chem. 5, 263–275 (2013) 10.1038/nchem.1589.DOI
- [38]Li, Q., Zhou, Q., Shi, L., Chen, Q. & Wang, J. Recent advances in oxidation and degradation mechanisms of ultrathin 2D materials under ambient conditions and their passivation strategies. J. Mater. Chem. A 7, 4291–4312 (2019) 10.1039/c8ta10306b.DOI
- [39]Gao, J. et al. Aging of Transition Metal Dichalcogenide Monolayers. ACS Nano 10, 2628–2635 (2016) 10.1021/acsnano.5b07677.DOI
- [40]Rao, R., Islam, A. E., Campbell, P. M., Vogel, E. M. & Maruyama, B. In situ thermal oxidation kinetics in few layer MoS₂. 2D Mater. 4, (2017) 10.1088/2053-1583/aa6532.DOI
- [41]Bussolotti, F. et al. Role of S-Vacancy Concentration in Air Oxidation of WS₂ Single Crystals. ACS Nano (2024) doi:10.1021/acsnano.3c10389 10.1021/acsnano.3c10389.DOI
- [42]Atkin, P. et al. Laser exposure induced alteration of WS₂ monolayers in the presence of ambient moisture. 2D Mater. 5, 015013 (2017) 10.1088/2053-1583/aa91b8.DOI
- [43]Kotsakidis, J. C. et al. Oxidation of Monolayer WS₂ in Ambient Is a Photoinduced Process. Nano Lett. 19, 5205–5215 (2019) 10.1021/acs.nanolett.9b01599.DOI
- [44]Wang, S., Zhao, W., Giustiniano, F. & Eda, G. Effect of oxygen and ozone on p-type doping of ultra-thin WSe₂ and MoSe₂ field effect transistors. Phys. Chem. Chem. Phys. 18, 4304–4309 (2016) 10.1039/c5cp07194a.DOI
- [45]Schmidt, H., Giustiniano, F. & Eda, G. Electronic transport properties of transition metal dichalcogenide field-effect devices: surface and interface effects. Chem. Soc. Rev. 44, 7715–7736 (2015) 10.1039/c5cs00275c.DOI
- [46]Li, W. Electrical transport limited by electron-phonon coupling from Boltzmann transport equation: An ab initio study of Si, Al, and MoS₂. Phys. Rev. B - Condens. Matter Mater. Phys. 92, 1–10 (2015) 10.1103/PhysRevB.92.075405.
- [47]Aryeetey, F., Ignatova, T. & Aravamudhan, S. Quantification of defects engineered in single layer MoS₂. RSC Adv. 10, 22996–23001 (2020) 10.1039/D0RA03372C.DOI
- [48]Fang, H. et al. Degenerate n-Doping of Few-Layer Transition Metal Dichalcogenides by Potassium. Nano Lett. 13, 1991–1995 (2013) 10.1021/nl400044m.DOI
- [49]Smets, Q. et al. Ultra-scaled MOCVD MoS₂ MOSFETs with 42nm contact pitch and 250μA/μm drain current. Tech. Dig. - Int. Electron Devices Meet. IEDM 2019-Decem, 2019–2022 (2019) 10.1109/IEDM19573.2019.8993650.DOI
- [50]Liu, H., Neal, A. T. & Ye, P. D. Channel length scaling of MoS₂ MOSFETs. ACS Nano 6, 8563–8569 (2012) 10.1021/nn303513c.DOI
- [51]Kappera, R. et al. Phase-engineered low-resistance contacts for ultrathin MoS₂ transistors. Nat. Mater. 13, 1128–1134 (2014) 10.1038/nmat4080.DOI
- [52]Lockhart De La Rosa, C. J. et al. Highly efficient and stable MoS₂ FETs with reversible n-doping using a dehydrated poly(vinyl-alcohol) coating. Nanoscale 9, 258–265 (2017) 10.1039/c6nr06980k.DOI
- [53]Abraham, M. & Mohney, S. E. Annealed Ag contacts to MoS₂ field-effect transistors. J. Appl. Phys. 122, (2017) 10.1063/1.4991961.DOI
- [54]Park, Y., Baac, H. W., Heo, J. & Yoo, G. Thermally activated trap charges responsible for hysteresis in multilayer MoS₂ field-effect transistors. Appl. Phys. Lett. 108, (2016) 10.1063/1.4942406.DOI
- [55]Wang, X., Feng, H., Wu, Y. & Jiao, L. Controlled synthesis of highly crystalline MoS₂ flakes by chemical vapor deposition. J. Am. Chem. Soc. 135, 5304–5307 (2013) 10.1021/ja4013485.DOI
- [56]Smithe, K. K. H., Suryavanshi, S. V., Muñoz Rojo, M., Tedjarati, A. D. & Pop, E. Low Variability in Synthetic Monolayer MoS₂ Devices. ACS Nano 11, 8456–8463 (2017) 10.1021/acsnano.7b04100.DOI
