Phase Equilibria and Crystal Growth in the Alkali Antimonate Systems Sb2O4–NaSbO3/ Sb2O4–KSbO3/, and Sb2O4–NaSbO3–NaF*

Phase equilibrium diagrams have been constructed from experimental data for the systems Sb2O4–NaSbO3, Sb2O4–KSbO3, and Sb2O4–NaSbO3–NaF. The system Sb2O4–NaSbO3 contains only an intermediate pyrochlore type solid solution with a maximum melting point of 1490 °C at a Na:Sb atom ratio of 3:5. The Sb2O4–KSbO3 system contains in addition to the pyrochlore phase a compound 3K2O • 5Sb2O5 which melts congruently at about 1450 °C and two polymorphs of K2O • 2Sb2O5. The low temperature form of K2O • 2Sb2O5 was found to be monoclinic P21/c with a = 7.178, b = 13.378, c = 11.985 A, β = 124°10′. The melting point of Sb2O4 was found to be 1350 ± 5 °C and NaSbO3 and KSbO3 both melt congruently at 1555 ± 5 °C and 1410 ± 5 °C respectively. The previously reported cubic form of KSbO3 was found to be a K+ deficient phase stabilized by reaction with atmospheric moisture. A similar cubic phase which appears to be a good Na+ ion conductor can be synthesized in the ternary system NaSbO3–Sb2O4–NaF.


Introduction
The search for potential candidates for ionic con~~lC~ors together with the lack of detailed phase eqUlhbnum data has served as an im petus to investigate the alkali antimonate systems.
In the system Na20-Sb20 4 -0 2 the compound N aSb03 was reported to occur by Schrewelius [1]1 and to be hexagon al with an ilmenite structure, a = 5.316, c= 15.95 A. A pyrochlore solid solution was found to occur ?y Steward and Knop [2] . No melting data was prevIO usly reported.

Specimen Preparation and Test Methods
In order to minimize the effect of foreign anion contamination in Sb20 4 , high purity antimony m etal was ground and oxidized on Pt se tters in air at elevated temperatures. It was found that the form ation of a thin antimony oxide coating at 450 °0 prevented further surface reaction of an timony with the platinum. Once this coating was form ed the temperature could be raised to 500 °0 for relatively rapid oxidation without reaction with the platinum setter. Spectrographic analysis of Sb20 4 indicated that platinum contamination was only 2 ppm. The only other metals found in quantities greater than the detectable limit were traces of Zr and Tb .
Mixtures of Sb20 4 with sodium or potassium carbonate were prepared by weighin g to th e nearest ±0.1 mg in sufficient quantities to yield a ] g batch. No corrections were made for p ercen tage purity except lo ss on ignition. Each batch was hand mixed under acetone with a mortar and pestle. The mixtures were placed on se tters fab ricated from platinum foil and calcined in air at 500 °U for 60 hs. Following this preliminary calcine th e mixtUl'es received a second calcine at 700 DC for 60 hs. In the K20-8b20 4 system the specimens received a third calcining at higher temperatures in a small platinum tube for 1 h. To minimize volatility at high er temperatures, sealed platinum tubes approximately 2 cm long were employed for all succeeding experiments unless otherwise stated. About one-third of the volume was occupied by the specimen and the remainder of the tube was flattened prior to sealing. At elevated temperatures the expansion of the flattened portion of the tube provided the necessary additional volume for expansion of the vapors without rupture. The actual pressure inside the tube is unknown. At elevated temperatures the time of the experiment was shortened to minimize "soaking in" of 8b20 4 into the platinum. By employing this procedure it was felt that the best approach to equilibrium was achieved. 8ub-solidus and melting point values were obtained by quenching specimens sealed in platinum tubes and examining them at room t emperature.
An electrically heated vertical tube furnace was used for quenching. The temperature was controlled to about ± 2 DC. Temperatures were measured with a Pt versus Pt 10 percent Rh thermocouple which was calibrated several times during the course of the work. Due to the marked volatility of the antimonates and the reactivity of the alkaline materials at elevated temperatures, thermocouple contamination sometimes resulted. To minimize this contamination problem the thermocouples were changed frequently. The overall accuracy of the reported temperature is estimated to be about ±5 DC.
The first sign of glazing of the specimen surface established the solidus values. The few liquidus values that are reported were established by the formation of a concave meniscus. No attempt was made to obtain liquidus values in the 8b20 c rich portion of these systems because of the high vapor pressure.
Equilibrium is generally considered to have been obtained when x-ray diffraction patterns of specimens successively heated for longer times and/or at higher temperatures show no change. X-ray powder diffraction patterns were made using a high angle recording Geiger co unter diffractometer and nickelfiltered copper radiation with a scan rate of 1/4 D 20/min and a chart speed of 1/4 in/min. The unit cell dimensions reported can be considered accurate to about ± 5 in the last decimal place listed .

