CO2 exposure

Journal of Power Sources 275 (2015) 612e620

Contents lists available at ScienceDirect

Journal of Power Sources journal homepage: www.elsevier.com/locate/jpowsour

Phase transition in lithium garnet oxide ionic conductors Li7La3Zr2O12: The role of Ta substitution and H2O/CO2 exposure Yuxing Wang, Wei Lai* Department of Chemical Engineering and Materials Science, Michigan State University, East Lansing, MI 48824, USA

h i g h l i g h t s
Measures were taken to eliminate/minimize Al-contamination and air exposure.
In-situ impedance measurement to probe phase transition temperature of LLZ.
A minimum of 0.6 mol of Ta is needed to stabilize the cubic phase LLZ.
H2O interacts with LLZ through a proton exchange mechanism.
CO2 interaction with LLZ worsens the grain boundary conduction.

a r t i c l e i n f o

a b s t r a c t

Article history: Received 8 September 2014 Received in revised form 11 November 2014 Accepted 12 November 2014 Available online 13 November 2014

High Li-content lithium garnet oxides are promising solid electrolyte materials for lithium batteries. Being the highest Li-content lithium garnet oxides, Li7La3Zr2O12 has been reported to crystallize in either the tetragonal or cubic phase with no consensus on the exact conditions under which these two phases are formed, which may be due to unintentional Al contamination and air exposure. In this work, the effects of Ta substitution and H2O/CO2 exposure have been studied under Al-contamination free conditions with minimal air exposure. We showed that 1) the Ta-substitution induced phase transition occurred through a two-phase mechanism and a minimum 0.6 mol of Ta substitution to Zr is needed to stabilize the cubic phase; 2) H2O and CO2 can individually induce the tetragonal-cubic phase transition in Li7La3Zr2O12 through proton exchange and Li extraction, respectively, which can have great influence on the transport properties of Li7La3Zr2O12. © 2014 Elsevier B.V. All rights reserved.

Keywords: Garnet oxide Li ion conductor Phase transition Ta substitution Proton exchange Transport property

1. Introduction The lithium garnet oxide ionic conductors have become one of the most promising solid electrolyte candidates [1e3] for the potential application in all-solid-state, LieS, and Li-air batteries due to their high ionic conductivity and good electrochemical stability [4e9]. The capability of Li transport in the garnet system was discovered by Weppner et al. in 2003 in Li5La3Ta2O12 [4]. Later, another garnet compound Li7La3Zr2O12 (LLZ) was found to possess very high ionic conductivity (5  104 S cm1) [6]. The LLZ was first identified to crystallize in the cubic symmetry (Ia3d, No. 230). However, single crystal X-ray and neutron diffraction suggested that the LLZ crystallizes in the tetragonal symmetry (I41/acd, No.

* Corresponding author. E-mail address: [email protected] (W. Lai). http://dx.doi.org/10.1016/j.jpowsour.2014.11.062 0378-7753/© 2014 Elsevier B.V. All rights reserved.

142) [10]. The tetragonal I41/acd space group is a subgroup of the cubic Ia3d space group. A complete ordering of Li-vacancies occurs in the tetragonal LLZ accompanying the symmetry reduction. As a result, the ionic conductivity of the tetragonal LLZ is about 2 orders of magnitude lower than the cubic counterpart. Many studies were devoted to the synthesis and characterization of LLZ due to the high ionic conductivity of the cubic phase. The phase information from these studies is summarized in Table 1 [6,10e19]. Clearly, considerable inconsistencies exist. It has been found that cation substitution can stabilize the cubic phase by creating extra vacancies and lowering the Li content [16,20,21]. A large number of various garnet compounds have been synthesized by deliberate cation substitution and the best garnet-type Li ion conductors so far are derived from cation substitution of LLZ [7,9,20,22e24]. Therefore, the knowledge of the minimum substitution level for the stabilization of the cubic structure is crucial. The minimum substitution level of Al or Ga was experimentally

Y. Wang, W. Lai / Journal of Power Sources 275 (2015) 612e620 Table 1 A summary of literature reports on Li7La3Zr2O12. Author

Synthesis temperature

Phase

1 2 3 4 5 6 7 8

Murugan et al. Shimonishi et al. Awaka et al. Janani et al. Xie et al. Awaka et al. Wang et al. Geiger et al.

1230  C 1180  C 1250  C 700  C 750  C 980  C 1000  C 1100  C

9

Kokal et al.

10

Buschmann et al.

<705  C >774  C 1130  C

11

Larraz et al.

