Redox Exfoliation of Layered Transition Metal Dichalcogenides - ACS

Dec 29, 2016 - For example, B2g is observed in MoSe2 (353 cm–1) and WSe2 (301 cm–1).(38, 39). UV–vis extinction measurements of the dispersed la...
0 downloads 9 Views 9MB Size
Redox Exfoliation of Layered Transition Metal Dichalcogenides Ali Jawaid, Justin Che, Lawrence F. Drummy, John Bultman, Adam Waite, Ming-Siao Hsiao, and Richard A. Vaia* Materials and Manufacturing Directorate, Air Force Research Laboratory, Wright-Patterson AFB, Ohio 45433-7702, United States S Supporting Information *

ABSTRACT: Transition metal dichalcogenides (TMDs) have attracted considerable attention in a diverse array of applications due to the breadth of possible property suites relative to other lowdimensional nanomaterials (e.g., graphene, aluminosilicates). Here, we demonstrate an alternative methodology for the exfoliation of bulk crystallites of group V−VII layered TMDs under quiescent, benchtop conditions using mild redox chemistry. Anionic polyoxometalate species generated from edge sites adsorb to the TMD surface and create Coulombic repulsion that drives layer separation without the use of shear forces. This method is generalizable (MS2, MSe2, and MTe2) and effective in preparing high-concentration (>1 mg/mL) dispersions with narrow layer thickness distributions more rapidly and with safer reagents than alternative solution-based approaches. Finally, exfoliation of these TMDs is demonstrated in a range of solvent systems that were previously inaccessible due to large surface energy differences. These characteristics could be beneficial in the preparation of high-quality films and monoliths. KEYWORDS: transition metal dichalcogenides, MoS2, redox exfoliation, surface oxidation, surface reactivity, liquid-phase exfoliation

L

exfoliation. For example, MoS2 dispersions of few-layer to monolayer flakes at concentrations up to 3 mg/mL can be formed quiescently in both protic (ethanol, methanol) and aprotic solvents (acetonitrile, acetone). Furthermore, we demonstrate that this method is general and offers a low-cost, low-energy, comprehensive framework for exfoliation of MS2-, MSe2-, and MTe2-layered TMDs from groups IV, V, VI, and VII. Traditional chemical exfoliation strategies exploit the ability of the interlayer gap to accommodate uncharged polar molecules or cationic guest species.6−9 Intercalate mobility, however, depends on the electronic structure of the layered TMD, requiring strong alkali earth or organometallic species for effective penetration.8 Thus, these processes require inert atmosphere and additional engineering controls to accommodate the reactivity of the intercalants. Further, due to phase changes during guest intercalation (i.e., semiconducting 2H to metallic 1T),10 additional restorative annealing steps are often required after exfoliation to generate single-domain materials. Recently, liquid-phase exfoliation (LPE) has emerged as an alternative approach to prepare large quantities of few-layer

ayered transition metal dichalcogenides (TMDs) are an emerging class of two-dimensional materials due to their diverse property suite, which ranges from semiconducting and semimetallic to metallic and superconducting.1 Their lamellar structure consists of a transition metal layer (M) sandwiched between two chalcogen layers (X) with strong M− X intralayer bonding. These layers are separated by a weak van der Waals gap. At the few-layer to monolayer limit, coupling between layers is reduced, in-plane confinement dominates, and the band structure changes. This affords opportunities for chemical sensing, catalysis, spintronics, single-photon emission, infrared optics, nanocomposites, coatings, and printable inks for nanoelectronics.2−5 Thus, a suite of processing methods has evolved to satisfy application-specific requirements, such as defect density, scale, cost, and integration. Top-down exfoliation of powders affords large volume use and compliments molecular-based monolayer growth for microfabricated devices. Although various methods for solution exfoliation of powders have been developed, reproducibility and scale-up in a broad array of solvents using limited amounts of surfactant stabilizers have been issues due to an incomplete understanding of the general organic−inorganic chemistry of layered TMDs. Herein, we address this challenge by demonstrating the use of mild oxidants to generate peroxometalate precursors from the TMDs. These form anionic polyoxometalates (POMs) in situ, which adsorb to the surface of the TMD and result in © 2016 American Chemical Society

Received: October 13, 2016 Accepted: December 29, 2016 Published: December 29, 2016 635

