您当前的位置： 首页 > 网页快照
 Remote modulation doping in van der Waals heterostructure transistors - Nature Electronics Abstract . Doping is required to modulate the electrical properties of semiconductors but introduces impurities that lead to Coulomb scattering, which hampers charge transport. Such scattering is a particular issue in two-dimensional semiconductors because charged impurities are in close proximity to the atomically thin channel. Here we report the remote modulation doping of a two-dimensional transistor that consists of a band-modulated tungsten diselenide/hexagonal boron nitride/molybdenum disulfide heterostructure. The underlying molybdenum disulfide channel is remotely doped via controlled charge transfer from dopants on the tungsten diselenide surface. The modulation-doped device exhibits two-dimensional-confined charge transport and the suppression of impurity scattering, shown by increasing mobility with decreasing temperature. Our molybdenum disulfide modulation-doped field-effect transistors exhibit a room-temperature mobility of 60?cm 2 ?V –1 ?s – 1 ; in comparison, transistors that have been directly doped exhibit a mobility of 35?cm 2 ?V –1 ?s – 1 . You have full access to this article via your institution. Download PDF Download PDF Main . Doping while suppressing Coulomb scattering is essential when fabricating high-performance electronic devices based on conventional semiconductors 1 , 2 . However, substitutional doping introduces excess carriers that generally leave behind ionized dopants in the channel, which disturb the transport of charge carriers. To avoid this, a modulation doping technique in which ionized dopants are spatially separated from the channel has been widely used in epitaxial heterostructures based on semiconducting nanowires 3 , complex oxides 4 , 5 and compound semiconductors 6 , 7 . Two-dimensional (2D) carrier gases created in this way have also been used as a platform to fabricate high-electron-mobility transistors (HEMTs) and to study quantum transport phenomena 8 , 9 . Charge carrier transport in 2D semiconductors, such as transition metal dichalcogenides (TMDCs), is strongly affected by both intrinsic defects and extrinsic effects, because their carriers are intrinsically confined to their atomically thin bodies 10 , 11 , 12 , 13 , 14 . Unlike 2D carrier gases, which are artificially created in bulk semiconductor heterostructures, carriers in 2D materials are substantially affected by scattering events via long-range Coulomb interactions induced by the surrounding environment 15 , 16 , 17 . This makes the investigation of the fundamental physics and physical properties of 2D materials challenging. In particular, to experimentally explore collective quantum phenomena—including the quantum Hall effect 18 , 19 , 20 , valley Hall effect 21 , 22 , 23 , magic-angle superconductivity 24 , 25 and moiré interlayer excitons 26 , 27 —high-quality materials and interfaces that can eliminate unwanted scattering are required. In addition, charged impurity scattering degrades device performance, such as mobility, limiting the practical applications for high-speed 2D electronics. Previous studies have shown that external scattering sources of 2D materials can be suppressed by encapsulation with hexagonal boron nitride (hBN) or a polymer 28 , 29 , 30 . However, reducing intrinsic scattering caused by impurity states, such as ionized dopants, remains a challenge 16 , 31 . Doping in 2D semiconductors can be achieved by either atomic substitution of elements or molecular surface doping via charge transfer 32 , 33 , 34 , 35 . Substitutional doping is prevalent in bulk semiconductors and has also been extensively explored in 2D semiconductors. Charge transfer doping is a simple and powerful approach that is uniquely applicable to atomically thin van der Waals (vdW) semiconductors 36 , 37 . Neither method can, however, avoid charged impurity scattering because ionized dopants exist within the atomically thin lattice or on its surface close to the channel. In this Article, we report modulation doping in 2D semiconductors using vdW band engineering and remote charge transfer doping. Our heterostructure consists of a molybdenum disulfide (MoS 2 ) embedded channel, an hBN tunnel barrier and a tungsten diselenide (WSe 2 ) layer. The WSe 2 layer is chemically