980  C

Cubic Cubic Cubic Cubic (Al free) Cubic (Al free) Tetragonal Tetragonal Cubic (Al2O3 vessel) Tetragonal (Pt vessel) Cubic Tetragonal Tetragonal (Al free) Cubic (Al-doped) Tetragonal (fresh sample) Cubic (aged sample)

determined to be 0.2 mol per formula unit [20], which corresponds to 0.4 mol of extra vacancies as the trivalent Al3þ/Ga3þ enters the Li site creating twice as many vacancies. Ta or Nb substituted LLZ were reported to crystallize in the cubic phase with substitution level of less than 0.2 mol (each substitution creates one vacancy) [7,9]. A recent study by Thompson et al. concluded that the critical Ta substitution level is 0.5 mol for Al-free samples [25]. It is suspected that these inconsistencies stem from contamination during the garnet sample preparation. A typical procedure involves high-temperature firing in Al2O3 vessels in air. The unintentional introduction of Al from the vessel was claimed to be responsible for the observed cubic phase in some reports [11,16]. Nevertheless, Al-free cubic LLZ has also been reported [13,14]. It was recently found that the tetragonal LLZ can be transformed into cubic phase through aging in air [19]. Similar tetragonal-cubic phase transformation was also observed in naturally aged Li7La3Sn2O12 samples [26]. Therefore, it is crucial to control both the Al contamination and air exposure during the synthesis in order to truly understand the phase transition of LLZ. However, no such attempts have been made so far. In addition, it is unclear how the exposure to H2O/CO2 affects the transport properties of Li conducting garnets. This knowledge is of great technological significant for the application of garnet-type solid electrolytes in lithium batteries. It should be noted that aforementioned reports on the role of H2O and CO2 employed a post-mortem approach, i.e., the analysis was done on naturally aged sample, so the H2O/CO2 effect were neither separated nor studied in real-time. In this study, we take measures to eliminate the Al contamination and air exposure during the garnet sample preparation. These samples were confirmed to be contamination-free by thermal gravimetric analysis (TGA) and Energy-dispersive X-ray spectroscopy (EDS). The minimum substitution level of Ta to stabilize the cubic structure is determined by powder X-ray diffraction (PXRD). The effects of H2O and CO2 separately on the phase transition and transport properties of porous LLZ samples are investigated through in-situ impedance spectroscopy.

loss during the synthesis. 5 mm yttria-stabilized zirconia balls were used as grinding media. The mixture was then dried and transferred to MgO crucibles for calcination at 900  C for 10 h with a heating rate of 3  C min1 and cooling rate of 2  C min1. To eliminate the air effect, all powders were heat-treated in a tube furnace at 750  C for 2 h under constant Ar flow. A MgO tray was used to hold the garnet samples. The samples were transferred to an Ar-filled glovebox with moisture level <0.1 ppm for storage immediately after the heat-treatment. To prepare porous LLZ pellets, the dried starting mixture was pressed into pellet form with stainless steel dies and a hydraulic press. The pellets were sintered at 1050  C for 24 h in MgO crucibles. The density and open porosity were measured with the Archimedes method. All samples have about 30% of porosity. Gold paste was applied to both sides of the pellet as blocking electrodes. A 700  C heating step was employed to improve adhesion between the garnet pellet and gold electrodes. The phase of the garnet powders and pellets were characterized by XRD using CuKa radiation operated at a voltage of 40 kV and current of 40 mA (Bruker D8 ADVANCE). The measurement range was 10 e70 with a step interval of 0.02 and a scan rate of 0.04 s1. Rietveld refinement implemented by the GSAS [27,28] was employed to extract the lattice parameters and relative phase fractions. TGA (TA instrument, TGA 500) with a temperature range of room temperature to 800  C under N2 flow was performed on the garnet samples. The elemental information was characterized by EDS (Carl Zeiss, EVO LS25). A setup combining a tube furnace and a potentiostat (Bio-Logic SP-200) as shown in Fig. 1 was used for the in-situ impedance measurement of LLZ pellets under various gas flows. Prior to the measurement, the pellets were heat-treated in Ar at 750  C for 2 h. This step ensures that the whole system is free of H2O/CO2. The tube was then cooled to measurement temperatures. The system was allowed to equilibrate before turning on the H2O or CO2 gas flow. Ar gas was passed through a bubbler to generate the moisture gas flow. A tank of compressed Ar with 5% CO2 was used to supply the CO2 flow. A sinusoidal voltage signal with amplitude of 20 mV was applied for the frequency range of 5 MHz to 1 Hz.

3. Results and discussion 3.1. Sample purity The purity of garnet samples can be severely compromised if the Al-contamination and air exposure are allowed during the synthesis. We highlight two measures taken to eliminate these effects: the use of MgO crucible and heat-treatment in Ar. In our prior experience, discoloration of the garnet samples is commonly observed especially when heated to high temperatures (>1000  C) if Al2O3 crucible were used. Using MgO crucibles, no discoloration was observed for the sintered LLZ pellet at 1050  C indicating very little or no interaction between the samples and crucibles. This is confirmed by EDS analysis (Fig. 2) which showed no observable

2. Experimental The garnet powder Li7-xLa3Zr2-xTaxO12 (LLZTx, x ¼ 0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6) was prepared by the solidestate reaction method. LiOH$H2O (AlfaeAesar) was dried at 200  C before use. A stoichiometric amount of LiOH, La2O3 (SigmaeAldrich), ZrO2 (AlfaeAesar) and Ta2O5 (AlfaeAesar) were wet-milled on a roller mixer for 24 h in polyethylene jars filled with isopropyl alcohol. 10% excessive lithium precursor was added to compensate for lithium

613

Fig. 1. Illustration of In-situ impedance measurement setup.

