SYNTHESIS OF ALKALI-METAL SILICOBERYLLATES AND BARIUM ALUMINOSILICATE AND STUDY OF THEIR ELECTRICAL PROPERTIES AT HIGH TEMPERATURES

V. I. ODELEVSKII and R. N. STREL’TSYNA

The development of radio engineering, in particular of high-power generator tubes, is imposing ever-increasing requirements on dielectrics. There is already a need for dielectrics capable of withstanding the action of direct voltage and a high-frequency field at not excessively small gradients and at temperatures of 500–1000°C.

In connection with this, studies aimed at developing inorganic dielectrics with high heat resistance and good electrical-insulating properties at elevated temperatures acquire substantial importance.

Proceeding from certain general regularities, it is possible in advance to reduce substantially the range of substances to be investigated.

Proceeding from the requirement of sufficient stability in vacuum at high temperature and sufficient refractoriness, we must exclude all halide compounds, sulfides, etc.

In practice, one has to confine oneself to oxides, since nitrides, carbides, and the like have high electrical conductivity.

Taking into account the requirement of low electrical conductivity, we must exclude simple and complex oxides of most metals (for example, the series from Ti to Zn, from Zr to Cd, etc.), since they, as a rule, have increased electronic conductivity, and also compounds containing alkalis, because of their high ionic electrical conductivity.

After excluding the comparatively low-melting compounds of boron and the volatile compounds of phosphorus, the oxides of aluminum, silicon, beryllium, magnesium, and the alkaline-earth metals remain.

An important factor is the density of the structure, the absence in the crystal lattice of sufficiently wide channels along which ions could move.

Dense structures may be divided into two groups: structures with dense anion packing (spinel \( \mathrm{MgAl_2O_4} \), beryllium oxide, aluminum oxide, etc.) and structures with loose packing of anions, in which large cations serve as the “densifying agent” (for example, celsian \( \mathrm{BaAl_2Si_2O_8} \)). In the present work, substances belonging to the second group of structures are investigated.

The work is divided into two parts: in the first, the results of investigation of new compounds—silicoberyllates—are presented; in the second, the conditions for obtaining various modifications of barium aluminosilicate and the electrical properties of barium aluminosilicate at high temperatures are studied.

Investigation of Silicoberyllates of Alkaline-Earth Metals

All modifications of silicon dioxide—quartz, cristobalite, and tridymite—are distinguished by a loose structure with a large number of channels (see the structure of cristobalite, Fig. 4, p. 269).

One may attempt to densify the structure by partial substitution

\[ \mathrm{Si}^{4+} \rightarrow \mathrm{Be}^{2+} + \mathrm{R}^{2+}, \]

where \(\mathrm{R}^{2+}\) is a large alkaline-earth ion, \(\mathrm{Ca}^{2+}\), \(\mathrm{Sr}^{2+}\), or \(\mathrm{Ba}^{2+}\). In accordance with the magnitude of the ionic radii, the small beryllium cations should take the place of silicon ions, while the large \(\mathrm{R}^{2+}\) cations should be located in the cavities and densify the structure.

The composition of the product of such substitution corresponds to the molecular formula

\[ \mathrm{RO \cdot BeO \cdot SiO_2 = RBeSiO_4}. \]

Analogous alkali aluminosilicates are known, such as hexagonal kaliophilite

\[ \mathrm{KAlSiO_4}, \]

or pseudocubic carnegieite

\[ \mathrm{NaAlSiO_4}; \]

naturally, in connection with their high alkali content, they are distinguished by relatively high electrical conductivity.

Only one compound of the type mentioned can be named in which divalent ions are located within the cavities: this is trimerite \((\mathrm{MnCa})\mathrm{SiO}_4\) [1].

Trimerite is a very rare mineral.

The existence of a natural mineral does not yet imply the possibility of obtaining a crystalline product of this kind by synthesis at atmospheric pressure.

Synthesis of Silicoberyllates, Determination of Refractive Index, Density, and Thermal Expansion

Initially, in the synthesis, alkaline-earth metal carbonates of “chemically pure” grade, beryllium oxide “for getter,” and rock crystal were used.

