Fast Water-Assisted Lithium Ion Conduction in Restacked Lithium Tin Sul ﬁ de Nanosheets

: While two-dimensional (2D) materials may preserve some intrinsic properties of the corresponding layered bulk material, new characteristics arise from their pronounced anisotropy or con ﬁ nement e ﬀ ects. Recently, exceptionally high ionic conductivities were discovered in 2D materials such as graphene oxide and vermiculite. Here, we report on the water-assisted fast conduction of lithium ions in restacked lithium tin sul ﬁ de nanosheets. Li 0.8 Sn 0.8 S 2 exfoliates spontaneously in water and can be restacked into homogeneous ﬁ lms in which the lithium content is decreased, and a partial substitution of sulfur with hydroxyl groups takes place. Using a recursive supercell re ﬁ nement approach in reciprocal space along with real-space pair distribution function analysis, we describe restacked lithium tin sul ﬁ de as a partially turbostratically disordered material composed of lithium-containing and lithium-depleted layers. In humid air, the material takes up multiple layers of water that coordinate lithium ions in the space between the layers, increasing the stacking distance and screening the interaction between lithium ions and the anionic layers. This results in a 1000-fold increase in ionic conductivity up to 47 mS cm − 1 at high humidities. Orientation-dependent impedance spectroscopy suggests a facile in-plane conduction and a hindered out-of-plane conduction. Pulsed ﬁ eld gradient nuclear magnetic resonance spectroscopy reveals a fast, simultaneous di ﬀ usion of a majority and a minority species for both 7 Li and 1 H, suggesting water-assisted lithium di ﬀ usion to be at play. This study enlarges the family of nanosheet-based ionic conductors and helps to rationalize the transport mechanism of lithium ions enabled by hydration in a nanocon ﬁ ned 2D space. and EDX measurements and SEM micrographs, and additional impedance measurements; and PFG NMR data (PDF)


■ INTRODUCTION
Several technological applications require improved development and understanding of materials with fast ion transport that can be easily fabricated on a large scale. In this regard, twodimensional (2D) materials have garnered significant attention as their anisotropic properties and miniature dimensions are attractive for various fields including nano(opto)electronics, energy conversion and storage, catalysis, membranes, or nanofluidics. 1,2 When such materials contain nanoconfined fluids, they can exhibit rapid ion diffusion and transport kinetics. 2 Deeper insight into the mechanisms underlying this behavior is desired.
Compared to their bulk counterparts, 2D single layers and restacked materials generally show different physical properties, 3−7 often developing new and useful characteristics due to their pronounced anisotropy and/or confinement effects. 6,8,9 The most prominent example of this behavior is graphene. 3,10−13 For instance, lithium shows fast chemical diffusion in bilayer graphene and water flows exceptionally fast in artificial nanochannels built from graphene. The fast flow of water is associated with an increased structural order in nanoconfined water and ionic motion which depends on the interaction of the hydration shells with the channel walls. 13,14 Besides, graphene oxide (GO), a derivative of graphene, also shows unusual diffusion properties. 15−19 GO films that are formed by restacking GO nanosheets take up water from the environment, which is embedded in the nanoconfined space between the layers. These hydrated systems show exceptionally fast conduction of different ions such as H + , K + , Mg 2+ , and Ca 2+ . The ionic conductivities at low electrolyte concentrations (<50 mM) in GO films even exceed the conductivity of the cations in aqueous salt solutions. 16 In this case, conductivity is rather independent of electrolyte concentration, which is characteristic of surface-charge governed ionic transport. Similar results were observed for GO samples exposed to a high relative humidity (RH). These show fast proton conduction up to σ = 4 × 10 −4 S cm −1 for thick multilayer films at 60% RH and σ = 7 × 10 −3 S cm −1 for single GO layers at 95%. 17,20 It is assumed that the hydrophilic surface groups such as −O−, −OH, and −COOH attract protons from adsorbed water, which propagate through hydrogen-bonding networks along the adsorbed water film. 20 Ion conduction properties have been examined in only a few other restacked nanosheet materials. Vermiculite layers were assembled into a nanofluidic device showing high proton conductivity using near-neutral solutions (σ = 6 × 10 −3 S cm −1 ) with the advantage of higher thermal stability than GO. In this system, the conductance of lithium ions stemming from a LiCl solution and being transported through the vermiculite layers was also demonstrated. 21 Besides, a remarkably high in-plane hydroxyl conductivity of 10 −1 S cm −1 was observed in singlelayer, layered double hydroxides. 22 Moreover, thin films fabricated from the exfoliated and restacked phosphatoantimonates HSbP 2 O 8 23 and H 3 Sb 3 P 2 O 14 , 24 which show a large swelling upon water exposure, exhibit an increase in proton conductance of several orders of magnitude upon exposure to relative humidities from 0 to 100%. This increase is larger than the change observed for the non-exfoliated, bulk layered material, pointing to higher ionic mobilities in the restacked materials compared to their bulk counterpart, which makes them interesting candidates for photonic humidity sensing applications. 25 In this study, the fast, water-assisted conduction of lithium ions in restacked lithium tin sulfide (Li-TS) is reported. Its parental bulk material Li 0.8 Sn 0.8 S 2 is a solid lithium ion conductor. 26,27 Under exposure to humidity, Li 0.8 Sn 0.8 S 2 forms defined hydrates that exhibit order of magnitudes higher ionic conductivity than in the dry state. The impact of hydration on the structure and transport properties on Li 0.8 Sn 0.8 S 2 was first preliminarily shown by Holzmann 28 and then fully disclosed by Joos et al. 27 Moreover, Li 0.8 Sn 0.8 S 2 can be easily exfoliated into individual layers in liquid water and restacked into homogeneous films either on a substrate or even free-standing (cf. with Figure 1). 24,29 Due to the ultra-high refractive index of n = 2.5, thin Li-TS films have been utilized as stimuli-responsive 1D photonic crystals and upon hydration by water vapor, a strong swelling of the films was reported by Szendrei-Temesi et al. 30 In line with previous reports correlating swelling upon humidity exposure and ion conduction, we find that the total conductivity of restacked Li-TS is indeed highly dependent on the RH of the environment and reaches up to 47 mS cm −1 . Orientationdependent measurements indicate a preferred in-plane conduction (in between the layers). Moreover, we develop a structural model for the dry and hydrated restacked Li-TS and demonstrate by pulsed field gradient nuclear magnetic resonance spectroscopy (PFG NMR) that the diffusion of the lithium ions and water takes place simultaneously.

