Prof. I. F. Ponomarev.

Glassy State and Devitrification.

I. Two States of Matter.

Substances, in passing from the molten liquid state into the solid state, form either crystals or an amorphous mass, depending on the properties of the substance and on the conditions of transition from one phase to another.

During crystallization the particles of a substance occupy strictly definite places in the spatial lattice that we imagine—there results a crystalline structure of the substance. The particles are distributed in space in a definite order and not uniformly in all directions. As a result of the non-uniform distribution of the particles, the properties of the substance—cohesion, thermal conductivity, electrical conductivity, double refraction, etc.—are different in different directions; e.g., quartz, sulfur, diamond, metaphosphoric acid—HPO₃, arsenious anhydride—As₂O₃, Sb₂S₃, etc.

When, however, a substance on cooling yields an amorphous, solid mass, the particles fill all space uniformly, remaining in the same position in which they were in the liquid state. Therefore the properties of the substance are identical in all directions; e.g., quartz glass, the glassy modification of metaphosphoric acid and arsenious anhydride¹), boric acid anhydride—B₂O₃, and many other substances, which on cooling form glass.

Thus solid aggregate states may be divided into 2 groups:

I. isotropic—the amorphous state, and
II. anisotropic—the crystalline state.

The isotropic state differs from the anisotropic in that all the properties of an isotropic substance do not depend on direction in the substance—these properties are scalar. Some properties of a crystalline substance depend on direction—they are vectorial.

Comparatively little attention is given to the study of the glassy state in comparison with other work in the field of studying the properties of solids, which, however, is also not numerous²). The obstacle is the difficulty of research in the field of glasses and the indefiniteness of the concept—what is glass.

II. Glass.

Ordinarily, a glass is understood to be a melt solidified in the amorphous state. Other properties of glass—conchoidal fracture, brittleness, transparency, luster, and surface properties—are not as characteristic as the amorphous structure of glass. However, not all substances with an amorphous structure are glasses. Typical amorphous substances are also gels, which possess certain properties common to glasses, but differ from them in water content and in structure—cellular, sponge-like, etc.³)

The characteristic properties of glass are:

1) capacity for supercooling, 2) viscosity, 3) gradual transition from the liquid phase to the solid and back, 4) a greater store of energy than in crystalline substances: a) specific volume higher, specific gravity lower, b) heat capacity greater, c) heat of dissolution greater.

¹) D. I. Mendeleev.—Principles of Chemistry (1906), pp. 683 and 688.
²) B. P. Weinberg.—The Task of Solid-State Physics. Zh. R. F. Kh. O. (1913). XLV; Phys. Sec., issue 2, p. 67.
³) J. M. van Bemmelen. Zeiff. f. anorg. Chemie 18, 14, 23, 85, 30, 279.

Besides ordinary glass, which consists of silica fused with various oxides, there are a number of other compounds that are found in a glassy state: boric acid anhydride—$B_2O_3$ and salts of boric acid—borates, phosphoric acid and its salts—phosphates, sulfuric acid at low temperature ¹), and a number of organic compounds: betol, piperine, resins, glue, rosin, Canada balsam, etc.

G. Tammann investigated 153 substances, and of these 59 substances, or 38%, could be obtained in the glassy state ²).

Dependence on chemical composition.

Substances possessing high valency, both as anions and as cations, have a great tendency to form glass. Glass is produced chiefly by the following anions:

In salts the following dependence is observed: with an increase in the valency of the cation and anion, the tendency toward supercooling and toward the formation of glass increases. Thus, metaborates of K and Na crystallize readily; metaborates of Ca and other divalent metals give glass. Silicates of divalent metals crystallize readily, e.g. wollastonite—$\mathrm{CaSiO_3}$, augite—$(\mathrm{CaMg})\mathrm{SiO_3}$, whereas alumina silicates crystallize much more difficultly.

