Introduction

Demands for large-scale rechargeable batteries for electric vehicles (EVs) and energy storage systems are increasing interest in ultimately safe all-solid-state lithium batteries (ASSLBs) using inorganic solid electrolytes (SEs)1,2,3,4. A composite-type ASSLB, in which the electrode layer has a three-dimensional (3D) structure consisting of particulate mixtures of active materials, SEs and conductive additives, has attracted much attention as being close to commercialization in the future5. To date, promising performance has been demonstrated mostly by composite-type all-solid-state rechargeable batteries with sulfide SEs2,3, which is attributed to several important characteristics. First, sulfide SEs with high ionic conductivities of 10−3–10−2 S cm−1 have been developed, including a thio-lithium superionic conductor (thio-LISICON, Li3.25Ge0.25P0.75S4, 2.2 × 10−3 S cm−1)6, glass–ceramic 70Li2S–30P2S5 (Li7P3S11, 3.2 × 10−3 S cm−1)7 and Li10GeP2S12 (1.2 × 10−2 S cm−1)2. Second, ductile sulfide SEs can make intimate contact with active particles, providing good ionic conduction pathways by simple cold pressing3,8. This is in contrast to “hard” oxide SEs such as garnet-type Li7La3Zr2O129. Electrodes using the oxide SEs need to be sintered in order to make two-dimensional (2D) contacts with active particles rather than point contacts, which inevitably forms an undesirable Li+ ion barrier layer at the interface, resulting in failure during operation of the battery10,11.

In order to develop ASSLBs that can eventually compete with conventional lithium-ion batteries, several issues should be addressed. First, because the Ge included in LGPS is expensive, highly conductive alternative SEs should be further developed. Second, without a metal oxide coating such as Li4Ti5O1212, LiNbO32, Li–Si–O13, or Al2O314, sulfide SEs are not compatible with conventional layered or spinel oxide cathodes such as LixMO2 (M = Co, Ni, Mn). This can be attributed to the intrinsically low oxidation onset potential of the sulfide SEs (~3 V)15 and incompatibility between LiMO2 and the sulfides12,16. Third, 3D design of the electrode layer is critical for the performance of ASSLBs. In the configuration of lithium-ion batteries, numerous studies on the design of the electrode have focused on how to achieve good electronic conduction pathways, short Li-ion diffusion pathways in active particles by downsizing and so on4. These approaches are based on the fact that liquid electrolytes can readily wet the active materials and thus transport of Li+ ions in LEs is much faster than solid-state Li+ ion diffusion and electronic conduction within the active materials. In ASSLBs, however, the solid nature of SEs adds one more important consideration in designing composite electrodes, which is ionic conduction through the percolated SE particles and at the interfaces between active materials and SEs. More creative approaches to distributing SEs homogeneously in order to ensure good percolation and favorable contact with active particles will advance the success of ASSLBs. For example, adding a thin SE coating by laser evaporation has been demonstrated to improve the performance of ASSLBs significantly8,17.

In choosing alternative cathode materials for ASSLBs, elemental sulfur has been considered due to its high theoretical capacity (1675 mA h g−1)18,19. Sulfur also does not suffer from polysulfide dissolution with SEs18,19,20,21. However, the extremely low conductivity and severe volume change during charge and discharge (theoretically ~ 179% for S + 2Li+ + 2e− ↔ Li2S) are still problematic. Although extensive research efforts focused on intercalation/deintercalation of Li in the layered transition metal dichalcogenides in the 1970s22, transition metal sulfides were eventually not commercialized. They operated at lower voltages than the competing LixMO2, initially required a lithiated anode, had a compatibility issue with the liquid electrolytes and so on23,24. Conversely, when it comes to ASSLBs using sulfide SEs, the transition metal sulfides can be attractive candidates because their mild operating voltages (less than ~3 V) are suitable for sulfide SEs, they have potentially good compatibility with sulfide SEs and they exhibit a highly reversible intercalation reaction with Li. In particular, promising performance of ASSLBs has been reported using TiS225,26, which can be attributed to the aforementioned advantages and metallic nature of TiS2. One of the remaining challenges for using transition metal sulfides for ASSLBs is that their limited capacity is not enough to compensate for the low working voltage. In a previous report by Lee et al., decreasing the particle size of TiS2 from ~10 to 0.1 μm by ball-milling was demonstrated to increase the power density of ASSLBs25. The underlying mechanism is thought to be associated with the shortened Li+ ion diffusion length in the TiS2 and the increased interfacial contact area. Considering that the SEs are particles, there should be opportunities for further improvements by size control of the SE and better mixing of SE and TiS2 in terms of homogeneous distribution and intimate contact. Some control of the TiS2 and SE interface also may affect the electrochemical behavior.