614

Y. Wang, W. Lai / Journal of Power Sources 275 (2015) 612e620

Fig. 2. (a) EDS spectrum of Li7La3Zr2O12 powder prepared in MgO crucibles. (b) TG curves of Li7La3Zr2O12 powders as prepared in air and heat-treated in Ar at 750  C.

peaks corresponding to the Mg K or Al K radiation. In comparison, the sample prepared by Shimonishi et al. [11] at 1180  C in Al2O3 crucibles contained 0.23 mol of Al; the sample prepared by Geiger et al. [16] contained around 0.2 mol of Al most likely coming from the preheating step of La precursor at 1000  C in Al2O3 crucible. Our experiments clearly demonstrate the superiority of MgO crucible for the preparation of low-contamination garnet samples. To test the effectiveness of the heat-treatment in Ar, TGA results on the heat-treated LLZ and as-prepared LLZ (in air) are compared (Fig. 2). Both curves feature three weight loss events at around 250  C, 400  C and 600  C, indicating the same nature of interaction with air for the heat treated and the as-prepared sample but different in the extent of interaction. The TG curves are in qualitative agreement with the report by Larraz et al. [19] who attributed the three weight loss steps to the release of H2O, H2O þ CO2, CO2, respectively. For the as-prepared LLZ (but stored in inert atmosphere), there was about 1% weight loss upon heating to 800  C. Note that although the quantity of H2O and CO2 is less than that in samples naturally aged in air reported by Larraz, interaction with air is inevitable at the calcinations step as garnet samples cool down, which could have significant impact on the Li content. In comparison, the heat-treated LLZ had only about 0.1% weight loss upon heating to 700  C thus the absorption of H2O and CO2 may be neglected. The weight loss above 700  C may be related to decomposition of excessive Li2CO3. TGA and XRD results show that the synthesis procedure employed in this study is capable of eliminating external factors (cation contamination and air exposure) that complicate the phase identification of garnet

compounds. It is interesting to note that for both samples, there was a noticeable weight gain below 200  C which can be attributed to the uptake of H2O or CO2 during the TG measurement.

3.2. Minimum substitution level of Ta for the stabilization of cubic LLZ The powder XRD patterns of seven garnet compounds Li7(x ¼ 0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6), are shown in Fig. 3. The pattern of the Ta-free sample (x ¼ 0) matches well with that in the literature [10], suggesting a pure tetragonal phase is formed. The lattice parameters (Table 2) are nearly identical to those determined from single crystal measurement. The pattern of the Li6.4La3Zr1.4Ta0.6O12 sample is typical of cubic garnet phase and the lattice parameter is again in good agreement with literature report [22]. Pure tetragonal and cubic phase were formed for the LLZ and LLZT0.6 samples, respectively. For the intermediate compounds (x ¼ 0.1e0.5), the patterns contain peaks from both tetragonal and cubic phases. No impurities were observed in all patterns. As the tetragonal I41/acd is a subgroup of the cubic Ia3d space group so the peaks in the tetragonal pattern can be derived from splitting the corresponding cubic peaks. For instance, the (400) peak in the cubic phase is split into (400) and (004) peaks which are about 1 apart in 2q for CuKa radiation. As seen from the zoomed-in region of cubic (400)/tetragonal (400)&(004) peaks, it is possible to separate the tetragonal and cubic peaks despite slight overlapping. Upon Ta-doping, the ratio of cubic peak and tetragonal

xLa3Zr2-xTaxO12

Fig. 3. (Left) PXRD patterns of Li7-xLa3Zr2-xTaxO12 (x ¼ 0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6). (Right) Zoomed-in patterns of the encircled region.

Y. Wang, W. Lai / Journal of Power Sources 275 (2015) 612e620

615

Table 2 Rietveld refinement results of Li7-xLa3Zr2-xTaxO12 (x ¼ 0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6) from powder XRD data.

