研究領域

High quality material is the most essential issue when we work on novel nanoelectronics. In our group, we start our research with the synthesis of novel 2D materials, including graphene and transition metal dichalcogenides such as WSe2, WS2, and MoS2. We aim at a fundamental understanding of the growth mechanism, defect formation, and crystalline phases so as to manipulate the material properties that meet the need of high-performance nanoelectronic devices. The growth is carried out using chemical vapor deposition which can be well integrated with standard CMOS processing technology. Currently, we have more than 10 chemical vapor deposition systems, each of which is designed for different materials or growth conditions.

Figure 1:
(a) Atomic model of 2H and 1T WS2. (b) Calculated formation energy of 2H and 1T phase with (green curve) and without (purple curve) the adsorption of alkali metal atoms. (c) Schematics of the experimental setup of halide-assisted chemical vapor deposition. (d) SEM image of monolayer WS2 grown on Sapphire. Butterfly-like grains are highlighted by yellow-dotted circles. (e) SEM images of butterfly-like grains in different joint angles. Scale bars are 10 μm. (f) XPS survey spectrum of WS2 nuclei formed in the early stage of nucleation in the halide-assisted CVD. The inset highlights the signal of Na on the WS2 nuclei. (g) XPS spectra of the W4+4f(5/2,7/2) peaks of the 1T and 2H WS2.

Figure 2:
(a) An low-magnification STEM image of a single-layer WS2 butterfly. The red dotted line indicates the boundary between the left and right domains. The nucleation core is marked by a red circle. (b) An ADF image of the WS2 butterfly at the boundary taken from the yellow rectangular in (a). The left domain shows the 2H phase, while the right domain shows the 1T phase. The phase boundary is indicated by red dotted line. (c) The ADF intensity profile along the green dotted line in (b). The two overlapped S atoms in the 2H phase shows c.a. 2.2 times higher intensity than the two misaligned S atoms in the 1T phase. (d) A high-magnification ADF image of the 2H phase. The overlapped S atoms around a W atom are highlighted by a red triangle. (e) A high-magnification ADF image of the 1T phase. Six S atoms around a W atom are highlighted by a yellow hexagon. Scale bars are 0.25 nm. (f),(g) The simulated STEM images of 2H and 1T WS2 corresponding to (d),(e). (h) The ADF image of the overgrown tri-layer WS2 near the nucleation core, sharing the phase boundary with the monolayer region (red dotted line). (i) Local area (surrounded by cyan dotted lines) phase transformation from the 1T to the 2H phase after 3 min of e-beam scanning.

 

The tuning of carrier concentrations in 2D semiconductors such as graphene or TMD is at the heart of novel nanoelectronic and optoelectronic applications. Molecular doping, that is, taking charges from the adsorbed molecules, shows promise as a means by which to change carrier density in graphene while retaining relative high mobility. However, poor control over doping concentrations is a major obstacle to practical applications. Besides, band structure by design in 2D layered semiconductors is highly desirable, with the goal to acquire the electronic properties of interest through the engineering of chemical composition, structure, defect, stacking, or doping. For atomically thin transition metal dichalcogenides, substitutional doping with more than one single type of transition metals is the task for which no feasible approach has been proposed. In one of our research topics, we demonstrate a doping methodology which is able to incorporate multiple kinds of impurities into the host lattice via a non-equilibrium pathway in chemical vapor deposition.

Figure 1 STEM and EELS characterization of WS2 with high-entropy doping of transition metals. a, Colored ADF image of WS2 with high density dopants. The dopants are indicated by white dotted circles. b-e, ADF images, simulated images, and atomic models of Cr, Fe, Nb, and Mo. Scale bars are 0.25 nm. f-i, ADF intensity profiles in comparison with simulated profile of the Cr, Fe, Nb, and Mo dopants. j-m, The corresponding EEL spectra of Cr, Fe, Nb, and Mo dopants.

