Barrier‐Assisted Plasma Doping for Spatially Selective Resistance Engineering in MoS₂ Transistors

ABSTRACT

Atomically thin transition‐metal dichalcogenides (TMDs) are promising channel materials for low‐power logic. However, the absence of scalable and region‐selective doping techniques leads to excessive local resistances that hinder the technological readiness of 2D transistors. Here, a barrier‐assisted NH₃ plasma process is demonstrated that enables degenerate n‐type doping of monolayer MoS₂ while preserving its crystallinity. The ultrathin pV3D3/Al₂O₃ dielectric stack not only blocks plasma‐induced damage but also functions as a chemical filter that permits NH_x radicals to diffuse through. Through this doping process, an electron density of 4.3 × 10¹³ cm⁻² is achieved, yielding a contact resistance of 1.45 kΩ·µm. Density functional calculations show that NH₂ radicals adsorbed on the pristine MoS₂ surface are the main source of n‐type doping while NH radicals can heal S‐vacancy defects. Leveraging the spatial selectivity of this approach, mobility and on‐current are enhanced by 5.8‐fold with negligible threshold‐voltage shift. Extension‐region activation further suppresses series resistance, increasing the on‐current by 260‐fold (V _DS = 0.05 V) while maintaining enhancement‐mode operation. These findings establish barrier‐assisted NH₃ plasma doping as a promising approach for enabling high‐performance n‐type 2D transistors and advancing future energy‐efficient 2D CMOS technology.

Barrier‐assisted NH₃ plasma doping enables damage‐free, spatially selective carrier modulation in monolayer MoS₂ transistors. An ultrathin pV3D3/Al₂O₃ protective‐dielectric stack acts as a chemical filter that allows NH_x radicals to reach the MoS₂ surface while blocking plasma damage. This process achieves a high electron concentration of 4.3 × 10¹³ cm⁻², enabling effective resistance engineering in 2D transistors.

1. Introduction

Two‐dimensional (2D) transition metal dichalcogenides (TMDs), such as MoS₂ have attracted considerable interest as channel materials for next‐generation electronics due to their atomic‐scale thickness, excellent electrostatic gate control, and resistance to short‐channel effects [1, 2, 3]. These attributes make 2D FETs attractive for both device downscaling and advanced integration schemes, including monolithic three‐dimensional (3D) architectures [4]. Despite these advantages, the practical performance of 2D transistors remains limited by large contact and series resistances in the extension regions, which critically degrade current drivability and switching speed. A primary factor underlying high contact resistance is strong Fermi level pinning at the metal‐semiconductor interface, resulting in substantial Schottky barrier heights that hinder efficient carrier injection [5, 6, 7]. In underlapped top‐gate devices, the intrinsically low carrier density in the extension regions becomes a dominant source of series resistance, since these areas are not electrostatically modulated by the gate [8, 9]. Therefore, an additional doping process is indispensable for enabling low resistance, high‐performance switching in 2D transistors. However, unlike silicon, 2D materials cannot rely on conventional ion implantation or diffusion, making it difficult to achieve scalable and selective doping for complementary metal‐oxide‐semiconductor (CMOS) integration [10, 11, 12]. As a result, alternative approaches molecular, dielectric‐assisted, and plasma‐based doping strategies have been pursued, yet each suffers from key limitations in selectivity and scalability [13, 14, 15, 16]. Surface molecular doping can achieve high carrier densities without structural degradation; however, it generally lacks spatial selectivity and uniformity over large areas, limiting its practical applicability [17, 18]. Dielectric‐assisted doping using defective oxides or nitrides offers improved stability, but the additional dielectric layers undesirably increase overall device thickness and introduce parasitic capacitance, potentially degrading switching performance and hindering further scaling [19, 20]. Plasma‐based doping, which in principle enables scalable and spatially selective carrier modulation, has been widely investigated as an effective approach for tuning the carrier densities in 2D semiconductors [21]. Among various plasma‐based methods, NH₃ plasma can serve as an effective n‐doping approach by supplying –NH_x radicals that can chemically bind to the MoS₂ surface and act as electron donors [22]. However, direct exposure of MoS₂ to plasma often induces permanent structural damage, leading to severely degraded device performance [23, 24, 25, 26]. To address these limitations, we propose a barrier‐assisted NH₃ plasma doping process that enables high‐density, spatially confined n‐type doping of MoS₂ without compromising crystallinity. The critical challenge is to introduce desired dopant species, such as –NH₂, onto the MoS₂ surface without causing lattice damage. If the barrier is too thick, it blocks dopant species from reaching the underlying MoS₂ and suppresses the intended doping, whereas if it is too thin, uniform barrier deposition cannot be achieved, leaving pinholes and incomplete protection.

