THE BIKARY SYSTEM Li20-SiOz BY F. C. KRACEK
Introduction By virtue of the unique position held by lithium in the periodic table, the properties of its compounds are sharply differentiated from those of the other alkalies, and yet are not wholly analogous to those of the alkaline earths, with which they are often compared. This transition character of lithium is exhibited by its compounds with silica in a striking manner, and hence the study of the system described in this paper is of particular interest in the broader problem of the chemistry of silicate formation and stability. The intermediate position of the lithium silicates between those of the alkalies and the alkaline earths is evidenced in the nature of the compounds formed and in their properties, such as the melting points, optical constants, crystallizability, solubility in water. While the alkalies, because of their high basicity, tend to form stable compounds of relatively high silica content, such as the disilicates and the tetrasilicates (K, Rb, Cs), the alkaline earths, with the exception of Ba, do not form disilicates; on the other hand, the orthosilicates of the alkalies are relatively unstable, while those of the alkaline earths are very stable. Lithium forms a highly stable metasilicate, with a higher melting point than that of any other alkali metasilicate. Lithium orthosilicate is also stable at a higher temperature than the other alkali orthosilicates, but the disilicate is only stable over a narrow region of composition in contact with the liquidus; at much higher pressures its field must disappear altogether. The liquidus curve of cristobalite is also characteristic in this respect, as shown recently by Kracek;' this curve exhibits a regular periodic trend from Mg to Cs. I n the magnesium, calcium, and strontium systems, liquid immiscibility is encountered in the cristobalite region. Greig,2 who made a very thorough study of this question, also showed that in the barium system the immiscibility region disappears, but that in this case the cristobalite liquidus takes the form of a pronounced reverse S curve. According to Kracek this reverse S curvature persists in the case of lithium and sodium, and only very slightly in the case of potassium, disappearing completely in the rubidium and cesium systems. I n the lithium system the curvature is very pronounced, and the departure from ideality is greater than with the other alkalies. I.
Because lithium silicate mixtures devitrify with much greater ease 2. than other silicate glasses, it is possible to obtain the approximate location of the melting point curves by the method of thermal analysis. This method 1 2
F. C. Kracek: J. Am. Chem. Soc., 52, 1436 (1930). J. K. Greig: Am. J. Sci., 13, I (1927).
2642
F. C. KRACEK
waB used by Rieke and Endell,' Ball6 and Dittler,2 Schwarz and Sturm,3 Wallace4 and van K l o o ~ t e r . ~I n spite of the fact that the melting relations in this system are the simplest to be found in silicate work, the results of individual workers did not agree with those of the others, indicating that the method as used is not adequate for refined results. Possibly the greatest cause of uncertainty in these results is the prevalent use of cooling curves with a high cooling rate. Thus, undercooling with Li,SiO, is reported to be anything between 1 5 and 100'. The results of heating curves, on the other hand, are more generally reliable, provided the heating rate is controlled, and not too high. The only reliable informat'ion on the system to date is given by Jaeger and van KloosterJ6who employed heating curves together with the method of quenching developed at this Laboratory, controlling their results by microscopic examination. Experimental The mixtures studied were made up from highly purified Li2C03, and selected quartz. The quartz gave a residue of 4.5 mg on evaporating a I O g sample with hydrofluoric and sulfuric acids. The Li2C03contained only traces of CaO and ;?;a20,and only enough A1 to give a faint preciphate with KH40H, when dissolved as chloride. The ingredients were mixed in the desired proportions by weight; the mixtures were then gently heated without fusion, until most of the COS was driven off. After this they were brought t'o a complete fusion in order to allow for equalization of the composition. The entire procedure took place in electric furnaces, to avoid contamination. Mixtures containing more Li,O than that corresponding to the ratio ILi20:ISi02tend to lose weight by volatilization of this component, and at t'he same time attack the Pt crucible with the formation of an olive gray deposit, presumably through the agency of absorption of O 2 from the atmosphere; hence it is necessary to avoid excessive overheating of such mixtures. 3. Unlike other alkali silicate mixtures, the compositions in the neighborhood of Li4Si04are not hydroscopic and do not rapidly absorb CO1 from the atmosphere, resembling alkaline earth silicates in this respect ; such mixtures crystallize so rapidly that the glass can not be retained on cooling. Compositions in the neighborhood of Li2SiOacan be quenched to a glass in small samples, but in general, the glasses devitrify very readily; only mixtures containing more than 7 0 per cent Si02 could be obtained in the form of fairly stable glasses. Because of this tendency toward devitrification, it was necessary to modify somewhat the usual procedure employed in determination of the liquidus. 'Rieke and Endell: Sprechsaal, 43, 683 (1910); 44, 97 (1911). ZBall6 and Dittler: Z. anorg. Chem., 76, 42 (1912). 3 Schware and Sturm: Ber., 47, 1737 (1914). 4R.Wallace: Z. anorg. Chem., 63, I j (1909). 5 H. S. van Klooster: Z. anorg. Chem., 69, 136 (1911). 6 Jaeger and van Klooster: Proc. Amst. h a d . Sei., 16, 857 (1914).
