Intercalation behaviour of Li and Na into 3-layer and multilayer MoS2 flakes

Lithium (Li) and sodium (Na) intercalation into molybdenum disulfide (MoS2) flakes with layer thicknesses of 2.2 nm (3 layers) and 51 nm (ca. 82 layers) was investigated in situ under potential control via a combination of Raman spectroscopy and optical microscopy. A Raman frequency shift indicative of reduced strain along the MoS2 sheet during Na intercalation compared with Li intercalation is observed, despite the atomic radii of Na being larger than Li, r(Na þ ) 1.02 Å> r(Li þ ) 0.76 Å. Overall, the shift of Raman bands exhibited similar trends in trilayer and multilayer flakes during lithiation. A combination of strain and electron doping was used to explain the observed Raman frequency shifts. The differences between lithiation and sodiation in MoS2 flake were also observed visually by optical microscopy, whereby Li inserted into MoS2 via a pushed-atom-by-atom behaviour and Na via a layer-by-layer behaviour. Variation of the insertion behaviour between lithiation and sodiation in MoS2 was further investigated via galvanostatic intermittent titration technique, in which the diffusion coefficient as a function of x in MxMoS2 (M1⁄4 Li or Na) suggested a stable intermediate phase existed in NaxMoS2 during sodiation, whereas this stable intermediate phase was absent in LixMoS2. © 2019 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). Layered materials such as graphite and transition metal dichalcogenide (TMD) have been widely investigated as energy conversion and storage materials, particular in Li and Na ion batteries [1e8]. Understanding the diffusion of Li and Na ions, the chemistry of intercalation and structural change of electrode materials are of great importance for high-performance ion batteries. MoS2 is a TMD compound in which each layer of MoS2 consists of a sandwich-like configuration with a layer of Mo atoms in between two layers of S atoms, and the MoS2 layers are bound together via van der Waals forces. The result of this weak binding is that the interlayer gap may easily be intercalated by alkali metal atoms or polymers [9e12]. The intercalation reaction is generally accompanied by charge transfer from the intercalant species to the lowest unoccupied conduction band of the host materials and in turn, changes the electronic properties of the host materials [13]. Similarly, intercalation causes the host materials’ optical properties to change. The ability to electrochemically tune the electronic,

Layered materials such as graphite and transition metal dichalcogenide (TMD) have been widely investigated as energy conversion and storage materials, particular in Li and Na ion batteries [1e8]. Understanding the diffusion of Li and Na ions, the chemistry of intercalation and structural change of electrode materials are of great importance for high-performance ion batteries. MoS 2 is a TMD compound in which each layer of MoS 2 consists of a sandwich-like configuration with a layer of Mo atoms in between two layers of S atoms, and the MoS 2 layers are bound together via van der Waals forces. The result of this weak binding is that the interlayer gap may easily be intercalated by alkali metal atoms or polymers [9e12]. The intercalation reaction is generally accompanied by charge transfer from the intercalant species to the lowest unoccupied conduction band of the host materials and in turn, changes the electronic properties of the host materials [13]. Similarly, intercalation causes the host materials' optical properties to change. The ability to electrochemically tune the electronic, magnetic, and optical properties of intercalation compounds makes layered materials attractive for applications like electrochromic displays, optical switches and photovoltaic devices [14e17]. The superconductivity of Na and potassium (K) intercalated MoS 2 compounds at low temperature (T c~1 .3 K for Na compounds and 4.5 K for K compounds) has also attracted wide research interests [13,18]. Transmission electron microscopy, X-ray diffraction and differential optical microscopy have been previously applied to study the process of alkali intercalation into TMDs [19e25]. Through a combination of in situ and ex situ observations critical insight regarding the degradation mechanisms, alkali ion diffusion, and 2H to 1T MoS 2 structure evolution have been provided. These works have helped to understand the role of chemistry and crystal structure on alkali ion diffusion and its concentration dependence, which is of crucial importance in tuning the electronic, magnetic, and optical properties and improving charge and discharge capabilities.
