- Email: [email protected]
+ alumina—A novel Li+ solid electrolyte
Li+-Nat
BETA ALUMINA-A NOVEL SOLID ELECTROLYTE
Lit
G. C. FARRINGT~N and W. L. ROIH General Electric Research and Development Center, P.O. Box 8, Schenectady, (ReceioeB
NY 12301, U.S.A.
1 September
1976)
is one of several monovalent cations known to completely replace Nat in Na+ beta alumina. The exchange occurs readily to approximately Xl%, producing a physically stable Li+ solid Abstract-Li+
electrolyte having a conductivity of approximately 10e3 (Q cm)-’ at 25°C. The Li+ transport number through Li+-Ni’ beta alumina is nearly 1 for compositions having Li+/Na+ greater than 1. Li+ ions migrate through the solid electrolyte lattice without significantly altering its Na+ content. This paper discusses our work examining ionic equilibrium, Li’ transport, ionic conductivity, and stability of Lif-Na+ beta alumina. The extraordinary preferential Na* occupation of the beta alumina structure even in the presence of high Li+ activity is discussed and similar behavior predicted for another beta alumina composition
INTRODUCTJON
There has been considerable interest in identifying a Li” solid electrolyte of conductivity comparable to Nat beta alumina, which in single crystal form is 1.4 x lo-’ (0 cm)-‘[l] at 25°C. We now report that Na+ beta alumina in which greater than ca 50% of the Na+ content has been replaced by Lif is an excellent Li+ conductor, having a Li+ transport number close to unity. Li+ ions can be electrochemically passed through the structure without significantly affecting its Li+/Na+ ratio. This phenomenon, which arises from a striking preferential Na+ occupation of beta alumina even in the presence of high Lif activity, we have termed ‘co-ionic conductivity’[2,3]. Co-ionic Li+-Na+ beta alumina can be prepared by direct ion exchange of polycrystalline Na+-beta alumina without cracking or obvious ceramic degradation. It could find application in energy storage systems requiring a selective Li+ electrolyte of high ionic and very low electronic conductivity. More importantly, it appears to be only one example of general co-ionic behavior observed in beta alumina and pcrhaps in other structure types as well. Detailed exchange equilibria between mixed MN03-NaNO, melts at 350°C and Mf-Na+ beta alumina for M = Lif, Ag+, K+, Tl+, and Cs+ were reported by Yao and KummerC41 and are summarized in Fig. 1. The Li+ exchange is intriguing in that a sample of pure Na+ beta alumina immersed in a large excess of molten LiNOa at 350°C undergoes only approximately 50”/, ion exchange. The small, generally less than 0.5x, impurity level of Na* in LiNO, is sufficient to maintain a Li+/Na+ ratio in beta alumina of about 1. Yao and Kummer attempted to prepare pure Li ’ beta alumina by replacing Na+ in Na+ beta alumina with Ag *, which occurs readily, and equilibrating the Ag+ beta alumina with LiNO, saturated with LiCl at 350°C. Formation of AgCl is the driving force for complete ion exchange. Several values have been reported for the conductivity of Li+ beta alumina synthesized by the Ag+
route. Radzilowski. Yao and Kummer[S] used tracer diffusion and dielectric loss techniques to deduce the ionic conductivity of beta alumina powder in which an unspecified proportion of Na* was replaced by Li+. Whittingham and Huggin$6] carried out UC conductivity measurements on single crystals of beta alumina containing Li*. Table 1 lists these results. This communication details our work examining ionic equilibrium, Li+ transport, ionic conductivity, and stability of co-ionic LiC-NaC beta alumina. Also discussed are possible applications for the novel solid electrolyte and the occurrence of similar co-ionic behavior with other ion combinations. EXPERIMENTAL
Tubes of Na+ beta alumina approximately 1.1 cm o.d.. 1.0 cm i.d., and 9 cm in length were prepared
Molten
salt
composition
Fig. 1. Ion exchange equilibria between Na* beta alumina and various molten nitrates at 350°C; data taken from Yao and Kummer [4]_ 767
768
G.C.
FARRINGTON
Table 1. Li*-Na+ Tracer
diffusion[5]
AND
beta alumina conductivity
(a, b)
Dielectric loss[5] (a, b) AC conductivity[6]
This work (c)
W. LROTH
(b)
1.0 x 10-5(Rcm)-i
1.5 x 10-3(ficm)-i 1.3 x 10-4(Rcm)~1 5 x 10-3(Qcm)-1
4 Extrapolated to 298°K. (b) Precise Li+/Na+ ratio in sample unreported. (c) Li+/Na+ = 1.1. AU data are single crystal values at 298°K.