Rwp

c2

Cell parameters (I41/acd, Å) Cell parameters (Ia3d, Å) T-phase fraction

a c

LLZ

LLZT0.1

LLZT0.2

LLZT0.3

LLZT0.4

LLZT0.5

LLZT0.6

18.2 1.55 13.1211 12.6660 e

23.9 1.57 13.1075 12.6831 12.9432

16.2 1.47 13.1126 12.7172 12.9680

14.6 1.51 13.0894 12.7116 12.9430

15.0 1.45 13.0861 12.7271 12.9416

19.9 1.32 13.0672 12.7592 12.9269

14.6 1.50 e e 12.9263

0.819

0.654

0.410

0.313

0.154

1

peaks intensity increases monotonically. The positions of these peaks are largely independent of the compositions. Rietveld refinement was performed on all samples in order to obtain the relative phase fractions and lattice parameters. For the LLZ and LLZT0.6, the tetragonal phase model proposed by Awaka et al. [10] and a cubic phase 24d-48g model [15] were used, respectively. A two-phase model was used for the intermediate compounds. For the tetragonal models, all three Li sites (8a, 16f, 32g) are fully occupied; for the cubic models, as XRD is insensitive to Li atoms, the site occupancies of Li were not refined. Instead, an estimated site occupancy of 0.7 for the 24d site was used [14]. It has been demonstrated that the use of anisotropic atomic displacement factors for octahedral Li and La atoms improves the fitting [15,29], however, for the current laboratory x-ray data, the improvement is insignificant. Therefore, isotropic ADPs were used for all atoms. The lattice constants of the tetragonal and cubic phases were also used as fitting parameters, to accommodate possible effects of interfacial regions on the lattice distortion due to the mismatch in elastic moduli or due to the extra interfacial energy. The refinement results are summarized in Table 2. In Table 2, the a/c ratio of the tetragonal phase decreases with increasing Ta contents, indicating that the tetragonal phase becomes more cubic. The goodness of fit was low for all samples whereas the relatively large Rwp values are due to low counting statistics of the laboratory x-ray diffractometer. For the purpose of determining lattice parameters and relative phase fraction, we consider the quality of fit reasonable (Fig. 4). The fraction of tetragonal phase can be predicted assuming a two-phase mechanism with LLZ and LLZT0.6 being the end members. The experimental values determined by Rietveld refinement are in good agreement with the predicted values except for x ¼ 0.3 (Fig. 5). Cation ordering is a common phenomenon in host structures [30e34]. Simple lattice models have been employed to successfully predict the phase stability as a function of composition and temperature [35]. The order-disorder transition can be explained by minimization of configurational energy characterized by pair-wise interaction. In many systems, the ordering of cation-vacancy occurs at a small range near a particular cation concentration. For instance, in the layered form of LixCoO2, Li ordering occurs for 0.45 < x < 0.55 (near 0.5) according to experimental evidences [31] and firstprinciples investigation [35]. Accompanying the ordering of Li, a lattice distortion drives the hexagonal-monoclinic phase transition. In Li conducting garnets, the Li concentration is tuned by cation substitution. There are a total of 72 Li sites in a unit cell including 24 tetrahedral sites and 48 octahedral sites. When Li content is at 7 (56 Li in a unit cell), Li ions are ordered so 1/3 of tetrahedral sites and all octahedral sites are occupied. In this type of arrangement, tetrahedral sites are occupied/vacant alternately which avoids mutual occupation of adjacent Td-Oh-Td sites thus short LieLi interaction. Similar to LixCoO2, lattice distortion occurs causing the cubictetragonal transition in the lithium garnet oxides. When the Li content sufficiently deviates from 7, the disordered/cubic phase dominates. This explains the effect of Ta-doping on the phase

0

transition of LLZ. However, there is no consensus on the minimum amount of cation substitution for the stabilization of the cubic phase. MD simulation based on DFT calculation has shown that the tetragonal/ordered form of LLZ is stable at room temperature and a minimum 0.4 mol of vacancy is needed for the cubic/disordered form to be stable [36]. This is largely consistent with previous experimental evidences [20,25]. However, there is no lacking of reported cubic phase garnets with much less cation doping levels [9,20,22]. Again, the obtained cubic phase could be a result of poor control of actual compositions (Al contamination or air effect). Our

Fig. 4. Observed (þ), calculated (red line) and difference patterns for the Rietveld refinement from the powder XRD data of the nominal composition Li6.7La3Zr1.7Ta0.3O12. The short vertical lines correspond to the all possible peak positions of the tetragonal phase (magenta, upper) and the cubic phase (light blue, lower). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 5. Compositional dependence of the tetragonal phase fraction.

616

Y. Wang, W. Lai / Journal of Power Sources 275 (2015) 612e620

present results showed that 0.6 mol of Ta substitution is more realistic for the stabilization of the cubic phase. This new experimental evidence agrees with our previous atomistic simulation [15] where we have shown a tendency of Li ordering in LLZTx series with 0.5 mol of Ta substitution. 3.3. Intrinsic phase transition A porous LLZ pellet was used for the impedance measurement in the temperature range of 200  C to 750  C with a complete coolingeheating cycle. Fig. 6 shows the impedance plots of the cooling cycle. For the ‘low’ temperature datasets (200  C and 300  C), a portion of a semicircle corresponding to the bulk process can be seen at the high frequency end. For high temperature datasets, no semicircles were observed for the bulk processes due to the decrease of bulk resistance thus smaller time constant; instead, the intercept at the real axis roughly corresponds to bulk resistance. The bending below the real axis was due to the inductance effect of silver wires (roughly 0.5 m) used as test leads (Fig. 1). At temperatures above 630  C, the intercepts move slightly to the right upon cooling; from 630  C to 610  C, the intercept value more than doubled. This sudden change of electrical properties is clear indication of phase transition between 630  C and 610  C. We refer to this type of phase transition as ‘intrinsic’ transition because it involves no change of stoichiometry (or Li content); rather it is the consequence of increasing contribution of the entropic term to the overall free energy with increasing temperature. A direct consequence of the cubic-tetragonal phase transition upon cooling is the sharp drop in Li ionic conductivity as Li becomes ordered. The experimental data were modeled using two equivalent circuits shown in Fig. 6. The grain boundary semicircle was not observed for the current temperature range; in fact, it only appeared upon further cooling from 200  C. The L and Cstray account for the inductance and stray capacitance effects of the test leads. The test leads resistance was determined to be 0.2 U from a short circuit and subtracted. Rb is the bulk resistance. A constant phase element Qb which takes the form of [(jw)nQ]1 was used for the ‘low’ temperature data to model the bulk capacitance. Other elements are associated with various processes at the electrodes. The Arrhenius plot of LLZ is shown in Fig. 7. For the cooling cycle, all the data points sit on a straight line above 630  C. Between 630  C and 610  C, the conductivity decreases sharply, indicating the cubic-tetragonal phase transition. In comparison, Matsui et al. determined the phase transition temperature to be between 600  C and 650  C through high-temperature XRD and impedance spectroscopy [37]; Larraz et al.’s high-temperature XRD results showed