DOI: 10.1021/acsnano.6b06922 ACS Nano 2017, 11, 635−646

Article

www.acsnano.org

Article

ACS Nano MoS2 and WS 2 flakes.11 Notably, metastable colloidal dispersions are obtained by matching interfacial energy between the layered TMD and solvent, followed by disruption of the interlayer bonding through aggressive mechanical treatment. NMethyl-2-pyrrolidone (NMP) is the most effective solvent for surfactant-free exfoliation, yielding dispersions with a distribution of layer thicknesses from 2 to 12 nm after extended probe−tip sonication (400 W, >24 h).12 LPE has been shown to be successful for MoX2 and WX2 (X = S, Se) but has been largely uninvestigated beyond these group VI layered TMDs. Introduction of surfactants can reduce processing times and allows for use of alternative solvent systems, although single- to few-layer yields remain low (3%)13 and additional purification steps are required before subsequent surface functionalization or implementation for electronics.14 Very recently, the chemistry underlying LPE of MoS2 in NMP was uncovered, shifting the focus from surface energy matching to redox chemistry.15 Redox-active species form in situ during LPE through an autoxidation pathway that converts NMP to N-methyl succinimide by 5-hydroxy-N-methyl-2pyrrolidone intermediates. Promotion of this autoxidation pathway under ambient conditions, such as via thermal treatment or sonication, results in oxidation to MoVI and MoV species and exfoliation without mechanical forces. Suppression of autoxidation via processing in an inert atmosphere or under anhydrous conditions frustrates exfoliation even with extreme probe−tip sonication. The underlying mechanism was hypothesized to include dissolution of edge sites to form anionic molybdate ions in situ. Subsequent adsorption of these species facilitates exfoliation via Coulombic repulsion. In parallel, other work16,17 has shown that strong oxidants completely dissolve MoS2, while mild oxidants, such as 5-hydroperoxy methylpyrrolidine, appear to selectively etch edge sites.15 The redox mechanism should be general for any layered TMD given an appropriate weak oxidizing agent and in situ formation of anionic polyoxometalate.

Figure 1. General scheme for redox-based exfoliation of layered TMD powders, demonstrated in MoS2. Initially, bulk MoS2 powder is suspended in solvent (e.g., ethanol or acetonitrile) and treated with an oxidant (e.g., cumene hydroperoxide, CHP), resulting in partial dissolution of MoS2. This results in peroxometalate precursor formation and TMD sedimentation (A). Subsequent slow addition of a reducing agent (e.g., NaBH4) initiates POM formation via condensation of the metal oxide species (B). These POMs likely adsorb to the basal surface of MoS2 sheets and act as a stabilizing, charged surfactant, creating a Coulombic repulsion facilitating layer delamination and exfoliation (C). Sedimentation and redispersion in fresh solvent yields few-layer to monolayer MoS2 flakes that can be transferred to polar solvents. Concentrations can be further increased with the application of shear/ acoustic force as seen by increased viscosity.

NMP. This suggests that etching alone does not drive exfoliation. The yellow product can be transferred to nonpolar solvents (toluene and dichloromethane), suggesting that a weakly charged or neutral species is formed. This is consistent with peroxidation of molybdenum compounds in organic media.25 Following the initial CHP oxidation step, the addition of a reductant (e.g., NaBH4 (0.10 M, 400 μL, 0 °C)) transforms the supernatant from yellow to blue and finally to brown. This is consistent with prior reports of the reduction of MoVI to a mixed MoVI−MoV species and, finally, to MoIV, respectively.26,27 In concert with the color change, the MoS2 flakes disperse. The solubility of the initial sediment increases substantially as the solution transforms to blue but then drastically reduces upon further reduction to brown. Mass spectra (Figure S4) of the blue supernatant indicates the presence of large species (ca. 3000 g/mol), consistent with the formation of mixed-valency polyoxomolybdate (POMo) via the condensation of peroxo-MoVI species through Mo−O−Mo bridges. Additionally, the colormetrics (i.e., yellow → blue for Mo species)18,19,28,29 coincide with ζ-potential measurements that show a transition from a broad, weak ζ-potential (−5 mV) for the few flakes dispersed in the initial yellow solution to large negative values (−40 mV) for the substantial number of dispersed flakes following reduction (Figure S6). Thus, the increased solubility and delamination of MoS2, as the valency within the POMo shifts from MoVI to MoV, suggest that increasing charge incorporation in the POM clusters and basal surface adsorption provides sufficient Coulombic repulsion to initiate and drive layer separation. Note that the dispersed flakes are negative for all the TMDs investigated (Figure S6). Collectively, these characteristics are very unlikely to arise from basal surface absorption of solvent molecules, such as previously attributed to stabilizing flakes in NMP after

RESULTS AND DISCUSSION Figure 1 shows the general methodology for redox exfoliation of layered TMDs. Initially, preferential edge oxidation of the TMD forms high-valency transition metalates. Partial reduction of these metalates in solution induces condensation to form POM species, which absorb onto the TMD. Layer separation then occurs via Coulombic repulsion due to the ability of POM clusters to accommodate large amounts of charge.18 Note that many transition metals (e.g., Mo,19 W,19 V,20 Re,21 Ta, and Nb22) are known to form anionic POMs. Using MoS2 as an example, Figure 1 demonstrates quiescent exfoliation following this redox process using cumene hydroperoxide (CHP) (5:1 mol/mol CHP/MoS2) in CH3CN (ACN) at 0 °C. Note that CH3CN was chosen as our solvent system due its resistance to oxidation as the nitrile carbon is in the same oxidative state as carboxylic acids (R-COOH) and, unlike NMP, does not form thermal or oxidative fragmentation products.23 Anhydrous CH3CN was used and stored under molecular sieves to prevent hydrolytic side reactions. Specific reagent concentration and procedures are detailed in the Experimental Section and the Supporting Information. Initially, the CHP/MoS2 solution is yellow, reflecting the presence of peroxo-MoVI species.24 Similar behavior occurs in other polar solvents, such as acetone and ethanol (Figure S1). The initial oxidized flakes do not disperse, even after aggressive sonication (Figure 1A) nor upon sedimentation and transfer to 636

