Chemical vapor deposition (CVD) can produce wafer-scale transition-metal dichalcogenide (TMD) monolayers for the integration of electronic and optoelectronic devices. Nonetheless, the material quality of the CVD-grown TMDs still remains controversial. Here, we compare the quality of representative WSe2 monolayers grown by CVD compared to that obtained by other synthesis methods: bulk-grown-chemical vapor transport (CVT) and flux. Through the use of a deep-learning–based algorithm to analyze atomic-resolution scanning transmission electron microscopy images, we confirm that Se vacancies (VSe) are the primary defects in WSe2, with a defect density of ∼5.3 × 1013 cm−2 in the CVD-grown sample, within the same order of magnitude of other methods (∼3.9 × 1013 cm−2 from CVT-grown samples and ∼2.7 × 1013 cm−2 from flux-grown samples). The carrier concentration in field-effect transistors at room temperature is ∼5.84 × 1012 cm−2 from a CVD-grown sample, comparable to other methods (6–7 × 1012 cm−2). The field-effect mobility of the CVD-grown sample is slightly lower than that of other synthesis methods, together with similar trends in on-current. While the difference in photoluminescence intensity is not appreciable at room temperature, different intensities of defect-related localized states appear below 60 K. We conclude that the wafer-scale CVD-grown samples can be utilized without loss of generality in the integration of electronic/optoelectronic devices.

The peculiar optical and electronic properties of two-dimensional (2D) van der Waals layered materials, including strong Coulomb interaction, large spin–orbit coupling, and superb carrier mobility, have been pioneering the development of future electronics and optoelectronics.1–4 Particularly, the direct-bandgap nature of semiconducting transition-metal dichalcogenide (TMD) monolayers fosters the development of highly sensitive optical devices.5,6 These unique properties have been primarily explored using monolayer TMDs mechanically exfoliated from single-crystal bulk, synthesized by chemical vapor transport (CVT) and flux methods. However, despite their outstanding qualities, the small flake size (reaching only the micrometer scale) limits further industrial applications.

On the other hand, chemical vapor deposition (CVD) offers several advantages such as scalability, cost-effectiveness, high throughput, and controllability. More significantly, CVD can produce TMD monolayers on a wafer scale.7–11 Recently, 14-inch van der Waals heterostructure films were fabricated through CVD growth, offering large-scale integrated circuits for electronic and optoelectronic devices for inverters, NAND gates, ring oscillators, and photodetector arrays.11–14 Nevertheless, some issues remain to be elucidated for the CVD-TMDs to be suitable for industrial applications due to their limited material quality. CVT or flux-TMD monolayers often exhibit large on-currents and high carrier mobility with a high on/off ratio, whereas CVD-TMD monolayers typically show values that are several orders of magnitude lower.15 Furthermore, the optical properties such as photoresponsivity, detectivity, and quantum efficiency in optoelectronic devices are subpar compared to CVT or flux-grown materials. Despite the relative inferiority of CVD TMD monolayers to those grown by CVT or flux methods, no study has directly compared their electrical and optical properties.

In this study, we offer a direct comparison of the defect density and electrical/optical properties of monolayer WSe2 grown via CVD, flux, and CVT methods. To accurately identify the defect types and estimate their concentrations in WSe2 monolayers, we conduct a deep-learning–based quantitative analysis utilizing a substantial dataset of annular dark-field scanning transmission electron microscopy (ADF-STEM) images. We observe predominant mono Se vacancies (VSe), whereas other vacancies and antisite defects are seldom detected. The corresponding electrical and optical properties of WSe2 are meticulously analyzed through a field-effect transistor (FET) device and temperature-dependent photoluminescence (PL) measurements.

Three different methods including CVD, CVT, and flux are employed to synthesize WSe2 samples (see Sec. IV) [Figs. 1(a)1(c)]. Briefly, for the CVD synthesis of WSe2 monolayers, the liquid precursor [ammonium metatungstate hydrate (NH4)6H2W12O40 · xH2O aqueous solution] was spun on a SiO2/Si substrate, followed by selenization at 750 °C for 20 min [Fig. 1(a)]. For the CVT synthesis, W was transported from the hot zone to the cold zone by iodine vaporization via a two-zone furnace.16 W and Se were nucleated within the cold zone, resulting in the synthesis of WSe2 crystals [Fig. 1(b)]. For the flux method, NaCl was used as a flux medium, leveraging its suitable melting point (801 °C) and boiling point (1465 °C). At high temperatures (900–1100 °C), the NaCl flux drenched the W and Se powders to minimize the Se deficiency during WSe2 growth, followed by the precipitation of W and Se during the cooling process [Fig. 1(c)].17 

FIG. 1.

