Plasma-induced energy band evolution for two-dimensional heterogeneous anti-ambipolar transistors

As three-dimensional (3D) Si-based complementary metal oxide semiconductor (CMOS) technologies approach the miniaturization limit of channel length, two-dimensional semiconductors such as transition metal dichalcogenides (TMDs, such as MoS₂ and WSe₂), metal monochalcogenides (MMC, such as InSe and GeSe), elemental semiconductors (such as silicene, germanene, and phosphorene), and metal oxides (MO, such as CuO and SnO) have been considered as promising channel materials for next-generation energy-efficient nanoelectronics.^1–3 Owing to confined charge transport in sub-1-nm body thickness and naturally self-passivated surface, these 2D layered materials are immune of a variety of scaling-induced unavoidable issues, including finite thickness variation, surface roughness and dangling bonds, and substantial degradation of charge transport in Si.^4,5 Moreover, as more 2D materials being discovered, this layered material family possesses rich and diverse electronic properties that are comparable or superior to Si, such as lattice constant, bandgap, effective mass, carrier mobility, saturation velocity, and critical electric field.^6–8 Owing to these advantages, novel field-effect transistor (FETs) concepts based on 2D semiconductors have been proposed and demonstrated. Taking the most representative 2D semiconductor MoS₂ as an example, these transistor prototypes include ultrashort channel FETs,^9,10 tunnel FETs,^11,12 ferroelectric FETs,¹³ negative capacitance FETs,¹⁴ Dirac source or cold source FETs,^15–17 phase transition FETs,¹⁸ filament-based FETs,^19–21 spin FETs,²² single electron FETs,²³ ambipolar FETs,²⁴ and antiambipolar FETs.²⁵ 

To realize specific functions of these 2D transistors and improve the device performance, plasma treatment stands as a reliable, compatible, and effective process.^26,27 Composed of electrons, ions, excited molecules, and radicals, plasma as the fourth state of matter can interact with 2D materials through different physical and chemical mechanisms and generate various modifications of material properties, including etching-induced thinning, electron or hole doping, phase transition, vacancy healing, and passivation. These interactions can be well controlled by a series of parameters, including plasma source, gas, pressure, frequency, power, and time. Especially with the ultrahigh surface-to-volume ratio of 2D materials, the impact of the plasma treatment is more significant and effective compared to the case with 3D bulk materials. Taking the O₂ plasma treatment on 2D WSe₂ as an example, the process can form a surface oxide layer (WO_x, x approximates 3), introduce a hole doping effect, and consequently provides a variety of benefits, such as atomic-scale etching,²⁸ suppressed contact resistance,^29,30 improved transistor performance,^31,32 highly sensitive photodetection,^33,34 and formation of junction structures.^34,35 Especially in the junction formation, both in-plane homogeneous junction and out-of-plane heterogeneous junction structures have been proposed for device applications.^34,35 However, the evolution of energy band structure, interfacial states, and the corresponding carrier transport as a function of the O₂ plasma treatment has not been clearly revealed yet. With applied electric fields (perpendicular to or along the junction interface), these variations and their impacts on the device operation and performance become more complicated, constraining the design and implementation of the O₂ plasma treatment for the rise of 2D nanoelectronics.

In this work, we exploit the CMOS-compatible O₂ plasma treatment as an effective anionic substitution doping approach for 2D WSe₂ and demonstrate a heterogeneous WSe₂/MoS₂ junction as an antiambipolar FET with outstanding performance. Compared to conventional ambipolar FETs where the dominant charge carriers switch between electrons and holes with an applied gate voltage, the novel characteristic of the antiambipolar FETs is the convergence of both electron and hole branches to achieve a peaked conductivity at an intermediate gate voltage.^36,37 For 2D materials, the antiambipolar charge transport can be realized in 2D/2D or mixed-dimensional heterogeneous structures.^25,38–42 Here, we exploited a novel plasma-induced oxidization and doping to achieve a controllable enhancement of hole transport in WSe₂ through moderate or even degenerate doping. By incorporating with 2D MoS₂ dominated by electron transport as well as the applied in-plane and out-of-plane electric fields, an evolution of the energy band structure of the 2D heterogeneous junction can be obtained, and the corresponding charge transport, dominated by the Fowler–Nordheim (FN) tunneling, is comprehensively elucidated. As an antiambipolar FET, our prototype device exhibits outstanding and balanced performance including a superior peak-valley ratio (PVR, 2.4 × 10⁵) and a high current density (55 nA/μm). This work demonstrates the great potential of 2D materials and their doping engineering to feasibly integrate with the existing CMOS technology and eventually improve the efficiency of future nanoelectronics.