LITHIUM OXIDE AND SILICA
2643
The quenching method' commonly employed in such work with silicates ceases to be reliable with compositions richer in Li20 than the metasilicate and it was necessary to resort to the method of thermal analysis. The apparatus used for this purpose has already been described.2 To avoid loss of Li,O by volatilization the samples were sealed up in Pt capsules holding about I g of material, the thermocouple wires being fused directly to the capsule, the Pt-Rh wir.e at the bottom, and the Pt wire at the top, according to the method of Jaeger and van Klooster (op. cit.). In place of the small Pt bottles nhich they used, the capsules in my work were made from pure sheet Pt, by bending the sheet ( 2 X z em) to form a small cylinder, welding at the joint, and then folding the Pt at one end to form a small crucible. The Pt-Rh wire is inserted in the folds and the whole is welded up. The Pt wire is welded onto the side of this crucible. After the sample is introduced the top is pinched together to give a flat fold which can be easily sealed in a blowtorch flame using oxygen and illuminating gas. Such capsules can readily be made in 15 minutes, and serve the purpose quite as well as the expensive Pt bottles referred to above. Differential heating curves were taken under controlled conditions, using storage battery current and taking all necessary precautions to eliminate both thermal disturbances and electrical leakage. The breaks were found to be reproducible, provided the heating rate was not excessively high (the usual rate of heating was 3 to 4' per minute). It seems unnecessary again to caution against the use of cooling curves in thermal analysis, unless accompanied by heating curves; even in this system, in which crystallizability is comparatively high, cooling curves invariably yielded values I O to 50" lower than the heating curves; the results of the latter could, however, be checked against the quenching method to within o . ~ "which , is as close an agreement as can be hoped for in this kind of work. 4. The thermocouples used were Pt vs. 90 P t I O Rh, calibrated in accordance with accepted standards. The readings were converted to ten.>eratures on the gas thermometer scale3used in this laboratory.
Results All preparations containing up t o 67 wt. per cent SiOL were studied by the heating curve method, with the samples sealed up in capsules as described above; the remaining preparations were studied by the quenching method. Preparation 602, which is v e r y nearly pure LiLSiOJ,was studied by both methods, for a comparison of their applicability; the result was satisfactory, as the m.p. by quenching was 1 2 0 1 I' and by the capsule method 1zo1.5". A cooling curve (with the capsule) gave 1189" for the freezing point, with a cooling rate of about I' per minute. j.
Shepherd and Rankin: Am. J. Sci., 28, 293 (1909). F. C. Kracek: J. Phys. Chem., 34, 1583 (1930). 3 Day, Sosman and Allen: "High Temperature Gas Thermometry," Carnegie Institution of Washington Publ. No. 157; L. H. Adams: Int. Crit. Tables, 1, 5 ; .
2644
F. C. KRACEK
1 The compositions and significant temperatures for preparations studied (Lis0 = 29.88, SiOz = 60.06) T.4BLE
No. of
Per cent Si02
preparation weight
624 625
46.0
626
53.70
36.58
616
55.32 57.90
38.12 40.63
627
t"C
H' H
1226 i I
29.8
52.07 3 j . 0 8
61 j
60.jo
43.25
614 602
65.37 66.78
48.43 50.00
604 603 605 606 607 608
69.13
601
80 08
609 610
80,93 67.86 82.29 69.80
611
84.98
j2.70
j3.72 74.88 5 9 . 7 3 7 6 , 6 4 62.01 77.93 63.72 79.00 65.72 70.00
66.65
73.79
612
87.44
77.60
613
89.97
81.69
62 I 622 623
Method
mol
92.48 94.96 97.01
85.95 90.36 94.17
I255
*
I
H H H H H
1024 li: I
H
1024 i I
H
114j
j, I
H
1024
I
H H
Q Q Q Q Q Q Q Q Q Q Q G!
