Lithium Dissolution/Deposition Behavior of Al-Doped Li7La3Zr2O12 Ceramics with Different Grain Sizes

Lithium dissolution/deposition behavior of Al-doped Li7La3Zr2O12 (LLZ) was investigated in terms of grain size of LLZ-sintered bodies. One LLZ had smaller grain sizes of primarily less than 1 μm (LLZ_SG), and the other one had larger grain sizes of 5–20 μm (LLZ_LG). The total resistance of a Li symmetric cell using LLZ_SG was smaller than that using LLZ_LG at 100°C. The cell using LLZ_SG was stably cycled at 100°C without short-circuiting at a high current density of 1.3 mA cm−2, while the cell using LLZ_LG was stably cycled at current densities below 0.26 mA cm−2. Stronger bonding at grain boundaries for LLZ_SG was expected to contribute to the improvement of cyclability.

Conventional lithium-ion batteries based on organic liquid electrolytes have been widely studied, but their safety is still a concern. 1,2 Conversely, all-solid-state lithium batteries based on solid electrolyte have attracted attention for their safety. 3,4 Garnet-type Li-ion conductor Li 7 La 3 Zr 2 O 12 (LLZ) is believed to be a promising candidate as a solid electrolyte because of its high conductivity and chemical stability against molten lithium. 5,6 Lithium metal has a high theoretical specific capacity (3861 mAh g −1 ) and low redox potential (−3.045 V vs. SHE). Higher battery energy density is expected by the use of lithium metal anode. However, it is known that lithium dendrite penetrates liquid electrolyte. 7,8 The use of solid electrolyte has been proposed as a solution to this problem. Monroe et al. have suggested that Li dendrite should be suppressed if the shear modulus of solid electrolytes is more than twice that of metallic Li (∼3.4 GPa). 9 In the case of all-solid-state cells, it is expected that dense inorganic electrolytes with a high shear modulus can prevent the Li growth. Although LLZ, with a high shear modulus of ∼60 GPa, was applied as a solid electrolyte, Li penetrated LLZ, resulting in short-circuiting. 10,11 Therefore, there must be another factor that affects Li penetration besides the shear modulus. However, the factors remain unclear. It is considered that high LLZ/Li interfacial resistance leads to detrimental current focusing due to inhomogeneous current distributions, causing rapid initiation of lithium growth at interfaces. 12 Various attempts have been conducted to reduce the interfacial resistance. For example, Tsai et al. have used a thin buffer layer of Au to improve the contact between LLZ and Li. 13 Compared with the Li/LLZ interfacial resistance without Au buffer, the interfacial resistance dramatically decreased because of the formation of Li-Au alloys, resulting in good cycling stability. Similar phenomena have been reported by using Li alloys. [14][15][16] On the other hand, Cheng et al. examined the influence of grain size of a sintered body on Li/LLZ interfacial resistance and cycling stability of Li symmetric cells. 12 They prepared two types of sintered bodies; one was composed of grains with an average size of 20-40 μm, and the other one had grains of 100-200 μm. Li symmetric cells, using these sintered bodies, were fabricated and galvanostatic cycling tests were conducted. As a result, the cell containing LLZ with smaller grains performed much better in critical current density, overpotential, and cycling lifetime. They suggested that the difference in LLZ/Li interfacial resistances is attributed to the difference in grain sizes, which impacted the cycling stability. 12 However, the investigation of the relationship between grain size and cycle performance in a region of much smaller grain z E-mail: hayashi@chem.osakafu-u.ac.jp; tatsu@chem.osakafu-u.ac.jp size has not been studied. Very recently, we have investigated the relationship between grain size and conductivity using LLZ ceramics with much smaller grains than the previously reported LLZ. 17 We prepared two sintered bodies with different grain sizes; one had a larger grain size of 5-20 μm (LLZ_LG), and the other had smaller grain size of ≤1 μm (LLZ_SG); LLZ_SG exhibited higher ion conductivity than LLZ_LG.
In this study, the influence of grain size on Li/LLZ interfacial resistance and cycling stability was examined by using both LLZ_LG and LLZ_SG sintered bodies. Li symmetric cells using these sintered bodies were fabricated and galvanostatic cycling tests were conducted. After the cycling tests, the microstructures of the ceramics were analyzed.