Fig. 7. Arrhenius plot of Li7La3Zr2O12. The data and linear fit are shown as open circles and lines.

Fig. 8. Activation energies of cubic phase garnet compounds as a function of Li content.

that the phase transition temperature to be around 650  C [19]. The MD prediction of the phase transition temperature was between 800 K and 1000 K [36]. The conductivities show good linearity below 500  C. The heating cycle data largely match the cooling cycle, although a clear hysteresis was observed which is typical of first-order phase transition. The activation energies of the cubic phase (Eac ) and the tetragonal phase (Eat ) are determined to be 0.21 eV and 0.41 eV, respectively, using the Arrhenius equation s ¼ s0/T exp(Ea/kT). The activation energy of the tetragonal phase is lower than reported value (0.54 eV) in literature [10]. There was only one report of activation energy of the cubic phase LLZ at high temperatures [37]. The reported value 0.0117 eV is much lower than

Fig. 6. Impedance plots of Li7La3Zr2O12 samples cooled from 750  C to 200  C under Ar flow: (a) 500e200  C; (b) 750e550  C. A selected range of frequencies instead of the whole measured range (5 MHze1 Hz) is shown for clarity. The two equivalent circuits used are shown in (a). The upper and lower circuits are for high temperature (above 400  C) data and ‘low’ temperature (200  C and 300  C) data, respectively.

Y. Wang, W. Lai / Journal of Power Sources 275 (2015) 612e620

our result. However, another form of Arrhenius equation (s ¼ s0exp(Ea/kT).) was used in their report. If converted to the form in equation (1), the activation energy is 0.2 eV which is closer to our result. Fig. 8 compares the Eac with other cubic garnet compounds of lower Li content [4,7,9,22,38e40]. It is interesting to note the trend that the activation energy decreases as Li content increases almost as a linear fashion. A tentative explanation to this trend is provided. Assuming the change of lattice parameters is insignificant to cause large change in the bottleneck size, the energies at the transition state are similar regardless of the Li content. Since the number of available sites for Li is fixed in the garnet structure, an increase in the Li content results in the increase of overall LieLi interaction energy which destabilizes the Li sites. Therefore, the activation energy which is the difference between the site energy and energy at the transition state decreases with increasing Li content.

3.4. H2O-induced phase transition In-situ impedance measurements of LLZ pellets under moisture gas flow were carried out at various temperatures to evaluate the effect of H2O on the transport property and phase transition of LLZ. The impedance plots at 150  C, 200  C and 250  C are shown in Fig. 9(aec). The high-frequency arcs in all the plots correspond to the bulk resistance. The bulk resistance values were fitted using equivalent circuits shown in Fig. 6 and it's time dependence is plotted in Fig. 9(d). To complement the impedance results, we characterized the phase information of LLZ upon H2O exposure through ex-situ XRD and weight loss (corresponding to amount H2O uptake) of exposed sample through TGA (Fig. 10). There are several possible mechanisms of interaction between LLZ and H2O: 1) physical absorption; 2) hydration; 3) proton exchange of Li. We eliminated the possibility of physical absorption of H2O by conducting the experiment at temperatures at least 150  C, which is

617

confirmed by the absence of weight loss below 200  C in TGA. The presence of hydration water is highly unlikely because no major change in the garnet framework or the lattice parameters were observed in the XRD. Also, there has been no report of hydration water in garnets to our knowledge. We propose the following proton exchange reaction under the current experiment conditions. Li7La3Zr2O12 þ x$H2O 4 Li7-xHxLa3Zr2O12 þ x$LiOH

(1)

There have been reports of Liþ/Hþ exchange in lithium garnets in proton exchange medium [26,41,42]. In addition, the existence of hydrogarnets, for instance, Ca3Al2(O4H4)3 through proton exchange of Ca3Al2Si3O12, is widely known [43]. We will show next that the proton exchange reaction is supported by our experimental evidences. As discussed in section 3.2, the Li ordering occurs at a narrow composition range as a result of configurational energy minimization. The proton exchange of Li, first of all, lowers the Li content. Secondly, since the protons stay much closer to O atoms than Li, the introduction of Hþ greatly alters the charge distribution of the oxygen framework [44]. Therefore, the protons in the tetragonal LLZ will disrupt the Li-ordering and induce the tetragonal-to-cubic phase transition. Indeed, the XRD patterns of LLZ after passing moisture for 1 h at 150  C (Fig. 10(a)) showed that part of the tetragonal LLZ phase was transformed to cubic phase; after 20 h, the tetragonal phase was completely transformed to cubic phase. Note that in the absence of H2O, the phase transition occurs at a much higher temperature (between 610  C and 630  C). The presence of LiOH peaks in the 20 h sample XRD corroborates the proposed reaction. The XRD patterns for 200  C and 250  C samples, Fig. 10(b and c), show lesser degree of tetragonal-cubic phase transition. The increase of cubic phase fraction coincides with increasing H2O uptake as seen from the increasing weight loss in TG results (Fig. 10(d)). The H2O uptake due to the conversion of LiOH

