Ultrasmall and Monolayered Tungsten Dichalcogenide Quantum Dots with Giant Spin-Valley Coupling and Purple Luminescence.

Monolayered tungsten dichalcogenide quantum dots (WS2 QDs) have various potential applications due to their large spin-valley coupling and excellent photoluminescence (PL) properties. What is expected is that with the decrease in lateral size of QDs, the stronger quantum confinement effect will dramatically strengthen the spin-valley coupling and widen the band gap. However, ultrasmall monolayered WS2 QDs prepared by ion intercalation unavoidably undergo the problem of structural defects, which will create defect levels and significantly change their properties. In this study, we report that by annealing defective monolayered WS2 QDs in sulfur vapor, pristine monolayered WS2 QDs with an ultrasmall lateral size of ca. 1.8-3.8 nm can be obtained. The results show that the ultrasmall monolayered WS2 QDs exhibit a giant spin-valley coupling of ca. 821 meV. Moreover, the pristine ultrasmall monolayered WS2 QDs show purple PL centered at 416 nm, and the defect PL peaks in defective WS2 QDs can be effectively removed by annealing. All of these results afford the ultrasmall monolayered QDs various applications such as in optoelectronics, spintronics, valleytronics, and so on.

Introduction

Two-dimensional (2D) semiconducting transition metal dichalcogenides (TMDs) have aroused great interest recently due to a wide range of physical and chemical properties.[1−6] The structure of monolayered TMDs is MX2 (M = Mo or W and X = S, Se, or Te), consisting of an intermediate layer of M atoms sandwiched between two layers of X atoms with strong ionic–covalent bonds.[4] Recently, 2D TMDs with large spin–valley coupling, which will introduce a strong coupling between the spin and valley degrees of freedom, have aroused continual and tremendous interest for potential applications in spintronics, valleytronics, and so on.[7−10] For example, monolayered MoS2 and MoSe2 have been confirmed to own the spin–valley coupling of ca. 150 and 180 meV,[9,11] respectively. However, the applications are hampered due to the modest spin splitting, which necessitates superpurity materials, low temperature, and the long channel lengths.[12,13]

Interestingly, due to the strong spin–orbit coupling (SOC) in 5d orbitals of tungsten and the broken inversion symmetry, monolayered WS2 sheets exhibit a much larger spin–valley coupling of up to 426 meV.[12,14,15] Moreover, previous studies have shown an indirect-to-direct band gap transition in monolayered WS2 sheets due to the missing interlayer interaction.[5,6] Therefore, the large spin–valley coupling combined with the presence of direct band gap in monolayered WS2 sheets affords them different applications in optoelectronics, spintronics, valleytronics, and so on.[16−19] By further reducing the lateral size of monolayered WS2 sheets, monolayered WS2 quantum dots (QDs) can be obtained. Normally, the WS2 sheets whose lateral size is under 100 nm can be named as WS2 quantum dots (QDs).[20,23,24] The monolayered WS2 QDs can exhibit more striking properties than monolayered sheets due to the strong quantum confinement effect.[20−24] In general, with the decrease in lateral size of QDs, the stronger quantum confinement effect will dramatically strengthen the spin–valley coupling and widen the band gap.[20,21,25,26] Therefore, investigations of the spin–valley coupling and photoluminescence (PL) of ultrasmall monolayered WS2 QDs are favorable to both the fundamental study and their applications.

In this article, we show that the pristine monolayered WS2 QDs with the ultrasmall lateral size of ca. 1.8–3.8 nm and purely semiconducting phase can be obtained by annealing Li-intercalated WS2 QDs in sulfur vapor. The results show that the pristine ultrasmall monolayered WS2 QDs exhibit a giant spin–valley coupling of ca. 821 meV and two clear peaks at 416 and 342 nm in PL spectra. Compared with the monolayered WS2 sheets and QDs with general lateral size, the ultrasmall monolayered WS2 QDs show significant improvement in spin–valley coupling and a great blue shift in PL, making them highly valuable for optoelectronics, spintronics, valleytronics, and so on.

