Nature and Origin of Unusual Properties in Chemically Exfoliated 2D MoS_2

MoS_2 in its two-dimensional (2D) form is known to exhibit many fundamentally interesting and technologically important properties. One of the most popular routes to form extensive amount of such 2D samples is the chemical exfoliation route. However, the nature and origin of the specific polymorph of MoS_2 primarily responsible for such spectacular properties has remained controversial with claims of both T and T' phases as well as metallic and semiconducting natures. We show that a comprehensive scrutiny of the available literature data of Raman spectra from such samples allow little scope for such ambiguities, providing overwhelming evidence for the formation of the T' phase as the dominant metastable state in all such samples. We also explain that this small band-gap T' phase may attain substantial conductivity due to thermal and chemical doping of charge-carriers, explaining the contradictory claims of metallic and semiconducting nature of such samples, thereby attaining a consistent view of all reports available so far.


Introduction:
The discovery of atomically thin layer of graphene from three dimensional graphite crystal by Geim and Novoselov in 2004, 1 opened up a new avenue of research in two dimensional (2D) layered materials.
Overwhelming attention has been focused on the study of the analogous layered materials since then.
Among these the long known, well-studied and technologically important is molybdenum disulphide (MoS2), a member of the transition metal dichalcogenides (TMDs) family. Although investigations on MoS2 can be traced through decades due to its natural abundance in the earth's crust, important catalytic properties 2-4 and extensive usage as solid state lubricant [5][6][7] , it has seen an exponential increase in number of publications recently. 8 Unlike graphene, each layer of MoS2 is three atomic layer thick with a thickness of 6.2 Å, 9 in which the planes of Mo atoms are sandwiched between two atomic layers of S with strong in-plane covalent bonding and between Mo and S planes, while such layers of MoS2 with three atomic planes are vertically stacked via weak van der Waals interactions. This allows for easy mechanical exfoliation of single or few layers of MoS2, ideal for investigating 2D form of MoS2. The large van der Waals gap can also allow different ions to readily intercalate between the MoS2 layers. 8 This route has been often used to chemically exfoliate MoS2 into the 2D form, since such intercalation typically expands the interlayer separation greatly, reducing the coupling between successive MoS2 layers to an insignificant level. Bulk MoS2 is semiconducting with an indirect band-gap of 1.2 eV 10 whereas a monolayer of MoS2 is a direct band-gap (1.8 eV) semiconductor. 11 One of the interesting features of MoS2 is that it can exist in several polymorphic forms, shown in Fig. 1, depending on how the three hexagonal layers of S-Mo-S are stacked above each other. A-B-A type of stacking, with the top and the bottom S layers being directly above each other, gives rise to the thermodynamically stable polymorph H (Fig. 1a), with six S atoms oriented around the central Mo atom in a trigonal prismatic coordination. In contrast, the unstable T form, shown in Fig. 1b, has the A-B-C type stacking with an octahedral coordination of S atoms around Mo. 8 This T polymorph can undergo various Jahn-Teller type distortions leading to the formation of superlattices with different metal-metal clustering patterns, such as a0  2a0 with dimerized zig-zag Mo chains (T') in Fig. 1c, 2a0  2a0 with tetramer Mo-Mo clusters in a diamond formation (T'') in Fig. 1d and 3a0  3a0 with a trimerized clustering (T''') in Fig. 1e. 12,13 Such diverse polymorphic forms are of great importance, since their electronic properties vary greatly, with the metastable T', T'' and T''' phases being semiconductors with varying band-gaps and the T phase being metallic. Because of this tuneability of electronic properties, ranging from wide gap insulator to metal, MoS2 has emerged as a potential candidate for an extraordinarily diverse range of novel applications in different fields, such as transistors, 14,15 optoelectronics, 16 catalysis, 2,17-19 photodetectors, 20 supercapacitors, 21 secondary batteries, 22,23 and even as superconductors. 24,25 MoS2 can be easily transformed to its various metastable states using different routes. These have been extensively studied and reported in literature, such as, plasma hot electron transfer, 26,27 mechanical strain, 28,29 and electron-beam irradiation. 30,31 However, the chemical routes to achieve such transformation have proven to be the most facile and, therefore, popular ones. Chemical routes in turn involve chemical, 32,33 electrochemical alkali metal intercalation, 34,35 or expansion of the interlayer distance by hydrothermal synthesis. 36,37 Although, through all the above-mentioned processes, the stable H phase is known to be transformed to one of the metastable states, the structure and electronic properties of the resultant phase have still remained highly contentious with many conflicting claims and ambiguities. Theoretical calculations predict that the Jahn-Teller distorted T' and T'' are small bandgap semiconductors 38,39 and T''' is a ferroelectric insulator. 40 Interestingly, the undistorted T phase is theoretically predicted to be dynamically unstable as phonon dispersion of this phase shows an instability at the zone boundary. 38 Despite such distinct properties expected of each variant, most of the experimental papers, dealing with such chemically treated samples do not clearly identify the specific phase formed, often using the term T or in few cases T' in a generic manner to denote a metastable phase. There are also several reports where instead of identifying any crystallographic phase, the additional phases formed due to such chemical treatments are classified by their presumed electronic or transport properties and termed as metallic or semiconducting MoS2. 33,[41][42][43] Unfortunately, the generic use of T to denote the metastable form and the frequent claim of a metallic nature have created an impression in the community that the metastable state formed is predominantly the metallic, undistorted T phase and not one of the small band-gap semiconducting, distorted T', T'' or T''' phase. We critically scrutinize this dogma by looking at all relevant data already available in literature to arrive at the contrarian view in this perspective.