(a)
from Alcoa XB-2 beta alumina powder by an electrophoretic deposition technique previously described in detail by Powers [7]_ Included in the ceramic were 1.0% Y *03, 0.5% MgO, and l.Oo/,Zr&. The samples were fired in an oxygen atmosphere at 1700°C into single phase, fine-grained, polycrystaltine beta alumina of !G9O/,theoretical density. Typical grain size was 4-10 q. Li+ exchange was carried out by immersing samples in molten LiNO, held in a Pt crucible heated by a thermostated sand bath. Fisher reagent LiNOB, reported by the manufacturer to have less than 0.2% total alkali ion impurity, was used as received. Before and after exchange, samples were washed with anhydrous methanol in a Soxhlet extractor and dried at 500°C. Sufficient excess LiN03 was used for each exchange so that the total Na+ extracted from the beta alumina never exceeded 0.5% of the Li’ content of the bath. RESULTS Thermal
ion exchange
Our results indicate that approximately 50% of the Na+ content of polycrystalline Naf beta alumina is readily replaced by Li+ when immersed in molten LiNO, at 3501350°C. This essentially confirms the observations of Yao and Kummer. The extent of exchange increases with temperature and with repeated equilibration with fresh LiN03 as the following data show. A tube of polycrystalline Nat beta alumina was immersed in LiNO, for 12 h at 450°C. The tube underwent 1.84% weight loss, corresponding to 52.1% Na+ exchange by Lif. Extended immersion produced no further weight change, indicating stability of the sample towards general attack by LiN03 and implying that 52.1% exchange represents equilibrium between the solid and the melt. Sufficient LiNOa was used that its Na* content increased only 0.13% during the exchange. Four samples of polycrystalline Na* beta alumina were then immersed for three successive 20 h periods in fresh LiNOJ at 350°C. The extents of exchange, determined by weight loss, agreed closely: 71.5, 71.5, 70.1 and 73.1%. Two additional exchanges, each for 10 h in fresh LiN03 at 5OO”C,increased the exchanges to 87.4, 85.2, 85.5 and 86.3%, respectively. All these experiments assume that complete Nat exchange by Li’ in these samples results in a 3.68% weight loss. This was determined by observing the weight gain of four representative Na+ beta alumina samples upon complete Naf replacement by Ag+. Yao and Kumtner have shown that Ag+ readily and
completely replaces Na + in beta alumina immersed in AgN03 at 350°C Four Nat beta alumina samples were equilibrated with molten AgNO, at 350°C. Each showed a weight gain of 18.6%, consistent with a chemical formula of 1.32Naz0. 1 lA1203. Li+ electrolysis throtqh Li+-Na+
beta alumina
Extended Li+ electrolysis through Li+-NaC beta alumina from a LiNOJ melt in equilibrium with the composition of the solid electrolyte does not significantly alter the Li+/Na+ ratio in the solid electrolyte. The Li+ transport number of Lif-Na+ beta alumina is nearly 1 under these conditions. This was shown in two electrolysis experiments using a Lif-Na+ beta alumina tube which had first been equilibrated with molten LiNO, at 450°C and found to have undergone 52.1%Na+ replacement. The Li+-Na+ beta alumina tube was then placed in the electrolysis apparatus shown in Fig. 2. The identical LiNOJ used in preparing the partially exchanged material was used in the electrolysis. Two electrolyses were carried out, the first at an average cd of 5 mA/cm2. A total of 505 C was passed through the sample, 2.97% of the 1700 C theoretically required to replace its entire Na+ and Lit content. But, subsequent weighing indicated that electrolysis had only increased the extent of Lif exchange from 52.1% to 54.3%. A second electrolysis at an average cd of 20mA/cm’ for 1055 C, or 62% of the total mobile ion content, increased the exchange to 55.8%. In neither experiment did the mobile ion composition of the sample change significantly.
Fig. 2. Molten
salt electrolysis apparatus.
769
Li +-Na + beta alumina 2000
I
I
I
I
0
-
1500-
500 CL
l l 1, I. 0.4
I 0.2
0
I 0.6
I 0.8
I .O
CLi+l
Composition, [Lit
No*7
Fig. 3. Intragrain and grain boundary resistivities as a function of composition for typical samples of polycrystalline Li*-Naf beta alumina; 0 Grain Boundary, 0 Intragrain Resistivity. Li+-Na’
beta alumina
conductivity
The dc resistivity of Li+-Na+ beta alumina increases continuously with increasing Li+/Na+ ratio, Resistivity at dc is actually comprised of intragranular and grain boundary components. lntragrain resistivity approximates that of a single crystal. Grain boundary resistivity generally is dominant at tem-
Fig.
4. Ideal
resistivity response of
the
peratures less than approximately 100°C. Figure 3 is a plot of intragrain and grain boundary resistivity at 25°C for Li+-Na+ beta alumina as a function of Li*/Na+ ratio in the solid electrolyte. Li+ exchange produces only a small increase in each resistivity component, so that 85% Li+-Na* beta alumina has intragrain and grain boundary resistivities only 34 fold greater than Na+ beta alumina at 25°C. Single crystal specimens of Li+-Na+ beta alumina were unfortunately not available for this investigation. Therefore, resistivity data were obtained with a group of polycrystalline Na+ beta alumina tubes converted to Li+-Na+ beta alumina by ion exchange in LiNO,. A 4 probe galvanostatic technique, previously described [S,9], was used to separate intragrain and grain boadary resistivity components. DC resistivity of polycrystalline beta ahunina can be modeled by a resistor (R,), representing intragrain transport, in series with a simple resistor (Rb) and capacitor (C,) in parallel, representing grain boundary resistivity (See Fig. 4). For certain compositions of polycrystalline beta alumina, as in this investigation in which Y is present in the grain boundary phase, the parallel RC grain boundary time constant is of the order of 10m5 s. This circuit predicts a resistivity/ time response upon -perturbation by a constant current square wave similar to that shown in Fig. 4. The observed response for Li+-Na+ beta alumina, shown in Fig. 5, is much less than ideal; so, the values of single crystal resistivity cited in Fig. 3 are approximate. Clear definition of the resistivity/composition relationship of this electrolyte requires detailed single crystal measurements. Li/‘Li+-Na+ beta nEumina/Fk,
cell
One potential use of Li+-Naf beta alumina is as a separator in secondary non-aqueous electrochemical cells. The electrolyte would prevent random diffusion between half cells but permit easy Lit migration. Other Lit solid electrolytes which have been reported are generally too low in conductivity or too soluble for use in non-aqueous electrolyte battery applications. To demonstrate one such application, a disc of polycrystalline Li+-Na+ beta alumina was prepared by thermal exchange and incorporated into a test Li/
circuit perturbation.
