Crystal Structure of Garnet-Related Li-Ion Conductor Li7-3x Ga x La3Zr2O12: Fast Li-Ion Conduction Caused by a Different Cubic Modification?

Li-oxide garnets such as Li7La3Zr2O12 (LLZO) are among the most promising candidates for solid-state electrolytes to be used in next-generation Li-ion batteries. The garnet-structured cubic modification of LLZO, showing space group Ia-3d, has to be stabilized with supervalent cations. LLZO stabilized with Ga3+ shows superior properties compared to LLZO stabilized with similar cations; however, the reason for this behavior is still unknown. In this study, a comprehensive structural characterization of Ga-stabilized LLZO is performed by means of single-crystal X-ray diffraction. Coarse-grained samples with crystal sizes of several hundred micrometers are obtained by solid-state reaction. Single-crystal X-ray diffraction results show that Li7-3x Ga x La3Zr2O12 with x > 0.07 crystallizes in the acentric cubic space group I-43d. This is the first definite record of this cubic modification for LLZO materials and might explain the superior electrochemical performance of Ga-stabilized LLZO compared to its Al-stabilized counterpart. The phase transition seems to be caused by the site preference of Ga3+. 7Li NMR spectroscopy indicates an additional Li-ion diffusion process for LLZO with space group I-43d compared to space group Ia-3d. Despite all efforts undertaken to reveal structure-property relationships for this class of materials, this study highlights the potential for new discoveries.

Introduction

Since the initial studies by Murugan et al. in 2007, Li7La3Zr2O12 (LLZO) has received much scientific attention as a solid electrolyte for “Beyond Li Ion Battery” concepts, such as Li–air or Li–S batteries.[1] LLZO provides a high Li-ion conductivity σ (10–3 to 10–4 S cm–1 at ambient temperature) and a Li+ transference number approaching 1, superior chemical stability against high voltage cathodes, and electrochemical inertness in a wide potential window of up to 6 V.[2] In particular, its stability against Li-metal as well as its thermal and mechanical robustness makes LLZO garnet exceptionally well-suited to be used as a protecting layer for Li-metal-based batteries.[1−3]

For pure LLZO, two structural polymorphs are described in the literature. A low-temperature tetragonal modification with a completely ordered arrangement of Li+ crystallizes in space group (SG) I41/acd (no. 142). This structure type is described as garnet-related framework with two types of dodecahedral LaO8 polyhedra (8b and 16e) and ZrO6 octahedra (16c). Li+ occupies three different sites: the tetrahedral 8a site as well as distorted octahedral 16f and 32g sites.[4]

In contrast to the tetragonal modification, the cubic high-temperature modification exhibits a disorder in the Li+ distribution. This cubic modification shows SG Ia-3d (no. 230) and has the garnet structure composed of a framework of 8-fold coordinated LaO8 dodecahedra (24c) and 6-fold coordinated ZrO6 octahedra (16a). Li+ is located at interstitial sites, showing tetrahedral (24d), octahedral (48g), and distorted 4-fold (96h) coordination.[5]

For use as Li-ion conductors, the cubic modification is much more desirable as its Li-ion conductivity is 2 orders of magnitude higher than for the tetragonal modification.[3,4,6]

The cubic modification of pure LLZO is not stable at room temperature (RT). However, it can be stabilized at RT by the introduction of supervalent cations, which cause a reduction of the Li+ content that leads to the introduction of additional vacancies at the Li+ sites. This increases the entropy and reduces the free energy.[7] The stabilization of the cubic modification was originally achieved by the incorporation of Al3+.[6,8−11] Further research led to the discovery of other cations that are capable to stabilize the cubic modification.[12−23]

Among these, Ga3+ turned out to be a promising candidate which has been studied extensively.[21,22,24−30] Ga3+ is incorporated at the Li+ sites according to 3 Li+ → Ga3+ + 2 □Li, i.e., creating two vacant sites, □Li, in the Li-sublattice.

Ga-stabilized LLZO is characterized by a Li+ conductivity of 1.3 mS cm–1 at RT, which is twice as high as for LLZO stabilized with Al.[28] The reason for the higher Li-ion conductivity of Ga-stabilized LLZO is, however, not yet fully understood. As noted above, supervalent cations are needed to stabilize the cubic modification. Because of their blocking effect, some of these ions are suspected to hinder the long-range Li-ion transport and, therefore, to reduce Li-ion conductivity. Allen et al. (2012) compared the properties of Li6.75La3Zr1.75Ta0.25O12. (LLZTO) with those of Li6.15La3Zr1.75Ta0.25M0.2O12 (M = Al, Ga); according to their study pure LLZTO shows the highest Li-ion conductivity, followed by LLZTO:Ga and LLZTO:Al.[21] On the basis of the crystal chemical considerations, they attributed this behavior to the different site preference of Al3+ and Ga3+. In particular, the larger Ga3+ prefers the 96h site, which seems to be less hindering for long-range Li-ion transport, compared to Al3+, located at the 24d site, which is a junction for Li-ion diffusion. However, the site preference of Ga3+ is still under discussion as 71Ga nuclear magnetic resonance (NMR) spectroscopy studies revealed different results.[28−31] The most recent study by Rettenwander et al. (2015), using very high magnetic fields (21.1 T), showed that the site preference of Ga3+ and Al3+ to occupy 24d and 96h voids in samples crystallizing with Ia-3d is practically the same.[31] Thus, despite all efforts, there is, so far, no satisfying explanation why some of the Ga-stabilized LLZO samples presented in literature show higher ionic conductivities.