Our approach employs a bilayer protective‐dielectric stack consisting of polymeric poly(1,3,5‐trimethyl‐1,3,5‐trivinyl cyclotrisiloxane) (pV3D3) grown by initiated chemical vapor deposition (iCVD) and of ALD‐grown Al₂O₃. The pV3D3 layer ensures uniform dielectric coverage and suppresses interfacial traps, whereas the ultrathin Al₂O₃ overlayer protects the underlying lattice from plasma‐induced damage. This unique bilayer design permits chemical doping through the barrier while preventing plasma‐induced etching, effectively acting as a precise chemical filter that both preserves the MoS₂ interface quality and enables controlled, damage‐free permeation of NH_x species essential for doping. Using this method, we demonstrate degenerately n‐doped MoS₂ with carrier densities approaching 4.3 × 10¹³ cm⁻². First‐principles calculations reveal that, while NH₂ radicals adsorbed on the MoS₂ surface act as the primary source of n‐type doping, NH radicals concurrently passivate S vacancy defects. This dual‐functional mechanism provides a theoretical basis for the experimentally observed transition from defect‐dominated hopping transport in untreated MoS₂ to high‐quality band‐like transport in the Al₂O₃‐protected pV3D3‐MoS₂. We further exploit the spatial selectivity of the process to target doping at the channel, contact, and extension regions in back‐gated, overlapped top‐gated, and underlapped top‐gated architectures. Selective contact doping substantially lowers the effective Schottky barrier height (from 226.7 to 73.7 meV), enabling improved carrier injection and a 5.8‐fold mobility enhancement without threshold‐voltage shift. Extension‐region doping suppresses parasitic series resistance in underlapped devices, yielding a 260‐fold on‐current increase at low drain bias (0.05 V) while preserving enhancement‐mode operation. These results demonstrate that barrier‐assisted plasma doping enables spatially selective resistance engineering in 2D transistors, providing a scalable pathway toward high‐performance and energy‐efficient CMOS integration.

2. Results

2.1. Mechanism of Barrier‐Assisted NH₃ Plasma Doping in MoS₂

To achieve a uniform dielectric stack on MoS₂, we first employed the iCVD process to deposit an ultrathin pV3D3 layer. The pV3D3 layer promotes uniform, pinhole‐free coverage on the chemically inert MoS₂ surface and reduces interfacial trap states, thereby improving both dielectric nucleation and interface quality [27]. However, a single pV3D3 layer does not provide sufficient protection against plasma‐induced damage to MoS₂ (Figure S1); therefore, an additional ultrathin Al₂O₃ overlayer is incorporated to reinforce the barrier functionality. This Al₂O₃ layer suppresses plasma‐induced etching of the pV3D3 while simultaneously allowing controlled diffusion of reactive radicals through the barrier [28]. In addition, this bilayer scheme yields superior surface roughness and film uniformity compared to a single Al₂O₃ layer (Figure S2). Consequently, the bilayer protective‐dielectric (PD) stack was formed by combining 2 nm of iCVD‐grown pV3D3 with 2 nm of ALD‐grown Al₂O₃. This stack serves a dual role: physical protection from plasma‐induced damage while enabling chemical access of reactive NH₃‐derived radicals to the underlying MoS₂. (Figure 1a)

FIGURE 1.

Mechanism of protective‐dielectric‐assisted NH₃ plasma doping in monolayer MoS₂. (a) Schematic illustration of NH₃ plasma exposure on monolayer MoS₂ without (top) and with (bottom) an ultrathin 2 nm Al₂O₃/2 nm pV3D3 protective‐dielectric (PD). (b) Optical microscopy (OM) images of MoS₂ and PD‐MoS₂ after NH₃ plasma. (c–e) Raman, photoluminescence (PL), and N 1s X‐ray photoelectron spectroscopy (XPS) characterizations of MoS₂ before (black) and after NH₃ plasma treatment through the PD (red). (f) Normalized XPS depth profile showing N 1s localized at the MoS₂ surface above S 2p/Mo 3d signals.

To validate the efficacy of this barrier design, we compared pristine MoS₂ and PD stack‐deposited MoS₂ (PD‐MoS₂) subjected to identical NH₃ plasma treatments. Figure 1b presents optical microscopy images of MoS₂ after plasma treatment. Even with mild plasma exposure, pristine MoS₂ suffers from pronounced etching and structural degradation. The resulting loss of crystallinity is unambiguously confirmed by Raman spectroscopy (Figure S3). In contrast, PD‐MoS₂ retains a well‐defined morphology even after 240 min of plasma treatment, and the presence of sharp Raman peaks further confirms that its crystallinity is well‐preserved, indicating that our dielectric stack effectively suppresses plasma‐induced damage.

To confirm that doping proceeds despite the physical barrier, we performed Raman spectroscopy as shown in Figure 1c. The E_2g mode, corresponding to the in‐plane vibrational mode, is primarily sensitive to strain, whereas the A_1g mode, associated with the out‐of‐plane vibrational mode, is predominantly affected by doping [29, 30]. Raman spectra of NH₃ plasma‐treated MoS₂ (NP‐MoS₂) exhibit a distinct 2.24 cm⁻¹ red shift of the A_1g peak, with negligible shift in the E_2g peak. These changes indicate enhanced electron density without inducing strain or major structural distortion, suggesting that doping occurs without introducing unnecessary structural perturbations [31, 32]. Consistent with these findings, the photoluminescence (PL) spectra in Figure 1d reveal that, after NH₃ plasma treatment, the main emission peak (∼1.8 eV) undergoes a clear red shift accompanied by a pronounced reduction in intensity, providing strong evidence of substantial n‐type doping [33, 34, 35]. These optical characterizations collectively demonstrate that electron density in MoS₂ can be effectively increased while preserving crystallinity, thereby validating the effectiveness of the PD stack.