DOI: 10.1021/acsnano.6b06922 ACS Nano 2017, 11, 635−646

Article

ACS Nano

Figure 2. Solution-phase characterization of few-layer to monolayer TMD dispersions. (A) Optical images of concentrated and dilute exfoliated dispersions after optimization of reaction conditions (see Experimental Section). UV−vis (extinction) and X-ray diffraction (XRD) spectra are color-coded to the identifying text of each TMD in the digital image. Concentrated (mass/vol) and dilute dispersions (o.d. of 1 cm path length at A-exciton): MoS2 (3.2 mg/mL, o.d. 0.4); MoSe2 (1.4 mg/mL, o.d. 0.4); MoTe2 (24 h (Figures 1B and S7). Specifically, MoS2 powder was initially oxidized with CHP to form a MoS2 slurry. Controlled reduction of this slurry via slow addition of NaBH4 (0.10 M, 400 μL, 0 °C) resulted in yields of suspended flakes in ethanol or ACN up to 80−90% with respect to the initial powder mass. Note we will use terminology to specifically differentiate between this initial, stable colloidal TMD solution (multilayer suspension) and the size-selected, fractionated solution of few-layer to monolayer flakes (few-layer to monolayer dispersions). To isolate few-layer to monolayer flakes in solution, centrifugation was utilized to

selectively sediment larger multilayer particles from the suspension (details provided in the Experimental Section). Such fractionation indicated that the optimized conditions yielded ∼10% by mass of few-layer to monolayer flakes. The use of mechanical shear after the quiescent redox exfoliation further increased yield. For example, applying a high-speed IKA mixer at 20 000 rpm for 1 h to the initial suspension increased solution viscosity substantially, consistent with an increased particle number density. The yield of dispersed few-layer to monolayer MoS2 flakes increased by 50%. The increased viscosity and layer exfolitation are reminiscent of process−solution correlations reported for the exfoliation of organically modified layered aluminosilicates (nanoclays).30,31 These prior studies emphasize that maximum exfoliation occurs via the simultaneous optimization of both chemistry and mechanical energy.31 In addition to the initial exfoliation of MoS2, sedimentation and redispersion can be repeated to continually remove excess reagent or fractionate the size of the dispersed layers. For 637

DOI: 10.1021/acsnano.6b06922 ACS Nano 2017, 11, 635−646

Article

ACS Nano

corresponds to out-of-plane sulfur vibrations, undergoes a positional blue shift upon reduction of layer thickness.36,37 Correspondingly, the E2g vibrational mode, arising from inplane sulfur vibrations, undergoes a red shift in energy. These shifts and their peak-to-peak difference (i.e., A1g−E2g) can be used to confirm the presence of few-layer to monolayer flakes relative to thick flakes found in bulk powder. For the TMDs dispersed via redox exfoliation (e.g., MoS2, MoSe2, and NbSe2), characteristic blue shifts in the A1g mode and change in A1g−E2g separation are observed (Figure S10). For instance, bulk MoS2 has a A1g−E2g separation of 26 cm−1, whereas redox-exfoliated MoS2 exhibits a A1g−E2g separation of 24 cm−1. This suggests that the average thickness of exfoliated MoS2 flakes is between 3 and 6 monolayers.37 In addition to these peak shifts, optically inactive modes (B2g) in bulk MoSe2 and WSe2 transform to active modes upon exfoliation. For example, B2g is observed in MoSe2 (353 cm−1) and WSe2 (301 cm−1).38,39 UV−vis extinction measurements of the dispersed layered TMDs in CH3CN (Figure 2B) are also consistent with the spectra of single- to few-layer structures. The direct transition for ReS2 and the shift from the indirect to the direct band gap of MoX2 and WX2 are observed. From the corresponding absorption spectra (Figure S11), the A-exciton transitions (2HMX2) are consistent with prior theory and experimental reports: MoS2 (1.85 eV,40 670 nm); MoSe2 (1.5 eV,41 805 nm); WS2 (1.95 eV,42 629 nm); WSe2 (1.63 eV,42 760 nm); ReS2 (1.55 eV,43 810 nm); MoTe2 (1.1 eV,44 1175 nm). Note that NbSe2 is featureless, reflecting its metallic character.45 Also, the optical scattering profile of these TMDs (Figure S11), which was derived from the separate absorption and extinction measurements, is consistent with the few available results of LPE MoS2.46 Prior reports on MoS2 have established an empirical relationship between the layer dimensionality and the intensity of the A-exciton maxima in the extinction spectra and the local minima at 345 nm, where the latter cross section (ε345 = 69 mL mg−1 cm−1) is independent of scattering. This procedure reduces the uncertainty that arises due to the subtraction of the size-dependent scattering background.46 This approach yields an average thickness of 4.3 nm (6−8 layers) for the redoxexfoliated MoS2 in CH3CN (A-exciton = 671 nm). This is similar to previously reported values for LPE of 2−12 nm,11 suggesting that redox exfoliation is as effective for exfoliation as LPE while simultaneously being quiescent, low cost, and flexible (i.e., solvent choice). Unfortunately, extinction and absorption measurements calibrated to flake morphology are not available for the other layered TMDs. Solution X-ray diffraction provides a compliment to extinction spectroscopy by direct structural measurement of exfoliation extent at higher concentrations (Figure 2C). As layers separate, correlations along the c-axis disappear, leading to attenuation of [00l] reflections.47,48 However, inplane [hk0] reflections remain due to lattice correlations within the layer and provide an internal signal-to-noise standard. Note that in contrast to extinction spectroscopy, which preferentially samples single- to few-layer flakes relative to the population of colloidally stable multilayer tactoids, X-ray scattering is the opposite with strong [00l] intensity from stacked tactoids and relatively weak in-plane [hk0] scattering. Therefore, X-ray scattering requires relatively large concentrations (>1 mg/mL) for the [hk0] signal to overcome background scattering from the solvent and sample cell.