Raman spectroscopy has been used to not only identify the thickness of graphene and TMDs, but study lattice vibration of TMDs under strain and electron doping [26e31]. In the process of Li intercalation of a graphite/graphene sample, with a less than~1.0% lattice increase, one can observe distinctive Raman spectra changes, including the frequency shift and the splitting of the G band [32]. During the early Li intercalation stage (0 < x < 1 in Li x MoS 2 ), there is up to 6% lattice constant increase in the MoS 2 , which will in turn cause strain along the basal plane [33,34], so one would expected that the Raman bands of MoS 2 , E 2g 1 and A 1g , which correspond to the in-plane and out-off-plane vibration, will exhibit prominent shift accordingly [26,35,36]. However, although Raman spectroscopy has been a key tool used to probe the change of physical and electronic properties during alkali atom insertion and extraction in graphite/graphene materials, there is no clear evidence on how E 2g 1 and A 1g bands of MoS 2 will respond to the Li intercalation. Herein the intercalation process of MoS 2 is revisited by using a carefully designed measurement setup, whereby a single MoS 2 flake, combined with a slow discharge/charge rate, and the application of in situ Raman are used to monitor the intercalation process and galvanostatic intermittent titration technique (GITT) is applied to compare the variation in diffusion kinetics of Li and Na intercalating into MoS 2 .

Preparation of MoS 2 single flake electrode
MoS 2 flakes (2D Semiconductors) were mechanically exfoliated onto a borosilicate glass cover slide (200 mm thickness) using the 'Scotch tape method'. The glass cover slide was cleaned in acetone, 2-propanol, and deionized (DI) water in ultrasonic bath, and then subjected to oxygen plasma cleaning to remove adsorbates from its surface. The newly exfoliated MoS 2 flakes on Scotch tape was brought in contact with the glass immediately after plasma cleaning and the tape was removed from glass slowly to leave MoS 2 flake on the substrate. The MoS 2 flake was then connected to Cu current collector using silver epoxy.

Determination of chemical diffusion coefficient
The chemical diffusion coefficients of Li and Na during the intercalation of MoS 2 were determined by using Galvanostatic Intermittent Titration Technique (GITT). Microcrystalline MoS 2 flake electrode were used instead of single flake electrode. The GITT measurement was carried out at discharging rate of 0.1C for 10 min and followed by resting for 10 min while cutting off the current. The ion diffusion coefficient was calculated by using equation (1) [37].
Here t is the duration of the current pulse (s); n m is the number of moles (mol) for the active material; V m is the molar volume of the electrode (cm [3]/mol); S is the electrode/electrolyte contact area (cm 2 ); DEs is the steady-state voltage change, due to the current pulse; DEt is the voltage change during the constant current pulse, eliminating the iR drop. Here the V m of 33.35 cm 3 /mol for MoS 2 was used instead of the V m of the whole electrode. The electrode/electrolyte contact area (S) was replaced by the surface area of the electrode (2.0 cm 2 ).

Confocal Raman spectroscopy and imaging
Confocal Raman measurements were carried out using a Renishaw inVia instrument (laser wavelength 532 nm, <19 kW/ cm 2 ). For Raman imaging, spectra were taken at an area of 50 mm Â 50 mm and then plotted using the intensity of A 1g after subtracting the baseline. Each image contains 50 pixel Â 50 pixel (2500 pixels) in the area of 50 mm Â 50 mm with each pixel having a Raman spectrum of a particular spatial position.

Results and discussion
Highly crystalline MoS 2 flakes of two different thicknesses, namely~2.2 nm (corresponding to 3 MoS 2 layers, denoted as trilayer MoS 2 ) and 51 nm (ca. 82 MoS 2 layers, denoted as multilayer MoS 2 ), were selected for this study. MoS 2 flakes were mechanically exfoliated onto a borosilicate glass cover slide using the 'Scotch tape method' [38]. Flakes of interest were chosen according to the following requirements; the flake should contain a thin and flat region of several square micrometres for monitoring of the intercalation behaviour with Raman spectroscopy, while being sufficiently large (~a few hundred micrometres) for facile connection to an electrode using silver epoxy. Fig. 1 illustrates the assembly process for the in situ spectroelectrochemical Raman cell. After the MoS 2 flake to be investigated was identified through atomic force microscopy (AFM), silver epoxy was used to create an electric connection between the flake and a copper current collector, leaving the area of interest pristine. Position of the flake was aligned to coincide with an aperture (ca. 1 mm diameter) located at the centre of the current collector for direct optical observation (Fig. 1a). The electrode was further assembled into a commercial test cell. Fig. 1b illustrates the configuration of the test cell, with MoS 2 flake acting as the working electrode and Li or Na metal as the counter electrode. 1 M LiPF 6 (for Li) or 0.5 M NaPF 6 (for Na) in 1:1 w/w ethylene carbonate/dimethyl carbonate was used as electrolyte.
