P. Jeevan Kumarae,
K. Nishimurab,
M. Senna*ac,
A. Düveld,
P. Heitjansd,
T. Kawaguchib,
N. Sakamotoab,
N. Wakiyaab and
H. Suzukiab
aResearch Institute of Electronics, Shizuoka University, 432-8561, Hamamatsu, Japan
bGraduate School of Engineering, Shizuoka University, 432-8561, Hamamatsu, Japan
cFaculty of Science and Technology, Keio University, 223-8522, Yokohama, Japan. E-mail: senna@applc.keio.ac.jp
dInstitute of Physical Chemistry and Electrochemistry, ZFM – Center for Solid State Chemistry and New Materials, Leibniz Universität Hannover, 30167, Hannover, Germany
eCentre for Nanomaterials and MEMS, Department of Physics, Nitte Meenakshi Institute of Technology (NMIT), Bangalore, 560064, India
First published on 17th June 2016
We present a novel approach to the solid-state synthesis of garnet-type cubic Li7La3Zr2O12 (c-LLZO) nanostructured particles with 1.0 mass% Al at 750 °C within 3 h. In contrast to conventional solid-state processes, a highly reactive precursor was prepared in two steps: (i) by homogenizing the stoichiometric mixture without Li, and (ii) subsequent addition of Li in the form of an ethanolic solution of lithium acetate. The actual composition determined by ICP analysis was Li6.61La3Zr2Al0.13O11.98. Sintering these nanoparticles at 1100 °C for 3 h in air after cold isostatic pressing brought a dense ceramic pellet with a relative density of 90.5%. The corresponding ionic conductivity with Au electrodes was 1.6 × 10−4 S cm−1 at room temperature. To study its electrochemical behavior as an electrolyte, a model cell of Li//(1 M LiPF6 + c-LLZO)//LiCoO2 configuration was constructed. Cyclic voltammetry of the cell delivered one set of redox couple with narrow voltage separation (15 mV) with a Li+ diffusion coefficient at room temperature of about 2 × 10−11 cm2 s−1 at the interface between LiCoO2 and 1 M LiPF6 + c-LLZO. The cell received an average discharge capacity of 64.4, 60.3, 56.1, 51.9 and 46.9 μA h cm−2 μm−1 at discharge rates 0.5C, 1C, 2C, 4C and 6C, respectively. The cell exhibited complete oxidation and reduction reactions with an average initial discharge capacity of about 64 μA h cm−2 μm−1, which is 92.7% of LiCoO2 theoretical value. These observations indicate the applicability of the present c-LLZO as an electrolyte for a solid-state Li-ion battery.
Within the variety of SSE including lithium superionic conductor (LISICON), thio-LISICON or sodium superionic conductor (NASICON), garnet-type c-LLZO is regarded as one of the most promising SSE due to its high ionic conductivity, stability with lithium and its less hygroscopic behavior.5,6 Weppner and coworkers have succeeded first in synthesizing c-LLZO by sintering at 1230 °C for 36 h and obtained a total (bulk + grain boundary) ionic conductivity of 2.44 × 10−4 S cm−1 at 25 °C.7 Thereafter, the material has received enormous research attention and a lot of work has been reported. LLZO garnet exhibits mainly two distinct phases, i.e. a tetragonal phase (t-LLZO) with low ion conductivity and a cubic phase with fast ion conductivity. The regulation of the Li site preference is found to be decisive for the phase purity of the cubic garnet structure and to yield fast ordered Li+ kinetics. This has been managed by the addition of few mol% of Al, Ga, Ta, Fe, Nb etc. for substituting Li and probably also some of the other cation species in LLZO. The reason(s) for this is(are) still under investigation. Ohta et al. reported that substitution of Nb for Zr in LLZO not only hinders the Li-ion motion but also lowers the Li-ion content more than all other additives. They found an ionic conductivity as high as 8 × 10−4 S cm−1.8 However, many reports suggested that Al or Ta is more preferable as compared to Nb from the viewpoint of chemical stability in the presence of Li.9 Recently, Bernuy-Lopez et al. reported an ionic conductivity as high as 1.3 mS cm−1 at room temperature (RT) by atmosphere controlled processing of Ga-doped LLZO.10 However, as very recently shown by single crystal studies by Robben et al. and briefly after by Wagner et al. Ga-doping is accompanied by a reduction of the LLZO cubic garnet symmetry.11,12 Buschmann et al. reported that Ga substituted LLZO/Li interface showed higher area specific interfacial resistance (ASR) (6000 Ω cm2) than 0.9 wt% Al substituted LLZO/interface (2800 Ω cm2).13 Geiger et al. also suggested that a minor concentration of Al in LLZO (0.10 to 0.15 Al per formula unit) as stabilizing agent could compensate the additional charge emergence due to Li+ vacancies and then reduce the free energy associated with ordering of the Li sub-lattice in LLZO core symmetry.14