■ EXPERIMENTAL SECTION
Preparation of Restacked Nanosheets. The precursor material Li 0.8 Sn 0.8 S 2 was prepared and exfoliated according to Kuhn and Holzmann et al. 26,29 The nanosheet suspension was washed once by centrifuging at 20000 rpm for 10 min and then redispersing in purified water. Since during exfoliation H 2 S evolves, exfoliation should be conducted in a fume hood. The pH of the suspensions was measured with a pH indicator paper. For thin films of restacked nanosheets on a substrate, 4 mL of the suspension (1 mg mL −1 ) was drop-cast onto alumina substrates (CRYSTEC, 10 × 10 × 0.5 mm, single crystal cleaved along (0001), polished on one side) placed in a plastic reservoir with a volume of 12.6 cm 3 and slowly dried at room temperature. By this procedure, thin films with a thickness of about 1−3 μm were obtained. The thickness of the films on a substrate was measured by scanning electron microscopy with a Merlin SEM (ZEISS). To fabricate freestanding films and powder samples, the suspension was poured in a vessel made of polypropylene and dried between 25 and 60°C in an oven for several days. Subsequently, the films could be simply peeled off the vessel. The thickness of the free-standing films was measured by evaluation of images made using an optical microscope (LEICA DM 2500 M).
X-ray Measurements. X-ray powder diffraction (XRPD) experiments were conducted using a STOE STADI P diffractometer (Mo K α1 radiation, Ge(111) monochromator, MYTHEN 1 K Detector) in a Debye−Scherrer geometry. For humidity-dependent measurements, capillaries were stored under a water-saturated atmosphere before sealing. Rietveld refinements were carried out with the program TOPAS v. 6.0. 31 Total scattering measurements were carried out using the highenergy Powder Diffraction and Total Scattering Beamline P02.1 of PETRA III at the Deutsches Elektronen-Synchrotron (DESY, more details are in the Supporting Information). X-ray total scattering data were collected in the rapid acquisition mode (RAPDF). 32 A large-area 2D PERKIN ELMER XRD1621 detector (2048 × 2048 pixels, 200 × 200 μm 2 each) was used at a sample-to-detector distance of approximately 304 mm. Samples were loaded into 1.8 mm ID and 1.9 mm OD polyimide capillaries (Cole-Parmer) and measured at room temperature. The incident energy of the X-rays was 59.858 keV (λ = 0.20713 Second, after washing the suspension by centrifuging, homogeneous thin films either on an alumina substrate (a thickness of about 1 μm, magnification in Figure S5e) or free-standing (a thickness of about 10 μm) can be fabricated by drop-casting and drying at RT. By grinding the free-standing films and compacting the powder, a pellet with a random orientation of nanosheets is created. The black arrows indicate the orientation of electrodes during conductivity measurements.
Chemistry of Materials pubs.acs.org/cm Article Å). Data integration and reduction to the pair distributions functions (PDFs) were performed using fit2D 33 and xPDFsuite, 34,35 respectively. Structure refinements were performed using both PDFgui 36 and Diffpy-CMI. 37 Impedance Spectroscopy. Electrochemical impedance spectroscopy (EIS) was performed with an IVIUM compactstat.h (24 bit instrument) in an enclosed two-electrode setup. The RH of the stagnant gas was controlled using saturated salt solutions. 38,39 Each sample was equilibrated at the respective RH for 24 h. After this time, no further change in conductivity was measurable. For in-plane measurements, the films on alumina substrates were sputter-coated with 80−100 nm gold or platinum electrodes, while the free-standing films contacted with carbon foils in a home-built measurement cell. The relative density of the films of about 91 ± 8% was calculated from the measured weight and thickness of the films and normalized on the crystallographic density of LiSnS 2 . The measurement cell for the inplane measurement was a gastight Schott glass with a total volume of 250 mL equipped with humidity and temperature sensors from Sensirion installed inside the chamber. The chamber was submerged in water and kept at a defined temperature with a JULABO FP50ME thermostat. For out-of-plane measurements, the samples were pressed between two pieces of carbon foil in a TSC Battery cell from RHD Instruments, humidified for 24 h in a desiccator at the respective humidity, and then sealed gastight. The same films as for in-plane measurements were also sample out-of-plane as well as pressed pellets (8 mm diameter, pressed with 40 kN; the relative density normalized on crystallographic density of LiSnS 2 is about 85 ± 3%). The applied rms AC voltage was 100 mV for films and 10 mV for the pellets. The setup applied for measuring the dry pellet (with sputtered ruthenium metal electrodes) and under varying partial water pressure is described in the Supporting Information, Section 7.1. Impedance spectra were analyzed using the RELAXIS3 software (RHD Instruments).
