New Series of Ternary Metal Chloride Superionic Conductors for High Performance All-Solid-State Lithium Batteries


 Understanding the relationship between structure, ionic conductivity, and synthesis is the key to the development of solid electrolytes for all-solid-state Lithium batteries. Here, we investigate chloride solid electrolytes with compositions Li3 − 3xM1+xCl6 (-0.14 < x ≤ 0.5, M = Tb, Dy, Ho, Y, Er, Tm). When x > 0.04, a trigonal to orthorhombic phase transition occurs in the isostructural Li-Dy-Cl, Li-Ho-Cl, Li-Y-Cl, Li-Er-Cl and Li-Tm-Cl solid electrolytes. The new orthorhombic phase shows a four-fold increase in ionic conductivity up to 1.3×10− 3 S cm− 1 at room temperature for Li2.73Ho1.09Cl6 (x = 0.09) when compared to the trigonal Li3HoCl6. For isostructural Li-Dy-Cl, Li-Y-Cl, Li-Er-Cl and Li-Tm-Cl solid electrolytes, about one order of magnitude increase in ionic conductivities are observed for the orthorhombic structure compared to the trigonal structure. Using the Li-Ho-Cl components as an example, detailed studies of its structure, phase transition, ionic conductivity, air stability and electrochemical stability have been made. Molecular dynamics simulations based on density functional theory reveal that the different cations arrangement in the orthorhombic structure leads to a higher lithium diffusivity as compared to the trigonal structure, rationalizing the improved ionic conductivities of the new Li-M-Cl electrolytes. All-solid-state batteries of In/Li2.73Ho1.09Cl6/NMC811 demonstrate excellent electrochemical performance at both room temperature and − 10°C. As relevant to the vast number of isostructural halide electrolytes, the present structure control strategy provides guidance for the design of novel halide superionic conductors.


Introduction
The development of high-performance all-solid-state batteries is contingent on the nding and synthesis of solid-state electrolytes (SSEs) with high ionic conductivity, good chemical and ambient air stability, wide electrochemical window, and desirable mechanical properties, especially for the applications in electric and hybrid electric vehicles [1][2][3][4] . Recently, several families of SSEs have attracted signi cant interest, such as metal sul des or oxides with polyanionic frameworks (PS 4 3− , PO 4 3− etc.) and metal halides with close packed anion sub-lattice structures [5][6][7][8][9][10] . Among these candidates, a promising family of metal chloride SSEs generally possess a wide electrochemical stability window (~ 4 V), good chemical stability towards ambient air and cathode materials (e.g. LiCoO 2 ). Some of them can be even synthesized at large scales from aqueous solutions 6,[11][12][13][14] . Although chloride-based SSEs have been developed over the past decades [15][16][17][18] , their use has been limited due to their low ionic conductivities. Until now, only a few metal chloride SSEs have achieved high room-temperature (RT) ionic conductivities over 10 − 4 S cm − 1 , including Li 3 YCl 6 14 , Li 3 InCl 6 19, 20 , Zr doped Li 3 MCl 6 (M = Y, Er) 21 , Li 3 Y 1 − x In x Cl 6 22 , Li x ScCl 3+x 23 and Li 2 Sc 2/3 Cl 4 24 . In the search for new metal-chloride SSEs, it is unclear whether the lack of good chloride conductors is intrinsic due to the different anionic framework structure or a result of the fact that they may be di cult to synthesize, which demands better understanding of the relationship between structure and ionic conductivity 19,25,26 .
Ternary chloride with the composition of Li 3 M(III)Cl 6 , where M(III) represents a trivalent rare earth metal, can crystallize in three types of structures including monoclinic (C2/m), trigonal (P-3m1), and orthorhombic (Pnma) phase 6,18 . Because the ionic radius of Cl − is typically much larger than that of the M 3+ ions, the Cl − sub-lattice forms the framework of these structures, where the interstitial sites are occupied by the Li + and M 3+ ions. The monoclinic structure occurs with relatively small metal radii, such as In or Sc. The Cl − sub-lattice forms a cubic close packing (ccp) stacking. In both the trigonal and orthorhombic structures, the Cl − sub-lattice is stacked in a hexagonal close packing (hcp) fashion. The hcp halogen stacking is interesting in view of structural diversity and Li-ion conductivity as it provides 6 octahedral and 12 tetrahedral interstitial sites in the Cl − sub-lattice per formula unit, with only 4 cations to occupy them. The speci c cation and vacancy arrangements imposed by the different symmetry of the trigonal and orthorhombic structures can be expected to result in distinct Li + transport properties.
