Insights into the MoS₂ lithium-ion battery lithiation mechanism via big data electrochemical analysis

Abstract

Molybdenum disulfide (MoS₂) is a potential alternative to the commercial graphite anode in lithium-ion batteries (LIBs) due to its higher theoretical capacity (670 mAh g⁻¹) and natural earth abundance. However, the deep discharge potential range (3.00–0.01 V) enabling MoS₂ high specific capacities is highly debated and suffers from a rapid degradation process (<40 cycles). To investigate the MoS₂ deep lithiation mechanism, long-term electrochemical testing, restrictive potential ranges after an initial deep discharge, and in situ electrochemical impedance spectroscopy (EIS) were applied. Electrochemical analysis reveals that the three cathodic reactions (labelled G, H, and I) are either parallel or series-independent, forming reversible redox couples with the single anodic peak (W). The inclusion of each additional cathodic peak increases cell capacity but accelerates capacity decay. As the redox peaks responsible for the high initial capacities vanish, we verify that capacity is stored entirely through capacitance in the long-term (>40 cycles). Our electrochemical analysis highlights that current MoS₂ lithiation mechanism proposals are unable to accurately describe the phenomena observed when restricting the potential range of MoS₂ after deep lithiation. Therefore, we propose a set of three electrochemical conditions that must be justified by any lithiation mechanism hypothesis.

Parallel or series-independent reaction scheme of molybdenum disulfide (MoS₂) during deep discharge lithiation dQ/dV.

Introduction

Fossil fuels are currently the central energy source for humankind,¹ causing undeniable climatic consequences²via unrestricted worldwide emissions, disproportionately generated by the world's richest nations. Due to the global environmental impact and stringent control of oil sources by a handful of nations, many governments have pivoted towards electric vehicles (EVs)³ and renewable energy sources. However, the lithium-ion battery (LIB) production infrastructure on which these technologies rely is dominated by Chinese manufacturers and supply chains, which are subject to governmental export restrictions on battery materials.⁴ Therefore, high-energy density materials available at scale that can diversify the anode side graphite monopoly via direct drop-in substitution, are attracting academic interest.

Molybdenum disulfide (MoS₂) is an abundant, cheap, chemically stable, and non-toxic transition metal dichalcogenide with an established global supply chain (∼5280 T per year)⁵ as a solid lubricant.⁶ MoS₂ was first conceived as a cathode material for LIBs in the 1980s,⁷ but since the 2010s was targeted to replace graphite (∼360 mAh g⁻¹) as a next generation anode.⁸ Its high theoretical capacity (∼670 mAh g⁻¹) is calculated assuming a four-electron conversion reaction.^9–12 However, commercially available bulk MoS₂ flakes cannot sustain such high capacities long-term due to poor cycling stability.^9,13,14 Additionally, the study of MoS₂ is complicated further by the existence of a complex lithiation mechanism^{9–11,14–26} and the amorphization of the electrode material upon first lithiation.²⁷ Therefore, most studies focus on the synthesis of intricate MoS₂/carbon composites through expensive laboratory techniques⁸ to enhance the cycling stability of low loading electrodes, whereas electrode and cell level engineering design approaches to scale-up bulk MoS₂ available at the ton-scale have been largely omitted.^9,42

MoS₂ + xLi⁺ + xe⁻ ↔ Li_xMoS₂ (0 ≤ x ≤ 1; 160 mAh g⁻¹; 0.80 V) 1

Li_xMoS₂ + (4 − x)Li⁺ + (4 − x)e⁻ → 2Li₂S + Mo (∼510 mAh g⁻¹; 0.01 V) 2

MoS₂ + 4Li⁺ + 4e⁻ ↔ 2Li₂S + Mo (670 mAh g⁻¹) 3

16Li⁺ + 16e⁻ + S₈ ↔ 8Li₂S (1672 mAh g⁻¹) 4

The shallow MoS₂ potential range (3.00–0.80 V in Fig. 1a) is considered stable and delivers a low long-term capacity of 114 mAh g⁻¹ (Fig. 1b).⁹ The first shallow discharge includes a single plateau A (∼1.0 V), whilst all following charge (Y 2.0 V & X 2.6 V) and discharge (E 1.8 V & C 2.5 V) cycles contain two sloped plateaus each, giving a symmetric charge–discharge profile⁹ (Fig. 1a). The initial lithiation plateau A forms Li_xMoS₂ (eqn (1)),^{9,10,16,20,25} while the consensus literature shallow discharge mechanism suggests that all following cycles operate via a multi-step intercalation (Li_xMoS₂) and de-intercalation (MoS₂) process through the E/Y and C/X plateau pairs⁹ (Fig. 1a). The only alternative mechanism proposal by Stephenson et al. considers the shallow potential range to include both Li_xMoS₂ intercalation/de-intercalation across plateaus A/Y and lithium–sulfur (Li–S) conversion/reformation through (E + C)/X.²⁸

Fig. 1. LIB MoS₂ electrodes (∼47 µm) cycled at a galvanostatic constant current density of 100 mA g⁻¹versus a Li metal counter electrode under different potential ranges. First and second cycle potential profiles of (a) shallow 3.00–0.80 V and (c) deep 3.00–0.01 V discharge ranges. (b) Discharge capacity plots of potential windows 3.00–0.80 V, 3.00–0.60 V, 3.00–0.40 V, and 3.00–0.01 V.

On the other hand, the deep discharge potential range (3.00–0.01 V) exhibits an extremely high initial capacity of 697 mAh g⁻¹ (Fig. 1c) but degrades rapidly in the first 40 cycles^9,12,13,27 (Fig. 1b). The initial deep discharge includes plateaus A and B (0.5 V in Fig. 1c).⁹ It is generally agreed in the literature that plateau B involves the formation of amorphous Li₂S and molybdenum metal nano clusters^{9–11,15,17–20,26,29,30} (eqn (2)), although our own ex situ first cycle studies of coin cell electrodes found a large degree of radial lithiation inhomogeneity (Fig. S1), material inactivity (pristine 2H MoS₂ detected), product variation (amorphous 1T/2H MoS₂ present), and depth compound differences.^27,31

The presence of deep discharge plateau B onsets an irreversible change in the electrochemical behaviour of the cell, causing the following cycles to exhibit asymmetric profiles (Fig. 1c), containing a single elongated charge plateau W (∼2.2 V) and three smaller severely sloped discharge plateaus G (∼1.9 V), H (∼1.1 V), and I (∼0.4 V).⁹ The chemical reaction mechanism behind the W, G, H, and I plateaus is highly debated, with proponents siding with either a reversible MoS₂ (REV) or irreversible Li–S (IRR) mechanism. The reversible mechanism considers MoS₂ lithiation into Li_xMoS₂ at (G + H) (eqn (1)), conversion of Li_xMoS₂ into Li₂S + Mo at I (eqn (2)), and MoS₂ reformation from Li₂S + Mo at W (eqn (3)).^10,17,18,26

Alternatively, the irreversible mechanism assumes that Li₂S + Mo formation at B is irreversible (eqn (2)) and all reversible capacity in the cell thereafter originates from (G and/or H)/W through incomplete Li_xS_y/S or complete Li₂S/S (eqn (4)) sulfur reactions.^{9,11,15,16,22,28,32} Additionally, the irreversible mechanism relies on a single (I)⁹ or two-step (H + I)²⁸ lithiation and conversion reaction plateau assignment for leftover pristine MoS₂, to account for the remaining plateaus. There also exist mixed (MIX) mechanism proposals that consider the presence of elements from both the REV and IRR reaction schemes at the same time.^11,22

Despite the shallow (A, Y, X, C, & E) and deep (A, B, W, G, H, & I) discharge electrochemical plateaus (Fig. 1) being identified on multiple occasions^9,11–13,16 and their development/fading being well documented in the literature,⁹ MoS₂ lithiation studies typically focus on ex situ, in situ, or operando material characterisation of the first cycles,^{9,11,16,19,20,29} without considering simple information beyond the specific capacity (mAh g⁻¹) that can be extracted from basic electrochemical techniques, data treatment methods, potential ranges, and long-term cycling.

Although multiple plateau assignments have been proposed in the literature,^9–11,13,26 to the best of our knowledge, thus far no attempt to systematically assign or isolate redox pairs according to proposed electrochemical reaction schemes has been conducted. Presently, the only detailed electrochemical studies of MoS₂ focus on in situ electrochemical impedance spectroscopy (EIS) of the first two cycles,¹⁴ electrode parameter selection (coating thickness and binder),¹³ observation of a surface gel-like layer,³³ and establishing that morphology dictates the long-term capacitance-based storage.²⁶ Hence, we believe that big data electrochemical characterisation of the shallow and deep discharge systems is crucial to further expand our understanding of the MoS₂ lithiation mechanism.

Therefore, herein we investigate the effects of potential window (3.00–0.80 V and 3.00–0.01 V) and mass loading (9–200 µm; 1.0–8.3 mg cm⁻²) on MoS₂ long-term stability. We observe that the shallow discharge suffers from a gradual build-up of resistance due to continuous electrolyte consumption, whereas the deep discharge loses its electrochemical plateaus quickly and relies on low-capacity capacitive storage long-term.²⁶ Applying restrictive potential ranges,^13,14 we electrochemically demonstrate that the deep discharge plateaus G/W, (G + H)/W, and (G + H + I)/W all form viable redox pairs. Therefore, we suggest that the deep discharge lithiation mechanism (<40 cycles) must rely on a parallel or series-independent reaction scheme producing a single product or product class, that reforms through the single anodic plateau W. Hence, we provide a set of electrochemical conditions that must be satisfied by lithiation mechanism hypotheses, as MoS₂ LIB ex situ material characterisation alone has been shown to be strongly affected by electrode disc radial and depth positioning.^27,31

Experimental details

Materials

MoS₂ powder (98+%) with an average flake size of 6 µm was purchased from Sigma-Aldrich (CAS: 1317-33-5), carbon black additive (Super P®, 99+%) was purchased from Alfa Aesar (CAS: 1333-86-4), and Solef® PVDF (Polyvinylidene Fluoride) was purchased from Solvay®. NMP (2-methyl-pyrrolidone) was purchased from Sigma-Aldrich (CAS: 872-50-4). Copper foil with an average thickness of 11 µm was purchased from Cambridge Energy Solutions. 1 M LiPF₆ in ethylene carbonate/di-methyl carbonate (EC/DMC 3 : 7 vol.) electrolyte was purchased from PuriEL (SoulBrain MI). Lithium chips were purchased from MTI Corporation.