Nuclear Magnetic Resonance Spectroscopy. 1 H, 7 Li, 6 Li, and 119 Sn solid-state NMR spectra were recorded on a BRUKER Avance III 400 MHZ instrument at Larmor frequencies of 400.1, 155.6, 58.8, and 149.2 MHz, respectively (B 0 = 9.4 T). 1 H spectra are externally referenced to tetramethylsilane (TMS, δ iso = 0.0 ppm and 7 Li spectra to a 1.0 M aqueous solution of LiCl δ iso = 0.0 ppm). 119 Sn spectra are referenced to tetramethyltin (Sn(CH 3 ) 4 , δ iso = 0.0 ppm) with solid SnO 2 used as a secondary chemical shift standard (δ iso = −603.0 ppm relative to Sn(CH 3 ) 4 ). 40 Magic-angle spinning (MAS) experiments were performed in ZrO 2 spinners at a spinning frequency of 5−14 kHz using a BRUKER 4 mm double-channel probe. The temperature of the samples was controlled using a BRUKER BVT3000 temperature controller and was calibrated using the 207 Pb signal of Pb(NO 3 ) 2 . 41 For pulsed field gradient (PFG) NMR measurements, the samples were sealed in a glass tube with an additional fitted glass rod inside to reduce the dead volume and to avoid desorption and evaporation of water during the experiments. 7 Li and 1 H PFG NMR measurements were performed on a BRUKER Avance III 400 MHz instrument (B 0 = 9.4 T; 1 H Larmor frequency of 400.1 MHz, 7 Li Larmor frequency of 155.56 MHz), equipped with a diff60 single-gradient diffusion probe. The probe allows for pulse field gradients g of up to 30 Tm −1 and variable temperature measurements up to +150°C. The diffusion measurements were accomplished using a stimulated echo pulse sequence. 42 The echo attenuation curves S(g, δ, Δ) were processed using a biexponential form of the Stejskal−Tanner equation, 43 S(g, δ, Δ) = S 0A exp ((−γ 2 δ 2 g 2 )D A (Δ − δ/3)) + S 0B exp ((−γ 2 δ 2 g 2 )D B (Δ − δ/3)), accounting for at least two independently moving nuclei (species A and B) with distinct effective diffusion coefficients D A and D B . Here, γ is the 1 H or 7 Li gyromagnetic ratio, respectively, g is the strength, δ is the duration of the pulse field gradient, and Δ is the time interval between the field gradients that defines the diffusion time scale. The signal intensity of each 1 H and 7 Li spectrum was obtained by applying peak area integrals. The PFG data were analyzed with TOPSPIN3.6 (BRUKER).

■ RESULTS AND DISCUSSION
Exfoliation and Chemical Composition. Exfoliation of Li 0.8 Sn 0.8 S 2 in water can be described by the formal removal of Li 2 S, which is realized by evaporation of H 2 S and the formation of LiOH in aqueous solution according to eq 1 This exfoliation process is distinct from the exfoliation of Liintercalated Li x SnS 2 , featuring partly reduced Sn(II), in aqueous solution into exfoliated SnS 2 , LiOH, and H 2 . 44−46 As previously shown by Kuhn et al., 29 the in-plane structure of the exfoliated material is reminiscent of a "defective" variant of SnS 2 , with a lower electron density at the S positions indicating the presence of sulfur vacancies. This suggests a (partial) filling of the sulfur vacancies by hydroxide ions according to eq 2 Following the pH during exfoliation gives more information about this process: instead of obtaining a basic pH as expected from eq 1, the stable nanosheet suspension is neutral at pH 7 and after long periods of time, the pH shifts into the acidic regime (pH = 4−5 after months in a sealed bottle), indicating an acid− base process to be at play. These pH changes can be rationalized by the gradual incorporation of hydroxyl groups (or oxygen) into the sulfur vacancies in [Sn 0.8 S 2−x ] −(0.8−2x) over time. During this process, protons are released into the suspension according to eq 2. The insertion of hydroxyl groups is consistent with the appearance of signals at −740, −660, and −585 ppm in the 119 Sn NMR spectra that are clearly distinct from SnS 2 and SnO 2 (Figure S1−S2). Raman spectra also support the insertion of hydroxyl groups ( Figure S4). The shift of the most intense signal of Li-TS to higher wavenumbers with respect to SnS 2 is consistent with bonding interactions between tin and oxygen. Moreover, 2D 1 H− 119 Sn NMR correlation spectra of the dry sample indicate that residual protons are present near oxygen− tin bond species, again suggesting that hydroxyl ions replace sulfur ions ( Figure S3a).
After the initial hydroxide insertion step, slow hydrolysis of the nanosheets proceeds. After several months of long storage in aqueous solution or as a dry material in air, thermodynamically stable SnO 2 particles start to form on the surface of Li-TS films, as shown in Figures S5c,d. To prevent further hydrolysis and to remove excess LiOH (cf. Figure S5a,b), the suspension was centrifuged and washed with purified water once before further investigations. The restacked Li-TS was used as freshly prepared as possible and investigated in two different states: the "dry" state and the "hydrated" state (see schematics in Figure 2). Upon exposure to humid air, the layers reversibly adsorb water likely driven by the large hydration energy of lithium ions as already demonstrated by ellipsometric porosimetry. 30 The dry state is obtained by heating the sample overnight at 120°C under vacuum and subsequent handling under Ar. In this state, most of the adsorbed water is removed, and only 1.5 wt % residual water is present ( Figure S6b, data point at 0% RH). The restacked Li-TS retains about 44 ± 8% of the initial lithium content (Table S1) and shows one broad signal in the 6 Li MAS NMR spectrum in the dry state and two distinct sharper signals in the hydrated state (cf. Figure 4c). Moreover, as already suggested by solid-state 119 Sn NMR, the elemental analysis (Supporting Information, Section 3 and Tables S1−S3) also indicates a partial substitution of sulfide with hydroxide (OH/S Chemistry of Materials pubs.acs.org/cm Article = 0.19) in the anionic network, lowering the S/Sn ratio toward ≤2 while preserving the initially present tin vacancies. The exact composition varies slightly for each batch. At high RHs (100% for 24 h) Li-TS adsorbs up to 32 wt % water (TGA data in Figure S6b−d). This corresponds roughly to 9 water molecules per lithium ion and 2.6 water molecules per sum formula. In summary, the exfoliation process changes the tin−sulfur−oxygen content significantly, while substantial amounts of lithium ions are washed out, making the exfoliated, restacked material chemically distinct from both pristine Li 0.8 Sn 0.8 S 2 and SnS 2 .
Structural Characterization. Restacked Li-TS films prepared by drop casting show a homogeneous thickness between 20 nm and several μm (see Figures 1 and S5b,d). To better understand the 3D structure, the powdered materials were further investigated using XRPD and PDF analysis. The diffraction patterns for the dry and hydrated material on a lab diffractometer (Mo K α1 radiation) are shown in Figure S6, and the synchrotron measurement with Rietveld refinement 47 is shown in Figure 3a.