In this study, we demonstrate an abnormally increased capacity without loss of the rate capability by using a scalable simple mechano-chemical process for fabricating an interfacial architecture consisting of TiS2 and Li2S–P2S5 SE for ASSLBs. Combined analyses by ex situ X-ray diffraction (XRD), field-emission scanning electron microscopy (FESEM), detailed high-resolution transmission electron microscopy (HRTEM) with energy dispersive spectroscopy (EDS), ex situ Raman spectroscopy, ex situ X-ray photoelectron spectroscopy (XPS) and various other electrochemical techniques show how the microstructure of electrodes results in the extra capacity while retaining good electronic and ionic conduction pathways.

Results and Discussion

We prepared an electrode by controlled ball-milling of both TiS2 and SE particles, denoted as BMe (ball-milled electrode). Because TiS2 is metallic, carbon additives are not needed. As a counter reference, another electrode was prepared by manual mixing of TiS2 and SE particles, denoted as HMe (hand-mixed electrode). A Li3PS4 (or 75Li2S–25P2S5) glass-ceramic SE with a conductivity of 1.0 × 10−3 S cm−1 at 30°C was selected because it is free from unreacted Li2S, which may disturb the interpretation of the capacity (Figure S1 (Supplementary Information)).

Figure 1a shows the FESEM images of the HMe and BMe powders and their corresponding elemental maps of Ti (red) and P (green) with EDS. As seen in the image of HMe, the regions of TiS2 and SE particles with sizes of a few to tens of micrometers are clearly separated, whereas there is no clear separation observed in the BMe images. The particles in the BMe are much smaller than those in the HMe. Also, the signals of both Ti and P are homogeneously distributed in BMe, suggesting better mixing and more favorable contacts between TiS2 and SEs. Figure 1b compares the XRD patterns of HMe and BMe. Three features are noticeable in the BMe as compared to the HMe. First, the intensity of the main peaks for TiS2 is significantly decreased. This result indicates not only reduced crystallinity25 but also a possible consumption of TiS2 by the reaction between TiS2 and Li3PS4 SE during ball-milling. The crystallite size in the [001] direction using the Scherrer equation is ~37 nm27. Second, the characteristic peaks for the SEs (the carat marks, ∧) are still visible, implying that the ionic conduction pathways are intact. The slightly decreased intensities of the SE peaks (denoted as ‘∧’ in the inset) in the BMe supports the possible reaction between TiS2 and SE during the ball-milling process. Third, the position of the TiS2 peaks after ball-milling is negatively shifted, indicating partial lithiation of TiS228,29. At this point, the source of the lithium should be the SE, suggesting that there must be some reaction between TiS2 and SE during ball-milling.

Figure 1
figure 1

(a) FESEM images and corresponding elemental maps with EDS (red: Ti, green: P) of electrode powders prepared by hand-mixing (HMe) and by ball-milling (BMe) TiS2 and solid electrolyte (SE). (b) XRD patterns of HMe and BMe.