MoS₂ electrode manufacture

MoS₂/Super P/PVDF electrodes (80 : 10 : 10 wt%) were manufactured following the recipe provided in our previous study.²⁷ A 5 wt% PVDF in NMP mixture was formed by adding 3.00 g of PVDF powder to 57 mL of NMP and stirring magnetically for 24 h to form a homogenous binder dispersion. Then, in a typical process 100 mg of MoS₂ powder, 12 mg of Super P powder, and 280 mg of PVDF (5 wt%)/NMP mixture (≈14 mg PVDF) were added to a plastic THINKY container, giving a slurry formulation of 80 : 10 : 10 wt%.

The mixture was homogenized with planetary mixer (THINKY ARE-250) in a two-step process, including 0.5 min at 500 rpm and then 15 min at 1500 rpm. The sequence was repeated a second time to avoid aggregates in the slurry. The slurry was then cast onto copper foil with a doctor blade with an automatic coater (Elcometer 4340). The wet electrode was dried using a hot plate at 60 °C until visibly dry. The electrode thickness was measured with a manual micrometre (Mituyo) and the current collector average thickness (11 µm) was subtracted.

Electrodes were cut into 11 mm or 15 mm diameter discs. The average mass of a blank copper disc is ∼7.2 mg and ∼15.3 mg (∼8.7 mg cm⁻²), respectively. The average electrode laminate coating thickness and active mass loading is displayed in Table S3.

Material characterisation

SEM was performed with a Zeiss EVO MA 10 SEM with a tungsten thermionic electron source and a scintillator-type Everhart–Thornley secondary electron detector. The accelerating voltage was set to 10 kV.

XRD patterns were collected on a Bruker D2 Phaser benchtop XRD in reflex geometry, with a Cu source tube (∼1.541 Å) operating at 30 kV and 10 mA. All samples were scanned from 1–80°, with a 2.00 mm incident slit, a 1.00 mm blade gap, a 0.01° step increment, and 0.5 s/step exposure. Samples were directly placed on a silicon holder.

Raman spectroscopy was measured in a Renishaw In-Via microscope with a 785 nm laser wavelength (300 mW), operating with a ×50 short distance working objective and a 1200 L mm⁻¹ diffraction grating. Scans were carried out at 1% power, 3 accumulations per scan, and 10 s exposure per accumulation. Data was normalized relative to the maximum intensity of each dataset and offset for comparison. Figures are split at 500 cm⁻¹ for readability, so that the MoS₂ E¹_2g and A_1g modes do not dominate the less intense peaks at the higher end of the spectrum.

Coin cell assembly & electrochemical testing

Coin cell assembly was carried out in an inert glovebox environment (UCL – MBraun; H₂O, O₂ < 0.5 ppm). The MoS₂ electrodes were dried further under vacuum at 60 °C in a Buchi tube oven before transferring into the glovebox. Hohsen CR2032 were assembled using a Celgard separator (19 mm diameter and 25 µm thickness), 100 µL of 1 M LiPF₆ in EC/DMC (3 : 7 vol.) without any additives, an oversized lithium metal counter (MTI Corporation), a 1 mm spacer, and a spring. Each MoS₂ electrode was weighed before assembly and its mass is displayed in Table S4.

Galvanostatic constant current cell cycling (100 mA g⁻¹) was carried out on a Biologic BCS (UCL) or Neware (QMU) cycler in the potential ranges of 3.00–0.80 V, 3.00–0.60 V, 3.00–0.40 V, and 3.00–0.01 V with 5 minute rests between charge/discharge. Restrictive potential ranges of 1.80–0.01 V, 3.00–1.50 V, and 3.00–0.80 V were also carried out after an initial deep discharge (D1 0.01 V). Cyclic voltammetry (CV, 0.1 mV s⁻¹) was carried out for potential ranges of 3.00–1.30 V, 3.00–1.10 V, 3.00–0.80 V, and 3.00–0.01 V on a Biologic BCS (UCL) cycler, with intermittent galvanostatic constant current density (100 mA g⁻¹) cycles to speed up the process. In situ EIS was carried out on a Biologic BCS (UCL) (10 mV; 10 mHz–10 kHz) before and after charge/discharge.

All electrochemical data processing was carried out utilizing custom Matlab (R2022b) scripts designed to rapidly extract and compare large sets of cell data. ZView was utilized to fit EIS spectra and Origin (2022b) PeakAnalyser was used for dQ/dV and CV peak energy analysis.

Results & discussion

Mass loading & potential window influence on capacity

Experimental MoS₂ electrodes are usually manufactured with an 80 : 10 : 10 wt% ratio of active material (i.e. MoS₂) to conductive additive (i.e. carbon black) to binder (i.e. polyvinylidene fluoride – PVDF) (Table S1) and low active loadings (0.4–2.0 mg cm⁻² in Table S2). Therefore, standard MoS₂/Super P/PVDF electrodes (80 : 10 : 10 wt%) were prepared by casting a 2-methyl-pyrrolidone (NMP) based slurry onto a copper current collector (11 µm) with a doctor blade set to a desired thickness (10–200 µm; Tables S3 and S4).^13,27

Although PVDF/NMP slurry formulations are being phased out due to their toxicity, environmental impact, and cost in favour of water-based formulations (e.g. carboxymethyl-cellulose – CMC)³⁴ or dry-processing altogether,³⁵ PVDF/NMP electrodes remain the most widely used for MoS₂ lithiation studies (Table S1). The as-cast MoS₂ electrode laminates were validated for consistency visually (Fig. S2) as well as with scanning electron microscopy (SEM; Fig. S3), X-ray diffraction (XRD; Fig. S4), and Raman spectroscopy (Fig. S5).

LIB electrochemical testing is often conducted at a constant current density of 100 mA g^{−129,36–39} (Table S2) versus a lithium metal Li⁺/Li counter electrode, with a polypropylene (e.g. Celgard) separator^9,10,15 (Table S5), and an organic-based electrolyte (e.g. 1 M LiPF₆ in EC/DEC in Table S5).^9,11,40 Hence, shallow (3.00–0.60 V) and deep discharge (<0.60 V) potential ranges were evaluated in CR2032 coin cells with a Celgard separator, a 1 M LiPF₆ in EC/DMC (3 : 7 vol.) electrolyte, and a lithium metal counter, as described in our previous studies^27,31 (Table S4). We introduce the following nomenclature for discharge (D1, D2, D3⋯D500) and charge (C1, C2, C3⋯C500) cycles, respectively.

During the first discharge (D1) the shallow potential range (>0.6 V) shows a single cathodic plateau A^9,10,16 (Fig. 1a and S6a), whereas the deep discharge (<0.6 V) includes an additional plateau B (Fig. 1c and S6b). Therefore, the D1 specific capacity increases (Fig. 1b and S7) as the potential limit increases from 0.80 V (180 mAh g⁻¹) to 0.60 V (186 mAh g⁻¹), 0.40 V (675 mAh g⁻¹), and 0.01 V (698 mAh g⁻¹)¹⁴ (Fig. S8).

During the first charge (C1 3.00 V), the shallow discharge exhibits two plateaus (Y and X in Fig. 1a) that deliver 132 mAh g⁻¹, whereas the deep discharge exhibits a single flat plateau (W in Fig. 1c) that allows for 584 mAh g⁻¹. The associated first cycle efficiency increases from the shallow (0.80 V, 73%; 0.60 V, 71%)⁹ to the deep (0.40 V, 87%; 0.01 V, 84%)¹³ discharge (Fig. S9), due to irreversible trapping of lithium inside the MoS₂ structure during the shallow discharge.^9,10,16,20 Additionally, the first cycle efficiencies of both potential ranges are hindered by the high amount of conductive carbon additive (10 wt%) and inhomogeneous lithiation in the electrodes.²⁷

Shallow discharge cells reduce in capacity by 42 mAh g⁻¹ from D1 to D2 and by 67 mAh g⁻¹ from D5 to D350 (Fig. 1b, S10 and S11a), with multiple cells beginning to fail between D350 and D500 (Fig. S10). Previous studies concluded that the shallow discharge is extremely stable^9,14 as they overlooked the effects of long-term cycling (>100 cycles) and the influence of electrode coating thickness (Fig. S11a). Only thin electrodes (∼12 µm) have stable shallow cycling long-term (∼80 mAh g⁻¹ in D500), whereas intermediate (∼36 µm, ∼39 mAh g⁻¹) and thick (∼87 µm, ∼2 mAh g⁻¹) electrodes suffer from severe capacity loss (Fig. S11a).

On the other hand, the elevated initial capacity (>600 mAh g⁻¹) of the deep discharge is short lived,^9,12,13 reducing drastically by D30 (124 mAh g⁻¹) and D40 (97 mAh g⁻¹) for potential limits of 0.01 V and 0.40 V, respectively (Fig. 1b and S10). Within a short cycling-period, deep discharge cells (78 mAh g⁻¹ in D50) perform worse than shallow discharge cells (125 mAh g⁻¹ in D50), as verified previously.¹⁴ After their extreme initial capacity loss, deep discharge cells have a stable low capacity long-term (75 mAh g⁻¹ in D500) with minimal change from D50 to D500.^13,26

In our previous study, for the same electrode samples we observed a strong lithiation inhomogeneity with increasing electrode thickness (∼12–∼89 µm in Fig. S1) and associated active mass loading (1.11–8.15 mg cm⁻²), which lead to corresponding lower stability long-term²⁷ (Fig. S11). Areal capacity analysis (Fig. S12) showed that irrespective of the electrode coating thickness (mass loading) and depth of discharge, MoS₂ cells cycled at 100 mA g⁻¹ tend towards areal capacities <0.2 mAh cm⁻² in the long-term (D500).

