Mat . Res . Bull . Vol. I I , pp. 203-220 , 1976. P e r g a m o n P r e s s , Inc. P r i n t e d in the Uni ted S t a t e s .

FAST Na + - ION TRANSPORT IN SKELETON STRUCTURES*

J. B. Goodenough, H. Y-P. Hong, and J. A. Kafalas Lincoln Laboratory, Massachusetts Institute of Technology

Lexington, Massachusetts 02173

(Rece ived D e c e m b e r 17, 1975; C o m m u n i c a t e d by J . B. Goodenough)

ABSTRACT Skeleton s t ructures have been explored experimental ly for fast Na +- ion t ransport . A skeleton s t ructure consists of a rigid skeletal a r r a y of atoms stabilized by electrons donated by alkali ions part ial ly occupying sites in a three dimensionally linked in ters t i t ia l space. Fast Na +- ion t ranspor t was demonstrated in several s t ructures , and the system Nal+xZr2P3_xSixO12 has a Na +- ion res is t iv i ty at 300°C of P300~<5fl-cm for x,~2, which is com- petitive with the best S" - alumina. An activation energy 8a~0 .29 eV is about 0.1 eV larger than that of ~" -alumina.

Introduction 1 +

The discovery of fast Na - ion t ranspor t in 8 - and 8"-alumina has st imu- lated in teres t in the use of solid electrolytes in ceils and thermoelect r ic genera tors . 2 The best compositions have res is t iv i t ies for fast Na +- ion t ransport at 300°C of P300 = 4fi-cm with an activation energy for the mobility E a =0.16 eV 3. The volatil i ty of sodi- um together with the re f rac tory charac ter of A1203 has made awkward economic fab- r ication of ceramic membranes , but this problem appears to be solved. 4 Neverthe- less , ~-and B"-alumina are layer compounds in which the Na + ions are constrained to move in only two dimensions. Anisotropic thermal expansion 5 can reduce the life of thermaUy cycled membranes , and confinement of the Na + ion to widely sepa- rated layers reduces sharply the fraction of the membrane volume that t ransports N a ; ions. Therefore , the design of a mater ia l that can provide equivalently fast Na -ion t ranspor t in three dimensions has challenged solid state chemists . We report here on a s t ra tegy and related experiments that have enabled us to meet this challenge.

Skeleton Structures

Our s t ra tegy was to investigate alkali- ion t ranspor t in cubic "skeleton" s t ructures . A skeleton s t ructure consists of a rigid subarray with interconnected in ters t i t ia l space in which ions move in three dimensions. Ionic mobility requires the existence of an interconnected space of part ial ly occupied lattice sites having equiva-

*This work was sponsored by the Defense Advanced Research Projects Agency, NASA, and NS F/RANN

203

204 J . B . GOODENOUGH, et a l Vol. 11, No. 2

lent (or near ly so) site-occupation energies . For fast ion t ransport , the activation energy for an ion jump from one site to another must be small . Therefore , the site- interface "bott lenecks", whether shared faces or shared edges, must be open enough for easy ion passage.

According to the simple Einstein theory for ion mobility, the conductivity is given by

= zNc(1-c)e2a; - v /kT (1)

where z is the number of neares t -neighbor alkali- ion sites of density N, c is the con- centrat ion of M + ions on the available set of interconnected si tes, e is the c a r r i e r charge, a£ is the distance between si tes. The jump frequency is

v = v o exp (S/k) exp (-ga/kT) (2)

where the Gibbs free energy for a jump is ~a " TS. The factor c(1-c) defines the requirement for par t ia l site occupancy, and the activation energy ~a is a measure of the "bottleneck ba r r i e r " to ion t r ans fe r between equivalent lattice si tes. The factor a~v o imposes the need for short jump distances, since ~o is l imited by the elast ic vibrat ional-mode frequencies. Short-range correla t ion may decrease the pre- exponential factor . 6

With these principles in mind, we have investigated four c lasses of skeleton s t ructure: the Im3 phase of h igh-pressure KSbO3, the defect-pyrochlore s t ructure i l lus t ra ted by RbMgA1F 6, the carnegiei te s t ructure of high- temperature NaA1SiO 4, and the system Nal+xZrzP3_xSixO12, 0 ~ x < 3.

A. The Im3 Phase 7

A primit ive-cubic Pn3 phase of KSbO 3 was f i r s t discovered by Spiegel_berg. The s t ruc ture is composed of an (SbO3)- subarray, the skeleton, with an ordered a r rangement of K + ions within <111> tunnels that t r ave r se the skeleton, intersect ing at the origin and body-center positions, see Fig. 1 The <111> tunnels a re made up of face-shared octahedral si tes that a re compressed along the tunnel axis so as to provide a large spacing between the bridging O 1 and O 2 oxygen within a shared face. Along any <111> axis, three octahedral si tes separate an origin and body-center position, and the order of the shared faces is O1-O2-O2-O 1 , see Fig. 2. Alkali-ion sites having 01 and 02 faces a re labeled M l sites, those with only 02 faces are M 2 sites. The unit cell con-

tains K12Sb12036, the K ÷ions occupying the M 1 and M 2 positions in an ordered manner so that each tunnel segment contains e i ther two M 1 cations or one M 2 cation, and an M 2 cation has only M 1 - cation neares t neighbors in the neighboring tunnel segments.

The stoichiometric , a tmospher ic -p ressure phase of KSbO 3 has the rhombohedral i lmenite s t ructure . During an investigation of the s t ruc tura l relat ion- ships among several A~B5+O 3 compounds, we 8 found a cubic h igh-pressure phase with space group Im3. This Im3 phase has the same skeleton as the primit ive-cubic Pn3 phase discovered by Spiegelberg; but it contains K + ions randomly distr ibuted among

the M 1 and M 2 positions.

Three charac ter i s t ics of this s t ructure are of interest : (1) Although the M 1 and M 2 sites a re not crystaUographical ly equivalent, their si te-occupation energies appear to be near ly so. (2) The distance in KSbO 3 between M 1 and M 2 positions of a ttm~el segment is only a12 =ol. 83.~, that between M 1 positions of neighboring tunnel segments is only a l l = 2 .66£. (3) The distance from the center of an 02 tr iangle to an

Vol. II, No. 2 FAST Na+-ION TRANSPORT 205

FIG, 1 The (SbO3) a r r a y of cubic t(SbO 3.

:f , - , ,-,

FIG. 2 Schematization of the eight <111> tunnel segments from the origin to neighboring body-centered positions in Im3-cubic KSbO 3. In the ordered Pn3 structure, each singly occupied branch has an M 2 cation and the neares t -neighbor branches are doubly occupied by M 1 cations.

1001 )

FIG. 3 The (110) projection of a ( ~

cubic pyrochlore Bi2Rh207 showing the (Rh206) skeleton of co rne r - sha red octahedra and 8b-16d-8b passageways paral le l to [110].