Synthesis of WSe2 samples by CVD, CVT, and flux methods. (a)–(c) Schematic illustrations of the synthesis of monolayer WSe2 and WSe2 bulk crystals through (a) CVD, (b) CVT, and (c) flux methods. (d)–(f) Photographs of (d) CVD-grown monolayer WSe2 film and (e) CVT- and (f) flux-grown WSe2 bulk crystals. (g)–(i) Optical images of monolayer WSe2 stemmed from CVD, CVT, and flux samples, respectively, and (j) corresponding Raman spectra, obtained from the respective blue, red, and green dots in (g)–(i). (k)–(m) Raman mapping images of the full-width-at-half-maximum (FWHM) value of E12g mode.

FIG. 1.

Synthesis of WSe2 samples by CVD, CVT, and flux methods. (a)–(c) Schematic illustrations of the synthesis of monolayer WSe2 and WSe2 bulk crystals through (a) CVD, (b) CVT, and (c) flux methods. (d)–(f) Photographs of (d) CVD-grown monolayer WSe2 film and (e) CVT- and (f) flux-grown WSe2 bulk crystals. (g)–(i) Optical images of monolayer WSe2 stemmed from CVD, CVT, and flux samples, respectively, and (j) corresponding Raman spectra, obtained from the respective blue, red, and green dots in (g)–(i). (k)–(m) Raman mapping images of the full-width-at-half-maximum (FWHM) value of E12g mode.

Close modal

The photograph shows that the CVD method yielded a centimeter-scale WSe2 film, whereas millimeter-scale bulk WSe2 crystals were grown via flux and CVT methods [Figs. 1(d)1(f)]. The flux-grown crystals exhibited typically larger sizes than the CVT-grown WSe2 crystals, owing to the slower crystallization in the flux method. To investigate the electronic and optical properties of the monolayer WSe2 obtained through various synthetic routes, the bulk crystals grown by flux and CVT methods were exfoliated to monolayers on SiO2/Si substrates using a gold-mediated exfoliation method, reaching a few tens of micrometer flakes (see Sec. IV).18 The optical contrasts of the WSe2 monolayers obtained from all three different methods are nearly identical, implying that the thickness of those samples is monolayer [Figs. 1(g)1(i)].

To evaluate the crystal quality of WSe2 flakes, Raman measurements were carried out. The phonon modes of E12g (∼249 cm−1) and A1g (∼258 cm−1) from all three samples were distinctly identified, whereas the B12g mode (∼308 cm−1) was not observed [Fig. 1(j)].19 The absence of the B12g mode suggests that all WSe2 samples are monolayers.20 The full-width-at-half-maximum (FWHM) values of E12g phonon mode for the flux, CVT, and CVD samples were 2.4, 2.5, and 2.7 cm−1, respectively. These variations in FWHM values could be attributed to the slight differences in crystal quality or structural disorder among the samples. The contrast of FWHM values in E12g mode was uniform over the whole region, indicating uniform crystal quality within each sample [Figs. 1(k)1(m)]. Furthermore, strong exciton emission at ∼1.65 eV in PL spectra and a thickness of ∼0.7 nm in AFM images were confirmed in all samples (Figs. S1 and S2 in the supplementary material).19 

Atomic-resolution ADF-STEM images showed that all WSe2 samples exhibited a typical hexagonal lattice structure with W, Se, and occasional Se defects [Figs. 2(a)2(i)]. To improve the image clarity, a deep learning denoising model was used to remove the statistical background noise from the images. The accuracy of this denoising model was proven to be above ∼98%.21 After the sequential processing of 25 ADF-STEM images from each sample to quantify sites, we obtained the histograms describing the defect distributions of the WSe2 samples prepared through three different synthetic routes [Figs. 2(j) and S3–S5 in the supplementary material]. The average content of VSe defects gradually increased from 2.7 to 5.3 × 1013 cm−2 following the order of flux, CVT, and CVD samples. Nevertheless, the VSe content ranged within 0.1%–0.2%, indicating reliable CVD synthesis comparable to CVT or flux synthesis. This is well contrasted with previous reports of large deviations in defect density from synthesis methods.22,23 Other types of defects such as W vacancy (VW) and di Se vacancy (VSe2) were scarcely found. No significant antisite defects (WSe and SeW) were detected, suggesting that their probable content was much lower than the detection limit or simply absent.

FIG. 2.

Quantitative analysis of atomic defects in monolayer WSe2 samples. (a)–(c) (top panel) Denoised ADF-STEM images and (bottom panel) corresponding atomic site mapping of monolayer WSe2 samples grown by each method. The scale bars indicate 2 nm. (d)–(f) Enlarged ADF-STEM images of monolayer WSe2 samples with VW, VSe, and VSe2 in their lattice, respectively. (g)–(i) Comparison of the intensity profiles in the experimental images and simulated ADF-STEM images. Each profile was obtained along the vertical direction marked by the arrows in (d)–(f). (j) VW, VSe, and VSe2 densities and the percentage of VSe in each sample.