Furthermore, it should be noted that so far the mass production of ultrasmall monolayered WS2 QDs is mainly realized by Li-ion intercalation.[27,28] However, the as-prepared QDs generally have plenty of structural defects such as high oxidation state and plenty of sulfur vacancies, which will unavoidably create defect levels in the electronic band.[20] In addition, there are both metallic phase and semiconducting phase in these WS2 QDs.[2,29,30] The facts that the structural defects in our as-prepared QDs have been restored and the metallic phase has been discarded after annealing in sulfur vapor confirm the effectiveness of our method for the preparation of pristine WS2 QDs with a purely semiconducting phase.

Results and Discussion

Figure a shows the transmission electron microscopy (TEM) image of as-prepared WS2 QDs (a-WS2 QDs). Particle size analysis shows that the diameter distribution of these dots is approximately 2–4 nm with an average lateral size of ca. 3.2 nm (inset of Figure a). The lattice spacing is 0.22 nm (inset of Figure a), stemming from the (103) plane.[31] After annealing, the diameter distribution is approximately 1.8–3.8 nm with an ultrasmall average lateral size ca. 2.9 nm (inset of Figure b). The high-resolution transmission electron microscopy (HRTEM) image presents a clear lattice fringe with lattice spacing of 0.22 nm (inset of Figure b), similar to that of a-WS2 QDs. Shown in Figure c,d are the atomic force microscopy (AFM) images of a-WS2 QDs and annealed WS2 QDs. The topographic heights of a-WS2 QDs and annealed WS2 QDs are ca. 0.5 and 0.4 nm, respectively (insets of Figure c,d), indicating that all of these quantum dots are monolayered. All of the results imply that annealing did not significantly change the size and thickness of a-WS2 QDs.

Figure 1

TEM images of (a) a-WS2 QDs and (b) annealed WS2 QDs. The lower left insets show diameter distribution, and the upper right insets show HRTEM images. AFM images of (c) a-WS2 QDs and (d) annealed WS2 QDs. Insets are the corresponding height profiles of the white lines.

X-ray photoelectron spectroscopy (XPS) measurements were performed to investigate the chemical environment of a-WS2 QDs and annealed WS2 QDs. Shown in Figure a are the XPS survey spectra over a range of binding energies of a-WS2 QDs and annealed WS2 QDs. As found, there is a striking oxygen XPS peak for a-WS2 QDs, similar to the reported results.[22,32−34] This indicates a massive number of oxygen groups in a-WS2 QDs. By contrast, the oxygen peak almost disappeared completely for the annealed sample after annealing. The contents of elements W, S, and O, respectively, are 24.2, 42.1, and 33.7% for a-WS2 QDs and are, respectively, 31.5, 64.7, and 3.8% for the annealed sample. Clearly, the content of O shows a drastic decrease from 33.7% to only 3.8% after annealing, demonstrating further that oxygen groups in a-WS2 QDs have been effectively removed after annealing. It is considered that a massive number of oxygen groups in a-WS2 QDs mainly results from oxidation of water during the process of chemical exfoliation.[28] Reasonably, the infinitesimal oxygen content in annealed WS2 QDs may arise from the physical absorption of CO2 and O2 on the surface of QDs. Accordingly, one can calculate the S/W atomic ratio, which is ca. 1.74 in a-WS2 QDs. The ratio is much lower than the stoichiometric ratio of 2 in perfect monolayered WS2, implying the formation of sulfur vacancies during chemical exfoliation, which may be attributed to the partial oxidization of W. However, the calculated ratio of ca. 2.05 for the annealed sample is very close to 2, suggesting that the oxygen groups in a-WS2 QDs can be effectively removed and the sulfur vacancies can be well repaired after annealing.

Figure 2

(a) XPS survey spectra over a range of binding energies (0–600 eV) and (b) high-resolution W 4f spectra of a-WS2 QDs and annealed WS2 QDs.