To investigate the origin of the observed n‐type doping, we employed X‐ray photoelectron spectroscopy (XPS) to analyze the surface states of MoS₂ before and after NH₃ plasma treatment, as shown in Figure 1e. The most prominent difference is the emergence of a new peak in the N 1s region. PD‐MoS₂ shows only the Mo 3p signal in the N 1s range (∼395.5 eV), whereas NP‐MoS₂ exhibits a prominent new peak near 399.5 eV, characteristic of –NH_x species such as –NH and –NH₂ [36, 37]. Concurrently, the Mo 3d spectrum of NP‐MoS₂ exhibits a clear blue shift, indicative of increased electron density and the onset of n‐type doping (Figure S4). This observation suggests that the newly detected –NH_x peak in the N 1s spectrum is likely the primary origin of this n‐doping behavior. To pinpoint the location of the newly formed –NH_x bonds and gain insight into the doping mechanism, we performed XPS depth profiling. The analysis revealed that the N 1s peak is predominantly localized on the MoS₂ surface, followed sequentially by the S 2p and Mo 3d peaks in the underlying layers (Figure 1f). This indicates that –NH_x species selectively bind to the outermost atomic layer of MoS₂ while preserving the underlying lattice in an undamaged state. EDS mapping further verifies that, despite the presence of the PD stack, nitrogen atoms are successfully localized on the MoS₂ surface (Figure S5). Collectively, these results demonstrate that NH₃ plasma treatment with a PD stack enables NH_x radicals to penetrate the dielectric layer and bind to the MoS₂ surface while preserving the integrity of the underlying lattice, thereby enabling controlled electron doping without structural degradation. Furthermore, the clear polarity conversion from p‐type to n‐type transport observed in a monolayer WSe₂ FET after NH₃ plasma treatment suggests that this barrier‐assisted doping strategy may be extendable beyond MoS₂ to other TMD semiconductors (Figure S6).

2.2. Dual Role of NH_x Radicals in Doping and Defect Passivation of MoS₂

To elucidate the microscopic origin of the n‐type doping under NH₃ plasma treatment of PD‐MoS₂, we performed density functional theory (DFT) calculations to analyze the energetics and electronic structure of NH_x‐adsorbed MoS₂ (NH_x/MoS₂). A 5 × 5 unit cell model of MoS₂ was employed, in which adsorption of a single NH₂ radical (see below) corresponds to a doping concentration of 4.7 × 10¹³ cm⁻², comparable to the experimentally observed doping level (Figure S7) [38, 39, 40]. We first examined the formation energies of ${\mathrm{NH}}_x^{\ast }$ (x  = 0, 1, 2) species on a pristine MoS₂ surface generated from N and H radicals that permeate through the PD stack. The formation energy was defined as [38–40]

$$E_{\mathrm{form}}=E_{\mathrm{N}{\mathrm{H}}_x/\mathrm{Mo}{\mathrm{S}}_2}-\left(E_{\mathrm{Mo}{\mathrm{S}}_2}+E_{\mathrm{N}}+x{E}_{\mathrm{H}}\right)$$ (1)

where $E_{\mathrm{N}{\mathrm{H}}_x/\mathrm{Mo}{\mathrm{S}}_2}$, $E_{\mathrm{Mo}{\mathrm{S}}_2}$, E _N, and E _H are the total energies of the NH _x ‐adsorbed MoS₂, pristine MoS₂, isolated N radical, and isolated H radical, respectively. The calculated reaction pathways presented in Figure 2a (see also Table S1) include: (i) stepwise adsorption of N and H atoms on a surface S site (black dashed lines) (ii) gas‐phase formation of NH₂ followed by adsorption (blue dashed lines), and (iii) the resulting adsorption energies of NH_x species (red dashed lines; see also Table S2). In all scenarios, ${\mathrm{NH}}_2^{\ast }$ is the most energetically favorable configuration. Transient N*and NH* species are exothermically hydrogenated to ${\mathrm{NH}}_2^{\ast }$, whereas further hydrogenation of ${\mathrm{NH}}_2^{\ast }$ leads to NH₃ formation and desorption, indicating that ${\mathrm{NH}}_2^{\ast }$ is the stable surface dopant species responsible for charge transfer.

FIGURE 2.

DFT investigations of NH_x adsorption on pristine and V_S‐containing MoS₂. (a,b) Calculated reaction‐energy profiles for the NH_x radical formation and adsorption (a) on the pristine MoS₂ surface and (b) on the MoS₂ containing a sulfur vacancy (V_S). (c–e) PDOS of (c) NH₂‐adsorbed MoS₂ (${\mathrm{NH}}_2^{*}$/MoS₂), (d) NH₂‐passivated V_S(N_SH₂/MoS₂), and (e) NH‐passivated V_S (N_SH/MoS₂), respectively. The insets show the partial charge densities of the in‐gap states marked by red arrows. For NH₂‐adsorbed MoS₂ and N_SH₂/MoS₂ produce shallow donor states and deep trap states, respectively, whereas N_SH/MoS₂ exhibits the recovery of pristine MoS₂bandgap.

The electronic structure of ${\mathrm{NH}}_2^{*}$/MoS₂ is shown in Figure 2c. The projected density of states (PDOS) [39, 40] reveals donor states introduced below the conduction band minimum (CBM), accompanied by the Fermi‐level shift toward CBM, consistent with n‐type doping. The inset showing the corresponding partial charge density indicates that the donor states are primarily delocalized over the MoS₂ lattice rather than localized on the NH₂ moiety, confirming substantial electron transfer from NH₂ to the MoS₂ conduction band. Spin‐polarized calculations (Figure S8) show a minor exchange splitting (∼0.03 eV), originating from the odd‐electron NH₂ radical, while the MoS₂ substrate remains essentially non‐magnetic.