Understanding the interrelationship of the structures and the Li + conducting behaviors can help us to design favorable structures towards higher ionic conductivities. Also, in the competition between the hcp stacked trigonal and orthorhombic structures, the average metal radius has been suggested to play a decisive role 6 , where the structures crystallize in the Pnma space group with smaller trivalent metal ions (Yb, Lu) and in the P-3m1 the space group with larger metal ions (Tb, Dy, Ho, Er, Tm) at room temperature. Only Li 3 YCl 6 has been reported in both space groups, while the Pnma phase is metastable. That was explained by an order-disorder phase transition connecting the two phases, with a phase transition temperature close to room temperature 18 .
Herein, we investigate the interplay between composition, structure, and Li-ion conductivity through the preparation of a series of Li-M(III)-Cl SSEs synthesized by co-melting of LiCl and MCl 3 . Using Li-Ho-Cl as example, by controlling the ratio between Li and Ho, a new series of orthorhombic-structured Li 3 − 3x Ho 1+x Cl 6 (0.04 < x ≤ 0.2, space group Pnma) were synthesized for the rst time. This orthorhombic Li-Ho-Cl material shows a cold-pressed ionic conductivity up to 1.3×10 − 3 S cm − 1 at room temperature. That is over four-fold greater than the Li 3 HoCl 6 with a trigonal structure (P-3m1). Bene ting from the high ionic conductivity as well as the wide electrochemical window of Li-Ho-Cl materials, the all-solid-state batteries with an In/Li-Ho-Cl/NMC811 con guration demonstrate excellent electrochemical performance at room temperature and − 10°C. A detailed structural investigation is combined with molecular dynamics simulations to reveal the relationship between structure and Li-ion conductivity. Moreover, similar trigonalto-orthorhombic phase transition phenomenon is reproducible in all Li-M(III)-Cl (M = Y, Er, Dy, Tm) structure with a P-3m1 space group except Li-Tb-Cl component. The structural transition induced by altering the ratio of Li + to M 3+ cations indicates that not only the average M 3+ radius but also the average cation (including both Li + and M 3+ ) size plays an important role in determining the structure. About one order of magnitude different in ionic conductivities is observed in the isostructural Li-Dy-Cl, Li-Y-Cl, Li-Er-Cl, and Li-Tm-Cl compositions. The increase in ionic conductivity from the trigonal to the orthorhombic phases is rationalized by the fact that the more regular orthorhombic phase results in easier Li + transport along the c-lattice direction (between the 8d1 and 8d2 Wykoff sites) which is a critical step in the diffusion network. These new insights into the relationship of ionic conductivity, chemical composition, and structure provide a new opportunity for halide solid electrolytes design and for the ultimate pursuit of highly conductive, stable, and processable solid electrolytes as required for all-solid-state batteries.

Results And Discussion
The halide SSEs based on the Li-Ho-Cl system were synthesized from a stoichiometric mixture of binary compound precursors (LiCl and HoCl 3 ) directly by co-melting at 650°C for 24 h. Figures 1a, b and S1 show the X-ray diffraction (XRD) patterns of Li 3 − 3x Ho 1+x Cl 6 SSE materials over the range of x from − 0.14 to 0.5. When the value of x is lower than 0.02, such as Li 2.95 Ho 1.017 Cl 6 ( Figure S1 x = 0.017), Li 3  quite different from that of the trigonal structure of Li 3 HoCl 6 . Peaks indexing of the XRD pattern reveals that the new phase has an orthorhombic unit cell with cell parameters of a = 12.9 Å, b = 11.2 Å, and c = 6.0 Å. Obedience to the extinction rules of h00 : h = 2n, 0k0 : k = 2n, 00l : l = 2n, 0kl : k + l = 2n + 1, hk0 : h = 2n + 1 is characteristic of the Pnma space group.
