Structural Evolution of Electrochemically Lithiated MoS2 Nanosheets and the Role of Carbon Additive in Li-Ion Batteries.

Understanding the structure and phase changes associated with conversion-type materials is key to optimizing their electrochemical performance in Li-ion batteries. For example, molybdenum disulfide (MoS2) offers a capacity up to 3-fold higher (∼1 Ah/g) than the currently used graphite anodes, but they suffer from limited Coulombic efficiency and capacity fading. The lack of insights into the structural dynamics induced by electrochemical conversion of MoS2 still hampers its implementation in high energy-density batteries. Here, by combining ab initio density-functional theory (DFT) simulation with electrochemical analysis, we found new sulfur-enriched intermediates that progressively insulate MoS2 electrodes and cause instability from the first discharge cycle. Because of this, the choice of conductive additives is critical for the battery performance. We investigate the mechanistic role of carbon additive by comparing equal loading of standard Super P carbon powder and carbon nanotubes (CNTs). The latter offer a nearly 2-fold increase in capacity and a 45% reduction in resistance along with Coulombic efficiency of over 90%. These insights into the phase changes during MoS2 conversion reactions and stabilization methods provide new solutions for implementing cost-effective metal sulfide electrodes, including Li-S systems in high energy-density batteries.

Layered transition metal dichacogenides[1] (MX2 where M = Mo, W, Ti; X = S, Se, Te) have recently been applied in solar cells,[2] photodetectors,[3] hydrogen storage,[4] catalysis,[5] Li-ion battery,[6] supercapacitors,[7] transistors,[8] and lubricants.[9] For Li-ion batteries, MoS2 offers almost 3-fold improvement in capacity[10,11] (>800–1000 mAh/g) compared to currently used graphite anodes[12] (theoretical capacity ∼370 mAh/g). MoS2 comprises sandwiched S–Mo–S layers with an interlayer spacing[13] of ∼6.7 Å, which allows Li-ion insertion[14] between layers, similar to graphite. However, in MoS2 electrodes, from ∼1.1 V versus Li+/Li onward, Li+-ions begin to react with sulfur atoms, and MoS2 gradually changes from a trigonal prismatic (2H-MoS2) to an octahedral (1T-LiMoS2) phase.[15] After this phase transition, conversion reactions are intensified at ∼0.5 V versus Li+/Li, causing MoS2 to fragment and disintegrate into LiS2 and Mo nanoparticles. When these reactions coincide with electrolyte decomposition (∼0.5–0.1 V), most of the electrode turns into a gel-like matrix containing Mo nanoparticles. The overall reaction typically leads to low capacity and poor columbic efficiency because, after reactions with Li+-ions, MoS2 electrodes are enriched with polysulfide species (as reaction products), and partially dissolve in the battery electrolyte.[16,17] Low electron conductivity is particularly problematic with the use of standard conductive additives (e.g., carbon particles ∼50–200 nm in diameter), which tend to lose electrical contact with the active particles during the conversion reactions. In order to mitigate these shortcomings, carbon nanotubes (CNTs) have been investigated as a conductive additive resulting in composites with improved electronic conductivity.[18,19] Despite suggested drawbacks related to multiwalled carbon nanotubes (MWCNTs) shedding carbon layers[20] and leading to an increase in solid–electrolyte interface (SEI), CNTs offer superior electrochemical performance as conductive additives. For example, CNT-MoS2 sheets[21] were shown to offer a capacity of ∼512 mAh/g at 100A/g, and the highest first cycle capacity was reported for MoS2-MWCNTs[20] composites reaching up to ∼2774 mAh/g at 100 mA/g, and more recently, hierarchical MoS2 tubular-CNT structures were shown to deliver 800 mAh/g at 5A/g (1000 cycles).[22] These values are larger than those reported for MoS2/carbon nanoboxes,[23] MoS2/polyaniline nanowires,[24] MoS2 nanowall/cellulose,[25] MoS2-graphene,[26] MoS2-polyaniline,[24] MoS2-carbon coatings,[27] MoS2-highly ordered mesoporous carbon,[28] MoS2-disordered graphene-like carbon,[29] and MoS2 hierarchical structures.[14] Although this fact highlights CNT’s commercial potential, it is still unclear how CNT additives cope with metal sulfide conversion reactions that cause large structural degradation and strain in electrodes during battery charge–discharge cycles. In order to understand this, we performed a detailed DFT analysis of likely intermediates that determine reversible and nonreversible reaction pathways during charge–discharge cycles. This is particularly challenging since the Li-MoS2 reaction entails a wide range of off-stoichiometric/intermediate compounds that are short-lived and not often identified but yet significantly influence the battery capacity and rate performance.[30] For example, detection of Mo and MoLi metallic nanoparticles in lithiated MoS2 electrodes poses considerable difficulties because they tend to be considerably amorphous[31] and therefore go unnoticed even in in situ XRD measurements, whereas an ab initio density-functional theory (DFT) analysis[32] reported that, after lithiation of MoS2, the reformation of MoS2 from the converted mixture of Mo and LiS2 is less energetically favorable, indicating that Li-MoS2 is an irreversible reaction. It was also suggested using both DFT and electrochemical analysis that MoS2-Li batteries behave similarly to lithium–sulfur (Li–S) batteries after the first discharge cycle. The exact reaction that metallic Mo nanoparticles undergo during successive cycles is however still not clear.