An optical microscope image of the anti-ambipolar FET based on a WSe₂/MoS₂ heterogeneous stacking structure is shown in Fig. 1(a). First, a MoS₂ flake is mechanically exfoliated from its bulk crystals, and then transferred on a Si wafer (0.001–0.005 Ω cm) which has a 285-nm thick SiO₂ layer. Next, a WSe₂ flake is exfoliated as well and transferred on top of the MoS₂ flake to form a partially overlapped heterostructure. A van der Waals (vdW) gap is expected at the heterogeneous interfaces, and it acts as electronically transparent due to the ultrathin spacing (about 0.3 nm). Then, multiple metal electrodes (Ti/Au, 1/100 nm) are patterned and deposited via electron beam lithography and evaporation (Elionix ELS-G100 and Kurt J. Lesker AXXIS). Taking one prototype device as an example, the thickness of MoS₂ and WSe₂ flakes is measured to be 45 and 25 nm, respectively, using an atomic force microscopy (AFM, Bruker Dimension Icon). The mapping of contact potential difference (CPD) is carried out using Kelvin probe force microscopy (KPFM) at room temperature to extract information such as surface potential and work functions. The channel length and width (L and W) of MoS₂, WSe₂, and WSe₂/MoS₂ FETs are 2 and 9 μm, 2 and 11 μm, as well as 11 and 9 μm, respectively. With a grounded source, drain current (I_D) is measured as a function of the applied drain voltage and gate voltage (V_D and V_G) and is normalized as drain current density (J_D = I_D/W) for benchmarking.

Anionic substitutional doping via the O₂ plasma treatment is applied on 2D materials and heterostructures using an inductively coupled plasma reactive ion etcher (ICP-RIE, Trion Tech Phantom III). The plasma treatment condition is optimized with 20 mTorr of the pressure, 11 W of the RIE power, and 98/20 SCCM of the O₂/Ar gas flow, leaving the exposure time as a solo variable parameter. For WSe₂, the O₂ plasma treatment bombards the surfaces with ions and creates anionic Se vacancies. Meanwhile, O free radicals occupy these vacancy sites and react with W to form WO_x. Due to a relatively higher work function (∼6.5 eV compared to ∼4.6 eV of WSe₂),^43–45 WO_x acts as an effective p-type dopant and dope the underneath WSe₂ with holes.

Raman characterization (NanoBase XperRam, 532 nm of wavelength) is performed at room temperature, as shown in Figs. 1(b) and 1(c). Pristine multilayer WSe₂ exhibits an in-plane mode (E¹_2g) at around 250 cm⁻¹ and an out-of-plane mode (A_1g) at 258 cm⁻¹. After the O₂ plasma treatment for 90 s, a blue shift of the E¹_2g mode is obtained, suggesting a p-type doping effect.^46,47 In contrast, pristine multiplayer MoS₂ possesses E¹_2g and A_1g modes at 382 and 406 cm⁻¹, respectively. Both the modes remain consistent after the same plasma treatment, implying negligible impacts of the O₂ plasma treatment for MoS₂ in this work. Especially for WSe₂, the surface potential variation under O₂ plasma treatment is extracted as a function of the etching time, as shown in Figs. 1(d)–1(f). Being different from the reference SiO₂ surface which CPD declines (from ∼300 mV to about zero) as the plasma treatment continues, the CPD of WSe₂ increases substantially from ∼10 to 200 mV, suggesting a significant change of the work function.⁴⁸ This result is also consistent with the doping effect as observed from the Raman spectroscopy analysis.