Full size image

The all-solid-state cells of TiS2–SE/SE/Li0.5In employing HMe and BMe were cycled between 1.5 and 3.0 V and between 1.0 and 3.0 V at 50 mA g−1 at 30°C. The capacities are based on the weight of TiS2. The first two discharge–charge voltage profiles are represented in Figure 2. Two electrodes of BMe with ball-milling times of 3 and 9 min are also compared. The first discharge capacity of HMe almost reaches the theoretical capacity of TiS2 (239 mA h g−1, TiS2 + Li+ + e− ↔ LiTiS2) between 1.5 and 3.0 V in Figure 2a. After ball-milling for 3 min (Figure 2b), a slight increase in capacity and a change of voltage shapes are seen. Upon further ball-milling for 9 min (Figure 2c), the changes are significant. Although the first discharge capacity is almost the same, the electrode surprisingly exhibited significantly increased capacity at the first charge and during the subsequent cycle. Also, the voltage curves become smoother. Table 1 summarizes the first cycle capacities and Coulombic efficiency of HMe and BMe. When ball-milling is performed for 9 min, the first charge capacity is almost doubled from 220 mA h g−1 to 416 mA h g−1, which far exceeds the theoretical capacity. Also, the couloumbic efficiency is increased from 93% to 158%. A couloumbic efficiency of more than 100% indicates that the amount of extracted lithium is more than the inserted amount. The BMe with the 9-min ball-milling (denoted as BMe-9-min), which is subjected to charge (delithiation) first, as seen in the inset in Figure 2c, affords 151 mA h g−1. This value corresponds perfectly with the difference between the first charge capacity and discharge capacity (416 − 263 = 153 mA h g−1) of the same cell under normal cycling conditions (discharge or lithiation first). From the position of the (001) peak of LixTiS2 (15.3°) in Figure 1b, d spacing value is 5.78 Å. The ex situ XRD results of the HMe cell in Figure S2 describes the relationship between the degree of lithiation of TiS2 (x in LixTiS2) and d28. When 5.78 Å of d is applied to the fitted curve in Figure S2c, x in LixTiS2 for the BMe sample becomes 0.05. Complete delithiation from Li0.05TiS2, however, accounts for only ~12 mA h g−1 at maximum, which is far lower than the first charge capacity of 151 mA h g−1 shown in the inset of Figure 2c. Also, the origin of the abnormally increased reversible capacity of 177 mA h g−1 (first charge capacity − theoretical capacity = 416 − 239) of the BMe-9-min cell as compared to that of HMe is questionable, which will be discussed in detail later.

Table 1 Electrochemical results HMe and BMe electrodes
Full size table
Figure 2
figure 2

First two discharge–charge voltage profiles of HMe and BMe at 50 mA g−1.

Left and right column show the curves between 1.5–3.0 and 1.0–3.0 V, respectively. The inset in c shows the first-cycle voltage profiles when charge (delithiation) is carried out first.

Full size image

Decreasing the lower cutoff voltage down to 1.0 V, the changes in electrochemical behavior as a function of ball-milling time is more dramatic. For HMe, no noticeable reaction occurs in the 1.0–1.5 V range (Figure 2d). As the ball-milling time increases (Figures 2e and 2f), however, a new long plateau at ~1.3 V develops, resulting in a significant increase in the reversible capacity up to ~840 mA h g−1. This value is ~3.6 times higher than the theoretical capacity of TiS2. Another counter reference electrode was prepared by manually mixing ball-milled TiS2 for 9 min and SE particles, denoted as HMe-BM-TiS2. The XRD pattern of HMe-BM-TiS2 exhibited a lower intensity for TiS2 but no shift in peak positions (Figure S3a). Also, the capacity is almost the same as that of HMe (Figure S3b). This control experiment strongly suggests that some reaction between TiS2 and SE during ball-milling is the key to explaining the extra capacity in the BMe samples. The results for the BMe samples were obtained using a weight ratio of TiS2 to SE of 0.5. This condition might be regarded as unbalanced in terms of energy density based on the total weight of the electrode. For a weight ratio of 1.0 (TiS2/SE), however, the evolution of the abnormally increased capacity with increased ball milling time is also confirmed, as seen in Figure S4. For example, ball-milling of a 1:1 weight ratio of TiS2 and SE for 27 min resulted in first charge capacities of 442 mA h g−1at 1.5–3.0 V and 909 mA h g−1 at 1.0–3.0 V. These capacities are equivalent to 221 mA h (g of electrode)−1 and 454 mA h (g of electrode)−1, respectively. Note that HMe with even higher weight ratio of TiS2 to SE of 2.0 produced only 140 mA h (g of electrode)−1 ( = 210 mA h g−1). Therefore, it is fair to say that the controlled ball-milling of TiS2 and SE together is a viable method for significantly increasing the capacity and energy density of ASSLBs.