model
electrochemical
shown
during
fast-rise,
constant current
770
G. C.FAKKINGTON
and W. L. ROTH
Fig. 5. Observed resistitity response of polycrystalline Li+-Na+ beta alumina during stant current perturbation; 25°C; figure indicates y0 Li+ in samples. Li+-Na+ beta alumina/Brz secondary cell. Propylene carbonate containing LiClO, was employed as the liquid electrolyte and the Li electrode was formed in situ by deposition of Li on a Ni foam substrate. No increase in cell resistance attributable to the solid electrolyte was observed during a discharge at 2.2 mA/cm* of disc area over several days or during subsequent recharge at a similar rate. During this time, sufficient charge was passed to replace 224% of the total Li+ content of the disc. DISCUSSION
Our data have shown Li*-Na* beta alumina, for compositions having Li+/Na+ greater than approximately 1, to be an excellent Li+ solid electrolyte in which the Li* transport number is nearly unity. They also demonstrate that the residual Na+ content of the solid electrolyte is not trapped within its structure by slow exchange kinetics. Rather, as Yao and Kummer have indicated, the extraordinary preferential Na+ occupation of Na+ beta alumina in contact with molten LiNOS appears to be an equilibrium state. A very small Nat concentration in the melt supports a large Na+ concentration in the solid. Electrochemical passage of Lif through the solid electrolyte, then, represents only a small perturbation from the equilibrium rate of ion exchange occurring spontaneously between the solid and liquid electrolytes. Equilibrium is maintained as long as the rate of electrochemical ionic drift imposed externally is much smaller than the rate of thermally activated ion exchange. The composition of Li+-Naf beta alumina, then, remains unchanged despite extended unidircctional Lit transport. To demonstrate how small the rate of electrochemical drift is compared to equilibrium ion exchange, we simplify the arrangement of mobile ions in the beta alumina conduction plane, distributing them into equivalent potential wells separated by energy barriers of about 0.17 eV (3800 &/mole). This is the activation energy of conduction typically
four
probe, con-
observed for Nat beta alumina. When a constant electric field, E, is externally imposed on the system, the probability of Na+ migration in the direction of the field, W,, is Wp = oexp -
((AG -
l/2 e&)/U)
(1)
and the probability opposing field, WB, is W, = uexp - ((AG + i/2 eaE)/kT)
(2)
In (1) and (2), v is the characteristic vibration frequency of Na+ within a potential well, e the electronic charge, and a the ionic hopping distance. In Nat beta alumina at 3OO”C, an electric field, E, of 1 V/cm sustains an ionic current of approximately 1 A/cm’. The maenitude of a is of the order 10-s m. Thirefore, the pr:duct enE is approximately lo-’ eV compared to AC which is 0.17 eV. Normal elechochemical currents, therefore, are extremely small in comparison to the rate of thermal diffusion and pose no threat to solid-liquid equilibrium. This treatment assumes a high concentration of vacancies within the structure, a factor true for Na+ beta alumina. The key to the curious ion exchange behavior of beta alumina lies in its structure. Ionic transport through beta alumina occurs in two dimensions along parallel conducting planes. Three non-equivalent crystallographic sites for mobile ion occupation have been identified in the beta alumina unit cell. The relative ionic occupation of these sites varies with ion type and temperature and is influenced by the overall fixed beta alumina structure and by interactions among the mobile ions themselves. We have previously reviewed [3] data currently available describing ionic distribution in the conducting plane and stated that we believe long range ion transport through beta alumina to proceed by ion exchange among these energetically distinguishable sites. The existence of co-ionic compositions having properties not directly proportional to those of their completely substituted end members clearly indicates that when two or more mobile ions are present simultaneously in beta alumina, they are not randomly distributed
Li+-Na+ beta alumina in the conduction plane. Rather, there is preferential occupation of specific sites by specific ions. It is this sitesite non-equivalency and mobile ion interaction that results in the non-linear concentration dependence of co-ionic transport in Li+-Na+ beta alumina and the striking variation in activity coefficients of M + in the solid and the melt observed by Yao and Kummer for (MN03pNaN03),i,,id/(M*-Na+ beta aluminaj,,l,d equilibria. Li*-Na+ beta alumina itself remains an excellent, high conductivity Li+ solid electrolyte. Assuming partial Li+ substitution produces no unexpected changes in its gross chemical behavior, it should be stable in a wide range of chemical environments, as is Na+ beta alumina, and have an extremely low electronic conductivity. Li+-Na+ beta alumina can be produced by direct ion exchange of polycrystalline Na’ beta alumina, for which there already exists a considerable fabrication technology, and can be made in a variety of physical shapes such as discs and tubes. It, therefore, may find application in ambient temperature rechargeable non-aqueous and solid-state Li batteries. Use in direct contact with molten Li+ in metal is probably not practicable. We have observed that samples of Na+ beta alumina are reduced at least within 1 h and probably more rapidly when in contact with molten Li at 450°C. A significant consideration