Luminescent transition metal dichalcogenide nanosheets through one-step liquid phase exfoliation

Liquid phase exfoliation (LPE) from the bulk is an adequate method for the mass-production of thin nanosheets of transition metal dichalcogenides (TMDCs). However, making suspensions in which the extraordinary properties of mechanically exfoliated TMDCs are observable remains a challenge. We describe a mild LPE method to produce luminescent suspensions of MoS2 and WS2 in N-methylpyrrolidone or isopropanol/water mixtures, without the need for a purification step. The key differences in our experimental procedure compared to previously reported LPE methods are the use of mild bath sonication at controlled temperature and the low initial concentration of the parent TMDC. Spectroscopic and AFM data confirm that an overwhelming majority of the sample is composed of ultrathin nanosheets. HREM data support the formation of the luminescent 2H polytype. The ultrathin nanosheets can be transferred to pure water and cell culture medium. Confocal fluorescence microscopy experiments on MCF-7 breast cancer cells exposed to LPE WS2 show that the cells are viable and the photoluminescence of the nanosheets is detectable.

Ultrathin TMDCs can be obtained by synthesis, through chemical vapor deposition [22,24,37], or by exfoliation from the bulk. Exfoliation is carried out either in dry phase through the mechanical action of an adhesive tape or in wet phase via dispersion in a liquid, with or without the help of a chemical reaction. Although mechanical exfoliation produces large area and nearly defect-free TMDCs flakes, the process lacks scalability [38,39]. Chemical exfoliation is usually based on the lithium intercalation (LI) method, which was developed in the 70 s [40][41][42] and consists of two steps: intercalation of the Li atoms and reaction of the intercalated Li with water, to produce exfoliation. To intercalate the Li, the bulk TMDC is reacted with nbutyllithium (BuLi) [43], typically for long reaction times and/or at high temperature; alternatively, Li can be intercalated electrochemically or with the help of ultrasounds [44,45]. While LI results in large quantities of single-layer TMDCs, due to the harsh reaction conditions, the TMDCs nanosheets obtained undergo a phase transition from trigonal prismatic (2H) to octahedral (1T), which is accompanied by a change in electronic properties from semiconducting to metallic [46,47]. Although metallic 1T-MoS 2 has proven useful as electrode for the construction of supercapacitors [48], the semiconducting character of 2H TMDCs is usually of more interest, especially in applications exploiting the intrinsic band gap of TMDCs such as optoelectronics or fluorescence microscopy. The photoluminescence (PL) of MoS 2 and WS 2 is centered at around 660 nm, in the optical window of biological tissues, which opens up the possibility of bioimaging applications [49]. Therefore, the production of suspensions of luminescent TMDCs (2H polytype) is of great interest. Indeed, a method for LI exfoliation followed by 1T to 2H phase reversal has been recently reported [50].
Liquid-phase exfoliation (LPE) techniques are one of the most versatile processes to obtain few-layer 2D nanosheets [51]. The best known LPE method to obtain few layer TMDCs is the direct ultrasonication of bulk TMDC powder in adequate solvents, pioneered by Coleman and co-workers [52,53], and later investigated by several other groups [54,55]. Shear exfoliation methods have also been investigated, in particular for large-scale production [56]. Aqueous suspensions can also be obtained with the help of surfactants [57,58]. LPE of layered materials generally results in a large polydispersity in thickness and lateral size, and unusual phase transitions have also been observed [59]. Therefore, it is very challenging to preserve the unique properties of single-layer nanosheets, reported for mechanically exfoliated TMDCs, in dispersions prepared by LPE. In particular, the strong PL routinely observed in mechanically exfoliated MoS 2 [60] has, to the best of our knowledge, only been described for LI exfoliated MoS 2 after deposition and annealing at 300°C [61], and for suspensions of LPE MoS 2 after purification of the single-layers through iterative centrifugation and redispersion cycles that severely increase the preparation time and thus the production cost of the nanosheets fabricated by this method [62].
Here, we present a mild exfoliation method to produce MoS 2 and WS 2 suspensions with an outstanding density of ultrathin luminescent layers in a single step, without the need for post-treatment or purification. The method works well both in N-methylpyrrolidone (NMP) and in isopropanol (IPA)/water mixtures, and diluted suspensions of the luminescent nanosheets can be obtained in pure water as well as, remarkably, in cell culture medium. Incubation experiments with breast cancer MCF-7 cells demonstrate the biocompatibility of the nanosheets and their potential use in biological applications.