The System Sb20 4-NaSb03
The system between the compositional limits of N a8bOa and 8b20 4 has been examined in detail. The phase equilibrium diagram, figure 1, has been constructed from the data given in table 1. When 8b20 4 is re acted at low temperature (500-1000°C) with alkali carbonate it generally loses CO2 and gains oxygen from the atmosphere to satisfy the equilibrium conditions of the phases formed, which may involve oxidation of the antimony ions. It is therefore understood that the phase diagrams determined in the antimonate systems reported here may not be strictly binary.
The compound N a8bOa (ilmenite-type) was found in this work to melt at about 1555 ± 5 DC. An intermediate pyrochlore solid solution exists from about 37.5 mol percent Na20:62.5 mol percent 8b20 4 to 24 mol percent Na20 :76 mol percent 8b20 4 at 1350 DC. The 1:3 composition probably does not really correspond structurally to [N a8b+3]8b2 H07 although the 3: 5 composition may be written as [Nau8btt]8b2+5065-see section 6.1. The 3Na20: 58b20. composition apparently melts congruently at 1490 ± 5°C. The solidus curve falls from this temperature to about 1340 ± 5°C at 24 mol percent Na20:76 mol percent 8b20 •. A two phase region exists between the pyrochlore solid solution and 8b20 4• An unknown phase was found to occur in the system which could be made approximately single phase by calcining the composition 15 mol percent Na20:85 mol percent 8b20. at 750 DC and reheating in a sealed Pt tube to 1000 DC for 64 h in the presence of Pt02. This phase has an as yet unindexed x-ray diffraction pattern with the four strongest lines occurring at d values equal to 2.283, 2.798, 3.453, 8.23 A.
In the 8b20 4 rich portion of the system from 10 percent Na20 (or Kzo):90 percent 8b20 4 to 100 percent 8b20 4 experimental interpretation at or near the liquidus is exceedingly difficult since the conventional picture of solid and liquid is not evident. At the composition 15 mol percent Na20 (or K 20): 85 mol percent 8b20 4, quenched liquid plus solid is evident. From this data the solidus can be delineated . However at or near 8b20 4 , the solid appears to transform to vapor with no indication of the liquid phase. The most likely interpretation of the data is shown in the circular insert in figure 1, indicating that solid 8b20 4 + solid pyrochlores8 melts to solid pyrochlore s8 and liquid. Within experimental error, the sublimation and eutectic points appear to be at the same temperature and the field 8b20 4 + Liq (labeled 81 + L) is not seen.

NaSb03
The compound N a8bOa was first reported by 8chrewelius [1] to be hexagonal, a = 5.316 and c= 15.95 A with an ilmenite structure. This compound was found in the present work to melt congruently at about 1555 ± 5°C. No other stable polymorphs were encountered .