The impact of the O₂ plasma etching on a WSe₂ FET is elucidated in Fig. 2(a). The O₂ plasma converts the exposed surface of WSe₂ to WO_x which extracts the electrons from the underneath WSe₂ layers and consequently generates a hole doping effect. According to I_D–V_G transfer characteristics, the pristine WSe₂ FET shows ambipolar charge transport, and the charge neutral point locates near zero V_G. As the O₂ plasma treatment is performed cumulatively from 15 to 60 and 120 s, the charge neutral point becomes increasingly positive until WSe₂ becomes degenerately doped with a suppressed gate modulation, as shown in Fig. 2(b). The hole current is promoted by about ten times at V_G = –80 V, and about 10⁵ at zero V_G, compared to the pristine case. I_D–V_D output characteristics possess a good linear current–voltage relationship before and after the plasma treatment, indicating an Ohmic contact condition at metal–semiconductor interfaces, as shown in Fig. 2(c). As a comparison, the response of a MoS₂ FET to the O₂ plasma etching is illustrated in Figs. 2(d)–2(f). The pristine MoS₂ FET shows typical electron-dominant unipolar charge transport starting at about V_G = –60 V. This is due to intrinsic doping and Fermi level pinning near the conduction band edge.⁴⁶ Instead of forming an oxide layer, the O₂ plasma treatment slightly damages the MoS₂ crystal structure and aggravates the formation of defects, dangling bonds, and vacancies. Although an Ohmic contact condition can still be maintained after the plasma treatment, the saturated electron current is decreased by about 50%. As a summary, the performance metrics of both WSe₂ and MoS₂ FETs, including the threshold voltage (V_th), the maximum I_D (I_D,max), and on/off ratio are presented as a function of the plasma treatment time, as shown in Figs. 2(g)–2(i). For the WSe₂ FET, V_th continues to evolve positively as the etching time increases, although I_D,max at V_G = –80 V reaches a saturation after 15 s of etching. The on/off ratio is degraded due to the degenerate doping which lowers the gating efficiency. In contrast, the MoS₂ FET is degraded and stabilized after 15 s of etching.

With the understanding of the individual WSe₂ and MoS₂ FETs, the evolution of the heterogeneous WSe₂/MoS₂ FET as a function of the plasma etching time is shown in Fig. 3. A clear antiambipolar charge transport peak is developed as the etching time increases, and the features including the maximum I_D and corresponding V_G of the antiambipolar peak (I_D,peak and V_G,peak) as well as the PVR value are summarized. For both forward and reverse sweeps, I_D,peak evolves to near 10⁻⁶ A after 120 s of etching, and V_G,peak shifts positively for about 20 V. The PVR is improved with the plasma treatment and reaches to the maximum up to 2.4 × 10⁵ after 60 s of etching. At this condition, the heterogeneous WSe₂/MoS₂ FET is optimized for antiambipolar charge transport since the less etching time (e.g., 15 s) causes a lower I_D,peak and the longer etching time (e.g., 120 s) leads to a doping-induced higher off current (or valley current).