Results and discussion
Optimization of TMDC liquid phase exfoliationspectroscopy and microscopy characterizations MoS 2 LPE in NMP The ultimate objective of this work was to develop a purification-free method to prepare suspensions of luminescent TMDC nanosheets. As a first step towards that goal, we optimized LPE exfoliation of MoS 2 in NMP in terms of MoS 2 dispersion initial concentration. We used UV-vis extinction measurements and transmission electron microscopy (TEM) imaging to characterize the optical properties (figure S1) and estimate the concentration of the final suspensions (figure S2), and analyze the size distribution of the corresponding nanoparticles (figures S3 and S4). Based on the results obtained, an initial MoS 2 concentration of C 0 = 1 mg ml −1 was found to afford the best compromise between the concentration (high enough, ∼3 μg ml −1 ) and the dimensions (low thickness, ∼850×470 nm 2 as the lateral size) of the colloidal material. So this C 0 concentration was used to further investigate the effect of sonication time on the PL spectra of the suspensions. Figure 1(A) shows the PL spectra of MoS 2 samples sonicated for different times between 5 and 240 min. To get a fair comparison of the relative amount of luminescent nanosheets, all samples were diluted to matching extinction at the wavelength of excitation (425 nm). Samples sonicated for longer times than 30 min display two emission bands at 655 nm and 605 nm that can be ascribed to A and B excitons respectively (the latter PL arising only from bi-to few-layer MoS 2 species, which is a first indication on the composition of the corresponding colloids) [62][63][64]. As the sonication time increases beyond 60 min, the PL intensity decreases, while no change is observed in the position of the bands, which points to a decrease in the relative concentration of the luminescent species, rather than to a reduction in their lateral size [62]. Note that further PL characterization of colloidal suspensions prepared by sonication of C 0 =0.5, 2.0 and 5.0 mg ml -1 MoS 2 initial dispersions showed almost no PL as compared to the above described samples, thus confirming our initial choice ( figure S5).
To gather more information on the source of the PL signal, we carried out PL excitation (PLE) measurements, and compared them with the extinction spectrum ( figure 1(B)). For λ em =655 nm, the PLE spectrum shows two bands, the most intense of which is located at 400-450 nm (C exciton). The second band is slightly weaker and broader, and it is found at 500-600 nm, corresponding to the B exciton. The relative intensity of the PLE maxima is consistent with the PL signal originating at least partially from fewlayered MoS 2 , besides single-layered nanosheets [64]. It is also interesting to note that the PLE position of all three excitons is significantly blue-shifted with respect to the extinction spectrum, where they appear at 464, 617 and 678 nm, respectively. This observation can be interpreted on the basis of the results reported by Coleman and co-workers: they observed such shifts in the extinction spectra of different MoS 2 suspensions to be especially larger as the average dimensions (number of layers and lateral size) of the nanosheets in the sample increase [62]. As a consequence, our spectroscopic results indicate that our sample is a mixture of nanosheets of different thicknesses and sizes in a particular distribution. The population of largest and thickest flakes is substantial enough regarding the population of mono-to-few-layer nanosheets to dominate the extinction, which explains the ∼20 nm blueshifts of the PLE bands as compared to the absorption bands. However, as only monolayers (and, to a far lesser extent, bi-and few-layers) [63] exhibit PL, we can deduce from the appreciable fluorescence signal detected that the relative concentration in smallest and thinnest flakes is significantly high.
Nevertheless, as the PL quantum yield (QY) of MoS 2 monolayers remains very weak and the samples exhibit large scattering, artefacts could be mistaken for a PL signal. We thus performed a PLE intensity map as a complementary measurement (figure 1(C)), which confirmed undoubtedly the PL nature of the previously observed band, with a signal centered at 655 nm, i.e. the direct bandgap emission wavelength of MoS 2 monolayers.