Given that fabrication‐induced defects are unavoidable in CVD‐grown MoS₂, we further investigated dopant‐defect interactions by considering a V_S. The formation energies similarly calculated for this case are summarized in Figure 2b (see also Figure S9) [41, 42]. When NH₂ is captured at the V_S site, the optimized NH₂/MoS₂‐V_S (N_SH₂/MoS₂) structure relaxed into a double‐bridge configuration in which the N atom is bonded to two H atoms and two neighboring Mo atoms, reflecting the sp³‐like bonding nature of N. This configuration is lower in energy than the NH₂ adsorption on an MoS₂ on‐top S site (${\mathrm{NH}}_2^{*}$/MoS₂) by 2.78 eV, indicating that V_S sites will act as strong trapping centers for mobile ${\mathrm{NH}}_2^{*}$ species and thereby suppress their long‐range interfacial migration. The corresponding PDOS of N_SH₂/MoS₂ (Figure 2d) show deep donor states, indicationg that NH₂ trapped at the V_S site will act as scattering centers.

We then discuss the NH‐passivated V_S configuration (N_SH/MoS₂). As shown in Figure 2b, N occupies the V_S site with a significantly larger binding energy (‐6.19 eV) compared to its adsorption on an MoS₂ surface S site (−1.70 eV). We found that subsequent hydrogenation further stabilizes the S vacancy‐passivating NH configuration (N_SH/MoS₂, −4.07 eV) but in this scenario the second hydrogenation to N_SH₂/MoS₂ is prohibited due to the hydrogen evolution mechanism to H₂ $+$ N_S/MoS₂. Then, the N_SH/MoS₂ PDOS demonstrates restoration of the pristine MoS₂ bandgap or the elimination of vacancy‐induced mid‐gap states (Figure 2e; Figure S9). Temperature‐dependent PL measurements further support this defect‐passivation mechanism (Figure S10). After NH₃ plasma treatment, the defect‐related localized emission peak, commonly referred to as the L peak, was strongly suppressed, as evidenced by the reduced I _L/I _A ratio over the measured temperature range. In addition, the A peak exhibited a red shift after NH₃ plasma treatment, consistent with increased electron density and n‐type doping. These temperature‐dependent PL results provide additional optical evidence that NH₃ plasma treatment suppresses V_S‐related localized defect emission while inducing electron doping in MoS₂. Collectively, these results indicate that NH₃ plasma treatment of PD‐MoS₂ induces robust n‐type doping primarily through NH₂ adsorption on a surface S site, while simultaneously healing sulfur vacancies via NH passivation, thereby enhancing carrier density and allowing band‐like transport without interruption by deep defect states.

2.3. Channel and Contact Resistance Engineering in NH₃ Plasma‐doped MoS₂ FETs

We next investigated how such chemically induced carrier accumulation manifests in transistor‐level device operation. We first fabricated back‐gated (BG) devices employing a 15 nm Al₂O₃ layer as the gate dielectric. Subsequently, our PD stack was deposited on top of the MoS₂ channel, as illustrated in Figure 3a. NH₃ plasma was then selectively applied to the channel region under mild plasma conditions (5 W), and the transfer characteristics were measured after cumulative exposure times ranging from 20 to 240 min to examine time‐dependent doping effects. (Figure 3b) The gray curve in Figure 3b corresponds to the transfer characteristics prior to plasma treatment, showing that the devices retain a high on/off current ratio even after deposition of the PD stack. This confirms that the PD stack deposition alone does not induce significant doping, and that the observed n‐doping effect originates solely from the subsequent plasma process. Upon increasing the plasma exposure time, a progressive negative shift in the threshold‐voltage (V_TH) and a noticeable increase in the on‐current were observed. Notably, after more than 180 min of NH₃ plasma exposure, the on/off ratio collapsed to below 2 while the on‐current increased by more than seventeen‐fold, confirming the emergence of strongly degenerate n‐type behavior. To quantify the induced carrier concentration, the two‐dimensional carrier concentration (n_2D) was extracted from the V_TH shift using the relation n _2D = C _oxΔV _TH/q (Figure 3c). Even after only 20 min of plasma exposure, the electron concentration exceeded 1 × 10¹³ cm⁻², and it eventually saturated at 4.3 × 10¹³ cm⁻² after 180 min of treatment. These results demonstrate that the carrier density can be deliberately tuned simply by adjusting the plasma exposure time, indicating that the doping level is externally controllable. Based on these observations, a plasma treatment duration of 180 min was adopted as the standard condition for all subsequent experiments. This controllable doping behavior was further supported by statistical analysis of 20 BG devices, in which the distributions of V_TH and field‐effect mobility consistently confirmed the NH₃ plasma induced n‐type doping effect through the PD stack (Figure S11 and Table S3).

FIGURE 3.

Localized NH₃ plasma doping of MoS₂ FETs for channel and contact resistance engineering. (a) Schematic of back‐gated MoS₂ FETs used to assess channel‐selective plasma doping under a PD. (b) Transfer curves of devices subjected to increasing plasma exposure durations. (c) Extracted 2D carrier concentration as a function of plasma exposure time. (d) Schematic of a modified device architecture in which both the channel and source/drain contact regions are exposed to plasma. (e) Transfer curves of undoped, channel‐only doped, and dual‐region doped devices. (f) Output characteristics of dual‐region‐doped devices; the inset highlights the linearity in the low‐bias regime. (g) SEM image of the TLM structure. h Total resistance as a function of channel length extracted from the TLM structure. (i) Temperature‐dependent transfer curves of doped devices.

