Introduction

Recently, interest in carbon neutrality has promoted a reassessment of energy devices, motivating a shift from gasoline vehicles to battery electric vehicles (BEVs) [1]. For this purpose, all-solid-state lithium batteries have attracted significant attention owing to their high theoretical energy density, and the implementation of bipolar structures [2] and thicker films [3] has been considered. To utilize all-solid-state batteries as an energy source for BEVs, sulfide solid electrolytes (SEs) with lithium-ion conductivity comparable to or higher than that of organic electrolytes have been developed to ensure a high output density [4,5,6,7]. Disadvantageously, sulfide SEs are unstable in atmospheric moisture, which leads to decreased lithium-ion conductivity and the evolution of toxic H2S gas, thereby increasing process control costs [8,9,10,11,12,13,14]. In addition, sulfide SEs are unstable on oxide-based cathodes, resulting in a shorter cycle life owing to material degradation and lower power output owing to increased interfacial resistance between the SE and cathode active material [15,16,17,18,19]. To overcome these issues, halide SEs have become an attractive alternative. In 2018, Asano et al. reported that Li3YCl6 and Li3YBr6 have high ionic conductivities (0.51 and 1.7 mS/cm, respectively). In addition, all-solid-state batteries fabricated with these halide SEs and an oxide cathode active material exhibited stable charge–discharge cycling [20]. As conventional sulfide SEs do not have a sufficiently wide oxidation potential window, the oxide cathode must be coated with another material, such as LiNbO3. In contrast, halide SEs are thought to have inherent stability against oxidation [21, 22]. However, metal chlorides can undergo side reactions, such as reacting with moisture to release HCl gas. In 2019, Li et al. reported that Li3InCl6, which was synthesized in aqueous solution, showed high water resistance in atmospheric environments [23]. Subsequently, many studies have focused on improving the ionic conductivity of Li3InCl6 by doping with different elements [24,25,26,27,28]. Nevertheless, a precise understanding of the reactivity of halide SEs with H2O is required for rational compositional or structural design. Based on in-depth experiments and observations, Wang et al. discussed the reaction mechanism responsible for the H2O sensitivity of halide SEs [29]. They demonstrated that when Li3InCl6 comes into contact with atmospheric moisture, it becomes Li3InCl6・2H2O, which eventually deliquesces to form an acidic liquid. However, there have been very few reports on gas generation and thermal (exothermic/endothermic) behavior, even though it is essential for the production of halide SEs for commercial applications. Very recently, Chen et al. have reported chemical stability under ambient air and dry room conditions for chloride solid electrolytes, and confirmed irreversible decomposition of Li3YCl6, while Li3InCl6 show less reactive against water and form stable hydrate which can be recovered to Li conductive Li3InCl6 phase by heating at 260 °C under vacuum[30].

In this study, the gas evolution behavior of typical halide SEs, Li3YCl6 and Li3InCl6, during heating in a low-moisture and inert atmosphere was investigated using temperature-programmed desorption mass spectrometry (TPD-MS). In addition, the amount of HCl gas evolved was evaluated using a detector tube gas analyzer. The pH of the SEs was also investigated upon exposure to H2O in air or in an atmosphere with a dew point of -30 °C. Furthermore, first-principles calculations were used to examine the stability of the SEs in the presence of moisture.

Methods

Experimental procedures

Synthesis and characterization of SEs

SEs were synthesized using mechanochemical methods. A stoichiometric mixture of LiCl (99.9%, Sigma-Aldrich, St. Louis, MO, USA) and YCl3 (99.9%, Sigma-Aldrich, St. Louis, MO, USA) or InCl3 (99.99%, High Purity Chemical Laboratory, Saitama, Japan) was added to a 45 mL ZrO2 pot. The mixture was milled with 34 ZrO2 balls (diameter = 8 mm) using a planetary ball mill (P-7 classic-line, Fritsch Japan K.K., Tokyo, Japan) at 500 rpm for 1 h. and then allowed to rest for 15 min. For Li3InCl6, the ball-milled sample was subsequently heat treated at 300 °C for 3 h to enhance Li ion conductivity according to literature [24].