In order to analyze the electrochemical behaviors more closely, the voltage curves in Figure 2 are represented by differential capacity plots in Figure 3. As the ball-milling time increases for the 1.5–3.0 V range, it is seen that new reactions between ~1.5 and 1.7 V (denoted as α/α′) and between ~2.3 and 3.0 V (denoted as β/β′) develop. In the case of 1.0–3.0 V, not only the α/α′ and a β/β′ redox couples but also huge redox peaks (denoted as γ/γ′) are added. Overall, comparing HMe and BMe-9-min, it is postulated that the reaction from TiS2 is overlapped by unknown reactions assigned as α/α′, β/β′ and γ/γ′ that might have originated from the phase formed by a reaction between TiS2 and SE. Polarization at several states of charge for the first two cycles was recorded by galvanostatic intermittent titration techniques (GITT) and the results are shown in Figure S5. Corresponding well with the differential capacity analyses in Figure 3, relatively larger polarizations (indicated by short dashed circles) are seen in the voltage ranges responsible for α/α′, β/β′ and γ/γ′. This observation implies a very different reaction mechanism for the additional Li+ storage from the intercalation/deintercalation of Li in TiS230.

Figure 3
figure 3

First two discharge–charge differential voltage profiles of HMe and BMe obtained from the voltage profiles in Figure 2.

Full size image

In an initial attempt to elucidate the reaction mechanism responsible for the extra capacity of BMe, ex situ XRD patterns were obtained and the results are shown in Figure S6. The (001) peak of LixTiS2 is shifted initially from 15.3° to 14.4° by discharge down to 1.5 V and held at the same position down to 1.0 V. Charging up to 3.0 V from 1.5 V and from 1.0 V results in a positive shift, affording a (001) peak at 15.6°, which corresponds well with the peak for fully delithiated TiS2. However, the XRD patterns obtained do not provide much information about the abnormally increased capacity, implying that the phase responsible for the extra capacity is not crystalline.

A detailed microstructural analysis of the BMe was carried out by HRTEM. An approximately 100-nm-thick specimen of the BMe was prepared by focused ion beam (FIB) sectioning and the results are shown in Figure 4a. An SE region of ~2 μm in size and unique textures surrounding the SE region are seen. The textured region is represented by a dark-field scanning transmission electron microscopy (STEM) image and corresponds to the elemental maps of P and Ti with EDS in Figure 4b. Noticeably, both elements are homogeneously distributed, indicating that TiS2 may be mixed with SE (Li3PS4) on a nanoscale level or that it extensively reacts with SE. From the bright-field TEM image in Figures 4c, it is observed that the unique textures are composed of layered bundles in an amorphous matrix. The HRTEM images in Figures 4d and 4e clearly show the lattice fringes of the layered regions. The spacing is measured to be ~5.8 Å, which matches nicely with the d spacing values obtained from the (001) peak of Li0.05TiS2 in BMe in the XRD pattern (2θ = 15.3°, d = 5.78 Å) (Figure 1b). The elemental analysis in the amorphous region on the nanoscale level was carried out by EDS, as shown in Figure 4f. Importantly, the amorphous phase includes not only P and S but also Ti. Along with the XRD results in Figure 1b, the elemental analysis by HRTEM with EDS evidences the partial reaction between TiS2 and Li3PS4 SE during ball-milling. The different composition in the two regions in Figure 4f (LixTi1.00P1.34S3.88 and LiyTi1.00P1.14S2.47), however, indicates inhomogeneity of the partial reaction. It is not surprising that the ball-milling condition is controlled so that not all of the TiS2 and Li3PS4 SE are consumed.