Electrochemical intercalation of metal ion between the MoS 2 layers was induced through cyclic voltammetry, while Raman spectra from the flakes were collected at pre-determined intervals during the reaction. A discharge rate of 0.025 mV/s was applied from the open circuit potential (OCP) down to 1.2 V (vs. Li þ /Li or Na þ /Na), in which range little change in Raman signal is observed, as will be shown below. On the other hand, shifts in Raman spectra are observed somewhere between 1.2 V and 0.5 V. Therefore, a slower rate of 0.005 mV/s was applied at this range for closer investigation of the different quasi-equilibrium states. Successful intercalation was also confirmed by optical microscopy from the strong colour changes of the flakes, which is known to be caused by the intercalant and decomposition of MoS 2 . Fig. 2 shows AFM images, height profiles, and Raman spectra of two representative MoS 2 flakes used in this study. The flakes have thicknesses of 2.2 nm (Fig. 2a) and 51 nm (Fig. 2b), which corresponds to 3 layers and~82 layers, respectively. At excitation of 532 nm, MoS 2 exhibit two main Raman bands, namely E 2g 1 and A 1g bands (Fig. 2c). The trilayer MoS 2 shows peak position of 383 cm À1 and 406 cm À1 . A frequency gap of 23 cm À1 between the two peaks matches well with that previously reported for trilayer MoS 2 [29,39]. The multilayer flake shows A 1g band at a slightly higher frequency (408 cm À1 ), which also is in agreement with previous literature. Fig. 3 and Fig. S1 shows changes in the Raman spectra of the trilayer and multilayer MoS 2 flakes during Li intercalation. For both type of flakes, as the potential was tuned from OCP to 1.1 V, the position of the E 2g 1 band remained stable but the position of the A 1g band shifted to a slightly lower frequency. The softening of A 1g mode suggests reduction of interlayer van der Waals forces (i.e., decoupling effect), leading to weaker restoration force in the vibrational mode. As the potential decreased below 1.1 V, both E 2g 1 and A 1g bands shifted to higher frequencies while decreasing in intensity. The E 2g 1 band displayed a shift of up to 3 cm À1 from its original position (383 cm À1 to 386 cm À1 ) for both trilayer and multilayer flakes. On the other hand, while the A 1g band of the trilayer flakes returned to its original position, the A 1g band of the multilayer flakes continued to shift up to 2 cm À1 from its original position (408 cm À1 to 410 cm À1 ). Changes in the Raman spectra during Na intercalation ( Fig. 4 and Fig. S2) displayed several differences compared to that during Li intercalation. First, the E 2g 1 band position remained fairly consistent throughout the reaction for both trilayer and multiplayer MoS 2 flakes. Second, upon reaching a potential of 0.885 V, the A 1g band showed a sudden shift toward lower frequency. Furthermore, the shift of A 1g band was more prominent with the multilayer flake (6 cm À1 ) compared to that with the trilayer flake (3 cm À1 ). The Raman shift of G band (ca. 1580 cm À1 ) in graphite intercalated compounds has been previously investigated using the combined effects of strain and electron doping [32]. Raman spectra of MoS 2 flakes have been reported to be sensitive to both strain and electron doping. For example, a biaxial compressive strain applied to trilayer MoS 2 caused upshift of both A 1g and E 2g 1 bands. With an applied strain of 0.2%, the E 2g 1 and A 1g modes were found to shift bỹ 3 and~2 cm À1 , respectively, indicating that the E 2g 1 mode were more influenced by strain [35]. Raman shift caused by electron doping was reported in a recent study by Sood et al. [40] Using in situ Raman scattering from a single-layer MoS 2 electrochemically top-gated field-effect transistor (FET), the authors demonstrated softening and broadening of the A 1g phonon with electron doping,  whereas the other Raman-active E 2g 1 mode remained inert. The combined impact of strain and electron doping can be used to explain the Raman band shifts observed in Figs. 3 and 4, as discussed below.