MoS 2 LPE in IPA/water
The use of pyrrolidone-based solvents presents many drawbacks related to their toxicity and high boiling points. Now, LPE of layered materials in mixtures of solvents with low-boiling points has already been demonstrated to be an effective strategy for high yield exfoliation of graphite and TMDCs [54,65,66]. Hence, the exfoliation of MoS 2 was carried out in different mixtures of IPA and water, based on the conditions optimized in NMP: ultrasonication for 60 min at C 0 =1 mg ml -1 .
The mild exfoliation process also produces fluorescent samples without purification in this case. However, significant PL is obtained from the IPA/water 7/ 3 mixture only (v/v, figure 1(D)), which is also the composition that leads to the highest concentration in colloidal material ( figure S6). Indeed, such a mixture has been reported by Ajayan et al to maximize MoS 2 dispersion [55]. Figure 1(E) shows the corresponding PL, PLE and extinction spectra. In analogy with the suspensions in NMP, the maximum PL is obtained for 60 min sonication. The blue-shift of the excitons' bands in the PLE spectrum with respect to the extinction data is again observed in this case, due to the differences in the absorption and emission properties between thin/small nanosheets and thicker/larger ones, as already explained in detail above (see figure S7 for the distribution in lateral dimensions obtained from TEM imaging). As in the previous case, the excitation-emission contour map shown in figure 1(F) allows for the unambiguous attribution of the fluorescence signal to the PL of MoS 2 monolayers. However, the PL intensity here is weak enough to observe artefacts whose positions change according the excitation wavelength (streaks), and most probably caused by the non-negligible light scattering observed in the extinction spectrum (figures 1(D) and S6(a)).

WS 2 LPE in IPA/water
Our mild LPE conditions produce WS 2 nanosheet suspensions in IPA/water mixtures as well (figure S8). As in the case of MoS 2 , luminescent colloids can be obtained without the need for purification; but quite the opposite, significant PL can be detected in a wider range of IPA volume ratio, from 50% to 90% ( figure 1(G)). Nevertheless, the IPA/water 7/3 mixture remains the optimal solvent to get suspensions that are enriched in luminescent colloidal material.
The PL spectra of the above-mentioned mixtures exhibit two emission bands at 610 and 655 nm. The former can be easily attributed to the A exciton of monolayered WS 2 [64,67,68], whereas the latter needs to be considered more carefully. As opposed to the MoS 2 case, this second emission appears at a higher wavelength than that of A exciton, hence it cannot be assigned to B exciton PL (observed at a far lower wavelength of ∼516 nm in WS 2 ); and as for the indirect bandgap emission from bi-and few layers, it occurs at wavelengths higher than 700 nm [67]. Otherwise, previous studies on WS 2 exfoliated either mechanically [67] or in liquid phase [68] showed a red-shifted A exciton PL due to bi-to five-layer nanosheets. Nevertheless, we discarded this possibility too, since the corresponding emissions do not exceed ∼630 nm; additionally, as the PL QY decreases rapidly with the number of layers (by a factor ∼10 2 for bilayers to ∼10 3 for five layers) [67], such an assumption would not match with the PL relative intensities of the bands at 610 and 655 nm. Slight PL shifts, due to sample/solvent interactions in LPE and as compared to most of literature data on mechanically exfoliated samples, neither would account for our observation. Actually, the higher-wavelength PL would be in better agreement with a significantly red-shifted and intense A exciton-type emission coming from defects and/or edges, a phenomenon already observed at ∼650 nm both with monolayer and few-layer nanosheets prepared by LPE [69] or CVD [62,70]. This would be due to the tendency of excitons in WS 2 to concentrate (hence the intense PL) in the above-mentioned lattice regions; the strong exciton localization would result in a higher exciton binding energy, leading in turn to a lower-energy PL emission [70]. Other spectroscopic features of WS 2 (extinction spectrum, PLE spectrum recorded at the emission wavelength of monolayers; figure 1(H)), further examined with the sample obtained from the IPA/water 7/3 mixture, are analogous to those previously described with the MoS 2 sample, leading to a similar interpretation (see figure S9 for details on the particle size distributions).
As a further characterization of the second PL band at 655 nm, a PLE spectrum of the sample was also measured at λ em =655 nm. In that case, an intense (∼10 times the PLE intensity recorded at 610 nm) and broad (FWHM ∼90 nm) signal centered at 590 nm is observed (figure S10). This last result seems to confirm the edge/defect emission hypothesis. On one hand, it does not match with the characteristic PLE expected for bilayer (and few-layer) WS 2 [64], which should be quite similar to that of the monolayer (exhibiting B and C exciton absorptions). On the other hand, it reveals that the 655 nm emission is related to high absorption over a large range of wavelengths, including the A exciton wavelength. This would be consistent with excitons originating from the whole structure of mono-to multilayers (several A exciton populations differing in their binding energies), then moving to be trapped in the local potential wells of edges or defects. There they would acquire a stronger binding energy, and consequently relax by emitting a lower-energy photon.
Finally, figure 1(I) confirms the most intense fluorescence signal is the one we attribute to the defects of the nanostructure. Additionally, it evidences the emission originating purely from WS 2 monolayers is appreciable only when exciting at 425 nm.