Within the firing-temperature range \(1200—1400^\circ\mathrm{C}\), the X-ray diffraction patterns of fired preparations corresponding to the equimolecular composition \(\mathrm{RO \cdot BeO \cdot SiO_2}\) remain unchanged and do not show lines characteristic of modifications of \(\mathrm{SiO_2}\), beryllium oxide, oxides \(\mathrm{RO}\), or alkaline-earth-metal beryllium silicates.

Microscopic examination of the fired preparations, carried out on powders by the immersion method [2], showed that in each of the samples the overwhelming majority of grains belong to one principal phase; there is a small number of grains of other phases, about 2%, which is probably connected with impurities.

Thus, the existence of the compounds \(\mathrm{CaBeSiO_4}\), \(\mathrm{SrBeSiO_4}\), and \(\mathrm{BaBeSiO_4}\) may be considered established.

Further investigations were carried out using specially purified raw materials; silicon dioxide was obtained by decomposition of distilled tetraethoxysilicate \(\mathrm{Si(OC_2H_5)_4}\), while beryllium oxide and alkaline-earth metal carbonates were purified by ordinary procedures [3].

Density was determined by the pycnometric method [4].

Thermal expansion was determined with a sensitive quartz dilatometer, with a magnification factor of 6000.

The measurement results are given in Table 1 and in Fig. 1 (thermal expansion).

Table 1

Physical properties of silicoberyllates

Fig. 1. Thermal expansion of silicoberyllates

1. Ca Be SiO$_4$; 2. Sr BeSiO$_4$; 3. Ba Be SiO$_4$.

Determination of interplanar spacings and intensities of reflected X-rays

X-ray diffraction photographs were taken in Debye cameras, 114.6 mm in diameter, by the asymmetric method.

The powder under study was packed into a celluloid sleeve 0.7 mm in diameter.

Cu$_{k-\alpha}$ radiation with a nickel filter was used to attenuate the $\beta$ lines.

The voltage on the BSV–Cu tube was 30 kV, and the current through the tube was 10 mA.

The values of the interplanar spacings and the relative intensities of the reflected rays are given in Tables 2–4.

Table 2

Interplanar spacings “d” and relative intensities of reflected X-ray beams for the silicoberyllate CaBeSiO₄

Table 3

Interplanar spacings “d” and relative intensities of reflected X-ray beams for the silicoberyllate SrBeSiO₄

Table 4

Interplanar spacings “d” and relative intensities of reflected X-rays for the silicoberyllate BaBeSiO₄

Study of Electrical Conductivity and Dielectric Losses in Silicoberyllates at High Temperatures

Samples for measurements were prepared from previously synthesized silicoberyllates by firing, in a silite furnace, disks pressed from powder. The synthesis temperature and firing temperature were 1400–1420°C.

Silver electrodes were applied to the samples by firing on the corresponding paste.

The electrical conductivity was determined on a conventional setup with a mirror galvanometer. To avoid the development of electrolytic phenomena, in the intervals between measurements the electrodes were short-circuited.

The dielectric loss tangent was determined at a frequency of 1 MHz, using a Q-meter. To eliminate the stray field in the furnace, a coaxial lead-in was designed. Testing showed that with this lead-in design the Q factor of the circuit does not depend on the furnace temperature.

The measurement results are given in Figs. 2 and 3.

As can be seen, the specific resistances of all three silicoberyllates are close to one another and exceed the specific resistance of quartz glass by two orders of magnitude.

Additional experiments showed that replacement of pure raw material by technical-grade material has almost no effect on the magnitude of the electrical conductivity (and also of the dielectric losses).

The dielectric losses at temperatures below 400° are small (\(\operatorname{tg}\delta\) on the order of 0.1–0.2%); at 600–800° they increase substantially (\(\operatorname{tg}\delta\) 3–7%), but without showing a progressive increase (convex curves).

The smallest dielectric losses in this temperature region are possessed by barium silicoberyllate, and the largest by the corresponding calcium compound.