doped with an n-type molecular dopant triphenylphosphine (PPh 3 ), and we show that this doping can modulate the charge carrier density in the underlying MoS 2 channel by remote charge transfer, without degrading the mobility. The resulting MoS 2 modulation-doped field-effect transistors (MODFETs) exhibit a room-temperature mobility of 60?cm 2 ?V –1 ?s –1 , compared with a mobility of 35?cm 2 ?V –1 ?s –1 for directly doped devices. Temperature-dependent mobility measurements show that almost complete suppression of charge scattering can be achieved in the doped MoS 2 channel. Band-modulated vdW heterostructure for remote doping . Figure 1a shows a schematic of the vdW heterostructure used to demonstrate modulation doping in a 2D semiconductor. To fabricate the WSe 2 /hBN/MoS 2 heterostructure, individually exfoliated 2D constituent flakes are stacked in a layer-by-layer manner in sequence using standard transfer methods. In particular, the electrical contacts using either metal or graphene were made only on MoS 2 before the stacking of hBN and WSe 2 layers (the fabrication details are available in Methods and Supplementary Fig. 1 ). The cross-sectional transmission electron microscopy image (Fig. 1b ) confirms the constitution and interfaces of a representative WSe 2 /hBN/MoS 2 heterostructure, consisting of four-layer WSe 2 , trilayer hBN and trilayer MoS 2 . The thickness of each layer was determined by atomic force microscopy before stacking (Supplementary Fig. 2 ). The trilayer MoS 2 channel was primarily used in this study because it is still thin enough to be modulated by the back gate, and small contact resistance can be achieved even at low temperatures 38 . In this heterostructure, with the backbone of a WSe 2 /MoS 2 type-II junction, the electrons introduced into the top WSe 2 layer via molecular doping spontaneously transfer to MoS 2 due to the conduction band offset. This consequently leads to the doping of the confined MoS 2 channel (Fig. 1c ), which is similar to the 2D electron gas (2DEG) in traditional bulk MODFETs or HEMTs 6 . In particular, WSe 2 was chosen as the top layer because of the easier charge transfer doping than hBN with a large bandgap in addition to the lower electron affinity than MoS 2 . Also, note that the insertion of a vdW hBN spacer layer between WSe 2 and MoS 2 not only enables the effective confinement of doped carriers but also reduces remote optical phonon scattering from WSe 2 and extrinsic scattering arising from the alloy disorder and interface roughness typically existing in bulk material systems 39 . Fig. 1: Band-modulated WSe 2 /hBN/MoS 2 heterostructure for remote doping. a , b , Schematic ( a ) and cross-sectional bright-field scanning transmission electron microscopy image ( b ) of the representative heterostructure consisting of WSe 2 , hBN and MoS 2 layers. Scale bar in b , 2?nm. c , Band diagram of the doped WSe 2 /hBN/MoS 2 FET consisting of MoS 2 as the active channel layer, hBN as the interfacial layer, WSe 2 as the chemically doped layer and PPh 3 as the n-type dopant. This structure leads to the confinement of electrons in the conduction band of MoS 2 . E C , E V and E F indicate the conduction band edge, valence band edge and Fermi level energy, respectively. Full size image Remote modulation doping in MoS 2 FETs . We first reveal the charge transfer interactions in the abovementioned heterostructure by optical characterization, including photoluminescence (PL) and Raman spectroscopy. Figure 2a shows a schematic (top) and optical microscopy image (bottom) of the representative WSe 2 /hBN/MoS 2 heterostructure with multi-terminal graphene contacts. In the PL spectra for each layer, characteristic peaks—the A exciton ( A Mo ) at 1.83?eV and B exciton ( B Mo ) at 2.00?eV for trilayer MoS 2 and the indirect transition ( I W ) at 1.45?eV and A exciton ( A W ) at 1.57?eV for four-layer WSe 2 —are observed (Fig. 2b ). Primarily, in the WSe 2 /hBN/MoS 2 junction area, the intensities of all the peaks decrease as a consequence of the interlayer charge transfer process of the photogenerated excitons 40 . The PL mapping at 1.83?eV ( A Mo ) shows a homogeneous reduction in intensity over the overlapping area, tentatively implying that the charge transfer interaction occurs uniformly over the area owing to the high interface