The structure of the Pnma phase was con rmed by Rietveld re nement of the XRD patterns ( Figure S2 octahedra with a short Ho-Li distance (Fig. 2d, e). It is interesting to note that the orthorhombic structure can only be obtained within certain limits of x in Li 3 − 3x Ho 1+x Cl 6 SSEs (0.04 < x ≤ 0.2). Compared to stoichiometric Li 3 HoCl 6 , the orthorhombic phase of Li 3 − 3x Ho 1+x Cl 6 (0.04 ≤ x ≤ 0.2) can also be regarded as a Ho-rich structure due to the doped amount of Li + by Ho 3+ , which leads to the introduction of more vacancies in the cation sub-lattice. Thus, even if they share a similar basic orthorhombic structure, the Ho 3+ , Li + , and vacancy occupation in Li 3 − 3x Ho 1+x Cl 6 (0.04 ≤ x ≤ 0.2) is slightly different from that of stoichiometric Li 3 HoCl 6 . For Li 3 HoCl 6 having a trigonal structure, there are three Ho sites (Fig. 2c). One is fully occupied (Ho1, Wyckoff 1a site) and the other two are partially occupied. There are two Li sites, the fully occupied Li1 layer (Wyckoff 6g site) and half occupied Li2 layer (Wyckoff 6d site) along the c axis.
As exhibited in Fig. 1g Figure S5). Feff modeling of each Hocentered atomic pairs was performed for scattering phase and magnitude. A ve-path structure model for R space curve tting was developed with scattering paths from corresponding coordination shells predicted by the theoretical Pnma structure of Li 3 HoCl 6 to be degenerated into 5 individual scattering paths (Table S2). The calculation predicted coordination number (CN) and bond distances (averaged from all paths of those corresponding shells) were used to guide the R space curve tting 28 . Table S2 shows the comparison between R space curve tting result and theoretical Pnma model in structure parameters. Figures S6, 7 compared the experimental data and Feff modeling in FT (mag & im part of FT; total and individual paths) and k 3 χ(k) (total and individual paths). It can be concluded that R space curve tting is well consistent with the initial Feff modeling. In addition, Finite Difference Method for Near Edge Structure (FDMNES) modeling 28 was performed to develop the corresponding theoretical XANES based on the same cluster (Fig. 3f). Comparison is made between the modeling and experimental data in XANES (Fig. 3e) and their rst derivative spectra (inset of Fig. 3e), revealing good agreement. The modeling data in k space for kχ(k) ( Figure S8) is also consistent with the initial Feff modeling (Fig. 3b). Thus, the Ho occupation at octahedral sites of the Li 2.73 Ho 1.09 Cl 6 framework was proven by the initial Feff modeling, R space curve tting, and XANES theoretical modeling.
Ionic conductivities of the cold-pressed Li 3 − 3x Ho 1+x Cl 6 SSEs were measured by temperature-dependent alternating-current (AC) impedance using a carbon/SSE/carbon cell. The conductivity isotherms as a function of frequency ν of the Li 2.73 Ho 1.09 Cl 6 SSE are presented in Fig. 4a. The conductivity plateau at 25°C corresponds to ~ 1.3×10 − 3 S cm − 1 is associated with the long-range ion transport. The 3D ionic transport in the structure of Li 2.73 Ho 1.09 Cl 6 SSE is re ected by the dispersive regime at higher frequency at -15°C with a κ value of 0.67 according to the Jonscher's power law 29,30 (σ (ν) ∝ ν κ ), as shown in Fig. 4a. Figure 4b shows the comparison of representative Nyquist plots at 25°C for the Li 2.73 Ho 1.09 Cl 6 (Pnma phase) and the Li 3 HoCl 6 (P-3m1 phase) SSEs. The equivalent circuit is presented in Fig. 4b are shown in Fig. 4c, and the extracted activation energies (E a ) with RT ionic conductivities in Li 3 − 3x Ho 1+x Cl 6 SSEs are presented in Fig. 4d. For the Pnma structures (0.04 < x ≤ 0.2), the RT ionic conductivities gradually increase with decreasing x, where the maximum ionic conductivity of 1.3 × 10 − 3 S cm − 1 is obtained at x = 0.09. In contrast, the ionic conductivity decreases dramatically when decreases further to x = 0.02. As aforementioned, Li 3 − 3x Ho 1+x Cl 6 changes from Pnma to P-3m1 when x ≤ 0.02.