Investigations on Ga-stabilized LLZO single crystals might help us to shed light on this question. Up to now, very little research has been performed on LLZO single crystals.[4,5,8,32] The synthesis of applicable LLZO single crystals is delicate. To the best of our knowledge, no single-crystal X-ray diffraction (SC-XRD) study on Ga-stabilized LLZO has been published yet.

In the present study, another cubic modification of LLZO, showing the acentric SG I-43d, has been observed for Ga-stabilized LLZO for the first time. Coarse-grained Ga-stabilized LLZO samples were synthesized via solid-state reaction and characterized by a rich portfolio of techniques including X-ray powder diffraction (XRPD), single-crystal X-ray diffraction (SC-XRD), neutron powder diffraction (NPD), scanning electron microscopy (SEM)/ backscattered electron (BSE) imaging, energy-dispersive X-ray spectroscopy (EDX), and 7Li NMR spectroscopy.

Experimental Section

Synthesis

A series of Li7–3GaLa3Zr2O12 with intended Ga contents xGa = 0.10, 0.15, 0.20, 0.30, 0.40, 0.50, 0.60 pfu was synthesized by high-temperature sintering in air. For reason of comparison, also one sample of Al-stabilized LLZO with an intended Al content xAl of 0.20 Al pfu (Li6.4Al0.2La3Zr2O12) was prepared under the same conditions. Results obtained by Cheng et al. concerning the densification and coarsening of LLZO samples were taken into consideration for this synthesis.[33] Li2CO3 (99%, Merck), Ga2O3 (99.0%, Aldrich), Al2O3 (99.5%, Aldrich), La2O3 (99.99%, Roth), and ZrO2 (99.0%, Roth) were used as reagents. The starting materials were weighed out in their intended stoichiometric proportions with an excess of 10 wt % Li2CO3 to compensate the loss of Li2O during sintering. The reagents were mixed in an agate mortar under addition of isopropyl alcohol and then cold-pressed into pellets with a uniaxial press. The resulting pellets were put in an alumina crucible. To avoid undesired contamination with Al from the crucible, the samples were placed on a pellet of pure LLZO. The samples were heated to 850 °C with a rate of 5 °C/min and calcinated for 4 h. The resulting pellets were then removed from the furnace, ground in an agate mortar, and ball-milled for 1 h under isopropyl alcohol (FRITSCH Pulverisette 7, 800 rpm, 2 mm ZrO2 balls). After drying under air, the powder was again cold-pressed into pellets. The sample pellets were again put into an alumina crucible. To avoid incorporation of Al3+ from the crucible, the samples were again placed on a pellet of pure LLZO. To suppress formation of extra phases due to Li loss during sintering, the sample pellets were covered with a pellet of pure LLZO. For the final sintering, the pellets were heated in a muffle furnace in air with a rate of 5 °C/min to 1230 °C and sintered for 6 h. For XRPD investigations, small fragments of the sintered pellets were ground using an agate mortar. Pellets used for NPD were stored in a glovebox under Ar atmosphere to avoid reaction with CO2 and H2O; also the whole processing after sintering, including grinding using an agate mortar and filling of the sample container has been performed under Ar atmosphere. For SEM analysis, polycrystalline chips from the pellets were embedded in an epoxy holder. The surface was ground and then polished using diamond paste. In addition, SEM analysis was also performed on unpolished pellet fragments. Single grains were isolated from the pellets for SC-XRD investigations.

SEM

Images were taken using a Zeiss Ultra Plus device. In particular, we put emphasis on the investigation of the grain size, morphology, phase composition, and the chemical homogeneity, i.e., the distribution, of Ga, La, Zr, using a backscattered electrons detector (BSE) and energy-dispersive X-ray spectroscopy (EDX) measurements, respectively.

XRPD

Patterns were recorded using a Bruker D8 Advance DaVinci Design diffractometer with a Lynxeye solid-state detector using Cu Kα radiation. The synthesis products were characterized regarding the presence of extra phases as well as to determine the symmetry and unit-cell dimension of the samples. Data were collected between 10° and 80° 2θ. Evaluation of XRPD data was performed by Rietveld refinement using the program Bruker DIFFRACplus TOPAS (version 4.2).