 

Figure 2 Density of states (DOS) of transition-metal-doped WS2. The calculations were performed using an 8×8 unit cell with one single substitutional impurity: a Cr-doped WS2, c Fe-doped WS2, e Nb-doped WS2, and f Mo-doped WS2. The energy is referenced to the vacuum level, and the DOS for pristine WS2 is shown in black solid lines for comparison. A Gaussian smearing function with a full width at half maximum (FWHM) of 0.1 eV is used in the DOS plots. Filled areas correspond to occupied states. The charge density for a representative impurity state in the energy gap [as marked in a and c by arrows] is shown in b for Cr-doped WS2 and in d for Fe-doped WS2, respectively. Both the top and side views are shown, with cyan, yellow, brown, and pink balls representing W, S, Fe, and Cr atoms, respectively. The constant density surfaces shown correspond to 0.0015 e/Bohr.

The most pressing barrier to the development of advanced electronics based on two-dimensional (2D) layered semiconductors stems from the lack of site-selective synthesis of complementary n- and p-channels with low contact resistance. Here, we report an in-plane epitaxial route for the growth of interlaced 2D semiconductor monolayers using chemical vapor deposition with a gas-confined scheme, in which patterned graphene (Gr) serves as a guiding template for site-selective growth of Gr-WS2-Gr and Gr-WSe2-Gr heterostructures. The Gr/2D semiconductor interface exhibits a transparent contact with a nearly ideal pinning factor of 0.95 for the n-channel WS2 and 0.92 for the p-channel WSe2. The effective depinning of Fermi level gives an ultralow contact resistance of 0.75 and 1.20 kum for WS2 and WSe2, respectively. Integrated logic circuits including inverter, NAND gate, static random access memory, and five-stage ring oscillator are constructed using the complementary Gr-WS2-Gr-WSe2-Gr heterojunctions as a fundamental building block, featuring the prominent performance metrics of high operation frequency (> 0.2 GHz), low-power consumption, large noise margins, and high operational stability. The technology presented here provides a speculative look to the electronic circuitry built on atomic-scale semiconductors in the near future.

Figure 1:
Figure CMOS inverter made of 2D FETs. a. OM image of a 2D integrated circuits array. b-c, OM images of an individual CMOS inverter and the zoomed active channel region. The corresponding circuit diagram is sketched. d, Voltage transfer characteristics with a supply voltage V_DD varying from 0.5 to 2 V. The inset shows the static power consumption as a function of input voltage and the NOT gate symbol. e, Corresponding voltage gain with different V_DD. f, Transient response of the inverter. Clocked swings of V_OUT shows the inverting state synchronized to the V_IN at a frequency of 1 kHz (top panel) and 100 kHz (bottom panel) under V_DD = 2 V.

Figure 2:
Figure NAND gate and SRAM made of 2D FETs. a. OM images of a NAND gate array and the zoomed active channel region. Scale bar: 100 µm. b. OM image of an individual NAND gate. c. Circuit diagram of the NAND gate. d. Output voltage of a NAND gate as a function of time, with the four combination of input states: (0,1), (1,1), (0,0), and (1,0). e. OM image of a 6T-SRAM with its circuit diagram shown in f. g. Static noise margin characteristics of the SRAM cell. The inset shows a circuit diagram illustrating the SRAM cell with two inverters connected in a closed loop. h. Output voltage V_OUT of the SRAM cell in time domain.

Figure 3:
Figure Five-stage ring oscillator made of 2D FETs. a. OM image of a five-stage ring oscillator. b. Circuit diagram of the ring oscillator. c. Output voltage as a function of time at V_DD = ± 2.3 V. d. The power spectrum shows that the fundamental oscillation frequency is at 217 MHz and an overtone at 432 MHz.

The novel 2D materials studied in our group are mostly in the form of single crystal, sometimes doped with other elements for modifying their electrical or optical properties. After the growth of these novel materials, the first thing we have to do is to perform comprehensive characterizations, including its material, electrical, and optical properties. We learn from these characterizations and get to understand more when we make our electronic devices out of these novel materials. Typical characterizations carried out in our group include electrical transport, Raman spectroscopy, X-ray photoelectron spectroscopy, optical conductivity, and high-resolution transmission electron microscopy measurements. Following is an example of TEM and Raman characterizations of single-crystal twisted bilayer graphene:

Figure 1:
TEM images and supercells of twisted bilayer graphene. a, Optical photograph of BLG grains transferred onto a Si substrate, with coordinate markers and scaffolds that encircle the grains for Raman measurements and later for TEM imaging and SAED. b, Low-resolution TEM image of the same structure transferred onto a Cu grid. c-f, HR-TEM images of BLG with different twist angles. Moiré patterns are formed due to the misorientation of the two layers. The insets for each figure present the fast Fourier Transform patterns. All the scale bars are 1 nm. g, Schematic of a commensurate twist of two graphene layers. h, Angle-dependent number of C atoms in the supercell of a commensurate twist. The marked points denote the commensurate twists of the HR-TEM images shown in c-f.