To further enhance carrier injection and reduce the total device resistance beyond what is achievable with channel‐only doping, we expanded the doping region to encompass not only the gated channel but also the source/drain contact regions. This strategy was enabled by a slight architectural modification in which monolayer MoS₂ was transferred onto pre‐patterned metal electrodes, thereby allowing the contact regions to be directly exposed to the NH₃ plasma. The reconstructed device configuration is shown in Figure 3d, and the corresponding transfer curves before and after dual‐region (channel and contact) doping are presented in Figure 3e. In comparison with channel‐only doped devices (Figure 3a), dual‐region doped devices exhibited more than a two‐fold enhancement in both on‐current and transconductance, demonstrating that simultaneous doping of the channel and contacts more effectively reduces the overall device resistance. These results confirm that the proposed barrier‐assisted NH₃ plasma process enables controlled electron doping in both regions, yielding a substantial improvement in carrier injection efficiency. Figure 3f shows that the dual‐region doped device exhibits highly linear output characteristics across the full bias range. This linear behavior is attributed to ohmic‐like carrier injection, which arises from substantial thinning of the Schottky barrier due to the increased electron concentration on the contact region. Separate stability measurements on NH₃ plasma‐doped BG MoS₂ FETs further confirmed the robustness of the doping effect under ambient exposure, 250°C annealing, and gate‐bias stress, showing negligible changes in the extracted doping level and on‐current (Figure S12).

In order to reliably quantify the reduced channel and contact resistance values after doping, transmission line method (TLM) structures with nine different channel lengths were fabricated. The SEM image of the fabricated TLM pattern is shown in Figure 3g. From the linear fit of total resistance versus channel length, we extracted a low contact resistance of 1.45 kΩ·µm (Figure 3h), which is substantially reduced from the undoped value of 47 kΩ·µm. These results demonstrate that our barrier‐assisted process achieves substantial carrier accumulation without structural degradation typically associated with direct plasma exposure.

The fundamental change in electronic quality after the NH₃ plasma treatment is further elucidated by temperature‐dependent electrical measurements. Prior to doping, the devices exhibit a “hopping transport” mechanism as shown in Figure S13, where on‐current increases with temperature—a characteristic behavior of insulating states dominated by V_S and interfacial defect‐induced localized states [43]. In stark contrast, the doped device exhibits a reversal of this trend: as shown in Figure 3i, the on‐current decreases as temperature increases across the entire gate bias range.

This temperature‐inverted behavior is a hallmark of “band‐like transport” in highly degenerate semiconductors. Notably, our DFT investigations provided the microscopic explanation for this transition. While NH₂ adsorption introduces the necessary donor states for n‐type doping, NH radicals concurrently passivate V_S by occupying vacant sites, thereby eliminating mid‐gap trap states and restoring the pristine bandgap. This dual‐functional mechanism simultaneous charge donation and defect healing fundamentally shifts the transport mechanism, enabling band‐like transport in MoS₂ [42, 44]

2.4. Schottky Barrier Engineering via Contact‐Selective NH₃ Plasma Doping

Building upon the previously demonstrated channel‐level doping results, we next extended our barrier‐assisted NH₃ plasma doping strategy to top‐gated MoS₂ FETs. The primary objective was to selectively enhance the contact regions while preserving the low‐power, enhancement‐mode operation of the device. Figure 4a schematically illustrates the device architecture and the corresponding doping procedure. Our devices adopt a source/drain‐first, top‐gate configuration, in which MoS₂ is transferred onto pre‐patterned Au source and drain electrodes. This architecture inherently mitigates Fermi‐level pinning at the metal–semiconductor interface by eliminating direct metal deposition onto MoS₂, thereby suppressing the formation of metal‐induced gap states (MIGS) and achieving cleaner electrical contacts [45]. Moreover, since the MoS₂ layer resides atop the metal electrodes, its exposed top surface remains accessible for subsequent interface engineering, including doping and passivation treatments. Following MoS₂ transfer, a dual‐layer gate dielectric stack (2 nm pV3D3/15 nm Al₂O₃) was conformally deposited. Previous reports have shown that pV3D3 effectively minimizes interfacial defect states and charge scattering, making it an ideal seed layer for top‐gated MoS₂ FETs [27].

FIGURE 4.

Selective contact‐region NH₃ plasma doping in top‐gated MoS₂ FETs. (a) Schematic of a source/drain‐first top‐gated device in which monolayer MoS₂ is transferred onto pre‐patterned Au electrodes. (b) Cross‐sectional illustration of the NH₃ plasma treated device structure, delineating L₁ (plasma‐accessible contact region), L₂ (source/drain–gate overlap region), and L₃ (gated channel region). (c–d) Transfer and transconductance characteristics of the MoS₂ FETs before and after NH₃ plasma treatment. e Schottky barrier height (SBH) as a function of V_TG of MoS₂ FETs before and after NH₃ plasma treatment.

To achieve contact‐selective doping, it was essential to expose the source/drain MoS₂ regions to the NH₃ plasma environment. In the completed top‐gate structure, however, the dielectric layer above the contact regions was significantly thicker than our designed PD stack, which could hinder NH₂ radicals from reaching the underlying MoS₂ surface. Therefore, an additional thinning step of the Al₂O₃ layer was required to enable a controlled doping process to the contact areas. However, plasma‐based dry etching for dielectric removal risks inducing lattice damage and unintentional doping in MoS₂ [8]. Therefore, we adopted a wet‐etching approach using AZ300 MIF, where the gate metal served as a self‐aligned mask. This method selectively removed only the excess Al₂O₃ above the contact regions, leaving only the 2 nm pV3D3/2 nm Al₂O₃ bilayer on top of the MoS₂, which functioned as an effective PD stack during the NH₃ plasma treatment. The AZ300MIF‐based Al₂O₃ etch‐rate characterization and cross‐sectional EDS verification of the etched top‐gated structure are provided in Figure S14. Interestingly, even prior to plasma exposure, a slight positive V_TH shift and moderate mobility enhancement were observed after Al₂O₃ thinning (Figure S15), which can be attributed to the removal of positively charged defect states in the etched Al₂O₃ layer [46, 47]