In the case of Li intercalation into MoS 2 , the metal ion insertion has been reported to induce an increase of the lattice at the basal plane by 6% [33,34]. Under the applied experimental conditions, it is expected that Li intercalation will result in the expansion of MoS 2 flakes in the in-plane direction along with the out-of-plane lattice, and it is also likely the flake will undergo compressive strain due to the constraint from the glass substrate or structure change caused by phase transformation, leading to the Raman E 2g 1 and A 1g bands shifting to higher wavenumber. Meanwhile, the electron doping will cause the Raman E 2g 1 and A 1g modes softening to lower wavenumber [40]. With a density of 1.8 Â 10 13 e À /cm 2 doping, the A 1g demonstrated a downshift of 4 cm À1 and the change in frequency of the E 2g 1 was not appreciable [40]. shifted more obviously than A 1g in both trilayer and multilayer MoS 2 samples, which agrees with the results of both experimental and first-principles plane-wave calculations based on density functional perturbation theory (DFPT) that E 2g 1 is more sensitive to strain [35,36]. The shift of the A 1g band in multilayer MoS 2 flake is more prominent in comparison to the trilayer sample, likely caused by a less significant decrease of interlayer van der Waals interaction. In the case of Na intercalation, the shifts of the E 2g 1 and A 1g bands are different from those in Li intercalation, most notably the A 1g band shifted in the opposite direction. Previously detailed electron diffraction and XRD studies examined the dependence of the lattice parameters changing against the concentration of intercalant Li or Na in MoS 2 . It has been clearly shown that the lattice parameter 'a' increased monotonically with x up to x z 1 (in Li x MoS 2 or Na x-MoS 2 ) and the maximum lattice parameter 'a' change occurred during lithiation was a 6% increase and in sodiation was only a 1.5% increase [34,42]. These reported volume expansions are contrary to the relationship of the size of Li and Na; as the relatively smaller atom, Li, causes a more prominent in-plane lattice expansion during intercalation. The large expansion in the in-plane lattice can also be proved by the decomposition of MoS 2 into small fragments by TEM under fast Li intercalation (0.1 V/s) [19]. It is reasonable to assume that the dominating factor affecting the E 2g 1 band and A 1g band shifts during Na intercalation is electron doping, since there is only a 1.5% in-plane lattice expansion. It has already been demonstrated that A 1g is more sensitive to electron doping than E 2g 1 is.
Thus, when the electron doping becomes the dominating factor, the overall Raman spectra exhibited no change in the E 2g 1 band and the A 1g band shifted towards lower wavenumbers. The reason why A 1g band shifted more prominently in multilayer MoS 2 flakes than trilayer sample still requires further investigation. The Raman data highlights the differences of structural expansion of MoS 2 during Li and Na intercalation, in agreement with previously reported XRD and TEM results [34,42]. The widely reported 2H to 1T phase transition has also been confirmed by Raman spectra during ion insertion into the multilayer MoS 2 flake (Fig. S3) with the observation of the appearance of weak peaks at around 150 (J1) and 325 (J2) cm À1 at potentials below 1.13 V for Li and 0.90 V for Na [43,44]. Furthermore, the change in intensity ratios of A 1g /E 2g 1 showed different trends between lithiation and sodiation. The E 2g 1 band is the in-plane vibration of S and Mo, and the A 1g band is the out-of-  plane vibration of S atoms (Fig. 5a). The intensity ratio of A 1g /E 2g 1 remains consistent and both bands decreased proportionally during Li intercalation (Fig. 5b). In contrast, during Na intercalation, the intensity of the A 1g band decreased more rapidly than that of the E 2g 1 band and the intensity of the A 1g decreased to approximate a third of that of the E 2g 1 . Strain is unlikely to affect the intensity of the Raman bands [36], therefore the intensity ratio change of E 2g 1 and A 1g results from e À doping effects. Although the e À doping effect exists in both Li x MoS 2 and Na x MoS 2 intercalated compound, doping is the dominant factor causing the observed Raman shift in Na x MoS 2 and the intensity ratio change.