Pyrochlore Solid Solution
One intermediate phase, a cubic pyrochlore solid solution was characterized in the system. The compositional range varies from approximately Na 2 0: 38b204 to 3Na20 :58b204 with unit cel.l dim ensions varying from 10.289 to 10.286 A respec tIv el~. 8ince the pyrochlore is a. tunnel st~u~ture and thiS pyrochlore is the only sodmm contam~ng pyrochlo~e reported that can be formulated by direct .synt~es~s it was worthy of further study as a pOSSible IOlllC Sb 2 0 4 MOL % Na20 NaSb0 3 F IG U RE 1. Phase equilibrium dia gram f or the system Sb20 c NaS b0 3.
• -melting X-no melt ing conductor. :For ionic condu ctivity measurements dense m a teri als were n eeded and sever al experiments were condu cted with N a20:2Sb20 4 in an effort to det ermine the stability of the pyrochlore solid solution under high pressure and temper ature. Samples in sealed pl atimun tubes were heated at 1100 °C and 4000-5000 psi 2 for sever al hours. The resultin g specimens are single phase pyrochlore which appear to be very dense. The aver age density of four measured fragments was 5.26 ± 0.05 g/cm 3 .
. 2 The usc of psi, bar, and kbar follows tho current comnl011 practice of w orkers In the field. Note that I bar = IOO N/m' (or pascal) .. IO' dyn/C1ll2 " O.9869 atm '" 14.504 pSI. The accepted Intern at IOnal standard (S1) uni t of pressure is t he pascal or newton per meter sq uared.
For ionic conductivity m easurements, pellets of N a20:2Sb204 (1.9 cm in di ameter) were placed in sealed platinum foil envelopes and hot pressed by a commercial company at 1100°C and 5,000 psi. The pellets were single phase m ateri al with a density of 96 per cen t theoretical (see sec. 6.1). The ionic condu ctivity of these pellets was measured at NASA L ewis R esear ch Cen ter [81 and they were found to be essentially insul ators.
The distribution of the various ions (i. e., N a+, Sb+3, 8b+ 5 , 0 -2 ) in the Na20:2Sb204 specimen is currently being determined at NBS from single crystal structure analysis. Until the results of this an alysis are forthcoming it may be assumed that the "lone pair" electrons associated with Sb+ 3 will not allow the passage of Na+ through the channels.   Two stable polymorphs of Sb20 4 have been reported in the literature. They are a-Sb20 4 , which is orthorhombic [6], a= 5.436, b= 11.76 , c= 4.810 A and {3-Sb20 4, which is monoclinic [7], a= 11.905, b= 4.834, c=5.383 A and {3 = 101022' . From table 2a it can readily be seen that specimens quenched from a temperature-composition region represented on the phase diagram , figure 1, as Sb20 4 + pyrochlore may contain either a-Sb20 4 and/or {3-Sb20 4 when quenched from high temperatures and ambient pressures and examined at room temperature. From this seemingly inconsistent data it would appear that a-Sb20 4 and {3-Sb20 4 have a polytypic relationship.

Composition
Heat Treatment To help resolve this problem a high resolution electron microscope study should be done . From the data in table 2b it appears that the {3 form is the equilibrium high pressure form of Sb20 4 • Insufficient dat a have been collect ed to establish if an equilibrium boundary curve exists between a-Sb20 4 and {3-Sb20 4 at various t emperatures and pressures. When specimens are sealed and heated under pressure in the presence of PtOz in either Pt or Au tubes single phase {3-SbZ0 4 is obtained . However when heated under pressure without the PtOz, a two phase specimen results, {3-Sb20 4 and the dense high pressure form of Sb20 3 (valentinite). A similar polytypic relationship probably exists for the two polymorphs of SbzOa. " " 900 a + i3 950 " " ~I The phases identified are given in the order of the amount present (greatest amount first) at room temperature. These phases are not necessarily those present at the temperature to which the specimen was heated. a refers to a-Sb 2 0 4 polymorph and i3 to the i3-Sb204 polymorph. bl -Material placed on platinum slide and heated and examined by x-ray diffraction at various temperatures .

