Wet chemical thinning of molybdenum disulfide down to its monolayer

Recent intense studies on two-dimensional (2D) crystals derived from layered van der Waals systems have led to discovery of numerous phenomena that are unique to this emerging class of materials.¹ Atomically thin layers of group 6 transition metal dichalcogenides (TMD) such as MoS₂ and WSe₂ are semiconductors with physical properties that make them attractive for novel electronic and optoelectronic devices.² Studies have demonstrated their potential applications in flexible electronics,³ photodetectors,⁴ light harvesting,⁵ valleytronics,⁶ electrocatalysis,⁷ and gas sensing.⁸ 

Despite the rapid progress of the field, efficient and controlled synthesis of high quality 2D crystals still remains a major challenge. Methods for producing monolayers of group 6 TMDs include top-down approaches such as mechanical “Scotch tape” exfoliation,⁹ intercalation-assisted exfoliation,¹⁰ liquid phase exfoliation,¹¹ as well as bottom-up approaches such as sulfurization of metal thin films,¹² chemical vapor deposition,^13–15 and physical vapor deposition.¹⁶ Exfoliation-based techniques are simple but typically yield flakes with a range of thicknesses and sizes. While the vapor deposition techniques are promising for growth of oriented thin films, current protocols lack control over thicknesses and deposition areas.

Post-deposition etching of multilayer TMD crystals down to single layers is a powerful route to producing 2D crystals in large quantities. Recently, laser thinning of MoS₂ flakes to monolayer was demonstrated by Castellanos-Gomez et al.¹⁷ This technique involves temporary illumination of focused laser beam onto thin MoS₂ flakes. Rapid heating due to laser beam absorption results in vaporization of upper few layers of the flake with the bottom-most single layer remaining intact due to efficient heat dissipation into the substrate. It was found that this technique is applicable to flakes that are ∼12 nm (∼20 layers) or less in thickness. Huan et al.¹⁸ reported chemical etching of MoS₂ flakes down to the last monolayer with xenon difluoride (XeF₂). This technique does not require the initial flake to be thin but the etching is not self-limiting and demands precise control on the reaction rate to achieve monolayers. Similarly, layer-by-layer thinning of MoS₂ by Ar⁺ plasma reported by Liu et al.¹⁹ also requires the thickness of the starting material to be uniform. In this contribution, we report on facile wet chemical approach to etching thin flakes of MoS₂ down to its monolayer. We show that hot nitric acid (HNO₃) efficiently thins MoS₂ flakes on the substrate down to the last one or two layers under optimized etching conditions. The process is self-limiting to some degree and the technique can be applied to thick flakes (>200 nm). We further demonstrate that the monolayer and bilayer flakes obtained by this method retain high electronic quality comparable to that of mechanically exfoliated counterparts.

Natural single crystals of MoS₂ (SPI supplies) were mechanically exfoliated onto silicon substrates with 300 nm thermal oxide using the standard scotch-tape method. The substrate was then floated on top of hot HNO₃ with the MoS₂ flakes in contact with the liquid as schematically shown in Figure 1. The temperature of HNO₃ was kept between 80 and 90 °C and the MoS₂ flakes were allowed to react with the acid over 1 h. The distribution of the etched flakes was found to be non-uniform across the substrate and the yield of mono- and bilayers depended on the reaction temperature and time. Lower reaction temperatures typically resulted in low yield of etched flakes while prolonged reaction led to complete etching of monolayers. We found that reaction temperature of 80–90 °C and reaction time of 1 h result in a reasonable yield of mono- and bilayer flakes. After the reaction, the substrate was cleaned with deionized water and blow-dried with nitrogen for subsequent characterization. Fluorescence images were obtained with a microscope (Olympus, BX51) equipped with a mercury lamp as the excitation light source. Raman and photoluminescence spectra were obtained using 532 nm excitation laser at excitation power of less than 150 μW to avoid sample damage. The electrical properties of the samples were studied in a back-gated field-effect transistor configuration. Source and drain contacts (50 nm Au/Ti) were deposited by thermal evaporation following e-beam lithography. Electrical measurements were conducted in nitrogen atmosphere using a parameter analyzer (Agilent, BA1500A).

Figures 2(a) and 2(b) show the optical images of thin MoS₂ crystals (∼100 nm) before and after etching in HNO₃. The flakes are clearly reduced in lateral size after etching, suggesting etching initiated from the edges. Chemical reactivity of the edge sites can be attributed to the presence of uncoordinated atoms.²⁰ A slight change in the contrast of the flake suggests reduction in thickness also occurred during the reaction. However, the reduction in thickness is negligibly small compared to the etching in the lateral direction, reflecting the highly anisotropic nature of the process.

To our surprise, the etching process yielded atomically thin flakes that were remains of the original flake. We found that these flakes luminesce in red under excitation by a green light (∼540 nm). This red photoluminescence is a signature of monolayer MoS₂, which is a direct gap semiconductor with a band gap of about 1.9 eV, and indicates that the residual flakes are indeed monolayer MoS₂. Height analysis by AFM revealed that the thinnest flakes were ca 0.8 nm thick, comparable to the thickness of monolayer MoS₂ (0.67 nm). We found that most of the residual flakes were monolayer, bilayer, or their mixtures. Figure 3 shows an example of a flake with both monolayer and bilayer regions. These flakes were up to 10 μm in lateral size.

The Raman spectra of chemically etched and mechanically exfoliated flakes are compared in Figures 4(a) and 4(b). The spectra are nearly identical for both monolayer and bilayer samples with the two characteristic Raman modes |$E_{2g}^1$|$E2g1$ (or formally E′) and A_1g (or A₁) appearing at the same frequencies and exhibiting similar peak widths. The Raman signature of MoO₃, which typically exhibits a strong peak near 820 cm⁻¹ (Ref. 21), was not observed. These observations indicate that the crystal and chemical structure of MoS₂ are preserved in chemically etched samples despite the exposure to a strongly oxidizing condition.