quality of the stacked WSe 2 /hBN/MoS 2 heterostructures (Supplementary Fig. 3 ). Notably, the insertion of hBN preserves the charge transfer and resultant PL quenching behaviour observed in the WSe 2 /MoS 2 heterojunction because the trilayer hBN between WSe 2 and MoS 2 is sufficiently thin for the tunnelling of charge carriers (Supplementary Fig. 4 ). Fig. 2: Remote modulation doping in WSe 2 /hBN/MoS 2 heterostructures via charge transfer. a , Schematic (top) and optical microscopy image (bottom) of the representative WSe 2 /hBN/MoS 2 FET. Red, blue, green and grey solid lines represent the edges of the MoS 2 , WSe 2 , hBN and graphene flakes, respectively. Scale bar, 10?μm. b , PL spectra of the regions including WSe 2 /hBN/MoS 2 (black solid line), MoS 2 (red solid line) and WSe 2 (blue solid line). E ph , I W , A W and B Mo indicate the photon energy, indirect transition of WSe 2 , A exciton of WSe 2 and B exciton of MoS 2 , respectively. c , Electrical sheet conductivity at V D ?=?0.1?V, σ ?=?( I D / V D )( L / W ), as a function of the gate voltage ( V BG ) for the MD (red lines) and DD (black lines) trilayer MoS 2 FETs before (dashed lines) and after (solid lines) PPh 3 doping. The inset shows a schematic of the MD and DD devices. d , Field-effect mobility ( μ ) as a function of the electron concentration ( n e ) of the FETs before (open symbols) and after (filled symbols) doping. The light blue (light red) and blue (red) filled squares (circles) represent the DD (MD) FETs after doping once and twice, respectively. The error bars indicate the standard deviation obtained from measuring more than five devices. Full size image More importantly, to verify the doping capability via charge transfer, we investigated the electrical characteristics of the MoS 2 field-effect transistor (FET) when the device was doped with PPh 3 . Two different types of MoS 2 FET are presented in this study: the band-modulated WSe 2 /hBN/MoS 2 heterostructure for remote doping and bare MoS 2 for direct doping as a control. For studies on the device characteristics depending on the doping concentration, the devices were fabricated using metal (Supplementary Figs. 5 and 6 ) and graphene (Fig. 2 ) electrodes, which exhibited Ohmic-like linear current?voltage ( I D ? V D ) characteristics at room temperature due to a low Schottky barrier. All the FETs were fabricated on hBN/SiO 2 /Si substrates to minimize external scattering from the surface roughness, charged impurities and polar optical phonons of the substrate. Figure 2c compares the representative gate-dependent transport characteristics between the modulation-doped (MD) WSe 2 /hBN/MoS 2 and directly doped (DD) MoS 2 FETs, which are doped with the same concentration of 45% PPh 3 solution ( I D ? V D curves and mobility characteristics are shown in Supplementary Fig. 7 ). Both devices similarly show negative shifts in the threshold voltage ( V th ) after PPh 3 treatment, indicating an n-type doping behaviour. The 2D sheet electron density ( n e ) is calculated as n e ?=? I D L /( qWV D μ ) using the field-effect mobility ( μ ) extracted from I D ? V BG curves, μ ?=? G m L /( V D C OX W ), where L is the channel length, W is the channel width, q is the electron charge, G m is the transconductance and C OX is the gate capacitance per unit area 41 . On doping, n e at V BG ?=?0?V increases from 1.1?×?10 12 to 4.8?×?10 12 ?cm –2 and from 8.1?×?10 11 to 5.6?×?10 12 ?cm –2 for the MD and DD devices, respectively. Variations in n e are similar for the two cases despite slightly less doping in the MD device. This implies that PPh 3 molecules can modulate the carrier density of the underlying MoS 2 channel via remote charge transfer across the WSe 2 /hBN layers without a substantial loss in the doping efficiency. The electron density of both MoS 2 devices can be controlled according to the PPh 3 concentration and the number of dopings, as confirmed by the control experiments, and its upper limit is 1.5?×?10 13 ?cm –2 and 1.1?×?10 13 ?cm –2 for the DD and MD samples, respectively (Fig. 2d and Supplementary Fig. 6 ). Note that the electron transport occurs dominantly through MoS 2 and the contribution of WSe 2 in the stacked heterostructures is negligible because pristine WSe 2 intrinsically exhibits p-type characteristics with a high series resistance (Supplementary Fig. 8 ). Even after doping, WSe 2 shows much lower electron conductivity than MoS 2 (Supplementary Fig. 9 ). In addition to the ability to modulate the carrier density, the notable aspect of remote modulation doping is presented in Fig. 2d , which plots μ as a function of n e at V BG ?