Preparation of these oxide powders with complex garnet structure is mostly based on the conventional sol–gel or solid state syntheses, in which powders were calcined usually at temperatures above 850 °C. Xie et al.15 and Kokal et al.16 have succeeded in synthesizing phase pure c-LLZO nanopowders at lower temperatures of 750 °C, preceded by a solution and/or a sol–gel procedure. However, their product showed unfavorable ionic conductivities in the order of ∼10−6 S cm−1 at room temperature (RT). Zhang et al. claimed higher ionic conductivity of 3.3 × 10−4 S cm−1 at RT for nano-sized Al-doped LLZO powders prepared via a novel wet-chemical method combined with post heating at 900 °C.17 In an attempt to fabricate LIB with high capacity and reliability, Cheng et al. explored solid state synthesis of LLZO with controlled particle size.18 By combining mechanochemical processing and subsequent heating, Düvel et al. succeeded in preparing quasi phase pure c-LLZO.19 Table 1 exhibits the results reported on Al-substituted c-LLZO synthesized by conventional solid-state methods. Solid-state routes generally demand higher calcination temperatures and longer heating duration. Phase stability of c-LLZO is influenced by the Al and Li concentration, which are highly sensitive to prolonged heat treatment due to usage of Al2O3 crucibles and higher calcination temperatures, respectively.20 Simultaneous formation of a pyrochlore phase often takes place, leading to the inferior ionic behavior of the ceramic. Based on those facts, our attempts were focused on the reduction of the calcination temperature and duration. Once c-LLZO powders are successfully prepared, it is often necessary to sinter them to obtain a dense pellet or a bulk ceramics for all-solid lithium ion battery.21–23 Bulk sintered LLZO is also useful as a target for a physical vapor deposited thin film.24 For the preparation of high-density sintered pellets, various additives were incorporated.25 In order to increase the sinter density, using the starting mixture with smallest possible primary particles is another important viewpoint, since the associated high surface energy serves as an important driving force of sintering.26–28
Reference | Precursors | Calcination treatment | Relative density | Activation energy | Ionic conductivity (S cm−1) at RT |
---|---|---|---|---|---|
Sakamoto et al.39 | LiCO3,La(OH)3, ZrO2, Al2O3, no excess Li | 1000 °C for 4 h | 98% | 0.26 eV | 4 × 10−4 |
Kotobuki et al.22 | LiOH, La(OH)3, ZrO2, Al2O3 | 900 °C for 6 h | 2.4 × 10−4 | ||
Chen et al.35 | Li2CO3, La2O3, ZrO2, Al2O3, 10% excess Li | Two steps 900 and 1125 °C for 10 h each | 93% | 2.48 × 10−4 | |
Cheng et al.18 | Li2CO3, La(OH)3, ZrO2, Al2O3, no excess Li | 1000 °C for 12 h | 94% | 0.29 eV | 2.3 × 10−4 |
Buschmann et al.13 | LiOH, La2O3, ZrO2, Al2O3, 10–20% excess Li | Calcination at 1130 °C for 12 h | 0.34 eV | 4 × 10−4 | |
Allen et al.9 | Li2CO3, La(OH)3, ZrO2, Al(OH)3, 3 wt% Li excess | 1000 °C for 3–4 h | 98% | 0.30 eV | 3.7 × 10−4 |
Lee et al.23 | LiOH, La2O3, ZrO2, 10 wt% excess Li | 900 °C for 12 h in alumina crucibles | 86% | 0.34 eV | 4.9 × 10−4 |
Present work | CH3COOLi, La(OH)3, ZrO2, 1 wt% AlOOH, no excess Li | 750 °C for 3 h | 91% | 0.36(1) eV | 1.6 × 10−4 |
In this paper, we report a novel methodology to synthesize phase pure c-LLZO nanostructured particles at lowest possible temperatures. For this purpose, we propose a two-step reaction scheme for the preparation of the precursor. In the first step, a precursor of La and Zr, with their electronegativity slightly higher as compared to that of Li is prepared under mechanical stress. Subsequently, a stoichiometric Li source is added as an ethanolic solution. In the latter process, lithium can penetrate further to its ordering into the sub-lattice sites in Zr octahedral and La tetrahedral environments of garnet structures. Fig. 1 schematically illustrates our concept of preparing c-LLZO in comparison to the conventional solid-state process. We further observed microstructural evolution of nanoparticles on cold isostatic pressing (CIP) followed by conventional sintering. After determining the Li-ion conductivities of the sintered pellet at different temperatures, a prototype of Li-ion battery has been crafted using thin c-LLZO sintered pellet as an electrolyte in combination with the two polymer separators soaked into a non-aqueous electrolyte. Finally, the preliminary electrochemical performances were examined by using these prototypic LIB.