example, redispersion of the exfoliated MoS2 can be achieved in a range of polar solvents (e.g., ethanol, methanol, NMP, dimethylformamide (DMF), acetone, ACN) (Figures 1D and S5 and Experimental Section). Figure 2 demonstrates the redox-exfoliation method for seven group V−VII layered TMDs. Optical images of concentrated and diluted solutions show homogeneous, welldispersed flakes in CH3CN (Figure 2A). Analogous to MoS2, layered TMDs with metal centers that form highly anionic POMs (e.g., M = Mo, W, Nb, Re)14−17 were exfoliated in CH3CN with excess CHP (5:1 mol/mol CHP/TMD), followed by partial reduction with dilute NaBH4 to drive dispersion. For example, the transformation from colorless to orange is consistent with the formation of anionic PONb species32 (Figure S2). The suspensions were allowed to react at 273 K under quiescent ambient conditions for 24 h, where the concentration and volume of NaBH4 added were determined by titration of the supernatant until the colormetic response corresponding to mixed-valency POM was obtained19,22,24 (Figure S2). The resulting suspensions were centrifuged at 10 000 rpm for 15 min to sediment the slurry and remove the reaction solvent. The suspensions were then reconstituted in anhydrous CH3CN (or other polar solvents) and fractionated via centrifugation. Optimization of the reduction rate for each layered TMD resulted in highly concentrated (upward to 3 mg/mL) dispersions of few-layer to monolayer flakes (Figure 1C) that are stable in solution for more than 2 months (Figure S7). Note that layered TMDs containing transition metals that do not have accessible higher valent states, such as Ti,33 did not show a colormetric response nor was dispersion possible (e.g., TiS2). The low charge34 and high bascisity33 of Ti-oxometalates leads to TiO2 nanoparticle formation in the absence of appropriate capping agents rather than mixed-valency POMs. Thus, delamination via the redox process does not occur. The presence and removal of POM species were confirmed by X-ray photoelectron spectroscopy (XPS) of drop-cast MoS2 films prepared during each step of the reaction (Figure S8). Initially, the MoS2 suspensions in acetonitrile contain small amounts of MoVI species from stirring in solvent (acetonitrile), similar to previous studies that reported the formation of MoO42− due to the presence of dissolved O2.35 Addition of CHP results in the additional formation of MoVI, upward of 12% of Mo by mass. Subsequent reduction of the peroxomolybdate−MoS2 slurry results in a decrease in MoVI species (12.10 to 6.64%). After the flakes are washed with fresh solvent, the MoVI species further deceases from 6.64 to 1.95%. In all cases, the Mo/S ratio is higher than the expected 1:2 ratio, whereas the MoIV/S ratio remains relatively constant at 1:2 (Figure S9). The XPS results suggest that the integrity of the MoS2 flakes remains intact and that the MoVI species are distinct and not within the MoS2 flake. The largest deviations from the expected Mo/S ratio occur during the oxidation step, where the MoVI abundance is the highest. Following reduction and washing, the abundance of MoVI decreases substantially, suggesting that the MoVI species are weakly bound and can be removed. Similar trends are observed in MoSe2 films (Figure S9), where excess Mo was associated with MoVI, and after washing, the abundance of MoVI decreases (16.5 to 8.3%). Overall, these data are consistent with the proposed redox mechanism. Raman spectroscopy of films formed from the few-layer to monolayer dispersions is also consistent with few-layer to monolayer flakes. In MoS2, the A1g vibrational mode, which 638