Figure 2:
Two-dimensional Raman mappings of bilayer graphene. a,b are the Raman intensity maps for, respectively, AB stacked and 28° twisted BLG using 488 nm excitation. The spectral regions of the intensity maps are: G peak (1560~1610 cm-1); 2D peak (2650~2750 cm-1); and D and R peaks (1335~1385 cm-1). A satellite BLG appears near the AB stacked BLG in a and shows properties similar to those of the 28° twisted BLG in b. All the maps have the same scale bar.

Mesoscopic physics which locates between the macroscopic world of classical physics and the microscopic world of quantum mechanics is concerned with electronic properties of systems which are, on the one hand, large enough to use statistical methods, but on the other hand, are sufficiently small that the quantum mechanical phase has to be included. Many intriguing physical phenomena appear as electron or spin propagates in such length scale. One part of the research themes in my group is focused on electrical transport in mesoscopic conductors. Ongoing topics include
– Half-integer quantum Hall effect in single-domain graphene hexagons
– Electron scattering at the graphene domain boundaries in quantum Hall regime
– Superconductivity in single domain graphene bilayer with low (0-5 degree) twisting angles

Figure:
Half-integer quantum Hall effect in single-domain graphene grown by chemical vapor deposition. (a) Optical photograph of device fabricated on silicon substrate with 300 nm oxide as a back gate dielectric. (b) SdH oscillation of longitudinal resistance as a function of magnetic field. (c) 2-dimensional plot of Hall resistances as a function of back gate voltage and magnetic field. (d) Hall resistances as a function of magnetic field. Plateaus at filling factor ±2, ±6, and ±10 can be clearly seen.

Performance of 2D photodetectors is often predominated by charge traps that offer an effective photogating effect. The device features an ultrahigh gain and responsivity, but at the cost of a retarded temporal response due to the nature of long-lived trap states. In this topic, we devise a gain mechanism, which originates from massive charge puddles formed in the type-II 2D lateral heterostructures. This concept is demonstrated using graphene-contacted WS2 photodetectors embedded with WSe2 nanodots. Upon light illumination, photoexcited carriers are separated by the built-in field at the WSe2/WS2 heterojunctions (HJs), with holes trapped in the WSe2 nanodots. The resulting WSe2 hole puddles provide a photoconductive gain, as electrons are recirculating during the lifetime of holes that remain trapped in the puddles. The WSe2/WS2 HJ photodetectors exhibit a responsivity of 3×10^2 A/W with a gain of 7×10^2 electrons per photon. Meanwhile, the zero-gate response time is reduced by 5 orders of magnitude as compared to the prior reports for the graphene-contacted pristine WS2 monolayer and WS2/MoS2 hetero-bilayer photodetectors due to the ultrafast intralayer excitonic dynamics in the WSe2/WS2 HJs.

Figure 1: Schematic of WSe2 hole puddle formed in WSe2 nanodot embedded in WS2 matrix due to the type-II band alignment at the WSe2/WS2 heterojunction.

Figure 2: Optoelectronic characteristics of the WSe2/WS2 HJ photodetector. (a) Responsivity as a function of back gate voltage at different incident power densities. (b) Photocurrent as a function of light power density at different gate voltage. (c) Responsivity as a function of light power density at different gate voltage. (d) I_on/I_off ratio and specific detectivity as a function of back gate voltage.Figure 5. Optoelectronic characteristics of the WSe2/WS2 HJ photodetector. (a) Responsivity as a function of back gate voltage at different incident power densities. (b) Photocurrent as a function of light power density at different gate voltage. (c) Responsivity as a function of light power density at different gate voltage. (d) I_on/I_off ratio and specific detectivity as a function of back gate voltage.