- [57]Smithe, K. K. H., English, C. D., Suryavanshi, S. V. & Pop, E. Intrinsic electrical transport and performance projections of synthetic monolayer MoS₂ devices. 2D Mater. 4, 1–8 (2017) 10.1088/2053-1583/4/1/011009.DOI
- [58]Xiao, J. et al. Record-high saturation current in end-bond contacted monolayer MoS₂ transistors. Nano Res. 15, 475–481 (2022) 10.1007/s12274-021-3504-y.DOI
- [59]English, C. D., Shine, G., Dorgan, V. E., Saraswat, K. C. & Pop, E. Improved contacts to MoS₂ transistors by ultra-high vacuum metal deposition. Nano Lett. 16, 3824–3830 (2016) 10.1021/acs.nanolett.6b01309.DOI
- [60]Rai, A. et al. Air Stable Doping and Intrinsic Mobility Enhancement in Monolayer Molybdenum Disulfide by Amorphous Titanium Suboxide Encapsulation. Nano Lett. 15, 4329–4336 (2015) 10.1021/acs.nanolett.5b00314.DOI
- [61]Yokoyama, M. et al. Extremely-thin-body InGaAs-On-Insulator MOSFETs on Si fabricated by direct wafer bonding. Tech. Dig. - Int. Electron Devices Meet. IEDM 46–49 (2010) doi:10.1109/IEDM.2010.5703286 10.1109/IEDM.2010.5703286.DOI
- [62]Alian, A. et al. Impact of the channel thickness on the performance of ultrathin InGaAs channel MOSFET devices. Tech. Dig. - Int. Electron Devices Meet. IEDM 437–440 (2013) doi:10.1109/IEDM.2013.6724644 10.1109/IEDM.2013.6724644.DOI
- [63]Sebastian, A., Pendurthi, R., Choudhury, T. H., Redwing, J. M. & Das, S. Benchmarking monolayer MoS₂ and WS₂ field-effect transistors. Nat. Commun. 12, 1–12 (2021) 10.1038/s41467-020-20732-w.DOI
- [64]Zou, X. et al. Interface engineering for high-performance top-gated MoS₂ field-effect transistors. Adv. Mater. 26, 6255–6261 (2014) 10.1002/adma.201402008.DOI
- [65]Schmidt, M. et al. Mobility extraction in SOI MOSFETs with sub 1nm body thickness. Solid. State. Electron. 53, 1246–1251 (2009) 10.1016/j.sse.2009.09.017.DOI
- [66]Jiang, J., Xu, L., Qiu, C. & Peng, L. M. Ballistic two-dimensional InSe transistors. Nature 616, 470–475 (2023) 10.1038/s41586-023-05819-w.DOI
- [67]Bandaru, N. et al. Effect of Pressure and Temperature on Structural Stability of MoS₂. J. Phys. Chem. C 118, 3230–3235 (2014) 10.1021/jp410167k.DOI
- [68]Mounet, N. et al. Two-dimensional materials from high-throughput computational exfoliation of experimentally known compounds. Nat. Nanotechnol. 13, 246–252 (2018) 10.1038/s41565-017-0035-5.DOI
- [69]Talha-Dean, T. et al. Nanowiring van der Waals Heterostructures toward Electrically Controlled Quantum Dots. ACS Appl. Mater. Interfaces 16, 31738–31746 (2024) 10.1021/acsami.4c03639.DOI
- [70]Kaushik, V. et al. Charge Transport in 2D MoS₂, WS₂, and MoS₂-WS₂ Heterojunction-Based Field-Effect Transistors: Role of Ambipolarity. J. Phys. Chem. C 124, 23368–23379 (2020) 10.1021/acs.jpcc.0c05651.DOI
- [71]Yun, S. J. et al. Synthesis of centimeter-scale monolayer tungsten disulfide film on gold foils. ACS Nano 9, 5510–5519 (2015) 10.1021/acsnano.5b01529.DOI
- [72]Chen, J. et al. Synthesis of Wafer-Scale Monolayer WS₂ Crystals toward the Application in Integrated Electronic Devices. ACS Appl. Mater. Interfaces 11, 19381–19387 (2019) 10.1021/acsami.9b04791.DOI
- [73]Lan, C., Li, C., Yin, Y. & Liu, Y. Large-area synthesis of monolayer WS₂ and its ambient-sensitive photo-detecting performance. Nanoscale 7, 5974–5980 (2015) 10.1039/c5nr01205h.DOI
- [74]Aji, A. S., Solís-Fernández, P., Ji, H. G., Fukuda, K. & Ago, H. High Mobility WS₂ Transistors Realized by Multilayer Graphene Electrodes and Application to High Responsivity Flexible Photodetectors. Adv. Funct. Mater. 27, 1703448 (2017) 10.1002/adfm.201703448.DOI