Similar comparison of the photoluminescence spectra is presented in Figures 4(c) and 4(d). Chemically etched monolayer MoS₂ exhibits distinct emission peak at 662 nm in agreement with the known optical gap of monolayer MoS₂.^22,23 Compared to the emission spectrum of mechanically exfoliated monolayer, the emission peak of the etched sample is slightly blueshifted and the emission intensity is enhanced by an order of magnitude. These observations suggest that the chemically etched sample is less electron-doped. This can be attributed to the oxidizing effect of HNO₃ that leads to the withdrawal of electrons from the typically electron-rich MoS₂.²⁴ Since the n-type behavior of MoS₂ is likely to be caused by sulfur vacancies, we attribute the observed weaker n-type doping to the passivation of the sulfur vacancies by oxidation. Bilayer samples obtained by etching exhibited emission intensity similar to that of mechanically cleaved pristine samples. This is in agreement with the earlier work that reported that the photoluminescence intensity of bilayer MoS₂ is not strongly susceptible to doping.²⁵ 

Field-effect transistors with chemically etched monolayer MoS₂ (Figure 5(a)) exhibited device performance that is comparable to that of devices fabricated with mechanically exfoliated counterparts. The linear behavior of I_d-V_d output curves shown in Figure 5(b) indicates good ohmic contact. The transfer characteristics (Figure 5(c)) shows a typical n-type behavior commonly observed in exfoliated samples.²⁶ The on-state device conductance of the etched samples was comparable to that of mechanically exfoliated samples. It is worth noting that the threshold voltage is shifted towards the positive gate voltages for the etched samples, indicating weaker n-type character. This behavior is in agreement with the photoluminescence results, which showed lower electron doping in the etched samples. We studied over 10 different samples and found that all devices consistently showed similar behaviors. As discussed above, we speculate that the sulfur vacancies, which are expected to give rise to the n-type behavior of MoS₂,²⁷ were passivated through oxidation by HNO₃. It should be noted that the weaker n-type doping is preserved even after annealing at 200 °C for 2 h in nitrogen atmosphere, suggesting that the effect is unlikely to be due to physisorbed species.

Carrier mobility is one of the key figures of merits for evaluating the electronic quality of the flakes. The field-effect mobility μ_FE of the device was extracted from the transfer curve assuming a linear change of conductance with charge carrier density in the on-state:

Here, C_ox = 12 nF/cm² is the back gate capacitance per unit area and L and W represent the length and width of the channel, respectively. The calculated mobility was found to be 30–50 cm²/V s, which is comparable to that obtained from our mechanically exfoliated monolayer MoS₂. Comparison of 2-terminal and 4-terminal measurements revealed that we underestimate the mobility in 2-terminal devices due to contact resistance²⁸ (see Figure S1 of the supplementary material²⁹). In 4-terminal configuration, field-effect mobility was calculated to be ∼60 cm²/V s. Figure 4(d) shows the contact resistance as a function of gate voltage. At high gate voltages, the contact resistance of 17 kΩ μm, which is comparable to the reported values,²⁸ was achieved. These results further verify the high quality of the monolayer MoS₂ obtained by wet chemical etching.

Finally, we discuss the possible mechanism behind the thinning of bulk MoS₂ crystals down to single monolayer. The etching process relies on the oxidizing power of HNO₃. When sufficient heat is provided, HNO₃ reacts with MoS₂ at the edges to form MoO₃. The oxide can further react with HNO₃ to form molybdic acid (H₂MoO₄), which is soluble in acid:

The formation of MoO₃ as an intermediate product was verified in a modified experiment where the MoS₂ flakes were exposed to hot vapor of HNO₃. In this experiment, the spontaneous dissolution of MoO₃ into the liquid phase is prevented. Indeed, we observed numerous particles of MoO₃ as confirmed by Raman spectroscopy (see Figures S2 and S3 of the supplementary material²⁹). We found that these MoO₃ particles dissolve in either HNO₃ or NaOH.

In order for the bottom monolayers to survive the etching process, the heat from the acid needs to be dissipated efficiently. Since the substrate is in contact with the air and is at a lower temperature than the acid, the substrate acts as a heat sink for the bottom layers that are in contact with the substrate. The etching rate of the bottom layers therefore depends on the heat dissipation efficiency. This is similar to the case of laser-induced thinning where rapid heat dissipation by the substrate plays an important role.¹⁷ To study this effect, we conducted chemical etching experiments with the substrate submerged in HNO₃, thereby keeping the substrate at the same temperature as the acid and avoiding heat sink effect. This resulted in the same lateral etching of the flakes but did not yield monolayers, therefore supporting our hypothesis. In other words, the heat gradient along the c-axis of the MoS₂ crystal, which is determined by the substrate heat dissipation rate, is responsible for the slower etching of the bottom monolayers. To achieve larger monolayer flakes with this method, further optimization of heat dissipation needs to be considered.

In summary, we have designed a facile wet chemical method for thinning MoS₂ crystals down to mono- and bilayers. We showed that hot concentrated HNO₃ reacts with the edge sites of MoS₂ and etches the material through oxidation. We attribute the formation of mono- and bilayer sheets to the heat sink effect of the substrate, which significantly lowers the etching rate for the bottom layers. Raman and photoluminescence spectroscopy along with electrical characterization of these flakes revealed that they preserve their crystal and chemical structure. Our wet chemical top-down approach represents a new unique technique for the preparation of 2D crystals.

G.E. acknowledges Singapore National Research Foundation for funding the research under NRF Research Fellowship (NRF-NRFF2011-02).