in the use of Lif-Naf beta alumina in battery applications is the stability of its composition. The Li+/Na+ ratio in the solid must change as the activity coefficients of Lif and Naf in the media in which it is in equilibrium change. The precise Li+/Na+ ratio in the solid in contact with 1% Na+ in LiNC& at 350°C will differ from that in contact with 1% Na+ in LiC104 dissolved in propylene carbonate at 25°C. But, excepting environments in which Na’ is strongly and preferentially complexed compared to Li +, for example by formation of an insoluble Nat compound, a fair conjecture is that stability will pose little problem in actual use. Regardless, our measurements indicate that the ionic conductivity of Li+-Na+ beta alumina decreases by a factor of 12 as the exchange extent is increased from 50 to 89%. So changes in composition produce only small conductivity effects. The small variation in ionic conductivit’i with relative composition actually has only been observed to 85% Li + in Li+-Na+ beta alumina. Pure Li+ beta alumina has not been prepared in this work nor its conductivity determined. Previous measurements of the conductivity of nominally pure Lit beta alumina prepared by exchanging Na* beta alumina with AgNOX, then replacing Ag+ with Li+ in a LiNOXLiCl melt, disagree markedly as Table 1 shows. A portion of this disagreement is surely due to the different techniques used to measure conductivity. These have been reviewed in the introduction. It is also possible that the different samples measured by various groups varied considerably in composition. Exchange 01 Na+ beta alumina with Ag+ is very favorable. Subsequent equilibration with LiNO,-LiCl may induce complete AgC removal by forming AgCl in the melt; but, there is no guarantee that only Li+ replaces it in the solid electrolyte. Residual Na+ in the LiNO,-LiCl can easily and preferentially migrate into the beta alumina. Whether all the Agf itself is
771
indeed replaced is also unclear. All these possibilities raise doubts about the actual compositions previously investigated as pure Li+ beta alumina. Preparation of the compound by direct ion exchange appears to require Li+ salts of extremely high Li+ purity. The properties of Li+-Na+ beta alumina arc interesting partly because there are many potential applications for a new Li+ solid electrolyte. More important, however, may be its implication that co-ionic behavior, or the mutual interaction of two or more mobile ions producing effects not predicted by simply averaging the characteristics of their completely substituted extremes, is a general phenomenon in solid electrolytes. This is true at least with beta alumina as exchange data for Ag+, TI+, K+, Li+ and Cs+ into Na+ beta alumina of Yao and Kummer in Fig. 5 show. Just as the exchange data show that a less than ca 1% Na* in LiNO, sustains ca 50% Na+ occupation of Li*-Na’ beta alumina, so they indicate that ca 1% Ag+ in NaN03 supports cu 55% Ag+ in Na+the transport Ag+ beta alumina. Extrapolating characteristics of Li l-Na+ beta alumina, Na+-Ag l beta alumina can be expected to have a Na+ transport number of nearly unity for compositions in which Na*/Ag+ is greater than 1. Other monovalent cations such as K+ and Tl’ resemble the behavior of Ag+ to lesser cxtcnts. Cs+ and Rb+ act similarly to Li+. Co-ionic interactions in the beta alumina structure are therefore not at all unique to the Li+Na+ composition. Indeed, co-ionic behavior may be generally observed in other structure types as well. Further investigation will test this concept. SUMMARY
AND CONCLUSIONS
When Na+ beta alumina is equilibrated with molten LiN03 at 350-500°C only partial Li * replacement of Na* takes place. The very small, typically less than OS%, concentration of Na+ in the melt supports a large concentration of Na+ in the solid. Polycrystaline Na+ beta alumina can be directly converted to LiC-Nat beta alumina by direct ion exchange without cracking or obvious physical degradation. Li +-Nat beta alumina having a Li + /Na + ratio greater than approximately one is”an exdellent Li+ solid ionic conductor with a Li+ transport number of nearly one. It should find a number of practical applications in energy storage systems. What is most extraordinary about Li+-Na+ beta alumina is the preferential Na* occupation of the beta alumina conducting plane even in the presence of high Li+ activity. Li+ ions move through the structure without significantly altering its Na* content. We have termed this behavior co-ionic cmdudvity and predicted that it is a general phenomenon in beta alumina and perhaps in other structure types as well. Specifically, Na+-Ag* beta alumina has been predicted to be a Na+ conductor exhibiting preferential Ag+ occupation even in the presence of high Na+ activity. REFERENCES 1.
M. S. Whittingham and R. A. Huggins, J. chern. Phys. 54. 414 (1971).
772
G.C.
FARRINGWN
2. G. C. Farrington and W. L. Roth. Ir~rvv~frrrir~~crl Conferencr on Suprrioitrc C‘orxdltctors, (Edited by G. D. Mahan and-W. L. Roth) p. 418: Plenum Press, New York (1977). 3. W. L. Roth and G. C. Farrington, article submitted to Science. 4. Y. F. Yao and J. T. Kummer, J. inorg. nucl. Chem. 29. 2453 (1967).
AND W. L.Rom 5. R. H. Radzilowski, Y. F. Yao and J. T. Kummer, J.. appl. Phys. 40, 4716 (1969). 6. &I: S. Vi’hitti&ham -and .R. A. Huggins, Solid State Chemistry, p. 139. N,B.S. Spec. Pub. No. 364, (1972). 7. R. W. Powers, J. dectrochem. Sot. 122, 490 (1975). 8. G. C. Farrington, J. elrctrochem. Sot. 121, 1314 (1974). 9. G. C. Farrington, J. electruchem. Sot. 123, 1213 (1976).