![]() | ||
Fig. 1 Conceptual representation of novel approach to solid state synthesis approach in contrast to conventional solid reaction method. |
The external morphology of the powder samples calcined and the cross-sectional view of the pellet sintered were observed by field emission scanning electron microscopy (FESEM) (Model JSM-700F, JEOL Ltd., Tokyo, Japan) with an accelerating voltage of 15 kV. For the measurements the samples were prepared by attaching the powder/broken pellet to an aluminum stub with silver paste, which then was coated with osmium by sputtering to prevent surface charging. The microstructure of calcined LLZO samples was investigated using scanning transmission electron microscopy (STEM) (Model JEM-2100F; JEOL Ltd.) equipped with energy dispersive spectroscopy (EDS) operating at an accelerating voltage of 200 kV. For TEM evaluations powder samples were dispersed in ethanol using ultrasound for 10 min and one droplet was dropped on a carbon coated grid and dried overnight before analysis.
The particle size distribution of nanoparticles was measured in ethanol solvent using an electrophoretic scattering photometer, i.e. dynamic light scattering (DLS) (Photal SELS-800Y; Otsuka Electronics Co. Ltd. Tokyo, Japan). Inductively coupled plasma optical emission spectrometry (ICP-OES) studies were performed to determine the concentration of trace elements in synthesized LLZO powders by Optima 2100DV (PerkinElmer, USA).
Temperature-dependent impedance spectroscopy (IS) measurements were conducted at temperatures ranging from 273 K to 513 K. A broadband dielectric spectrometer (Novocontrol, Concept 41) was used for the measurements in the frequency range between 10−2 Hz and 107 Hz with voltage ramp of 100 mV s−1. Electrochemical tests of the cells were performed in Swagelok type cells using a thin slice of LLZO as electrolyte with lithium foil (area = 0.8 cm2) negative electrode. RF magnetron sputter deposited LiCoO2 thin films on Pt/Si substrates have been used as working electrode. The active area of the working electrode was estimated to be 0.8 cm2 and its thickness was 1.6 μm. Li-ion conductive polypropylene porous separators were introduced in between electrode(s) and LLZO electrolyte to minimize the unavoidable electrolyte-to-electrode(s) interfacial resistance. These polymer sheets were actually soaked in 1.0 M LiPF6 in an ethylene carbonate (EC)–dimethyl carbonate (DMC) mixture (1:
1 ratio by volume), overnight. The cell designed was annotated as Li//(1 M LiPF6 + c-LLZO)//LiCoO2 cell. All cells were assembled in an Ar-filled dry glove box (O2, H2O below 5 ppm). The series of room temperature potentiostatic and galvanostatic electrochemical measurements such as cyclic voltammetry and chronopotentiometry were conducted with a Versastat 4 electrochemical analyzer (Tokyo corp., Japan) within the potential range 3.0–4.2 V and at various charge/discharge current rates.