DOI: 10.1021/acsnano.6b06922 ACS Nano 2017, 11, 635−646

Article

ACS Nano

Figure 3. AFM images of few-layer to monolayer TMDs. LPE of MoS2 is included as a quality reference. In general, lateral flake size scales with initial powder size (see Table S1). The distribution of layer thickness of redox exfoliates is very narrow (within 1−2 layers). Intermediate step sizes in flake thicknesses were not observed. In many samples, small particulates (approximately 1−2 nm) are observed on the surface (i.e., ReS2, WSe2), likely reflecting dried POM. These can be removed through repeated centrifugation/redispersion cycles. All scale bars are 300 nm. Comparison of AFM and XRD thicknesses is provided in Figure S14. For line cuts and thickness distributions, see Figures S17−S19.

Figure 4. High-resolution TEM micrographs for TMDs after redox-exfoliation strategy. The 2H crystalline domains are preserved as indicated by the single-crystalline domains as well as fast Fourier transforms resolving [100] and [110] lattice planes (A−G; scale bars 3 nm). Table S2 summarizes obtained indices that are in good agreement with those in the literature (see Supporting Information). (H) Through-focus exitwave reconstruction of the [211] zone axis of MoS2 indicates minimal differences in contrast for simulated 1−10 layer flakes (I), indicating limited thickness resolution for flakes, despite the presence of apparent atomically thin single-crystalline domains. 639

DOI: 10.1021/acsnano.6b06922 ACS Nano 2017, 11, 635−646

Article

ACS Nano

Figure 5. Orientation analysis of thick (0.1−1 μm), ordered films fabricated by reassembly of few-layer to monolayer flakes. (A) Twodimensional XRD scans and associated Herman’s orientation parameter, P2, showing high orientation along the [002] crystallographic axis. (B) SEM cross sections of films showing thin flakes stacking along the [002] axis. Films formed from bulk powder have random order. Scale bars are 500 nm. (C) Models of bulk and turbostratic films depicting the angle between [00l] and [h0l] reflections, denoted as ω (15°). Tilting the sample at this angle enhances intensities from [h0l] planes ([103] for 15°). (D) One-dimensional XRD scans of the highly ordered films show only [00l] reflections (highlighted in beige). (E) XRD spectra of MoS2 films measured at a 15° tilt (along the [103] and [105] axes). A significant reduction of the [103] and [105] intensities in the redox films as compared to bulk is indicative of a turbostratic structure with stacking faults along the [002] axis, which is consistent with the cross-sectional SEM and high P2 orientation parameters.

the [002] reflection from solution XRD. For MoS2, the thickness obtained from XRD is slightly larger than that obtained from UV−vis and atomic force microscopy (AFM) analysis discussed below (7.9, 4.2, and 4.4 nm). This is consistent with the sampling bias of the measurement techniques, as well as the expected increase in agglomeration between dilute (UV−vis and AFM) and concentrated (XRD) solutions. Figure 3 summarizes AFM of few-layer to monolayer TMD flakes on silicon wafers. Consistent with edge-site oxidation, flake edges appear eroded.16 Similar flake morphologies were observed with scanning electron microscopy (SEM) and transmission electron microscopy (TEM) imaging (Figures

For all of the solutions with few-layer to monolayer TMDs, sharp [hk0] in-plane peaks are observed whereas [002] reflections are substantially attenuated, confirming a loss of interlayer ordering even at these extreme concentrations (Figure 2C). The extent of layer separation scales inversely with the intensity ratio of [00l]/[hk0]. 47 Figure S12 summarizes the [002]/[110] intensity ratio for bulk powders and various exfoliated TMDs in solution. The intensity ratio is reduced to less than 50% for highly concentrated, exfoliated systems. Additionally, the [002] peaks are broadened, indicating a decrease in the average crystallographic size orthogonal to the layer (Figure S12). Figure S14 summarizes the average flake thickness determined from Scherrer analysis of 640

DOI: 10.1021/acsnano.6b06922 ACS Nano 2017, 11, 635−646

Article

ACS Nano

the reassembly resulting in lateral re-registry between layers and a reconstruction of the three-dimensional-layered TMD crystal. As a final point of reference, the redox process was compared to traditional liquid-phase exfoliation methods. A concentrated few-layer to monolayer dispersion of MoS2 (2 mg/mL determined by UV−vis) in NMP was prepared with probe− tip sonication at a power of 150 W (20%, 750 W) for 90 h according to previously reported methods.12 These conditions were chosen as they are currently the highest concentrations affordable via LPE.12 Comparable dispersions were prepared via the quiescent redox method in ACN and solvent transferred to anhydrous NMP (24 h of total processing). XRD data of the solutions were very similar (Figure S12). For example, average MoS2 thicknesses obtained from solution XRD by fitting the [002] peak were 7.8 and 6.0 nm for redox and LPE methods, respectively. AFM also confirmed similar flakes, where LPE exhibited a thickness distribution (2.3 ± 2.2 nm) broader than that of the redox process (4.4 ± 0.7 nm) (Figures S14 and S17). However, redox dispersions were stable for than 60 days with no sign of sedimentation, whereas traditional sonicated LPE dispersions were metastable with complete sedimentation in 60 days (Figure S7).