First, we will discuss the structure of the dry Li-TS sample (for details, see the Supporting Information, 5.1−5.2 and Figures S7−S14). The patterns show broad, triangular-shaped "Warrentype" peaks characteristic of orientational (i.e., turbostratic) disordered layers, as visualized in Figure 3a. 48 Such peaks are often observed when layered materials such as tetravalent sulfides are restacked from suspension. 49 Additionally, the 003 basal reflection is split into two overlapping Bragg peaks, suggesting a modulation of the interlayer distance, likely by interstratification effects. Thus, by applying a supercell approach, 50,51 the microstructure of the stacking-faulted dry Li-TS was approximated by a combination of stacking faults and turbostratic disorder as summarized in the following.
The structure of dry Li-TS contains basic stacking motifs related to Li-containing Li 4x Sn 1-x S 2 (x > 0) (with an interlayer distance of 6.137(50) Å) and "defective", Li-depleted SnS 2 optimized for the C19-type stacking (with an interlayer distance of 5.90(10) Å). The former motif can be approximated and modelled by the structure of anhydrous LiSnS 2 52 with a reduced lithium and tin occupancy similar to the parental material. This Li 0.8 Sn 0.8 S 2 model has a typical CdCl 2 -like C19 stacking with lithium residing between the layers. Exfoliation in water presumably removes lithium ions that were originally located on Sn sites (inside the Sn−S layers) and lithium ions in the interlayer galleries between the Sn−S layers in the parental material. This removal of lithium ions leaves behind lithium vacancies within the Sn−S layers. However, no information on the distribution of voids within the layers and on the presence of intralayer lithium could be extracted from the refinements. In 6 Li MAS NMR of restacked dry Li-TS, inter-and intralayer lithium positions are no longer distinguishable (cf. Figure 4c), hinting to broad dispersion of lithium positions caused by the structural disorder. Consequently, the question whether intralayer lithium is removed entirely cannot be unambiguously resolved from the MAS NMR and PXRD data. The other motif is more similar to pure SnS 2 . The model exhibits a CdI 2 -like C6 stacking with vacant octahedral, interlayer cation positions in the hexagonal close-packed anion lattice (see Figure S12). This presence of SnS 2 -like regions is likely caused by exfoliation and restacking which generates lithium vacancies in the interlayer space, leading to the large overall lithium deficiency in Li-TS.
However, the sole combination of faultless Li 0.8 Sn 0.8 S 2 and SnS 2 structure models is not sufficient to describe the structure of dried Li-TS (cf. Figure S14, R wp = 12.21%). To obtain a satisfactory description of the experimental data, both faulting between these two motifs and turbostratic disordering of each layer are required (R wp = 3.41%). Dry, restacked Li-TS thus contains a high faulting concentration (∼40%) of CdI 2 -type (C6) (SnS 2 -like) within CdCl 2 -type (C19) (Li 0.8 Sn 0.8 S 2 -like) Figure 2. Upon exposure to water vapor, water molecules adsorb between the restacked nanosheets and lead to swelling of the material. Chemistry of Materials pubs.acs.org/cm Article interlayer associations, depending on the local presence or absence of lithium, as visualized in Figure 3b. This is in agreement with the loss of 40% of initial lithium content in the refined Li-TS sample (Table S2, sample 11). The layer orientations are significantly (although not completely) modulated by turbostratic disorder, with the refinement converging to random in-plane translations of ±42.5% of the ab-dimensions.
Turning to the pattern from the hydrated Li-TS sample ( Figure 4a, and further explanation in Supporting Information Figures S15−S19), the basal reflection is shifted to a lower angle than that seen for the dry sample. This shift indicates a significant increase in interlayer distance due to the intercalation of water. Depending on the sample and its history, interlayer distances ranging from 9.3 to 11.9 and 12.8 Å were observed (vs 6.1 Å in dry Li-TS), which could be associated with the intercalation of roughly two or three layers of water, respectively (Supporting Information, Figure 10). The difference between 11.9 versus 12.8 Å possibly stems from the change of octahedral to tetrahedral coordination of the lithium ions  Figure S16). The diffraction patterns in Figure S15 indicate further structural characteristics in the hydrated state and can only be sufficiently well described by the inclusion of two phases of hypothetical Li 0.8 Sn 0.8 S 2 ·3H 2 O representing different local stacking orders (abc-stacked α-Li 0.8 Sn 0.8 S 2 ·3H 2 O + ab̅ ab̅ -stacked α-Li 0.8 Sn 0.8 S 2 ·3H 2 O, Figure  S19) as well as regions containing dry Li 0.8 Sn 0.8 S 2 and dry SnS 2 (cf. Figure 4a and Supporting Information 5.3). First, the interlayer distances are modulated by random variations in both the number of intercalated water layers and the relative orientations of the Sn−S layers. This is supported by the fact that the basal reflections 003 are more broadened than the 100 reflection, which would not be expected for simple turbostratic disorder with identical layer distances. 53 Second, the characteristic triangular Warren-type peak shape indicates that the hydrated samples contain a high density of stacking faults as with the dry sample. Third, there are domains without any water intercalation, as indicated by the larger-than-expected intensity of the secondorder basal reflection (006) of the hydrated state, due to the overlap of the dry-state basal reflection. From these considerations, a model for the highly hydrated material was derived. Due to the high number of possible structural defects, we could not use the approach for fitting the XRPD pattern that was applied for the dry samples. Instead, we used a multiphase approach, similar to the one applied for SnTiO 3 , 54 in which each phase represents certain microstructural features of the sample. Three layers of water (Li 0.8 Sn 0.8 S 2 ·3H 2 O) were used to describe the average hydrated domains. Since tin and sulfur are by far the strongest scatterers in the structure, shifts between staggered and eclipsed stacking of the SnS 6/3 -octahedra layers have the strongest impact on the diffraction pattern. The intercalation of three layers of water can lead to a C19-type ( Figure S18,a) or C6-type-like stacking ( Figure S18,b) of the SnS 6/3 -octahedra layers, so phases for both types were included. Phases of both non-hydrated Li 0.8 Sn 0.8 S 2 and SnS 2 were also included to describe the non-hydrated faulting domains. The final fit and a visualized representation of the different domains in the material are shown in Figure 4a,b.