SC-XRD

Single-crystal X-ray diffraction data were collected on a Bruker SMART APEX CCD-diffractometer. Samples for single-crystal studies were selected readily after removing of samples from the furnace, and data collection was done within 24 h after synthesis. So we exclude that the change in SG symmetry is due to the incorporation of H+ in the structure as it was suggested in literature. Single crystals were carefully hand-picked under the binocular loupe from the gently crushed pellets. For each composition, more than 10 crystals were selected on the basis of their optical properties (regular shape, clearness, and homogeneity in color). Selected crystals were glued on top of glass capillaries (0.1 mm Ø) and tested on the diffractometer. A full set of intensity data was collected on 2−4 crystals for each composition to obtain good statistics and data sets for crystal chemical interpretation, resulting in a total of 21 data sets. Generally all tested crystals were of high quality with sharp diffraction peaks. Intensity data were collected with graphite-monochromatized Mo Kα X-radiation (50 kV, 30 mA); the crystal-to-detector distance was 30 mm, and the detector was positioned at −30° and −50° 2θ using an ω-scan mode strategy at four different ϕ positions (0°, 90°, 180°, and 270°) for each 2θ position. In total, 630 frames with Δω = 0.3° were acquired for each run. With this strategy, data in a large Q-range up to minimum d-values d = 0.53 Å could be acquired. This is necessary for accurate determination of anisotropic displacement parameters and to reduce correlation effects between atomic displacement parameters and site occupation numbers. Three-dimensional data were integrated and corrected for Lorentz-, polarization, and background effects using the APEX2 software.[34] Structure solution (using direct methods) and subsequent weighted full-matrix least-squares refinements on F2 were done with SHELX-2012 as implemented in the program suite WinGX 2014.1.[35,36]

NPD

Neutron powder diffraction data for a sample with a nominal Ga3+ content xGa = 0.20 pfu were collected at the Maier-Leibnitz Zentrum (MLZ), FRM-II, Munich, Germany. Powder diffraction data were acquired at 298 K in constant wavelength mode using the high-resolution powder diffractometer SPODI with Ge551 monochromatized neutron radiation (λ = 1.5482 Å). Experiments were performed in a 2θ range of 3° ≤ 2θ ≤ 154°, step width of 0.045°. Data treatment of powder diffraction data sets as well as a combined refinement of neutron powder diffraction data and single-crystal X-ray diffraction data (with special emphasis of Li-cationic distribution) was performed using the FULLPROF-suite of programs.[37]

7Li NMR

Variable-temperature 7Li NMR spectra and spin–lattice relaxation rates were recorded with a Bruker Avance III solid-state spectrometer connected to a 7-T magnet (Bruker). The resonance frequency was 116 MHz; 90° pulse lengths ranged from 2 to 3 μs depending on temperature. We used single-pulse and solid-echo experiments to record NMR line shapes. The spin–lattice relaxation rates in the laboratory frame were measured by means of the saturation recovery sequence; the rates in the rotating frame were acquired by using the spin-lock technique, see Epp et al. for details.[38] To protect the samples permanently from humidity during the NMR measurements small pieces of the sintered garnets were put into glass ampules that were sealed under vacuum.

Results

Crystal Size and Morphology

Polished samples and pellet fragments were examined by SEM-BSE. For all compositions, grain sizes of >100 μm were achieved. Single LLZO grains show an isometric shape. Minor amounts of extra phases were documented by SEM-BSE due to a contrast in brightness and a different morphology. As already observed visually, extra phases are commonly found in the peripheral part of the pellets. Single grains are commonly separated by gaps, and small amounts of extra phases are located in these gaps. LLZO grains do not show inclusions of extra phases. Only small open voids with a size of <10 μm were found as inclusions within LLZO grains. No zoning within LLZO grains was observed by SEM-BSE. A representative SEM-BSE image of a pellet fragment is shown in Figure .

Figure 1

Comparison of XRPD patterns for samples Ga10 (lowermost pattern), Ga15, Ga20, Ga30, Ga40, Ga50, Ga60 (uppermost pattern). Pattern contributions from extra phases are marked with colors: La2Zr2O7 (blue), La(OH)3 (green), LiGaO2 (violet). The peak at 2θ = 21.65° (red) could neither be attributed to cubic LLZO with SG Ia-3d nor to any extra phase. The inset shows a representative SEM-BSE image of a pellet fragment of sample Ga40 to demonstrate the grain size and morphology.

Chemical Composition According to SEM-EDX

Standard-free SEM-EDX analysis of LLZO grains showed the presence of La, Zr, and Ga. The La/Zr ratio of LLZO grains was slightly below the theoretical values of 3:2, which is in accordance with site occupation refinements from SC-XRD data and microprobe measurements of Ga-stabilized LLZO by Rettenwander et al.[29] Ga contents of LLZO grains were partially lower than the target stoichiometry; additionally, a slight variation of the Ga content within single samples was noted. EDX mapping of LLZO grains confirmed their homogeneity with regards to the distribution of La and Zr within single grains. Extra phases observed by SEM-BSE can be divided due to their chemical composition seen by EDX into three different phases. An extra phase with a La/Zr ratio of 1:1 and another extra phase showing a La signal only were mainly found in the peripheral part of the pellets. For samples with nominal xGa = (0.50, 0.60) per formula unit (pfu), another extra phase that only showed a Ga signal in EDX was found.