By adjusting the current density from 100 mA g⁻¹ to 0.5 mA cm⁻² to avoid unfavourable skewing of mass-based results towards thin electrodes²⁷ (Fig. S13), we explored cell stability (Fig. S14 and S15) across a wider range of thicknesses (9–204 µm in Note SI I). However, again irrespective of the coating thickness or the depth of discharge (0.80 V vs. 0.01 V) all cell capacities tend to areal capacities below <0.2 mAh cm⁻² after 500 cycles. Therefore, as areal capacity no longer scales with mass loading (Fig. S12a), only a thin region of the electrode is electrochemically active due to either increased interfacial resistance, loss of electronic percolation, electrolyte depletion, surface passivation, or contact loss through cracking and delamination.⁴¹

Shallow discharge mechanism (3.00–0.80 V)

Electrochemical potential profiles, differential capacity voltage numerical data (dQ/dV), cyclic voltammetry (CV), and EIS data from shallow (3.00–0.80 V) cells were analysed to gain insight into the reaction mechanism behind the initial cycling stability and long-term cell failure.

Shallow discharge potential profiles

During the first shallow discharge (D1 0.80 V), cells exhibit a large flat plateau A (1.03 V) associated with the formation of Li_xMoS₂ (ref. 9, 10, 16, 20 and 25) (Fig. 2a). In the subsequent charge (C1), the potential profiles display two pronounced sloped plateaus Y and X beginning to form^9,11 (Fig. 2b). In the ensuing D2, plateau A becomes steeper (labelled A′) and is joined by two higher potential sloped plateaus C and E. Over the course of the first 10 cycles, the anodic plateaus Y and X stabilize at 2.00 V and 2.55 V (Fig. 2b), respectively, whereas on the cathodic side plateau A fades away and is replaced by plateaus C and E solidifying at 2.49 V and 1.77 V (Fig. 2), respectively.

Fig. 2. Electrochemical galvanostatic data from MoS₂ electrode (∼47 µm) cycled at a constant current density of 100 mA g⁻¹versus a Li metal counter electrode, within a shallow potential range 3.00–0.80 V. Capacity-potential profiles from (a) discharge and (b) charge, and (c) dQ/dV numerical analysis.

The shapes of the shallow charge and discharge potential profiles are highly mirrored over the next 100 cycles, across a range of electrode thicknesses (Fig. 2, S16, and S17), suggesting the redox pairs C/X and E/Y are reversible.⁹ However, beyond D100 thicker electrodes (≥36 µm) begin to lose capacity as both discharge and charge plateaus shorten, causing cells to eventually fail (Fig. 2, S16, and S17).

Shallow discharge dQ/dV analysis

dQ/dV analysis (Fig. 2c, S18, and S19) of the galvanostatic cycling data can further clarify the reaction development and fading with cycling. Initially the shallow cell clearly exhibits a rapid evolution of peak A (1.03 V) into an upshifted A′ peak (∼1.13 V) throughout D2 to D5 that pairs with the anodic peaks Z and X, as the cathodic peaks C and E require several cycles to develop (Fig. 2c). Within the first ten cycles, as the redox couples C/X and E/Y gradually form, peak A′ fades from the dQ/dV profile (Fig. 2c, S18, and S19).

Furthermore, dQ/dV analysis detects the presence of minor cathodic D (1.84 V) and anodic Z (1.93 V)⁹ peaks (Fig. 2c), which could not be noticed in the potential profiles (Fig. 2a and b). The D and Z peaks reveal a stepped structure in the larger plateaus E and Y, suggesting a multi-step lithium intercalation process in MoS₂.⁹ Hence, the complete shallow discharge mechanism redox pairs appear to be C/X, D/Z, and E/Y.

Between D100 and D500, the shallow discharge thicker electrodes (≥36 µm in Fig. S19) experience a loss of stability of all redox couples resulting in a decreased capacity and eventually cell failure (Fig. S11a). Four interconnected electrochemical traits denote loss of electrochemical stability, including plateau shortening, plateau potential shift, dQ/dV peak reduction, and dQ/dV redox peak pair separation (Fig. 2). The latter phenomenon indicates an increase in polarization with cycling, ultimately leading to cell failure.

On the other hand, thin electrodes (∼12 µm) retain sharp constant potential dQ/dV peaks long-term without experiencing any peak separation due to increased polarization (Fig. S19a–c). Therefore, thin electrodes can be studied further with ex situ or in situ material characterisation techniques to understand how shallow discharge degradation can be mitigated by electrode engineering.

Shallow discharge cyclic voltammetry

In situ cyclic voltammetry (CV; 0.1 mV s⁻¹) is employed in shallow discharge cells (3.00–0.80 V) to validate the constant current observations, through detailed CV sampling on key cycles. In all other cycles, regular constant current (CC) cycling (100 mA g⁻¹) is utilized to speed up cell evaluation (see Note SI II).

The shallow discharge is activated by peak A (0.82 V) in D1 (Fig. 3a), after which the C1 anodic peaks Z (1.94 V), Y (2.03 V), and X (2.53 V) pair against the only cathodic A′ peak (1.10 V) present within the initial cycles (<3 cycles). As peak A′ fades away between D2 and D10, the cathodic peaks E (1.77 V), D (1.85 V), and C (2.49 V) develop (Fig. 3a) to form the complete C/X, D/Z, and E/Y redox pairs (Table S6). Subsequently, the redox pairs remain stable until D90 with the cathodic (1.77 V, 1.86 V, and 2.50 V) and anodic peaks (1.95 V, 2.00 V and 2.55 V) retaining their positions.

Fig. 3. Cyclic voltammetry data from MoS₂ electrodes versus a Li metal counter electrode, within a shallow potential range 3.00–0.80 V at a scan rate of 0.1 mV s⁻¹, for a coating thickness of (a) ∼47 µm and (b) ∼36 µm. Select cycle ranges displayed.

During the initial cycles (D1–C5), the large separation between peaks A′ and Y–X (ΔV > 0.90 V) suggests the system is highly polarizing until the MoS₂ electrode structurally alters^16,25 to allow reversible lithiation through C/X, D/Z, and E/Y (Fig. 3a). This coincides with the noticeable capacity drop and poor first cycle coulombic efficiency (72%) of the shallow discharge (Fig. 1b), indicating that improving the overall electrode conductivity could highly benefit the shallow mechanism.⁴¹

Tracing the critical shallow discharge capacity loss between D50 and D300 (Fig. 3b), all redox pairs reduce in intensity, broaden, and shift away from their respective redox peaks relative to D50. Therefore, clearly corroborating the previous constant current and dQ/dV analysis observations of plateau shortening and peak separation. Noticeably, by reducing the shallow discharge potential range (3.00–1.10 V) it is observed that the shallow discharge redox pairs can be induced even without reaching the potential of plateau A (<1.00 V). Even a cutoff of 1.10 V (Fig. S20a) is sufficient to trigger the gradual growth of the C/X, D/Z, and E/Y redox pairs. Electrochemical activation is only avoided by restricting the potential to 1.30 V (Fig. S20b–d).

Shallow discharge electrochemical impedance spectroscopy

To unravel the origin of shallow discharge cell (3.00–0.80 V) failure, a cell was cycled under a CC density of 100 mA g⁻¹ with periodic EIS (10 mV; 10 mHz–10 kHz) scans after every discharge and charge. The cell displays the development of the standard C/X, D/Z, and E/Y redox pairs (Fig. 4a) and a typical stable discharge capacity (Fig. S21) of 84 mAh g⁻¹ in D200.

Fig. 4. Electrochemical galvanostatic data from MoS₂ electrodes (∼42 µm) cycled at a constant current density of 100 mA g⁻¹versus a Li metal counter electrode, within a shallow potential range 3.00–0.80 V. Select cycle range displayed including the 1st, 2nd, 5th, 10th, 25th, 50th, 100th, 200th, and 300th cycles. (a) dQ/dV numerical analysis, (b–e) EIS data from charge and discharge (10 mV; 10 mHz–10 kHz), and (f) resistance contribution change with cycle number.

Two equivalent circuits (Fig. S22) from Deng et al.¹⁴ were employed to fit the two types of EIS spectra acquired. The former includes a semi-circle and a straight line (Fig. S22a), whilst the latter contains two semi-circles and a straight line (Fig. S22b). In each case, the hypothetical high frequency x-axis intercept represents the contact resistance (R_c) from the coin cell casing, current collector, and electrode materials, the first semi-circle depicts the charge transfer resistance (R_ct) originating from the electrochemical reaction(s), the second semi-circle if present describes the impedance from an electrode surface film (R_sf), and the straight line is the Warburg resistance (W₁) from ion diffusion.¹⁴

Upon assembly (OCP 3.39 V in Fig. 4b; Note SI III & Fig. S23) the shallow cell possesses a significant resistance originating almost entirely from R_ct (84 Ω), due to bulk MoS₂ being a semiconductor. R_ct increases after D1 0.80 V (114 Ω) and then decreases after C1 3.00 V (42 Ω in Fig. 4b). In the ensuing cycles D2–C100 (Fig. 4c–e and S24), the shallow discharge continues to swing between a low resistance after charge (R_ct ∼7 Ω) and a high resistance after discharge, indicating that it is a critical feature of the shallow potential window.

The initial high discharge resistance (R_ct) reduces across D1–D5 (Fig. 4f) coinciding with the slow development of the of the dQ/dV cathodic peaks (E, D, & C) while peak A′ slowly fades (Fig. 4a). Similarly, the subsequent discharge R_ct continuously decreases across D5–D100 (Fig. 4f) matching the period where the cathodic dQ/dV peaks remain stable (Fig. 4a). On the other hand, the charge R_ct is only noticeable in C1 (Fig. 4f), rapidly decreases by C2 and remains negligible until C100, without a clear correlation to dQ/dV. However, from D100 onwards both the charge and discharge R_ct and R_c resistances increase drastically (Fig. 4f), while the dQ/dV pairs reduce and separate from their redox pairs (Fig. 4a).