To understand the phase purity of the calcined LLZO polycrystalline powders (750 °C for 3 h) and the corresponding sintered LLZO pellet, unpolarized micro-Raman spectroscopic studies were conducted in the wavenumber range from 100 to 550 cm−1. The results are plotted in Fig. 5A and B, respectively. In general, t-LLZO exhibits more complicated spectral features than the cubic phase either due to the lower symmetry of the tetragonal garnet (I41/acd space group) or due to more disordered Li+ ion arrangements, as summarized by Tietz et al.36 According to Orera et al.,37 Raman modes confined to the region between 330 and 600 cm−1 might be particularly sensitive to the Li-ion distributions and its changes in the symmetry. The Raman signals visible in the low frequency region, i.e. <150 cm−1, are mainly of La based vibrational bands. Fig. 5A and B clearly exhibit overall similarities in Raman excitations at wave numbers around 107, 121, 144, 209, 251, 361, 410 and 514 cm−1 for both, calcined and sintered, LLZO powders. No splitting of Raman bands was observed in the low-frequency region, i.e., at wavenumbers 107 (T2g) and 121 (Eg) cm−1 nor in Li+ ion bonding sensitive regions, i.e. at 361 (T2g), 410 (Eg/T2g) and 514 (T2g/Eg) cm−1, respectively. These observations indicate the phase purity of both the samples ascribed to the characteristic cubic garnet structure with I4d space group. Again, these observations were in good agreement with the ones reported in the literature.37 It is also noticed that the Raman spectrum of calcined LLZO precursor powder exhibited additional spectral features with minute Raman intensities at wavenumbers 155 and 190 cm−1 (Fig. 5A). These Raman signals were ascribed to the T2g and Eg symmetries of t-LLZO garnet structure which later disappeared after sintering the CIP pressed LLZO powders (Fig. 5B) at elevated temperature (1100 °C for 3 h). A slight blue shift in all Raman excitations was observed for the sintered pellet which may be due to the development of strain in the confined grain growth. Structural rearrangement might have built up for the tiny inhomogeneous calcined precursor powders upon thermal treatment. The presence of additional phase was not reflected in the XRD observations due to its low sensitivity to the light weight elements. However, Raman scattering measurements could be able to clearly distinguish the presence of various phases in LLZO that are difficult to detect by XRD since they are present in very small concentrations.
![]() | ||
Fig. 5 Unpolarized micro-Raman spectra obtained from polycrystalline LLZO (A) after calcining the precursor B at 750 °C for 3 h and (B) the sintered LLZO pellet at 1100 °C for 3 h. |
Fig. 6A and B show SEM images of the calcined LLZO powders at different magnifications. Fig. 6C shows the corresponding particle size distribution obtained from the DLS spectrum. In Fig. 6A and B, we observe the irregular block with the typical particle size distribution of 1–2 μm. At the same time, we recognize much smaller particulate units densely agglomerated. On the other hand, DLS measurements from the supernatant gave rise to an average size of ca. 90 nm. The average crystallite size estimated from the X-ray diffraction peak width was around 145 nm. This altogether gave us consistent information that our as-calcined LLZO powders have a primary particle size in the order of about 100 nm, are, however, agglomerated to particle sizes in the order of 1 to 2 μm. As shown in Fig. 6D, the particle sizes after sintering remain practically unchanged from those of agglomerated as-calcined powders, indicating that the sintering took place preferentially within the agglomerates.
The pellet density was estimated by weighing the pellets' mass and measuring their diameter and height. It is about 4.606 g cm−3 which is about 90.5% of c-LLZO theoretical density (5.098 g cm−3).32 For comparison, the relative density of the pellet was also examined by the conventional Archimedes method and is almost identical to the projected value. Morphological features at fractured surface of the sintered pellet are shown in Fig. 6D. From the figure it is clear, that there are residual pores between and in the crystallites. However, the sample seems to be quite dense and is in agreement with the calculated high relative density without significant voids. The fracture of the pellet appears to be transgranular, indicating the grain boundary strengthening effect. All the ceramic grains exhibited irregular disc shaped morphology with an average grain size estimated to be around 4 μm.
Fig. 7A–E presents the STEM images of the as-calcined LLZO powders and corresponding EDS elemental mapping. Results indicate a fairly homogeneous distribution of La, Zr, and Al. The actual cationic concentration for the powders synthesized estimated by ICP was Li6.61La3Zr2Al0.13O11.98, where the oxygen concentration was calculated to maintain charge neutrality. These observations consistently support the XRD and Raman results, demonstrating the cubic phase purity of the calcined powders.