As with the preceding etching process, subsequent NH₃ plasma treatment was performed using the gate metal as a self‐aligned mask, thereby confining doping to the MoS₂ regions covered by the PD stack above the source and drain contacts. Our goal was to suppress unnecessary diffusion into the channel region while selectively enhancing electron injection at the contact areas. This approach aimed to improve carrier mobility without compromising the enhancement‐mode operation. Figure 4b presents a cross‐sectional schematic of the device together with the corresponding energy band diagrams before and after doping. The contact area can be conceptually divided into three regions: (1) Region L₁, which is protected by the PD stack and is the area where the NH₃ plasma can directly interact with the MoS₂ surface. (2) Region L₂, which overlaps the gate metal and is therefore not directly exposed to the plasma. Although this region does not undergo intentional plasma doping, partial carrier diffusion from L₁ may still occur, which can contribute to reducing the contact resistance. (3) Region L₃, the channel region located beneath the gate metal. While doping in this region can effectively reduce channel resistance, it also leads to degraded switching behavior, including a reduced on/off ratio and a pronounced deterioration in subthreshold swing (SS). In contrast, selective doping of the L₁ region significantly lowers the Schottky barrier, thereby enhancing carrier injection while maintaining the electrostatic control of the channel [48] Therefore, our approach focuses on selective L₁ doping, which enhances carrier injection while sustaining the switching characteristics and preserving enhancement‐mode operation of the device.

As shown in Figure 4c,d the transfer and transconductance characteristics of the device before and after doping. For accurate mobility extraction, the gate capacitance (C_ox) of the pV3D3/Al₂O₃ stack was determined from V_TH shifts under back‐gate bias in a separate underlapped top‐gated MoS₂ device (Figure S16), which eliminates overlap‐induced parasitic capacitances and ensures a reliable value. Based on this capacitance value, the two‐probe field‐effect mobility was extracted from the transconductance in the linear regime using the following Equation (2):

$${\mu }_{\mathrm{FE}}=\frac{L}{W{C}_{\mathrm{ox}}{V}_{\mathrm{DS}}}\frac{d{I}_{\mathrm{DS}}}{d{V}_{\mathrm{GS}}}$$ (2)

where L and W are the channel length and width, respectively, C _ox is the gate capacitance per unit area, V _DS is the drain voltage, and dI _DS/ dV _GS is the transconductance extracted from the top‐gate transfer curve. The extracted field‐effect mobility increased by 5.8‐fold after plasma treatment. The on‐current also increased by 5.8‐fold. Importantly, the V_TH and on/off ratio remained nearly unchanged, confirming that the doping was confined to the contact (L_1,2) regions with only minimal diffusion into the channel. To elucidate the physical origin of performance enhancement, we further conducted temperature‐dependent electrical measurements. As shown in Figure 4e, the effective Schottky barrier height was significantly reduced from 226.7  to 73.7 meV following plasma doping. This reduction originates from Schottky barrier thinning due to increased electron concentration in the contacts, explaining the enhanced mobility and on‐current after NH₃ plasma treatment. The reproducibility of this contact‐doping effect was further verified by statistical analysis of 20 overlap top‐gated MoS₂ FETs, where the distributions of SS, V_TH, and field‐effect mobility confirmed improved carrier injection while preserving gate‐controlled switching characteristics (Figure S17 and Table S4). In addition, these results also support the uniformity of the AZ300MIF‐based Al₂O₃ thinning step prior to NH₃ plasma treatment.

2.5. Extension Resistance Engineering in Underlapped MoS₂ FETs

The ungated extension regions remain the bottleneck for carrier injection, especially under low‐power operation. To address the high series resistance inherent to ungated extension regions, we fabricated underlapped top‐gated MoS₂ FETs and applied our selective NH₃ plasma doping strategy to these areas. Figure 5a illustrates the architecture of our underlapped top‐gated MoS₂ FETs. In this structure, the channel beneath the gate metal is efficiently modulated by the top‐gate voltage, whereas the exposed extension regions exhibit intrinsically high series resistance due to the low carrier density of MoS₂ in these areas. However, the extension regions were initially covered with a thick Al₂O₃, necessitating a controlled wet‐etching step to enable subsequent NH₃ plasma doping. Using the gate metal as a physical mask, we employed a wet etching process to thin down the extension regions, leaving behind a controlled barrier layer suitable for subsequent NH₃ plasma doping. The actual dimensions of the fabricated device were confirmed through SEM analysis (Figure 5b), revealing extension region lengths of 694 nm and 655 nm on the left and right sides, respectively. Even though the extension regions are physically short relative to the channel length (L _ch = 12 µm), their high resistance is considered to severely limit the overall performance of the device. To examine the effect of series resistance, we compared the transfer characteristics before and after doping under varying back‐gate biases.

FIGURE 5.

Extension‐region activation with reduction of series resistance in underlapped top‐gated MoS₂ transistors. (a) Process flow for the fabrication of underlapped top‐gated MoS₂ FETs. (b) SEM image of the fabricated device showing extension lengths of 694 nm and 655 nm on the left and right sides, respectively. (c–d) Transfer and transconductance characteristics of the undoped device measured under varying back‐gate biases. e Output characteristics before (blue) and after (red) extension doping. (f–g) Transfer and transconductance characteristics of the doped device measured under varying back‐gate biases. h Transfer characteristics measured at different drain biases for undoped (blue) and doped (red) devices.