FIG. 2.

Quantitative analysis of atomic defects in monolayer WSe2 samples. (a)–(c) (top panel) Denoised ADF-STEM images and (bottom panel) corresponding atomic site mapping of monolayer WSe2 samples grown by each method. The scale bars indicate 2 nm. (d)–(f) Enlarged ADF-STEM images of monolayer WSe2 samples with VW, VSe, and VSe2 in their lattice, respectively. (g)–(i) Comparison of the intensity profiles in the experimental images and simulated ADF-STEM images. Each profile was obtained along the vertical direction marked by the arrows in (d)–(f). (j) VW, VSe, and VSe2 densities and the percentage of VSe in each sample.

Close modal

The electrical properties of each sample were investigated by fabricating back-gate two-terminal FETs using e-beam lithography (see Sec. IV). The transfer characteristics of the monolayer WSe2 FETs manifested clear p-type behaviors in all samples [Fig. 3(a)]. The on-current from the CVD sample was lower than that of the CVT or flux samples but fell within one order of magnitude [Fig. 3(b)], which is in contrast with the 1–2 order of magnitude difference in the previous reports.24–29 This trend is consistent with the relationship between the VSe density variation among the samples. The carrier concentrations were extracted using the parallel capacitance model, n = Cox (VgsVth)/q, where n, Cox, Vgs, Vth, and q represent the respective carrier concentration, gate capacitance, gate voltage, threshold voltage, and charge. The carrier concentration slightly varied within 5–7 × 1012 cm−2 for all samples at a gate voltage (Vgs) of −60 V [Fig. 3(c)]. The field-effect mobility was ∼1.24 cm2 V−1 s−1 from the CVD sample and slightly improves to 3–4 cm2 V−1 s−1 in CVT and flux samples (Table S1 in the supplementary material). Such a small variance in electrical properties indicated that the CVD sample is feasible for further integration of devices on a wafer scale.

FIG. 3.

Electrical properties of monolayer WSe2 grown by CVD, CVT, and flux methods. (a) Current–voltage (I–V) transfer characteristics of monolayer WSe2 FETs. (b) Extracted on-state currents, (c) carrier concentrations, and (d) field-effect mobilities of monolayer WSe2 FETs as a function of VSe density.

FIG. 3.

Electrical properties of monolayer WSe2 grown by CVD, CVT, and flux methods. (a) Current–voltage (I–V) transfer characteristics of monolayer WSe2 FETs. (b) Extracted on-state currents, (c) carrier concentrations, and (d) field-effect mobilities of monolayer WSe2 FETs as a function of VSe density.

Close modal

We conducted temperature-dependent PL measurements for all samples. WSe2 monolayers were transferred onto hexagonal boron nitride (hBN) flakes to prevent charge transfer from the SiO2/Si substrate.30 At room temperature (300 K), only neutral exciton (X0) and trion (XT) with a binding-energy difference of 30 meV were detected in the PL spectra [Figs. 4(a)4(c)]. A slight red-shift of 0.004 eV in the CVD sample, relative to the CVT and flux samples, could be attributed to a p-doping effect introduced by the wet-transfer process. Strong X0 intensities were observed in all samples, but the intensity gradually decreased from flux to CVD sample, consistent with carrier-concentration variation associated with charge screening [Fig. 4(d)]. By contrast, several peaks appeared at a low temperature (10 K) in all samples. The PL spectrum was deconvoluted by six Lorentzian curves, identifying neutral exciton (X0), intervalley trion (XTT, triplet trion), intravalley trion (XST, singlet trion), negatively charged biexciton (XX), and two localized states (L1 and L2) (Fig. S6 in the supplementary material).31–33 These peaks were present in all samples, although with different intensities.

FIG. 4.

Optical properties of monolayer WSe2 grown by CVD, CVT, and flux methods. (a)–(c) Normalized PL spectra measured at room temperature (300 K) and low temperature (10 K) on (a) CVD, (b) CVT, and (c) flux-grown WSe2 monolayers. (d) Intensities of neutral exciton in each sample. (e) Exponent (α) value extracted from the power law (IPα) for each PL characteristic peak, where I and P are the PL intensity and excitation laser power, respectively. (f) Relative spectral weights of localized states (L1 and L2) as a function of VSe density. The spectral weights were extracted by the areal ratio of L1 (or L2) to neutral exciton (X0) from the normalized PL.

FIG. 4.