Li 3 HoCl 6 (P-3m1) exhibits a RT Li + conductivity of 2.9 × 10 − 4 S cm − 1 (the conductivity isotherms are shown in Figure S9). The further decrease of ionic conductivity along with the decreasing of x value could be due to the formation of a LiCl impurity phase, which is also present in the XRD results. Moreover, the activation energies of the Li 3 − 3x Ho 1+x Cl 6 SSEs present the opposite trend (Fig. 4d). An activation energy below 0.4 eV is achieved for Li 3 − 3x Ho 1+x Cl 6 SSEs with the Pnma phase (0.02 < x ≤ 0.2). The minimum activation energy occurs at x = 0.09, which is consistent with the highest ionic conductivity of the resulting material. When x is lower than 0.02, where Li 3 − 3x Ho 1+x Cl 6 has the P-3m1 phase, the activation energy notably increases with decreasing x. This indicates that the cation arrangement of the Pnma symmetry supports a lower activation energy for Li-ion transport as compared to P-3m1, both having the same anion sub-lattice.
The electronic conductivity of Li 2.73 Ho 1.09 Cl 6 (x = 0.09) SSE was measured by chronoamperometry. Figure   S10 shows the current response for Li 2.73 Ho 1.09 Cl 6 (x = 0.09), with the initial response for both Li + and electron transport and the steady-state response for electron transport. The electronic conductivity of Li 2.73 Ho 1.09 Cl 6 (x = 0.09) is less than 1 × 10 − 9 S cm − 1 as calculated by Ohm's law at the voltage range from 0.1 to 0.4 V. At the same time, the stability of Li 3 − 3x Ho 1+x Cl 6 SSEs in dry air was evaluated by thermo gravimetric analysis, differential scanning calorimetry ( Figure S11) test, and XRD measurements ( Figure S12, 13) after exposure in a lithium battery dry room for 24 h. The results con rm that Li 3 − 3x Ho 1+x Cl 6 SSEs are stable in dry air and can be exposed for hours in the lithium battery dry room, which facilitates practical application.
In order to gure out whether the phase transition from a trigonal-to-orthorhombic is unique in the Li-Ho Dy, Tm), this is the rst time to achieve its orthorhombic phase, which not only extends its contents and application but also makes more clear understanding of the chemistry and structure formation of ternary rare earth chloride. These trigonal-to-orthorhombic phase transition by changing the ratio of Li and M(III) cations indicates that not only the average metal radius plays a role, but also the average cation size (including both Li + and M 3+ ). At the same time, the increased conductivity going from the trigonal to the orthorhombic phase can be found in all Li 3 − 3x M 1+x Cl 6 (-0.14 < x ≤ 0.5) SSEs. As shown in Table 1, a high ionic conductivity up to 7 × 10 − 4 S cm − 1 at room temperature can be achieved in orthorhombic phase Li 2.73 Y 1.09 Cl 6 (x = 0.09), which is more than 12 times higher than that of the trigonal phase Li 3 YCl 6 (x = 0).
The ionic conductivity of the new orthorhombic phase of Li 2.73 Er 1.09 Cl 6 (x = 0.09) is also seven times higher than that of trigonal phase of Li 3 ErCl 6 (x = 0) at room temperature (6.4 vs. 0.9 × 10 − 4 S cm − 1 ).