Graphene is being intensively explored as a primary channel material applied to high-speed electronics, especially on pliable platforms. However, the potential performance of graphene-based RF devices is primarily limited by two adverse factors within the device fabrication process: charge trapping defects within dielectric layers/graphene interface as well as relatively large access resistance. In this work, we have devised a method for mushroom-shaped top-gate using a stacking technique to improve performance up to a remarkable maximum oscillation frequency of 21.5 GHz. Our method relies on the use of a gate dielectric formed by dry aluminum self-oxidation in pure O2 chamber; this helps achieving higher breakdown voltage and higher gating ability. By using mushroom-shaped gate electrodes, we are able to deposit thicker self-aligned metal contacts, which can effectively reduce the access resistance, thereby enhancing the transconductance. Our devices demonstrate unity-current-gain frequency (fT) and maximum oscillation frequency (fmax) of 22.1 GHz and 13.3 GHz respectively when subjected to strain levels up to 2.5%, for this work, we clearly demonstrate the potential of flexible RF graphenetronics for microwave integrated circuit applications such as low noise amplifiers and frequency mixers, seen as valuable components for next generation wireless communication systems.

Figure 1:
(a) The schematic diagram of the G-FET on a flexible substrate, like PET. (b) The optical image of the large-area G-FET array on an extremely bendable PET substrate. (c) Fmin vs. VTG and VDS contour plot showing suppressed noise output regime. (d) Small signal current gain |h21 | vs. frequency at VSD = 0.7 V, highlighting an intrinsic and extrinsic cutoff frequency of 33 GHz and 20 GHz respectively, corresponding to gate length= 200 nm. (e) The intrinsic available power gain, MAG, shows maximum oscillation frequency exceeding 20 GHz. (f) Application on Doubler. There are input and output signal at 5GHz and 10 GHz respectively. (g) The schematic diagram of the G-FET based receiver system. (h) The snapshot of output spectrum with LO input of 5.98 GHz and RF input of 9.78 GHz at equal power adjusted to 0 dBm, with device bias at VSD = 0.65 V, and VTG = -0.05 V, in which linear sub-harmonic at 3.8 GHz (IF) emerges distinctly, corresponding to the RF mixer frequency limit, since it depends strongly on the “gm” and the extrinsic fT of the device.

Figure 2: Fabrication and transport properties of WS2 FETs with T-gate. (a) Process schematics showing the fabrication of Al/AlOx T-gate. (b) Schematic illustration of the top-gated FET comprising of WS2 n-channel parallel stitched with graphene contacts. The Al/AlOx T-gate allows for the deposition of self-aligned Au contacts (M3 level) on the pre-deposited Au/Cr contacts (M1 level), minimizing the ungated graphene segments, as shown in the false-colored AFM and SEM images in the (c) and (d), respectively. (e) Output characteristics of the WS2 FET at V_TG ranging from 0 to 4 V with a step voltage of 1 V. (f) Transfer characteristics and transconductance of the WS2 FET with various applied drain biases V_DS.

Figure 3: Radio-frequency WS2 FET with channel length of 1 µm. (a) Short circuit current gain |h_21 | as a function of frequency. The inset shows the as-measured S-parameter of the transistor. (b) Power gain (MSG/MAG) as a function of frequency.

Development of novel ALD technology for ultrathin dielectrics and area-selective metal gap filling has been the most pressing barrier in advanced semiconductor manufacturing. In our research group, we have been working on the chamber and flow design of novel ALD systems in the past few years. The novel ALD system incorporates external stimuli into the conventional layer-by-layer deposition process. This design is protected under US patent: US 10522361 B2 (2019) and Taiwan patent: TW 202003896 A (2020). This novel deposition route breaks through the conventional deposition restrictions and provides area-selective metal gap filling using commercial precursors down to the state-of-the-art 5-nm silicon manufacturing node (N5). It should be noted that this development is sponsored mainly by our industrial collaborators, with consecutive joint research projects. The follow-ing figure shows the TEM image of the conformal deposition of SiNx dielectric layer.