Starting
.£1 Poorly crYstalline as received Sb 2 0 4 was heated 750°C -24 hours and the same specimen which was never ground was reheated at 800°C -24 hours, then 900°C -64 hours and finally 950°C -24 hours.

The System Sb20 . 1-KSb03
This system has b een examined between the compositional limits of K8b03 and 8b20 4 • The results are given in the data presented in table 3 from which the phase relationships have been established as shown in figure 2.

. Compounds in the System
The compound K8b03 with an ilmenite structure a = 5.361, c= 18.21 3, was previously reported [3] and was found in this work to melt congruently at 1420 ± 5°C. A body centered cubic solid solution phase originall y reported as K8b03 [3] has been found to occur metastably at about 47.5 percent K 20. The 3K 20 :58b20 5 compound was found to melt congruently at about 1450 °C . The K 20 :2Sb20[ compound was found to have a phase transition at about 1000°C and to dissociate to pyrochlore plus 31(20:58b205 at about ll50 °C. The low temperature form of K 20 :28b20 5, labeled P2[ /c, represents a mono clinic phase with a= 7.178, b= 13.378, c= 11.985 A and ,6 = 124°10' . 8ingle crystals of th is phase were grown by flux evaporation from the composi tion 501(20 :5Sb20 4 :45Mo03• The unit cell and space group were determined from th ese crystals and confirmed by least square indexing of th e powder diffraction pattern of the low temperature form of the compound K 20 :28b20 5 • The pyrochlore solid Heat Treatment Temp Time --.
--------solution exists at 1150°C from about 15 mol percent K 20:85 mol percent Sb20 4 to greater than 30 mol percent 1(20 :70 mol percent 8b20 4. The melting characteristics of these phases have been partially determined as shown in table 3 and figure 2.

Hydroxyl Ion Stabilization of Cubic Potassium Antimonate
The compound KSb03 was reported previously as being cubic at ambient conditions after treatment at high temperatures and pressures [9] .
In the current work, occasional small amounts of a cubic phase were seen in the x-ray powder diffraction pattern of 1(8b03 ilmenite heated at ambient pressure. For these reasons, specimens of 1: 1 and 3 : 5 mol ratios 1(20 :Sb20 4 were equilibrated in air at 750°C for 60 h to oxidize and form the phases KSb03 and K 3S b50[4 and then reheated for 1 h at 1200°C to drive off all excess moisture. X -ray diffraction patterns of th ese specimens showed singl e phase ilmentite and th e 3K 20 :5Sb20 5 compound. Portions of these 1200°C calcines were then weighed and mixed in acetone in the appropriate ratios to yield composition s of 46, 47, 47.5, 48 and 49 mol percent K 20. E ach of these specim ens was dried at 240°C fo), 1 h and heated in open Pt tubes at 1200°C for 1 h . Only th e x-ray pattern of the 46 percent specimen showed a small amoLint of 3K20 :5Sb20 5, the others contained only th e cubic phase. A n ew specimen of ResultsE al .
-a-Sb 2 0 4 prepared by the oxida tion of Sb at 530·C on Pt tray. This material was reheated a t 800·C -60 hr bl -The phases identified are given in the order of the amount present (greatest amount first) at room temperature. These phases are not necessarily those present at the temperatures to which the specimen was heated.  These phases are not necessarily those present at the temperature to which the specimen was heated. 1:2 -K 2 0.2Sb 2 0 s ; 3:5 -3K 2 0'5Sb 2 0 s and 1:1 -KSb0 3 -ilmenite structure.
~/ Non-equilibrium mixture -see Discussion in text.
~ The phase was indexed from single crystal x-ray precession data which has shown the compound is monoclinic  Not necessarily a true binary system Q -melting X -no melting 58-solid solution 1 :2-K20:2Sb20; 3 :5-3K20 :5Sb20 S P2Ifc-Iower temperature form of K 20:2Sb20; of 48 mol percent K 20 was prepared in the same way except the Pt tube was sealed. After 1 h at 1200°C, the x-ray pattern of the specimen showed only about 50 percent cubic. A new specimen of 48 percent K20 was prepared by weighing the 1: 1 and 3: 5 phases immediately after removal from the 1200°C furnace and sealing the material in a flattened Pt tube within 1-2 min. This tube was then inflated at 1200°C for a few minutes and the material mixed by shaking in a "wiggle-b'Jg." The sealed specimen was then heated for 64 h at 1200 °C. The resultant specimen had exceedin gly large grain growth indicating considerable solid state recrystallization but showed no cubic phase. The conclusion is inescapable that access to atmospheric moisture is probably necessary for the formation of the cubic phase at atmospheric pressure.
. A paper entitled "Flux Synthesis of Cubic Antimonates" was published by the present authors during the course of this work [10]. In addition to the discovery that the F-ion stabilized the formation of the body centered cubic phase of potassium antimonate it was disclosed that the cubic antimonate could also be obtained by reacting KSb03 with a small amount of other cations with small radii like B+3, SiH , etc. It now appears obvious that in this reaction the boron or silicon (etc.) actually ties up some of the K+ ion in a second phase and allows tl~e K + deficient antimonate to react with atmosphenc moisture to form the cubic antimonate previously thought to be "KSb03·"