The dynamics of the metal ion intercalation was further examined by the colour changes within the flakes during the reaction. In the case of Li intercalation, a black frontier formed at the edge of the flakes and progressed inwards (Fig. 6). After being held at a low voltage below 1 V for prolonged hours, the whole flake turned black in colour. No Raman peak was observed from the blackened area (Fig. 6f). Even after the voltage was brought back to 3.0 V, the A 1g and E 2g 1 Raman bands did not recover, indicating that the process was irreversible. Since Li intercalation between the MoS 2 layers eventually causes the intercalated compounds to undergo a conversion reaction, the appearance of a black area can be considered as the result of Li x MoS 2 decomposition (equation (2)), explaining the disappearance of the Raman bands for MoS 2 .
On the other hand, Raman bands of the inner area remained stable until eventually turning black ( Fig. 6e and f). The distinct boundary between lithiated regions and non-lithiated regions suggests that the diffusion of Li in MoS 2 is limited to the adjacent site of intercalation. In other words, the intercalation frontier moves inward via an atom-by-atom fashion, where the intercalated Li ions are pushed inwards as more ions are inserted into the MoS 2 layers. The visual observation of the intercalation process was performed only with multilayer MoS 2 flakes, since the colour change within the trilayer MoS 2 flakes was hard to detect due to its transparency.  Interestingly, the intercalation of Na was found to proceed in a very different fashion. As illustrated in Fig. 7, no clear frontier was formed during the reaction. Rather, the whole area of the flake suddenly turned silvery at 0.891 V (Fig. 7b), and then gradually changed to a dark-bluish colour (Fig. 7cee) as the potential was continuously lowered. An abrupt change in the E 2g 1 /A 1g ratio was observed from the Raman spectra taken at 0.889 V and 0.885 V (Fig. 7f). Eventually both Raman peaks disappeared at 0.850 V. The lack of distinct boundary between sodiated regions and nonsodiated regions suggests that the intercalation of Na into MoS 2 occurs in a layer-by-layer fashion, in which the Na ion is welldistributed throughout a relatively large area instead of resting near the initial intercalation site at the edges. Galvanostatic intermittent titration technique (GITT) study was performed to further investigate the diffusion kinetics of Li and Na intercalating into MoS 2 (Fig. 8). In the case of Li, the ion discharge profile showed a flat plateau (Fig. 8b), suggesting that no stable intermediate Li x MoS 2 phase is formed during the intercalation. On the other hand, in the case of Na, the diffusion coefficient profile and ion discharge profile indicates the formation of a stable Na x-MoS 2 (x ¼ 0.5e0.6) intermediate phase (Fig. 8 d and e, Fig. S4). The stable intermediate phase of Na x MoS 2 is referred the as b phase, the Na poor phase of Na x MoS 2 as a phase and the Na rich phase of Na x MoS 2 as g phase. The Na diffusion coefficient (D Na þ ) at x ¼ 0.06 (a phase) is 5.29 Â 10 À11 cm 2 s À1 , at x ¼ 0.6 (b phase) is 2.22 Â 10 À11 cm 2 s À1 and at x ¼ 1.0 (g phase) is 4.84 Â 10 À12 cm 2 s À1 . In the aþb phase and bþg phase D Na þ show a "U" shape and according to the modified theory of GITT for phase-transformation electrodes, these decreased diffusion coefficients are the apparent coefficients, which are usually 2e3 orders of magnitude lower than the real coefficient [45,46]. The decrease of diffusion coefficient is likely caused by the structural change associated with strain and unfavourable energy transitions between stable intercalated compounds [42]. These results confirmed one of several thermodynamically stable phases of Na x MoS 2 during sodiation previously identified by DFT calculations [47]. The process of Li intercalation ( Fig. 8a and b; Fig. S5) clearly exhibited different feature from Na intercalation. The same kind of different behaviour between Li and Na ion also exists in TiS 2 : structure studies on Na x TiS 2 have shown that at least 3 different phases exist in the range 0 < x < 1; on the contrary, Li x TiS 2 did not show staging property [48]. Consolidating all the evidence together, the proposed schematic diagrams of Li and Na diffusion within MoS 2 layers are illustrated in Fig. 8c and f: Li ion intercalates into MoS 2 via pushed-atom-by-atom behaviour and Na ion slides into MoS 2 via a layer-by-layer fashion . In general, the vanishing of Raman bands was considered as the characteristic feature for the MoS 2 decomposition. Indeed, in the Li/ MoS 2 system, the vanishing of Raman bands