Q Q Q Q
Q Q Q Q Q Q
1143
*
2
1024 f I 1028 i I
1095
=
*
1
1199 f I 1201.5
1201
*I
Crystal Phases
R? L E L E L L E L E L
LizO Ili4Si04 Li Si01 Li4Si04 LizSiOs Li S i 0 Li4SiOl Li2Si03 Li2Si03 IdzSi03 Li4Si04 Li2Si03 LiQSiO3 Li4Si04 Li2Si03 LizSi03 LizSi03 LizSi03 I,i2Si03 LizSiOS Li2Si03 Li2Si03 Li2Si03 Li2Si03 LizSi205 LizSiOs Liz Si 03 Li2Si03 Li2Si205 I,izSi205 Tridymite Li2Si206+tridymite Tridymite Li2Si205+tridymite Tridymite LizSi205+ tridymite Tridymite LizSi20s+tridymite Cristobalite Cristobalite Cristobalite
hl bl
L L I125 1 L 1092 i 2 L 1070 * 2 L I052 * I L 1033 zt 0 . 5 R 1195 f 1 1188 2
* *
L
I034 i O . 5 I033 0.5 1032 & 0 . 5 1037 i z 1028 k o . 5
R L L E
1217 i 2
L
i0.5 I340 5 1028 k 0 . 5 I434 5
E L E
1028 h 0 . 5
E
I535 1581 I t 3 1622 i 5
L
=
1028
*
* =5
H = heating curve method. Q = quenching method. 2
Type
R = reaction temperature, incongruent melting. L = liquidus temperature. M = melting point. E = eutectic temperature.
L L L
+
+ +
+ +
+
+
LITHIUM OXIDE ASD SILICA
2645
The method of thermal analysis was employed also in searching for heat effects below the liquidus, particularly in the disilicate region. These will be described in the appropriate place (section I O ) . The results of the experiments are collected in Table I, giving the compositions in weight and mol per cent, the temperatures. and the solid phase
170(
t'C
/ 1M)I
Cristobolite -L.
1%
l a
IWC
IZW
ll0C
5
1 mix
IOOC
IOZa'
cr
5t
0 + Tridyrnite
LI
I
mir crqstals 50
60
70
80
9 C u t percent Si&
FIG.I The Equilibrium Diagram for the System Li20-SiOl. Black circles represent the author's results.
F. C. KRACEK
2646
stable at the liquidus. Collateral data, on optical properties of the crystals, the refractive indices of the glasses, and the invariant points are given in Tables 11, 111, and 11'. TABLE I1 Invariant Points in the System Li20-SiOz wt. per cent Si02
+ Li4Si04 + liquid Li4Si04+ Li2SiO3, eutectic
t"C
(I) LizO
50.9 f0.2
1255
(2)
55.3 f 66.78
1024
f f
I201
f I
1033
f 0.5 f 0.j
(3) LizSi03,melting point (4) LizSi03 LizSi20: liquid (5) LizSiz06 tridymite, eutectic (6) Tridymite cristobalite, inversion (;) Cristobalite, melting point
+ + +
+
80.1
0 . I
f 0.05
I I
82.2 f O . 1
1028
91.o f 0 .I
1470 i
IO
1713
5
100.0
f
FIG.2 Typical heating curves on lithium silicate preparations in the orthosilicate region. The numbers refer to compositions in Table I. Note the absence of the eutectic break with preparation 624.
The equilibrium diagram is shown in Fig. I, typical heating curves in Fig. 2 , and a graph of the refractive indices of the glasses is given in Fig. 3 . 6. It will be noticed from the diagram and from Table I that the study was confined to that portion of the system from 46 weight per cent Si02 to IOO per cent SiO,. No attempt was made t o study the liquidus of Lizo. The reason for this will be obvious when it is remembered that Li,O is rapidly lost from mixtures high in this component, when heated in air, resulting in considerable uncertainty as to composition. I n order to study the liquidus curve of LinO (and other alkali oxides as well) the experimental conditions must be modified to eliminate the disturbing effect of O2(and possiblyN2 also).