Figure b shows the W 4f fine-scan spectra of the as-prepared and annealed samples. After the subtraction of Shirley background, the spectra were well fitted with the Gaussian–Lorentzian sum function (80% Lorentzian and 20% Gaussian). The spectrum of a-WS2 QDs can be deconvoluted to seven subpeaks: 1T-W4+ 4f7/2 (31.96 eV), 1T-W4+ 4f5/2 (34.06 eV), 2H-W4+ 4f7/2 (32.76 eV), 2H-W4+ 4f5/2 (34.86 eV), W6+ 4f7/2 (35.68 eV), W6+ 4f5/2 (37.80 eV), and W 5p3/2 (38.47 eV), similar to previous works.[32,35] After annealing, there are only three subpeaks at 32.92, 35.10, and 38.62 eV, corresponding to 2H-W4+ 4f7/2, 2H-W4+ 4f5/2, and W4+ 5p3/2 core energy levels, respectively. The high valence state of W6+ is attributed to the oxidation states in a-WS2 QDs,[35,36] in good agreement with the high content of the bonded oxygen. However, the absence of W6+ peaks in the annealed sample implies that the hexavalent tungsten has been completely converted to a tetravalent form. Moreover, it is known that there is a metallic 1T phase (octahedral) in WS2 QDs obtained by Li intercalation, which will be transferred to the semiconducting 2H phase (trigonal prismatic) after annealing.[2,29,30] Obviously, the 1T-W4+ 4f7/2 and 1T-W4+ 4f5/2 peaks completely disappear in annealed WS2 QDs, revealing the exhaustive phase-transition from 1T to 2H. Therefore, semiconducting WS2 QDs with a high-purity 2H phase are obtained.

To further investigate the structural changes before and after annealing, Raman measurements were performed on the as-prepared and annealed samples. Considering that the Raman signals of WS2 QDs are much weaker than those of bulk WS2,[37] 532 nm excitation wavelength was used to enrich the spectra, which would reveal many second-order peaks.[6,38,39] As shown in Figure , the first-order modes LA(M) and A1g(Γ) are located at 176 and 418 cm–1, respectively. The strongest peak at 352.5 cm–1 is attributed to the second-order mode 2LA(M) at 351.5 cm–1 and the first-order mode E2g1(Γ) at 355.8 cm–1.[38] It is found that the A1g(Γ) mode weakened after annealing. Therefore, we carried out the Gaussian–Lorentzian fitting to separate their individual contributions and calculate the relative intensity I2LA/IA1g. The intensity ratios of a-WS2 QDs and annealed WS2 QDs, respectively, are 1.5 and 2.3, which further indicate that both of the samples are monolayered.[38] One can find that the increase in the intensity ratio from 1.5 to 2.3 attributes to the small reduction of the A1g(Γ) mode. Considering the existence of high content of oxygen groups in a-WS2 QDs, those oxygen groups bonded to tungsten or sulfur atoms are more likely to increase the out-of-plane phonon restoring force, which will lead to the enhancement of the A1g(Γ) mode.[38] After annealing, oxygen groups are wiped off and thus the A1g(Γ) mode decreases.

Figure 3

Raman spectra obtained with 532 nm excitation. Colorful lines are the measurements, and black lines are Gaussian–Lorentzian peak fits.

All of these results indicate that by annealing of defective a-WS2 QDs, one can (i) restore their structural defects and (ii) discard the residual metallic 1T phase. Our monolayered WS2 QDs with an ultrasmall lateral size of ca. 1.8–3.8 nm are favorable to investigate the intrinsic properties of pristine, ultrasmall, and monolayered WS2 QDs with a purely semiconducting phase.