is related to this irreversible process. However, the Raman signal will recover in the Na/MoS 2 system at certain intercalation stage. A series of experiments to examine the reversibility of Raman intensity were designed. During discharge the voltage was held at 0.885 V, 0.840 V and 0.820 V respectively for at least 2 h to allow the diffusion of Na to equilibrium before taking the Raman spectra at a 50 mm Â 50 mm area on the MoS 2 flake. Then after the voltage was brought back to 2.0 V, another set of Raman spectrum were recorded in the same area. At 0.885 V, the intensity of A 1g band has almost gone, while it was recovered fully after the flake was charged back to 2.0 V (Fig. 9aec, Fig. S6). Fig. 9def shows the Raman intensity of MoS 2 flake recovered from sodiation at 0.840 V, implying the intercalation is still reversible at this voltage. So far, the disappearance of the Raman bands shall be attributed to the e À doping effect and the influence of the intercalants, and these results confirm that disappearance of Raman band is not necessarily related to the decomposition of MoS 2 . The Raman cell was discharged further to 0.820 V before bringing the voltage back to 2.0 V, and this time most of the area on the flake did not show any characteristic peak of MoS 2 , suggesting the decomposition of Na x MoS 2 (Fig. 9gei, Fig. S7). When the intercalated compound Na x MoS 2 decomposing, the value of x is estimated to be 0.98 based on charge/discharge profile of the microcrystalline MoS 2 electrode (Fig. S8). In previous work, it was observed that microcrystalline flake graphite electrode exhibited lower overpotential during lithiation compared to single graphite flakes, likely due to improved electronic contact. Therefore, it is Fig. 9. The reversibility of the sodiation process at different voltage. The microscopic images of the MoS 2 flake at a) 0.885 V, d) 0.840 V and g) 0.820 V after holding at that voltage for at least 2 h, and the yellow square (50 mm Â 50 mm) indicating where Raman spectra were taken from. After Raman spectra were taken, Raman mapping images b), e) and h) were plotted out using the intensity of A 1g . band (the brighter the red colour the more intense the band). Finally, the voltage was brought back to 2.0 V and hold for 2 h before Raman spectra were taken again in the same area and Raman mapping images c), f), and i) plotted out using the intensity of A 1g (scale bar is 50 mm). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) expected that potentials at which the E 2g 1 and A 1g bands changing in single flake samples and microcrystalline MoS 2 electrode samples will have a discrepancy. Therefore, the value of x is only approximately close to the real value. It is important that the distinctive difference between Li and Na intercalation in MoS 2 flake have been captured visually and spectroscopically. These results demonstrate the irreversible nature of alkali metal intercalation into TMDs and highlight the limits to which Li or Na can be reversibly intercalated. Furthermore, the results on trilayer MoS 2 provide useful information for future studies for the comparison of the intercalation behaviour in stacked few layer graphene/MoS 2 hybrid materials.

Conclusions
An in situ Raman spectroscopy study of the electrochemical lithiation and sodiation into large MoS 2 flakes with two different thicknesses, trilayer (2.2 nm) and multilayer (51 nm), revealed a transient Raman shift during Li and Na intercalation due to structural changes of host MoS 2 flakes. The MoS 2 flake with various thickness showed similar trends in Raman frequency shift during lithiation and sodiation, however, the shifts exhibited distinctive difference between lithiation and sodiation. A combination of strain and electron doping was used to elucidate the observed frequency change of the Raman bands during Li and Na intercalation. Raman spectra highlight that the effect of volume change during Li intercalation of the MoS 2 flakes. Furthermore, differences in the diffusion behaviour between Li and Na intercalating into the MoS 2 single flake was observed. GITT measurements highlighted the presence of a stable intermediate phase during sodiation only. Accordingly, it is proposed that Li inserted into MoS 2 via a pushedatom-by-atom process and Na via a layer-by-layer behaviour. The irreversibility of alkali intercalation of TMDs is a barrier for their practical use as negative electrodes in alkali metal-ion batteries. This study highlighted the limits to which one can reversibly insert Li or Na into MoS 2 and revealed kinetic and mechanistic information of electrochemical ion insertion of Li and Na into MoS 2 .