The synthesized samples were characterized using X-ray diffraction (XRD; MiniFlex 600, Rigaku Corp., Tokyo, Japan). The ionic conductivity was measured at 298 K using the AC impedance method (VSP electrochemical analyzer, BioLogic, Grenoble, France) at an applied voltage of 10 mV and frequencies of 102–107 Hz. For these measurements, pellets (diameter = 10 mm, thickness = ~ 0.50 mm) were prepared by sandwiching the SE powder between stainless steel pins and compressing at 509 MPa. All procedures were performed under a dry Ar gas atmosphere maintained at a dew point of less than − 80 °C.

TPD-MS measurements

The gas evolution behavior of the synthesized SEs was evaluated using TPD-MS. For TPD-MS, an MS instrument (GC/MS QP2010Plus, Shimadzu Corporation, Kyoto, Japan) was directly connected to a heating device with a temperature controller. The concentration change at the mass number corresponding to each gas generated by the sample during heating was calculated as a function of temperature and time (Fig. S1). For 15 min before measurement, dry He (purity 99.999%, dew point <-70 °C) was used as the carrier gas at a flow rate of 50 mL/min was used to displace the gas inside the apparatus. The same flow rate was used during the measurement to maintain a low-moisture atmosphere with a dew point of less than − 60 °C. During measurement, the temperature was increased from room temperature to 500 °C at a rate of 10 °C/min.

FT-IR measurements

The vibrational states of water molecules in solid-state samples were investigated using attenuated total reflection Fourier transform infrared spectroscopy (ATR-FTIR, VERTEX 70v, Bruker Corporation, Billerica, Massachusetts, USA), where incident angle is set as 45 degrees. Samples are sealed In air-tight containers in a dry Ar gas atmosphere at a dew point of less than − 80 °C, and transferred to the analyzer. The measurements were performed under dry N2 atmosphere immediately after opening the containers in the sample chamber of the instrument.

Detector tube gas analyzer measurements

The gas evolution behavior of the synthesized SEs was also investigated using a detector tube gas analyzer. The synthesized SEs were sealed in a dry Ar gas atmosphere at a dew point of less than − 80 °C and then moved to an Ar gas atmosphere at a dew point of -30 °C and placed in a beaker (Fig. S2). Subsequently, the sample was heated to a predetermined temperature on a hot plate and maintained at that temperature for 10 min. The amount of evolved gas was determined using a detector tube gas analyzer (GC-100 S, GASTEC, Ayase, Japan) and a gas detector tube (No 14 L, GASTEC). A sealed environment was maintained to prevent the gas from dispersing during heating and standing. The amount of gas was calculated from the concentration measured using the gas detector tube and the volume of the glass container, and the results were reported as the amount of gas produced per gram based on the sample weight.

Static test of SE powder using pH test paper

The synthesized SEs were sealed in a dry Ar gas atmosphere at a dew point of less than − 80 °C. A portion of the SE powder SE sample was transferred to an Ar gas atmosphere at a dew point of -30 °C or to air (45% humidity at 25 °C) and placed on pH test paper (universal indicator paper pH 1–14, MACHEREY-NAGEL GmbH & Co. KG, Düren, Germany). In addition, to confirm the color corresponding to each pH value, the pH test papers were dipped in aqueous solutions adjusted to various pH values using ion-exchanged water, HCl, and NaOH.

The remaining portion of the SE powder was placed in an atmosphere at a dew point of -30 °C for 60 min, and the products were characterized by XRD.

Calculation procedures

Density functional theory (DFT) calculations were performed to investigate the reaction of the halide SEs with H2O. Structure models for Li3InCl6 and Li3YCl6 were referred to the the inorganic crystal structure database (ICSD), where ID numbers of Li3InCl6 and Li3YCl6 are 17,638 and 29,969, respectively. Li3InCl6 belongs to monoclinic (C2/m). while Li3YCl6 belongs to trigonal (P-3 m). Superstructure models for Li3InCl6 and Li3YCl6, corresponding to Li48In16Cl96 and Li72Y24Cl144, respectively, were prepared in accordance with a previous study [28]. This study investigated the diffusivity of Li in Li3InCl6 and Nb- and/or Zr-doped compounds.