(fro)

®

• Bismuth

0 Rhodium

O Oxygen

I

J

0 2 ion is 2.5 A, indicating that the "bottleneck" for Na + - ion t ranspor t within a segment is not severe , and t rans fe r between M 1 sites to neighboring tunnel segments is constrained by the crysta l l ine fields to go through the t r iangular face created by an O1-O 1 edge and the origin or body-center position. These charac ter i s t ics were suf- f iciently promising that we decided to measure alkali- ion t ranspor t in compounds with this (SbO3)- skeleton even though the chemical stabil i ty of the skeleton in the presence of molten alkali metals is probably not adequate for pract ical applications.

206 J . B . G O O D E N O U G H , e t al Vol . 11, No. 2

B. The Defect Pyrochlore AB2X 6

The cubic pyrochlore s t ruc ture cor responds tO the chemical formula A2B2X6X', where A is a large cation and B is a sma l l e r cation octahedral ly coordinated by six X anions. The B2X 6 subar ray forms a skeleton of c o r n e r - s h a r e d octahedra, see Fig. 3. In the interpenetrat ing A2X' subar ray , each X' anion is te t rahedral ly coordi- nated by four A cations, and each A cation has two neares t -ne ighbor X' ions on oppo- site s ides. With space group Fd3m and the origin chosen at a B si te, the A and X' positions a r e identified as 16d and 8b, the cubic unit cel l containing AI6BI6X48X' 8. The B and X positions a re the 16c and 48f posit ions of the space group.

Defect pyrocholores have a stable B2X 6 skeleton with vacancies introduced into the A2X' subar ray . In PbRuO 3 and AgSbO3, for example, the X' a toms at the 8b s i tes a re missing. 9, 10

Babel et al. I I p repared a number of f lourides A+B2+B '3+ F 6, where A = Cs, Rb, or K. T h e s e f l u o r i d e s have a B2+B'3+F6 skeleton identical to the B2X 6 pyro- chlore skeleton, the B and B' a toms being randomly distr ibuted. In place of the A2X' subarray of a pyrochlore , only the 8b positions a re occupied by a large A + cation. In seve ra l oxides, d isplacement of the A cation from an 8b toward a 16d (to occupy 32e) site has been signaled. 12 Singerl3 encouraged us to explore these compounds for fast alkali-ion transport. Since the A site is a strongly squashed octahedron presenting a minimal bottleneck barrier for ion transfer between 8b and 16d positions, we decided t~pulr~ ' t~$ strategies: (I) Establish vacancies on the 8b positions by stabilizing

AI_xBI_xBI+xF 6 compounds, thus making non-zero the factor c(l-c) in Eq. (i) for 8b occupancy since the 16d sites are not energetically equivalent. However, this strategy suffers from a long jump distance abL between 8b positions. (2) Increase the cell size

u 4- until the crystal-field preference of the M ions for 16d sites balances the size prefer- ence for 8b sites, thus making the 8b and 16d positions energetically equivalent. This latter strategy seemed preferable since the jump distance aab is only half ahb. More- over, this should be the condition for minimum ga" This strategy was complicated by water contamination, in the case of K +- ion transporters, and by excess Na20 in the case of Na +- ion transporters (see below).

C. Carnegiei tes

The aluminosi l icates a re c lass ic skeleton s t ruc tures . The zeoli tes, for example, form molecu la r s ieves . As these la t ter s t ruc tu res a re stabil ized by water , they a re unsuirnble for solid e lec t ro ly tes that a re to be in contact with molten alkali metal . Moreover , the openings a re too large for optimal alkal i- ion t ranspor t . The h igh- tempera ture form of NaAISiO 4, carnegiei te , is a more interest ing possibil i ty.

Carnegiete has a cubic (AISiO4)- skeleton having the s t ruc ture of cubic SiO2, see Fig. 4. In the idealized s t ruc ture , the Al and Si a toms form a cubic zinc- blende a r r a y with oxygen a toms on every A1-Si bond axis. This a r rangement provides a network of c o r n e r - s h a r e d te t rahedra that, in the r ea l carnegiei te s t ruc ture , becomes dis tor ted to a pr imi t ive-cubic a r r a y by interact ion with the Na + ions in the in ters t i t ia l space. 14 In the idealized s t ructure , the inters t i t ia l space is identical to the space occupied by the skeleton, as is wel l i l lus t ra ted by the two interpenetrating, ant icar - negieite sublat t ices of Cu20. 15 Transla t ion of the skeleton by ao/2 along a cube edge would super impose it on the inters t i t ia l space. Thus the in ters t i t ia l space has two distinguishable s i tes , M 1 and M 2, corresponding to the Al and Si s i tes of the skeleton, and any anion excess would be located on the M I - M 2 axis in the shared face of the M 1

Vol. II, No. 2 FAST Na +-ION TRANSPORT 207

iiiii",,/i"

FIG. 5 l=S i O=AI @ -0

FIG. 4 The (A1SiO4)- subar ray of ca rne- gieite.

0 0

V / k v ~ : - -J ° \ t / - _ ~ " ~ ,

\ / /ok k/ / \ \ / / A V • . / _ Z L _ ~ . ~ t " , , I P ° \ / ¢ ~ , , I °" \ . ! ..

..... IX z X ........ :,*- ......

o o

View of rhombahedra l R3c s t ruc ture of

NaZr2P3012 showing the (ZrP3012)- units para l le l to c r and the Na+ ions in ~ positions octahedra l ly coordinated by

- ions. The M 1 positions a r e also oc- tahedral ly coordinated by empty M 2 posi- tions (not shown) in the same basal planes as the nea res t -ne ighbor 02- ions.

3

2 " "-., / .....

2 i

3

(a) R3c FIG. 6 (b) C2/c

Basal or a m - b m plane of Zr -oc tahedra and P- te t rahedra neighboring an M 1 showing the octahedral coordination of M 2 sites (c i rc les) .

si te (M)

and M 2 s i tes . This face consis ts of a hexagon having its six sides a l t e rna te ly the edges of SiO~ and A1OA te t r ahedra . In the absence of excess anions, which would block M + - ion transfer:, the b a r r i e r to M + - ion motion would seem to be a minimal c rys t a l - f i e ld energy.