To investigate the optical properties of WS2 QDs before and after annealing, we measured the UV–vis extinction spectra. As known, there are in general three absorption peaks at A (625 nm), B (550 nm), and C (450 nm) in monolayered WS2 sheets.[40] The A and B peaks were considered to originated from two kinds of transitions from the spin-splitting valence band to the conduction band at the K point of the Brillouin zone,[12,40] whereas the absorption peak C may be attributed to the optical transitions between the density-of-state peaks in the valence and conduction bands.[40,41]Figure a shows the UV–vis extinction spectra of a-WS2 QDs and annealed WS2 QDs. It is found that greatly different from the case of a-WS2 QDs where no clear absorption peak can be seen, there are three clear absorption peaks at ca. 379, 303, and 269 nm in annealed WS2 QDs, which can be assigned, respectively, to the excitonic absorption peaks A, B, and C. The difference between the spectral shape of a-WS2 QDs and annealed WS2 QDs may show the phase transition from 1T to 2H.[42] The excitonic transitions of A (379 nm) and B (303 nm) are clearly recognizable in the extinction spectrum of annealed WS2 QDs (2H phase), whereas they are absent in a-WS2 QDs with a high content of 1T phase. Moreover, due to a large proportion of the 1T metallic phase, the extinction spectrum of a-WS2 QDs exhibits a broader absorption across a wide spectral region. Clearly, the A and B absorption peaks are blue-shifted by ca. 14 and 30 nm, respectively, with respect to those of QDs with a lateral size of 8–15 nm, as reported in refs (20). As reported, due to the quantum confinement and edge effects, the blue shift in extinction spectra may appear in QDs with a decrease in the lateral size.[37,43,44] Reasonably, the ultrasmall lateral size of our WS2 QDs may be responsible for the observed blue shift. As known, 2D materials with spin–valley coupling have a pair of degenerated valleys at the high-symmetry K and K′ points of the hexagonal Brillouin zone.[13,14] The spin-splitting valence bands V1 and V2 are schematically illustrated in Figure b. The presence of distinct A and B absorption peaks in annealed WS2 QDs may also suggest that their electronic band structures become more intrinsic after annealing. One can find that the direct band gap labeled according to absorption peak A, 393 nm (3.27 eV), is much larger than that of monolayered WS2 sheets (1.98 eV) and QDs (3.16 eV) with the general lateral size, as reported in ref[20, 40]. Reasonably, the wider band gap of our QDs may be attributed to the ultrasmall size due to the stronger quantum confinement effect.[20,21,25,26] Most importantly, it is found that the great energy difference (ΔSO) between A and B excitonic absorption peaks in annealed WS2 QDs is large, up to ca. 821 meV, implying the appearance of a giant spin–valley coupling. As reported, monolayered MoS2 and MoSe2 have spin–valley couplings of ca. 150 and 180 meV,[9,11] respectively. By contrast, due to the strong SOC in 5d orbitals of tungsten, monolayered WS2 sheets exhibit a much larger spin–valley coupling of about 426 meV.[12,14,15,40] Notably, this value in our ultrasmall monolayered WS2 QDs is much larger than 570 meV for general WS2 QDs with a lateral size of ∼8–15 nm.[20]

Figure 4

(a) UV–vis extinction spectra of a-WS2 QDs and annealed WS2 QDs. (b) Diagram of the band structure of pristine monolayered WS2 QDs near the K point.

To further study the optical properties of WS2 QDs before and after annealing, we measured the PL spectra. As shown in Figure a,b, the PL spectra of a-WS2 QDs and annealed WS2 QDs were measured with excitation wavelength from 290 to 350 nm. Both of them exhibit an excitation-dependent PL behavior. It is found that with the increase of the excitation wavelength from 290 to 350 nm, the intensity of the main PL peak increases first and then decreases. On excitation at 310 nm, the PL spectra show a strongest peak at 416 nm. PL excitation (PLE) spectra of the two samples were investigated with a detection wavelength of 416 nm, which was determined by PL investigations of a-WS2 QDs and annealed WS2 QDs with various excitation wavelengths (Figure a,b). It should be noted that the influence of the N-methyl-2-pyrrolidone (NMP) solvent on PL spectra has been eliminated (as illustrated in Figures S1 and S2). Shown in Figure c are the normalized PLE spectra with the detection wavelength of 416 nm. The same peak position at 315 nm was detected in both samples, which is considered to be the strongest excitation wavelength for both samples.

Figure 5

PL spectra of (a) a-WS2 QDs and (b) annealed WS2 QDs with various excitation wavelengths. (c) Normalized PLE spectra with the detection wavelength of 416 nm and (d) normalized PL spectra at the excitation wavelength of 315 nm of a-WS2 QDs and annealed WS2 QDs.

The PL properties of monolayered WS2 sheets and WS2 QDs with the general lateral size have been well investigated. As reported, there is one strong emission peak A at ca. 638 nm and a weak emission peak B at ca. 530 nm for WS2 sheets[6,16,40] and one strong emission peak A at ca. 461 nm and a weak emission peak B at ca. 369 nm for WS2 QDs with the general lateral size of 8–15 nm.[20]Figure d shows the normalized PL spectra of a-WS2 QDs and annealed WS2 QDs using the excitation wavelength of 315 nm. For the case of a-WS2 QDs, the highest emission peak A is located at ca. 420 nm. It is found that there are only two distinct emission peaks at 416 and 342 nm for the annealed WS2 QDs, corresponding to A and B emission peaks. It is clear that compared with the general WS2 QDs our ultrasmall QDs show a great blue shift in both A and B emission peaks, respectively, of ca. 45 and 27 nm in the PL spectrum.