S15 and S16), and lateral size is summarized in Table S1. The average lateral size of redox-exfoliated MoS2 was similar to that obtained from LPE (224 and 149 nm, respectively). Note that while LPE can give larger flake sizes, the extended sonication necessary to provide a similar concentration of dispersed flakes induces flake scission.12 Overall, flake diameters scaled with the size of the bulk powders (Table S1), similar to prior bulk processing reports.12,49 For larger initial powders, such as MoSe2 and MoTe2 (>10 μm), exfoliated flakes were larger (200−400 nm). For smaller bulk powders, such as MoS2 (∼2 μm), exfoliated flakes were small (100 μL of NaBH4). After ca. 100 μL of NaBH4, MoS2 flakes slowly started to exfoliate into the solution. Further addition of NaBH4 increased the yields of exfoliated MoS2 (ca. 350 μL), after which point the exfoliation yields subsequently decreased with further addition of NaBH4. At this point, a brown precipitate was occasionally observed and immediate flocculation of MoS2 was observed. Optimization of MoS2 Exfoliation. Maximum exfoliation for each TMD can be optimized by varying amounts of the oxidant (CHP) with constant amount of reductant (NaBH4). As an example, dispersions with constant MoS2 concentration (3 mg/mL in CH3CN, 10 mL total) were treated with varying amounts of CHP (0.1 to 10 mmol) and reduced with a constant amount of NaBH4 to tune in situ POM formation. The solution was analyzed for mass of suspended flakes (i.e., no centrifugation) and mass of exfoliated flakes (centrifugation 1500 rpm, 30 min) via UV−vis spectroscopy techniques previously discussed. For low loading ratios of CHP (1:10 CHP/MoS2), suspension and exfoliation were low (15 and 0.6%, respectively). As the amount of CHP increased, the concentration of suspended flakes also increased, to almost quantitative (88%) suspension with a CHP/MoS2 ratio of 5 (mol/mol). This also yielded the maximum amount of monolayer to few-layer flakes (∼10%). Further increase of CHP reduced the amount of suspended MoS2 until a critical threshold (approximately 10:1 CHP/MoS2), where upon no exfoliation occurred. Interestingly, addition of NaBH4 to suspensions of high loading ratio CHP/MoS2 failed to give a blue colormetric response, suggesting that formation of anionic POMo is suppressed. In these cases, the supernatant turns into a yellow-brown color, and addition of NaBH4 affords a brown precipitate. MoSe2 Exfoliation. Surface oxidation was carried out in a manner identical to that for MoS2. The resultant supernatant was an optically clear yellow solution. The stirrer was removed, and the flakes sedimented (ca. 45 min). The volume was reduced to 4 mL, and the temperature was reduced to 0 °C. This temperature was maintained for the duration of the reduction step. After equilibration (30 min), 100 μL of a 0.010 M NaBH4 solution was added quickly. Subsequently, 10 μL of 0.10 M NaBH4 was added in 60 min intervals (5×) and stirred for a total of 24 h (at 0 °C). The viscosity increased, and the color turned from a dull gray to a dull red color, indicating exfoliation of MoSe2. After 24 h, the ice bath was removed and the solution stirred for another 24 h at 25 °C. Afterward, the stirrer was removed, large flakes were allowed to settle, and the supernatant was pipetted out and centrifuged at 10 000 rpm for 30 min to sediment all the flakes. The optically clear, red supernatant was discarded, and anhydrous solvent (CH3CN, NMP) was added to suspend the exfoliated flakes. After homogenization via a vortexer, the mixture was centrifuged at 1500 rpm for 45 min. The resultant supernatant contained few-layer to monolayer MoSe2 flakes, and the top 2/3 of the supernatant was pipetted and analyzed by XRD, AFM, SEM, and TEM. Note: Addition of NaBH4 in excess of 110 μmol results in unrecoverable flocculation of MoSe2. MoTe2 Exfoliation. Oxidation and reduction were carried out in a manner exactly the same as that for MoSe2. The supernatant after oxidation was clear light yellow. After reduction, the supernatant turned into a clear black solution. WS2 and WSe2 Exfoliation. Surface oxidation was carried out at 333 K (60 °C). All other conditions were identical to those for MoS2 oxidation. After oxidation, the heat was removed and allowed to equilibrate to room temperature before proceeding. The powder was allowed to sediment (ca. 15 min), and the volume was reduced to 4 mL. The temperature was reduced to 273 K, and 20 μL of a 0.10 M NaBH4 solution was added. The viscosity increased slowly, and the solution was stirred for 24 h at 0 °C. After 24 h, 100 μL of a 0.10 M NaBH4 was added and allowed to stir for an additional 24 h. This was repeated until there was a total of 400 μL of 0.10 M NaBH4, after which the supernatant turned from a gray to a dull green color (WS2) or dull red (WSe2) color, indicating exfoliation of material. The temperature was increased to 25 °C after NaBH4 addition was