The XRPD patterns of Li7–3GaLa3Zr2O12 (GaX) with nominal Ga3+ concentration xGa = 0.10, 0.15, 0.20, 0.30, 0.40, 0.50, 0.60 pfu are shown in Figure . The phase compositions of the syntheses were evaluated by Rietveld analysis. All garnet compositions exhibit reflections indicating cubic symmetry. For sample with xGa = 0.10, tetragonal LLZO was still present with a share of 18 wt % as determined by quantitative phase analysis using the Rietveld method. For samples with xGa ≥ 0.15, no tetragonal LLZO phase was observed. La2Zr2O7 and La(OH)3 were identified as extra phases and confirmed results of SEM-EDX investigations. Considering the SEM-BSE results, these extra phases occur predominantly in the rim of the pellets and formed due to evaporation of Li during high temperature sintering. The presence of extra phases was accepted as they facilitate the extraction of single crystals. For sample with xGa = 0.60, also LiGaO2 was observed (1.2 wt %). It has to be noted that for most samples, an additional weak reflection at 2θ = 21.65° (equal to d = 4.10 Å) was observed. This peak cannot be explained with LLZO showing SG Ia-3d or any extra phase.

Crystal Structure Determination from SC-XRD

The experimental data and results of structure refinement for selected samples are reported in Table and Table , while the fractional atomic coordinates, occupation numbers and equivalent isotropic, and anisotropic atomic displacement parameters are given in Table S1. Crystallographic information files (CIF) with full structural information are deposited as Supporting Information.

Table 1

Summary of SC-XRD Results of Selected Samples

sample	SG	xGa_ref (pfu)	a0 (Å)	Li1–Li2 (Å)	Li1–Li3 (Å)	Li2–Li3 (Å)	Li2–Li2 (Å)	VZrO6 (Å3)	VLaO8 (Å3)
Ga10_1	Ia-3d	0.05(1)	12.9844(2)	1.64(4)	b	b	0.73(6)	12.43	28.33
Ga30_1	I-43d	0.20(1)	12.9736(1)	b	1.640(12)	2.349(12)	b	12.42	28.28
Ga40_2	I-43d	0.36(1)	12.9658(1)	b	1.653(13)	2.330(13)	b	12.39	28.22
Al20_7	Ia-3d	0.19(1)a	12.9652(2)	1.66(2)	b	b	0.71(3)	12.43	28.18

xAl value instead of xGa_ref; fixed to the data obtained from EDX analysis for sample Al20_7.

Not applicable for this modification.

Table 2

Experimental Setup and Results of Refinement for Selected LLZO Samples

sample ID		Ga10_1	Ga30_1	Ga40_2	Al20_7
depository no. (FIZ Karlsruhe)		CSD-430602	CSD-430603	CSD-430604	CSD-430571
diffractometer		Bruker SMART APEX
radiation		Mo Kα
wavelength (Å)		0.71073
scan mode		ω-scan at 4 different ϕ positions (0°, 90°, 180°, 270°) for each 2θ position
absorption correction type		empirical
crystal system		cubic
space group		Ia-3d	I-43d	I-43d	Ia-3d
a (Å)		12.9844(7)	12.9736(1)	12.9658(1)	12.9652(4)
volume (Å3)		2189.1(4)	2183.64(5)	2179.71(5)	2179.4(2)
Z		8	8	8	8
ρcalc (g/cm3)		5.024	5.12	5.167	5.160
crystal size (μm)		120 × 110 × 110	160 × 150 × 110	160 × 140 × 70	170 × 165 × 160
temp (K)		298(2)	298(2)	298(2)	295(2)
theta range (deg)		3.88–33.46	3.85–37.33	3.84–39.03	3.85–29.96
index range	h	–20...20	–21...22	–22...22	–17...18
	k	–20...20	–22...22	–22...22	–18...17
	l	–20...19	–22...22	–22...22	–17...18
resolution dmin (Å)		0.64	0.53	0.56
absorption coefficient		13.14	13.64	13.97	13.44
abs. corr. Tmin/Tmax		0.31/0.33	0.15/0.26	0.14/0.29	0.68/0.75
reflections collected		25935	32907	34027	12617
independent reflections		366	960	1055	268
Rint (%)		4.19	4.20	3.32	2.46
no. of free parameters		25	46	48	24
R1 (all data)		3.75	1.67	1.76	1.94
wR2 (all data)		6.67	3.21	3.38	3.55
GooF on F2		1.454	1.222	1.385	1.518
extinction coefficient		n/a	0.00078(4)	0.00202(6)	0.00034(4)
largest diff. peak/hole		1.058/ −0.974	0.495/ −0.802	1.072/ −0.985	0.432/ −0.466

Single-crystal X-ray intensity data processing gave strong evidence for the cubic crystal system for all data collected. For the data of sample Ga10, with nominal Ga concentration xGa = 0.10 pfu, SG determination yields Ia-3d symmetry as is widely accepted for LLZO garnets.[5,6] However, for all samples with nominal Ga concentrations xGa ≥ 0.15 pfu, intensity statistics and systematic extinctions did not result in SG Ia-3d but yielded an acentric SG I-43d (no. 220). This SG was consistently obtained using all the different SG determination tools as implemented in WinGX.[35] Subsequent structure solution tests for samples with xGa ≥ 0.15 pfu using commonly occurring SG Ia-3d or tetragonal SG I41/acd and SG I41/a failed. Simulated precession images of the 0kl layer of samples Ga10_1 and Ga40_2, displayed in Figure , obviously show the presence of Bragg peaks with k = odd and l = odd in sample Ga40. These peaks are forbidden in SG Ia-3d (as is also the case for sample Ga10_1) but allowed in SG I-43d.