Fig. 9. Impedance plots of Li7La3Zr2O12 under moisture flow at (a) 150  C, (b) 200  C and (c) 250  C. (d) Bulk resistivity evolution with time at all three temperatures.

618

Y. Wang, W. Lai / Journal of Power Sources 275 (2015) 612e620

Fig. 10. (aec) XRD patterns and (d) TGA curves of Li7La3Zr2O12 pellets upon exposure to Ar/H2O at different temperatures.

(product of the proton exchange reaction) to LiOH$H2O can be ruled out since the water vapor pressure of LiOH$H2O is calculated to be 4 bar at 150  C [45]. Therefore, the release of H2O in TG measurement can be solely attributed to the proton removal reaction. The sample exposed to H2O for 20 h at 150  C shows 9% weight loss upon heating to 800  C corresponding to 4.7 mol of H2O per formula unit. The extent of proton exchange that is sufficient to induce tetragonal-cubic phase transition has not been studied, but it is highly likely that the amount of protons in this sample is much larger than the minimum level for stabilizing the cubic phase. The amount of H2O uptake decreases with increasing temperatures as seen from TG results on the 200  C, 30 h or 250  C, 20 h samples. This can be explained by the shifting of the proton exchange reaction equilibrium. In all samples, the weight loss begins at around 200  C. Therefore, at temperatures above 200  C, a dynamical equilibrium will be established and higher temperature favors the reverse direction. The proton exchange mechanism is also supported by the impedance evolution. As seen in Fig. 9, at all three temperatures, the bulk resistance first dropped upon passing moisture gas. This is because the small amount of protons partially disrupt the Li ordering, activating the fast conduction mechanism of cubic Li garnets. As the exposure time increased, the bulk resistivity began to increase after reaching a minimum. The increase of bulk resistivity is more significant for low temperature data sets, which is consistent with greater extent of proton exchange as seen from the TG results. For the 150  C sample, the bulk resistivity at 20 h increases by more than four orders of magnitude. We attribute the deterioration of transport property to the blocking effect of protons. Neutron diffraction studies of proton containing garnets have shown that protons are located slightly above the face of O tetrahedron (looking outward from the 24d site) and inside O octahedron [44,46]. As our recent MD simulation has shown, the

conduction path of Li ion is through a triangular oxygen bottleneck between a tetrahedron and an octahedron [47]. The locations of protons clearly hinder the Li hopping. Although there has been no direct characterization of proton conductivity in hydrogarnets, it is well known that proton conduction in solid crystals is much more difficult than alkali ions [48]. Fast proton conduction at this low temperature in a neutral atmosphere is probably not favored. Therefore, in the highly protonated garnet samples, both Liþ and Hþ possess poor mobilities. Similar phase transition of tetragonal garnets, although occurring at a much slower rate, has been observed through natural aging [19,26]. We expect that the reaction mechanism at ambient condition should be similar to that in this experiment so naturally aged garnet samples should also suffer from the blocking effect of protons. Therefore, storage or packing of lithium garnet materials in inert atmosphere may be crucial to the preservation of fast ion conduction.

3.5. CO2 induced phase transition Similar to the study of the H2O effect, we employed in-situ impedance spectroscopy, ex-situ XRD and TGA to study the CO2 effects on the LLZ. The porous garnet pellets were exposed to CO2 at various temperatures. We found that the characteristics of impedance evolution at different temperatures are similar but evolutions at low temperature are much slower than the high temperature sets. Therefore, only the 350  C data are shown here. The XRD results (Fig. 11) show that CO2 exposure induces tetragonal-cubic phase transition in the 40 h sample although a significant amount of tetragonal phase is still present whereas the pattern of the 2 h sample is almost identical to the tetragonal LLZ. These results suggest that CO2 can individually induce tetragonalcubic phase transition in the absence of H2O. Since no impurity

Y. Wang, W. Lai / Journal of Power Sources 275 (2015) 612e620

619

Fig. 11. (a) XRD patterns and (b) TGA curves of LLZ pellets after passing CO2 gas for 2 h and 40 h at 350  C.