. The System NaSb03-NaF
To determine if N aF additions will stabilize the body-centered cubic phase, similar to the 6KSbOa: KF-phase [10], NaF was added to NaSbOa in the ratio of 3NaSbOa:NaF, 4NaSbOa:NaF, 5NaSbOa: NaF and 6NaSbOa:NaF. After heating at 750°C and 1000 °C in sealed Pt tubes, the x-ray patterns showed only ilmenite and NaF, however after heating at "'1150°C all the compositions contained some body centered cubic-type phase. The compositions 3N aSbOa:N aF and 4N aSb03:N aF, when heated in sealed Pt tubes at ",1250°C, did not contain ilmenite and appeared to be the closest to single phase cu bic. The small crystals of 4N aSb03 :N aF prepared by quen ching in a small sealed tube appeared to be well-formed truncated octahedrons. However, the room temperature x-ray diffraction pattern of the material had somewhat diffuse lines, with the exception of the hOO lines which were reasonably sharp, suggesting rhombohedral symme try. This material was placed on a hot stage microscope slide and analyzed by x-ray diffraction from room temperature up to 220°C. At 190°C the material appeared to start to go cubic and by 220 °C a good quality cubic x-ray diffraction pattern was obtained (a = 9. 353 A). When the material was cooled to room temperature the symmetry was again noncubic. As the hOO lines deteriorate somewhat on cooling, the true symmetry of the room temperature form is probably no higher than monoclinic or triclinic rather than rhombohedral. It was therefore not unreasonable to expect that a bod y centered cubic phase could be obtained by direct synthesis with N aF without the necessity for N a+ ion exch ange.

The Ternary System NaSb03 :Sb z0 4 :NaF
X-ray diffraction patterns (single crystal and powder) of selected N aF -flux syn thesized [11] washed crystal s show only a truly cubic body centered phase (a = 9.334 A). It must be postulated that the composition formed by this technique is slightly different from that made essentially single phase at 4N aSb03 :N aF in a sealed tube. In an attempt to obtain a fluorine-substituted body centered cubic phase which exists at room temperature the compositions shown in table 4 were prepared and show the reported phases wh en quenched from 1250°C. Equilibrium was not obtained in overnigh t heat treatments at 1200 °C. At 1350 °C th e body centered cubic phase started to decompose. Th e composition 68N aSbOa :4Sb20 4 :28N aF (mol %) was chosen as the best composition for furth er studies on cer amic procedures [11] . The phases found in the specimens heated at "'1250°C are summarized in " equilibrium" diagrams for the quaternary system N aSb03-Sb203-Sb20 s-N aF (fi g. 3) and the ternary plane of this sys tem N aSbOa-Sb20 4-N aF ( fig. 4).   T AB L E 4. Experimental data f or the ternary system NaSbOa-Sb2Oc NaF