D. The System Nal+xZr2P3_xSixO12

From gene.~al considerat ions , an open skeleton allowing th ree -d imens iona l

208 J.B. GOODENOUGH, et al Vol. 11, No. 2

ionic transport would have skeleton anions linked to at most two skeleton cations. The stronger these skeletal bonds, the more stable the skeleton. The Ira3 and defect- pyrochlore skeletons are composed of linked octahedra, the carnegieite skeleton of linked tetrahedra. A remaining alternative is a skeleton of linked tetra~edra and octa- hedra. To explore this possibility, we initially chose (PO4)3" and (SO4)'~" tetrahedra to be linked by a cation with an octahedral-site preference. Since oxides of the octa- hedral-site cation should not be reduced by moke~+alkali metals, the list of suitable ions is limited. Of the quadrivalent ions, the Zr ion was a natural candidate for the octahedral-site cation. One of us 16 preheated a mix of Na2CO 3 + ZrO 2 +2NH4H2PO 4 at 170 ° C to decompose NH4H2PO4, again at 900 ° C to decompose Na2CO 3, and heated to 1600 ° C for 3 hrs before turning off the furnace. Cube-shaped crystals about 0. Imm on an ec~ge were examined crystaHographically and shown to be NaZr2P3OI2 with space group R3c. It was also found that all the phosphor may be replaced by silicon, giving the system Nal+xZr2P3.xSixOl2 , 0 < x < 3. All structures are rhombohedral R3c, ex- cept in the range I. 8 < x -< 2.2 where a distortion to the monoclinic C2/c space group is found at room tempera ture . We subsequently d iscovered that the s t ructure~7of 8 NaZr2P3OI~ and Na4Zr2Si3012 had a l ready been repor ted in the l i tera ture . '

In the (Zr2P3_~SixO12)(l÷x)-skeleton, each Zr 4+ octahedron shares its six corners with te t rahedra , a n o e a c h te t rahedron shares its four corners with octahedra, see Ref. 16 and Figs. 5, 6. Thus each anion bonds s trongly to a te t rahedra l and an octahedral cation of the skeleton. In the rhombohedral phase, two M +- ion si tes a re distinguishable in the inters t i t ia l space: c lose-packed-hexagonal M 2 layers in the basa l plane connected by one- th i rd 'as many M 1 s i tes between the M 2 layers . Thus the inters t i t ia l space is con- nected in three dimensions, and the anion bott leneck between M 1 and M 2 positions con- s i s t s of a puckered hexagonal ring having as its six sides ake rna te ly t e t rahedra l - s i t e and oc tahedra l -s i te edges of the skeleton. The monoclinic distort ion distinguishes two M 2 s i tes : 3M 2 -~ M 2 +2M 3 as shown in Fig: 6(b). The jump distances for monoclinic Na3Zr2PSi2012 lie in the range 3.51 - 3 . 8 A .

Transpor t Measurements

Transpor t measurements were made on dense (in excess of 85 percent theoret ical) , polycrysta l l ine ce ramic disks approximately I /2" in d iameter and 0.05" to 0. i " thick.

Ionic conductivity was measured with an ac vec tor - impedance me te r (5 Hz to 500 kHz). Gold o r platinum blocking e lec t rodes , which permi t e lec t ron t ranspor t but r e s t r i c t ionic t ranspor t to a displacement current , were used to determine any electronic compo- nent. Nonblocking e lec t rodes , consist ing of a coating of colloidal graphite on both sides of the sample, permit Na +- ion discharge from the e lect rolyte , polarization at higher fre- quencies general ly remaining t r iva l to f requencies as low as 500 Hz. For compar ison purposes , some dc and low-frequency ac measu remen t s were made with molten NaNO 3 on both sides of the specimen.

Resuks

A. NaSbO 3

A metastable "NaSbO3" phase having the cubic Im3 s t ruc ture was o t ~ i n e d by ion exchange, the high - p r e s s u r e Im3 phase of KSbO 3 or ofT~SbO 3 being immersed for a few hours in molten NaNO 3. S ingle-crys ta l s t ruc ture analysis , obtained by ion exchange of a c rys t a l of TgSbO3, located the Na + ions randomly distr ibuted among the octahedral M 1 and M 2 si tes of Fig. 2. This analysis produced three significant findings:

Vol. I I , No. 2 F A S T Na + - I O N T R A N S P O R T 209

(1) The apparent sodium concentra t ion is in excess of s to ichiometry , suggesting a chemica l formula Nal+xSbO 3 with x ~ 0 . 2 9 . This apparent excess could be at least par t ia l ly due to an incomplete exchange of the heavier T ~ + ions, since only a single exchange bath was used.

(2) T h e r e was no measurab le e lec t ron density in the tunnel in tersec t ions at the origin and body-cen te r positions where excess 02- ions were antici- pated.

(3) the disordered Im3 phase, separations of only 230 appar- / /

ently occur , especia l ly as the occupancy fac tors of the M 1 and M 2 si tes were in the rat io 0 . 8 2 / 0 . 2 8 r a the r than 2/1.

Since the re was no measurab le e lec t ronic component to the conductivity at 300 ° C, it is concluded that the ion-exchanged composit ion must be NaSbO 3 • yNa20 , y<-O. 14, and that the excess oxygen may be displaced toward empty M 1 si tes , thus blocking a f rac t ion of the tunnel segments to Na+ - ion t ranspor t .

High p r e s s u r e stabi l izes the cubic Im3 phase of KSbO 3 because this phase is more dense than the i lmenite KSbO 3. For NaSbO 3 the situation is r eve r s ed , which l imits the p r e s s u r e that can be applied for densffication of the metas table , cubic NaSbO 3. Never the less , we were able to obtain specimens of 92-98 percent theore t ica l density by adding 2 weight percent NaNH 2 flux to a mixture of fine (< l~zm) powder placed in a tung- s ten-carb ide d ie-and-pis ton assembly . After evacuation to < 10 -3 T o r r , a p r e s s u r e of 30, 000 psi was applied and the t empera tu re was ra i sed to 600 ° C for an hour o r so before

the p r e s s u r e was r e l eased and the specimen was baked out at 500 ° C.

Conductivity measuremen t s showed no measurab le e lec t ronic contribution and an ionic r e s i s t iv i ty at 300 ° C of 0300 = 18 ~ c m at lkHz with graphite e lec t rodes . The fact that this r e s i s t iv i ty is only a fac tor four l a rge r than that of the best 8"-a lumina sam- ples at 300 ° C demonst ra ted to us the validity of the ske le ton-s t ruc tu re concept.

1 B. NaSbO 3 • ~ NaF

Since the tunnel - in te rsec t ion si tes a re empty in NaSbO3, it suggests that they repe l posit ive ions and a t t r ac t negative ions. If so, the Na + ions jump f rom M 1 site to M 1 site of neighboring tunnel segments without passing through the tunnel in tersec t ions . Moreover , p lacement of an anion at the in te rsec t ion si tes could stabil ize a cubic phase of KSbO 3, thus explaining Spiegelberg 's stabil ization at a tmospher ic p r e s s u r e of the Pn3 phase by annealing for three weeks at 1000 ° C in a porcela in crucible . Roth 19, suspect- ing si l icon contamination of Spiegelberg 's c rys ta l , was the f i r s t to investigate impur i ty stabil izat ion of cubic KSbO 3. He subsequently t r i ed stablization of KSbO~ by f luorine sub-

- • 2 0 s " " " sntut lons and repor ted the atmo pherlc - p r e s s u r e p repara t ion of K 1 gbO.3. ^--Fv" Since he fai led to r epo r t any chemical analys is , it is probable that his phase-wXas I(S-~O3 • xKF, 0 < x < 1/6. We independently investigated the a t m o s p h e r i c - p r e s s u r e phase KSbO 3. 1/6 KF, since we were in te res ted not only in the possibi l i ty of el iminating a h igh -p re s su re step in the prepara t ion of NaSbO 3 • 1/6 NaF, but also in whether the exis tence of an anion at the tunnel in te rsec t ions would aid o r hinder Na + - ion mobility.