![]() | ||
Fig. 7 (A) STEM images of the LLZO calcined powders and (B–E) area distribution of elements of Al, La, O and Zr measured by EDS, respectively. |
![]() | (1) |
![]() | ||
Fig. 8 AC conductivity of the sintered LLZO with Au blocking electrodes with respect to frequency at different temperatures. |
Typical complex impedance plots for LLZO at various temperatures ranging from 293 K to 513 K are presented in Fig. 9, in Nyquist representation. In both the cases, the experimental complex impedance plot consists of a depressed semicircle accompanied by a straight line on the low frequency side which is suggestive of electrolyte–electrode polarization at the Au blocking electrodes. The depression of the semicircle may be ascribed to the presence of distribution in relaxation times within the bulk response. Again, the radius of the depressed semicircle is observed to decrease with increasing temperature due to the increase in the conductivity of the sample. The equivalent circuit (physical model) appropriate to these plots is shown in the inset of Fig. 9. The impedance data at all the temperatures can also be fitted with the above equivalent circuit. The equivalent circuit consists of a CPE2, acting as a blocking double layer capacitance in series with the parallel combination of bulk resistance (R2) and constant phase element (CPE1), where CPE is generally considered as a leaky capacitor (i.e. hybrid between a resistor and a capacitor). Ideally the impedance assembly related to the process of charge transport in the bulk specimen is represented by parallel combination of Rb and Cb,8 where Rb represents the bulk resistance and capacitance Cb arises due to the electric relaxation process. But in this case we have replaced Cb with CPE1, which accounts for the observed depression of the semicircle and also the non-ideal electrolyte symmetry.9 CPE2 in the equivalent circuit accounts for the formation of the inclined straight line in the low frequency region. The ion conductivity at each temperature was obtained by using the relation
![]() | (2) |
![]() | ||
Fig. 9 Nyquist plots of the sintered LLZO with Au blocking electrodes around room temperature (inset: the equivalent circuit i.e. physical model appropriate to Nyquist plots). |
![]() | ||
Fig. 11 Schematic diagram of the hybrid electrochemical cell constructed in Li//(1 M LiPF6 + LLZO)//LiCoO2 configuration. |
The galvanostatic and potentiostatic measurements on the cell are displayed in Fig. 12A and B, respectively. A cyclic voltammogram of cell measured in the cut-off potentials range from 4.2 to 3.0 V with a slow scan rate of 0.05 mV s−1 is shown in Fig. 12A. The cell exhibited completely reversible electrochemical reactions with one set of redox peak currents located at 3.898/3.913 V with a potential separation of 15 mV vs. Li/Li+ reference electrode. The higher current densities at redox peak potentials with complete reversibility indicate the signature of reduced IR drop and faster Li+ kinetics of the hybrid cell designed with the combination of organic and crystalline electrolyte. These results are well congruent with the previous results reported on the Li//LiCoO2 cell constructed with conventional organic electrolyte.45 Like the results reported on Li//LiCoO2 organic cell, two sets of tiny shoulder like redox couples were 4.081/4.077 V and 4.174/4.163 V (labeled in Fig. 12A as dashed lines I, II) corresponding to a phase transition of positive electrode.46 As can be seen, besides higher ionic conductive LLZO solid electrolyte, the conventional organic electrolyte separator sheet had a lower cathodic/anodic interfacial resistance in the cell as well. Lower resistance of the hybrid combination of LLZO + organic electrolyte structure resulted in higher performance of the lithium cell. From the CV plot (Fig. 12A), the average Li+ ion diffusion co-efficient in the electrochemical reaction can be obtained by the following equation47
Ip = 2.69 × 105An3/2CoDL1/2υ1/2 | (3) |
The potential–capacity relationship after 5 cycles is shown in Fig. 12B. Its inset presents the C-rate performance of the hybrid cell discharged for about 50 cycles after initial 5 cycles. At the end of each cycle, a study time of 1 h was maintained. Thus, it is evident that the hybrid combination of the cell exhibits the best rate performance with an average discharge capacity of 64.4, 60.3, 56.1, 51.9 and 46.9 μA h cm−2 μm−1 at 0.5C, 1C, 2C, 4C and 6C, respectively. Even after, the cell performance was observed to be consistent, when the cell was continued to run at 0.5C rate and delivered an average discharge capacity of 64.4 μA h cm−2 μm−1 (∼128 mA h g−1). This value is about 92.7% of the theoretical capacity of LiCoO2.45 These observations are consistent with the work reported by Iriyama et al. on c-LLZO/LiCoO2 interface modification with thin layer of Nb.48
Thus, the present c-LLZO based prototypic LIB exhibits fairly higher discharge capacity with higher coulombic efficiency with the hybrid combination of liquid and solid interface. These encouraging results are probably due to the combination of crystalline/organic electrolyte which derived the enhanced electronic conductivity. These observations are in good agreement with the CV measurements i.e. voltage plateaus observed in charge discharge curves are well matched with redox peak potentials observed from the CV curve.
A fairly large capacity fade (∼0.5 μA h cm−2 μm−1 per cycle) as compared to the cell designed with organic electrolytes after 50 cycles (0.12 μA h cm−2 μm−1 per cycle)49 remains as a week point of our prototypic LIB. We actually observed some traces of LiCoO2 cathode peel off, after the electrochemical measurements. Development of an all solid state LIB, now being in progress in our group will be an option to solve the problem.
This journal is © The Royal Society of Chemistry 2016 |