Figure 2

Comparison of simulated precession images of the 0kl layer. (A) Ga10_1 showing SG Ia-3d. (B) Ga40_2 showing SG I-43d. The presence of Bragg peaks with k = odd and l = odd in (B) is evident. Two sections are marked with yellow rectangles to serve as an example.

Structure solution with direct methods using I-43d symmetry resulted in a structure model which could be refined down to wR2 values for all data <4%. The Li-positions were located from difference Fourier map calculations, in the final structure all atoms could be treated with anisotropic atomic displacement parameters. Our structure model has been found consistent with the one given by Lager et al. and Galven et al.[39,40] It is, however, unexpected here as their samples involved “hard” Li+ leaching and replacement by H+ and up to now, no doubt on the violation of Ia-3d symmetry for Ga-stabilized LLZO was reported. Samples of Al-stabilized LLZO, which are produced with the very same synthesis strategies, show Ia-3d symmetry up to nominal compositions of xAl = 0.30 pfu in single-crystal structure refinements. Different to the Ia-3d structure of sample Ga10 with a nominal Ga content xGa = 0.10 pfu, which exhibits 5 different atomic positions in the asymmetric unit, the crystal structure of Ga-stabilized LLZO with a nominal Ga content xGa ≥ 0.15 pfu, showing SG I-43d, exhibits seven different atomic positions: La3+ occupies the 8-fold coordinated 24d position (site symmetry 2..), and Zr4+ is located at the octahedrally coordinated 16c position (site symmetry .3.). These positions are similar in the two different structures. Li+ is found distributed over three different positions in I-43d: two of them correspond to the regular tetrahedral coordinated site of the silicate garnets (24d in Ia-3d), nevertheless, they split into two positions, namely Li1 (12a) and Li2 (12b), both with site symmetry −4.., and differ in both bond lengths and polyhedral distortion. The third Li+ position, namely Li3, has been found located on general position 48e, site symmetry 1, which is only partly occupied as is also common in LLZO compounds. In contrast to common silicate garnets, the acentric structure exhibits two independent O2− positions, both in general position 48e. Table shows a summary of the SC-XRD results of selected samples. More detailed information is given in Table , while a comparison of the two structural models Ia-3d and I-43d is given in Table S1. A graphical representation of the Ia-3d and I-43d structures, including the Li-network, is displayed in Figure .

Figure 3

(A) Crystal structure of Ga-stabilized LLZO with xGa = 0.10 and SG Ia-3d. Blue dodecahedra represent 8-fold coordinated La3+ (at the Wyckoff position 24c); green octahedra 6-fold coordinated Zr4+ (16a). The red spheres correspond to tetrahedrally coordinated Li+ at the 24d (Li1) site, yellow spheres correspond to distorted 4-fold coordinated Li+ at Wyckoff position 96h (Li2). (B) Crystal structure of Ga-stabilized LLZO with xGa = 0.30 and SG I-43d. Blue dodecahedra represent 8-fold coordinated La3+ (at the Wyckoff position 24d); green octahedra 6-fold coordinated Zr4+ (16c). The red spheres correspond to tetrahedrally coordinated Li+ at the 12a site (Li1), orange spheres represent tetrahedrally coordinated Li+ at the 12b site (Li2); yellow spheres correspond to distorted 6-fold coordinated Li+ at Wyckoff position 48e (Li3). (C) Projection of Li-network for SG Ia-3d (left) and SG I-43d (right).

Site Occupation Refinements from SC-XRD

Both, for the Ia-3d and the I-43d structure, site occupation refinements yielded the octahedral 16c sites to be fully occupied by Zr4+, no evidence was found for a Ga3+ substitution onto this site, so during refinements, the occupation of Zr4+ was fixed to the ideal value. For both structures, however, there is a small deficiency of La3+, and thus, the site occupation factor was allowed to vary freely during the refinements. The finding of a small La3+ deficit is consistent with results from EDX analysis. The amount of vacancies varies between 0.08(2) in Ga3+-poor samples to 0.04(2) in Ga3+-rich samples, and there are some evidence that the amount of vacancies at the La3+ site steadily decreases upon increasing of Ga3+ content, see Figure A.

Figure 4

Occupation of different sites as a function of the refined Ga content xGa_ref. (A) The occupation of La site increases at higher Ga contents. (B) Ga is almost exclusively located at the smaller Li1 site (12a), only for samples with high Ga contents, a small amount of Ga is located at the Li2 site (12b). (C) As a function of the decreased total Li content for samples with higher Ga contents, the occupation of the Li3 site decreases.