Figure 4
figure 4

(a) FESEM image of FIB-sectioned BMe. (b) Dark-field STEM image and corresponding EDS signals of P and Ti of the rectangle region in (a). (c) Bright-field TEM images of the rectangular region in (a). (d and e) HRTEM images of enlarged region in (c). (f) HRTEM image showing the crystalline Li0.05TiS2 and the amorphous region. Chemical compositions of Ti, P and S of the rectangular regions were obtained by EDS signals.

Full size image

The amorphous Li-Ti-P-S phase confirmed by HRTEM with EDS can explain not only the extra capacity but also the couloumbic efficiency of more than 100% of BMe at the first cycle, as observed in Figure 2 and Table 1. The latter shows that lithium in the newly formed product may be extracted in the first charge (delithiation). In an attempt to further identify the amorphous Li-Ti-P-S phase, the ex situ Raman spectra at different discharged and charged states were obtained, as revealed in Figure S7. Unfortunately, the results exhibited signatures from LixTiS231 and PS43−7 only. Also, it should be noted that any signal related to sulfur-rich LiyTiSx, such as LiyTiS3 and LiyTiS432 and sulfur or LixS (0 < x ≤ 2) was not observed. Overall, the inhomogeneous composition of the amorphous Li-Ti-P-S phase (as displayed in Figure 4f), the lack of data from the ex situ XRD and Raman experiments along with the extreme diversity of transition-metal thiophosphates33 and lack of spectroscopic database restrict the clearer identification of the amorphous Li-Ti-P-S phase.

From the ex situ XRD, HRTEM with EDS and ex situ Raman analyses, it is confirmed that the composite electrode comprised crystalline Li0.05TiS2, Li3PS4 SE and amorphous Li-Ti-P-S, which accounts for the unusual extra capacity. As another scenario for the extra capacity, it should not be ruled out that a thin region of Li3PS4 SE in contact with the conductive Li0.05TiS2 may be electrochemically activated and participate in additional Li+ storage. As a control sample, a mixture of nanosized carbon (super P) and SE with a weight ratio of 0.1 (carbon/SE) was ball-milled under the same condition as for BMe (at 2000 rpm for 9 min). Note that carbon will not react with SE, in contrast to the partial reaction between TiS2 and SE. The first two discharge-charge voltage profiles of the carbon-SE electrode are represented in Figure S8. Very small capacity of 1–2 mA h (g of carbon)−1 implies negligible contribution of Li3PS4 SE to the abnormally increased capacity.