At the same time, during D1 a surface film develops as evident by the second semi-circle in the EIS scan (Fig. 4b; R_st 10 Ω), corroborating the previous results by Deng et al.¹⁴ However, upon charge the additional semi-circle disappears, indicating that the surface film is unstable. The deposition and removal of the surface film occurs continuously (Fig. 4b–e and S24), demonstrated by the low D1–D2 surface film resistance (R_sf) increasing significantly by D5 (Fig. 4f), peaking in D10, and maintaining a persistent contribution until D200 (R_sf ∼25 Ω), without a surface film presence on charge. Hence, a continuous unstable surface reaction is occurring, likely involving the deposition and removal of an unstable SEI on every cycle due to the EC/DMC electrolyte.³³

Overall, the high R_ct resistances in D1/C1 & D2–D5 and the sluggish cathodic dQ/dV development indicates that MoS₂ requires lithium induced structural altering¹⁶ before being able to support the shallow discharge. Whereas the persistent R_sf resistance on discharge, lack thereof on charge, and drastic increase in contact resistance in C200 (Fig. 4f), suggests that the long-term resistance build-up (Fig. S25) and cell failure originate from continuous electrolyte decomposition¹³ until depletion.

Although, in situ EIS was only carried out on a single electrode of ∼42 µm thickness, a clear correlation is observed between increasing EIS resistances and dQ/dV peak fading and separation (Fig. 4). Assessing the larger dQ/dV electrode thickness dataset (Fig. S19), thick electrodes (≥36 µm) experience deterioration in both dQ/dV resistance indicators, whereas thin electrodes (∼12 µm) do not experience significant dQ/dV peak fading or potential shift within 500 cycles (Fig. S19). Thus, from the dQ/dV data higher active mass loading promotes resistance build-up in the shallow discharge potential.

However, as the active material increases, the volume of electrolyte in the cell remains unchanged (100 µL), leading to a larger degree of electrolyte excess for the thin electrodes. In such a case, 500 cycles might simply be insufficient to reach electrolyte depletion in thin electrodes. Alternatively, thin electrode stability could suggest that the thin electrode coating mechanical or conductive properties are superior due to limitations in MoS₂ lab-scale slurry formulation or doctor blade coating when increasing the mass of active material.

Hence, future research on the improvement of the MoS₂ shallow discharge should focus on reducing the first cycle efficiency loss (72%) and the resistance build-up within the cell, to allow stabilization of the dQ/dV redox pairs in the long-term (>500 cycles). Stabilization may be achieved through electrolyte solvent selection, electrolyte additive trial (e.g. FEC),³³ and coating mechanical stability testing.⁴¹ Additionally, detailed ex situ study of thin shallow discharge electrodes could hold the key to long-term cycling stabilization, clarifying whether continuous electrolyte consumption or electrode properties (mechanical or conductive) limit the shallow discharge.

Shallow discharge electrochemical mechanism hypothesis

For the shallow discharge (3.00–0.80 V), the experimental potential profiles (Fig. 2a, b, S16, and S17), dQ/dV (Fig. 2c, S18, and S19), and CV (Fig. 3) all showcase the development of a stable reversible redox pairs C/X, D/Z, and E/Y, which with long-term cycling reduce in intensity and separate from their respective electrochemical pair leading to the cell's untimely failure. This degradation process is further supported by an increase in cell resistance as highlighted by in situ EIS (Fig. 4). Furthermore, the deterioration process is sped up and amplified by increasing the MoS₂ electrode coating thickness (Fig. S12a).

The MoS₂ shallow discharge is activated by peak A (0.82 V) in D1 (Fig. 5), which causes the lithiation of the pristine MoS₂ structure into Li_xMoS₂,⁹ as we have observed in our previous ex situ material characterisation study,²⁷ and is well documented in the literature with ex situ^9,16,25 and operando^9,20 techniques. Hence, MoS₂ retains its overall structure upon lithiation, although it is slightly altered and loses crystallinity with lithium insertion.²⁷

Fig. 5. Schematic of shallow discharge MoS₂ reaction mechanism based on CV scans. Alt. MoS₂ denotes altered MoS₂, which is formed after irreversible electrochemical amorphization.

During charge (C1+ in Fig. 5), the anodic peaks Z, Y, and X (1.95 V, 2.02 V, and 2.55 V) enable de-lithiation. In the initial cycles (D2–D5), the anodic peaks pair with the upshifted A′ cathodic lithiation peak (1.10 V); however, in the long-term the cathodic peaks E, D, and C (1.77 V, 1.85 V, and 2.49 V) complete the reversible redox couples. The C/X, D/Z, and E/Y shallow lithiation/de-lithiation process is stable and continues to operate for several hundred cycles (Fig. 5).

The leading shallow discharge mechanism^9,16 (Fig. 5), considers the reversible intercalation and de-intercalation of structurally altered MoS₂. Intercalation of pristine MoS₂ occurs through peak A in D1 to produce Li_xMoS₂, which is subsequently de-intercalated via peaks Z, Y, and X. However, pristine MoS₂ is never fully recovered on charge. Instead, the material crystalline size remains reduced¹⁶ leaving the material with an amorphous character, and some lithium remains permanently trapped within the structure.^9,10,16,20 The recovered altered MoS₂ on charge is available for intercalation via peaks A′ or C, D, and E to form Li_xMoS₂ in the following discharges. The lithiation/de-intercalation shallow discharge mechanism can be represented in two manners (Fig. 5), either considering the reactions E/Y, D/Z, and C/X as sequential or parallel. However, without further data we cannot state which reaction scheme is conclusively taking place.

The only alternative literature proposal by Stephenson et al. considers the shallow discharge as partial conversion of leftover MoS₂ and reversible Li–S²⁸ operation, due to the proximity⁴² of peaks C (2.49 V), E (1.77 V), and X (2.55 V) to the Li–S peaks K (2.3 V), L (1.9 V), and U (2.3 V)–T (2.5 V),⁴³ respectively (Fig. S26). In this assignment (Fig. S27), a minor presence of peak B in the lower potential region of peak A is considered the source of Li₂S for peak X. However, the cathodic peaks D (1.85 V) and E (1.77 V) are too low in potential to fall within the range of Li–S operation (1.9–2.5 V), the combined capacity of peaks E + D + C far exceeds the capacity of peak X, and there is no justification for the presence of peaks Z and Y (MoS₂ de-intercalation) in D50 after the disappearance of peak A′ (MoS₂ intercalation) in the initial cycles.

Even considering the redox couples D/Z and E/Y as multi-step MoS₂ intercalation/de-intercalation and only assigning C/X as a Li–S redox couple (Fig. S27), fails to withstand long-term CV data. During the initial CV cycles (Fig. 5), whilst peak X forms immediately its cathodic pair C takes several cycles to appear and during this initial period the shallow discharge operational redox pair is A′/(Z + Y + X). Hence, it is highly improbable that the single cathodic peak A, responsible for Li_xMoS₂ formation as confirmed by XRD,^9,16,27 is also responsible for sulfur conversion until D10 at such a low potential. Additionally, by restricting the shallow discharge to a potential range of 3.0–1.1 V, the redox pairs C/X, D/Z, and E/Y still form despite the omission of peak A (Fig. S20a). In this instance, a minor triggering of peak B can be considered implausible, as peak A itself is not present. In such a case, there is no viable Li₂S/sulfur source for the development of an Li–S reaction scheme.

Therefore, we refute the possibility that the shallow discharge is governed by Li–S operation. Instead, the shallow discharge potential range (3.00–0.80 V) operates as reversible intercalation and de-intercalation of altered MoS₂ (Fig. 5), continuously consuming electrolyte to form an unstable SEI layer on discharge that strips on charge, thus depleting the cell electrolyte in the long-term.

Deep discharge (3.00–0.01 V)

Potential profiles, dQ/dV numerical data, cyclic voltammetry (CV), and EIS data from deep discharge (3.00–0.01 V) MoS₂ cells with a lithium metal counter (100 mA g⁻¹) were investigated utilising a range of standard and restrictive potential conditions^13,14 to decouple the intertwined reactions producing short-lived high capacities.

Deep discharge potential profiles

In contrast to the shallow discharge, the deep discharge range experiences the inclusion of plateau B (0.51 V) after A (0.98 V) in D1 (Fig. 6a), ascribed as amorphous Li₂S and Mo metal nanocluster formation in the literature,^{9–11,15,17–20,26,29,30} which completely alters the subsequent charge and discharge potential profiles (Fig. 6a, b, and S28). The first charge (C1) is governed by a large single plateau W (2.21 V in Fig. 6b), whilst the subsequent discharges display three sloped plateaus G (1.96 V), H (1.14 V), and I (0.35 V)^9,11,13 (Fig. 6a).

Fig. 6. Electrochemical galvanostatic data from MoS₂ electrodes (∼47 µm) cycled at a constant current density of 100 mA g⁻¹versus a Li metal counter electrode, within a deep potential range 3.00–0.01 V. Capacity-potential profiles from (a) discharge and (b) charge, and (c) dQ/dV numerical analysis.

Within C1–C10 the charge (W) and discharge (G, H, and I) plateaus are present (Fig. 6a and b), but by D20 they are significantly faded corresponding to a sharp decrease in cell capacity (300 mAh g⁻¹ in Fig. 6b). Although, the plateau shapes remain present in thinner electrodes for longer (∼12 µm in Fig. S29a–c), irrespective of coating thickness by D40 no trace of the deep discharge electrochemical plateaus (W, G, H, & I) can be observed (Fig. S29). Therefore, the large initial capacities (>670 mAh g⁻¹) of deep discharge MoS₂ (Fig. 1b) are due to the electrochemical plateaus (W, G, H, & I), whereas the long-term low capacity is due to surface capacitance.

Deep discharge dQ/dV & cyclic voltammetry analysis

Similar to the potential profiles, the dQ/dV profiles show the rapid appearance (C1/D2) and subsequent disappearance of the W, G, H, and I peaks before D20 and D40 for thick (≥36 µm) and thin electrodes (∼12 µm), respectively (Fig. 6c, S30, and S31). Additionally, two minor peaks are identified through dQ/dV that could not be seen in the potential profiles. Firstly, the cathodic F peak (2.13 V), a high potential shoulder to peak G. Secondly, the anodic V peak (1.69 V) at a lower potential than peak W (Fig. 6c).