BETA ALUMINA-A NOVEL SOLID ELECTROLYTE
Lit
G. C. FARRINGT~N and W. L. ROIH General Electric Research and Development Center, P.O. Box 8, Schenectady, (ReceioeB
NY 12301, U.S.A.
1 September
1976)
is one of several monovalent cations known to completely replace Nat in Na+ beta alumina. The exchange occurs readily to approximately Xl%, producing a physically stable Li+ solid Abstract-Li+
electrolyte having a conductivity of approximately 10e3 (Q cm)-’ at 25°C. The Li+ transport number through Li+-Ni’ beta alumina is nearly 1 for compositions having Li+/Na+ greater than 1. Li+ ions migrate through the solid electrolyte lattice without significantly altering its Na+ content. This paper discusses our work examining ionic equilibrium, Li’ transport, ionic conductivity, and stability of Lif-Na+ beta alumina. The extraordinary preferential Na* occupation of the beta alumina structure even in the presence of high Li+ activity is discussed and similar behavior predicted for another beta alumina composition
INTRODUCTJON
There has been considerable interest in identifying a Li” solid electrolyte of conductivity comparable to Nat beta alumina, which in single crystal form is 1.4 x lo-’ (0 cm)-‘[l] at 25°C. We now report that Na+ beta alumina in which greater than ca 50% of the Na+ content has been replaced by Lif is an excellent Li+ conductor, having a Li+ transport number close to unity. Li+ ions can be electrochemically passed through the structure without significantly affecting its Li+/Na+ ratio. This phenomenon, which arises from a striking preferential Na+ occupation of beta alumina even in the presence of high Lif activity, we have termed ‘co-ionic conductivity’[2,3]. Co-ionic Li+-Na+ beta alumina can be prepared by direct ion exchange of polycrystalline Na+-beta alumina without cracking or obvious ceramic degradation. It could find application in energy storage systems requiring a selective Li+ electrolyte of high ionic and very low electronic conductivity. More importantly, it appears to be only one example of general co-ionic behavior observed in beta alumina and pcrhaps in other structure types as well. Detailed exchange equilibria between mixed MN03-NaNO, melts at 350°C and Mf-Na+ beta alumina for M = Lif, Ag+, K+, Tl+, and Cs+ were reported by Yao and KummerC41 and are summarized in Fig. 1. The Li+ exchange is intriguing in that a sample of pure Na+ beta alumina immersed in a large excess of molten LiNOa at 350°C undergoes only approximately 50”/, ion exchange. The small, generally less than 0.5x, impurity level of Na* in LiNO, is sufficient to maintain a Li+/Na+ ratio in beta alumina of about 1. Yao and Kummer attempted to prepare pure Li ’ beta alumina by replacing Na+ in Na+ beta alumina with Ag *, which occurs readily, and equilibrating the Ag+ beta alumina with LiNO, saturated with LiCl at 350°C. Formation of AgCl is the driving force for complete ion exchange. Several values have been reported for the conductivity of Li+ beta alumina synthesized by the Ag+
route. Radzilowski. Yao and Kummer[S] used tracer diffusion and dielectric loss techniques to deduce the ionic conductivity of beta alumina powder in which an unspecified proportion of Na* was replaced by Li+. Whittingham and Huggin$6] carried out UC conductivity measurements on single crystals of beta alumina containing Li*. Table 1 lists these results. This communication details our work examining ionic equilibrium, Li+ transport, ionic conductivity, and stability of co-ionic LiC-NaC beta alumina. Also discussed are possible applications for the novel solid electrolyte and the occurrence of similar co-ionic behavior with other ion combinations. EXPERIMENTAL
Tubes of Na+ beta alumina approximately 1.1 cm o.d.. 1.0 cm i.d., and 9 cm in length were prepared
Molten
salt
composition
Fig. 1. Ion exchange equilibria between Na* beta alumina and various molten nitrates at 350°C; data taken from Yao and Kummer [4]_ 767
768
G.C.
FARRINGTON
Table 1. Li*-Na+ Tracer
diffusion[5]
AND
beta alumina conductivity
(a, b)
Dielectric loss[5] (a, b) AC conductivity[6]
This work (c)
W. LROTH
(b)
1.0 x 10-5(Rcm)-i
1.5 x 10-3(ficm)-i 1.3 x 10-4(Rcm)~1 5 x 10-3(Qcm)-1
4 Extrapolated to 298°K. (b) Precise Li+/Na+ ratio in sample unreported. (c) Li+/Na+ = 1.1. AU data are single crystal values at 298°K.