Optical properties of monolayer WSe2 grown by CVD, CVT, and flux methods. (a)–(c) Normalized PL spectra measured at room temperature (300 K) and low temperature (10 K) on (a) CVD, (b) CVT, and (c) flux-grown WSe2 monolayers. (d) Intensities of neutral exciton in each sample. (e) Exponent (α) value extracted from the power law (IPα) for each PL characteristic peak, where I and P are the PL intensity and excitation laser power, respectively. (f) Relative spectral weights of localized states (L1 and L2) as a function of VSe density. The spectral weights were extracted by the areal ratio of L1 (or L2) to neutral exciton (X0) from the normalized PL.

Close modal

To identify the origin of each emission, the coefficient (α) for the power law (IPα) was obtained from the PL intensity (I) as a function of the excitation power (P) [Figs. 4(e) and S7 in the supplementary material]. For X0, XTT, and XST peaks, a clear linear dependence (α ∼ 1) was observed, suggesting the radiative recombination of excitons and trions. The XX peak, on the other hand, exhibited super-linear behavior at α ∼ 1.5, which is consistent with the previous report.34 Conversely, sub-linear excitonic behavior at α < 1 was observed in L1 and L2 peaks, indicating that these peaks arise from localized states.31 These localized states in monolayer WSe2 could be attributed to crystal imperfections such as defects or strain.35 The relative spectral weights of these localized states were further analyzed by calculating the areal ratio of L1 (and L2) to X0 in the PL spectra [Fig. 4(f)]. In the flux and CVT samples, small spectral weights of 2%–5% were observed for both L1 and L2. On the other hand, the CVD sample showed a higher areal ratio of 22% and 45% for L1 and L2, respectively. The spectral weights for both L1 and L2 localized states increased gradually with elevating VSe density, implying that VSe induces the localized states in PL spectra. Nevertheless, these emissions were only observed at very low temperatures (<60 K), implying that the optical properties of the CVD-grown monolayer WSe2 are comparable to those of CVT and flux-grown WSe2 at room temperature.

In conclusion, we conducted a comparative analysis of the structural, electrical, and optical properties of monolayer WSe2 samples grown by CVD, CVT, and flux methods. Through statistical ADF-STEM analysis, we found that VSe was prevalent in all samples, while VW, VSe2, and antisite defects were rarely observed. The CVD-grown sample exhibited a slightly higher VSe density (5.3 × 1013 cm−2) compared to CVT and flux-grown samples (2–4 × 1013 cm−2). The carrier concentrations, on-currents, field-effect mobilities in FETs, and room-temperature PL intensities of CVD-grown samples were slightly degraded compared to flux and CVT samples but still retained the values within the same order of magnitude. Our findings suggest that the quality of the CVD-grown WSe2 films is currently acceptable for wafer-scale integration in industrial applications such as thin film transistors for display, flexible electronics, and chemical sensors.

0.01M liquid W precursor was prepared by dissolving ammonium metatungstate hydrate powder [(NH4)6H2W12O40 · xH2O, Sigma-Aldrich] in deionized water. This liquid W precursor was spin coated on a 300 nm thick SiO2/Si substrate at 3000 rpm for 30 s. Both the precursor-coated substrate and 0.1 g of Se (229 865, 99.99%, Sigma-Aldrich) were separately loaded into a two-zone furnace. WSe2 was grown in an atmosphere of N2 (99.9999% purity) and H2 (99.999% purity) with respective flow rates of 600 and 15 SCCM at atmospheric pressure. The Se zone temperature was elevated to 400 °C with a ramping rate of 50 °C/min, while the growth substrate zone was heated to 750 °C with a ramping rate of 100 °C/min. After a 20 min growth period, both furnaces were opened and allowed to cool naturally to room temperature.

W, Se, and I (326143, 99.99%, Sigma-Aldrich) powders were mixed in a 1:2:0.3 ratio and then placed in a quartz tube. The quartz tube was then sealed under vacuum at a pressure of 10−5 Torr. For the growth process, the sealed quartz tube was placed in a two-zone furnace. The temperature of the hot zone was set to 1100 °C, and that of the cold zone was set to 1000 °C. These temperature conditions were maintained for one week, after which the system was cooled naturally.

W (357421, 99.99%, Sigma-Aldrich) and Se (229865, 99.99%, Sigma-Aldrich) powders were mixed in a 1:2 ratio with NaCl (57653, 99.5%, Sigma-Aldrich) in an alumina tube. The alumina tube was then sealed in a quartz tube under vacuum at a pressure of 10−5 Torr. The prepared quartz tube was placed in a muffle furnace, and the temperature of the furnace was increased until it reached 1100 °C. This temperature was maintained for 12 h and then reduced at a rate of 0.5 °C/h to 900 °C. After completing all these processes, the quartz tube was taken out of the furnace and immediately quenched in cold water.