Fig. 12. (a) Impedance plots of Li7La3Zr2O12 under CO2 flow at 350  C. (b) Bulk and grain boundary resistivity evolution with time.

phase was observed, the possibility of decomposition of the garnet phase can be ruled out. The absorbed CO2 is most likely to exist in the form of Li2CO3 by extracting Li from the garnet structure and the phase transition is driven by the created extra vacancies at the Li sites. Based on this analysis, we propose the following reaction for the LLZ under CO2 exposure: Li7La3Zr2O12 þ x$CO2 / Li7-2xLa3Zr2O12-x þ x Li2CO3

(2)

The TG curve of CO2-exposed LLZ shows no weight loss below 400  C, contrary to the H2O-exposed LLZ which loses most of H2O below 400  C. At above 400  C, the weight loss occurs in two steps, one at 500  C and the other at 700  C. Larraz suggested the presence of La2O2CO3 based on their Raman spectra of aged sample which could account for the weight loss at 500  C. We attribute the weight loss at 700  C (Fig. 11) to the restoration of the tetragonal LLZ phase through the reverse reaction of (2) as our XRD results on heat-treated samples (at 750  C) always show pure tetragonal phase regardless of the history. The amount of reaction product Li2CO3 present in the 40 h sample can be roughly estimated to be 0.4 mol per mole of garnet phase based on the weight loss at 700  C step. Note that the lithium vacancy concentration (0.8 ¼ 0.4  2 mol) is higher than the theoretically predicted critical value (0.4 mol) discussed in section 3.2. The incomplete tetragonal to cubic transition is possibly caused by the slow and/or nonhomogeneous reaction kinetics. The absence of Li2CO3 in the XRD, in contrast to the case of H2O exposure, is probably due to that Li2CO3 is in an amorphous state or the quantity is below the detection limit.

For the impedance response of LLZ at 350  C, Fig. 12(a), before the CO2 gas flow was turned on, an arc corresponding to the bulk resistance was observed. Once CO2 gas flow started, the bulk resistance kept decreasing so the bulk arc eventually disappeared in the measured frequency range. This can be explained by the creation of extra vacancies via the formation of Li2CO3. When the quantity of vacancies accumulates to the critical amount, the tetragonal phase is transformed to cubic phase which further reduces the bulk resistance. After 8 h, a small semicircle appeared and grew continuously. This semicircle can be assigned to the grain boundary resistance that is probably related to the deposition of Li2CO3. For the data that contains grain boundary semicircle, an R-Q (Q is a constant phase element) was added to the equivalent circuit model. The evolution of bulk and grain boundary resistances is summarized in Fig. 12. The bulk resistivity displays exponential decay with time whereas the grain boundary resistivity increase is largely linear after 8 h.

4. Conclusion We prepared contamination-free Li7La3Zr2O12 and Ta-doped Li7La3Zr2O12 samples by taking careful measures to eliminate Alcontamination and air exposure. The minimum substitution level of Ta for the stabilization of the cubic phase is determined to be 0.6 mol per formula unit. Two-phase coexistence was observed for the intermediate compositions. The intrinsic phase transition temperature of Li7La3Zr2O12 is between 610  C and 630  C determined by high-temperature impedance measurement. In-situ impedance measurement at various temperatures under H2O or

620

Y. Wang, W. Lai / Journal of Power Sources 275 (2015) 612e620

CO2 gas flow showed that H2O and CO2 can individually induce the tetragonal-cubic phase transition of LLZ. The exposures to H2O and CO2 are in general detrimental to the transport properties of LLZ. This study provides clarification on the tetragonal-cubic phase transition of LLZ and will be helpful to further research on the optimization of Li ion conduction in lithium garnet oxides. Acknowledgment This work was financially supported by the Ceramics Program of National Science Foundation (DMR-1206356). References [1] J.W. Fergus, J. Power Sources 195 (2010) 4554e4569. [2] E.J. Cussen, J. Mater. Chem. 20 (2010) 5167e5173. [3] V. Thangadurai, S. Narayanan, D. Pinzaru, Chem. Soc. Rev. 43 (2014) 4714e4727. [4] V. Thangadurai, H. Kaack, W.J.F. Weppner, J. Am. Ceram. Soc. 86 (2003) 437e440. [5] V. Thangadurai, W. Weppner, J. Power Sources 142 (2005) 339e344. [6] R. Murugan, V. Thangadurai, W. Weppner, Angew. Chem. Int. Ed. 46 (2007) 7778e7781. [7] Y.X. Wang, W. Lai, Electrochem. Solid State Lett. 15 (2012) A68eA71. [8] M. Kotobuki, H. Munakata, K. Kanamura, Y. Sato, T. Yoshida, J. Electrochem. Soc. 157 (2010) A1076eA1079. [9] S. Ohta, T. Kobayashi, T. Asaoka, J. Power Sources 196 (2011) 3342e3345. [10] J. Awaka, N. Kijima, H. Hayakawa, J. Akimoto, J. Solid State Chem. 182 (2009) 2046e2052. [11] Y. Shimonishi, A. Toda, T. Zhang, A. Hirano, N. Imanishi, O. Yamamoto, Y. Takeda, Solid State Ionics 183 (2011) 48e53. [12] J. Awaka, A. Takashima, K. Kataoka, N. Kijima, Y. Idemoto, J. Akimoto, Chem. Lett. 40 (2011) 60e62. [13] N. Janani, S. Ramakumar, L. Dhivya, C. Deviannapoorani, K. Saranya, R. Murugan, Ionics 17 (2011) 575e580. [14] H. Xie, J.A. Alonso, Y.T. Li, M.T. Fernandez-Diaz, J.B. Goodenough, Chem. Mater. 23 (2011) 3587e3589. [15] Y.X. Wang, A. Huq, W. Lai, Solid State Ionics 255 (2014) 39e49. [16] C.A. Geiger, E. Alekseev, B. Lazic, M. Fisch, T. Armbruster, R. Langner, M. Fechtelkord, N. Kim, T. Pettke, W. Weppner, Inorg. Chem. 50 (2011) 1089e1097. [17] I. Kokal, M. Somer, P.H.L. Notten, H.T. Hintzen, Solid State Ionics 185 (2011) 42e46. [18] H. Buschmann, J. Dolle, S. Berendts, A. Kuhn, P. Bottke, M. Wilkening, P. Heitjans, A. Senyshyn, H. Ehrenberg, A. Lotnyk, V. Duppel, L. Kienle, J. Janek, Phys. Chem. Chem. Phys. 13 (2011) 19378e19392.