Atomic force microscopy (AFM)
To complement the information provided by the spectroscopic data, we performed AFM measurements. Figure 2(A) shows a typical large area AFM image of a dropcast from an IPA/water 7/3 MoS 2 suspension onto a freshly cleaved mica substrate (see figure S11 for a zoomed-in micrograph). Mica has been selected for its atomically flat surface, in order to facilitate the thickness determination without artefacts arising from the substrate corrugation. Statistical analysis over 358 nanosheets is shown in figures 2(B) and (C). Raw height data show that ultrathin nanosheets are an overwhelming majority ( figure 2(B)). Our data show that more than 60% of the flakes are thinner than 3 nm (∼4-5 layers), and are expected to show PL. More remarkably, our analysis shows that 25% of the flakes have a thickness of 1 nm or less, a thickness value that is compatible with singlelayer MoS 2 . The inset in figure 2(B) shows a fit of the height histogram to multiple Gaussian curves spaced by 0.65 nm, used to determine the population of single-, bi-, tri-and multi-layers in our MoS 2 samples. Note that the Gaussians have been offset by 0.4 nm in order to obtain the best fit, this offset accounts for the adsorbates between the flake and substrate that yield to an artificially higher thickness in single layer MoS 2 prepared by liquid phase exfoliation. In fact, even though the thickness of a pristine single-layer of MoS 2 should be ∼0.65 nm, the measured height might vary with adsorbed solvent, etc and reported height values of single-layer LPE MoS 2 can be as high as 1.9 nm in the literature [62]. The area versus height data in figure 2(C) again shows that the vast majority of flakes show heights <3 nm. However, a significant cluster of large flakes of area ca. 10 4 nm 2 and of height between 4 and 5 nm also appears, and accounts for the extinction data.
Panels D-F of figure 2 show the same analyses for the WS 2 sample prepared in the same conditions. In this case, the size distribution is even narrower. In fact, more than 55% of the flakes are thinner than 2 nm, which would correspond to flakes only 1-3 layers in thickness, which is in very good agreement with the intense PL observed. The inset in figure 2(E) shows the fit of the height histogram for WS 2 samples to multiple Gaussian curves spaced by 0.65 nm to determine the population of single-, bi-, tri-and multi-layers. Here a height offset of 0.5 nm has been used to optimize the fit and account for the adsorbates between the flake and substrate.