For organic compounds Tammann ³) observed the following dependence:

1) the presence of hydroxyl groups favors the formation of glasses, and with an increase in the number of OH groups the tendency to form glass increases,

2) among aromatic compounds the tendency to supercooling proceeds in the following order: greatest of all in meta-compounds, then in ortho-compounds, and least of all in para-compounds.

Metals cannot be kept for a long time below the melting temperature in the liquid state; the number of nuclei formed during supercooling is very large; the same is observed also in metallic alloys; therefore it has not been possible to obtain a metal in the glassy state. Only one metal has been obtained in an amorphous state, but it was not entirely pure. This is so-called explosive antimony, which is separated from a solution of $\mathrm{SbBr_3}$ during electrolysis and contains several percent of bromine. Apart from this, no metal is known that would be obtained in the amorphous state in the form of a compact mass. But in 1908 Cahen and Olie proved that this view of Herard’s concerning the amorphousness of antimony was incorrect ⁴).

Difference between the glassy state and the crystalline state

Depending on whether a substance is in the glassy (anisotropic) state or in the crystalline (isotropic) state, this substance, with one and the same chemical composition, has a different store of energy and occupies in space, for the same mass, different volumes, i.e. the specific gravities of the given substance in the two states are different ⁵).

These properties are expressed by the curves of isotherms (Fig. 1) and isobars (Fig. 2); on these curves the crystalline state is indicated by a solid line, and the glassy state by a dashed line. The isobar clearly depicts the process that occurs with a substance during cooling, when the substance passes from the liquid phase into the solid. If during such a transition from one phase to another crystals are formed, then at the crystallization temperature there is observed a sharp jump

¹) E. V. Biron, Zh. R. F. Kh. O. XXXI. 1899, p. 171.
²) G. Tammann, Zh. R. F. Kh. O. 1902, physical part XXXIV, section I, p. 83.
  » » Kristallisieren u. Schmelzen 155.
³) Tammann. Zeit. f. Electrochemie 1904 X. s. 537.
⁴) W. Guertler Metallographie s. 134.
⁵) Kohlson. Fizika III. p. 316. Tammann. Kristallisieren und Schmelzens 9, 12.

in properties: crystals occupy a smaller volume than the liquid; this is accompanied by a decrease in the store of energy, which is expressed in the release of “hidden heat of fusion.” If glass is formed, then the properties of the substance change continuously, and therefore the curve for glass runs higher than that for crystals. But these curves intersect in the region of lower temperatures, and after this point of intersection of the curves the density of the glass may be greater than the density of the crystals.

Stein,¹) studying the specific gravities of silicates, observed both cases, as expressed in the following table:

At ordinary temperature the density of crystalline metasilicates is higher than the density of glassy ones, excluding Zn SiO₃; in it the reverse is true. The same phenomenon is also observed in metaborates.

The store of energy of a substance can be measured from its heat capacity and from the heat of solution.

a) The heat capacity of a substance in the amorphous state—\(C'\)—is higher than in the crystalline state—\(C''\).

\[ C' > C'' \]

which was verified for the following substances:

b) The heat of solution of a substance in the amorphous state is greater than that of the same substance in the crystalline state, as is seen from the following table, in which the specific gravities are also given.²)

This can explain the fact that glass dissolves more readily than crystals. This was verified on silicates; Lemberg also pointed out the significance of this property of silicates for geology. It is difficult to measure the rate of solubility of a substance in the isotropic state, since this phenomenon proceeds differently than in crystals; it resembles the way in which swelling substances—colloids—go into solution.

III. Conditions for the formation of glasses.

To obtain a substance in the glassy state, it is necessary that it possess the following properties:

1. a great tendency toward supercooling.
2. great viscosity,

¹) Stein, Darstellung einiger Silikate. [Zeit. f. anorg. Chemie 55. (1907) s. 157]
²) Tammann, Kristallisieren und Schmelzen s. 61.

1. low spontaneous capacity for crystallization,
2. low rate of crystallization.
   Let us consider how each of these properties affects the formation of glass.