Our DFT based investigation performs structure searches from random starting atomic locations over a range of stoichiometries, complementing a previous study[28] which biased their predicted structures toward lithiated layered MoS2 phases. Our approach ensures that the structures found are low in energy for a given stoichiometry, allowing the prediction of average voltages. They are highly likely to be thermally accessible at finite temperature and in a disordered system are likely to provide ground-state structural motifs.

By combing DFT simulation and electrochemical analysis we demonstrate that MoS2-Li-ion battery electrodes inherently suffer from a progressive insulating behavior and structural instability from the first charge/discharge. Therefore, the choice of carbon additive which better mitigates the above issues is critical to the performance of these batteries. We carefully compared equal loading of Super P Carbon and commercial double wall carbon nanotubes (DWCNTs) as a conductive additive (≤10 wt %), see Experimental Section. We deduced that these 1D materials form an entangled network that confines the converted products in electrochemically active regions, which otherwise are degraded via reactions with the battery electrolyte as discussed earlier. This is evidenced by a 2-fold capacity enhancement and improved Coulombic efficiency compared to Super P carbon.

Results and Discussion

As shown in Figure a, the MoS2 sheets have an average thickness of ∼7 nm based on measuring tens of individual sheets in HRTEM while the lateral average dimensions are estimated to be around ∼100 nm. The UV–Vis static absorption data of MoS2 dispersion is characteristic of excitons (A and B) originating in 2D MoS2 nanosheets (Figure b) corresponding to photon energies (at ∼1.83–2.05 eV), while higher energies (at ∼2.5–2.9 eV) are associated with transitions taking place between the valence and conduction bands via deep states. Raman spectroscopic data show two characteristic peaks at ∼383 cm–1 (E2g1 in-plane mode) and ∼408 cm–1 (A1g out-of-plane mode), consistent with the S–Mo–S vibration in MoS2 (Figure c). The XRD patterns obtained from MoS2 nanosheet films (on silicon substrates) can be indexed to pure phase MoS2 as shown in Figure d. The samples were characterized to be fully crystalline, and neither side products nor different MoS2 phases were detected. In Figure e the Raman spectra of DWCNTs and Super P carbon (∼50 nm nearly spherical particles) are compared, showing as expected a substantially higher G/D ratio for the DWCNTs. Finally, Figure f shows an SEM image of a DWCNT-MoS2-PVDF slurry after dilution in hexane, casting and drying along with an inset TEM image of the pristine DWCNTs that are used as conductive additive.