Notably, it is found that there is an obvious peak at 466 nm (2.66 eV) and several other shoulder emission peaks at greater than 420 nm in the as-prepared QDs, similar to the results reported in ref (20). These shoulder emission peaks were considered to originate from the transitions from the conduction band to the new defect levels above the valence band (as illustrated in Figure b).[20] It is clear that defect-related emission peaks in a-WS2 QDs were eliminated due to the restoration of an intrinsic band structure. Moreover, one can find that broad peaks in a-WS2 QDs were narrowed after restoration (Figure d). All of these results indicate further the structural restoration of a-WS2 QDs after annealing, leading to the perfect structure with an intrinsic electronic band.

As mentioned above, the strong quantum confinement effect will dramatically strengthen the spin–valley coupling and increase the band gap of QDs.[20,21,25,26] Therefore, one can propose that both the giant energy difference (ΔSO) and the purple PL in our annealed WS2 QDs may be attributed to the ultrasmall lateral size. We should note that there is no clear spin–orbit splitting observed in a-WS2 QDs. As reported, structural defects in monolayered TMDs would induce intervalley electronic scattering and defects are the common limiting factors that can provide the required momentum for intervalley scattering due to their short-range nature.[45] Therefore, even if the giant spin–valley coupling is in pristine WS2 QDs due to the suppression in the intervalley scattering, the features expected from the valence band may be absent due to the high-density defects in defective a-WS2 QDs. As we noted, the defect PL peaks have been removed completely, which may lead to the shift of the highest main PL emission position. Compared with WS2 QDs with the general lateral size, our WS2 QDs with the ultrasmall lateral size have stronger spin–valley coupling and greatly blue-shifted emission, which afford them high potential for various applications.

Conclusions

In conclusion, we have synthesized pristine, monolayered, and semiconducting WS2 QDs with an ultrasmall lateral size of ca. 1.8–3.8 nm by annealing defective WS2 QDs in sulfur vapor. The results showed that after annealing (i) a giant spin–valley coupling up to ca. 821 meV appeared; (ii) the ultrasmall monolayered WS2 QDs showed purple PL centered at 416 nm, and the defect-related areas in PL spectra of defective WS2 QDs disappeared. The fact that the excellent properties of pristine QDs are absent in the as-prepared ones suggests the necessity of sulfur annealing. Compared with the general WS2 sheets or QDs reported, the ultrasmall monolayered WS2 QDs show a significant improvement in spin–valley coupling and a great blue shift in PL. The spin–valley coupling and purple PL are obtained with the ultrasmall lateral size of our pristine WS2 QDs. All of the results afford these ultrasmall monolayered QDs various applications in optoelectronics, spintronics, valleytronics, and so on.

Methods

Materials

a-WS2 QDs were purchased from Nanjing XFNANO Materials Tech Co., Ltd. Sulfur powder (99.998% purity) was purchased from Sigma-Aldrich. N-Methyl-2-pyrrolidone (NMP) was purchased from Nanjing Chemical Reagent Co., Ltd.

Sample Preparation

a-WS2 QDs (100 mg) and 1 g of sulfur powder were put in two quartz boats, respectively. They were then placed inside a quartz tube in sequence with sulfur powder on the upstream of the gas flow. Before heating in a tubular furnace, the tube was emptied by a vacuum pump and fed with Ar several times to ensure that there is no residual oxygen. After that, the restoration process was performed at 450 °C for 30 min with Ar flow at 20 sccm. Finally, annealed WS2 QDs were collected after cooling down to room temperature.

Sample Characterizations

The morphologies were investigated by transmission electron microscopy (TEM, FEI Tecnai-F20) and atomic force microscopy (AFM, Veeco dimension 3100). X-ray photoelectron spectroscopy (XPS) measurements were performed on PHI-5000 VersaProbe using Al Kα radiation. Raman spectra were obtained by a confocal Raman microscope (LabRAM Aramis, Japan) using a laser excitation of 532 nm. The extinction and PL spectra were recorded at ambient conditions by an ultraviolet spectrophotometer (Shimadzu UV-3600, Japan) and fluorescence spectrophotometer (Shimadzu RF-5301PC, Japan), respectively. For optical spectrum investigation, samples were prepared at a concentration of 0.1 mg/mL using NMP and were ultrasonically dispersed for 1 h.