In the undoped device, the high resistance of the extension regions dominates the channel at zero or negative back‐gate bias, resulting in early current saturation due to current limitation through the resistive extensions. Furthermore, when a small positive back‐gate bias (V_BG = 10 V) was applied, the extension resistance no longer dominated the electron transport, leading to a sharp increase in the on‐current. (Figure 5c) This result underscores the often‐overlooked impact of extension region resistance in limiting the performance of 2D FETs, particularly where current injection is most sensitive to series resistance. In contrast, the doped device maintains linear current increase over a broader gate voltage range, indicative of improved carrier injection and reduced series resistance. (Figure 5f) These trends are further confirmed by the transconductance characteristics (Figure 5d–g). While the undoped device exhibits an abrupt rise in transconductance at V_BG = 10 V, the doped device shows a gradual and steady increase without such a dramatic transition. This contrast confirms that selective NH₃ plasma doping effectively activates the ungated extensions, ensuring carrier injection. Importantly, even after doping, the device retains enhancement‐mode operation without a significant shift in V_TH. This suggests that the applied NH₃ plasma treatment mainly affects the exposed extension regions, while unwanted diffusion into the gated channel appears to be largely suppressed. The output characteristics (Figure 5e) further highlight the distinct difference between the undoped and doped devices. The undoped device displays unstable conduction and a low on‐current resulting from limited carrier injection, while the doped device exhibits stable, linear output behavior consistent with ohmic‐like contact formation. Figure 5h further highlights the importance of extension resistance in low‐power operation by comparing the transfer curves at various drain voltages. At low V_DS (0.05 V), the undoped device shows rapid current saturation at very low current levels, while the doped device achieves an on‐current that is approximately 260 times higher. This confirms that sufficient doping in the extension region is essential for maintaining normal device operation under low‐power bias conditions, where series resistance becomes increasingly dominant. The fact that enhancement‐mode operation is preserved even after doping emphasizes the advantage of our method in reducing localized resistance without disrupting gate controllability. Selective extension doping effectively reduces injection‐limiting resistance while preserving strong gate control. This capability is crucial for stable top‐gated MoS₂ FET operation without relying on global electrostatic gating. The reproducibility of this extension‐region doping effect was further confirmed by statistical analysis of 20 underlap top‐gated MoS₂ FETs, where the distributions of SS, V_TH, and on‐current showed consistent enhancement in current drivability and improved effective gate‐controlled switching behavior (Figure S18 and Table S5). Compared with previously reported spacer‐doping approaches, which frequently induce excessive carrier injection into the channel and lead to large V_TH shifts that hinder low‐power operation, our plasma‐based method enables spatially confined conductivity enhancement while preserving enhancement‐mode switching [8, 15, 49, 50]. This controllable doping profile allows the device to maintain a low‐power operating regime without relying on back‐gate electrostatic compensation. To quantitatively compare our barrier‐assisted NH₃ plasma doping approach with previously reported doping strategies for 2D TMD transistors, key parameters including doping concentration, contact resistance, and field‐effect mobility are summarized in Table S6. Collectively, the proposed barrier‐assisted plasma doping achieves spatially confined conductivity enhancement in the underlapped access regions, mitigating the trade‐off between series resistance reduction and V_TH stability. By suppressing parasitic resistance without compromising gate electrostatics, the approach enables high‐performance, low‐power 2D logic and is directly compatible with scalable 2D CMOS integration.

3. Discussion

We have developed a barrier‐assisted NH₃ plasma doping strategy that achieves high‐density, spatially confined n‐doping of monolayer MoS₂ without structural degradation. A bilayer stack consisting of 2 nm pV3D3 and 2 nm Al₂O₃ acts as a “chemical filter”, enabling the controlled diffusion of reactive –NH_x radicals while blocking plasma‐induced kinetic damage. This approach yields degenerate carrier densities ($\sim 4\times {10}^{13}\mathrm{c}{\mathrm{m}}^{-2}$). DFT calculations revealed a dual‐functional mechanism: while NH₂ adsorption provides the primary donor states near the CBM, NH radicals effectively passivate sulfur vacancy defects by occupying the vacancy sites. This passivation is critical as it eliminates mid‐gap trap states and restores the pristine electronic environment of the MoS₂ lattice. Consequently, the doped devices exhibit a transition from the defect‐dominated “hopping transport” characteristic of untreated MoS₂ to “band‐like transport,” confirming the high electronic quality of the doped regions.

The spatial selectivity of the process allows for precise, region‐specific resistance engineering: Contact doping successfully thins the Schottky barrier (from 226.7 to 73.7 meV), enhancing field‐effect mobility and on‐current by 5.8‐fold with only negligible V_TH shift. Furthermore, in underlapped top‐gated devices, extension‐region doping substantially mitigates the dominant parasitic series resistance, improving low‐bias on‐current by 260‐fold (V_DS = 0.05 V). Notably, this performance boost was achieved while maintaining enhancement‐mode operation, which is critical for minimizing static power consumption in logic circuits. Because this method is selective, scalable, and avoids irreversible lattice damage inherent to direct plasma exposure, it provides a practical and integrable route toward high‐performance, energy‐efficient, and 3D‐compatible 2D CMOS platforms.