Figure 1

(a) High resolution transmission electron microscopic (HRTEM) image of an individual MoS2 flake (scale bar 10 nm); inset showing MoS2 atomic lattice (scale bar 1 nm). (b) UV–vis static absorption spectra of MoS2 nanosheets. (c) Raman spectroscopic data show two characteristic peaks at 383 (E2g1) and 408 (A1g) of MoS2 sheets. (d) X-ray diffraction patterns obtained on MoS2 samples indexed to molybdenite. (e) Raman spectra of Super P carbon and DWCNTs. (f) Scanning electron microscopic image of MoS2-DWCNTs-PVDF composite diluted for imaging (scale bar 500 nm) with an inset showing HRTEM image of the commercial CNTs used in this work (inset scale bar 5 nm).

Coin cell batteries (CR2032 type) were fabricated with MoS2 nanosheet electrodes containing equal loading of either DWCNTs or conventional Super P carbon, and the electrode thickness was maintained at around ∼30 μm (see Supporting Information section S4). Figure a shows a typical discharge profile from the first cycle of both MoS2/DWCNTs and MoS2/Super P electrodes. In the discharge curves, the sloping plateau at ∼1.8 V corresponds to the onset of Li+-ion insertion in MoS2 (according to MoS2 + xLi+ + xe– → LiMoS2). At ∼1.2 V versus Li+/Li, Li+-ions begin to react with S atoms to form a LiS2 matrix in which Mo metallic clusters are embedded (according to 2Li+ + S + 2e– → Li2S).[33] Below 0.5 V versus Li+/Li is a deep conversion zone along with the formation of solid electrolyte interface (SEI). The MoS2-DWCNT electrodes showed a specific capacity of ∼1400 mAh/g while the MoS2-Super P delivered ∼900 mAh/g at 100 mA/g. With equal wt % of carbon additives used, the capacity obtained using DWCNTs corresponds to the theoretical value obtained by DFT simulation (see Figure a and further), and this is often attributed to the ability of CNTs to electrically connect the active particles as they convert into various discharge products that tend to decrease electron/ion conductivities. In addition, it was proposed on the basis of a quantum chemical calculation[34] that in MoS2-CNT composites there is an energetic preference for effective charge-transfer from CNT to MoS2.

Figure 2

(a) Representative discharge profiles from the first cycle of MoS2-DWCNT and MoS2-Super P electrodes together with the average voltages calculated from DFT energies, relative to Li+/Li and assuming each reaction is 2-phase: (i) For the metastable Cmcm symmetry LiMo2S2 phase, the sheets are buckled. (ii) For MoS2, 2 clusters found in a sea of Li, the clusters Li3Mo2S4 and Li5Mo2S4 are found, with the sulfur being pulled away from Mo and toward the Li. (iii) The P21/c symmetry Li3MoS2 leading to segregated Li2S regions. (b) Density-functional theory (DFT) studies of MoS2 electrodes, convex hull of the low energy Li(MoS2) structures found using AIRSS.

In order to illustrate the MoS2-Li-ion battery discharge process (lithiation), the lowest energy structures at each stoichiometry were reoptimized using harder pseudopotentials generated by CASTEP as detailed in Supporting Information section S2. From these structures we obtained the convex hull and average voltages, depicting the formation energies of more stable intermediate phases as reported in Figure b. We find MoS2, LiMoS2, Li3MoS2, Li5MoS2, Li7MoS2, and Li on the hull with Li5(MoS2)2 and Li4MoS2 near to the hull, hence well within the error of DFT and expected to be accessible by thermal excitation at room temperature. The convex hull of LiMoS2 can be described by three distinct regions, x ≤ 1, 1 < x < 2.5, and x > 2.5, characterizing MoS2 battery discharge cycles. In the first region the lowest energy structures are all MoS2-like with various concentrations of Li intercalated between the layers (Li-ion insertion zone). In the second region, clusters of MoS2 with sulfur on the outside, directed toward the lithium, dominate (conversion zone). Finally, in the last region, sulfur is released from Mo. The sulfur bonds with the lithium, forming structures similar to Li2S, and with the remaining Li and Mo, forming Li–Mo alloys (deep conversion zone). Thus, the MoS2 electrodes undergo a series of conversion reactions with Li-ions as the battery is fully discharged at the end of first discharge cycle.