In order to clarify the mechanism, ex situ XPS analysis on the BMe sample was carried out before the cycle and after discharge at 1.5 and 1.0 V. In order to obtain bulk information, the specimens were sputtered with Ar ions for 60 min. The results are represented in Figure 5. As compared to the pristine BMe, the signal of Ti 2p is slightly positively shifted during discharge at 1.5 V. This subtle change agrees with the reference for TiS2 + Li+ + e− → LiTiS234. Ti in the amorphous Li-Ti-P-S can also contribute to this change. The signals of P 2p and S 2p remain almost the same after discharge at 1.5 V. In the process of further discharge at 1.0 V, several important changes are highlighted. First, a slightly negative shift of the Ti 2p signal is observed. Second, negative shifts of the signals of both P 2p and S 2p are noticeable. These results present strong evidence that the reduction center in the range 1.0–1.5 V in BMe is not only Ti but also P and S, which is in agreement with the observation from HRTEM with EDS presented in Figures 4d, 4e and 4f that the amorphous phase includes Ti, P and S. It was reported that the extra capacity observed in a Chevrel-phase CuxMo6S8−y electrode in ASSLBs at high temperature is related to the diffusion of Cu into the Li2S–P2S5 SE35. On a similar basis, some phase formed by a reaction between Cu and SE may be related to the extra capacity as well. The use of LiTi2(PS4)3 for lithium-ion battery applications, reported first by Goodenough et al., shows a large capacity in the liquid cell1,36. Recently, our group reported that LiTi2(PS4)3 in an all-solid-state cell can exhibit a first discharge capacity of ~1050 mA h g−1 in 1.2–3.5 V by a conversion reaction37. In this context, it is not surprising to learn that a significant contribution to the huge extra capacity occurs by formation of amorphous Li–Ti–P–S phases in BMe in this work. In order to investigate the electrochemical reactivity of the amorphous Li-Ti-P-S phase in BMe further, two amorphous Li-Ti-P-S powders to mimic the one in the BMe sample were prepared by ball-milling TiS2 and SE powders until the XRD peaks for TiS2 and SE disappear as seen in Figure S9a. The amorphous powders prepared from TiS2 and SE with the weight ratio of 0.5 and 1.0 (TiS2/SE) were denoted as a-BMe1 and a-BMe2, respectively. The a-BMe1 sample exhibits two orders of magnitude lower electronic conductivity (1.3 × 10−3 S/cm) than that of the BMe sample (7.0 × 10−1 S/cm), which indicates that almost all of TiS2 might be consumed for the reaction with the SE. In order to examine the electrochemical reactivity of the a-BMe samples, the a-BMe powders were hand-mixed with the Li3PS4 GC SE and conductive carbon additives (a-BMe:SE:super P = 20:30:3 weight ratio) and cycled in the range 1.0–3.0 V. Figure S9b represents the first two discharge-charge voltage profiles. Two features are in common in the electrochemical reactivity of the amorphous phase (a-BMe) in Figure S9b and the BMe in Figure 2f (BMe – 9 min). First, two reaction regions at ~2.3 V and ~1.3 V are noticeable in discharge (lithiation). Second, the amount of extracted lithium (charge) exceeds the one of inserted lithium (discharge) at first cycle. From the two common characteristics, it can be postulated that the amorphous powders prepared by the control experiment (a-BMe in Figures S9) possibly represent the amorphous Li-Ti-P-S phase in the BMe sample (Figure 2f). Most importantly, the a-BMe samples show the abnormally high capacity, which strongly supports that the amorphous Li-Ti-P-S phase in the BMe contributes to the extra capacity.

Figure 5
figure 5

(a) Discharge voltage curve of the BMe for ex-situ XPS experiment. (b) Ex-situ XPS results of Ti 2p, P 2p and S 2p before cycle and after discharge at 1.5 and 1.0 V.

Full size image

Figure 6a represents the Nyquist plots of HMe and BMe cells discharged to 2.1 V at the first cycle, which was obtained by electrochemical impedance spectroscopy (EIS). The EIS spectra are composed of RSE at high frequency, which corresponds with the resistance of the SE layer and Warburg tails at low frequency, which explains Li+ diffusion into the electrode materials13. Importantly, both signals include negligibly small resistance at mid-frequency, which can be assigned to the interfacial charge transfer resistance from TiS2/SE and LixIn/SE. This result highlights an advantageous feature of sulfide-based electrode materials in terms of not only mild operating voltage but also compatibility with sulfide SEs. In sharp contrast, significant charge transfer resistance develops in solid-state cells containing LiCoO213. Also, it is noticeable that both HMe and BMe show similar impedances, which suggests that controlled ball-milling does not degrade the kinetics in composite electrodes. The cycle performances of HMe and BMe at 50 mA g−1 are represented in Figure 6b. After 60 cycles, 95% of the initial charge capacity is retained in the range 1.5–3.0 V for both BMe and HMe. The significantly increased capacity of BMe in the range 1.0–3.0 V is offset by deterioration in cycle performance (76% of capacity retention). The large hysteresis in voltage profiles between discharge and charge as seen in Figures 2 and 3 strongly indicates the lithiation reaction in the range 1.0–1.5 V is likely a conversion-type reaction30,37. Large volume change and structural reorganization during lithiation-delithiation for the conversion reaction may explain the degradation of BMe cycled in the range 1.0–3.0 V37. The charge capacities of HMe and BMe as a function of cycle number with increasing current density in 1.5–3.0 V are represented in Figure 6c. It should be emphasized that the rate performance of BMe is only slightly poorer than that of HMe. This result is in line with the EIS results shown in Figure 6a.