The partial replacement of Cl ions by OH ions in Li48In16Cl96 and Li72Y24Cl96 was considered to model the reaction with H2O (as described later). After structural optimization, the total energy was evaluated using the Vienna ab initio software package (VASP) [30,31,32,33,34] with the projector augmented-wave method [35, 36] and a plane–wave basis set. The generalized gradient approximate exchange correlation function developed by Perdew, Burke, and Ernzerhof and modified for solid materials (PBEsol-GGA) was used [37]. The cut-off energy of the plane–wave basis was set to 500 eV. A 2 × 3 × 1 Monkhorst–Pack k-point grid was used for optimization calculations.

Results and discussion

Experimental results

The XRD patterns and AC impedance plots of Li3InCl6 and Li3YCl6 under dry conditions are shown in Fig. 1(a) and (b), respectively. All measurements were performed in a dry room (dew point < -30 °C) immediately after transfer from the dry box (dew point < -30 °C). Both the XRD patterns and AC impedance plots of the Li3InCl6 and Li3YCl6 samples agree well with those in previous studies [20, 28].

Fig. 1
figure 1

Characterization of the synthesized SE samples. (a) XRD patterns of SEs. Li3YCl6 was synthesized in a planetary ball mill and Li3InCl6 was synthesized by heating at 300 °C for 3 h after treatment in a planetary ball mill. (b) Cole–Cole plots of SEs (black: Li3YCl6, red: Li3InCl6)

Figure 2(a) and (b) show the TPD-MS results for H2O (m/z = 18) and HCl (m/z = 36), respectively. In addition, for comparison purposes, the profiles for both gases are shown for Li3YCl6 and Li3InCl6 in Fig. 2(c) and (d), respectively. Cl2 gas was not detected in the TPD-MS measurements. The total molar ratio of H2O to the theoretical molar amount of Cl in the SEs was 3.1 and 5.0 mol% for Li3InCl6 and Li3YCl6, respectively (Fig. 2(a)), suggesting that the SE samples either retained moisture during synthesis or adsorbed moisture from the TPD-MS measurement environment at a dew point below − 60 °C. The TPD-MS peak for H2O evolution from Li3InCl6 appeared at ~ 100 °C, equivalent to the boiling point of water. In contrast, for Li3YCl6, this peak appeared at a higher temperature (~ 140 °C). This difference may indicate differences in the chemical interactions between the SEs and trace H2O originating from the synthesis and/or analysis environments, even in a humidity-controlled atmosphere. We infer that H2O molecules are physically adsorbed on the surface of Li3InCl6 because of the agreement between the TPD-MS peak and the boiling point of H2O. In contrast, H2O penetrates the interior of the Li3YCl6 particles, resulting in H2O evolution at higher temperatures. The onset temperature for HCl evolution was much higher for Li3InCl6 (~ 220 °C) than for Li3YCl6 (~ 100 °C) (Fig. 2(b)). For Li3InCl6, no correlation is observed between the gas evolution behavior of H2O and HCl. However, for Li3YCl6, the temperatures of the H2O peak and HCl shoulder agree well (~ 140 °C). Hence, we assume that the HCl evolution reaction is triggered by H2O gas, as shown in Eq. (1):

$${\rm{L}}{{\rm{i}}_{\rm{3}}}{\rm{YC}}{{\rm{l}}_{\rm{6}}} + {\rm{ x}}{{\rm{H}}_{\rm{2}}}{\rm{O}} \to {\rm{L}}{{\rm{i}}_{\rm{3}}}{\rm{YC}}{{\rm{l}}_{{\rm{6 - x}}}}{\left( {{\rm{OH}}} \right)_{\rm{x}}} + {\rm{ xHCl}}$$
(1)