Gold-mediated exfoliation was conducted to obtain monolayer WSe2 flakes from CVT and flux-grown WSe2 crystals.18 The WSe2 crystals were attached to a Scotch tape, and a 150 nm-thick Au layer was deposited on the crystals using a thermal evaporator. The Au layer was delaminated from the WSe2 crystals using a thermal release tape. Owing to the strong interaction between Au and chalcogens, monolayer WSe2 flakes were easily separated from the bulk crystal. The Au/WSe2 sample attached to the thermal release tape was transferred onto a SiO2/Si substrate. The thermal release tape was then removed by thermal annealing at ∼130 °C. After the removal of the Au layer using a gold etchant (GE-8111, Transene), monolayer WSe2 flakes were obtained on the SiO2/Si substrate.

Raman and PL measurements were performed using an in-house Raman system with an Ar-ion laser (514 nm wavelength). The measurement was performed under the following conditions: laser power of 32 µW, laser spot size of 1 µm, and exposure time of 0.5 s. The final spectrum was obtained by averaging ten individual spectra. For laser-power-dependent PL measurement, the laser power was varied in the sequence of 2, 4, 8, 16, 32, 64, and 128 µW.

To transfer the WSe2 flakes onto the TEM grids, CVD-grown or exfoliated monolayer WSe2 samples on SiO2/Si substrates were coated with poly(methyl methacrylate) (PMMA C4, MicroChem) and then detached from the substrate by immersion in diluted hydrofluoric acid. The PMMA-supported samples were transferred to the TEM grids (PELCO, 200 mesh, copper, 1.2 µm holes), and the PMMA was removed by dipping in acetone for 5 min. Subsequently, the TEM grid was annealed at 180 °C in a vacuum chamber at ∼7.5 × 10−5 Torr for 12 h to prevent polymerization during STEM imaging. Atomic-resolution images of the WSe2 samples were obtained using the ADF mode in a probe-corrected STEM (ARM 200CF, JEOL Ltd., Japan) at 80 kV and 25 pA probe current, with a 23 mrad semi-convergence angle for the electron probe and an ADF detector collection angle range of 68–280 mrad. The irradiation conditions were verified to ensure that they would not damage the WSe2 structure.21 For the ADF-STEM image simulations, the qSTEM software package based on the multislice method was used with the same microscope parameters as in the experimental condition.36 

The quantitative ADF-STEM analysis was executed using a deep-learning-based algorithm that used two deep neural network models. These models include an image restoration model based on a convolutional neural network and an atomic site-classification model, which is a U-Net backbone model comprising U-Net based fully convolutional neural networks for semantic segmentation.37 It contains an encoding section for feature-map extraction from the input image and a decoding section for organizing these features according to their positions in the image for site classification.38 

WSe2 samples were transferred onto a highly p-doped silicon substrate with a 300 nm thick oxide layer using the PMMA method described above. Metal electrodes for the contacts were patterned by e-beam lithography, followed by the e-beam deposition of Cr/Au (5 nm/50 nm). All electrical experiments were conducted at room temperature within a high-vacuum environment (∼10−6 Torr) utilizing a Keithley 4200 SCS system for measurement. To obtain the carrier concentration, Vth was extracted by linear extrapolation at the maximum slope in transfer curves.39 

See the supplementary material for the representative PL spectra and AFM images of monolayer WSe2, denoised and atomic site mapping ADF-STEM images of CVD, CVT, and flux-grown monolayer WSe2, temperature-dependent PL spectra of flux, CVT, and CVD-grown monolayer WSe2, extraction of α values from the laser-power dependent PL intensity, and electrical properties of monolayer WSe2 grown by flux, CVT, and CVD methods.

S.H.C., K.K.K., and Y.H.L. acknowledge the support from the Institute for Basic Science (Grant No. IBS-R011-D1). K.K.K. acknowledges the support from Basic Science Research (Grant No. 2022R1A2C2091475) and the Next-generation Intelligence Semiconductor Program (Grant No. 2022M3F3A2A01072215) through the National Research Foundation of Korea (NRF), which is funded by the Ministry of Science, ICT & Future Planning. Y.-M.K. acknowledges the support from the NRF (Grant No. NRF-2023R1A2C2002403) funded by the Ministry of Science, ICT & Future Planning in Korea. This work was also supported by the Advanced Facility Center for Quantum Technology.

The authors have no conflicts to disclose.

S.H.C., S.-H.Y., and S.P. contributed equally to this work.