(a)
from Alcoa XB-2 beta alumina powder by an electrophoretic deposition technique previously described in detail by Powers [7]_ Included in the ceramic were 1.0% Y *03, 0.5% MgO, and l.Oo/,Zr&. The samples were fired in an oxygen atmosphere at 1700°C into single phase, fine-grained, polycrystaltine beta alumina of !G9O/,theoretical density. Typical grain size was 4-10 q. Li+ exchange was carried out by immersing samples in molten LiNO, held in a Pt crucible heated by a thermostated sand bath. Fisher reagent LiNOB, reported by the manufacturer to have less than 0.2% total alkali ion impurity, was used as received. Before and after exchange, samples were washed with anhydrous methanol in a Soxhlet extractor and dried at 500°C. Sufficient excess LiN03 was used for each exchange so that the total Na+ extracted from the beta alumina never exceeded 0.5% of the Li’ content of the bath. RESULTS Thermal
ion exchange
Our results indicate that approximately 50% of the Na+ content of polycrystalline Naf beta alumina is readily replaced by Li+ when immersed in molten LiNO, at 3501350°C. This essentially confirms the observations of Yao and Kummer. The extent of exchange increases with temperature and with repeated equilibration with fresh LiN03 as the following data show. A tube of polycrystalline Nat beta alumina was immersed in LiNO, for 12 h at 450°C. The tube underwent 1.84% weight loss, corresponding to 52.1% Na+ exchange by Lif. Extended immersion produced no further weight change, indicating stability of the sample towards general attack by LiN03 and implying that 52.1% exchange represents equilibrium between the solid and the melt. Sufficient LiNOa was used that its Na* content increased only 0.13% during the exchange. Four samples of polycrystalline Na* beta alumina were then immersed for three successive 20 h periods in fresh LiNOJ at 350°C. The extents of exchange, determined by weight loss, agreed closely: 71.5, 71.5, 70.1 and 73.1%. Two additional exchanges, each for 10 h in fresh LiN03 at 5OO”C,increased the exchanges to 87.4, 85.2, 85.5 and 86.3%, respectively. All these experiments assume that complete Nat exchange by Li’ in these samples results in a 3.68% weight loss. This was determined by observing the weight gain of four representative Na+ beta alumina samples upon complete Naf replacement by Ag+. Yao and Kumtner have shown that Ag+ readily and
completely replaces Na + in beta alumina immersed in AgN03 at 350°C Four Nat beta alumina samples were equilibrated with molten AgNO, at 350°C. Each showed a weight gain of 18.6%, consistent with a chemical formula of 1.32Naz0. 1 lA1203. Li+ electrolysis throtqh Li+-Na+
beta alumina
Extended Li+ electrolysis through Li+-NaC beta alumina from a LiNOJ melt in equilibrium with the composition of the solid electrolyte does not significantly alter the Li+/Na+ ratio in the solid electrolyte. The Li+ transport number of Lif-Na+ beta alumina is nearly 1 under these conditions. This was shown in two electrolysis experiments using a Lif-Na+ beta alumina tube which had first been equilibrated with molten LiNO, at 450°C and found to have undergone 52.1%Na+ replacement. The Li+-Na+ beta alumina tube was then placed in the electrolysis apparatus shown in Fig. 2. The identical LiNOJ used in preparing the partially exchanged material was used in the electrolysis. Two electrolyses were carried out, the first at an average cd of 5 mA/cm2. A total of 505 C was passed through the sample, 2.97% of the 1700 C theoretically required to replace its entire Na+ and Lit content. But, subsequent weighing indicated that electrolysis had only increased the extent of Lif exchange from 52.1% to 54.3%. A second electrolysis at an average cd of 20mA/cm’ for 1055 C, or 62% of the total mobile ion content, increased the exchange to 55.8%. In neither experiment did the mobile ion composition of the sample change significantly.
Fig. 2. Molten
salt electrolysis apparatus.
769
Li +-Na + beta alumina 2000
I
I
I
I
0
-
1500-
500 CL
l l 1, I. 0.4
I 0.2
0
I 0.6
I 0.8
I .O
CLi+l
Composition, [Lit
No*7
Fig. 3. Intragrain and grain boundary resistivities as a function of composition for typical samples of polycrystalline Li*-Naf beta alumina; 0 Grain Boundary, 0 Intragrain Resistivity. Li+-Na’
beta alumina
conductivity
The dc resistivity of Li+-Na+ beta alumina increases continuously with increasing Li+/Na+ ratio, Resistivity at dc is actually comprised of intragranular and grain boundary components. lntragrain resistivity approximates that of a single crystal. Grain boundary resistivity generally is dominant at tem-
Fig.
4. Ideal
resistivity response of
the
peratures less than approximately 100°C. Figure 3 is a plot of intragrain and grain boundary resistivity at 25°C for Li+-Na+ beta alumina as a function of Li*/Na+ ratio in the solid electrolyte. Li+ exchange produces only a small increase in each resistivity component, so that 85% Li+-Na* beta alumina has intragrain and grain boundary resistivities only 34 fold greater than Na+ beta alumina at 25°C. Single crystal specimens of Li+-Na+ beta alumina were unfortunately not available for this investigation. Therefore, resistivity data were obtained with a group of polycrystalline Na+ beta alumina tubes converted to Li+-Na+ beta alumina by ion exchange in LiNO,. A 4 probe galvanostatic technique, previously described [S,9], was used to separate intragrain and grain boadary resistivity components. DC resistivity of polycrystalline beta ahunina can be modeled by a resistor (R,), representing intragrain transport, in series with a simple resistor (Rb) and capacitor (C,) in parallel, representing grain boundary resistivity (See Fig. 4). For certain compositions of polycrystalline beta alumina, as in this investigation in which Y is present in the grain boundary phase, the parallel RC grain boundary time constant is of the order of 10m5 s. This circuit predicts a resistivity/ time response upon -perturbation by a constant current square wave similar to that shown in Fig. 4. The observed response for Li+-Na+ beta alumina, shown in Fig. 5, is much less than ideal; so, the values of single crystal resistivity cited in Fig. 3 are approximate. Clear definition of the resistivity/composition relationship of this electrolyte requires detailed single crystal measurements. Li/‘Li+-Na+ beta nEumina/Fk,
cell
One potential use of Li+-Naf beta alumina is as a separator in secondary non-aqueous electrochemical cells. The electrolyte would prevent random diffusion between half cells but permit easy Lit migration. Other Lit solid electrolytes which have been reported are generally too low in conductivity or too soluble for use in non-aqueous electrolyte battery applications. To demonstrate one such application, a disc of polycrystalline Li+-Na+ beta alumina was prepared by thermal exchange and incorporated into a test Li/
circuit perturbation.