Chemical exfoliation of MoS2:
Briefly, there are three distinct chemical exfoliation routes employed for MoS2, namely chemical intercalation, electrochemical intercalation and hydrothermal or closely related solvothermal synthesis.
Schematic representations of these two routes are shown in Fig. 2. Historically, intercalation has been applied to layered materials as a means of exfoliating individual 2D layers from their bulk counterparts in large quantities. Intercalation chemistry plays a key role in a majority of the liquid-based exfoliation methods which in contrast to the mechanical exfoliation, presents great advantages for the mass production of 2D materials. 44,45 The key principle for the intercalation-based exfoliation is to increase the interlayer spacing between individual layers by inserting foreign species. This weakens further the already weak interlayer van der Waals interaction and reduce the energy barrier of exfoliation. 44 Although some research on the intercalation of different alkali metals into MoS2 has been reported, [46][47][48][49] most of the attention has been focused on the intercalation of lithium (Li). This is based on the expectation that Li + ions, with the smallest ionic radius among all alkali metal ions, will easily enter the interlayer space and also because of the potential of such Li-intercalated materials as components of high-power rechargeable batteries. The chemical route of Li-intercalation, developed by Joensen and co-workers, involves treating MoS2 with n-butyl lithium (n-BuLi) in hexane as the intercalating agent followed by a water exfoliation step. 50 Schematic illustration of the procedure is shown in Fig. 3a. The key step of this procedure is the formation of LixMoS2 via a slow process, requiring Li-intercalation about 48 hours or more. The lithiated solid product is retrieved by filtration and washed with hexane to remove excess Li and organic residues of n-BuLi. In the next step of washing with water, the bare alkali ions are immediately solvated by water molecules which form number of layers in the Van der Waals gap facilitating the exfoliation process and also stabilizing the monolayers in the solvent.
However, for the solvated phase, the Li content is significantly lower than in the intercalated compounds prior to washing with water. 51 For the solvated phases of these type of compounds, the alkali metals remain almost fully ionised and the guest (alkali atoms) and host charges (residual negative charges on disulphide layers) remain separated by solvent layers. In the solvated phase, the expansion of interlayer spacing with respect to that of the pristine compound naturally depends on the number of solvent layers formed in the interlayer space, which in turn depends on the ionic radius of the intercalated alkali metal. b.

e.
Thus, the distance between the adjacent layers change in a stepwise manner, depending on whether the solvating molecules form a monolayer or bilayer, as demonstrated in Fig. 3b. 51,52 The hydration energy, characterised by charge/radius ratio, of Li + and Na + is greater due to the smaller radius compared to the other alkali metal ions, leading to the formation of two water layers in the intercalation compound for these two guest ions, whereas all the other cations stabilize with mono-layered packing of water in the interlayer space. While we focus our discussion primarily on Li intercalation route in this article, in view of its pre-eminence in the published literature as the most preferred route, we note in passing many other investigations of chemical intercalation driven exfoliation of MoS2, involving other alkali ions. 49,53,54 Electrochemical intercalation allows a considerably higher control on the amount of Li-intercalated, while also achieving a faster rate of intercalation for small quantities of the host, compared to the chemical intercalation route. In general, the Li + electrochemical intercalation (see Fig. 3c) is performed in a test cell using a Li foil as the anode, LiPF6 or LiClO4 in propylene carbonate as the electrolyte, and MoS2 as the cathode using galvanostatic discharge at a certain current density. 55 The advantage of this method is that Li + insertion can be monitored and precisely controlled, so that the galvanostatic discharge can be stopped at the desired Li content to avoid decomposition of the Li-intercalated compounds by optimizing the cut-off voltage.