The antiambipolar charge transport is directly introduced by the unique energy band structure of the heterogeneous WSe₂/MoS₂ junction and its response to the applied electrostatic gating. Therefore, the I_D–V_D output characteristics of the WSe₂/MoS₂ FET are carefully evaluated after each plasma treatment. The applied V_D provides an in-plane electric field to drive the charge injection and transport, whereas the applied V_G generates an out-of-plane electric field and capacitively modulates the doping level of 2D materials and, thus, the bending of energy band structures. First, the energy band structure of the WSe₂/MoS₂ junction before and after the O₂ plasma doping are compared, as shown in Fig. 4(a). Both MoS₂ and Wse₂ multilayers have the similar bandgaps of 1.2 eV. Owing to the WO_x-induced p-type doping, the work function of the WO_x/WSe₂ (Φ_WSe2) layers becomes larger, compared to the unaffected work function of MoS₂ (Φ_MoS2). As a result, a larger built-in potential energy barrier (qV_bi = Φ_WSe2–Φ_MoS2) at the WSe₂/MoS₂ interface is established as the etching time increases. Second, by assigning MoS₂ as grounded and applying a positive V_D at WSe₂, the I_D–V_D output characteristics in the antiambipolar regime (−40 to 80 V) reveal a polarity transition after the plasma etching, as shown in Fig. 4(b). Meanwhile, the knee voltage (or cut-in voltage) shifts from −0.3 to 0.41 V as 120 s of the etching is applied, suggesting a drastic change of qV_bi, approximately 0.7 eV. Third, the data are reconstructed into ln(I_D/V_D²) versus 1/V_D plots to reveal the charge transport mechanism, as shown in Fig. 4(c). Linear decay from the high V_D (or zero 1/V_D) center indicates the dominance of FN tunneling.^49,50 In the pristine state, the charge transport in the V_G range of −40 to 80 V is mainly dominated by the electrons [see Figs. 2(b) and 2(e)], and their FN tunneling injection at the WSe₂ end is enabled when the negative V_D and positive V_G synergistically form a thin triangular electron barrier on the conduction band. After the plasma-induced p-type doping, the dominant charge is switched to the holes, and their FN tunneling injection through a thin triangular hole barrier on the valence band is enabled by the cooperation of the positive V_D and negative V_G. As V_G increases positively, the triangular barrier turns into a thicker and greater barrier and consequently diminishes the quantum tunneling probability and, thus, the current injection efficiency. When a degenerate p-type doping in WSe₂ is achieved after the plasma treatment, WSe₂ becomes less modulated by V_G, and an electronically transparent contact is established at the WSe₂ end. The valence band offset reaches to the maximum and approximates to qV_bi. The hole FN tunneling occurs at the WSe₂/MoS₂ interface at the positive V_D and shows a V_G dependence due to the varying thickness of the MoS₂ triangular barrier under the electrostatic gating. Finally, the on and off states of the WSe₂/MoS₂ junction at different V_D and V_G are summarized in Fig. 4(d). A clear evolution of the energy band structure and the corresponding charge transport is illustrated as a moderate-to-degenerate doping is applied on WSe₂ through a tunable O₂ plasma etching treatment.

In addition to the WSe₂/MoS₂ FET where WSe₂ is transferred on top of MoS₂, we also fabricated the MoS₂/WSe₂ FET where the stacking order is reversed. It is anticipated that WSe₂ receives a stronger electrostatic gating in the MoS₂/WSe₂ FET, but only the exposed WSe₂ area can form WO_x by the O₂ plasma treatment and then develop a localized p-type doping effect. The experimental I_D–V_G transfer characteristics as a function of the plasma etching time are shown in Fig. 5. Similar to the WSe₂/MoS₂ FET, an evolution of the antiambipolar charge transport is obtained, suggesting the broad applicability of the O₂ plasma treatment for WSe₂-based nanoelectronic devices, even with a variety of architectures.

Among the WSe₂/MoS₂ antiambipolar FETs prepared in this work, the best device exhibits an outstanding PVR value up to 2.4 × 10⁵ at room temperature. The I_D–V_G transfer characteristics of the WSe₂/MoS₂ FET are presented in contrast with the individual WSe₂ and MoS₂ FETs after the identical plasma treatment, as shown in Fig. 6(a). It is clear that the antiambipolar charge transport behavior reflects the integration of the hole branch from WSe₂ and the electron branch from MoS₂. The channel current is constrained by the counterpart material with a lower conductivity in a series connection and, consequently, presents a peak at the transition between the hole and electron dominance. We also benchmark the PVR and I_D,peak metrics of our device with other heterogeneous antiambipolar FETs based on 2D materials including WSe₂, MoS₂, SnS₂, and ReS₂,^{25,37,51–62} as shown in Fig. 6(b). Our device shows superior and balanced performance, and it can be further improved by architectural and processing optimization, for example, by adopting high-k dielectrics (e.g., HfO₂) to enhance the electrostatic gating and utilizing asymmetric contacts (e.g., Pt contacts for WSe₂ and Bi contacts for MoS₂) to promote the antiambipolar charge transport.

In this work, we implement a CMOS-compatible O₂ plasma treatment to controllably introduce a p-type doping effect, ranging from moderate to even degenerate doping, on 2D semiconductor WSe₂. As the plasma treatment progresses, the energy band evolution of a WSe₂/MoS₂ heterogeneous structure is evaluated, and the novel charge transport behavior under the applied electric fields is investigated for realizing an antiambipolar FET with outstanding and balanced performance.

This work was mainly supported by the National Science Foundation (NSF) under Award No. ECCS-1944095 and partially supported by the New York State Center of Excellence in Materials Informatics (CMI). The authors acknowledge support from the Office of the Vice President for Research and Economic Development (OVPRED) at the University at Buffalo. A. Ahmed, A. Cabanillas, A. Chakravarty, J. Muhigirwa, and M. Enaitalla acknowledge support from the Arthur A. Schomburg Fellowship, the Presidential Fellowship, and the Collegiate Science and Technology Entry Program (CSTEP) at the University at Buffalo.