The p r e c u r s o r KSbO 3 • 1/6 KF was p repared by f i r ing a 2:1 by-weight mix- tu re of K2H2Sb207 • 4H20:KF at 900 ° C for 2 hours. A cubic product was separa ted f rom the excess KF by washing in water . This product was ion-exchanged th ree success ive t imes with NaNO 3 at 330 ~ C in a 20:1 by-weight mixture of NaNO3:product. The t r ip le -

210 J . B . G O O D E N O U G H , e t a l Vol . 11, No. 2

ion-exchanged product, which remained cubic, was ho t -p re s sed into disks for ion- t r anspor t measu remen t s .

In o r d e r to c h a r a c t e r i z e the products with s ing le -c rys t a l s t ruc tu re analysis , a mixture of 20g potassium pyro-ant imonate and 100 g KF was f i red in a covered platinum crucible at 1100 ~ C for 12 hours and then p rog rammed down at a ra te of 5 ° C/hour to room t empera tu re . Many well shaped single c rys t a l s about 0.05 mm in size were ob- tained. Powder diffract ion of the specimen showed a body-cente red-cubic pat tern. A c rys t a l measur ing 0.03 x 0.03 x 0.05 mm was se lected for s t ruc tu ra l analysis . The sodium analog was p repa red by ion exchange in molten NaNO 3. The sodium c ry s t a l se lected for s t ruc tu ra l analys is measured about the same size as that of the potasium analog.

Oscillation and Weissenberg photographs of both c rys t a l s showed diffract ion s y m m e t r y m3. The sys temat ic absence was hk~ for h + k +g = 2n + 1, which is con- s is ten with space groups I23, I213, and Im3. The space group Im3 was ass igned as de- t e rmined for the Im3-NaSbO 3 s t ruc ture . 8 Three -d imens iona l intensity data to 2e = 55 ° were col lected in the same fashion as in Ref. 8. A total of 605 ref lec t ions was measured for the KSbO 3 - KF product, 498 for the sodium analog.

A s t rong s imi la r i ty between the powder pat terns of Im3-NaSbO 3 and the new phase induced us to use an identical (SbO3)- skeleton for the initial ref inement . The l e a s t - s qua re s p rogram gave a re l iabi l i ty fac to r R=0. 17 for the KSbO 3 • KF product, R = 0. 14 for the ion-exchanged c rys t a l . From the calculated s t ruc tu re fac tors based on this model, a di f ference Four ie r map in both cases revea led two independent a lkal i - ion posit ions at (x, x, x): a l a r ge r e lec t ron density at x = 0.15 and a lower at x = 1/4. Unlike the Im3 - NaSbO 3 pattern, a considerable e lec t ron density appeared at (0, 0, 0), o r the tunnel - in te rsec t ion si tes 2a. We assigned F- ions to the 2a site in both cases , c o r t e s - pending to composi t ional fo rmulas KSbO 3 • 1/6 KF and NaSbO 3 • 1/6NaF, and obtained re l iabi l i ty fac tors R = 0. 077 and R = 0.064, respec t ive ly , a f te r a few ref inement cycles with anisot ropic t em pe ra t u r e fac tors . The final analyses a r e summar ized in Tables 1-3.

Comparison of the Im3 - NaSbO 3 and the NaSbO 3 • 1 /6NaF s t ruc tu res is in- s t ruct ive . The Im3-NaSbO 3 product of ion exchange with cubic KSbO 3 or T~SbO 3 contains excess e l ec t ron densi ty, p resumably indicating NaSbO 3" xNa20 . The apparent occupan- c ies of the sodium-atom sites a r e 29 percen t at 8c (M 2) and 82 percent of 16f (M1); the tunne l - in te rsec t ion si tes 2a a r e empty. The 16f positions form a simple cube about each 2a site, and the shor tes t 16f-16f (M1-M 1) separat ion is only 2.27.~ in Im3-NaSbO 3. This surpr is ingly shor t distance seems to imply the displacement of the excess 02- ions toward vacant 16f positions. In the NaSbO 3 • 1/6NaF s t ruc ture , on the other hand, each 2a site is occupied by an F" ion and there is no apparent sodium-atom excess over the chemical formula . The sodium-atom site occupancies a re 25 percent at 8c and 80 percent at 16f,

and the shor tes t Na+-Na + distance is 2.85~.

An anomalously high t empera tu re coefficient is observed for the Na + ions in positions 16f, the anisotropic t empera tu re coefficient indicating that the the rma l motion along cubic axes is six t imes l a rge r than that along cubic diagonals. I sos t ruc tu ra l KSbO 3 • 1/6KF, which also contains no M + - ion excess , has a t empera tu re coeff icient only one- th i rd as large. These findings a re consis tent with higher Na +- ion mobili ty in NaSbO 3 • 1/6NaF and with a hopping of the Na + ions between 16f positions of different tun- nels that not only is around the 2a s i tes , but also is e a s i e r than hopping within a tunnel

segment.

Vol. 11, No. 2 FAST Na + - I O N T R A N S P O R T 211

T A B L E 1

F i n a l A t o m i c P a r a m e t e r s f o r KSbO 3 • 1 / 6 K F *

Atom o.f.* ~ x y z S 11 q 22 S 33 S 12 ~ 13 ~ 23

K(1) 8c 0.32(3) I/4 1/4 1/4 0 .024(4) 0.024(4) 0.024(4) 0 "0 0

K(2) 16f 0.66(2) 0.1582(6) 0.1582(6) 0. 1582(6) 0.0072(6) 0.0072(6) 0.0072(6) 0.0038(5) 0.0038(5) 0.0038(5)

Sb 12e 0.8421(1) 0 1/2 0.0011(1) 0.0013(1) 0.0011(1) 0 0 0

O(1) 12d 0.363(1) 0 0 0.000(1) 0.001(1) 0.005(2) 0 0 0

0(2) 24g 0 0.346(1) 0.292(1) 0.005(1) 0.002(1) "0.000(1) 0 0 0.0004(9)

F 2a 0 0 0 0.004(2) 0. 004(2) 0.004(2) 0 0 0

*Number of K a toms pe r unit cel l = 1 3 . 1 -+ 0 . 3 ; Space group: Im3; ceil p a r a m e t e r : a = 9 . 606(5)h *Occupancy fac tor ,

T A B L E 2

F i n a l A t o m i c P a r a m e t e r s f o r NaSbO 3" 1 / 6 N a F *

Atom o, f. :' * x y z ~ 11 822 S 33 S 12 S 13 S 23

Na(1) 8c 0.25(4) 1/4 1/4 1/4 0.009(4) 0.009(4) 0.009(4) 0 0 0

Na(2) 16f 0.80(4) 0. 154(1) 0. 154(1) 0. 154(1) 0. 024(2) 0. 024(2) 0. 024(2) 0.004(1) 0.004(1) 0.004(1)

Sb 12e 0.8381(1) 0 1/2 0.0032(1) 0.0037(1) 0.0033(1) 0 0 0

O(1) 12d 0. 360(1) 0 0 0.000(1) 0.006(1) 0. 005(1) 0 0 0

0(2) 24g 0 0.3325(9) 0.2873(9) 0.006(1) 0.003(1) 0.005(1) 0 0 0.0003(9)

F 2a 0 0 0.004(1) 0.004(1) 0. 024(1) 0 0 0

*Number of Na a toms per unit ce l l = 14. 7 -+ 0 .6; Space group: Ira3; cel l p a r a m e t e r : a = 9. 306(1)h **Occupancy fac tor .