Fig. 2. Temperature dependence of specific resistance: 1. quartz ⟂ to the axis; 2. quartz ∥ to the axis; 3. quartz glass; 4. CaBeSiO₄; 5. SrBeSiO₄; 6. BaBeSiO₄.

Fig. 3. Temperature dependence of the dielectric-loss angle at \(f = 10^6\) Hz.

1. BaBeSiO₄; 2. SrBeSiO₄; 3. CaBeSiO₄.

Obviously, at temperatures above 800°C and a frequency of 1 Mc/s, the value of $\operatorname{tg}\delta$ will be determined mainly by electrical conductivity (Fig. 4); however, at frequencies of $10^7$–$10^8$ c/s the relative importance of conductivity losses decreases sharply, and additional losses acquire substantial significance.

The study of dielectric losses at temperatures of the order of 800°C and frequencies of $10^7$–$10^8$ c/s is in progress.

Fig. 4. Comparison of total dielectric losses with conductivity losses.
1–1′. $\mathrm{BaBeSiO_4}$; 2–2′. $\mathrm{SrBeSiO_4}$; 3–3′. $\mathrm{CaBeSiO_4}$.

Additional losses may be caused by relaxation of impurity ions or by some other similar relaxation mechanism. At present there are insufficient experimental data for a more definite conclusion.

Synthesis and study of various modifications of barium aluminosilicate

The mineral celsian, of composition $\mathrm{BaAl_2Si_2O_8}$, monoclinic, is structurally very close to anorthite $\mathrm{CaAl_2Si_2O_8}$ and albite $\mathrm{NaAlSi_3O_8}$.

In 1915 Ginzberg discovered a second modification of barium aluminosilicate—the hexagonal one [5].

Attempts were made to use synthetic celsian as a refractory.

In recent years interest in celsian has increased in connection with its good electrical properties and low thermal expansion [6].

For technical purposes, especially if internal lamp insulators are considered, it is precisely monoclinic celsian that is of interest, and not the hexagonal modification \(\mathrm{BaAl_2Si_2O_8}\), which we shall hereafter call “hexacelsian.”

A drawback of hexacelsian is its low resistance to thermal shock, associated with a modification transformation at a temperature of \(300^\circ\mathrm{C}\) [7].

Solid-phase synthesis from kaolin and barium carbonate at \(1300\text{—}1400^\circ\mathrm{C}\) always gives the monoclinic modification; however, kaolin contains alkaline impurities, which impair the electrical properties at high temperatures.

An attempt to obtain celsian by a solid-phase reaction from pure raw materials (purified aluminum oxide hydrate and rock crystal) gave a negative result: the product of synthesis proved to be hexacelsian (according to X-ray structural and dilatometric investigations).

The literature data on the conditions for the formation of celsian and hexacelsian are contradictory [5—10]; it was also not known which modification is thermodynamically stable in the range of the usual synthesis temperatures; therefore, a small investigation was carried out of the conditions of formation and of the comparative stability of the two modifications.

Influence of the starting materials on the structure of barium aluminosilicate

To determine the content of each of the modifications in the synthesis product, dilatometric analysis was used, which is widely employed in investigations of dinas [11]. Measurements were made on a highly sensitive dilatometer with a magnification coefficient of 6000.

It was established that the use of kaolin in the synthesis ensures the formation of celsian; a mixture of aluminum oxide hydrate with rock crystal gives hexacelsian on firing.

To establish the equilibrium modification, mixtures of hexacelsian with celsian were prepared, from which specimens were then made and fired.

In Fig. 5 are given the expansion curves of celsian, hexacelsian, and the fired mixtures.

Calculation shows that after a single firing the amount of hexacelsian decreased markedly: from 50 to 13% and from 75 to 27%; consequently, at the firing temperature (\(1400^\circ\mathrm{C}\)) the thermodynamically stable phase is celsian.

Thus, celsian specimens with a small alkali content can be obtained if the method of introducing a kaolinitic monoclinic seed is used in the synthesis of barium aluminosilicate from pure raw materials.