Figure 6
figure 6

(a) Nyquist plots of HMe and BMe at first cycle. The RSE term corresponds with the resistance of the SE layer. (b) Cycle performances and (c) rate capabilities of HMe and BMe. The voltage range in c is 1.5–3.0 V. The open and closed symbols in b are discharge and charge capacities, respectively.

Full size image

From the XRD results shown in Figure 1b and the FESEM image shown in Figure 4a, it is confirmed that the SE domain is active after ball-milling. In order to evaluate Li+ ion conduction in the electrode layer, Li+ ion conductivity was measured using a specially designed electron-blocking symmetric cell of Li/SE/BMe/SE/Li by DC and AC methods (Figure S10). A slightly smaller ionic conductivity for BMe (8.7 × 10−5 S cm−1) than for HMe (1.4 × 10−4 S cm−1) was recorded (Table 2). The electronic conductivities of HMe and BMe pellets measured by the van der Pauw method38 affords 1.2 × 101 and 7.0 × 10−1 S cm−1, respectively. The considerably lowered electronic conductivity of BMe also supports the claim that an appreciable amount of the metallic TiS2 is consumed by the reaction with SE during the controlled ball-milling process. Importantly, however, the electronic conductivity value of 7.0 × 10−1 S cm−1 indicates that the remaining LixTiS2 domains in the BMe provide the electronic conduction pathways. These ionic and electronic conductivity measurements demonstrate that the controlled ball-milling of TiS2 and SE is an effective way to maximize the capacity while maintaining the conduction pathways of the electrode.

Table 2 Ionic and electronic conductivity of TiS2–SE composite electrodes
Full size table

From all the results and analyses so far, a schematic diagram representing the microstructure of BMe and the role of each component is depicted in Figure 7. The controlled ball-milling of TiS2 and SE can not only downsize the particles and make intimate contacts between TiS2 and SE, but it also produces amorphous phases of Li–Ti–P–S that contributes to the abnormally increased capacity while retaining ionic and electronic conduction pathways. This strategy can be effectively employed to other electrode systems for ASSLBs.

Figure 7
figure 7

Schematic diagram representing the microstructure of BMe in ASSLB.

Electron and Li+ ion conduction is shown. The role of each component is explained in the figure.

Full size image

In conclusion, the controlled ball-milling of a sulfide-active material (TiS2) and an SE (Li2S–P2S5) for the electrode was demonstrated to achieve considerably increased capacity in ASSLBs without loss of the ionic and electronic conduction pathways. The ball-milled electrode exhibited the first charge capacity of 416 mA h g−1 in 1.5–3.0 V at 50 mA g−1 at 30°C, with excellent capacity retention of 95% after 60 cycles. The capacity was further increased to ~840 mA h g−1 in 1.0–3.0 V offset with a somewhat poorer cycle performance. The origin of the abnormally increased Li+ storage appeared to be associated with the formation of an amorphous Li–Ti–P–S phase during the controlled ball-milling process. We believe that these results are important in understanding the interfacial phenomena and in the development of high-performance ASSLBs.

Methods

Preparation of Li2S–P2S5 SE

75Li2S–25P2S5 (or Li3PS4) glass-ceramic SE powders were prepared by mechanical milling and subsequent heat treatment. A mixture of 2 g each of Li2S (99.9%, Alfa Aesar) and P2S5 (99%, Sigma Aldrich) were mechanically milled at 500 rpm (8.3 s−1) for 10 h at room temperature using a planetary ball mill (Pulverisette 7PL; Fritsch GmbH) with a ZrO2 vial (80 mL) and 115 g of ZrO2 balls (5 mm in diameter). The obtained glass powders were put into a glass ampoule and sealed under vacuum (≤40 Pa). The sealed ampoule was subjected to heat treatment at 243°C for 1 h. Measurements by a Li+ ion blocking cell (Ti/SE/Ti) with AC and a Li+ ion non-blocking cell (Li/SE/Li) with DC afforded an ionic conductivity of 1.0 × 10−3 S cm−1 at 30°C.