Akhough NaSbO 3" 1/6NaF has moderate tempera ture stability, it cannot be heated to tempera tures near the meking point and it cannot sustain very high p ressu res . At 5000 atmospheres and 600 ° C, there is a par t ia l disproportionation and t ransformat ion to the more dense i lmenite phase of NaSbO 3. Therefore , hot -pressed disks for ion- t ranspor t measurements were prepared in the same manner as the NaSbO 3 disks. Samples of 92-percent density had 1-kHz res is t iv i t ies at 300 ° C of 0300 = 13 and 17 Q-cm, respectivi ty, for measurements made with graphite and molten NaNO 3 elec- t rodes . With d .c . and molten NaNO 3 e lec t rodes ,a P300 = 18 Q-cm was obtained. Fig. 7 shows a plot of res is tance vs 1/T for a typical specimen of NaSbO 3" 1/6NaF. The slope yields an activation energy of 8a~0.35eV for a Na + - ion jump.

From these measurements we deduce three conclusions: (1) Cubic skeleton s t ruc- tures have the improvement in pre-exponential factor of Eqs. (1) and (2) anticipated for bulk vs layer Na +- ion conduction. (2) Placement of an anion at the t, mnel intersect ions does not impede the Na +- ion mobility, thereby confirming that the Na + ions avoid these positions in cubic NaSbO3. (3) Reduction in the Nat- ion activation energy 8 a for hopping would make the t ranspor t propert ies of cubic skeleton s t ruc tures superior to those of B-alumina.

212 J . B . G O O D E N O U G H , e t a l Vol . 11, No. 2

Bond D i s t a n c e s (in ~ )

Octahedron around Sb

S b - O(1) 2x - 0 (2) 2x

2x

Octahedron around alkali ion M(1)

M(1) - 0(2) 6x

Octahedron around alkali ion M(2)

M(2) - 0(1) 3x - 0 (2 ) 3x

Cube around F

F - M(2) 8x

T A B L E 3

in KSbO 3 • 1 / 6 K F and NaSbO 3 • 1 / 6 N a F

KSbO 3. 1/6KF NaSbO 3. 1/6NaF

2. 008(6) 1.994(6) 1. 960(9) 1.947(8) 1.995(9) 1.979(8)

2. 604(4) 2.474(3)

2. 913(6) 2.79(1) 2.686(9) 2.52(1)

2. 632(6) 2. 482(9)

10,000

I000

eJ

| 0 0

10

Of

I , I I I 2 3

1 0 0 0 / T

F I G . 7 Resis tance vs 1/T for a typical ceramic specimen of NaSbO 3 • NaF with graphite e lec t rodes at lkHz.

In o rde r to determine how ga changes with lattice parameter , we at tempted to study the Na+-ion conductivity as a function of p re s su re . Unfortunately the die broke before we could obtain quantitative data, but the data c lea r ly indicated that the conduc- t ivi ty dec rea se s with the latt ice pa rame te r (d~/da o >0). Since the pre-exponent ia l fac tor should be re la t ive ly p ressu re - insens i t ive , we may conclude that a smal le r 8a can be achieved by increas ing the lattice pa ramete r a . We have found that i sos t ruc-

Vol. II, No. 2 FAST Na+-ION TRANSPORT 213

rural KJ3iO 3 is hygroscopic, but it would be interesting to explore the system NaSbl-xBixO 3 • 1/6NaF to see how ga changes with cell size for those compositions where H20 contaminntion is not a problem. Since our objective was a pract ical elec- t rolyte stable against molten sodium, we did not pursue this investigation.

C. Defect Pyrochlore

initially we prepared the defect pyrochlore "KTaWO6" by conventional ceramic techniques. A nominal "NaTaWO6" was prepared by ion exchange in molten NaNO 3. At 300 ° C, the res is t iv i ty of this compound was about a factor 10 la rger than that of NaSbO 3- 1/6NaF, but a lower activation energy made it interesting. Because W 6+ ions would be reduced to W 5+ ions by molten sodium, we turned to a study of the chemically more stable pyrochlore A+Ta205F . In the course of this study, we found that the pyrochlores may be hydrated, and a subsequent investigation of nominal NaTaWO 6 showed it to be, in fact, part ial ly hydrated at room tempera ture : NaTaWO 6. xH20 with x ~ 0.5.

The compound RbTa205F can be synthesized direct ly. Equal amounts of dry RbF and reagent-grade Ta205 were mixed in a crucible and f i red at 750 ° C under an argon atmosphere for two days. Regrinding and ref i r ing two t imes produced single-phase powders. Single c rys ta l s of RbTa2OsF were prepared by addition of excess RbF as a flux. In one run, a 4:1 rat io of RbF and Ta205 was f i red at 1150 ° C for 30 minutes and quenched. The product contaIned crys ta ls about 0.1 mm across .

The K and Na analogs could be synthesized directly. If RbTa205 is placed ha molten KNO3, the ion-exchange product is KTa205F. Chemical analysis gave Rb:Ta and K:Ta rat ios of 1:2, even on revers ing the exchange. On exposure to a i r KTa20~F becomes hydrated to KTa20~F. H20. The lattice parameter of RbTa205F is 10.496~, of KTa205F" H20 is 10.605~. TGA measurements showed a revers ible , broad t ransi - tion from 50 ° to 200°C having a weight loss of one H20 molecule per K2Ta2OsF mole- cule. Ion exchange from RbTa205F with molten NaNO 3 did not yield the completely ex- changed product NaTa205F. It was necessary to ion exchange f i r s t with K + ions and then with Na + ions. Chemical analysis gave a Na:Ta ratio of x:2 where l < x < 3.

In order to determine the ionic distribution over the inters t i t ia l 16d and 8b sites, single c rys ta l s were selected for s t ructure analysis . Oscillation and Weissenberg photographs showed diffraction symmet ry m3m for the products of both potassium and sodium exchange. The systematic absence was hk0 for h + k =4n +1, which is consistent with the space group Fd3m reported for the pyrochlores. Moreover, an additional ab- sence occurs wherever the hk~ are all even and the index is not 4n, indicating that posi- tion 32e of Fd3m cannot be occupied. Thus, as in the normal pyrochlore s t ructure , the X 6 =O5F atoms are randomly distr ibuted over the 48f si tes, and the B---Ta atoms occupy the 16c sites of the BX 3 skeleton. To determine the alkali-ion distribution over the 16d and 8b sites, three-dimensional intensity data to 20 =60 ° were collected and analyzed as described in Ref. 8. Table 4 summar izes the resul ts of the s t ruc tura l ref inements .