Overall, the deep discharge dQ/dV peak pattern is highly asymmetric (Fig. 6b, S30, and S31) as validated further via in situ CV-CC measurements (Fig. 7a). Within CV scans, the peaks V (1.76 V), W (2.39 V), F (2.10 V), G (1.84 V), H (1.09 V), and I (0.29 V) form immediately following the D1 presence of peak B (0.37 V) (Fig. 7a).

Fig. 7. Cyclic voltammetry data from MoS₂ electrodes versus a Li metal counter electrode, within a deep potential range 3.00–0.01 V at a scan rate of 0.1 mV s⁻¹, for a coating thickness of (a) ∼47 µm and (b) ∼36 µm. Select cycle ranges displayed.

Within both the dQ/dV profiles (Fig. 6c, S30, and S31) and in situ CV (Fig. 7), after a short period of possible peak intensity increase (<D10) all peaks begin to decrease in intensity. This coincides with the beginning of the strong capacity fade of MoS₂ deep discharge cells (Fig. 1b). At this stage, the cathodic peaks generally downshift in potential, whilst the single anodic peak W upshifts in potential (2.69 V in C10; Fig. 7b). The separation between the cathodic and anodic peaks suggests an increase in polarization within the cells.

The dQ/dV and CV peaks continue to recede in intensity across D10–D30 (Fig. 7b) resulting in a capacitance-based storage mechanism developing in the long-term²⁶ (>D30 in Fig. S32). This trend holds true across all electrode thicknesses (9–97 µm in Fig. S31), which all tend towards similar areal capacities (≤0.2 mAh cm⁻² in Fig. S12b) suggesting that the long-term deep discharge mechanism is limited by reactions taking place only in a thin segment of the electrode.

Deep discharge electrochemical impedance spectroscopy

To elucidate the internal resistance development during deep discharge operation (3.00–0.01 V), an in situ EIS cell was cycled at a constant current density of 100 mA g⁻¹ with periodic EIS (10 mV; 10 mHz–10 kHz) scans at the beginning and end of every discharge and charge. The cell displays the formation of the standard W, G, H, and I peaks within the initial cycles, long-term peak loss (D25+) in favour of capacitance storage (Fig. 8a), and the typical long-term deep discharge capacities (Fig. S21).

Fig. 8. Electrochemical galvanostatic data from MoS₂ electrodes (∼42 µm) cycled at a constant current density of 100 mA g⁻¹versus a Li metal counter electrode, within a deep potential range 3.00–0.01 V. Select cycle range displayed including the 1st, 2nd, 5th, 10th, 25th, 50th, 100th, and 200th cycles. (a) dQ/dV numerical analysis, (b–e) EIS data from charge and discharge (10 mV; 10 mHz–10 kHz), and (f) resistance contribution change with cycle number.

Upon assembly (OCP 3.38 V), the deep discharge EIS cell possesses elevated internal resistance R_ct (66 Ω), which reduces after discharge to R_ct (22 Ω in Fig. 8b). A reduction in cell resistance after lithiation (D1 0.01 V) is most unexpected, as the more electrochemically stable shallow discharge experienced an increase in resistance in D1 0.80 V (Fig. 4b) and the expected D1 product Li₂S is more resistive than MoS₂.^44,45 Moreover, in contrast to in situ EIS studies by Deng et al.¹⁴ and the irreversible mechanism expectation of high resistance sulfur formation upon charge, cell resistance reduces further after the first de-lithiation process (C1 3.00 V R_ct 13 Ω in Fig. 9b).

Fig. 9. Electrochemical galvanostatic cycling data from MoS₂ electrodes (∼36 µm) versus a Li metal counter electrode at a constant current density of 100 mA g⁻¹. The first discharge is a deep discharge (3.00–0.01 V) followed by a restricted potential range (1.80–0.01 V). (a–c) dQ/dV numerical data and (d) galvanostatic discharge capacity comparing the restricted potential range (1.80–0.01 V) against standard (3.00–0.01 V) deep discharge cells.

Additionally, we previously observed the formation of a gel-like layer with ex situ SEM in D1 0.01 V,²⁷ which is validated with EIS (R_sf 33 Ω) with the appearance of a second semi-circle in the spectra (Fig. 8b). However, considering the drastic morphological surface change observed with SEM,²⁷ the surface film resistance is not significantly higher than the shallow discharge film resistance (R_sf 10 Ω), where the SEI is much thinner and not visible with SEM.²⁷ Furthermore, corroborating our previous ex situ results that showed the gel-like layer vanishing after charge,²⁷ the deep discharge cell EIS scan only possesses a single semi-circle at 3.00 V (Fig. 8b), indicating an absence of a surface film.

Nevertheless, like the shallow discharge the deep discharge exhibits a resistance swing between charge (3.00 V) and discharge (0.01 V), with the charged state being less resistive (Fig. 8b and c). However, opposite to the shallow discharge the deep discharge resistance swing inverts after D5 (Fig. 8d), more clearly seen in D10+ (Fig. 8e and S33). This indicates that although the charge product from peak W is not initially resistive, with cycling the product(s) from peak W becomes increasingly more resistive inhibiting cell operation.

Despite the initial resistance decrease within D1–D2, the increase in charge resistance across C5–C100 drives a response increase in the discharge resistance (Fig. 8f and S34). In particular, the charge R_ct is the main factor contributing to resistance build-up within the deep discharge cell. From C5 R_ct starts rapidly increasing, while the discharge R_ct is stable and R_sf remains negligible (Fig. 8f). Between D5 and C10, both charge and discharge R_ct contribute to increasing cell resistance, with dQ/dV peaks W and I being predominantly affected and subsequently causing deviation in cell performance (Fig. 8a).

By D25 the deep discharge cell reduces in capacity (Fig. S21), loses all electrochemical peaks (Fig. 8a), and exhibits only capacitive storage, yet its EIS signal continues to swing between a single semi-circle charge state and a two semi-circle discharge state (Fig. S33a). The resistance spectra swing is associated with the reaction pathway from the initial cycles still taking place. However, by D50 the discharge state is permanently altered, displaying a single semi-circle like the charged state. Therefore, the electrochemical process (G, H, and I) that creates the surface film is completely absent beyond D50. Ultimately, the resistance build-up in the deep discharge cell (Fig. 8f) closely matches the capacity reduction of the cell (Fig. S21), strongly suggesting that the source of the rapid cell decay is an unstable electrochemical reaction evolution.

Deep discharge restrictive potential ranges

To establish the connection between the major deep discharge peaks (W, G, H, and I), restrictive potentials of 1.80–0.01 V, 3.00–1.50 V, and 3.00–0.80 V^13,14 were employed to suppress the cell potential range after an initial deep discharge (D1 0.01 V). The initial deep discharge serves to activate the cell through peaks A and B, switching the operating mode from the shallow redox pairs to the transient mechanism (W, G, H, and I).^9,13,14

Limiting the charge potential to 1.80 V from the first charge (C1+) onwards results in the absence of any dQ/dV peaks (Fig. 9a–c) and a long-term stable capacitance-based capacity similar to the standard deep discharge (3.00–0.01 V) performance beyond D40 (Fig. 9d). Therefore, the presence of peak W during charge is required for MoS₂ LIB operation (G, H, and I). We note that in our previous ex situ study,²⁷ identical MoS₂ coin cells experienced highly inhomogeneous lithiation.

Since we have retained the same electrode laminates, cell configuration, electrolyte, and cycling conditions, the presence of unlithiated MoS₂ in the electrode centre is very likely. However, any underused active material present in the centre is electrochemically inaccessible in D2, as no further peaks A and/or B appear when a restricted potential range is applied (1.80–0.01 V in Fig. 9a–c). Similarly, if peak I represented leftover pristine MoS₂ lithiation in D2+ (Fig. 7a), then peak I should have occurred despite the absence of peak W in the reduced 1.80–0.01 V potential window (Fig. 9a).

On the other hand, restricting the lower potential range (3.00–1.50 V) to eliminate peaks H and I from the second discharge (D2+) onwards, results in a stable low capacity (Fig. 10a). In the absence of peaks H and I, a reversible redox pair forms between peaks G/W. This is the first instance where peak W is symmetric with a single cathodic peak (G), providing a stable low capacity long-term (∼82 mAh g⁻¹ in D200). Nevertheless, capacity reduces with cycling (Fig. 10a) as both peaks fade and performance in D250 is worse than in the standard shallow discharge (3.00–0.80 V).

Fig. 10. Electrochemical galvanostatic data from MoS₂ electrodes (∼47 µm) versus a Li metal counter electrode at a constant current density of 100 mA g⁻¹. The first discharge is a deep discharge (3.00–0.01 V) followed by a restricted potential range (3.00–1.50 V or 3.00–0.80 V). (a) Galvanostatic discharge capacity plots. Numerical dQ/dV analysis for a restricted potential range from the second cycle onwards for (b) 3.00–1.50 V and (c) 3.00–0.80 V.

Including the cathodic peak H (Fig. 10c), by adjusting the lower potential limit (3.00–0.80 V) after the initial activation deep discharge (D1 0.01 V), results in the growth of the anodic peak W and a corresponding increase in the initial capacities (Fig. 10a). Hence, peaks (G + H)/W form a reversible redox couple (Fig. 10c). However, the capacity boost gained by lowering the potential limit to include H is short lived as the expansion of the potential window results in a faster capacity fade (100 cycles in Fig. 10a) relative to the 3.00–1.50 V (G/W) restricted range.

Furthermore, by comparing dQ/dV data from the standard deep discharge (3.00–0.01 V) and the restricted potentials (C1+ 1.80–0.01 V, 3.00–1.5 V, and 3.00–0.80 V in Fig. 11), we identify that the addition of each subsequent cathodic peak (G, H, and I) increases the intensity of the single anodic peak W without affecting its potential (Fig. 11). At the same time, inclusion of each subsequent cathodic peak does not affect the intensity and potential of peaks G and H (Fig. 11). Hence, each potential range studied (3.00–1.50 V, 3.00–0.80 V, & 3.00–0.01 V) forms reversible redox pairs with the single anodic peak W (Fig. 11), expressed as G/W, (G + H)/W, and (G + H + I)/W, respectively. Whereby, inclusion of each subsequent cathodic peak causes faster cell decay (Fig. 10a and S35). Moreover, the minor anodic peak V only appears after the presence of peak I (Fig. 11).