In this reaction, the Cl ions are partially replaced by OH ions. Equation (1) is also supported by the ATR-FTIR spectra shown in Fig. 3. For Li3YCl6 (Fig. 3(a) and (b)), the peaks derived from hydroxyl group are broad. In contrast, the corresponding peaks for Li3InCl6 (Fig. 3(c) and (d)) are relatively sharp, which suggests that the H2O molecules are fixed in the lattice (i.e., crystalline water). Notably, a relatively strong peak appears at approximately 1600 cm− 1 before the exposure test for Li3InCl6, indicating that a trace amount of H2O, similar to crystalline water, is retained during the synthesis stage (dew point less than − 80 °C ). A schematic of the processes occurring at relatively low temperatures is shown in Fig. 4. Note that above H2O-involved reaction was conducted in humidity-controlled environments, where samples were kept in a glove box at the dew point of dry Ar (<-80 °C), whereas the TPD-MS measurements were conducted in a dry He environment. Nevertheless, trace H2O molecules may affect the generation of HCl gas from Li3YCl6 at ~ 140 °C. Furthermore, the lack of strong interactions between Li3InCl6 and H2O indicates that the HCl evolution reaction with H2O can be controlled by selecting appropriate metal ions for Li-containing chlorides.

Fig. 2
figure 2

TPD-MS results (temperature vs. gas generation rate curve) for the synthesized SEs. (a) and (b) Rates of evolution for H2O (m/z = 18) and HCl (m/z = 36), respectively (black: Li3YCl6, red: Li3InCl6). (c) and (d) Rates of evolution for Li3InCl6 and Li3YCl6, respectively (blue: H2O (m/z = 18), green: HCl (m/z = 36))

Fig. 3
figure 3

ATR-FTIR spectra of (a, b) Li3YCl6 and (c, d) Li3InCl6. The peaks attributed to OH vibrations are shown, where γ- and δ-OH indicate stretching and bending vibrations, respectively. The black and red lines correspond to the samples before and after exposure to an atmosphere with a dew point of -30 °C for 10 min

Fig. 4
figure 4

Schematic diagram showing the form in which the solid electrolyte retains moisture from the atmosphere

Figure 5 shows the amount HCl gas evolved during detector tube gas analyzer measurements at a dew point of -30 °C as a function of temperature. This method can specify the evolution of HCl gas, sensitively and inexpensively. Similar to the TPD-MS results (Fig. 2), HCl gas evolution from Li3YCl6 was detected at a lower temperature than that from Li3InCl6. The onset temperature of HCl gas evolution for Li3YCl6 was below 100 °C, and a trace amount of HCl gas was detected at room temperature. This behavior is consistent with the hypothesis that H2O penetrates the SE powder particles, even in a humidity-controlled atmosphere. The observed onset temperature for HCl evolution is lower with the detector tube gas analyzer than with TPD-MS because of the different operating atmospheres (dew points of -30 and − 60 °C, respectively). However, with Li3InCl6, HCl gas was observed at temperature above 150 °C, which is slightly lower than the temperature obtained using TPD-MS. Note that the same results are observed for Li3YCl6 heated at 300 °C for 3 h as shown in Figure S3 in the Supporting Information, so that the HCl gas evolution behavior does not change the heat treatment after ball milling. This behavior suggests that increasing the ambient moisture content not only causes water to be retained through surface adsorption but also promotes internal chemical reactions, as in Li3YCl6. HCl gas evolution above 200 °C occurred with both Li3YCl6 and Li3InCl6, as also observed in the TPD-MS measurements. Wang et al. proposed that decomposition to InCl3 and LiCl and subsequent hydrolysis reaction leads to the formation of HCl [29]. We infer that a similar hydrolysis reaction mechanism proceeds in Li3YCl6 at temperatures above 200 °C.

Fig. 5
figure 5

Amount of generated HCl, as measured using a detector tube gas analyzer at a dew point of − 30 °C (black: Li3YCl6, red: Li3InCl6)