Nevertheless, despite the possibility of such a solution, it was of interest to examine in more detail the influence of the raw materials on the structure of the synthesis product.

The corresponding data are given in Table 5. If kaolinitic raw material is excluded, the most favorable results were obtained when quartz sand and alumina were used.

Approximately equivalent is the variant with barium disilicate and alumina; however, technologically it is considerably more complicated because of the need for preliminary synthesis of barium disilicate.

Table 5

Effect of raw materials on the relative content of hexacelsian in the synthesis product

Fig. 5. Thermal expansion of barium aluminosilicate (effect of the amount of monoclinic additive). 1. hexagonal; 2. monoclinic; 3. 50% monoclinic additive; 4. 25% monoclinic additive.

Let us note in passing that, in the synthesis of barium aluminosilicate from oxides and barium carbonate, it is necessary to employ a holding time of 3–4 hours at 1200°C; otherwise difficulties arise because the reaction does not proceed to completion, readily fusible eutectics appear, the blanks spread, etc.

The facilitated formation of an equilibrium monoclinic phase when rock crystal is replaced by quartz sand may be explained by the mineralizing action of impurities (iron oxides, etc.), which are present in quartz sand in appreciably larger amounts than in rock crystal.

Reprocessing (grinding followed by firing) leads to a further decrease in the content of hexacelsian. In practice, one reprocessing is sufficient.

Study of the Electrical Properties of Barium Aluminosilicate at High Temperatures

Figs. 6 and 7 present the temperature dependences of the specific resistance and \(\tg \delta\) for various specimens of barium aluminosilicate.

Small amounts of alkaline impurities, on the order of hundredths of a percent, do not affect the electrical conductivity: specimens based on alumina and quartz sand proved to be, at the very least, equivalent to specimens prepared from purified aluminum oxide hydrate and rock crystal; however, tenths of a percent of alkaline contamination (specimens based on kaolin) already somewhat reduce the specific resistance.

Fig. 6. Temperature dependence of specific resistance. 1. Quartz glass. 2, 3, 4, 5—barium aluminosilicate. Raw materials: 2. \(\mathrm{Al_2O_3}\), quartz sand; 3. kaolin; 4. rock crystal, \(\mathrm{Al_2O_3}\); 5. \(\mathrm{Al(OH)_3}\), rock crystal; \(\Delta\) the same +5% barium disilicate.

The addition of a small amount of flux (5% barium disilicate) does not affect the electrical conductivity.

Barium aluminosilicate prepared using nondeficit and inexpensive alumina and quartz sand exceeds quartz glass by a factor of 100 in specific resistance at high temperatures.

Alkaline impurities have a stronger effect on dielectric losses than on electrical conductivity.

The lowest dielectric losses at high temperatures are exhibited by specimens prepared from pure raw materials (\(\operatorname{tg}\delta = 0.17\) at \(800^\circ\)C); next comes the specimen prepared from technical oxides (\(\operatorname{tg}\delta = 0.026\)) and, finally, the one prepared on kaolin (\(\operatorname{tg}\delta = 0.14\)).

As can be seen, the difference between the first two specimens is small.

Of particular note are the exceptionally low dielectric losses of specimens prepared from technical oxides and from pure raw materials at a temperature of \(500^\circ\)C: \(\operatorname{tg}\delta\) is only \(0.05\)–\(0.1\%\). The presence of \(10\%\) of a glassy phase (the specimen with barium disilicate), in contrast to the case of electrical conductivity, somewhat increases \(\operatorname{tg}\delta\) (from \(0.026\) to \(0.037\) at \(800^\circ\)C). When present in small amounts, glassy regions may be regarded as isolated inclusions, which explains this difference in the effect of the glassy phase on electrical conductivity (at constant voltage) and on dielectric losses.

Fig. 7. Temperature dependence of the angle of dielectric losses of barium aluminosilicate:
1. \(\mathrm{Al(OH)_3}\), rock crystal; 2. the same \(+ 5\%\) barium disilicate; 3. \(\mathrm{Al_2O_3}\), quartz sand; 4. \(\mathrm{Al_2O_3}\), rock crystal; 5. kaolin.