In Figure a,b, we find only one Li3MoS2 phase with low energy. It has P21/c symmetry, and is clearly segregated into Li2S-like and MoS-like regions. All low energy structures of Li4MoS2, Li5MoS2, and Li6MoS2 comprise regions of simple Mo chains and Li2S-like regions. Evidently, DFT shows that at the start of lithiation (LiMoS2x ≤ 1) this region comprises MoS2 sheets with lithium intercalated between them. The LiMo3S6 stoichiometry’s lowest energy structure has a P63/mcm symmetry, and 0.024 eV/f.u. above it lies a P6322 phase. Both contain layers of MoS2 sheets only differing in the lithium sites between adjacent layers. All of the lowest energy structures to at least 0.068 eV/f.u. above the ground state of LiMo2S4 were lithium intercalated MoS2. All of which contain flat MoS2 sheets except a Cmcm phase 0.014 eV/f.u. above the ground state which has buckled sheets (Figure a). The only Li2Mo3S6 phase we found within the cutoff was a layered P63/mcm phase. Also, in this region the lowest energy LiMoS2 found by ab initio random structure searching was the known P1̅ [15] layered phase. However, the electrodes do not suffer from severe structural distortions due to the lithiation of MoS2 layers, similar to Li-ion insertion in commercial carbon anodes[35][36] but offer only one-third of its total capacity (∼350 mAh/g).

After the Li+-ions insertion, the second lithiation stage (LiMoS2 1 < x ≤ 2.5) is a region where MoS2 mainly comprises clusters and chains. The lowest energy Li3Mo2S4 and Li5Mo2S4 structures belong to the I4/mmm and R3̅ space groups, respectively. They both comprise MoS2 clusters, containing 6Mo and 12S atoms toward the outside (see Figure a). These are reminiscent of well-known Mo-halogen/chalcogen clusters, although these normally comprise 6 Mo and 14 halogen/chalcogen atoms. The former has eight 3-fold and four 1-fold coordinated sulfur atoms, whereas the latter has six 3-fold and six 1-fold coordinations, indicating sulfur’s preference to bond with lithium, as lithium concentrations increase. The lowest energy Li2MoS2 phase contains MoS2 chains and has P1̅ symmetry. 0.018 eV/f.u. above the ground state is a C2 phase containing similar clusters to the I4/mmm phase mentioned above, but with clusters sharing some sulfur and containing one fewer Mo. As a result, the MoS2 particles undergo a drastic structural and morphological change, which causes the active particles to lose their interconnectivities and detach from carbon additive as the active particles swell and disintegrate. Then, as the lithiation continues toward completion, DFT indicates a final region LiMoS2 2.5 < x – Li2S where some sulfur atoms are bonded only to lithium. Therefore, in the battery this area appears to be highly amorphous, segregating into Li2S-rich and Mo-rich regions, leading to severe structural discontinuities within electrodes. If carbon additives present in electrodes are not able to retain interparticle connectivity, the batteries show a rapid capacity fade and eventually fail. We find only one Li3MoS2 phase with low energy. It has P21/c symmetry, and is clearly segregated into Li2S-like and MoS-like regions (see Figure ). All low energy structures of Li4MoS2, Li5MoS2, and Li6MoS2 comprise regions of simple Mo chains and Li2S-like regions. Therefore, our DFT results reveal that, at the end of the first discharge, the MoS2 electrodes lose their structural integrity and also accumulate sulfide-rich phases that tend to be electrochemically inactive and electric insulators[37] (sulfur ∼5 × 10–30 S cm–1 compared to 1–4 × 102 S cm–1 of graphite).