Figure 6(a) shows photographs of the Li3YCl6 and Li3InCl6 powders placed on pH test paper in a dry atmosphere (dew point of -30 °C) after 30 min as well as the colors of the pH test paper when dipped in aqueous solutions with various pH values. After 30 min, the pH test paper with Li3InCl6 showed little coloration, whereas that with Li3YCl6 showed a color change corresponding to acidic conditions, probably due to HCl gas evolution. These results are consistent with trace HCl gas being evolved without heating, as shown by the detector tube gas analyzer test (Fig. 5). To clarify the different gas evolution behavior of Li3InCl6 and Li3YCl6, the same test was conducted in an ambient atmosphere (25 °C and 45% relative humidity), as shown in Fig. 6(b). Li3InCl6 exhibited deliquescence, absorbing H2O from the atmospheric moisture and dissolving, and color change of the pH test paper indicated acidity, in good agreement with the system reported by Wang et al. [26]. In contrast, Li3YCl6 showed no deliquescence in the early stage of the reaction (~ 2 min); however, the pH test paper exhibited an acidic color change, which gradually spread without dissolution of the Li3YCl6 powder, even after 30 min. XRD analysis of the Li3YCl6 powder after 30 min in the ambient atmosphere showed peaks corresponding to Li3YCl6 and LiCl (Fig. 7). These results indicate that the initial reaction with water occurs without disrupting the matrix structure. We infer that Li3InCl6 dissolves easily in H2O when sufficient H2O molecules are present for solvation, whereas the partial replacement of Cl by OH ions is unlikely to proceed with a small amount of H2O. The opposite behavior occurs for Li3YCl6.

Fig. 6
figure 6

Photographs of pH test papers after exposure to the SEs for predetermined times. (a) and (b) Static exposure to an atmosphere with a dew point of -30 °C and atmospheric conditions, respectively. (c) Color change of the pH test paper dipped in aqueous solutions with various pH values

Fig. 7
figure 7

XRD patterns after exposure of the SE samples to an atmosphere with a dew point of -30 °C for up to 60 min. Panels (a) and (b) show the results for Li3InCl6 and Li3YCl6, respectively

Calculation results

Based on the experimental observations, we considered two types of reactions between Li3MCl6 and trace H2O during the early stages of decomposition, where the host structure of Li3MCl6 is maintained.

$${\rm{L}}{{\rm{i}}_{\rm{3}}}{\rm{MC}}{{\rm{l}}_{\rm{6}}} + {\rm{ x}}{{\rm{H}}_{\rm{2}}}{\rm{O}} \to {\rm{L}}{{\rm{i}}_{{\rm{3 - x}}}}{{\rm{H}}_{\rm{x}}}{\rm{MC}}{{\rm{l}}_{\rm{6}}} + {\rm{ xLiOH}}$$
(2)
$${\rm{L}}{{\rm{i}}_{\rm{3}}}{\rm{MC}}{{\rm{l}}_{\rm{6}}} + {\rm{ x}}{{\rm{H}}_{\rm{2}}}{\rm{O}} \to {\rm{L}}{{\rm{i}}_{\rm{3}}}{\rm{MC}}{{\rm{l}}_{{\rm{6 - x}}}}{\left( {{\rm{OH}}} \right)_{\rm{x}}} + {\rm{ xHCl}}$$
(3)

In Eq. (2), Li+ is replaced with H+, which results in the formation of alkaline LiOH. In contrast, in Eq. (3), Cl is replaced with OH, leading to the formation of acidic HCl. The experimental observation of HCl gas evolution suggests that Eq. (3) is likely to proceed. Therefore, defect formation energies, ΔEdef, for Eq. (3), where M = In and Y, were evaluated using DFT calculations (Table 1). Zero-point energy (ZPE) corrections are also considered by referring Materials Project datasets (molecule IDs, e67f3-H2O1-0-1 and b3c0-Cl1H1-0-1 for H2O and HCl, respectively) [38], where ZPE for H2O and HCl are 0.579 and 0.185 eV, respectively. Positive ΔEdef values were observed for both Li3InCl6 and Li3YCl6 because of the absolute zero-Kelvin assumption in this calculation and because the free energies of the gas-phase H2O and HCl molecules are largely reduced by the entropic effect. Hence, HCl gas evolution is more likely to occur from Li3YCl6 than from Li3InCl6, in agreement with the TPD-MS results at lower temperatures (< 200 °C), as shown in Fig. 2.