and
The resulting two environments for lithium are in line with the two lithium signals found in the 6 Li MAS NMR spectrum of hydrated Li-TS in Figure 4c. The signal at 0.26 ppm is shifted toward higher fields with respect to that of the dry sample, presumably due to the presence of a high amount of surrounding water screening the electrostatic interaction between lithium and sulfur, as visualized in Figure 4b. The peak at 0.95 ppm corresponds to a low water content, being similar to the dry material. Thus, these findings help to interpret the PFG NMR data in Figure 7 that suggest the presence of two lithium species in the hydrated samples: a mobile and a fairly immobile species. In the hydrated layers, where three layers of intercalated water are present, the lithium cations are fairly mobile. In contrast, in the non-hydrated layers, the lithium ions show a lower local mobility and therefore give an NMR signal similar to that of the dry sample. Notably, we associate the main influence on the chemical shifts of the two types of lithium ions with the difference in surrounding water content, but we cannot disregard the possibility that under hydration, lithium ions fill vacancies inside the S−Sn layers, as visualized in Figure 4d on the top. This might also explain the existence of two lithium species with distinct diffusion coefficients.
PDF data were measured and analyzed to support the reciprocal-space analysis and further investigate details of the atomic structure (see the Supporting Information for more details). Most notably, the PDFs for both dry and hydrated samples were very similar and could be suitably described by a single SnS 2 -layer model refinement. 55 This confirms the strongly turbostratic nature of the layers in both states, resulting in weak correlation between interlayer atom pairs, and further rules out any likelihood for ordered bi-or multilayer stacks. For the dry sample, a similar goodness of fit could be achieved for a turbostratic, 3D model with only three layers in the unit cell compared to the 2D, single-layer model. This further suggests that preferred, relative orientations likely exist depending on the amount and distribution of intercalated lithium. The preference of the data toward models with in-plane distortions is indicative of local distortions of the Sn−Sn distances due to the presence of discrete vacancies and/or Li substitutions, and a misfit of the second nearest neighbor shell suggests that the atoms in the turbostratically disordered layers also distort slightly out-ofplane. This was tested using supercells of both single-layer (2 × 2) and 3D (2 × 1 × 1) models by allowing each Sn and S atom to distort slightly off its symmetrically allowed position in only the z-direction. While distinctive features of the data could not be identified to determine whether Sn site vacancies (or possibly Li substitutions) are locally ordered or random, refinements suggest that the density of vacancies/substitutions is relatively low.
The hydrated material measured at the synchrotron had a lower water content compared to the samples of the Rietveld refinement due to drying during transport to the synchrotron. The layer distance of 9 Å associated with a bilayer of water was observed, but with lower relative intensity compared to the nonhydrated basal reflection in the same data set. However, a significant SAXS signal increasing toward the Q min (= 0.4 Å −1 ) of the measurement was observed, which did not appear in the dried sample. This may suggest smaller domain sizes in the hydrated samples or possibly scattering effects due to other microstructural effects such as delamination.
Transport Properties. Conductivity Measurements. The electrical properties of Li-TS are distinct from the parental Li 0.8 Sn 0.8 S 2 26 and SnS 2 56 but similar to the properties of Li 0.8 Sn 0.8 S 2 hydrates. 27 To assess the ion conduction properties of dry Li-TS first, a pellet pressed from the fine ground powder was measured under argon flow (cf. Supporting Information Section 7.1 and Figure S26). The obtained EIS semicircles are severely depressed and are not well described with a single Par(R, CPE) equivalent circuit (R: resistance, CPE: constant phase element).
Regarding the structural analysis of Li-TS and comparing it with the parental compound Li 0.8 Sn 0.8 S 2 , one might still expect the dominant mobile charge carrier to be Li defects after exfoliation. However, the increased complexity of the granular microstructure and structural disorder suggest prominent changes in diffusion paths, which directly affect the relaxation times. Therefore, rather than one defined process, we should consider a distribution of relaxation times for this system, which explains the severe depression of the observed EIS arc and requires a fitting with two Par(R, CPE) elements in series (cf. Figure S26). Hence, we will only consider the total (DC) resistance of the sample (i.e., the intercept of the semicircle with the Z re -axis at low frequencies), R 1 + R 2 = R Tot , corresponding to the total conductivity σ = d/(AR tot ) with d being the thickness and A the area of the sample. The conductivity of dry Li-TS has an activation energy of (0.55 ± 0.2) eV and a room-temperature magnitude of (3.4 ± 0.5) × 10 −8 S cm −1 . This conductivity is Chemistry of Materials pubs.acs.org/cm Article lower than that in the parent Li 0.8 Sn 0.8 S 2 compound (<10 −5 S cm −1 ), probably caused by the depletion of lithium ions in restacked Li-TS as well as structural changes. Furthermore, the restacked Li-TS shows ion conducting properties which are intriguingly sensitive to the humidity of the environment. Li-TS was measured as a function of water vapor pressure (p H 2 O ) at temperatures ranging from 30 to 100°C, as shown in Figure S27. At high temperatures and low humidities (≤2% RH at 100°C, ≤25% RH at 60 and 30°C), the same (R)(CPE)-(R)(CPE)-CPE model as for the dry Li-TS can be applied (the inset in Figure 5). However, upon a certain degree of hydration of the pellet (Figure 6b, starting from an RH of 15% at 60°C), the semicircle is fit only by one R-CPE element with a dielectric constant of 50 (being similar to the species with the low dielectric constant in dry Li-TS). Figure 5 shows the large impact of the hydration on the ionic conductivity at varying temperatures. At 30°C, the conductivity increases by several orders of magnitude from 10 −8 to 10 −4 S cm −1 , which is much more drastic than at 100°C. Since the structural characterization showed that water incorporates in the bulk of the material, the strong increase in conductivity is enabled by incorporated H 2 O. Since the interaction of lithium ions with the anionic Sn−S layers is screened by the coordination with water, we assume an increase in mobility of the lithium ions, leading to the high ionic conductivities. At 100°C, the humidity is low; consequently, only a small amount of water incorporates into the structure, leading only to a small increase in conductivity. Even though the formation of a surface layer is possible, which becomes conducting under very humid conditions, especially at low temperatures, the dominance of bulk conduction is supported by the structural analysis and surface effects were therefore neglected.