In RbTa205F, the 8b sites are completely occupied by Rb atoms and the 16d sites a re empty. The 8b site has 6 neares t -neighbor X atoms that form an ideal octahedron and 12 next-neares t -neighbor X atoms. The Rb-X distances are 3.35~ and 3.78~, respect ively. A lack of electron density at the 16d sites is consistent with the short distance, only 2.27~ from 8b to 16d.

In KTa205F" H20, x - r ay data show electron densit~ at both the 16d and 8b sites in ratio 2.5:1. There are two possible explanations: (1) KTions are randomly dis-

214 J . B . GOODENOUGH, et al Vol. I I , No. 2

tributed over 16d and 8b sites, or (2) K t ions occupy half the 16d sites and water occu- pies the 8b sites. Since the 02- ion has an 8b-site preference and since just one H20 molecule is accepted per molecule, the latter akernative can be inferred. Therefore, we conclude that the large cell size for the K analog corresponds to the hydrated com- pound KTa205F. H20 , the water occupying the 8b sites, even though the distance 2.'29~ between 8b and 16d sites gives a surprisingly short K-O separation.

Whereas chemical analysis of the sodium analog gave a Na:Ta ratio x:2, with x~2, Table 4 gives an apparent Na +- ion excess equivalent to x~3. We thought this discrepancy might be due to the presence of OFF ions giving Na2Ta205F" OH, but neith- e r IR data nor TGA experiments nor the preparation of "NaTa205F" from anhydrous KTa205F gave any evidence of the presence of OH- ions. Since electr ical measurements gave no measurable electronic component to the conductivity, we concluded that the ex- cess sodium is charge-compensated by 02- ions. Anion analysis (by vacuum fusion at MIT) of a sample having a Na:Ta ratio x:2, with x~1.55, gave Na1.55Ta205Fm00.28 corresponding to NaTa205F. 0.28 Na20. The excess 02- ions presumably occupy 8b positions. X-ray intensities indicate that some Na + ions also occupy the 8b sites.

Excess Na20 appears characterist ic of many fast Na+-ion conductors, in- cluding (l+x)Na20" IIAI203. In NaTa205F.xNa20, the amount of excess Na20 and the lattice parameter are found to vary With the temperature and duration of the ion- exchange process , giving 10.44 <ao< i0.48A. Similar results were obtained when mol- ten NaCO 3 was used for the ion exchange. The potential for oxidation was not even avoided by exchanging with NaNH 2, since it melts at a lower temperature where the

KTa205F • H20 is hydrated.

Preparation of hot-pressed, ceramic disks of the sodium compound is complicated by the fact that both temperature and pressure induce disproportionation to TaO2F and the perovskite NaTaO 3. Specimens approaching 82 percent theoretical density were prepared by the procedures used to make pellets of NaSbO 3. A P300 ~ 150 ~-cm and an

T A B L E 4

Final A tomic P a r a m e t e r s fo r M T a 2 0 5 F (M = Rb, K, Na x)

RbTa2OsF KTa205F NaxTa205F

Cell size a (~) i0. 496 (3) I0.605 (3) 10.470 (2)

48f x 0. 306 (2) 0. 305 (2) 0. 308 (1)

t 0. 1 (4) 0.6 (4) i. 8 (2)

16d o 0 0.5 0.96 (5)

t 17 (3) 5.1 (6)

16c ~ 0.65 (6) 0.98 (6) i. 08 (2)

8b o 1 H20 1.05 (7)

5.4 (5) 15 (6) 5.7 (8)

R 0.10 0 .10 0 .053

x = position parameter of O5F = isotropic temperature factor

o = occupancy factor R = discrepancy factor

Vol . I I , No. 2 FAST N a + - I O N T R A N S P O R T 215

.

T A B L E 5

L a t t i c e P a r a m e t e r s and H y d r a t i o n of D e f e c t P y r o c h l o r e AB2X 6

A = Li, Na, A g a n d B =Nb, Sb, Ta: ABWO 6 . H2Owith 1 0 . 2 7 < a ~ 1 0 . 4 2

2. A = K: KMgA1F 6 KNiA1F 6 KCuCrF 6 KSbWO 6 KNbWO 6 KTaWO 6 9.86 9.92 monocl. 10.23 10.33 10.34

KNiCrF 6 " H20 KCoCrF 6 • H20 KTaWO 6 • H20 KNbWO 6 • H20 10.45 10.48 10.48 10.50

KNiVF 6 • H20 KTa205F • H20 10.51 10.61

3. A =Rb, T1, Cs: AllAB2X 6 w i t h 9 . 9 4 ~ a < 10.57~ o

~a ~ 0 .4 eV were obtained with graphite e lec t rodes and lkHz. This finding is compati- ble wiZh 16d s i tes par t ia l ly occupied by Na + ions. Slight di f ferences in P300 and ?-a were observed from sample to sample, but no corre la t ion of the t ranspor t proper t ies with a o was at tempted.

Table 5 l is ts known defect pyrochlores of the type AB2X 6 and thei r lattice para- me t e r s . The Rb +, T t +, and Cs + compounds, which have A + ions only in 8b si tes (though perhaps displaced f rom the center) , a r e dehydrated. The K ÷ compounds, on the other hand, a re only dehydrated if the latt ice pa rame te r is smal l enough to retain the K + ions in 8b s i tes . Compounds with la rger lattice pa rame te r become hydrated on exposure to a i r . These findings suggest there is a c r i t i ca l lattice pa rame te r for hydra- t ion that is s m a l l e r than, but near the pa ramete r at which K + ions have equal preference for 8b and 16d s i tes . Defect pyrochlores with smal l A + ions (Lit , Na +, Ag ~) a re all e i ther hydrated 21 or contain excess A20 , the A+ions occupying 16d s i tes .

Ion exchange of RbMgA1F 6, a = 9.94A, gives anhydrous KMgA1F 6 with a = 9.86~ and a noncubic "NaMgA1F6" that can be r eve r s e ion exchanged. Location of the K + ions on 8b s i tes was not independently confirmed by s ing le -c rys ta l s t ruc ture analysis , but it can be inferred. Unlike the MTa205F sys tem, these mater ia l s ho t -p ress readily. After a p re l iminary bake at 200 ° C, vacuum press ing at 60,000 psi at 500 ° C for one hour yielded disks of over 99-percent theore t ica l density. The ionic conductivity of KMgA1F 6 is ex t remely low (P300 = 20,000 Q-cm), p resumably because the 8b a r r a y is fil led. In this case , t r ans fe r of a K + ion to a neighboring 16d position costs a large K+-K+elec t ros ta t ic energy, since each 16d position has two near-neighbor 8b positions at a distance of only 2.12~. Therefore , we introduced cation vacancies at the 8b posi- tions by prepar ing samples with chemical formula K l_xMgl .xAl l+ F~, x = 0.05 and 0.10. The pyrochlore s t ruc ture was unstable for x> 0 1 Cation v~ca~cies at 8b posi- tions would permi t K + - ion t r ans fe r f rom one 8b site to another via the 16d intermedi- a ry position. A dramat ic increase in conductivity was observed. A plot of K + - ion conductivity vs r ec ip roca l t empera tu re for a K0. 9Mg0 9All. 1F6 disk over 99-percent dense gave P300 ~ 800 Q-cm and an 8 a ~ 0.35 eV.