Table 1 Defect formation energies (ΔEdef) values calculated for Eq. (3). The values in brackets represent energies after zero-point energy correction

We also investigated the decomposition reaction energy with H2O, ΔEdecomp, which accompanies reconfiguration of the host structure of Li3MCl6. The following three reactions were considered:

$${\rm{L}}{{\rm{i}}_{\rm{3}}}{\rm{MC}}{{\rm{l}}_{\rm{6}}}{\rm{ + }}{{\rm{H}}_{\rm{2}}}{\rm{O}} \to {\rm{3LiCl }} + {\rm{ 1}}/{\rm{3MC}}{{\rm{l}}_{\rm{3}}} + {\rm{ }}1/3{{\rm{M}}_{\rm{2}}}{{\rm{O}}_{\rm{3}}} + {\rm{ }}2{\rm{HCl}}$$
(4)
$${\rm{L}}{{\rm{i}}_3}{\rm{MC}}{{\rm{l}}_6} + {\rm{ }}3{{\rm{H}}_{\rm{2}}}{\rm{O}} \to 6{\rm{HCl }} + {\rm{ }}1.5{\rm{L}}{{\rm{i}}_{\rm{2}}}{\rm{O }} + {\rm{ }}0.5{\rm{ }}{{\rm{M}}_{\rm{2}}}{{\rm{O}}_{\rm{3}}}$$
(5)
$${\rm{L}}{{\rm{i}}_3}{\rm{MC}}{{\rm{l}}_6} + {\rm{ }}3{{\rm{H}}_{\rm{2}}}{\rm{O}} \to 4.5{\rm{HCl }} + {\rm{ }}1.5{\rm{LiOH }} + {\rm{ }}0.5{{\rm{M}}_{\rm{2}}}{{\rm{O}}_3} + {\rm{ }}1.5{\rm{LiCl}}$$
(6)

The calculated values for the decomposition reaction energy per mole of H2O for reactions in Eqs. (4)–(6) are listed in Table 2. The decomposition energy for the reaction in Eq. (5) was high for both M = Y and M = In, indicating that the reaction is unlikely to proceed. This suggests that the formation of Li2O does not occur as readily as that of LiCl or LiOH. In contrast, the decomposition energy for the reactions in Eq. (4) and Eq. (6) when M = Y and M = In, respectively, was below 1 eV/H2O, which is comparable to the decomposition reaction energy value for the reaction in Eq. (3). The reaction in Eq. (3) involves the formation of a solid solution phase, Li3MCl6 − x (OH)x, without a significant change of the host structure. In contrast, reactions in Eq. (4) and Eq. (6) involve structural changes requiring the diffusion of all ions, making them kinetically slower than that in Eq. (3). Therefore, the reaction in Eq. (3) is dominant at low temperatures, resulting in the observation of HCl formation only for M = Y, which has a lower decomposition energy (Table 1). In contrast, the decomposition reactions in Eq. (4) for M = Y and Eq. (6) for M = In are expected to proceed at higher temperature. Thus, the HCl release reactions observed above 200 °C in the TPD-MS profile (Fig. 2) are considered to correspond to the reactions in Eqs. (4) and (6).

Table 2 Decomposition reaction energy with H2O (ΔEdecomp) values calculated for Eqs. (4)–(6). The values in brackets represent energies after zero-point energy correction

Conclusions

To promote the widespread use of all-solid-state batteries, it is important to design safe processes and properties, such as ion conductivity and formability. In this study, we focused on gas evolution resulting from chemical reactions between halide SEs and H2O in the atmosphere. Li3InCl6 and Li3YCl6 exhibited different interactions with H2O molecules and reaction mechanisms for HCl gas formation. Li3InCl6 was relatively resistant to HCl gas evolution because a small amount of H2O was physically adsorbed on the surface and the defect formation energy for Eq. (3) was higher. In contrast, Li3YCl6 underwent defect formation at lower temperatures because H2O was adsorbed within the bulk and the defect formation energy was lower. At high temperatures or in atmospheres containing more H2O, decomposition r proceeded for both Li3InCl6 and Li3YCl6. Though the knowledge on water tolerance for chloride electrolytes is limited to Li3InCl6 and Li3YCl6 in this study, the present study unveiled the reaction mechanism between chloride electrolytes and water. It is thought that penetration of water into the bulk of particles, leading to replacement of Cl by OH, is crucial for HCl gas release at low temperature. This phenomenon can be accelerated or suppressed, depending on the amount of H2O in the atmosphere, the metal ion or crystal structure in the Li3MCl6 composition.