I. Supercooling.

A cooling liquid, when supercooled, becomes continuously more and more viscous and, finally, becomes solid, passing into an amorphous solid body—forming glass. In such a transition no jumps are observed; this transition of a liquid into glass may be compared with the way in which, at the critical temperature, a gas and a liquid pass continuously into one another, i.e., without a change in volume and without absorption of heat¹.

Therefore, on the cooling curve of glasses there are no arrests corresponding to a constant crystallization temperature, as occurs in crystalline substances (Fig. 3, curve 1—1). If a crystallizing substance has a tendency to supercooling, but only down to a certain temperature, and then crystallization begins, the curve will have a sack-like depression before the horizontal segment (curve 2—2); for example, antimony is readily supercooled and gives such a curve.

If, during the solidification of a substance, not enough heat is liberated to raise the temperature of the melt and maintain this temperature all the time until the last crystal has formed, then two bends, \(a\) and \(b\), are visible on the cooling curve (curve 3—3): the upper bend \(a\) corresponds to the temperature of the beginning of solidification—the beginning of crystallization; the lower bend \(b\) corresponds to the temperature of the end of solidification.

Finally, the cooling curve of glasses [curve (4—4)] has a smooth form; on it there are neither arrests nor breaks, because in this case, during the transition from the liquid state to the solid, the molecules of the substance do not occupy in space the definite arrangement that is inherent in their crystalline form. In this transition there is not liberated that quantity of heat which must be liberated during the regrouping of molecules, and which is the heat of crystallization and is expressed on the cooling curve by a horizontal segment.

The molecules in glass do not have the stable position possessed by molecules in a crystalline substance. They have solidified in the position in which they were in the liquid. Glass is an unusually strongly supercooled liquid, in which the viscosity has increased so much that, in its numerical magnitude, it is no longer viscosity but hardness. Therefore glass is called a solidified liquid or even a solid liquid. Solid glass has properties analogous to those of a liquid, which is clearly seen from Figs. 1 and 2, where the curve of the glassy state is a continuation of the curve of the liquid state.

Such cooling curves were obtained by many investigators of silicates. These curves usually were the reason for discontinuing further study of silicate systems by means of fusibility curves. From the cooling curves obtained, it was possible to say only that glass is obtained; no data characterizing the substance under study could be extracted from these smooth curves. It was necessary to seek other methods for studying glasses; the method of thermal analysis proved unsuitable.

2. Viscosity.

With increasing supercooling of a liquid, the viscosity increases. The viscosity curve rises rather rapidly and has a continuous form.

The viscosity curve is of great importance for the technology of glass. From it one can judge the qualities of the glass—whether it is suitable for blowing. For blown—

glass must have a definite viscosity. Let us suppose that, in our diagram, this range of viscosity values lies between the lines aa and ee. The points of intersection of the viscosity curves with the lines aa—ee will indicate between which temperatures the working of the glass is possible. Curve (1—1), having the smaller temperature interval \(t_3 — t_4\), shows that this glass will be worse for working than the glass whose viscosity is represented by curve 2—2 with the interval \(t_1 — t_2\); it will be more difficult for the craftsman to seize upon the small interval in order to complete all the operations that are necessary in the manufacture of glass articles.

3. Spontaneous capacity for crystallization.

Observing the formation of crystals during the cooling of a substance, we can readily notice that crystallization does not proceed at once throughout the whole mass. The first crystals appear at separate points of the mass. And from these points proceeds the growth of crystals, which then fill the entire volume of our substance. Such spontaneous nucleation of crystals depends on the temperature of the liquid phase of our substance; it depends on the degree of supercooling, since above the melting temperature no nucleation of crystals can occur.

In some cases this property can be measured and expressed by a number. Tammann proposed the following method of measurement. A substance contained in a narrow glass tube with thin walls is heated to a certain temperature above the melting point and is quickly transferred into a bath at a temperature below the melting point. In the cooling substance, separate crystallites appear, which grow into spherulites. From the fixed spherulites thus formed, or nuclei, as Tammann calls them, one can count the number of crystallization centers that have formed.