Soo Ho Choi: Conceptualization (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal). Sang-Hyeok Yang: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Validation (equal); Writing – original draft (equal); Writing – review & editing (equal). Sehwan Park: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Validation (equal); Writing – original draft (equal). Byeong Wook Cho: Data curation (equal); Formal analysis (equal). Tuan Dung Nguyen: Data curation (equal); Formal analysis (equal). Jung Ho Kim: Data curation (equal); Formal analysis (equal). Young-Min Kim: Conceptualization (equal); Funding acquisition (equal); Project administration (equal); Resources (equal); Supervision (equal); Writing – original draft (equal); Writing – review & editing (equal). Ki Kang Kim: Conceptualization (equal); Funding acquisition (equal); Project administration (equal); Resources (equal); Supervision (equal); Writing – original draft (equal); Writing – review & editing (equal). Young Hee Lee: Conceptualization (equal); Funding acquisition (equal); Project administration (equal); Resources (equal); Supervision (equal); Writing – original draft (equal); Writing – review & editing (equal).

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

1.
Z. Y.
Zhu
,
Y. C.
Cheng
, and
U.
Schwingenschlogl
,
Phys. Rev. B.
84
(
15
),
153402
(
2011
).
2.
T. C.
Song
,
X. H.
Cai
,
M. W. Y.
Tu
,
X. O.
Zhang
,
B. V.
Huang
,
N. P.
Wilson
,
K. L.
Seyler
,
L.
Zhu
,
T.
Taniguchi
,
K.
Watanabe
,
M. A.
McGuire
,
D. H.
Cobden
,
D.
Xiao
,
W.
Yao
, and
X. D.
Xu
,
Science
360
(
6394
),
1214
1218
(
2018
).
3.
M.
Orlita
,
C.
Faugeras
,
P.
Plochocka
,
P.
Neugebauer
,
G.
Martinez
,
D. K.
Maude
,
A. L.
Barra
,
M.
Sprinkle
,
C.
Berger
,
W. A.
de Heer
, and
M.
Potemski
,
Phys. Rev. Lett.
101
(
26
),
267601
(
2008
).
4.
S.
Kim
,
A.
Konar
,
W. S.
Hwang
,
J. H.
Lee
,
J.
Lee
,
J.
Yang
,
C.
Jung
,
H.
Kim
,
J. B.
Yoo
,
J. Y.
Choi
,
Y. W.
Jin
,
S. Y.
Lee
,
D.
Jena
,
W.
Choi
, and
K.
Kim
,
Nat. Commun.
3
,
1011
(
2012
).
5.
K. F.
Mak
,
C.
Lee
,
J.
Hone
,
J.
Shan
, and
T. F.
Heinz
,
Phys. Rev. Lett.
105
(
13
),
136805
(
2010
).
6.
O.
Lopez-Sanchez
,
D.
Lembke
,
M.
Kayci
,
A.
Radenovic
, and
A.
Kis
,
Nat. Nanotechnol.
8
(
7
),
497
501
(
2013
).
7.
K.
Kang
,
S. E.
Xie
,
L. J.
Huang
,
Y. M.
Han
,
P. Y.
Huang
,
K. F.
Mak
,
C. J.
Kim
,
D.
Muller
, and
J.
Park
,
Nature
520
(
7549
),
656
660
(
2015
).
8.
S.
Boandoh
,
S. H.
Choi
,
J. H.
Park
,
S. Y.
Park
,
S.
Bang
,
M. S.
Jeong
,
J. S.
Lee
,
H. J.
Kim
,
W.
Yang
,
J. Y.
Choi
,
S. M.
Kim
, and
K. K.
Kim
,
Small
13
(
39
),
1701306
(
2017
).
9.
S. H.
Choi
,
H. J.
Kim
,
B.
Song
,
Y. I.
Kim
,
G.
Han
,
H. T. T.
Nguyen
,
H.
Ko
,
S.
Boandoh
,
J. H.
Choi
,
C. S.
Oh
,
H. J.
Cho
,
J. W.
Jin
,
Y. S.