Experimental
Two Al-doped LLZ powders with different particle sizes (DLZ-2 and Improved LLZ 2, Daiichi Kigenso Kagaku Kogyo Co., LTD.) were ball-milled for 45 min. The milled powders were pressed into pellets (diameter = 10 mm) under a pressure of 60 MPa. Furthermore, the pellets were isostatically cold-pressed under a pressure of 120 MPa. These green bodies were sintered at 1230 • C for 20 h, covered with mother powders of the same composition. Xray diffraction measurements (XRD, SmartLab; Rigaku Corp.) were performed with CuKα radiation. X-ray photoelectron spectroscopy (XPS) (Thermo Fisher SCIENTIFIC) of the LLZ surfaces was performed with an Al-Kα source (1486.6 eV). The observed binding energies were calibrated with the adventitious C 1s peak at 284.7 eV.
Symmetric Li/LLZ/Li cells were fabricated by sandwiching lithium metal foils of thickness 0.25 mm and stainless steel foils (current collector). To ensure good contact between metallic Li and LLZ, the cells were isostatically cold-pressed under a pressure of 100 MPa for 10 min. Alternating current (AC) impedance spectroscopy (Solartron SI-1260) was performed with an amplitude of 10 mV in the frequency range from 10 7 Hz to 0.1 Hz. For the Li dissolution/deposition experiment, square waves of current were applied to the cells at 100 • C using charge-discharge measuring devices (BTS-2004, Nagano Corp.). The current density was stepwise increased from 0.078 to 1.9 mA · cm −2 after every group of ten cycles for 1 hour and ten cycles for 2 hours. Scanning electron microscopy (SEM, SU8220, Hitachi Ltd.) observations were conducted at an acceleration voltage of 1.0 kV for the cross-section of the cycled cells.  Figure 1a shows the XRD patterns of LLZ_LG and LLZ_SG sintered bodies. Si powder was used as an external standard. No peaks of impurity phases were observed in either sintered body, indicating that pure cubic-LLZ was obtained. These grain sizes were estimated from the SEM images of thermally etched surfaces of each sintered body as shown in Figs. 1b and 1c: LLZ_LG had a larger grain size of 5-20 μm than LLZ_SG (grain size of ≤ 1 μm).