model
electrochemical
shown
during
fast-rise,
constant current
770
G. C.FAKKINGTON
and W. L. ROTH
Fig. 5. Observed resistitity response of polycrystalline Li+-Na+ beta alumina during stant current perturbation; 25°C; figure indicates y0 Li+ in samples. Li+-Na+ beta alumina/Brz secondary cell. Propylene carbonate containing LiClO, was employed as the liquid electrolyte and the Li electrode was formed in situ by deposition of Li on a Ni foam substrate. No increase in cell resistance attributable to the solid electrolyte was observed during a discharge at 2.2 mA/cm* of disc area over several days or during subsequent recharge at a similar rate. During this time, sufficient charge was passed to replace 224% of the total Li+ content of the disc. DISCUSSION
Our data have shown Li*-Na* beta alumina, for compositions having Li+/Na+ greater than approximately 1, to be an excellent Li+ solid electrolyte in which the Li* transport number is nearly unity. They also demonstrate that the residual Na+ content of the solid electrolyte is not trapped within its structure by slow exchange kinetics. Rather, as Yao and Kummer have indicated, the extraordinary preferential Na+ occupation of Na+ beta alumina in contact with molten LiNOS appears to be an equilibrium state. A very small Nat concentration in the melt supports a large Na+ concentration in the solid. Electrochemical passage of Lif through the solid electrolyte, then, represents only a small perturbation from the equilibrium rate of ion exchange occurring spontaneously between the solid and liquid electrolytes. Equilibrium is maintained as long as the rate of electrochemical ionic drift imposed externally is much smaller than the rate of thermally activated ion exchange. The composition of Li+-Naf beta alumina, then, remains unchanged despite extended unidircctional Lit transport. To demonstrate how small the rate of electrochemical drift is compared to equilibrium ion exchange, we simplify the arrangement of mobile ions in the beta alumina conduction plane, distributing them into equivalent potential wells separated by energy barriers of about 0.17 eV (3800 &/mole). This is the activation energy of conduction typically
four
probe, con-
observed for Nat beta alumina. When a constant electric field, E, is externally imposed on the system, the probability of Na+ migration in the direction of the field, W,, is Wp = oexp -
((AG -
l/2 e&)/U)
(1)
and the probability opposing field, WB, is W, = uexp - ((AG + i/2 eaE)/kT)
(2)
In (1) and (2), v is the characteristic vibration frequency of Na+ within a potential well, e the electronic charge, and a the ionic hopping distance. In Nat beta alumina at 3OO”C, an electric field, E, of 1 V/cm sustains an ionic current of approximately 1 A/cm’. The maenitude of a is of the order 10-s m. Thirefore, the pr:duct enE is approximately lo-’ eV compared to AC which is 0.17 eV. Normal elechochemical currents, therefore, are extremely small in comparison to the rate of thermal diffusion and pose no threat to solid-liquid equilibrium. This treatment assumes a high concentration of vacancies within the structure, a factor true for Na+ beta alumina. The key to the curious ion exchange behavior of beta alumina lies in its structure. Ionic transport through beta alumina occurs in two dimensions along parallel conducting planes. Three non-equivalent crystallographic sites for mobile ion occupation have been identified in the beta alumina unit cell. The relative ionic occupation of these sites varies with ion type and temperature and is influenced by the overall fixed beta alumina structure and by interactions among the mobile ions themselves. We have previously reviewed [3] data currently available describing ionic distribution in the conducting plane and stated that we believe long range ion transport through beta alumina to proceed by ion exchange among these energetically distinguishable sites. The existence of co-ionic compositions having properties not directly proportional to those of their completely substituted end members clearly indicates that when two or more mobile ions are present simultaneously in beta alumina, they are not randomly distributed
Li+-Na+ beta alumina in the conduction plane. Rather, there is preferential occupation of specific sites by specific ions. It is this sitesite non-equivalency and mobile ion interaction that results in the non-linear concentration dependence of co-ionic transport in Li+-Na+ beta alumina and the striking variation in activity coefficients of M + in the solid and the melt observed by Yao and Kummer for (MN03pNaN03),i,,id/(M*-Na+ beta aluminaj,,l,d equilibria. Li*-Na+ beta alumina itself remains an excellent, high conductivity Li+ solid electrolyte. Assuming partial Li+ substitution produces no unexpected changes in its gross chemical behavior, it should be stable in a wide range of chemical environments, as is Na+ beta alumina, and have an extremely low electronic conductivity. Li+-Na+ beta alumina can be produced by direct ion exchange of polycrystalline Na’ beta alumina, for which there already exists a considerable fabrication technology, and can be made in a variety of physical shapes such as discs and tubes. It, therefore, may find application in ambient temperature rechargeable non-aqueous and solid-state Li batteries. Use in direct contact with molten Li+ in metal is probably not practicable. We have observed that samples of Na+ beta alumina are reduced at least within 1 h and probably more rapidly when in contact with molten Li at 450°C. A significant consideration