In the bath where the crystals nucleate, the tube is kept for a definite time, for example, 2 minutes, and then the tube is quickly transferred into another bath with a low temperature, at which crystallization centers no longer form, while all the remaining liquid solidifies into a transparent glass.

If the temperature of the bath in which the crystals nucleate is varied, one can construct a curve of the number of nuclei. In these determinations the following remain constant: the time during which nucleation of crystals occurs, and the temperature to which the liquid is heated before immersion in the first bath. The numbers obtained are referred to a unit of time—1 min.—and to a unit of volume—1 cubic centimeter. Plotting along the abscissa axis the temperature of the bath where the crystallization centers form—the temperature of supercooling—and along the ordinate axis the number of nuclei formed in 1 min. in 1 cu. cm. of the substance under investigation, we obtain the curve of the number of nuclei. From this curve it is seen that the number of nuclei increases with the temperature of supercooling up to a certain maximum and then rapidly falls, approaching zero¹.

The number of nuclei may be used as a measure of the stability of the supercooled liquid.

The greater the number of nuclei, the more prone the substance is to a change in its state of aggregation. According to the curve of the number of nuclei, one may draw the conclusion that, with an increase in supercooling, the stability of the substance at first decreases, and then, upon further supercooling, increases considerably.

It is possible, in Tammann’s opinion, that considerably below the melting point some substances are more stable in the amorphous state than in the crystalline state; in the diagram (Fig. 2) this is expressed by the fact that the curve of the glassy state may, in the solid phase, have a point of intersection with the curve of the crystalline state and then run below this curve, i.e. the volume of the glass will be smaller than that of the crystal.

4. Rate of Crystallization.

This property of a substance, like the capacity for crystallization, manifests itself at temperatures below the melting point and, as a function of supercooling, is expressed by the curve shown in Fig. 6.

At the melting temperature the rate of crystallization is equal to 0; then, upon cooling of the substance, the rate of crystallization increases, reaching a maximum, which over a certain interval of temperatures remains constant, and then rapidly falls, so that in a highly supercooled substance the rate of crystallization is practically equal to 0.

To determine the rate of crystallization, the growth of crystals in one minute is measured in millimeters. For readily crystallizing substances the rate of crystallization is \(> 3\) mm per minute, and the curve of the rate of crystallization has a horizontal segment, as shown in Fig. 6. If, however, the rate of crystallization is \(< 3\) mm per minute, then the heat released during crystallization is so small that this horizontal segment on the curve (Fig. 6) may turn into a point.

The work of Kittl¹ on the rate of crystallization of silicates confirmed this. He took the minerals: olivine, leucite, wollastonite, diopside, hypersthene, nepheline. The curves of the rate of crystallization were obtained with a sharp maximum. The greatest value for hypersthene was equal to 2.5 mm per minute; the smallest—for nepheline—0.4 mm per minute.

For glasses these quantities have not been determined. One may calculate approximately the “average” rate of crystallization of glass over that interval of time during which, at the corresponding temperatures in the course of cooling of the glass, it crystallized. Let us take, for example, an ordinary lime-soda glass, close in composition to the formula

\[ \mathrm{Na_2O \cdot CaO \cdot 6SiO_2}. \]

Such glass is often melted in factories for the production of window glass and ware; in the form of a small impurity such glass contains FeO and \(\mathrm{Al_2O_3}\), in all about 2%.

Glass discharged from the tank into a pit cools for about 3 days. One may assume that the temperature range in which crystallization took place lasted about 50 hours, i.e. \(50 \times 60 = 3000\) minutes. During this time, spherulites with a radius of 5–10 mm grow from the centers of crystallization.

Hence the rate of crystallization of glass is approximately

\[ = \frac{5}{3000} - \frac{10}{3000}, \]

i.e. from 0.0017 to 0.0033 mm/min.