F rom these exper iments we draw the following conclusions: (1) The B2X 6 network of the A2B2X6 X' pyrochlere s t ruc ture is a suitable skeleton for fas t ion t ranspor t . (2) Direct prepara t ion of anhydrous defec t -pyrochlore A+B2X 6 requ i res large A + cations

216 J . B . GOODENOUGH, et a l Vol. I I , No. 2

(Cs +, Rb +, and -- for lattice parameters below a c r i t ica l size - - K+ions) that a re stabil ized in the 8b (or X') positions. (3) Fast ion t ranspor t of the K + ions among the 8b sites requires the introduction of 8b-site vacancies, as in K0.9Mg 0 9All. 1F6 . (4) In- direct preparat ion of defect pyrochlores A+B2X6 having smal le r ,~ cations randomly distr ibuted on 16d sites is plagued by the introduction on 8b sites of water or of excess anions. We failed to obtain a "clean" A+B2X 6 pyrochlore having A + ions randomly dis- tributed on 16d positions. However, Na ~ ion t ranspor t in partiaUy filled 16d positions occurs in NaTa2OsF • xNa20. It would be interest ing to compare the conductivity in this compound with that in a compound like NaW206 • F, if such a defect pyrochlore can be prepared.

Although these resul ts indicate that the defect-pyrochlore s t ruc ture is capable of providing fast alkali- ion t ransport , an activation energy of 0.35 eV is st i l l a factor two higher than desired. Therefore , we turned to other skeleton s t ruc tures .

D. Carnegieite Skeletons

Because cubic NaA1SiO 4 is unstable at room temperature , we prepared the related stable compound NaA1SiO 4 • xNa20, with x = 0. 25 and 0.5, and Na1+yMgyA11.ySiO 4 with y = 5/8. A ceramic disk was made from the f i r s t of these by adding NaF as a binder and H20 to form a s lur ry . The mixture was vacuum-baked at 150 ° C and then vacuum-f i red under 15,000 psi at 950 ° C. The result ing disk had a density of the order of 75 percent and a res i s t iv i ty P300 = 1900 f~-cm. The second was "meked" in a cruci - ble to form a disk-shaped lump filled with bubbles. Our suspicion that some form of decomposition occurred is supported by the measured res is t iv i ty P300 = 600 fl-cm, since the nominal amount of sodium should have fiUed al l the available sodium positions. The third mate r ia l was pressed into a disk by f ir ing at 820 ° C in a i r under 20, 000 psi. It had a 0300 = 200f~-cm. Although 0300 is more than an o rder of magnitude higher than that of NaSbO 3 • 1/6 NaF, it never theless indicates that the carnegieite skeleton may yield solid e lect rolytes of pract ical in teres t .

E. The System Nal+xZr2P3_xSixOl2

In o rder to obtain dense ceramic disks of compositions in the system Na l+xZr 2 SixP 2 . O,~, appropriate mixtures of Na,CO 3, ZrO 2, SiO 2, and NH4H2PO 4 were heat~d~o~16 hours at 170 ° C (to decompose the NH4H2PO 4) and calcined at 900 ° C for 4 hours. The product was cold-pressed at 80, 000 psi a f ter the addition of 4-percent polyethylene glycol as a binder. The cold-pressed peUets were f i red for 6 to 16 hours at t empera tures between 1200 and 1400 ° C. The f ir ing tempera ture of Na~Zr2PSi2012 , for example, was 1250 ° C, since at 12750 C there was appreciable decoml~osition of the specimen. Densities as high as 94 percent were obtained with this method. Transpor t measurements with blocking electrodes indicated a negligible electronic contribution, and prolonged immers ion in molten sodium showed chemical stabili ty to this environ- ment. Table 6 fists the res is t iv i t ies at 300 ° C for Na + - ion t ranspor t as obtained with graphite electrodes at lkHz on ceramic disks more than 85-percent theoret ical density. Fig. 8 shows a plot of T / R vs l / T , where R is the sample res is tance , for a ceramic disk of Na3Zr2PSi2012. The activation energy obtained from this plot is 8 a =0.29 eV.

The compound NaZr2P3012 has all the M 1 sites occupied by Na+ions, none of the M 2 sites. 16 Since the M 1 and M 2 sites are not energet ical ly equivalent, the factor c(1-c) approaches zero in Eq. (1), and the res is t iv i ty of this compound is very high. Substitution of Si for P is charge compensated by the introduction of Na + ions on M 2

Vol. i i , No. 2 FAST Na +- ION T R A N S P O R T 217

T A B L E 6

Resist ivit ies at 300 ° C and Activation Energies (vT vs T -I) of Dense (> 85% Theoretical) Ceramic Specimens in the System Nal+xZr2P3.xSixO12 with Graphite Electrodes at 500 kHz. *

Composition (x) Structure P300 (fi-cm) ~:a (eV)

0.4 R 1867 0.26 0.8 R 341 0.31 1.2 R 38 0.27 1.6 R 21 0.32 1.8 M 8 0.24 2.0 M 5 0.29 2.2 M 6 0.24 2.4 R 9 0.26 2.6 R 28 0.24 2.8 R 56 0.24

* (~ (500 kHz)/~ (10 kHz)~ 2. Higher frequencies used to reduce electrode polarization,

W

U

1.1

x

l g

W

p .

/ No 3 Zr 2 P Si2012

o'300, C • O. 2,.q, -1 cm -1 / o'T • A exp ( ' ( o / k T ), ( 0 "0 . 29eV

O. I [ I I I I 3.e 3.4 3.0 2.6 2.2 ~.e ~.~

103/T [K -1]

FIG. 8 Tempera ture - conductance prod- uct T / R vs 1/T for a typical cer - amic specimen of Na3Zr3PSi2012 with graphite electrodes at 500 ld-tz.

1.2

3 2

o

FIG. 9 Structure of CuTa206 showing orien- tations of the Square CuO 4 units.

s i tes , which are linked to one another via M 1 sites. Therefore , so long as the M 1 sit~s remain occupied, Na + - ion conductivity requires a corre la ted ionic motion: Na 2 +Na~÷c] 2 -*rn 2 +Na~+Na2+,where the subscripts re fer to sites M 1 and M 2 and [] represents an empty s i te : The-fact that the lowest values of 0300 occur where the rhombohedral axis c r and cell volume are a maximum 16 suggests that Na ÷ - Na ÷ elec-

218 J.B. GOODENOUGH, et al Vol. II, No. 2

t rostat ic interactions reduce the M 1 - site preference energy with increasing x. For x > 2, both M 1 and M 2 vacancies must coexist at 300 ° C. The fact that the c rys ta l -ce l l volume reaches a maximum with increasing x suggest that the e lectrosta t ic forces be- tween Na + ions at adjacent M 1 and M 2 sites may actually displace Na + ions toward bottleneck positions between M 1 and M 2 si tes.