in the use of Lif-Naf beta alumina in battery applications is the stability of its composition. The Li+/Na+ ratio in the solid must change as the activity coefficients of Lif and Naf in the media in which it is in equilibrium change. The precise Li+/Na+ ratio in the solid in contact with 1% Na+ in LiNC& at 350°C will differ from that in contact with 1% Na+ in LiC104 dissolved in propylene carbonate at 25°C. But, excepting environments in which Na’ is strongly and preferentially complexed compared to Li +, for example by formation of an insoluble Nat compound, a fair conjecture is that stability will pose little problem in actual use. Regardless, our measurements indicate that the ionic conductivity of Li+-Na+ beta alumina decreases by a factor of 12 as the exchange extent is increased from 50 to 89%. So changes in composition produce only small conductivity effects. The small variation in ionic conductivit’i with relative composition actually has only been observed to 85% Li + in Li+-Na+ beta alumina. Pure Li+ beta alumina has not been prepared in this work nor its conductivity determined. Previous measurements of the conductivity of nominally pure Lit beta alumina prepared by exchanging Na* beta alumina with AgNOX, then replacing Ag+ with Li+ in a LiNOXLiCl melt, disagree markedly as Table 1 shows. A portion of this disagreement is surely due to the different techniques used to measure conductivity. These have been reviewed in the introduction. It is also possible that the different samples measured by various groups varied considerably in composition. Exchange 01 Na+ beta alumina with Ag+ is very favorable. Subsequent equilibration with LiNO,-LiCl may induce complete AgC removal by forming AgCl in the melt; but, there is no guarantee that only Li+ replaces it in the solid electrolyte. Residual Na+ in the LiNO,-LiCl can easily and preferentially migrate into the beta alumina. Whether all the Agf itself is
771
indeed replaced is also unclear. All these possibilities raise doubts about the actual compositions previously investigated as pure Li+ beta alumina. Preparation of the compound by direct ion exchange appears to require Li+ salts of extremely high Li+ purity. The properties of Li+-Na+ beta alumina arc interesting partly because there are many potential applications for a new Li+ solid electrolyte. More important, however, may be its implication that co-ionic behavior, or the mutual interaction of two or more mobile ions producing effects not predicted by simply averaging the characteristics of their completely substituted extremes, is a general phenomenon in solid electrolytes. This is true at least with beta alumina as exchange data for Ag+, TI+, K+, Li+ and Cs+ into Na+ beta alumina of Yao and Kummer in Fig. 5 show. Just as the exchange data show that a less than ca 1% Na* in LiNO, sustains ca 50% Na+ occupation of Li*-Na’ beta alumina, so they indicate that ca 1% Ag+ in NaN03 supports cu 55% Ag+ in Na+the transport Ag+ beta alumina. Extrapolating characteristics of Li l-Na+ beta alumina, Na+-Ag l beta alumina can be expected to have a Na+ transport number of nearly unity for compositions in which Na*/Ag+ is greater than 1. Other monovalent cations such as K+ and Tl’ resemble the behavior of Ag+ to lesser cxtcnts. Cs+ and Rb+ act similarly to Li+. Co-ionic interactions in the beta alumina structure are therefore not at all unique to the Li+Na+ composition. Indeed, co-ionic behavior may be generally observed in other structure types as well. Further investigation will test this concept. SUMMARY
AND CONCLUSIONS
When Na+ beta alumina is equilibrated with molten LiN03 at 350-500°C only partial Li * replacement of Na* takes place. The very small, typically less than OS%, concentration of Na+ in the melt supports a large concentration of Na+ in the solid. Polycrystaline Na+ beta alumina can be directly converted to LiC-Nat beta alumina by direct ion exchange without cracking or obvious physical degradation. Li +-Nat beta alumina having a Li + /Na + ratio greater than approximately one is”an exdellent Li+ solid ionic conductor with a Li+ transport number of nearly one. It should find a number of practical applications in energy storage systems. What is most extraordinary about Li+-Na+ beta alumina is the preferential Na* occupation of the beta alumina conducting plane even in the presence of high Li+ activity. Li+ ions move through the structure without significantly altering its Na* content. We have termed this behavior co-ionic cmdudvity and predicted that it is a general phenomenon in beta alumina and perhaps in other structure types as well. Specifically, Na+-Ag* beta alumina has been predicted to be a Na+ conductor exhibiting preferential Ag+ occupation even in the presence of high Na+ activity. REFERENCES 1.
M. S. Whittingham and R. A. Huggins, J. chern. Phys. 54. 414 (1971).
772
G.C.
FARRINGWN
2. G. C. Farrington and W. L. Roth. Ir~rvv~frrrir~~crl Conferencr on Suprrioitrc C‘orxdltctors, (Edited by G. D. Mahan and-W. L. Roth) p. 418: Plenum Press, New York (1977). 3. W. L. Roth and G. C. Farrington, article submitted to Science. 4. Y. F. Yao and J. T. Kummer, J. inorg. nucl. Chem. 29. 2453 (1967).
AND W. L.Rom 5. R. H. Radzilowski, Y. F. Yao and J. T. Kummer, J.. appl. Phys. 40, 4716 (1969). 6. &I: S. Vi’hitti&ham -and .R. A. Huggins, Solid State Chemistry, p. 139. N,B.S. Spec. Pub. No. 364, (1972). 7. R. W. Powers, J. dectrochem. Sot. 122, 490 (1975). 8. G. C. Farrington, J. elrctrochem. Sot. 121, 1314 (1974). 9. G. C. Farrington, J. electruchem. Sot. 123, 1213 (1976).