Nevertheless, we cannot entirely rule out the formation of new mobile carriers, such as protons and hydroxide ions, that could contribute to the ionic conductivity (besides Li + ) and/or a more complex underlying transport mechanism (Supporting Information 7.1). Hydroxyl-substituted Sn−S can show acid/base chemistry in principle, but a change in pH of the nanosheet suspension during exfoliation and restacking only leads to a decrease in observed conductivity. As stated below, probing these non-structural charge carriers is challenging and NMR (vide inf ra) shows no additional signal for protons or hydroxides in the hydrated state, rendering the detection of the influence of additional charge carriers unfeasible.
To probe the dimensionality of ion transport, which is often preferred in the lateral plane of the nanosheets, as observed in other restacked and single-layer materials, 21,22 a foil-like freestanding film of Li-TS was measured using in-plane and out-ofplane geometries and compared to a pellet with a random orientation of layers (cf. Figure S28). To ensure full hydration, the samples were exposed to 100% RH and equilibrated for 24 h at RT. The in-plane conductivity exceeds the out-of-plane conductivity by more than 2 orders of magnitude, indicating facile transport along the layers. This suggests that the lithium cations move easily through the water-filled space between the Sn−S layers but have a limited opportunity to move perpendicular to the layers from one interlayer gallery to the one below or above. Remarkably, Li-TS films show ionic conductivity in the temperature range down to −25°C as shown in Figure S29, that is, below the freezing point of water, without any sharp change in value. This points to a melting point depression for water as observed in other nanoporous and nanoconfined 2D systems. 57−59 The pellet sample measured under these conditions has a conductivity of 0.1 mS cm −1 , which is lower than the conductivity in the in-plane but higher than that in the out-of-plane orientation of the free-standing film. This is consistent with a mixed orientation of layers, decreasing the conductivity.
To probe even thinner Li-TS films and to enhance the mechanical stability, Li-TS was coated onto an alumina substrate. The ionic conductivity as a function of RH of the film on the substrate is presented in Figure 6 in comparison to the pellet measurement at 30°C. With increasing RH (16 to 92% RH), the resistance of the thin film on alumina drops and the in-plane ionic conductivity increase drastically over 3 orders of magnitude from σ EIS = 0.03−47 mS cm −1 . Figure 6c shows a representative impedance spectrum fitted with the given model to extract the resistance of the spectrum. The model includes one Par(R, C) in series to another Par(R, CPE) to account for the high (R1 and C1) and low (R2 and CPE1) frequency semicircles that are attributed to the bulk properties and contributions from processes at the electrode, respectively. The polarization of ions at the interface of the blocking electrode was modelled by CPE2. The ionic conductivity (σ EIS ) is calculated from impedance spectra by applying σ EIS = l/(AR 1 ) with l being the distance between the electrodes, A the surface area of the sample calculated from the length of the electrode and the thickness of the film (cf. Figure 1 and values in the Supporting Information, Table S5), and R 1 . 60 On thin films, some transport of charge carriers on the surface could influence the EIS measurement, but as water clearly incorporates into the structure as shown above and the measurements of the pellet suggest, we attribute the measured conductivity to the bulk process within the film. The conductivity (cf. Figure 6) can be roughly assigned to three regimes associated with an increased H 2 O/Li ratio deduced from TGA measurements ( Figure S6b,c): with increasing H 2 O/ Li ratio, the conductivity increases. In the regime with 3.4 < H 2 O/Li, a maximum conductivity of 47 mS cm −1 can be found at 92% RH. It is on par with state-of-the art solid lithium ion conductors. 61 The activation energy of the thin film on alumina in the regime with 3.4 < H 2 O/Li of 0.29 eV as measured at 76% RH in the temperature range of 45−70°C (cf. Figure S30) is very similar to the activation energy of the Li 0.8 Sn 0.8 S 2 hydrates Figure 5. Total ionic conductivity of a Li-TS pellet as a function of RH at varying temperatures (30,60, and 100°C) clearly shows a large impact of hydration. The inset shows the impedance spectrum at 100°C and low RH. For these measurements, sputtered ruthenium was used to contact the pelletized samples.
Chemistry of Materials pubs.acs.org/cm Article and lithium montmorillonite and only slightly smaller than that in sodium vermiculite (0.50 eV). 27,62,63 Notably, the out-ofplane activation energy of 0.27 eV, determined from the measurement ( Figure S29) at 76% RH, is very similar to the activation energy of the in-plane measurement. The in-plane conductivity of 9 mS cm −1 of the free-standing film is a factor of 3 smaller than that of the film on a substrate. This might be an effect of the more heterogeneous microstructure or increased thickness (factor 10−30) of the free-standing films in comparison to the very uniform thin films on a substrate. Diffusion Study by PFG NMR. Having established a high ionic conductivity of hydrated Li-TS nanosheet films, the nature of ionic charge carriers cannot be unambiguously resolved by EIS. Measurements with ion-blocking electrodes cannot distinguish between different ionic charge carriers (e.g., Li + or H + ), while the use of Li metal as a Li-selective electrode proved impossible due to decomposition of Li in the presence of water. Similarly, proton-conducting materials such as Nafion are also known to conduct lithium ions. 64 The use of an aqueous lithium salt solution as done by Raidongia and Huang 16 was not feasible since the restacked Li-TS nanosheets redisperse in solution. Thus, to further elucidate the type of charge carriers and the corresponding diffusivities, 7 Li and 1 H pulsed field gradient (PFG) NMR was applied on powder samples. Similar to EIS,   7 Li attenuation curves of the normalized peak area as a function of (γ 2 δ 2 (Δ − δ/3)g 2 in 10 10 sm −2 ) compared to a monoexponential decay and to the 2D diffusion model according to Stoll et al. 77,78 (b) Diffusion coefficients of all components extracted from 1 H and 7 Li PFG NMR data. The diffusivity of 1 H is of the same order of magnitude as that of 7 Li. (c) The diffusion coefficients of components 1 only slightly depend on the diffusion time Δ, but components 2 show a stronger dependence, indicating some sort of inhibition on a larger distance.