Orthorhombic phase Li 2.73 Dy 1.09 Cl 6 (x = 0.09) can reach to 9 × 10 − 4 S cm − 1 at room temperature, while it can only achieve about 1.2 × 10 − 4 S cm − 1 at room temperature in trigonal phase of Li 3 DyCl 6 (x = 0). For Li 3 − 3x Tm 1+x Cl 6 SSEs, the orthorhombic phase structure also have a higher ionic conductivity at room temperature than that of trigonal phase (8.9 vs. 1.1 × 10 − 4 S cm − 1 ). In general, around one order of magnitude different in ionic conductivities is observed in the isostructural Li-Dy-Cl, Li-Y-Cl, Li-Er-Cl, and Li-Tm-Cl compositions. As a result, lots of new ternary metal chloride superionic conductors have been found and reported here. Moreover, the similar structural transition process in many kinds of Li 3 − 3x M 1+x Cl 6 (-0.14 < x ≤ 0.5) SSEs indicates a general phenomenon, which can be used to explore and achieve many new materials with unknown phase structures.  (Fig. 5a), as well as from the offset of the slopes in the plot illustrating the Arrhenius behaviour (Fig. 5b). The activation energies extracted from the slope of the Arrhenius plot are almost similar, which have a slightly difference in the experimental observations ( Fig. 4). To further investigate the reason for the higher conductivity we consider the equation for the diffusion coe cient: Here, g is the geometrical factor, f the correlation factor, v 0 the attempt frequency, and a the jump distance, which can all be summarized into the preexponential factor D 0 . The geometrical factor describes the effect of a porous network on diffusion. The correlation factor describes the percentage of back and forth movements of atoms between sites, and is de ned as where D Tracer is the tracer diffusion coe cient and D jump the jump diffusion coe cient.
The attempt frequency v 0 and amplitude of the Li + ions were extracted by the Fourier transform of the diffusion paths, as described elsewhere 33 . Both the frequencies and the amplitudes are similar in both phases. The increase in ionic conductivity is therefore not due to an increase in attempt frequency. Using the same algorithm for the attempt frequency, the Cl vibrations were analyzed as well. As can be expected considering the structural similarity of the hcp stacked chlorine atoms, the Cl vibrational amplitudes and frequencies are also similar. ( Figure S20). The jump distance a can be approximated by the distance between the different Wyckhoff sites. The a values are similar for both phases, around 3.1Å Hence, the difference observed in D 0 has to arise either from the geometrical factor or the correlation factor. It is therefore interesting to compare the jump rates between sites and the long-range diffusion path. Figures 5c and d illustrate the jumps between different Li sites. The blue circles are the Li sites, and the different sizes of the blue circles represent the occupancy of the site. The red lines represent jumps that occurred, with the thickness of the line indicating the frequency of the jump occurred. The green trajectories belong to the Ho atoms, and the Wyckoff sites are included. For the Pnma phase at 500 K (Figure 5c), there is a clear long-range diffusion path along the z-direction between the Wyckoff 8d1 and 8d2 sites. Even though jumps between the 8d1 and 8d2 sites are also possible in the xy-plane, these jumps only start occurring at 600K ( Figure S21). This indicates that the jumps along the z-direction have a lower activation energy than cross-plane jumps, thus indicating that the high conductivity is mainly due to these one-dimensional diffusion pathways. For the P-3m1 phase, the jumps in the z-direction between the 6h and 6g sites are the most frequently occurring jumps (Figure 5d and S22). Though in principle, the pathway along the z-direction is also possible, the clear long-range path as in Pnma is not present. Considering the structures, this is caused by the different Ho arrangements, leading to a different local structure which can be less favourable for jump processes. A different initial arrangement of Ho on the sites would probably lead to different diffusion pathways, but this was not investigated in the scope of this report. The effect of M(III) site occupancy in the P-3m1 phase has been previously studied for the P- 3m1 phase of Li 3 YCl 6 and Li 3 ErCl 6 6, 34 . It was found that the structures with more frequently occurring M1 and M3 distances (hence, a higher joint occupancy of neighbouring M1 and M3 sites), the diffusivity increases. This was explained by a continuous diffusion path in the z-y plane passing through one of the free tetrahedral sites. The present results indicate that the facile diffusion in the z-direction in the Pnma phase is more favorable for achieving high conductivities.