The authors have no conflicts to disclose.

H. Li conceived the project and wrote the manuscript. F. Yao and J. Liu supervised the project. S. Shahi, A. Ahmed, A. Cabanillas, and Y. Guo performed the device fabrication and measurement. R. Yang, Y. Fu, A. Butler, and C. T. Chen performed the material characterization. A. Chakravarty, M. Liu, H. N. Jaiswal, S. Jadeja, and H. Murugesan assisted the data analysis and discussion. J. Muhigirwa and M. Enaitalla participated in the sample preparation.

Simran Shahi: Investigation (lead). Asma Ahmed: Investigation (equal). Ruizhe Yang: Investigation (equal). Anthony Cabanillas: Investigation (equal). Anindita Chakravarty: Investigation (equal). Maomao Liu: Investigation (equal). Hemendra Nath Jaiswal: Investigation (equal). Yu Fu: Investigation (equal). Yutong Guo: Investigation (equal). Satyajeetsinh Shaileshsin Jadeja: Investigation (equal). Hariharan Murugesan: Investigation (equal). Anthony Butler: Investigation (equal). Chu Te Chen: Investigation (equal). Joel Muhigirwa: Investigation (equal). Mohamed Enaitalla: Investigation (equal). Jun Liu: Supervision (equal). Fei Yao: Supervision (equal). Huamin Li: Conceptualization (lead); Writing – original draft (lead); Writing – review & editing (lead).

Simran Shahi: Investigation (lead). Asma Ahmed: Investigation (equal). Ruizhe Yang: Investigation (equal). Anthony Cabanillas: Investigation (equal). Anindita Chakravarty: Investigation (equal). Maomao Liu: Investigation (equal). Hemendra Nath Jaiswal: Investigation (equal). Yu Fu: Investigation (equal). Yutong Guo: Investigation (equal). Satyajeetsinh Shaileshsin Jadeja: Investigation (equal). Hariharan Murugesan: Investigation (equal). Anthony Butler: Investigation (equal). Chu Te Chen: Investigation (equal). Joel Muhigirwa: Investigation (equal). Mohamed Enaitalla: Investigation (equal). Jun Liu: Supervision (equal). Fei Yao: Supervision (equal). Huamin Li: Conceptualization (lead); Writing – original draft (lead); Writing – review & editing (lead).

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Jun Liu received his Ph.D. in Materials Engineering from the University of Alberta, Canada in 2018. He was a Postdoc Fellow in Chemical and Biological Engineering at SUNY Buffalo (2018–2020). He is currently an Assistant Professor in Mechanical and Aerospace Engineering, University at Buffalo. He is also an Associate Editor of Energy Technology (Wiley). His research is focused on understanding and manipulating fundamental physics at materials surface and interfaces for energy and sensing applications. He is the recipient of many awards such as 2022 SONY Faculty Innovation Award and 2020 Microsystem and Nanoengineering (MINE) Young Scientist Award (Springer Nature).

Fei Yao received her dual Ph.D. degrees in Energy Science from Sungkyunkwan University, Korea and in Physics from Ecole Polytechnique, France, in 2013. She worked as postdoctoral researcher in the Center for Integrated Nanostructure Physics, Korea, and in the Department of Electrical Engineering, University of Notre Dame, USA. She is now an Assistant Professor in the Department of Materials Design and Innovation, University at Buffalo. Her research interests include low-dimensional materials synthesis and applications in energy storage and conversion.

Huamin Li has completed his Ph.D. from Sungkyunkwan University, Korea, and postdoctoral research from University of Notre Dame, USA. His expertise is in the exploration of two-dimensional materials and their application for high-performance energy-efficient nanoelectronics. He is currently serving as an Associate Editor Member in IEEE Access, Nano Express, and Materials Research Letters; a Technical Committee Member (Nanoelectronics) for Institute of Electrical and Electronics Engineers (IEEE) Nanotechnology Council; and an Executive Council Member for 2D Materials Technical Group in American Vacuum Society (AVS).