Relation of Structural Mechanisms of Non-Stoichiometry to Ionic Conductivity
It is probably generally accepted that a phase which exhibits unusual ionic conductivity must necessarily be structurally non-stoichiometric. Unfortun ately the opposite is not necessarily true. Nevertheless a crystall ographic understanding of non-stoichiometric phases is an obvious necessity to the tailoring of new alkali ion conductors. For this reason it is worthwhile to discuss the nature of the non-stoichiometry which has been observed in this study for those phases which seem to be of interest.

.1. Pyrochlore Pha ses
In the KTaOa-W03 system a pyrochlore phase occurs at abou t the 1 :1 ratio or K1. o[T a W]06 [11 , 12] . Unfortunately, the pyrochlore in this sy::; tem transforms to a tetragonal tungsten bronze (TTB) at high termperatures. Although it can be ion exchanged with N a+ to produce an ion conducting pyrochlore phase, this phase is not stable above about 450 °0 [11]. The only stable Na+ containing pyrochlore is the one in the Sb20.-N aSb04 system and apparently this one is not a good ionic conductor.
The distribution of Na+, Sb+ 3 , Sb+ s and 0 -2 ions in a pyrochlore single crystal is currently under evaluation by the Orystallography Section at NBS. However, certain assumptions can be made which may enable us to postulate the approximate distribution. The formula for th e composition s observed to result in a pyrochlore structure night be postulated to be [N aSb+ 3 JSb20 7 for the N a/Sb ratio of 1: 3. [N a133Sbo':~7JSb206 67 for 1 :2, and [N auSbo 5]Sb20 65 for 3 :5. However, these compositions do not illustrate the structural nature of pyrochlore nor account for the observation that the "Jone pair" electrons associated with Sb+ 3 will not allow 0 -2 ions to completely coordinate the antimony and result in apparent vacancies.
The structural formula of pyrochlore should be written as [A2X][B2X6] to emphasize the fact that the octahedral network of B2X6 is required to be complete if the structure is to be stable. The A2X ions fill th e intersec ting channels in this B2X6 framework. In our material the B2X 6 framework must be represented as [Sb2+ s 0 6J-2 and m'Llst be stoichiometric. All remaining Na+ and 0 -2 ions, as well as Sb+3, must be in the [A2X]+2 portion of the formula. All Sb H mnst be in B2X 6 and only Sb+ 3 in A2X. Furthermore th e maxim"-Lm number of the sum of N a+ 1 , Sb+ 3 , excess 0 -2 (beyond 0 6 -2 ) and "lone pair" electrons cannot exceed three. One can then write the general formula as [A20]+2[Sb20 6 ]-2 with [A20 ]+2 equal to [N ai/1+N ax +1+SbkX +3+0 v -2+ L.P.kX]:::;3 where k equals the ratio Sb/N a. Using the ionic valences and the sum of the ions equal to three, maximum densities can be calculated and compared with the observed to t est the structural hypothesis. The maximum density for the N a/Sb ratio of 1:3 repr esented by the formul a is calculated to be 5.469 g/cm 3 . For the N a/Sb ratio of 3:5 with the formula [N a+\ sSb+3 o . 50-20 sOo. s]+2[Sb20 6]-2 the density is calculated as 5.406 g/cm 3 . For the intermediate composition with the N a/Sb ratio of 1: 2 and a formula of [N an94 Sbd.~88 00:;2900588]+2 [Sb20 6]-2 the maxim'lLm density is found to be 5.481 g/cm 3 . The density found for our isostatically h ot pressed specimens is 96.0 percent of the maximum theoretical density. It should be remembered however that the true theoretical density of any given Sb/N a ratio will decrease with decrease in temperature. Thus the densities obtained on our hot pressed specimens are, in all probability, greater than 96 percent of theoretical in view of the expected increased oxidation of the Sb at the relatively low temperatures involved.