Fig. 11. dQ/dV analysis of electrochemical galvanostatic data from MoS₂ electrodes (∼36–pink or ∼47 µm – blue, red, and green) versus a Li metal counter electrode cycled at a constant current density of 100 mA g⁻¹. The first discharge is a deep discharge (3.00–0.01 V) followed by a restricted potential range (3.00–1.50 V, 3.00–0.80 V, or 1.80–0.01 V) from the first charge onwards (C1+). Second cycle displayed.

The information gathered from the restrictive potentials (see Note SI IV) electrochemically excludes the possibility that reactions G, H, and I are directly consecutive or parallel-series (Fig. S36), due to the consistent presence of peak W despite exclusion of peaks H and/or I and the reduction in peak W upon the removal of peaks H and/or I (Fig. 11). Additionally, restrictive potentials (1.80–0.01 V) prove that peaks H and I are not lithiation and conversion (Li₂S and Mo formation) of leftover pristine MoS₂, respectively, due to their appearance depending on the presence of peaks W and/or G (Fig. 9). Hence, the cathodic peaks G, H, and I must be forming the same product or product class, so that their product(s) can react synchronously during the single anodic peak W. Hence, we propose that the cathodic peaks of MoS₂ are not series reactions or parallel-series but form a parallel or series-independent system (Fig. S36).

Energy analysis of the deep discharge first cycle CV redox couples (Fig. S37a), due to the greater stability of CV data relative to dQ/dV data in the first cycle, indicates that redox reactions only account for 48% of capacity on discharge and 65% of capacity on charge. Capacitive storage accounts for a significant portion of the first cycle, due to the large capacitance tail below peak B and the flat capacitance region throughout the entire charge (Fig. S37a). However, the (A + B)/W redox pair has a symmetry of 88%, indicating a relatively symmetric redox process (Fig. S37a).

In the second cycle (Fig. S37b–d), dQ/dV energy analysis indicates that the redox couple G/W is symmetric in capacity (97%), whilst the redox couples (G + H)/W (66%) and (G + H + I)/W (80%) are asymmetric in capacity. The difference is caused by a large portion of capacitive storage coinciding with peaks H and I in the lower potential range (<1.2 V in Fig. S37c and d). Thus, a significant fraction of capacity on charge (C2 ∼351 mAh g⁻¹, 52%) and discharge (D2 ∼451 mAh g⁻¹, 63%) is capacitance-based (Fig. S37b). With redox peaks W, G, H, and I contributing ∼313 mAh g⁻¹, ∼109 mAh g⁻¹, ∼46 mAh g⁻¹, and ∼109 mAh g⁻¹, respectively.

Therefore, despite both C2 and D2 capacities exceeding the 4e⁻ theoretical capacity (670 mAh g⁻¹), their reaction-based energy storage involves less than 2e⁻ per MoS₂ molecule, whilst their capacitive storage accounts for the remaining approximate 2e⁻. Hence, capacitive storage has a major contribution in the MoS₂ mechanism. However, the overall charge to discharge capacity ratio (C_i/D_i) for the three redox couples G/W, (G + H)/W, and (G + H + I)/W is above 0.95 in D2 (Fig. S38), meaning that the combined capacitance and redox processes are symmetric (see Note SI V).

Shallow & deep discharge contrast

The shallow and deep discharge of MoS₂ have drastically different capacity performances (Fig. 1b), potential profiles (Fig. 2 and 6), and CV peaks (Fig. 3 and 7). Whereby, the shallow discharge has stable reversible redox couples, whilst the deep discharge is highly unstable losing redox behaviour within 40 cycles. Nevertheless, close dQ/dV inspection (Fig. S39) shows a significant resemblance in D2 (Fig. S39b) between the shallow discharge cathodic activation process (A′ + E) and the deep discharge G and H peaks.

Despite differences in potential and intensity, the similarity between the cathodic peaks A′ & H and E + D & G + F in D2 is significant (Fig. S39b). However, by D5 the similarity between the two potential ranges disappears (Fig. S39c). Furthermore, the anodic C2 process is completely different between the two potential ranges (Fig. S39a–c). After the inclusion of peak B in the deep discharge, there is no resemblance between the shallow anodic peak (Z) and the anodic deep discharge peaks (V + W). The minor deep discharge peak V, that has previously been claimed as MoS₂ de-lithiation,²⁶ resides at a lower potential than peak Z (Fig. S39b). Inversely, peak W sits at a higher potential than peak Z. Moreover, even excluding peak I with a restrictive potential range (3.00–0.80 V in Fig. S39d–f), the anodic process remains starkly different between the shallow and deep discharges.

The resemblance between the shallow and deep discharge cathodic peaks only occurs during the shallow discharge's activation process, which lasts approximately five cycles, terminating with the disappearance of peak A′ (Fig. S39c and f). However, the deep discharge peak H that resembles A′ remains clearly present until all deep discharge peaks fade around the 30th cycle (Fig. S30). Therefore, irrespective of the similarity between deep discharge peaks G & H and shallow discharge peaks E & A′, deep lithiation via peak B in D1 completely alters the anodic process of MoS₂ LIBs. Thereafter, even restrictive potential ranges (3.00–0.80 V) cannot return cell operation to anything resembling the shallow discharge, indicating that a permanent alteration of the MoS₂ system has taken place.

Deep discharge electrochemical mechanism hypothesis

Given the electrochemical restrictive potential isolation of the redox pairs G/W, (G + H)/W, and (G + H + I)/W, we reviewed all proposed MoS₂ lithiation mechanisms in the literature and have represented them to the best of our understanding visually via electrochemical reaction schemes (Fig. 12–14).

Fig. 12. CV data from a deep discharge (3.00–0.01 V) cell (∼36 µm) scanned at 0.1 mV s⁻¹versus a lithium metal counter. Reaction flow schematics of literature proposed MoS₂ lithiation mechanisms REV2 and REV3 considering reversible MoS₂ reforming.^{10,17–19,26}.

Fig. 13. Reaction flow schematics of literature proposed MoS₂ lithiation mechanisms IRR1–IRR3 considering irreversible formation of Li₂S & Mo and subsequent Li₂S/S operation.^{9,13,15,28,32}.

Fig. 14. Reaction flow schematics of literature proposed MoS₂ lithiation mechanisms considering a (MIX) mixed reversible MoS₂ reformation and Li₂S/S operation.^11,22.

We purposefully omit focusing on the long-term mechanism of MoS₂ lithiation beyond D20–D40, as by that point all redox pairs disappear from the CV (Fig. 7) and dQ/dV profiles (Fig. 6c), leaving behind a capacitive storage mechanism that is distinctly different from the initial 20 cycles. Based on our previous ex situ study²⁷ and EIS data (Fig. 8) we believe that long-term capacitance is dominated by the surface morphology mechanism proposed by Liu et al.,²⁶ after the development of the surface gel-like layer shown by Zhu et al.³³ The surface morphology mechanism appears to favour thin electrodes (Fig. S14b) where the surface is more highly connected to the current collector.

Furthermore, although we prefer to focus on literature mechanism proposals that rely on the use of commercial bulk 2H MoS₂ electrodes^9–11,15,16 that are more comparable across laboratory settings, we have considered lab synthesized nanostructure derived lithiation mechanisms as well, despite their wide range of MoS₂ morphologies, carbon composites, and cycling conditions.^{17–19,22,26,46}

Overall, three lithiation mechanism categories exist in the literature including the reversible (REV), irreversible (IRR), and mixed (MIX) mechanism schemes, which consider the (i) reversible conversion and reconstruction of nanosized MoS₂ and Li from amorphous Li₂S and Mo metallic nanograins (eqn (3) and Fig. 12), (ii) the irreversible formation of Mo metallic nanograins on D1 and the subsequent Li–S operation (Li₂S/S) of the cell (eqn (4) and Fig. 13), and (iii) the mixed occurrence of both the reversible and irreversible reactions (eqn (3) and (4), and Fig. 14), respectively.

Reversible (MoS₂) mechanisms

Reversible MoS₂ lithiation mechanism hypotheses consider that MoS₂ undergoes conversion into amorphous Li₂S and Mo metal upon lithiation and reconstitution into nano-sized MoS₂ during de-lithiation.

The earliest proposed MoS₂ lithiation mechanism from 2011 outlines a reversible mechanism (REV1) where MoS₂ reversibly forms on charge through peak W⁴⁷ and decomposes on discharge via an unspecified CV peak.^46,47 The authors establish that the distinction between the first and all following discharge profiles (Fig. 1c) is due to MoS₂ conversion into Mo metal and Li₂S and the formation of a gel-like SEI layer in potentials below peak B.^46,47 However, in the following cycles reactions are not ascribed directly to peaks G, H, and I,^46,47 instead the D2 lithiation capacity is described as a combination of MoS₂ conversion into Li₂S, Mo nanoparticle redox action, and reversible SEI growth. Thus, mechanism REV1 cannot be disproved or validated by electrochemical peaks, as there are no proposed reactions to contest or verify via the establishing of G/W, (G + H)/W, and (G + H + I)/W as complete redox peaks (Fig. 11).

On the other hand, the leading literature reversible lithiation mechanism (REV2)^10,17–19 considers the reversible reconstitution of altered nano-domains of MoS₂ from amorphous Li₂S and metallic Mo on charge (C1+) through peak W (Fig. 12). The nano-domains are XRD amorphous, complicating their isolation outside of microscopy techniques^17–19 and beamline X-ray absorption spectroscopy¹⁰ (XAS). In the following discharge (D2), the REV2 nano-sized MoS₂ is lithiated through peak G and decomposes into Li₂S and Mo via reaction I (Fig. 12). After which, nano MoS₂ is reformed by reaction W on charge. Within this proposed mechanism, the role of H is the lithiation of any pristine MoS₂ still found in the electrode until all the available pristine material is nano-sized.

However, we have shown by limiting the upper potential to 1.80 V after deep discharge in D1 (Fig. 9), that no electrochemically available preserved pristine MoS₂ can be lithiated after peak B. Thus, the reaction assigned to peak H is infeasible. Moreover, REV2 relies on consecutive reactions as both G and H require peak I to convert their lithiated Li_xMoS₂ into Li₂S and Mo for the subsequent charge cycle's anodic peak W. Therefore, REV2 fails to justify the operation of restricted potential ranges, which clearly function through the redox pairs G/W and (G + H)/W in the absence of peak I (Fig. 11).