Results and Discussion
It is known that LLZ reacts with H 2 O to produce LiOH, followed by reaction with CO 2 to form Li 2 CO 3 . 18,19 Although the reaction layer is very thin, it is enough to dramatically increase the interfacial resistance and degrade cycling stability. [20][21][22] Hence, it is extremely important to remove Li 2 CO 3 layers from the surfaces of the LLZ sintered bodies. In this work, LLZ surfaces were first conditioned by an automated polisher (ML182, MKL-250, Maruto Instrument Co., Ltd.) using 9-μm, 3-μm, and 1-μm diamond paste in air. The pellets were then dry-polished using sand paper with grit number 3000 in an Ar glove box. The surface layers on LLZ are too thin to be detected by X-ray diffraction (XRD) or Raman spectroscopy.
In this study, XPS was used to compare the surface chemistry of LLZ just after polishing with the sand paper and the LLZ after storage for 1 week in the Ar glove box. Figure 2 shows C 1s and La 3d spectra for these LLZ_SG sintered bodies. Only one peak was identified at binding energy of 284.7 eV in the C 1s spectrum of LLZ just after polishing. The peak was due to adventitious carbon (CH n ) in the vacuum chamber. In contrast, two peaks were observed at binding energies of 284.7 eV and 289.7 eV after storage for 1 week in the Ar glove box. The latter peak was assigned to carbonate in Li 2 CO 3 . 23 Two sets of La 3d doublets, which were not observed for LLZ after 1 week of storage, appeared in the spectra of LLZ just after polishing. These data suggest that LLZ just after polishing has no surface layers. Figure 3 shows Nyquist plots of Li symmetric cells using LLZ_LG and LLZ_SG at (a) 25 • C and (b) 100 • C. Two semi-circles were observed at 25 • C. A semi-circle in a higher frequency region was assigned to LLZ electrolyte resistance (R LLZ ), which includes bulk and grain boundary components, and a semi-circle at lower frequency region was assigned to interfacial resistance (R interface ), on account of their capacitances. 24 The cell using LLZ_SG showed a lower R LLZ and R interface than that using LLZ_LG. There was little difference between the ionic conductivities of both ceramics; LLZ_SG exhibited a conductivity of 4.4 × 10 −4 S cm −1 , and LLZ_LG had a conductivity of 3.6 × 10 −4 S cm −1 . 17 Electronic conductivities of Al-doped LLZ are reported to be from 10 −8 to 10 −10 S cm −1 in previous papers. 25,26 Because the composition and density of the prepared LLZ_LG and LLZ_SG are almost the same as those of the reported LLZ, the electronic conductivities of them are expected to be negligible compared to the ionic conductivities. Furthermore, the thicknesses of LLZ_LG and LLZ_SG were 740-750 μm and 570-580 μm, respectively. Thus, LLZ_SG showed lower R LLZ because its thickness was smaller. Even if the thickness was the same, the cell using LLZ_SG would further lower total cell resistance (R total ) than that using LLZ_LG due to its lower R interface . The reason why R interface for LLZ_LG was much larger than that for LLZ_SG has remained unclear. Both surfaces of LLZ_LG and LLZ_SG were pretreated in the exact same manner; they were first conditioned with an automated polisher and dry-polished using a sand paper with grit number 3000 in an Ar glove box. There would be thus no difference in the surface conditions between LLZ_LG and LLZ_SG. On the other hand, R total was so small at 100 • C that resistance components could not be separated. R total of the cell using LLZ_SG was much smaller than that using LLZ_LG. The difference in R total at 100 • C as well as that at 25 • C is determined by the difference in R interface .
Galvanostatic cycling tests for the Li/LLZ/Li cells were conducted at 100 • C. Figure 4a shows the result of the test for Li/LLZ_LG/Li and its enlarged one is shown in Figure 4b. The voltage profile of the cell was stable at current densities less than 0.26 mA cm −2 . In contrast, it gradually increased and suddenly dropped to ∼0 V at a current density of 0.39 mA cm −2 . Figure 4c shows the results of Li dissolution/deposition behavior of the cell using LLZ_SG, and its enlarged one is shown in Figure 4d. It was stably cycled at a high current density of 1.3 mA cm −2 for 2 hours, corresponding to an areal capacity of 2.6 mAh cm −2 . Yonemoto et al. also conducted a Li dissolution/deposition test at 100 • C. 27 Li films with a thickness of 2.0 μm were plated and stripped at a current density of 0.1 mA cm −2 (0.41 mAh · cm −2 ). Compared with this result, the cell with LLZ_SG  showed better cyclability even at high current densities. It showed variable overpotential and sudden potential drop at a current density of 1.9 mA cm −2 . This behavior indicated failure of the cell. The difference in these cycling stabilities should be related to the difference in the interfacial resistances at Li/LLZ, as mentioned in Figure 3.  Figures. 5a and 5c, respectively. A web-like structure was observed on the cross section of LLZ_LG. The diameter of one cell in the web-like structure was estimated to be 5-20 μm, which is similar to the grain size of LLZ_LG. Considering that Li grows through grain boundaries easily, the web-like structure is probably composed of metallic Li. 28,29 On the other hand, a web-like structure was not observed on the cross section of LLZ_SG, suggesting that Li growth was suppressed. It is reported that Li penetrates through interconnected pores and grain boundaries. 28 Li probably deposited in those areas for LLZ_SG after short-circuiting, but a trace of Li penetration was not observed in Fig. 5c because the areas were a quite minor region. Li growth, therefore, depends on the difference in the microstructure of the grain boundary. Recently, various attempts have been conducted to reduce the interfacial resistance between LLZ and Li by inserting Au buffer layers, because it is assumed that large interfacial resistance influences Li growth. In addition, physically preventing the Li growth is also effective. We have reported that LLZ_SG had stronger bonding at grain boundaries than LLZ_LG, 17 suggesting that Li growth would be avoided by enhancing the bonding at grain boundaries in LLZ electrolytes.
In this study, we prepared LLZ ceramics with much smaller grains than that reported by Cheng et al. 12 LLZ ceramic with smaller grains had stronger bonding at grain boundaries than that with larger grains. Li barely grew through the LLZ with smaller grains, resulting in good cycling stability.

Conclusions
LLZ ceramics with different grain sizes were applied to Li symmetric cells, and the relationship between grain size and cycling stability was investigated. Galvanostatic cycling tests were conducted at 100 • C. The cell using LLZ_SG with submicron grains had a smaller interfacial resistance than the cell of LLZ_LG with 5-20 μm grains. LLZ_SG was more stably cycled than LLZ_LG. A web-like structure consisting of dendritic Li was observed on the cross section of LLZ_LG after short-circuiting. In contrast, the morphology of the cross section of LLZ_SG was not changed by cycling, suggesting that the use of LLZ_SG with stronger bonding at grain boundaries is effective in preventing the Li penetration. We believe that these results lead to further understanding and improvement of all-solid-state lithium batteries with a lithium metal negative electrode.