In accordance with these DFT results, Figure a,b shows the cycling behavior of both the CNT and Super P electrodes. These electrodes were cycled at a high current density of ∼2A/g which is sufficiently high to cause rapid structural changes of the electrodes and therefore to test the ability of carbon additives to cope with MoS2 electrode reactions in the conversion zones (as in Figure ii,iii). DWCNT electrodes, with the equal loading of carbon additive as Super P carbon references, offer higher capacity and columbic efficiency (CE). The high capacity and CE of CNT electrodes indicates better kinetics and reversibility of the electrode reactions (Figure a,b). Super P in the same loading by wt % apparently fails to stabilize the MoS2 particles which disintegrate during conversion reactions, and tend to expose new surfaces which result in electrically inaccessible domains, which corroborates the measured low capacity and CE. In general, 10 wt % carbon additive is sufficient to achieve a good electrical percolation in electrodes at the start of battery cycling. However, for Super P carbon, this does not seem to compensate for the loss of electrical conduction due to the conversion reactions as described by our DFT analysis. For example, carbon content ∼>30% has widely been used in MoS2 composite electrodes[38,39,28] for achieving high reversible capacity and cyclability, whereas MoS2 electrodes containing ≤10 wt % carbon exhibited poor electrochemical activities and quick capacity fade.[40−42] Nevertheless, the use of carbon additives above 10% decreases volumetrically the loading of active particles, which decreases energy density of electrodes.

Figure 3

Cycling data: (a) MoS2-Super P carbon, (b) MoS2-DWCNTs, (c) electrochemical impedance analysis (EIS) of both types of electrodes, (d) schematics of a Super P carbon and CNT based electrode internal structures.

The MoS2–CNT electrodes show a stable cycling performance (Figure b) despite undergoing an initial fluctuation regime where the electrodes experience dramatic structural changes and produce Li-polysulfides. For example, the amount of strain that is induced by the conversion reactions has been shown to cause cracks in MoS2 electrodes, but CNTs prevent the complete electrical disruption by bridging cracks.[43] We note that the CE % fluctuation in cycling was found to be minimal at a low rate (Supporting Information). The fact that CNT electrodes (Figure b) are stabilized after initial fluctuations in galvanostatic cycling indicates that (i) the MoS2 electrodes undergo a drastic structural rearrangement from the early cycles as confirmed by our DFT, and (ii) the converted electrode particles are then effectively stabilized by the CNT network. The sulfur-rich intermediates, apart from leading to severe disruption of the electrical network within electrodes, can also dissolve in battery electrodes or diffuse between electrodes,[44] and therefore, their stabilization is critical to the battery life which can be improved by using CNTs as conductive additives.

Finally, we performed electrochemical impedance analysis (EIS) to obtain information on the electrode impedance.[45] The representative EIS Nyquist plots in Figure c show that the electrode resistance is substantially lower (∼145 Ω) for MoS2-DWCNT than MoS2-Super P carbon (∼350 Ω). We therefore conclude that CNTs offer a highly conductive scaffold (Figure d) that enhances the electrical conduction throughout battery cycling and increases the utility of Li polysulfide phases, resulting in high capacity and Coulombic efficiency.

Conclusion

This paper presents new insights into the chemistry of a conversion-type MoS2 nanosheet for Li-ion batteries. Using DFT simulations, we show that the Li-MoS2 phase diagram is very rich, especially for higher lithium concentration regions. Ab initio random structure searching predicts a series of Li-polysulfide intermediates for a wide range of stoichiometries. These sulfur-rich intermediates progressively insulate the MoS2 electrodes, and to cope with this, the choice of the conductive additive is critical. By benchmarking CNTs against Super P carbon, the ability of CNTs to stabilize conversion reactions is evidenced. We measured a 45% reduction in electrode resistance, a nearly 2-fold increase in capacity, and Coulombic efficiencies of over 90%. Our results on MoS2 phase transformation and the role of CNT additives in these batteries will facilitate the adoption of metal sulfide electrodes in high energy density batteries.