Fig. 8. Temperature dependence of the angle of dielectric losses:
1. quartz glass; 2. porous ultraporcelain; 3. barium aluminosilicate with zirconosilicate; 4. barium aluminosilicate from pure raw materials.

Barium aluminosilicate specimens have much lower dielectric losses at high temperatures than fused quartz, porous ultraporcelain, and porous talc mass, which are currently used for in-lamp insulators. A sufficiently good property is also exhibited by steatite mass to which barium zirconosilicate has been added to improve sintering (Fig. 8).

Conclusion

In the present work:

1) New compounds—calcium, strontium, and barium silicoberyllates—were obtained; their interplanar spacings and the corresponding intensities of X-ray reflections were determined.

2) The refractive indices, specific gravities, and coefficients of thermal expansion of the silicoberyllates were determined.

3) The electrical conductivity and dielectric losses of the silicoberyllates at high temperatures were investigated.

It was shown that the electrical conductivity of the silicoberyllates at 800°C is 100 times lower, and the dielectric losses 10 times lower, than the corresponding characteristics of fused quartz.

4) Conditions were found for obtaining the monoclinic modification of barium aluminosilicate from technical oxides.

It was shown that monoclinic celsian is thermodynamically stable at a temperature of about 1400°C.

5) The electrical properties of barium aluminosilicate specimens prepared using various types of raw materials were investigated.

It was established that the dielectric losses of barium aluminosilicate prepared from technical oxides are comparatively small even at 800°C: \(\operatorname{tg}\delta = 0.03—0.04\) at a frequency of 1 MHz, i.e., 20 times lower than in fused quartz.

The exceptionally low dielectric losses of barium aluminosilicate at a temperature of 500°C are especially notable \((\operatorname{tg}\delta = 0.0005—0.001\) at a frequency of 1 MHz).

The electrical conductivity of the above-mentioned specimens at 800°C is equal to \(10^8\ \Omega^{-1}\text{ cm}^{-1}\), i.e., one hundred times lower than that of quartz glass.

6) It may be expected that the use of monoclinic barium aluminosilicate (celsian) as a material for internal lamp insulators will be expedient, taking into account its good electrical properties at high temperatures and its relatively small coefficient of thermal expansion. Silicoberyllates, while possessing good electrical properties, have two shortcomings: increased thermal expansion and the presence in the batch of a scarce and expensive raw material—beryllium oxide; for this reason, they are unlikely to compete with celsian.

The work was carried out with the participation of engineers V. G. Prokhvatilov and E. I. Gindin (X-ray structural investigations) and Candidate of Technical Sciences N. V. Kozhevnikova (determination of refractive indices).

Literature

1. Dana E. S. Descriptive Mineralogy. L.–M., ONTI, 1937.
2. Bokii G. B. Immersion Method. Publishing House of Moscow State University, 1948.
3. Karyakin Yu. V. Pure Chemical Reagents. Goskhimizdat, 1947.
4. Zhukovskii G. Yu., Keller K. I., Antonovich N. K. Methods for the Investigation of Ceramic Materials, GONTI, NKTP, 1939.
5. Ginzberg A. S. Izv. Petr. Politekhn. In-ta 23, 34, 1915.
6. Polyakova N. L. Report of the Scientific-Technical Conference on Radioceramics, p. 51, Leningrad, 1955.
7. Ioshiiki V. and Matsumoto K. Journ. Amer. Cer. Soc. 34 [9] 283, 1951.
8. Eskola P. Amer. Journ. of Science [5th] ser. 4 [23], 931, 1922.
9. Tomas R. Journ. Amer. Cer. Soc. 33 [2], 35, 1950.
10. Toropov N. A., Galakhov F. Ya., Bondar I. A. DAN, 89, 1953.
11. Kainarskii I. S., Kukolev G. V., Smelyanskii I. S. Technology of Dinas, ONTI, NKTP, 1938.

ERRATA