2647
LITHIUM OXIDE AND SILICA
The results for preparations between 46 and 5 5 weight per cent Si02 established beyond doubt that the orthosilicate, Li4Si04,decomposes before its melting point is reached, the phase reaction occurring at Iz5j°C. Jaeger and van Klooster (op. cit.) determined the melting points in this region in open crucibles (using the closed capsules only in attempting to determine
TABLE I11 Approximate Refractive Indices of Lithium Silicate Glasses No. of Preparation
wt. per cent
RiOI
616 602 605 606 607 608 60 I 609 610 611 613 -
55.32
66.78 74.88 76.64 77.93 79 .oo
Refractive Index of Glass
1,567 * o o . o o 3 1,557 1.546 1.543 1,540 I . 538
80.08
1.535
80,93 82.29 84.98 89.97
1
1 0 0 .oo
'
533
1.529 I . 522 1 ,503 I ,459
TABLE IS' Optical properties of lithium silicates (examined by H. E. Merwin). L i t h i u m disilicate LiZSi205 appears to be orthorhombic with three cleavages a t right angles. One cleavage is micaceous, the other two are nearly perfect. The plane of the opt'ic axes is parallel to the micaceous cleavage, and y parallel to the intersection of the two best cleavages. Optical character is positive, 2V = j o o to 60'. Refractive indices: a = 1.547, j3 = 1.550, y = 1.558, all probably within + 0.001. Measurements on a preparation containing the disilicate and tridymite showed a = 1.545 + 0.002, y = 1.558 0 . 0 0 2 , +2V = ca. 56'. There is thus no solid solution indicated optically. L i t h i u m metasilicate LizSiOBwas studied both as cleavage splinters and as long prisms, with parallel extinction and negative elongation. Measurements of the refractive indices seem to indicate that the crystals are uniaxial with w = 1.591 and e = 1.611, both probably within =t0.001. When present with large excess of the disilicate, no solid solution was indicated optically ( w and e were observed to be 1.591 and 1.610 respectively). L i t h i u m orthosilicate could be studied only in the form of rounded grains, commonly with two lamellar twinnings oblique to extinction directions. a = 1.602, y = 1.610, both 0.002. The optical character and /3 could not be definitely determined.
*
*
2648
F. C. KRACEK
the melting point of Li,O), and although it does not seem reasonable to suppose that there was enough volatilization of Li20 from their preparations to completely account for the discrepancy between their results and mine, it must be concluded that their analyzed compositions do not represent the compositions present in the samples at their observed liquidus. They emphasize the fact that their mixtures altered on heating. I n my experiments with the sealed capsules the temperatures were definitely reproducible; the eutectic Li,Si04 Li-SiO:, liquid was located at 5 5 . 3 o I wt.per cent
+
+
*
FIG.3 Graph of the approximate refractive indices of lithium silicate glasses at room temperature. Determined by the immersion method.
+
+
S O 2 , 1024O, and the invariant reaction point Li,SiO, Li.0 liquidat + 0.2 wt. per cent S O a , I z j j ' . The liquidus for Li.0 rises steeply from this point. The melting point of Li,O according to Jaeger and van Klooster lies above 1625'. 7 . The liquidus curve of LilSiOB has a moderately rounded maximum, indicative of dissociation of the compound in the liquid state. The melting point of LiASiOsis one of the best known and useful secondary st,andard points for calibration of thermocouples. If we take the melting point of Au as 1063' (by definition, on the International Temperature Scale), the melting point of Li,Si03 is 1 2 0 1 ~f by quenching, and 1 ~ 0 1 . 5 " * 0 . j" by the heating curve method. The points on the liquidus, for compositions richer in Li,O than the compound, were determined by heating curves, and those for compositions richer in SiOl than the compound by quenching. 8. On the SiOI side, the liquidus of Li-SiOs descends smoothly, and encounters the liquidus curve of the disilicate, LiLSi-Os,almost exactly at t,he composition of the disilicate, certainly within 0 .I per cent SiO,. The quenching results obtained with preparation 601, which is 80.08 i 0 . o j wt. per cent SiO, (theoretical for Tli2SiPOj= 8 0 . 0 8 mt. per cent SiOe) indicate that 50.9
IO
LITHICM OXIDE .4ND SILICA
2649