.2. Body Centered Cubic Antimonates
A successful method of synthesizing cubic potassium antimonate by heating in molten KF was published by the present authors [10]. The major reason for the success in obtaining completely single phase fluorine stabilized cubic potassium antimonate is that the KSb03 ilmenite form is H 20 soluble and may be easily separated from the cubic material.
An examination of the structural model of the octahedral framework of the body centered cubic antimonate phase suggests that this structure must always have some anion (X) occupancy in the 000 and 1/2 1/2 1/2 positions. Th e structural formula thus appears to be [AI6X2]+12[SbI2036]-12 with the alkali ion in position (A) located at (or just off) the juncture of the open cages. However, it seems very likely from both structural reasons (bond lengths, etc.) and valency considerations that either or both of the nonframework positions will be nonstoichiometric. Valency considerations require that at least two out of 16 alkali ions must be missing and the structural formula then becomes This formula corresponds to the composition reported by Good eno ugh, et al. [ It seems quite likely, however, that this general formula does not completely account for all of tl:e preparations which have been observed to form thIS structure, whether body centered or primitive. The observation that a primitive phase can be formed, in air, by reaction with atmospheric moisture at a 48 :52 ratio suggests that this phase may well have considerably less than 14 alkali ions per unit cell. The formula must be compensated, in this case, by a substitution of a monovalent anion [(OH)-, F-] in the octahedral framework. The general formula then becomes [0 2+xAI4-xXd+ (l2-x) [8bI20 36 _ xXxt (12-x) . The composition found at ",48 :52 in the potassium antimonate system can be written (assuming a ratio of 11 :12 K/8b or 47.826% K20) : or [05KII (OH)2]+9[8bI20 33 (OH)3]-9 which also can be described as 6K8b03 :38b20 5 :5KOH The general formula describing the K + containing compositions is then [02+xKI4-XX' 2] + (12-X) [8b 12 +50 36 _ X X' x]-(l2-x) .
The above formula contains only pentavalent antimony and apparently does not completely explain the compositions which form a " stable" body centered cubic phase in the system Na8b03: 8b20 Hx :N aF. The only formula which does not involve the loss or gain of 0-2 (or F-) when the 8b20 4 is added in a sealed tube corresponds to: [02N al,F2] [8bv +38b;t;s.. v036-2vF2v] which is represented by the join 6:1-3 :7 on figures 3 and 4. There is really no place in the framework structure for 8b+3 and it is difficult to believe that octahedrally coordinated antimony can be 8b+3. However, for convenience, the formulas can be written involving 8b+3. The new formula would then have two variables: represented by the plane in the quaternary system N a8b03 : 8b20 3 : 8b20 5 : N aF bounded by the 6: 1-3 : 4 and 6:1-3:7 joins of figures 3 and 4. However the single phase region in this system actually appears to contain more N aF than described by this general formula. Apparently some O2 is evolved in the sealed Pt tubes, the amount depending on uncontrolled variables such as the amount of free volume in the tube and on changes from the original composition durino-treatment. The absolute maximum amount of N aF which can be accommodated structurally by the body centered cubic phase can be described by the formula which represents a line in the system shown by the join 3:1-3:8 in figure 4 and involves the evolution of one molecule of gas (02) per formula unit. The results of our investigations so far suggest that the body centered phase approaches this formula as a limit. The composition of the cubic phase in equilibrium with excess 8b20 4 and molten N aF actually appears to touch this line at approximately 10Na8b03:8b20 4:6NaF or The single phase distorted cubic material on the binary join Na8b03:NaF appears to have a composition between 6: 1 and 5: 1 or approximately 11Na8b03:2NaF or The compositions in the quaternary system thus probably lie on a join between these two end members.