For the Ia-3d structure, strong evidence was found that Ga3+ substitutes onto the 24d position. For the I-43d structure, the refinement strategy was modified slightly. In a first refinement cycle, the site occupation numbers of the Li1, Li2, and Li3 sites were allowed to refine freely without constraints, assuming only the scattering power of Li+. These preliminary refinements showed a surplus of electron density at the Li1 site (occupation of Li larger than 1.00, the value of full occupation), and thus, a larger scatterer—in this case Ga3+—must be present, whereas at the Li2 site, the refined occupation was close to 1.00 or even lower, indicating some vacancies at this site. Li2 site values >1.00 were observed only for higher overall Ga3+ contents. For occupation numbers >1.00, a mixed occupation with Li+ and Ga3+ was refined, assuming full occupation of these sites. With the model of site occupation refinement, we have strong evidence that Ga3+ shows a strong preference for the Li1 site, only very small amounts of Ga3+ are found at the Li2 site (Figure B). It should be noted that the Li2 site, thereby, is fully occupied in most cases and the Li-content is 1.44–1.48 pfu. We are aware that this model of full occupation of Li1 and Li2 sites may suffer small deficits. However, it yields total Ga3+ contents which are very close to or even higher than those determined from EDX analysis of the samples studied by SC-XRD. Assuming vacancy concentrations of 20% on both sites, this does not change the site preference of Ga3+ for the Li1 site but only increases the overall Ga3+ content by ∼15%. Additionally, there is good evidence from neutron diffraction data that the amount of vacancies indeed is below 20% in Ga-stabilized LLZO samples (see below). No changes in structural parameter are observed by assuming vacancies. The occupation of the Li3 site generally was refined freely and gave a partial occupation of this site. With increasing Ga3+ substitution, the occupation of the Li3 site decreases; that is, the trivalent cationic substitution reduces the amount of interstitial Li+ (see Figure C). The decrease of Li+ with increasing Ga3+ is even more evident and with smaller scatter in the data when the total Li+ content of the samples is taken into account. Therefore, with our high-quality, high-resolution X-ray data, we are able to deduce smooth trends in Li-occupation in the LLZO:Ga series.

For one composition (xGa = 0.20), the derived structural model as well as the cationic distribution were simultaneously refined against high-resolution neutron powder diffraction (NPD) and SC-XRD data. The best fit to the neutron data was indeed obtained with the I-43d model of this study, and the cationic distribution of X-ray data was proven. Results of refinement are displayed in Supplement Figure S1 and supplement Table S2. Also, in this combined simultaneous refinements, the La3+ site shows the presence of a small amount of vacancies. Again, there is no clear evidence for Ga3+ on the Li2 position, while it is enriched on Li1; in the combined refinement, the Ga3+ and Li+ occupation was refined freely without constraints: This strategy gave evidence for a rather low concentration of vacancies, both on Li1 (∼14%) and Li2 (∼12%) sites. The occupation of the Li3 site is somewhat higher than in the single-crystal study but still comparable within estimated standard deviations.

Crystal Chemistry

The substitution of Ga3+ into the structure of pure LLZO stabilizes the cubic structure. The lattice parameters thereby are reduced from 12.985 to 12.965 Å, and the decrease is almost linear, see Figure .

Figure 5

Correlation of lattice parameter a0 with the refined Ga content. The lattice parameter decreases with increasing Ga content. The phase transition from Ia-3d to I-43d does not cause a significant change in the lattice parameter.

For small Ga3+ concentrations, we find—as mentioned—the Ia-3d symmetry, while for refined Ga3+ contents xGa_ref > 0.07, a reduction in symmetry to SG I-43d takes place. The reduction in symmetry most probably is induced by the strong ordering of Ga3+ onto one of the two possible tetrahedrally coordinated Li sites, namely, onto Li1. The Li1 site in I-43d is smaller as compared to the corresponding Li1 site in Ia-3d, and it is also slightly smaller as compared to the Li2 site in I-43d, expressed by the smaller Li–O bond lengths and smaller volume. With increasing Ga3+ substitution, the Li1–O bond lengths successively decrease due to the smaller cationic size of Ga3+ as compared to Li+ in tetrahedral coordination (0.47 and 0.59 Å, respectively).[41] The Li2–O bond lengths remain constant, or increase slightly with increasing Ga3+ content (see Figure S2A, Supporting Information). This is seen as a further evidence that Ga3+ preferentially eenters the Li1 site (12a). Obviously, it is the smaller and more distorted character of the Li1 tetrahedron, which favors the Ga3+ substitution.

The Li3 site is a large cavity which has four Li–O bonds within 1.9 and 2.25 Å and two more distant Li–O bonds at 2.65 and 2.75 Å, and thus, one may consider the coordination polyhedron of the Li3 site as a strongly distorted octahedron. Each Li3 site shares two of its triangular faces with neighboring Li1 and Li2 sites, while both the Li1 and the Li2 site share all of their four triangular faces with Li3 sites. Thus, a three-dimensional network of face sharing Li-sites is present, which forms a diffusion pathway for Li+ ions. There are two different Li–Li distances within the Li-network: Li1–Li3 is the shorter one with distances of ∼1.6 Å, whereas the Li2–Li3 distance is around 2.3 Å. LaO8 and ZrO6 polyhedral volumes as well as Li–Li distances for selected samples are given in Table . Further crystal chemical considerations are given as Supporting Information.