We note that this analysis does not provide any indication of how intense or weak a specific symmetryallowed Raman signal may be; therefore, it is entirely possible that a symmetry-allowed Raman signal is not observed in an experiment due to its low intensity. In this sense, the above consideration helps to establish a rigorous upper limit on the number of peaks one may observe for a given phase of MoS2 and not the lower limit. Before turning to the available information on Raman spectra of these samples in the literature, we note the reason to consider the undistorted T phase as a possible candidate, despite the undeniable theoretical result that this is an unstable phase that will spontaneously distort itself into one of the lower energy T', T'' or T''' phases; in other words, theoretical analysis shows that it is purely unstable and cannot be a metastable state. However, such theoretical analysis bases itself on the longrange periodic structure of the T phase, whereas the observed metastable states of MoS2 coexist as small patches within the domains of the H phase; this leads to several additional effects, such as the finite size effect, strains generated across the two phase boundaries and possibilities of charge doping, making the real scenario very different from the idealized case considered by theoretical approaches; many of these additional effects may have the ability to make the finite-sized and embedded and/or charge-doped T phase a metastable rather than an unstable phase under the experimental realization.     Table 2, we have represented these results also in the form of a plot in Fig. 4. Table 2 and Fig. 4 together make evident a few interesting observations. First, it appears that the claim of the T phase formation was relatively more abundant till about 2017, while the claim of the formation of T' phase has become relatively more frequent in recent years. We also note that most of the Raman spectra reported in the literature invariably exhibits signatures of the H phase with peaks appearing at ~382 and ~405 cm -1 . This suggests that the conversion of bulk MoS2 to its few layered 2D form via chemical exfoliation generally does not lead to a complete transformation of the stable H phase to its various metastable T forms. Most importantly, we find that all Raman spectra published so far to provide evidence of metastable states in chemically From these figures, it is evident that both T'' and T''' phases have bandgaps (see Fig. 5e and 5f) of about 14 and 57 meV, significantly smaller than that in the H phase (Fig. 5b). The T' appears to have a Dirac cone formed between B and Γ points, as illustrated in Fig. 5d. However, it has been shown by some authors that incorporating spin-orbit coupling (SOC) within the Mo 4d in the calculation leads to the splitting of this Dirac cone and opening of a band-gap of ∼ 50 meV (Fig. 5g). 39 There are other reports suggesting the formation of similarly small band-gap (≤ 100 meV) for the T' phase by several groups. 32,87 Since some of these calculations do not involve SOC, the essential role of SOC in forming the band-gap in the T' phase is not fully established. SOC decreases the band-gap of the T'' structure, while the band-gap of the T''' structure is almost unaffected by the SOC showing in Fig. 5h and 5i, respectively.
We now focus on experimental investigations leading to claims of formation of specific polymorphs, other than the stable H phase, due to chemical exfoliations of MoS2 in the 2D forms. Apart from T and what is the electronic properties of that state? Is it metallic or semiconducting? Theoretical considerations, of course, suggest that T phase in the extended bulk form, is not even a metastable state but a dynamically unstable state; 38,40 in addition, the formation energy of the T phase is quite high compared to that of the T' phase, 39 making the formation of T phase in preference over T' phase unlikely.
As already shown in Table 2 g. h. i.
highly insulating regions, marked black on a surprisingly conducting major phase on the surface of a well-known insulating H phase, possibly arising from the limitations of the technique or the sample. On the other hand, more relevant to the present discussion, the map of the chemically exfoliated sample in Fig. 6b presents a highly conducting homogeneous surface. I-V diagram, also shown in Fig. 6b, was used to assert the metallic nature of the chemically exfoliated sample. In passing, we note that I-V curve of the exfoliated sample in Fig. 6b is unusual with a highly non-ohmic, switching behaviour for an extremely small threshold voltage followed by current saturations, not expected of a usual metal.