Chemistry of Materials pubs.acs.org/cm Article PFG NMR probes the long-range ion dynamics of a sample. PFG NMR measures diffusion processes in the 10−100 ms regime on a μm length scale (vide infra), and EIS measures the conduction of ions in a time domain of 1 × 10 −6 to 100 s over the whole sample thickness (mm length scale). PFG NMR has already been used for investigating the mobility of lithium ions and water in a variety of materials such as carbon nanotubes, 65 organo-functionalized GO, 66 hydrated zeolites, 67,68 and liquid and solid lithium ion conductors 69−73 among others. All the following 7 Li and 1 H PFG NMR measurements were conducted on the same sample to elucidate the interplay of lithium ions with the adsorbed water. In the dry state, where the conductivity is very low <10 −7 S cm −1 , no measurable diffusion of 7 Li or 1 H was found by PFG NMR. In the hydrated state, the typical signal of adsorbed mobile water is visible at 4.57 ppm in the 1 H MAS NMR in Figure S3b. Due to a lack of detectable protonic charge carriers (ca. 6.5−7 ppm), 74−76 the contribution of protons to the conductivity is neglected for the interpretation of the data and we assume molecular water to be the dominant mobile species. The PFG NMR data, after exposure to 100% RH for a week (42 wt % H 2 O), show a similar behavior for water and lithium ions following a biexponential decay (cf. Figure 7). In general, PFG NMR probes the self-diffusion of a nucleus by measuring the attenuation of the static NMR signal as a function of the field gradient strength g. While a monoexponential decay is expected for a single diffusing species with a 3D trajectory, our data cannot be explained based on this simple model (long dashed lines in Figure 7). Assuming 2D diffusion as observed by PFG NMR for layered zirconium beryllium hydrides 77,78 and the parent lattice hydrates of Li 0.6 [Li 0.2 Sn 0.8 S 2 ] 28 also does not reproduce our data satisfactorily (dashed curved lines in Figure  7). A reasonably good fit for both the 7 Li and 1 H data was obtained only by using a biexponential model including two distinct diffusion coefficients. The population distribution of the diffusion coefficients is roughly 20 to 80% (Supporting Information, Figure 31) for both nuclei. Due to differences in the relaxation behavior of the different nuclei, quantification of PFG NMR data is inherently difficult. Nevertheless, the existence of two diffusing species is supported by the ratio of 7% for the peak at 0.95 ppm (minority species, 20% in PFG NMR) and 93% for the peak at 0.26 ppm (majority species, 80% in PFG NMR) in the 6 Li MAS spectrum of Figure 4. From here on, we refer to the diffusion coefficients originating from the majority species as component 1 and the ones of the minority species as component 2 for both 7 Li and 1 H diffusivity measurements. The higher diffusion coefficient (component 1) always stems from the majority species. Variable temperature 7 Li MAS measurements show a slow exchange of the two species. Therefore, the origin of the biexponential decay in the PFG data can be attributed to the presence of two independent, mobile lithium species in the sample, which is in agreement with the presence of both hydrated and non-hydrated Li species from structural analysis (Figure 4b).
An alternative interpretation of the observed biexponential decay follows the model of Osti et al. 79 (visualized in Figure 4d), which was developed for different diffusion coefficients of water in nanoconfined spaces, for example, in vermiculite clays. The water interacting with the wall is significantly slowed down in comparison to fast water in the middle of the confined space. However, observing a similar biexponential decay for 1 H and 7 Li suggests a coupled movement of lithium ions and water similar to the movement of sodium ions with their hydration sphere in vermiculite. 63 In the latter case, water coordinates the sodium ions between the SiO 2 −Al 2 O 3 layers and sodium ions move by dragging their hydration shells along (i.e., their motion is accompanied by water motion). In restacked Li-TS, the high diffusion coefficients of the majority species (component 1) at 305 K are of the same order of magnitude for both nuclei: D H 2 O = 4.3 × 10 −10 m 2 s −1 is 2 times higher than D Li + , which is 2.1 × 10 −10 m 2 s −1 .
The value for D H 2 O is very close to diffusivities observed for water in Li vermiculite 80 (D H 2 O = 3.4 × 10 −10 m 2 s −1 ) and Li montmorillonite 81 (D H 2 O = 4 × 10 −10 m 2 s −1 ), which also show 2D confinement of water and have alkali cations between anionic layers. Unfortunately, the diffusion coefficients of the alkali cations were not measured in that study. The diffusion coefficients of a sample exposed to 100% RH for a shorter period of time (1 d (identical to EIS samples) vs 1 week) are slightly lower and almost identical for both nuclei (see Figure S32). Apparently, interlayer water gradually becomes more bulk-like, and the diffusion coefficient of water approaches the value of water in 1 M LiCl (of D H 2 O = 1.73 × 10 −9 m 2 s −1 ).
However, D Li + exceeds the diffusivity of lithium ions in the hydrated zeolite LiLSX and in Li-β-alumina single crystals as determined by PFG NMR by 1 order of magnitude. 67,68,82 Of note, the D Li + value is also 2 times higher than the diffusion coefficient of the fully hydrated Li 0.8 Sn 0.8 S 2 , 27 very similar to the diffusion coefficient of lithium observed for a LiCl solution in Nafion, 64 and only slightly lower than that of free lithium ions in 1 M LiCl solution (8.0 × 10 −10 m 2 s −1 ).