Transmission electron microscopy
Low magnification TEM reveals the presence of ultrathin nanosheets for both MoS 2 (in NMP: figures S3 and S4; in IPA/water 7/3: figure S7) and WS 2 (figure S9). The high resolution transmission electron microscopy (HRTEM) characterization provides a more detailed study of the nanosheets, revealing areas of different degree of crystallinity and orientation. A typical situation is shown in figure 3(A) where flat crystalline 2H-polytype areas (marked as i), as confirmed by the corresponding FFT, coexist with more disordered domains, even of amorphous nature, as seen in the edge (marked as ii). The periodicity measurement in the flat hexagonal areas (0.31×0.31 nm 2 ) is in agreement with the 2H unit cell along the [001] zone axis. In addition, a Moiré pattern is observed in the upper part of the figure. This can be explained on the basis of the stacking of hexagonal layers mutually rotated along the c direction. This can be better understood in the corresponding FFT, which shows two sets of spots in hexagonal arrangements rotated by 27°. The presence of small fragments exhibiting fringe-like features have also been observed as depicted in figure 3(B). It should be noticed that the measured periodicity of 0.63 nm fits well with the interlayer MoS 2 separation. Moreover, figure 3(C) shows a HRTEM image of the WS 2 sample that clearly evidences the 2H stacking. Notice that there are only 4-5 layers, confirming the low thickness of the nanosheets.
These spectroscopic and microscopy data thus point out that the straightforward process we have developed is effective in yielding monolayer-enriched TMDC colloids. Although it appears similar to other LPE procedures, to the best of our knowledge, the only LPE processes that have been reported to lead to fluorescent TMDC samples [62,68,71] required purification procedures. The key differences in our experimental procedure are: (1) the absence of additives, (2) the low initial concentration, (3) the controlled temperature during LPE, (4) the use of NMP or IPA/water mixtures as solvents. In comparison, the LPE methods reported by Coleman et al [62,68] are based on probe-tip sonication, during which the actual sonication efficiency is far higher than during bath sonication, even at the same operating power [72]. Additionally, they operate in aqueous solutions of surfactants and the suspensions obtained are subjected to either higher-speed or successive centrifugation steps, each of them lasting at least 2 h. As for the mild bath sonication conditions, they proved to lead efficiently to monolayered nanosheets (hence, exhibiting fluorescence in the case of MoS 2 ) from other previously reported LPE procedures [71,73]; in those cases however, and as above, the exfoliation was performed with the help of additional binding molecules.

Experiments in biological media
Finally, considering the growing interest in the application of low-dimensional TMDCs to the biomedical sciences [74,75], we explored the transfer of the WS 2 luminescent nanosheets produced in IPA/water to pure water and cell culture medium. The flakes are observable either in reflection (figure 4(A)) or in transmission mode under a confocal microscope, and were unambiguously identified as WS 2 nanosheets through Raman spectroscopy, which showed the characteristic E g 1 2 and A 1g Raman modes at 351 cm −1 and 419 cm −1 , respectively, in agreement with ultrathin nanosheets (inset figure 4(A)) [70]. Although the final concentration of the nanosheets in water and cell culture medium is too low for the PL to be detected in suspension, we could observe that the nanosheets showed red PL (λ em =570-700 nm) when explored under a confocal fluorescence microscope (λ exc =405 nm), as seen in figures 4(B) and (C).
When MCF-7 cells were exposed to dispersions of WS 2 nanosheets at a concentration of 5 μg ml -1 for 4 h, PL from the exfoliated nanomaterials was weaker, but still distinctly detected, as seen by confocal fluorescence microscopy (figures 4(E) and (F); details can be found in the section F of the supporting information). The decrease in the PL of the nanomaterial is attributable to the incubation of the nanosheets within the cell microenvironment, which seems to cause the formation of aggregates. This aggregration can be easily explained by the absence of specific functionalization of the particles that would ensure their dispersion in such a complex medium. Moreover, viability studies showed no cytotoxicity of the WS 2 nanosheets at concentrations up to 10 μg ml -1 in MCF-7 breast cancer cells (figures S12 and S13).