EXPERIMENTAL SECTION Materials. NMP, acetonitrile (CH3CN), DMF, ethanol (EtOH), acetone ((CH3)2CO), CHP (80%), molybdenum(IV) sulfide (MoS2), and tungsten(IV) sulfide were purchased from Sigma-Aldrich. Molybdenum(IV) selenide was purchased from Strem. Tungsten(IV) selenide, rhenium(IV) sulfide, and niobium(IV) selenide were purchased from Alfa-Aesar. Molybdenum(IV) telluride was purchased from Materion. NMP and DMF were distilled in the presence of CaH2 at 300 mTorr at 60 °C. CH3CN was distilled in the presence of CaH2 under an argon flow. All solvents were stored in argon-purged bottles under activated molecular sieves (3 Å). Molecular sieves were activated by heating for 24 h under vacuum (150 °C, 300 mTorr). Surface Oxidation of MoS2. A three-neck round-bottom flask was initially evacuated (300 mTorr) for 5 min and backfilled with argon gas. This was performed three times to ensure removal of O2. (Note: Because it has been shown that MoS2 dissolution occurs under ambient atmosphere,35 we ensured reactions were carried out with minimization of side reactions by utilizing anhydrous, distilled solvent and inert atmosphere. While these extra precautions have been taken, we have observed that the experimental results are insensitive to atmosphere and water content in solvent for all TMDs investigated.) Next, 0.625 mmol (100 mg) of MoS2 suspended in 5 mL of CH3CN was added through a septa seal into the purged round-bottom flask and stirred at 0 °C using an ice bath. After the temperature was equilibrated (30 min), 3 mmol of cumene hydroperoxide (562 μL of 80% CHP) in 5 mL of CH3CN was added dropwise to the MoS2 suspension over the course of 15 min. The mixture was stirred at 273 K (0 °C) and kept at this temperature throughout the addition of CHP. The supernatant of the suspension became discolored within 60 min (S1). After all CHP was added, the ice bath was removed and the reaction was allowed to stir at 25 °C for 24 h. Exfoliation of MoS2. The stirrer was removed after 24 h, and the MoS2 powder was allowed to settle (ca. 10 min); the volume was reduced to 4 mL by pipetting out the yellow supernatant. The temperature was reduced to 0 °C using an ice bath and allowed to equilibrate for 30 min. Following equilibration, 50 μL of a fresh, ice cold aqueous solution of 0.10 M NaBH4 (38 mg of NaBH4, 10 mL of H2O) was quickly added to the oxidized TMD solution at 273 K (0 °C) and stirred for 1 h. Additional portions of 100 μL of 0.10 M NaBH4 were added in 1 h intervals until a total of 350 μL was added. After NaBH4 addition was completed, the ice bath was removed and the solution was allowed to stir for 24 h at room temperature. The viscosity of the solution increased, and the suspension turned from a dull gray to a green color, indicating exfoliated material. At this point, the solution can be probe-sonicated (30 min) or mixed (IKA mixer, 20 000 rpm, 1 h) after the suspension was transferred to a 20 mL flatbottom vial. This further increases yields approximately 2× (solvents used: acetonitrile, acetone, ethanol). After addition of NaBH4 and optional probe sonication, the stirrer was removed and large, unexfoliated flakes were allowed to settle. The supernatant was isolated and centrifuged at 10 000 rpm for 15 min to sediment the exfoliated flakes. The optically clear, colorless supernatant containing excess CHP and possibly unreacted molybdates was discarded, and fresh anhydrous solvent (e.g., CH3CN, DMF, NMP) was added and the suspension homogenized in a vortexer. The suspension was subsequently centrifuged at 1500 rpm for 30 min. The resultant supernatant contained few-layer to monolayer MoS2 flakes and was isolated for analysis by XRD, AFM, SEM, and TEM. Weaker reducing agents (hydroquinone, ascorbic acid, or sodium tribasic citrate) were not able to reduce the yellow peroxomolybdate complex effectively, resulting in poor yields and quality. Titration of MoS2 with NaBH4. To determine appropriate amounts of NaBH4 to add for efficient exfoliation, a portion of each reaction was taken and titrated with NaBH4. After the surface oxidation step, a 1 mL aliquot was taken out of the reaction vessel and NaBH4 (0.10 M) was added in 20 μL portions (at 0 °C) and mixed for 5 min under mechanical stirring. Subsequently, the mixture was centrifuged at 1000 rpm (5 min) to monitor the exfoliation of MoS2 flakes. This stepwise titration of MoS2 was performed until flakes were 642