7Li NMR Spectroscopy

In Figure , the 7Li NMR line shapes and line widths of the central transition of Ga20 and Al20 are shown. Considering the 7Li NMR central lines shown in Figure B it is seen that the lines are stepwise narrowed with increasing temperature. As an example, at 213 K (− 60 °C) the NMR signal of Ga20 with xGa = 0.20 is composed of two components viz. a broad Gaussian shaped line and a motionally narrowed Lorentzian one. Those ions contributing to the latter are already exposed to sufficiently fast exchange processes with rates exceeding the rigid-lattice line, the latter is estimated to be in the order of 8 to 9 kHz. At even higher temperature, see the spectrum recorded at 273 K (0 °C), all of the Li ions participate in sufficiently fast Li diffusivity to average dipole–dipole couplings. Above 273 K, the line widths (see Figure A) reached the so-called extreme narrowing regime.

Figure 6

(A) Motional narrowing of the 7Li NMR central line of Ga20 and Al20; the inset shows a magnification of the quadrupole intensities of the two NMR lines recorded at 223 K (−50 °C) by means of a nonselective solid-echo experiment. (B) 7Li NMR lines recorded at 116 MHz via single-pulse excitation.

Essentially the same behavior is found for the sample Al20 stabilized with 0.20 Al pfu. The corresponding motional narrowing curve, however, is slightly shifted toward higher temperatures indicating somewhat lower diffusivity in the sample containing Al3+. Looking at the quadrupole intensities, visualized by the solid-echo technique (see the inset of Figure A), the contribution to the NMR line of Ga20 is slightly reduced. The satellite transitions reflect the interaction of the spin-3/2 nucleus with the electric field gradients at the Li sites in garnet-type LLZO. A reduction in intensity (note that the spectra shown are scaled such that they have the same height) might be explained by faster Li exchange processes in Ga-bearing LLZO. At very high temperatures, full averaging of the satellite intensities is seen for the two samples.

In order to quantify ionic motion, we recorded 7Li NMR laboratory-frame (1/T1) and rotating-frame (1/T1ρ) spin–lattice relaxation rates. The rates obtained are shown as a function of the inverse temperature in a semilogarithmic plot in Figure A; in Figure B, selected magnetization transients of the experiments in the rotating frame are displayed. Solid lines represent fits with stretched exponentials to extract the rates 1/T1ρ. The inset in Figure A shows the temperature dependence of the stretching exponent γ. As temperature decreases, the transients become more stretched. Interestingly, the sample Ga20 reveals a local minimum of γ at ca. 200 K. As has been observed for other Li-ion conductors this feature might correspond to a (local) maximum in the 1/T1ρ(1/T) plot.[6,38,42] Indeed, this behavior is seen in Figure B, see the vertically drawn arrow. It could be interpreted as an additional Li ion diffusion process that is absent for the sample stabilized by Al3+ instead of Ga3+. For Ga20, an activation energy of 0.39 eV can be roughly estimated.

Figure 7

(A) 1/T1ρ magnetization transients, ranging from 233 to 373 K in steps of 20 K, which were analyzed by stretched exponentials (solid lines). Inset: variation of the stretching exponent as a function of the inverse temperature. The arrow points to a local minimum in the case of the Ga20 sample. (B) 7Li NMR spin–lattice relaxation rates of the two samples investigated.

Besides this slight difference observed by rotating-frame spin-lock NMR, the two samples show two marked similarities that were also observed in earlier studies focusing on garnets stabilized by M3+ cations:6 (i) 1/T1ρ passes through an extremely broad rate peak from which the high-T flank, characterized by a mean activation energy of 0.41 eV, is only partly accessible. On the other hand, (ii) up to ca. 400 K, the rates 1/T1(1/T) follow linear behavior in the Arrhenius plot pointing to activation energies of 0.13 to 0.14 eV. These values characterize local Li ion jumps in the garnet structure, whereas those deduced from 1/T1ρ might already be influenced by long-range ion transport.[43]

Discussion

It is generally accepted that the garnets and garnet-related (synthetic) materials can adopt both cubic and tetragonal symmetries.[44] The inorganic crystal structure database reports almost 500 entries for garnets, with the vast majority of 95% showing the cubic space group Ia-3d. Few natural garnets like henritermierite Ca3(Mn3+, Al)2(SiO4)2(OH)4 and synthetic materials such as LLZO are known to show tetragonal I41/acd symmetry.[4,6,8,45,46] It is accepted that symmetry breaking from cubic Ia-3d to lower symmetry takes place—among others—as a consequence of Jahn–Teller distortion for Mn3+ bearing garnets, strain, and growth effects, cation ordering and incorporation of hydrogen atoms.[44] In pure LLZO, the symmetry breaking is explained by the complete ordering of Li+ onto tetrahedral site.[46] In a recent paper, Galven et al. report on the symmetry change from Ia-3d to I213 during Li+/H+ exchange in the Li7–HLa3Sn2O12 and Li7–HLa3Nb2O12 systems.[47] For Li6-xHCaLa2Nb2O12, a change from Ia-3d to the acentric space group I-43d is described as a consequence of Li+/H+ exchange upon leaching in acetic acid for 4 days by Galven et al.[40] Both studies have been performed on polycrystalline samples using neutron diffraction. Galven et al. were not the first to report onto the unusual acentric cubic space group: the structure was first reported and solved by Lager et al. in the unusual mineral compound katoite Ca3Al2(O4H4)3.[39] Using single-crystal X-ray diffraction, the authors describe a phase transition from Ia-3d to I-43d occurring at high pressures (above 5 GPa).