As presented in Figure 6a, cell shows a steady overpotential of 580 mV in the initial cycles, which gradually decreases to ~50 mV over 900 h. The reduced overpotential might be due to the increased electroactive surface area of the Li metal caused by dendrite growth in the grain boundary of SSEs layer, thus causing the localized current density to decrease, with a consequential decrease in overpotential.  Figure 7b and Figure S24, with speci c capacities of 125.5 mAh g -1 and 90 mAh g -1 remaining after 180 cycles at 25 °C and 100 cycles at -10 °C. The typical delithiation/lithiation behaviour of NMC811 is clearly visible in the differential capacity curves (Figure 7c). Three pairs of redox peaks are observed, corresponding to the phase transitions from hexagonal to monoclinic (H1→M), monoclinic to hexagonal (M→H2), and hexagonal to hexagonal (H2→H3). The high initial Coulombic e ciency and lack of redox reactions attributed to the Li 2.73 Ho 1.09 Cl 6 SSE indicate that Li 2.73 Ho 1.09 Cl 6 is stable towards NMC811 within the applied voltage range. The rate capabilities at different current densities ranging from 100 mA g -1 to 1000 mA g -1 (0.1 C to 1 C) at 25 °C are displayed in Figure 7d. The capacity gradually decreases along with an increase of current density, with 98 mAh g -1 achieved at 1 C (58% capacity retention of that at 0.1 C). Moreover, the capacity can be recovered upon returning to the initial 0.1 C rate.
To monitor the interfacial resistance contributions, EIS spectra were measured at different state-of-charge (SOC) and state-of-discharge (SOD) during the 3 rd charge/discharge process (Figure 7e, voltage evolution curve shown in Figure S25). The mid-frequency semicircle related to the NMC811/Li 2.73 Ho 1.09 Cl 6 interfacial resistance increased slightly during the charging process and decreased again during the discharging process. The change in the interfacial resistance might be caused by the mild volume shrinkage of NMC811 during delithiation and corresponding volume recovery once lithiated 36  composite at different charge/discharge states also showed no obvious change. Therefore, it can be concluded that there is a stable interface between NMC811 and Li 2.73 Ho 1.09 Cl 6 SSE regardless of the static physical contact or during charge/discharge cycling process.

Conclusion
In summary, the ternary metal chloride solid electrolyte series of Li 3 − 3x M 1+x Cl 6 (-0.14 < x ≤ 0.2, M = Tb, Dy, Ho, Y, Er, Tm) reveal a phase transition from trigonal (P-3m1) to orthorhombic (Pnma) upon increasing x values for M-rich environments. Both P-3m1 and Pnma structures consist of hcp framework of Cl − anions but differ in their cations (including Li + and M 3+ ) arrangement. Using Li 3 − 3x Ho 1+x Cl 6 (-0.14 < x ≤ 0.2) as an example, the relationship between structure and Li-ion conductivity is revealed by temperaturedependent EIS, X-ray and neutron diffractions and ab initio MD simulations. The highest RT Li + conductivity of 1.3 ×10 − 3 S cm − 1 is achieved for the orthohombic Li 2.73 Ho 1.09 Cl 6 , which is over four times higher than that of the trigonal Li 3 HoCl 6 (0.3 ×10 − 3 S cm − 1 ). About one order of magnitude difference in ionic conductivities is observed in the isostructural Li-Dy-Cl, Li-Y-Cl, Li-Er-Cl, and Li-Tm-Cl compositions.
The phase transition to Pnma triggers a signi cant increase in Li + diffusivity and reduces the activation energy barrier for Li + diffusion in all different Li-M-Cl systems. Considering the vast number of isomorphic structures, the synthesis strategy based on trigonal-to-orthorhombic phase transition phenomenon not only can be used to discover the fundamental chemical theories of the rare earth metal halides, but also can explore and achieve new materials with high Li + conductive. Ab initio MD simulations consistently derive a higher diffusivity and a reduce activation energy of the Pnma phase comparing to the P-3m1 phase. This is due to facile one-dimensional diffusion pathways in the z-direction, which is obstructed by the different Ho arrangement in the P-3m1 phase. All-solid-state batteries of In/Li 2.73 Ho 1.09 Cl 6 /NMC811 exhibit excellent electrochemical performances at both room temperature and low temperature (-10°C). This further demonstrates the applicability of this new Li 2.73 Ho 1.09 Cl 6 SSE with high ionic conductivity and a wide electrochemical window. These results provide guidance for the design of novel halide superionic conductors and contribute to the development of ASSLBs.