=?0 for all the devices studied in this work. Even after doping, μ remains virtually unchanged at approximately 60?cm 2 ?V –1 ?s –1 for the MD device, as depicted in Fig. 2d . However, the counterpart DD device shows a degradation in μ from 60 to 35?cm 2 ?V –1 ?s –1 , which is consistently observed for all the DD devices. This result partly supports the fact that charged impurity scattering is suppressed even at room temperature in the case of heterostructure MD FETs, and the detailed transport characteristics and scattering mechanisms are further discussed below. Temperature-dependent electron transport in MoS 2 MODFETs . To evaluate the charge transport properties, we performed the temperature-dependent electrical characterization of the graphene-contacted devices using four-probe measurements to eliminate the effect of contact resistance. Figure 3a shows four-probe conductivity ( σ 4p ) as a function of V BG for the MD device at various temperatures ranging from 300 to 10?K. As the temperature decreases, σ 4p increases for V BG ?>??50?V and vice versa for V BG ? \frac{{3k_{{{\mathrm{B}}}}T}}{q}\) , the equation can be reduced to $$I_{{{\mathrm{D}}}} = A_{2{{{\mathrm{D}}}}}T^{3/2}{{{\mathrm{exp}}}}\left[ { - \frac{q}{{k_{{{\mathrm{B}}}}T}}\left( {\phi _{{{\mathrm{B}}}} - \frac{{V_{{{\mathrm{D}}}}}}{\eta }} \right)} \right]$$ (see the detailed derivation in Supplementary Information ). As shown in Fig. 3c , using this equation, the ? B value of MD devices extracted from the slope of the Arrhenius plot approaches 0?eV as the unpinned Fermi level of graphene is effectively aligned to the conduction band of MoS 2 with an increase in the gate voltage (Fig. 3c , inset). Such a behaviour is similarly observed from the DD counterparts (Supplementary Fig. 10 ). Further, R c is determined as $$R_{{{\mathrm{c}}}} = \frac{1}{2}\left( {R_{2{{{\mathrm{p}}}}} - R_{4{{{\mathrm{p}}}}}\frac{{l_{{{{\mathrm{tot}}}}}}}{{l_{{{{\mathrm{in}}}}}}}} \right)$$ , where R 2p and R 4p are the two- and four-probe resistances, respectively, and l tot and l in are the total and inner channel lengths, respectively. Both MD and DD samples clearly show low R c ranging from 6 to 15?kΩ?μm at V BG ?=?0?V and low temperatures (Supplementary Fig. 11 ). These values of R c and ? B are comparable with or smaller than those of the previously reported graphene-contacted MoS 2 FETs 28 , 30 . Fig. 3: Low-temperature transport measurements of MD MoS 2 FETs. a , Four-probe sheet conductivity ( σ 4p ) as a function of V BG for the MD FET for temperatures varying from 300 to 10?K. Inset: I D ? V D curves at V BG ?=?0?V for the devices under consideration. MoS 2 is doped twice by the PPh 3 solution with a concentration of 45%. b , Arrhenius plots for the FETs as V BG varies from ?60 to 60?V with an increment of 10?V. The symbols and lines represent the experimentally measured values and linear fittings, respectively. c , Schottky barrier height ( ? B ) as a function of V BG extracted from the Arrhenius plot. Error bars are determined from the standard deviation of the linear fitting. For the extraction, V D ?=?0.1?V and η ?=?4.1 calculated from the ln( I D )? V D plot are used. The inset shows the energy band diagram under a high gate voltage; E C , E V and E F denote the conduction band edge, valence band edge and Fermi level, respectively. d , Four-probe sheet conductivity as a function of temperature at different V BG values ranging from ?70 to 70?V with an increment of 10?V. The red- and blue-shaded regions shown in a and d indicate the metallic and insulating states, respectively. Full size image Once reasonably good contacts were achieved, we could verify the metal–insulator transition (MIT) in our samples, which is shown in the σ 4p curves as a function of temperature for various V BG values (Fig. 3d ). Here σ 4p increases as the temperature decreases for V BG ?>??50?V, which is a characteristic of the metallic phase, and vice versa for V BG ? From：
 监测目标主题 (1) 系统抽取主题 (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1)