At 300 ° C, the t ranspor t propert ies of Na3Zr2PSi2012 are c lear ly equivalent to those of B"- alumina at t empera tures T > 300 ° C, but infer ior for T < 300 ° C. A much improved pre-exponential factor is consistent with the three-dimensional charac te r of the Na + - ion t ranspor t in Na3Zr2PSi2012. The fact that B" - alumina has an activation energy 0.1 eV smal le r than that of Na3Zr2PSi2012 leaves hope that three-dimensional Na + - ion t ranspor t in skeleton s t ruc tures may also have activation energies as low as 0.2 eV. With the improved pre-exponential factor of three-dimensional t ransport , such Na + - ion conductors would be competitive with liquid electrolytes at room tempera ture .

The sys tem Nal+xZr2P~ vSi~O12 appears to be eas ie r to fabricate into dense ceramics than 8' - alumina. We have made both open and closed-ended tubes of Na3Zr2PSi2012 at 90-percent theoret ical density with pre l iminary ext rus ion/ i sos ta t ic - press ing techniques.

The sys tem Nal+xZr2P3.xSi O12 is only one representat ive of a large class of com l~.+unds that can be stabilized w~lth this basic ~tructure. Ion exchange with Li +, Ag +, and K ions has been successful ly accomplished 1 . Other ions can replace the octa- hedral Zr 4 . ions. However, we were not able to replace Si by Ge, which is suggestive of a size constraint for the te t rahedra l ion.

We would like to thank ]. Singer for helpful discussion, K. Dwight for Figs. 5 and 6, Car l Anderson and David T racy for technical ass is tance .

Appendix

The defect perovskite CuTa206 has a (TaO3)- skeleton of corner-shared octahedra, but it is strongly distorted from the cubic ReO 3 structure. 22 Reinen and Propach 23 also deduced the Pro3 space group on the basis of optical and powder- dif~action data, but they did not determine the atomic positions. Single-crystal structure analysis 22 confirmed the existence of square-coplanar oxygen coordination about Cu 2+ ions and located them at positions 3d and 3c. Although it is remarkable that so small a cation as Cu 2+ should occupy the A site of a perovskite, this is made possible by the dis- tortion of the TaO 3 network that reduces the A-site anion coordination from twelve to

four nearest neighbors.

Fast Na +- ion transport is not found in cubic perovskite, presumably be- cause the (oxygen-square) bottlenecks at the shared Na - site faces have too small an area However, the Cu 2+ ion should be small enough to pass through such a bottleneck. More- over, the distortion of the (TaOs)- skeleton creates square-copl~n~r configurations at 3/4 of the cubic-perovskite A sites, see Fig. 11, and the Cu 2+ ions are randomly dis- tributed over these square-coplanar sites, which are 2/3-filled in CuTa206. However, the orientations of the square complexes are such that a Cu 2+ ion must pass through an edge of the square on jumping from site to site.

Powder samples are green. Single crystals grown from a copper-oxide flux are dark. Conductivity measurements on single-crystal specimens with copper electrodes gave an B a = 0.36 eV, obtained from (x T vs T -1, and a P300~1000 fl-cm. With gold or platinum electrodes the resistivity at 300°C was a factor of I0 larger. We

Vol. I I , No. 2 FAST Na +- ION TRANSPORT 219

conclude that our s ingle-crystal conductivity was mixed, but that the Cu 2+- ion conduc- tion dominated.

R e f e r e n c e s

1. Y.F. Yao and J.T. Kummer, J. Inorg. Nucl. Chem. __29, 2453 (1967); J.T. Kummer and N. Weber, Proc. 21st Ann. Power Sources Conf. (1967) and M.S. Patent 3,404, 035 (1968).

2. J.T. Kummer and N. Weber, U.S. Patent No. 3,475,223.

3. M.S. Whittingham and R.A. Huggins, NBS Special Publication 364, R.S. Roth and S.J. Schneider, Jr . , eds. (July 1972)p. 139; A.V. Virkar, G.J. Tennenhouse, and R.S. Gordon, J. Am. Ceram. Soc. 5_~7, 508 (1974).

4. R.S. Gordon, NSF Program Review, Dearborn, Mich. (Jan. 30, 1975).

5. R. Ridgeway, A. Klein, and W. O'Leary, Trans Electrochem. Soc. 7_0, 71 (1936).

6. D.B. McWhan, S.J. Allen, J r . , J.P. Remeika, and P.D. Dernier, Phys. Rev. Letters 35_._, 953 (1975).

7. P. Spiegelberg, Ark. Kemi 14A, 1 (1940).

8. J.A. Kafalas, NBS Special Publication 364, R.S. Roth and S.J. Schneider, Jr . , eds. (July 1972)p. 287; H. Y-P. Hong, J.A. Kafalas, and J.B. Goodenough, J. Solid State Chem. 9, 345 (1974).

9. J.M. Longo, P.M. Raccah, andJ.B. Goodenough, Mater. Res. Bulletin4, 191 (1969).

10. N. Schrewelius, Z. Anorg. Allgem. Chem. 238, 241 (1938).

11. D. Babel, G. Pausewang, and W. Viebahn, Z. Naturforsch. 22b, 1219 (1967).

12. C. Michel, D. Groult, and B. Raveau, Mat. Res. Bull. 8, 201 (1973); J. L. Fourquet, C. Jacobini, and R. de Pape, Mat. Res. Bull8, 393 (1973).

13. J. Singer, contract monitor.

14. W. Borchert and J. Keidel, Heidelberg Beitr. Mineral. u Petrogr. 1, 17 (1947).

15.

16.

17.

18.

19.

L. Pauling, The Nature of the Chemical Bond, 3rd Ed, Cornell Univ. Press (1960) p. 254.

H. Y-P. Hong, companion paper.

L. Hagman and P. Kierkegaard, Acta Chem. Scand. 2_~2, 1822 (1968).

R.G. Sizova, A.A. Voronkov, N.G. Shumyatskaga, V.V. Pynkhin, and N. V. Belov, Dokl, Akad. Nauk SSSR, Ser. 205, Issue 1, 90 (1972).

R.S. Roth, H.S. Parker, W.S. Brower, and D. Miner, NASA Contract C-50821-C Final Report, CR-134869 (July 1975).

220 J . B . GOODENOUGH, et al Vol. 11, No. 2

20. W.S. Brower, D.B. Minor, H.S. Parker, and R.S. Roth, Mater. Res. Bulletin 9, 1045 (1974).

21. C. Michel, D. Groult, and B. Raveau, J. Inorg. Nucl. Chem. 37, 247 (1975).

22. H. Y-P. Hong, Lincoln Laboratory, M . I . T . , Solid State Research Report (1972:2) p. 33.

23. D. Reinen and V. Propach, Inorg. Nucl. Chem. Letters 7_, 569 (1971).