The low diffusion coefficients of 7 Li and 1 H component 2 at 305 K in Figure 7b are of similar magnitudes with D H 2 O = 2.1 × 10 −11 m 2 s −1 and D Li + = 3.7 × 10 −11 m 2 s −1 and ca. 1 order of magnitude lower than that of component 1. The diffusion coefficients of 7 Li and 1 H as a function of temperature show an activation energy of 0.24 eV for the fast species and 0.4 eV for the slow species. The lower activation energy is close to the activation energy determined by EIS. A variation of gradient spacings and, hence, diffusion times Δ (Figure 7c) results in very similar diffusion coefficients. This is particularly the case for the fast-diffusing species, indicating essentially free lithium diffusion unimpeded by grain boundaries. This suggests a continuous diffusion path for hydrated Li throughout the sample. The slow species is impeded much more, possibly due to confinement within dry regions of the sample. The behavior of the 1 H diffusion coefficients with Δ is strikingly similar to that of the 7 Li coefficients, further strengthening the assumption of waterassisted transport of lithium ions. The isotropic displacement calculated by r D 2 rms NMR = Δresults in 4−6 μm for 7 Li and 1 H (component 1) and 1.5−2 μm for component 2. These distances span several layers and are thus consistent with lithium ion transport through the bulk of the material.
To evaluate whether the diffusion data probe the same process as the EIS measurements, the NMR conductivity (σ NMR ) was calculated based on the Nernst−Einstein equation σ NMR = D NMR tr nz 2 e 2 /k B T with n being the charge carrier concentration, e the elementary charge, k B the Boltzmann constant, and T the temperature using the diffusion coefficient D NMR tr obtained by PFG NMR (for details, see the Supporting Information). Assuming a lithium content of roughly 40% relative to the initial lithium content in the parental Li 0.8 Sn 0.8 S 2 based on ICP analysis, n = 4.6 × 10 27 m −3 is obtained. To account for the increase in interlayer spacing (d) upon hydration (dilution of charge carrier concentration), n is multiplied with d dry-Li-TS / Chemistry of Materials pubs.acs.org/cm Article d hydrated-Li-TS . From this, the conductivity is calculated as a weighted sum of the fast species with D comp1 (93%) and the slow species with D comp2 (7%) by σ NMR = (D comp1 n comp1 z 2 + D comp2 n comp2 z 2 )e 2 /k B T. In this way, a σ NMR of roughly 20 mS cm −1 is obtained, which is in good agreement with the in-plane conductivity obtained from EIS. We therefore conclude that (i) most of the lithium ions contribute to the in-plane conduction, and (ii) PFG NMR probes the same long-range transport phenomenon as EIS. Our findings are consistent with lithium ions being the dominant charge carriers, and possible proton or hydroxide conduction is judged to be minor.

■ CONCLUSIONS
In summary, we elucidated the exfoliation of Li 0.8 Sn 0.8 S 2 in water and determined the chemical composition of the restacked Li-TS material, which is clearly distinct from the parental Li 0.8 Sn 0.8 S 2 and SnS 2 . The number of lithium ions is severely decreased, and spectroscopic analysis suggests the incorporation of hydroxyl groups into sulfur vacancies during exfoliation. We established a 3D structural model based on XRPD and PDF analyses for the highly disordered restacked Li-TS. The structure forms through turbostratic stacking of layers, where the relative orientation of neighboring layers can be classified as akin to either Li 0.8 Sn 0.8 S 2 (C19, CdCl 2 -type) or SnS 2 (C6, CdI 2 -type), depending on the local presence or absence of lithium ions. Upon hydration, the increase in the stacking distance can be explained by the incorporation of water coordinating the lithium ions. However, a fraction of material remains in the dry state or contains lithium-depleted SnS 2 -type regions. The presence of these different phases gives rise to different lithium ion hydration states between the layers. By impedance spectroscopy, we show an increase of lithium ion conductivity over several orders of magnitude upon hydration and a preferred conduction in the inplane direction of the layers. This increase is presumably caused by the screening of the interaction of the lithium ions with the anionic Sn−S layers by the incorporated water molecules enhancing the lithium ion mobility. Furthermore, by PFG NMR, we show that the high conductivity upon hydration results from mainly co-diffusion of water and lithium ions. PFG NMR was applied to follow the lithium ion diffusion in nanosheets for the first time, and the similar, fast diffusion coefficients D Li + = 2.1 × 10 −10 m 2 s −1 and D H 2 O = 4.3 × 10 −10 m 2 s −1 at 305 K indicate a coupled movement of lithium ions and water. This points to water-assisted lithium ion conduction possibly similar to sodium-ion conduction in vermiculite. 63,83 Besides, the presence of both highly mobile and fairly immobile lithium ion species observed by PFG is rationalized by the distribution of lithium ions between hydrated and non-hydrated states in restacked Li-TS.
In conclusion, despite the differences in composition and a high degree of disorder, the fast water-assisted lithium ion conduction observed in the parental Li 0.8 Sn 0.8 S 2 is persevered in Li-TS after restacking. Moreover, the exfoliation process allows the fabrication of large oriented thin films exhibiting an even faster lithium ion conduction of up to 47 mS cm −1 in the in-plane direction of the layers. Thus, our study shows that alkali cations in restacked 2D Li-TS nanosheets are highly mobile depending on their hydration state and sheds light on the role of waterassisted lithium ion transport in this 2D system.
Additional powder X-ray diffraction data; PDF data; explanations and details on refinement models, thermal analysis, elemental analysis, ICP and EDX measurements and SEM micrographs, and additional impedance measurements; and PFG NMR data (PDF)

■ AUTHOR INFORMATION
Corresponding Author gratefully acknowledges support from BASF. We acknowledge DESY (Hamburg, Germany), a member of the Helmholtz Association HGF, for the provision of experimental facilities. Parts of this research were carried out at beamline P02.1. Open access funded by Max Planck Society.