the disilicate disappears at 1033', leaving very rare LisSiOJ and liquid. The Li2SiOadisappears from this preparation just below 1035'. The order of uncertainty in these determinations is very near the limit of experimental precision; the most probable interpretation is that Li,Si?05just misses having a stable melting point. The results place the invariant reaction point a t 80. I O i o.ojwt. per cent SiO,, 1033'. 9. The disilicate liquidus has only a brief range of stable existence, descending t o the eutectic Li-Si205 tridymite liquid, which is located at 82,z 0 . I wt. per cent S O ? , 1028'. I O . Microscope studies of the various preparations investigated by Jaeger and van Klooster led them to suppose that a certain amount of mixed crystal formation takes place when Li metasilicate crystallizes from melts containing excess SiO,. Dr. Merwin, who kindly examined some of my preparations, found, however, that no solid solution was indicated optically in either Li!SiOa or LizSi.Oj. On the other hand, the results obtained with the disilicates K,Si?Ojl and Sa2Si20S2indicate that these compounds take up small excess of either Si02 or of the basic oxide, depending upon the composition of the solution from which they separate, and that t)he solid solutions thus formed dissociate completely on cooling, at 814' and 992' in the case of K2Si-Ojrand at 706" and 768" in the case of NasSi20s. No optical evidence of these solutions can be foundat room temperature, both of these compounds passing thru enantiotropic inversions (at j 9 0 " and 678" respectively); the temperatures of the inversions are independent of the gross composition of the mixture, indicating the unnlixing is complete, the phases just above and just below the inversions being the pure compounds. To test whether an analogous situation exists in the case of the disilicate, Li2Si20j, a search was made for heat effects inseveralpreparationsinthe disilicate region. The pure disilicate (preparation 601) gave no indication of any phase reaction below the liquidus temperature. Preparations containing excess LinO (606, 607) gave a heat effect at 939O, and those containing excess Si02 (610, 612) at 960°, in presence of tridymite. Neither tridymite nor pure Li2Si03(preparation 602) gives heat effects in this region of temperatures. I t must therefore be concluded that Li2Hi205does not pass thru an enantiotropic inversion, as evidenced by the absence of breaks on the heating curves of the pure compound, but that i t presents a case analogous to the phenomena exhibited by KsSi20j and S a 2 S i 2 0 j above their inversion points. 1 1 . The liquidus for silica consists of two branches, that for tridymite extending smoothly from 82 . z i 0 . I wt. per cent SiO,, 1028', to 91 .o =t= 0 ,I wt. per cent SiO,, 1470'. Above this temperature3 cristobalite is the stable crystalline phase up to I i I ? ' ) its melting point.4 The cristobalite liquidus is of the reverse S type,jdrparting very decidedly from the ideal curve, indicated
*
'
-+
+
Kracek. Bowen and Morev: J. Phrs. Chem., 33, 1 8 j 7
* F. C. Kracek: J. Phvs. Cgem., 74,' 1583 (1930).
(1929).
C. S . Fenner: Am. j . Sci., 36, 3 7 1 ( 1 9 1 3 ) . J . W. Greig: Am. J. Sei., 13, I ( i t u j ) ; Ferguson and Merivin: 46, 41; (1918). F. C. Kracek: J. .4m. Chem. Soc:, 52, 1436 (1930).
2650
F. C . KRACEK
in the diagram by the dotted line thru 1713', derived from the cristobalite liquidus for Rb and Cs silicate mixtures. The departure from ideality of the cristobalite-tridymite liquidus in the lithium system is greater than in the case of any of the other alkalies, and is only exceeded in the alkaline earth silicate systems. I t is interesting to note the fact that the temperature of the eutectics with one of the forms of silica as a eutectic constituent is the lowest temperature at which liquid can exist in any of the (binary) silicate systems thus far studied with competent methods, the only known exception to this observation being found in the case of potassium, the eutectic K2Si205 K2Si409 liquid beinglocated at 75z'whereas the eutectic K2Si409 quartz liquidis at 764'.
+
+
+
+
Summary The system LinO-SiOz contains three compounds: Li4Si04,Li2Si03and Li2SizOj. The equilibrium relations along the liquidus curves and a t the invariant points were studied over the region from 46 to I O O weight per cent SiOp. The results show that Li4Si04decomposes at 1 2 j j', before its melting point is reached, the composition of the liquid phase being j o 9 wt. per cent SiO2. The eutectic point between LilSi04 and LizSi03is at I O Z ~ j~j ,. 3 wt. IO, and its liquidus curve meets the per cent SiOz. Li2Si03 melts a t 1 2 0 1 incongruent melting point curve of Li2Si20jat 1033' and 80 I wt. per cent SiO,, within 0 . 0 5 per cent of the LizSilOs composition. The Li2Si205 and tridymite eutectic point is a t 1028', 8 z z wt. per cent Sios. The composition of the liquid at the tridymite-cristobalite inversion (14;o') is 9 1 o wt. per cent SiOn. Acknowledgment I am indebted to H. E. Merwin for accurate determination of the optical properties of the compounds, as given in Table IT. Geophysacal Laboratory, Carnegw Imtztutaon of Washington, M a y , 1930.