Won
,
B. H.
Lee
,
S. J.
Yun
,
B. G.
Shin
,
H. Y.
Jeong
,
Y. M.
Kim
,
Y. K.
Han
,
Y. H.
Lee
,
S. M.
Kim
, and
K. K.
Kim
,
Adv. Mater.
33
(
15
),
2006601
(
2021
).
10.
S. H.
Choi
,
S. J.
Yun
,
Y. S.
Won
,
C. S.
Oh
,
S. M.
Kim
,
K. K.
Kim
, and
Y. H.
Lee
,
Nat. Commun.
13
(
1
),
1484
(
2022
).
11.
J. K.
Baidoo
,
S. H.
Choi
,
F. O. T.
Agyapong-Fordjour
,
S.
Boandoh
,
S. J.
Yun
,
L. A.
Adofo
,
A.
Ben-Smith
,
Y. I.
Kim
,
J. W.
Jin
,
M. H.
Jung
,
H. Y.
Jeong
,
Y. M.
Kim
,
Y. H.
Lee
,
S. M.
Kim
, and
K. K.
Kim
,
ACS Nano
16
(
6
),
8851
8859
(
2022
).
12.
H.
Wang
,
L.
Yu
,
Y. H.
Lee
,
W.
Fang
,
A.
Hsu
,
P.
Herring
,
M.
Chin
,
M.
Dubey
,
L. J.
Li
,
J.
Kong
, and
T.
Palacios
, in
2012 IEEE International Electron Devices Meeting (IEDM)
(
IEEE
,
2012
), pp.
4.6.1
4.6.4
.
13.
Y. R.
Lim
,
W.
Song
,
J. K.
Han
,
Y. B.
Lee
,
S. J.
Kim
,
S.
Myung
,
S. S.
Lee
,
K. S.
An
,
C. J.
Choi
, and
J.
Lim
,
Adv. Mater.
28
(
25
),
5025
5030
(
2016
).
14.
H.
Wang
,
L. L.
Yu
,
Y. H.
Lee
,
Y. M.
Shi
,
A.
Hsu
,
M. L.
Chin
,
L. J.
Li
,
M.
Dubey
,
J.
Kong
, and
T.
Palacios
,
Nano Lett.
12
(
9
),
4674
4680
(
2012
).
15.
M.
Lanza
,
Q.
Smets
,
C.
Huyghebaert
, and
L. J.
Li
,
Nat. Commun.
11
(
1
),
5689
(
2020
).
16.
J. D.
Franolic
,
J. R.
Long
, and
R. H.
Holm
,
J. Am. Chem. Soc.
117
(
31
),
8139
8153
(
1995
).
17.
D. H.
Keum
,
S.
Cho
,
J. H.
Kim
,
D.-H.
Choe
,
H.-J.
Sung
,
M.
Kan
,
H.
Kang
,
J.-Y.
Hwang
,
S. W.
Kim
,
H.
Yang
,
K. J.
Chang
, and
Y. H.
Lee
,
Nat. Phys.
11
(
6
),
482
486
(
2015
).
18.
S. B.
Desai
,
S. R.
Madhvapathy
,
M.
Amani
,
D.
Kiriya
,
M.
Hettick
,
M.
Tosun
,
Y.
Zhou
,
M.
Dubey
,
J. W.
Ager
III
,
D.
Chrzan
, and
A.
Javey
,
Adv. Mater.
28
(
21
),
4053
4058
(
2016
).
19.
R.
Zhang
,
D.
Drysdale
,
V.
Koutsos
, and
R.
Cheung
,
Adv. Funct. Mater.
27
(
41
),
1702455
(
2017
).
20.
P.
Tonndorf
,
R.
Schmidt
,
P.
Böttger
,
X.
Zhang
,
J.
Borner
,
A.
Liebig
,
M.
Albrecht
,
C.
Kloc
,
O.
Gordan
,
D. R. T.
Zahn
,
S.
Michaelis de Vasconcellos
, and
R.
Bratschitsch
,
Opt. Express
21
(
4
),
4908
4916
(
2013
).
21.
S. H.
Yang
,
W.
Choi
,
B. W.
Cho
,
F. O. T.
Agyapong-Fordjour
,
S.
Park
,
S. J.
Yun
,
H. J.
Kim
,
Y. K.
Han
,
Y. H.
Lee
,
K. K.
Kim
, and
Y. M.
Kim
,
Adv. Sci.
8
(
16
),
2101099
(
2021
).
22.
D.
Edelberg
,
D.
Rhodes
,
A.
Kerelsky
,
B.
Kim
,
J.
Wang
,
A.
Zangiabadi
,
C.
Kim
,
A.
Abhinandan
,
J.
Ardelean
,
M.
Scully
,
D.
Scullion
,
L.
Embon
,
R.
Zu
,
E. J. G.
Santos
,
L.
Balicas
,
C.
Marianetti
,
K.
Barmak
,
X.
Zhu
,
J.
Hone
, and
A. N.
Pasupathy
,
Nano Lett.
19
(
7
),
4371
4379
(
2019
).
23.
Z.
Zhang
,
H.
Liang
,
L.
Loh
,
Y.
Chen
,
Y.
Chen
,
K.
Watanabe
,
T.
Taniguchi
,
S. Y.
Quek
,
M.
Bosman
,
A. A.
Bettiol
, and
G.
Eda
,
Adv. Opt. Mater.
10
(
23
),
2201350
(
2022
).
24.
Z.
Wang
,
Y.
Nie
,
H.
Ou
,
D.
Chen