Alternatively, the last literature reversible mechanism (REV3) considers peak V as the de-intercalation of any unreacted Li_xMoS₂ in the electrode, peak W as MoS₂ reconstitution from Mo and Li₂S, peaks G and H as lithiation of MoS₂ into Li_xMoS₂, and peak I as conversion of Li_xMoS₂ into Li₂S and Mo²⁶ (Fig. 12). The sequential order of the cathodic reactions in mechanism REV3 suggests that by limiting the discharge potential (D2+), it should be possible to select between a complete conversion reaction (I) and solely lithiation (G + H). However, the restrictive potential (3.00–0.80 V) clearly demonstrates that removing peak I does not lead to the development of the shallow discharge peaks (Z, Y, & X) on charge nor the increase of peak V (Fig. 11). Instead, redox couples G/W and (G + H)/W operate reversibly in the absence of peak I, with no change in potential, and peak V completely disappears (Fig. 11).

Hence, no literature reversible mechanism^{10,17–19,26} can currently justify the electrochemical signal derived from cycling commercial MoS₂ electrodes. Albeit we highlight the fact that most reversible mechanism proposals originate from articles working with lab synthesized MoS₂ through production techniques such as hydrothermal,^10,18,19,47 alkali reductive chemistry,¹⁷ or liquid phase exfoliation,²⁶ which might significantly modify the operating mode of MoS₂ when comparing to commercial MoS₂.^{9–11,15,16,20,25,29}

Irreversible (Li–S) mechanisms

Irreversible MoS₂ lithiation mechanism hypotheses consider that the first discharge is unique and irreversibly transforms the mode of operation of the MoS₂ LIB cell. Electrochemical activation of the electrode though peak B causes the formation of amorphous Li₂S and metallic Mo.^{9–11,17–20,26,28–30} Thereafter, the metallic Mo remains unaltered with cycling whereas Li₂S becomes the active material of the cell.^9,13

The first literature irreversible lithiation mechanism (IRR1)²⁸ considers G/W to be the only sulfur chemistry redox couple (Li₂S/S), whilst peak H and I are considered consecutive reactions converting unreacted MoS₂ into Li₂S and Mo until all available pristine MoS₂ is exhausted (Fig. 13). However, we have established electrochemically via restrictive potentials that reactions H and I are non-consecutive (Fig. 11) and that there is no electrochemically accessible pristine MoS₂ available after the onset of peak B (1.80–0.01 V in Fig. 11).

Additionally, hypothesis IRR1 suggests that (G + I)/W is the complete Li₂S redox couple in the range 3.00–0.01 V. In such a case, peak H should provide excess capacity that should lead to a low charge to discharge capacity ratio (C_i/D_i), due to the asymmetry in the reaction scheme (see Note SI V). However, under all potential ranges considered C_i/D_i is highly symmetric (>0.95 in Fig. S38). Therefore, mechanism IRR1 fails to justify the electrochemical cycling data presented (Fig. S38), whilst our proposed complete redox couple (G + H + I)/W has a symmetric C_i/D_i.

An alternative irreversible theory (IRR2) again assigns G/W to be the only sulfur chemistry redox couple (Li₂S/S) in the system and proposes peaks H and I to be storage of Li⁺ on inactive metallic Mo nanograins^13,16,32 or Li_xS_y¹⁶ (Fig. 13). During charge, the immobilised Li⁺ is considered to be released in peak V and during the peak-less region before 2.0 V.^16,32 Hypothesis IRR2 indicates that G/W ought to be symmetric as H and I are not contributing to peak W. However, on the contrary our electrochemical isolation data showcases that reactions H and I both enhance reaction W's response (Fig. 11). Thus, mechanism IRR2 cannot accurately describe the electrochemical response of the MoS₂ system.

Lastly, the most widely supported irreversible hypothesis (IRR3) developed by Zhang et al. considers Li–S operation to function through the redox couple (G + H)/W,⁹ whilst peak I is the conversion of any leftover pristine MoS₂ (Fig. 13). The benefit of IRR3 is that by considering the Li–S process as a two-step process (G + H) it avoids the issues encountered by IRR1 and IRR2 when proposing consecutive reaction schemes. However, it remains unclear why under IRR3 the two-step lithiation (A) and conversion (B) process of pristine MoS₂ becomes a single step-process (I) at a much lower potential in D2+ (Fig. 13). Additionally, the restrictive potential of 1.80 V demonstrated that there is no electrochemically accessible pristine MoS₂ in the electrode beyond the onset of D1 peak B (Fig. 9), highlighting that the reaction ascribed to peak I in IRR3 is doubtful.

Mixed reversible (MoS₂) and Li–S mechanisms

The last category of MoS₂ lithiation mechanism hypotheses merges a mixture of both the reversible (MoS₂) and irreversible mechanism (Li–S) elements. Generally, mixed mechanisms consider that a part of the electrode reversibly decomposes and reforms into MoS₂, whereas the remainder operates Li₂S/S storage.

The first mixed theory (MIX1) considers the combination of REV2 and IRR1 (Fig. 14), whereby two redox couples G/W and (H + I)/W are considered.¹¹ Surprisingly, in such an instance the single anodic peak W acts as the redox couple for two distinct electrochemical processes. The G/W pair is assigned as a single step Li₂S/S reversible reaction, whereas the (H + I)/W couple is ascribed as consecutive MoS₂ lithiation and Li₂S + Mo/MoS₂ conversion. Evidently, the mixed theory encounters the same setback expressed for REV2. Notably, that the reactions H and I are in fact not sequential, as electrochemically (G + H)/W is a viable redox pair under a restricted potential range (Fig. 11), without introducing any de-intercalation peaks for Li_xMoS₂ on the anodic side.

The alternative mixed theory (MIX2) combines the basis of REV2 and IRR3, whereby (G + H)/W is considered Li₂S/S operation and I/V is assigned as MoS₂ conversion/reconstitution²² (Fig. 14). Although, this hypothesis withstands any consecutive reaction criticism, from dQ/dV analysis (Fig. 11) it is visible that the proposed peak couple I/V is highly asymmetric in capacity (see Note SI IV). Energy calculations reveal peak I to contribute approximately 109 mAh g⁻¹ on discharge, whereas peak V only delivers 13 mAh g⁻¹. Thus, making I/V an unlikely reversible pairing. On the other hand, the inclusion of the cathodic peak I contributes to a noticeable increase in the anodic peak W, indicating that the complete couple is (G + H + I)/W (Fig. 11).

Proposal for MoS₂ lithiation mechanism

The utilisation of the restricted potential ranges 1.80–0.01 V, 3.00–1.50 V, and 3.00–0.80 V after an initial activation deep discharge (D1 0.01 V) has provided significant information to guide the elucidation of the MoS₂ lithiation mechanism (Fig. 11). Although we cannot confirm the complete bulk MoS₂ deep discharge lithiation mechanism without further detailed ex situ, in situ, or operando material characterisation to validate reaction products, we are able to outline the mechanism's electrochemical boundaries to justify the data presented within this study. This includes (i) the parallel or series-independent nature of the cathodic reactions (G, H, and I), (ii) the cathodic peaks pairing with the single anodic peak W to form redox couples G/W, (G + H)/W, and (G + H + I)/W, and (iii) the fact that no electrochemically accessible preserved pristine MoS₂ can be detected beyond the presence of peak B in D1.

Based on these boundary conditions, we propose three mechanism flows (Fig. 15) that are able to justify the electrochemical signal from MoS₂ deep discharge LIB cells. We stress the fact that all our proposed mechanisms require further material characterisation or well-designed electrochemical testing (e.g. long-term MoS₂ + S electrode cycling in Fig. S40)¹¹ to validate or conclusively refute. Furthermore, we highlight that our mechanism proposals consider the minor anodic peak V, to be largely negligible in the deep discharge mechanism of bulk MoS₂, due to its limited 13 mAh g⁻¹ contribution and its presence being entirely associated with peak I (Fig. 11). However, we do highlight that in lab-based synthetic MoS₂ or MoS₂ composites, the role of peak V might be more significant and should be considered closely.

Fig. 15. Reaction flow schematics of proposed MoS₂ lithiation mechanisms based on: (OVER1) over-lithiation of the MoS₂ network, (MIX3) mixed MoS₂ reformation and Li₂S/S irreversible operation, and (IRR4) Li₂S/S irreversible operation.

Our first proposal (OVER1 in Fig. 15) assigns each of the cathodic peaks (G + H + I) as a further degree of MoS₂ lithiation (Li_xMoS₂ < Li_yMoS₂ < Li_zMoS₂) beyond the usual Li_xMoS₂ literature stoichiometry (x ≤ 1)¹² and peak W as MoS₂ de-intercalation. We propose an entirely lithiation-based series-independent mechanism due to our recent ex situ depth-profiling XPS identification of lithiated 1T MoS₂ in the electrode bulk after lithiation beyond peak B in D1 0.01 V and its preserved presence after charge past peak W in C1 3.00 V.³¹ The 1T MoS₂ could not be detected otherwise with XRD, suggesting it must be nano-sized and/or amorphous in nature.³¹

In over-lithiation (OVER1), all cathodic reactions (A, B, G, H, & I) are considered lithiation of pristine or altered MoS₂. As the shallow potential window has demonstrated,^16,27 peak A is sufficient to reduce the crystallinity of pristine MoS₂, whereas peak B results in complete amorphization. We consider that each subsequent intercalation reaction (G, H, and I) can induce further nano-sizing and expose fresh material boundaries, allowing for localised MoS₂ conversion to take place (Li₂S + Mo). Conversion is considered irreversible in OVER1, resulting in the gradual depletion of the nano-sized MoS₂ over approximately 20–40 cycles.