Experimental Section

Materials

Commercial DWCNTs (NC2000) were purchased from Nanocyl s.a. and used without any pretreatment. MoS2 crystals (micron sized) were obtained from Acros-Organics. Electrochemical grade propylene carbonate, ethylene carbonate, diethyl carbonate, N-methyl-2-pyrrolidone, polyvinylidene fluoride (PVDF), and Li metal foils were purchased from Sigma-Aldrich. Carbon powder (carbon Super P) was purchased from Alfa Aesar. Polypropylene layers were purchased from Cell Guard.

Exfoliation of MoS2

MoS2 crystals were exfoliated followed by a method described elsewhere.[46] Briefly, MoS2 crystals were mixed with N-methyl-2-pyrrolidone (NMP) and sonicated leading to a dispersion of MoS2 nanosheets. The resultant dispersion was centrifuged and redispersed in fresh NMP. The suspension was kept still for several hours to allow for the precipitation of partially exfoliated MoS2 fragments, and this process was repeated a number of times until a clear suspension containing MoS2 nanosheets with an average thickness of 7 nm was achieved.

Li-Ion Battery Fabrication

Commercial DWCNT powder was added to MoS2 nanosheets, and then mixed with PVDF in a weight ratio of 10:85:5 in NMP. The composite formed a thick slurry which then was coated onto a copper current collector and dried at 120 °C for 12 h. The 2032-type cells were assembled using a MoS2-CNT-PVDF slurry coated electrode, and pure Li metal was used both as reference and counter electrode. Polypropylene layers were used to separate the electrodes, and 1 M LiF6P in diethyl carbonate and dimethyl carbonate in volume ratio 1:1 was used as electrolyte.

Instrumentation

Scanning electron microscopy (SEM) images were acquired using a Leo Variable pressure SEM. High resolution transmission electron microscopy (HRTEM) measurements were carried out using an FEI Tecnai F20 operating under 200 kV field emission gun (FEG). UV–vis–NIR optical absorption spectroscopy was carried out using a PerkinElmer lambda 750 UV–vis–NIR instrument. All electrochemical measurements were carried out by using a VMP3 multichannel potentiostat/battery cycler equipped with a 5A current booster from Biologics. Raman measurements were carried out using a EZRAMAN-N instrument on films of samples on Si substrates. X-ray diffraction (XRD) measurements were carried out by using a Bruker D8 θ/θ (fixed sample) instrument with a position sensitive detector (LynxEye) and standard detector (SC).

DFT Simulation

To preduct the likely structures of LiMoS2 that form during lithiation, we used the ab initio random structure searching (AIRSS) method.[47,48] AIRSS has successfully predicted the structures of phases of Li-ion batteries (LIBs) containing sulfur,[49] phosphorus,[50] silicon, and germanium.[51] Not limited to ground-state crystal structures, AIRSS has been used to understand point defects[52] and continuous phase transitions in LIBs.[53] AIRSS searches were carried out on stoichiometries of LiMoS2 at ratios of less than of 8:1 of Li:MoS2 which corresponds to a theoretical capacity of up to 1688 mAh/g. To limit the number of atoms in the simulation cells, stoichiometries were constrained to x + 3y ≤ 11. Initial searches were carried out using the CASTEP 8.03 DFT code with the PBE[54] exchange-correlation functional and Vanderbilt “ultra-soft” pseudopotentials. The Brillouin zone (BZ) was integrated using a Monkhost–Pack grid spacing of 2π × 0.05 Å–1 or better. The basis set contained plane waves up to 300 eV. The stable structures at zero temperature were obtained by plotting the fractional atomic composition against the formation energy per atom (see SI section S2) and drawing a convex hull around the points from 0, 0 to 0, 1. The stable structures lie on these tie lines. To obtain the formation energies and average voltages, the lowest energy structures at each stoichiometry were reoptimized using harder pseudopotentials generated by CASTEP as detailed in Supporting Information section S1.