,
Y.
Cen
,
J.
Liu
,
D.
Wu
,
G.
Hong
,
B.
Li
,
G.
Xing
, and
W.
Zhang
,
Nanomaterials
13
(
8
),
1368
(
2023
).
25.
B.
Liu
,
Y.
Ma
,
A.
Zhang
,
L.
Chen
,
A. N.
Abbas
,
Y.
Liu
,
C.
Shen
,
H.
Wan
, and
C.
Zhou
,
ACS Nano
10
(
5
),
5153
5160
(
2016
).
26.
B.
Liu
,
M.
Fathi
,
L.
Chen
,
A.
Abbas
,
Y.
Ma
, and
C.
Zhou
,
ACS Nano
9
(
6
),
6119
6127
(
2015
).
27.
H. C. P.
Movva
,
A.
Rai
,
S.
Kang
,
K.
Kim
,
B.
Fallahazad
,
T.
Taniguchi
,
K.
Watanabe
,
E.
Tutuc
, and
S. K.
Banerjee
,
ACS Nano
9
(
10
),
10402
10410
(
2015
).
28.
Y.
Liu
,
S.
Liu
,
Z.
Wang
,
B.
Li
,
K.
Watanabe
,
T.
Taniguchi
,
W. J.
Yoo
, and
J.
Hone
,
Nat. Electron.
5
(
9
),
579
585
(
2022
).
29.
Y.
Wang
,
J. C.
Kim
,
Y.
Li
,
K. Y.
Ma
,
S.
Hong
,
M.
Kim
,
H. S.
Shin
,
H. Y.
Jeong
, and
M.
Chhowalla
,
Nature
610
(
7930
),
61
66
(
2022
).
30.
J. M.
Xue
,
J.
Sanchez-Yamagishi
,
D.
Bulmash
,
P.
Jacquod
,
A.
Deshpande
,
K.
Watanabe
,
T.
Taniguchi
,
P.
Jarillo-Herrero
, and
B. J.
Leroy
,
Nat. Mater.
10
(
4
),
282
285
(
2011
).
31.
J. N.
Huang
,
T. B.
Hoang
, and
M. H.
Mikkelsen
,
Sci. Rep.
6
,
22414
(
2016
).
32.
Z. P.
Li
,
T. M.
Wang
,
Z. G.
Lu
,
C. H.
Jin
,
Y. W.
Chen
,
Y. Z.
Meng
,
Z.
Lian
,
T.
Taniguchi
,
K.
Watanabe
,
S. B.
Zhang
,
D.
Smirnov
, and
S. F.
Shi
,
Nat. Commun.
9
,
3719
(
2018
).
33.
E.
Courtade
,
M.
Semina
,
M.
Manca
,
M. M.
Glazov
,
C.
Robert
,
F.
Cadiz
,
G.
Wang
,
T.
Taniguchi
,
K.
Watanabe
,
M.
Pierre
,
W.
Escoffier
,
E. L.
Ivchenko
,
P.
Renucci
,
X.
Marie
,
T.
Amand
, and
B.
Urbaszek
,
Phys. Rev. B
96
(
8
),
085302
(
2017
).
34.
M.
Barbone
,
A. R. P.
Montblanch
,
D. M.
Kara
,
C.
Palacios-Berraquero
,
A. R.
Cadore
,
D.
De Fazio
,
B.
Pingault
,
E.
Mostaani
,
H.
Li
,
B.
Chen
,
K.
Watanabe
,
T.
Taniguchi
,
S.
Tongay
,
G.
Wang
,
A. C.
Ferrari
, and
M.
Atature
,
Nat. Commun.
9
,
3721
(
2018
).
35.
S.
Tongay
,
J.
Suh
,
C.
Ataca
,
W.
Fan
,
A.
Luce
,
J. S.
Kang
,
J.
Liu
,
C.
Ko
,
R.
Raghunathanan
,
J.
Zhou
,
F.
Ogletree
,
J. B.
Li
,
J. C.
Grossman
, and
J. Q.
Wu
,
Sci. Rep.
3
,
2657
(
2013
).
36.
G.
Ghimire
,
K. P.
Dhakal
,
W.
Choi
,
Y. A.
Esthete
,
S. J.
Kim
,
T. T.
Tran
,
H.
Lee
,
H.
Yang
,
D. L.
Duong
,
Y. M.
Kim
, and
J.
Kim
,
ACS Nano
15
(
8
),
13770
13780
(
2021
).
37.
K.
Zhang
,
W.
Zuo
,
S.
Gu
, and
L.
Zhang
, in
IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
,
Honolulu, HI
,
21–26 July 2017
.
38.
O.
Ronneberger
,
P.
Fischer
, and
T.
Brox
, in
International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI)
,
Munich, Germany
,
5–9 October 2015
.
39.
F. J.
GarcõÂa SaÂnchez
,
A.
Ortiz-Conde
,
G.
De Mercato
,
J. A.
Salcedo
,
J. J.
Liou
,
Y.
Yue
, and
J.
Finol
, in
Proceedings of the XIV Design of Circuits and Integrated Systems Conference
,
Palma de Mallorca, Spain
,
1999
.

Supplementary Material