Conclusion
In conclusion, we have shown that it is possible to produce suspensions of MoS 2 and WS 2 sufficiently enriched in ultrathin nanosheets to observe their PL in a single LPE step, without the need for a timeconsuming purification. As a proof-of-principle application, we have transferred the luminescent nanosheets to water and cell culture medium and incubated them with human breast cancer cells. The cells remain viable and the PL of WS 2 can be observed with confocal fluorescence microscopy, which paves the way for their use in biological applications. The reach of our results, however, expands to any other application where ultrathin semiconducting TMDC nanosheets are required.
Optimization of MoS 2 LPE in NMP MoS 2 powder was dispersed in NMP at different initial concentrations: 0.5, 1.0, 2.0 and 5.0 mg ml -1 . The solutions were sonicated for various periods of time from 5 to 480 min To exclude the effect of temperature and avoid changes in the composition of the solvent due to partial evaporation or decomposition due to an increase in the temperature of the water bath during sonication, a cooling system was connected to the bath sonicator to perform the experiments at 20°C. Aliquots of the suspensions were centrifuged at 3000 rpm (990 g) for 30 min at 25°C. The supernatant was carefully collected.
Optimization of MoS 2 and WS 2 LPE in IPA/water mixtures MoS 2 and WS 2 nanosheet suspensions were prepared via LPE technique by dispersing 1 mg ml -1 of MoS 2 or WS 2 powder in different volume ratios of IPA/water (0/10, 1/9, 3/7, 5/5, 7/3, 9/1, 10/0). The solutions were sonicated for 60 min. As described previously, the experiments were thermostated at 20°C. Aliquots of the suspensions were centrifuged at 3000 rpm (990 g, 25°C) for 30 min. The supernatants containing suspended MoS 2 and WS 2 respectively were carefully collected. WS 2 nanosheets for biocompatibility and bioimaging WS 2 nanosheet colloidal suspensions were prepared by LPE in a 7/3 v/v mixture of IPA and water, following preceding procedure. The supernatant collected after centrifugation was evaporated under vacuum to remove isopropanol and redispersed in Milli-Q water at different concentrations.

Colloidal nanosheet characterization
Determination of nanosheet concentration Nanosheet concentration was determined by weighing the solid residue obtained after filtration of the different colloidal suspensions. Suspensions in NMP (100 ml) were filtered through a PTFE membrane (0.1 μm nominal pore size) and dried at 150°C overnight, whereas suspensions in IPA/water were filtered through a PVDF membrane (0.1 μm nominal pore size) and dried at 100°C overnight.

UV-Visible spectroscopy
The extinction spectra were measured in a quartz cuvette (path length=1 cm) with a Cary 50 UV-visible spectrophotometer. Three independent samples were analyzed for each dispersion type.

Fluorescence spectroscopy
Photoluminescence excitation (PLE) intensity maps, PL and PLE spectra were performed on a Fluorolog ® -3 HORIBA spectrofluorometer. All PL spectra were recorded at an excitation wavelength λ exc =425 nm. The samples were diluted so that their extinction at the emission wavelength does not exceed 0.1, in order to avoid possible re-absorption phenomena. The spectrum shown for each TMD dispersion is the average of the measurements carried out at least on three different and independent samples. Details on the different measurement parameters can be found in the dedicated section D of the supporting information.
Atomic force microscopy Samples were prepared by depositing a drop (10 μl) of the suspensions on freshly cleaved mica substrates and dried under vacuum. The images were acquired using a JPK NanoWizard II AFM working in dynamic mode. NT-MDT NSG01 silicon cantilevers, with typical values of 5.1 N m -1 spring constant and 150 kHz resonant frequency, were employed under ambient conditions in air.

Transmission electron microscopy
Samples were prepared by drop casting of the dispersions onto a holey carbon mesh grid. Low resolution TEM observations were performed using a JEOL-JEM 2100 microscope (LaB 6 electron gun) operated at 200 kV. HRTEM images were obtained with a JEOL-JEM 2100F instrument (field-effect electron gun, 2.5 Å resolution).