4. Experimental Method

4.1. MoS₂ Synthesis and Transfer

The MoS₂ film was synthesized within the inner quartz tube (2 inches) situated in the outer quartz tube (4 inches) furnace. High‐purity argon (Ar, 120 sccm) was used as the carrier gas under atmospheric pressure (≈750 Torr). Sulfur powder (120–180 mg, Sigma–Aldrich 99.98%) was placed in the upstream region of the furnace, while molybdenum trioxide (MoO₃) powder (3.0–3.5 mg, Sigma–Aldrich ≥99.5%) and the Si/SiO₂ growth substrates were positioned downstream. Note that perylene 3,4,9,10‐tetracarboxylic acid tetrapotassium salt (PTAS) was applied as a seeding promoter on the growth substrate. During heating, the upstream and downstream temperatures were raised to 830°C and 320°C, respectively, enabling sulfur vaporization, MoO₃ reduction, and subsequent formation of monolayer MoS₂ on the downward‐facing substrates. For transfer, polystyrene (PS) was spin‐coated onto the as‐grown MoS₂ and soft‐baked at 85°C for 15 min to provide mechanical support. The PS/MoS₂ stack was then delaminated from the growth substrate in deionized water and transferred onto the target substrate. After transfer, the PS layer was removed with toluene, yielding a MoS₂ film suitable for device fabrication.

4.2. Device Fabrication

Si/SiO₂ substrates were cleaned in a piranha solution (H₂SO₄:H₂O₂ = 1:1) for 20 min and rinsed thoroughly with deionized water. For the back‐gated devices, Cr (5 nm)/Au (35 nm) gate electrodes were first deposited via thermal evaporation, followed by the formation of a 15 nm thick Al₂O₃ bottom‐gate dielectric using ALD at 150°C. Monolayer MoS₂ was then transferred onto the dielectric surface using a PS‐assisted wet‐transfer method. The active channel region was defined using GXR601 photoresist, and unprotected MoS₂ was removed by O₂ plasma etching using an ICP‐RIE system. After channel patterning, Cr (5 nm)/Au (35 nm) source and drain electrodes were deposited by thermal evaporation. The sequence between MoS₂ transfer and S/D formation was adjusted depending on the intended doping region: MoS₂ was transferred before S/D deposition when only the channel required doping, whereas S/D electrodes were formed first when the doping needed to extend into the contact regions. Following electrode formation, a PD stack was deposited on the device. First, a 2 nm pV3D3 layer was grown by iCVD using V3D3 (300 sccm) and TBPO (100 sccm) at a substrate temperature of 40°C and a filament temperature of 250°C. After this step, an additional 2 nm Al₂O₃ capping layer was deposited by ALD at 150°C. Top‐gate devices were fabricated using a similar process flow, with the primary distinction being the relative order of the MoS₂ channel and the S/D electrodes. In overlapped top‐gate structures, Cr (5 nm)/Au (35 nm) S/D electrodes were patterned on the Si/SiO₂ substrates prior to MoS₂ transfer. In underlapped top‐gate structures, MoS₂ was first transferred onto the substrate, and the S/D electrodes were deposited afterward. After defining the MoS₂ channel, the top‐gate dielectric was formed by sequential deposition of 2 nm pV3D3 (iCVD) and 15 nm Al₂O₃ (ALD, 150°C). Finally, Cr (5 nm)/Au (35 nm) gate electrodes were patterned using NR9 photoresist and deposited by thermal evaporation.

4.3. Electrical and Optical Properties Characterization

A parameter analyzer (Keithley 4200 SCS) integrated with a probe station (MS‐TECH, MST‐1000B) was used for standard electrical characterization, whereas a high‐resolution source measurement unit (Keysight B1500A) enabled precision measurements. All electrical tests were performed under dark conditions and in vacuum (≈5 × 10⁻² Torr) to suppress environmental effects. For in‐house micro‐photoluminescence (PL) and Raman spectroscopy, scattered light was collected using a spectrometer (Andor Kymera 328i) equipped with a charge‐coupled device detector (Andor iDus DU401ABVF). A 532 nm laser was used as the excitation source for both PL and Raman measurements, and identical exposure times were employed across samples to ensure consistent intensity comparisons.

4.4. Computational Method

We performed DFT calculations using the Vienna Ab initio Simulation Package (VASP) based on the projected augmented wave method [51, 52]. Plane waves were expanded with a kinetic energy cutoff of 400 eV with the self‐consistency energy criterion of 10⁻⁴ eV. Atomic structures were optimized using conjugate‐gradient approach until the Hellmann–Feynman forces were less than 0.01 eV/Å. The simulations were performed within the Perdew‐Burke‐Ernzerhof form of generalized gradient approximation [53] with weak long‐range dispersion interactions corrected within the DFT‐D3 scheme [54]. To model 2H MoS₂ monolayer, we adopted a 5 × 5 supercell with 30 Å vacuum spacing Monkhorst‐pack k‐point grid of 4 × 4 × 1 was employed for structural relaxation and 12 × 12 × 1 for density of states calculation. Further computational details can be found in Refs [55, 56].

Author Contributions

I. L. conceived the research, designed the experiments, carried out device fabrication and processing, and performed the electrical characterization with assistance from S. P., M. L., and H. L. J.P. and S.Y. performed the density functional theory calculations and contributed to the theoretical analysis. S.Yo. assisted with figure preparation. D. K., K. R., S. K., K. K., and W. A. contributed to device fabrication and participated in data discussion and analysis. K. Ka. provided guidance on manuscript preparation and contributed to the scientific interpretation. I.L. and J.P. wrote the manuscript with input and discussion from all authors. S.‐Y. C. supervised the overall project and Y.‐H. K. oversaw the computational part and they provided critical feedback and revisions to the manuscript. All authors discussed the results and reviewed the manuscript.

Conflicts of Interest

The authors declare no conflicts of interests.

Supporting information

Supporting File: adma74389‐sup‐0001‐SuppMat.docx.

Data Availability Statement

Relevant data supporting the key findings of this study are available within the article and the Supplementary Information file. All raw data generated during the current study are available from the corresponding authors upon request.