[19] G. Larraz, A. Orera, M.L. Sanjuan, J. Mater. Chem. A 1 (2013) 11419e11428. [20] J.L. Allen, J. Wolfenstine, E. Rangasamy, J. Sakamoto, J. Power Sources 206 (2012) 315e319. [21] E. Rangasamy, J. Wolfenstine, J. Sakamoto, Solid State Ionics 206 (2012) 28e32. [22] Y.T. Li, J.T. Han, C.A. Wang, H. Xie, J.B. Goodenough, J. Mater. Chem. 22 (2012) 15357e15361. [23] C. Deviannapoorani, L. Dhivya, S. Ramakumar, R. Murugan, J. Power Sources 240 (2013) 18e25. [24] S. Ramakumar, L. Satyanarayana, S.V. Manorama, R. Murugan, Phys. Chem. Chem. Phys. 15 (2013) 11327e11338. [25] T. Thompson, J. Wolfenstine, J.L. Allen, M.D. Johannes, A. Huq, I.N. David, J. Sakamoto, J. Mater. Chem. A (2014) 13431e13436. [26] C. Galven, J.-L. Fourquet, M.-P. Crosnier-Lopez, F. Le Berre, Chem. Mater. 23 (2011) 1892e1900. [27] A.C. L, R.B. V. Dreele, Los Alamos National Laboratory Report LAUR 86-748, 2000. [28] B.H. Toby, J. Appl. Crystallogr. 34 (2001) 210e213. [29] E.J. Cussen, Chem. Commun. (2006) 412e413. [30] H. Sato, R. Kikuchi, J. Chem. Phys. 55 (1971) 677. [31] J.N. Reimers, J.R. Dahn, J. Electrochem. Soc. 139 (1992) 2091e2097. [32] I.R. Gibson, J.T.S. Irvine, J. Mater. Chem. 6 (1996) 895e898. [33] G. Ceder, A. Van der Ven, Electrochim. Acta 45 (1999) 131e150. [34] Y. Harada, Y. Hirakoso, H. Kawai, J. Kuwano, Solid State Ionics 121 (1999) 245e251. [35] A. Van der Ven, M.K. Aydinol, G. Ceder, G. Kresse, J. Hafner, Phys. Rev. B 58 (1998) 2975e2987. [36] N. Bernstein, M.D. Johannes, K. Hoang, Phys. Rev. Lett. 109 (2012) 205702. [37] M. Matsui, K. Takahashi, K. Sakamoto, A. Hirano, Y. Takeda, O. Yamamoto, N. Imanishi, Dalton Trans. 43 (2014) 1019e1024. [38] V. Thangadurai, W. Weppner, J. Solid State Chem. 179 (2006) 974e984. [39] Y.X. Gao, X.P. Wang, W.G. Wang, Q.F. Fang, Solid State Ionics 181 (2010) 33e36. [40] Y. Li, C.-A. Wang, H. Xie, J. Cheng, J.B. Goodenough, Electrochem. Commun. 13 (2011) 1289e1292. [41] T. Luna, V. Thangadurai, Chem. Mater. 23 (2011) 3970e3977. [42] M. Nyman, T.M. Alam, S.K. McIntyre, G.C. Bleier, D. Ingersoll, Chem. Mater. 22 (2010) 5401e5410. [43] G.A. Lager, T. Armbruster, J. Faber, Am. Mineral. 72 (1987) 756e765. [44] C. Galven, J. Dittmer, E. Suard, F. Le Berre, M.-P. Crosnier-Lopez, Chem. Mater. 24 (2012) 3335e3345. [45] M.T. Thakker, C.W. Chi, R.E. Peck, D.T. Wasan, J. Chem. Eng. Data 13 (1968) 553. [46] B.C. Chakoumakos, G.A. Lager, J.A. Fernandezbaca, Acta Crystallogr. Sect. CCryst. Struct. Commun. 48 (1992) 414e419. [47] Y. Wang, M. Klenk, K. Page, W. Lai, Chem. Mater. 26 (2014) 5613e5624. [48] K.D. Kreuer, Ann. Rev. Mater. Res. 33 (2003) 333e359.