Despite these unusual aspects, these experiments point to a reasonable conductivity of the chemically exfoliated sample. Absorbance data of chemically exfoliated samples have also been used 33 to suggest the existence of the metallic T phase. This work 33 also reports the systematic change in the optical absorbance spectra, accompanying the conversion of the metastable phase to the stable one as a function of the annealing temperature. These absorbance spectra are shown in Fig. 6c. Spectral features, preferentially present in the sample with higher proportion of the large band-gap H phase following the higher temperature annealing, are interpreted as excitonic features. It was argued that the absence of any excitonic peak in the exfoliated sample at room temperature indicated that, the as synthesized, exfoliated sample was metallic and, consequently, existed in the T polymorphic form. However, as pointed out in Table 2, the as-synthesized sample in this case exhibited J1 (151 cm -1 ), J2 (229 cm -1 ), E1g (300 cm -1 ) and J3 (332 cm -1 ) peaks in its Raman spectrum and this is inconsistent with high symmetry of the undistorted T phase. In this context, we note that there is a significant level of absorbance, extending to the longest wavelength displayed in Fig. 6c, for the chemically exfoliated sample without annealing. This featureless absorbance must be associated with the metastable phase formed, since the absorbance is found to decrease systematically with an increasing conversion of the metastable state to the stable H phase with a large band-gap (~ 1.8 eV or ~ 690 nm) with the successively higher annealing temperatures. This allows for the possibility of a small band-gap existing for the metastable phase, since the excitonic peak for that phase will appear in the vicinity of its band-gap. Considering that the various estimates of the band-gap of T' phase is smaller than 100 meV, the spectral features in Fig. 6c do not exclude the possibility of the metastable state being T' phase with its observable excitonic features lying outside of the wavelengths probed and presented in Fig. 6c.
It is well-known that the issue of metal/insulator property can be most easily probed by photoelectron spectroscopy, as it maps out electron states directly. Thus, it characterises a metal by the presence of a finite photoelectron spectral intensity at the Fermi energy, while an insulator is characterized by a finite energy gap between the Fermi energy and the onset of the finite spectral intensity. In order to enhance the contribution from the metastable phase to photoelectron spectra, we performed 32 scanning photoelectron microscopy experiment with a photon beam size down to 120 nm. The photon beam was positioned on the sample to maximize the contribution of the metastable state and the valence band spectra were obtained from the same spot. A magnified view of the energy region around the Fermi energy is shown in the inset of Fig. 6d. Clearly, the spectral intensity at the Fermi energy is negligibly small for both the H phases, termed meMoS2, and the chemically exfoliated sample (ceMoS2), establishing the semiconducting nature of the metastable state, consistent with the interpretation of the formation of the T' phase. The above interpretation, however, does not explain why several past experiments found evidence of substantial conductivity for such metastable states. One possibility of course is that the thermally excited charge carriers are substantial for such a small band-gap semiconductor, whereas such charge carriers will be entirely negligible for the large band-gap H phase.
In addition, we need to consider the possibility of charge carrier doping of such a small band-gap  c. d.

e.
semiconductor as a plausible origin of the observed conductivity. In order to address this possibility, we prepared samples with different extent of Li + ions present in it by simply varying the water washing cycle after Li-intercalation. 78 The valence band spectra from the two extremes of washing, namely no washing at all (termed 0W) and after 12 cycles of washing with water (12W) are shown in Fig. 6e.
These spectra with essentially zero intensity at the Fermi energy, establish both samples as small band gap semiconductors. We also find that the valence band spectrum obtained from 12W sample is shifted towards the higher binding energy side by almost 0.2 eV, indicating electron doping of MoS2 by the Li ions. Therefore, it is possible that such charge doping of the chemically exfoliated MoS2 samples contribute to the conduction though the polymorphic phase remains T', as suggested by the Raman frequencies and low photoelectron spectral intensity at the EF.
In summary, we first discussed different routes to chemical exfoliation of bulk MoS2 that provide the most convenient ways to synthesize copious amounts of 2D MoS2. Such chemical exfoliation has been shown to give rise to several polymorphs of MoS2 in addition to the most stable H phase and many interesting properties and device applications of such samples have been attributed in the past literature to the presence of these additional phases. Surveying the existing literature, we help to focus on the ambiguities present in identifying the dominant polymorphic phase present in such samples and show that the existing literature in terms of Raman spectra provides an overwhelming evidence in favour of the T' phase being present, rather than the often stated T phase. Since the T' phase is known to be semiconducting, we then address the puzzling issue of several more direct probes of the electronic structures of these samples appear to point to a highly conducting state of such exfoliated samples. We show that the substantial conductivity has to be understood in terms of thermal and dopant induced charge-carrier dopings of the small band-gap T' phase, rather than in terms of the formation of the unstable, metallic T phase.