In this study, we observed the acentric SG I-43d for Ga-stabilized LLZO samples with nominal Ga3+ contents xGa ≥ 0.15 pfu. This deviates from Al3+ stabilized LLZO prepared under the same conditions, which still shows SG Ia-3d. The present study indicates that for Li-oxide garnets, SG I-43d seems to be more common than expected. Ga-stabilized LLZO gets ordered in this acentric SG even without explicit aging or protonation. Previous studies reporting SG I-43d for Li-oxide garnets correlated this phase transformation with a protonation process.[40,48] As our samples were characterized immediately after the final calcination step, we exclude a phase transformation due to hydration caused by humidity from exposure to air as supposed by Larraz et al. for pure LLZO from I41/acd to Ia-3d.[49] Recently, Larraz et al. documented some additional weak reflections for pure LLZO that was stored for three years, washed, and then heated to 300 °C that cannot be explained using SG Ia-3d.[48] They already assumed that these reflections might be related to SG I-43d. In contrast to this, Ma et al. did not observe any phase transformation for cubic LLZO exposed to aqueous solutions and suggested that LLZO with SG Ia-3d is relatively stable, even at very high Li+/H+ exchange rates of 63.6%, which affected almost solely 4-fold-coordinated Li at the 96h position.[50] A particular focus of their study was on the presence of any other space groups such as I213 or I-43d, which was clearly denied by this paper. As the phase transition seems to be triggered by a splitting of the 24d position of SG Ia-3d, which is not affected by the Li+/H+ exchange, it is hardly imaginable that a phase transition can be caused by a Li+/H+ exchange at the 96h position only; so possibly other processes such as the heating to 300 °C might have caused the phase transition to I-43d mentioned by Larraz et al.

The additional reflections observed by Larraz et al. at 2θ = 21.5°, 40.3°, and 53.4°, respectively, can indeed be attributed to the (310), (530), and (710) reflections of SG I-43d, as these reflections are forbidden for SG Ia-3d. In this study, the reflection at 2θ = 21.65° (d = 4.101 Å) can be attributed to the (310) reflection of SG I-43d, which is expected to show a relative intensity of 1.2% compared to the strongest reflection at 2θ = 30.81° (d = 2.90 Å). The other additional reflections show an even lower relative intensity of 0.4% and 0.3%, respectively. These additional peaks might enable the identification of SG I-43d by means of XRPD. However, due to their low intensity, these peaks might have been overlooked by previous studies on Ga-stabilized LLZO or incorrectly attributed to LiGaO2, which is a common extra phase and shows a reflection at 2θ = 21.52° (d = 4.126 Å). For the detection of LiGaO2, the use of most intense reflection at 2θ = 22.55° (d = 3.94 Å) is more advantageous. For a definite determination of the SG of Ga-stabilized LLZO, single-crystal diffraction techniques are the most qualified methods.

Much experimental as well as theoretical effort has been undertaken to collect information on the site preference of Ga and its influence on Li-ion dynamics and Li-ion conductivity.[28−31] The position of these cations might influence the mobility of Li+ ions due to a possible blockage of the diffusion path. Several studies using 71Ga NMR spectroscopy lead to different interpretations, indicating either one or two Ga3+ positions. In this study, the refinement of SC-XRD and NPD data suggests that Ga3+ is preferentially located at a single position, namely the tetrahedral 12a position, and only minor amounts of Ga3+ occupy the 12b position.

As the coarse-grained LLZO samples in this study were prepared at high temperatures, we cannot assess whether Ga-stabilized LLZO prepared under different conditions (e.g., at lower temperatures) still orders in SG I-43d. Further research will be needed to clarify the influence of the preparation condition on the crystal structure of Ga-stabilized LLZO.

Despite that Ga-stabilized LLZO shows a different SG than Al-stabilized LLZO, the Li-ion mobility is still comparable or even better than for Al-stabilized LLZO with SG Ia-3d. It stands to reason that the structural differences will have a significant impact on the Li-ion dynamics. However, investigations on the influence of the new cubic modification on the Li-ion mobility are beyond the scope of this study and will therefore be reported in a subsequent paper.

Conclusions

The present study reveals that Ga-stabilized LLZO shows the acentric cubic SG I-43d (no. 220), which is different from other members of the Li-oxide garnet group that show SG Ia-3d (no. 230). In contrast to other studies which observed this SG due to Li+/H+ exchange, the structural changes might be caused by the site preference of Ga3+. The unit-cell parameter decreases slightly with increasing Ga3+ content. 7Li NMR relaxometry and line shape studies support the findings by impedance spectroscopy revealing enhanced ion dynamics in Ga-stabilized LLZO as compared to LLZO stabilized by Al. This study stimulates further research for an understanding of the structure−property relationship as a basis to improve the electrochemical capabilities of these electrolyte materials.