All-solid-state cells

Composite electrode materials were prepared by manually mixing the TiS2 and SE powders (denoted as HMe) or by ball-milling 30 mg of a TiS2/SE mixture using a Thinky Mixer (AR-100, THINKY Inc.) with a plastic vial (50 mL) and three ZrO2 balls (5 mm in diameter) at 2000 rpm (33 s−1, denoted as BMe) for 9 min. The weight ratio of the TiS2/SE was 0.5. Partially lithiated indium (Li0.5In, nominal composition) powders prepared by mixing In (Aldrich, 99%) and Li (FMC Lithium Corp.) powders were used as the counter and reference electrodes. SE pellets were formed by pressing 150 mg of the 75Li2S–25P2S5 glass-ceramic SE at 74 MPa. TiS2 composite electrode materials (5 mg) were then carefully spread on the top of the SE layer and the cell was pelletized by pressing at 370 MPa for 3 min. Finally, 100 mg of the prepared Li0.5In powders were attached to the back SE face at 370 MPa for 3 min. All pressings were carried out in a polyaryletheretherketone (PEEK) mold (diameter = 1.3 cm) with Ti metal rods as current collectors for both the working and counter/reference electrodes. All the processes for preparing the SEs and fabricating the all-solid-state cells were performed in an Ar-filled dry box. The galvanostatic discharge–charge cycling of the all-solid-state cells was performed at 50 mA g−1 at 30°C. The capacity is based on the weight of TiS2. The EIS study was performed using an Iviumstat (IVIUM Technologies Corp.). The AC impedance measurements were recorded using a signal with an amplitude of 10 mV and a frequency range from 500 kHz to 5 mHz. The cells were discharged to 2.1 V (vs. Li/Li+) at 50 mA g−1 and a constant voltage of 1.8 V was applied until the current decreased to 10 mA g−1. Then, the cell was allowed to rest for approximately 3 h.

Characterization of materials

For the XRD analyses, a specially designed cell was used, in which the air-sensitive materials such as the TiS2–SE electrode powders or disassembled all-solid-state cell pellets were placed on a beryllium window and hermetically sealed inside the Ar-filled dry box. Then, the prepared XRD cell was mounted on a D8-Bruker Advance diffractometer equipped with Cu Kα radiation (1.54056 Å). All the XRD patterns were recorded at 40 kV and 40 mA using a continuous scanning mode at 1.5° min−1. For the HRTEM and EDS analyses, the TiS2–SE electrode pellet was sectioned by using a 30 keV Ga+ ion beam. A cross-sectional sample with dimensions of 5 × 5 × 0.1 μm3 was recovered by the lift-out technique and then attached to the Cu TEM grid using a manipulating probe (Copper 3-Post, Omniprobe). Further thinning of the lift-out sample was performed by milling parallel to the cross-sectional plane with low ion beam currents (47 pA) until a final sample thickness of 80–100 nm was achieved. The HRTEM images and EDS elemental mapping were obtained using JEM-2100F (JEOL) with 200 kV and 0.105 mA. For the ex situ XPS measurements, the electrode samples collected by disassembling the all-solid-state cells in an Ar-filled dry box were directly transferred to another Ar-filled dry box connected to the XPS equipment (SIGMA PROBE, Thermo Scientific Inc.) without any exposure to air. Characterization with ex situ XPS was performed with a monochromatic Al Kα source (1486.6 eV), 100 W, 15 kV and 6.7 mA. The base pressure in the XPS analysis chamber was less than 1 × 10−7 Pa. XPS spectra were collected using a constant analyzer energy mode with a pass energy value between 30 eV. The step size was 0.1 eV.