DOI: 10.1021/acsnano.6b06922 ACS Nano 2017, 11, 635−646

Article

ACS Nano

Determination of TMD Yields. A solution of electrostatically stabilized and suspended TMD flakes contains a broad distribution of flake thickness, of which only a fraction are few-layer to monolayer exfoliates. There is no accepted standard to report yields of few-layer to monolayer flakes, and many times, values reported in the literature as “exfoliated” are really the total mass of TMD flakes suspended in solution and thus are much greater than the few-layer to monolayer population. To estimate the yield of suspended and few-layer to monolayer flakes, the following procedure was followed. The mass and concentration of suspended flakes were obtained by allowing the reaction vial to remain undisturbed for 1 h after the reduction step. This ensured unsuspended flakes sedimented. A known volume of the colloid solution was then pipetted out and placed in a preweighed vial where the solvent was removed via vacuum (e.g., acetonitrile at 200 mTorr). The measured mass and mass-to-colloidal solution volume represented the total mass and concentration of suspended flakes, respectively. For example, the initial concentration of MoS2 in typical redox exfoliations was 10 mg/mL. Five to 7 mg were typically suspended in 1 mL of colloidal solution, giving a suspension yield of 50−70%. The fraction of few-layer to monolayer flakes was then estimated based on centrifugal fractionation. A 1 mL aliquot from the stable colloidal solution was centrifuged at varying rpm, and the supernatant was examined by UV−vis and AFM as discussed above. For the TMDs examined herein, 10 000 rpm provided effective separation of multilayer from few-layer to monolayer flakes. After centrifugation of the aliquot at 10 000 rpm, the pellet was washed three times with ethanol to remove excess reagents (CHO, NaBH4), placed in a preweighed vial, and dried under vacuum (300 mTorr, 60 °C). The mass reflected the amount of multilayer, suspended flakes. For the same aliquot volume, the difference between the total suspended mass and multilayer mass represented the mass of few-layer to monolayer flakes. For example, the yields of few-layer to monolayer MoS2 are typically between 3 and 10%. Characterization. X-ray Diffraction Analysis. XRD patterns were recorded using a Smartlab system (Rigaku) with Cu Kα radiation (λ = 0.15418 nm). Solution XRD was carried out by depositing 200 μL of 2 mg/mL exfoliated TMDs suspended in NMP into 1.5 mm quartz capillary tubes (Charles Supper Company) and measured in transmission 2θ mode. After redox reactions were complete, the TMDs were dispersed in NMP after removal of the original solvent. Next, the dispersion was centrifuged at 1500 rpm for 45 min, collecting the supernatant, and concentrating the supernatant down to 200 μL. This was performed by centrifuging the exfoliated dispersion at high speed (6000 rpm, 45 min) to sediment the exfoliated flakes. The solvent volume was reduced to 200 μL, and loaded into a 1.5 mm quartz capillary tube. Thin films of TMDs were drop-casted (acetonitrile) onto Si100 wafers (Ted Pella, Inc.) and measured in θ/ 2θ mode. Average crystallite thicknesses along a specific (hkl) plane were calculated using Scherrer’s equation:

completed and stirred for an additional 24 h. Isolation of few-layer to monolayer flakes was performed in a manner identical to that for MoS2. Note: Addition of NaBH4 in excess of 400 μmol results in unrecoverable flocculation of WX2. NbSe2 Exfoliation. Surface oxidation, reduction, and few-layer to monolayer flake isolation were carried out in a manner identical to that for MoSe2. ReS2 Exfoliation. Surface oxidation was carried out at 333 K (60 °C). All other conditions were identical to that for MoS2 oxidation. After oxidation, the heat was removed and the mixture allowed to equilibrate to room temperature before proceeding. The powder was allowed to sediment (ca. 5 min), and the volume was reduced to 4 mL by pipetting the dark yellow optically clear supernatant. The temperature was reduced to 273 K, and 10 μL of a 0.010 M NaBH4 solution was added quickly. The color of the suspension slowly turned from gray to red, indicating exfoliation of ReS2. After 3 h, 10 μL of a 0.10 M NaBH4 solution was added, and the solution was stirred for 24 h, during which the viscosity increased. Isolation of few-layer to monolayer flakes was performed in a manner identical to that for MoS2. Note: Addition of NaBH4 in excess of 75 μmol results in unrecoverable flocculation of ReS2. General Experimental Note. After more than 400 different trials, incremental addition of NaBH4 at a reduced temperature (0 °C) provided the most reproducibility. Addition of reductant at elevated temperatures (25 °C) increased variability and reduced yield and flake quality (i.e., thick flakes (λa = 680 nm) and low yields (