Within OVER1 peak G is considered standard lithiation (Li_xMoS₂, x < 1), whilst peaks B, H, and I are all considered lithiation beyond unity (Li_zMoS₂, 1 < z < 2). Therefore, all reactions are producing a single product class, enabling for the reverse operation to function via the single anodic peak W. Reducing the degree of lithiation by restricting the potential range would slow down the degradation process by limiting the amount of MoS₂ conversion taking place, but it cannot prevent material degradation altogether. An upper lithiation (Li_zMoS₂) limit of 2 is selected for OVER1 due to the dQ/dV energy analysis indicating that the redox couples only account for approximately 2e⁻ of charge transfer with the remaining 2e⁻ originating from capacitance. Such a reliance on capacitive storage would justify the significantly better performance of thinner electrodes, where the surface to bulk ratio is smaller.

Verification of OVER1 would require a detailed ex situ, in situ, or operando study of thin MoS₂ electrodes at low current densities, where lithiation inhomogeneities are minimised to avoid partially lithiated species contamination. XAS and/or XPS will be necessary due to the difficulty in monitoring the MoS₂ mechanism with other techniques.²⁷ The objective would be to track the preservation of nano-MoS₂ within the respective G/W, (G + H)/W, and (G + H + I)/W redox pairs.

Alternatively, our second proposal considers a mixed hypothesis with parallel MoS₂ reformation (I/W) and Li–S operation (G + H)/W (MIX3 in Fig. 15). Choi et al.¹¹ previously suggested that two separate reactions cooccur during peak W (MIX1 in Fig. 14) based on XAS findings. Hence, in MIX3 we combine the schematically functional aspects of MIX1 and MIX2. By considering peak W as the sole anodic reaction, we avoid the peak capacity asymmetry encountered by MIX2 with the I/V redox couple. On the other hand, by assigning a parallel scheme for the (G + H)/W redox couple, we avoid having consecutive reactions between peaks H and I like MIX1 that could be interrupted by restricted potential ranges (Fig. 11).

However, despite our MIX3 process flow being robust in terms of its electrochemical reaction scheme, as it justifies operation within the restricted potential ranges, we cannot provide a viable reason for why reformed MoS₂ lithiation and conversion would occur at such a low potential (peak I ∼0.36 V) in a single step, rather than as a two-step process initiating around the potential range of peak A (∼0.80 V). Therefore, our MIX3 mechanism suffers from the same overall criticism as IRR3 and MIX2. Justification of MIX3 would require detailed material characterisation of the restricted potential range 3.00–0.80 V to verify sole Li–S operation in that potential region, whereas the full potential window 3.00–0.01 V should experience the inclusion of activity from Mo as well.

Our final proposal (IRR4) relies on completely irreversible Li–S operation (Fig. 15). Herein, we consider the cathodic reactions G, H, and I as completely parallel (Fig. S36), producing a single product (Li₂S) or class of products (e.g. different length polysulfides; Li_xS_y or molybdenum lithium sulfides; Mo_xS_yLi_z) that are all recovered through the single anodic peak W. This assignment enables all redox pairs G/W, (H + I)/W, and (G + H + I)/W to function with a single anodic Li–S peak W, whilst avoiding criticism due to consecutive reactions (IRR1), direct MoS₂ conversion (IRR3), or Li⁺ storage on Mo metal (IRR2 in Fig. 13).

Nevertheless, mechanism proposal IRR4 suffers from the assignment of Li–S operation to peak I (∼0.3 V) with a potential drastically below the literature Li–S range of operation (1.9–2.5 V). Even peak H at ∼1.1 V is considerably below the Li–S range of operation. However, adding sulfur to an MoS₂ electrode (MoS₂ + S), was shown by Choi et al.¹¹ to have no influence on the potential of the W peak, despite the presence of an additional sulfur peak during D1 (J in Fig. S40). Therefore, IRR4 can be validated by long-term MoS₂ + S electrode cycling in a standard LIB electrolyte (e.g. 1 M LiPF₆ in EC/DEC or EC/DMC). Currently only the first cycle has been explored (Fig. S40) and thus conclusions cannot be drawn regarding the critical second cycle.

MoS₂ lithiation mechanism outlook

Despite numerous studies on the improvement of MoS₂ performance to date and over a decade of MoS₂ lithiation mechanism investigation, the combination of material amorphization & lithiation inhomogeneity²⁷ impacting material characterisation and the type of MoS₂ used (nano vs. bulk) influencing the electrochemical signal have led to multiple proposed mechanisms. Given that even the better understood shallow discharge of MoS₂ is unstable in the long-term (500 cycles), what outlook does MoS₂ hold for energy storage?

Firstly, as identified with in situ EIS both the shallow and deep discharge suffer from resistance build-up. The former occurs from continuous electrolyte consumption, while the latter happens due to an unstable reaction. Both sources could potentially be resolved via electrolyte selection, electrolyte additive trials,³³ electrode mechanical improvement,⁴¹ or replacement of the lithium metal counter electrode.⁴⁸ For both depths of discharge, in situ EIS should be used as the standard technique for improvement evaluation.

Secondly, although ex situ, in situ, and operando material characterisation of MoS₂ LIB has been challenging, knowledge of inhomogeneous lithiation in thicker electrodes and at high current densities²⁷ will aid design more consistent experimental studies. The use of a standard bulk MoS₂ material should also be adopted to make cross-lab research comparable. Lab synthesized nano MoS₂ or MoS₂ composites can be studied as a separate system. Although the first cycle is crucial in MoS₂, the second cycle represents the long-term mechanism that critically requires understanding. Hopefully, with the introduction of reaction schemes competing mechanism proposals can be easily compared.

Additionally, knowledge gained from the restrictive electrochemical conditions will enable future materials characterisation studies to select regions of interest more concisely. In the case of our proposed mechanisms this includes (i) detailed tracking of the species interacting in G/W, (G + H)/W, and (G + H + I)/W with XAS or XPS for OVER1, (ii) isolating solely Li–S operation in the restricted potential range 3.00–0.80 V with XAS for MIX3, (iii) and long-term operation of a MoS₂ + S electrode for IRR4.

Thirdly, even if MoS₂ technological development continues progressing slowly and cannot replace LIB's graphite anode in time for meaningful commercial impact, thanks to its multifunctional ability to store sodium ions as well, MoS₂ can be applied in upcoming sodium ion designs.²⁵ Thus, potentially offering an alternative to the sodium ion hard carbon anode, which suffers from several limitations.⁴⁹ Therefore, a promising development could involve redirecting knowledge gained from studying the lithiation mechanism towards understanding the sodiation mechanism of MoS₂.

Finally, MoS₂ is the most widely studied TMD; however, it is not the TMD with the highest theoretical capacity (TiS₂ ∼957 mAh g⁻¹),⁵⁰ the most environmentally friendly & abundant TMD (FeS₂),⁵¹ nor the cheapest TMD (FeS₂). Experimental knowledge and material understanding gained from MoS₂ studies can directly be transferred to similar materials such as WS₂, TiS₂, and FeS₂ or toward other amorphous battery materials, where conventional material characterisation techniques struggle.

Conclusion

Overall, we have validated the long-term capacitive storage of deep discharge MoS₂ (>40 cycles).²⁶ Whereby, over 500 cycles all electrode thicknesses (0.9–19.5 mg cm⁻²) tend towards a similar areal capacity (<0.2 mAh cm⁻²), as only a thin region of the electrode remains electrochemically active.

Furthermore, we applied dQ/dV peak tracking and in situ EIS to the shallow potential range (3.00–0.80 V), identifying the reason for long-term cycling failure as the build-up of resistance within the cell from continuous electrolyte consumption. Additionally, since the shallow discharge first cycle efficiency (72%) is so low, we believe that improving the electrode configuration and conductivity through process optimization could prove key in stabilizing the shallow MoS₂ discharge for low energy density applications.

Moreover, by applying a series of restrictive potential ranges after initial deep discharge lithiation (D1 0.01 V), we conclusively isolated the redox pairs G/W, (H + I)/W, and (G + H + I)/W as viable systems. This enabled us to systematically evaluate whether the reaction mechanisms proposed in the MoS₂ literature allow for all electrochemical pairs to exist. Only literature mechanisms IRR3 by Zhang et al.⁹ and MIX2 by Halankar et al.²² withstood identification of the G/W, (G + H)/W, and (G + H + I)/W redox pairs. However, both mechanisms consider the presence of preserved pristine MoS₂ after the first discharge, which contradicts the findings from the restricted potential range 1.80–0.01 V. Furthermore, both mechanisms lack conclusive justification for their transition from a two-step lithiation (A ∼0.8 V) and conversion (B ∼0.4 V) process in the first discharge to a one-step (I ∼0.3 V) process from the second discharge onwards. Moreover, MIX2 also considers a highly asymmetric capacity pairing between peaks I/V.

Therefore, we proposed three alternative mechanism flows (over-lithiation, mixed parallel reactions, and polysulfide species) that could be viable electrochemical options but require further material characterisation and specific electrochemical testing for confirmation. More importantly, we put together a conclusive list of electrochemical requirements for future studies to explain, when presenting lithiation mechanisms for commercial bulk MoS₂ with the same electrochemical response (W, G, H, and I), including:

• Cathodic reactions (G, H, and I) are parallel or series-independent;

• The single anodic peak (W) pairs with all cathodic peaks to form redox pairs G/W, (G + H)/W, or (G + H + I)/W, depending on the potential range;

• No electrochemically accessible pristine MoS₂ is preserved beyond the presence of peak B in D1.

Author contributions

ADM and PLC were responsible for the project conceptualization. ADM manufactured all the electrode laminates and characterised them. ADM and ARS developed the coin cell testing methodology and conducted cell cycling. ADM collected and analysed the battery data with custom Matlab scripts. ADM wrote the initial draft. ARS, CAH, and PLC reviewed and edited the manuscript.

Conflicts of interest

There are no conflicts to declare.

Data availability

The electrochemical data presented in this article, with either Biologic (.mpt) or Neware (.txt) file formats, is available at Science Data Bank repository: https://doi.org/10.57760/sciencedb.32974.

Supplementary information (SI): literature information, electrode details, coin cell details, material characterisation (pristine MoS₂ electrode digital images, SEM images, XRD diffraction patterns, and Raman spectroscopy scans), further cycling data, and further electrochemical analysis. See DOI: https://doi.org/10.1039/d6ta03464k.