How is glass formed? In Fig. 7 the curves present the conditions under which, upon cooling from the melt, we obtain crystals, and under which—glass. If the maximum of the spontaneous capacity for crystallization on curve 3—3 lies in the same temperature region as the maximum of the curve of the rate of crystallization 2—2 and coincides with the region of low viscosities, as is seen from curve 1—1, then crystals are formed.

If, however, the curve of the spontaneous capacity for crystallization 4—4 has a weakly developed maximum and intersects the curve of the rate of crystallization at its minimum values, then, with sufficiently rapid cooling, centers of crystallization are not formed, and we obtain glass.

The possibility of the appearance of crystals and their growth lies at such temperatures at which the viscosity, represented by curve 1—1, rapidly reaches very large values—the substance becomes solid; and in the solid state crystallization is impossible under ordinary conditions.

IV. Method of Forced Crystallization of Glass.

When considering the conditions for the formation of glasses, it is clearly seen how great are the difficulties in the thermal analysis of substances that give glass on solidification.

Finding this most advantageous temperature for the appearance of crystallization centers and their growth is extremely tedious—it is necessary to carry out many tests. Attempts have been made to seed crystals into the glass in order to accelerate its crystallization; and this does little to make the task easier.

At the Geophysical Institute of the Carnegie Institution (Washington), enormous work was carried out on the crystallization of glasses with the aim of investigating such systems.

The method I have developed greatly simplifies the finding of this most advantageous temperature and therefore makes it possible to investigate systems that give glass on solidification.

The furnace used for this purpose is shown in Fig. 8; it, like the method itself, is very simple. It is made of asbestos cardboard, which is then coated with a paste of water glass and clay. In the front wall of the furnace an opening is made, through which a test tube of refractory glass is inserted, resting on asbestos-cardboard supports. Into the test tube is inserted a platinum boat, into which the previously melted substance is poured.

The platinum boat is heated with a gas burner from one end in such a way that at this end the temperature is higher than the melting temperature of the substance under investigation, while at the cold end the substance solidifies into a glass; and in this state the boat remains throughout the entire experiment.

Along the boat a constant temperature drop is established, which is expressed by curve a in Fig. 9. Thus, in the boat, by itself, that temperature most advantageous for crystallization is realized, which is found with such difficulty by the quenching method or by other means.

Crystallization centers appear, and the nuclei that subsequently form grow in the direction of both ends of the boat with the substance contained in it. On the cold side the growth of crystals ceases where, with decreasing temperature, the rate of crystallization rapidly falls and becomes entirely imperceptible. In the direction toward the hot end of the boat, the crystals grow in this viscous, syrup-like liquid until they reach the region with a temperature corresponding to the equilibrium between crystals and melt; and then the sharply formed boundary between the crystallized part and the liquid no longer moves. It is possible to shorten the time required to reach this boundary. For this purpose we move the gas burner closer to the crystals, or push the platinum boat with the substance deeper in. Part of the crystals that have formed melts, and after a short time—15 minutes is sufficient—the equilibrium boundary between the liquid and crystalline phases is established again. We lower the junction of the thermoelement into the liquid at the boundary of the crystals and measure the temperature, which, as the equilibrium temperature between the liquid and crystalline phases, is the melting temperature of our substance.

The method of forced crystallization was tested on $\mathrm{Na_2SO_4}$ and $\mathrm{Na_2O . 2B_2O_3}$ by means of the cooling curve, as described in greater detail in my work “Acid Sodium Borates.” [J. R. Phys.-Chem. Soc. XLIX. 1917, issue 3–4, p. 235]. After this, the following glass-forming systems were investigated.

V. Investigation of Systems.

1. System $\mathrm{Na_2O . 2B_2O_3 — B_2O_3}$.

In this system the described method of forced crystallization of glasses was applied for the first time. The data obtained on the melting temperatures of mixtures intermediate between borax and boric acid anhydride are expressed in

the melting curve (Fig. 10). The two maxima C and E on this curve correspond to two chemical compounds discovered in this way.

\[ \begin{aligned} \mathrm{Na_2O\cdot 3B_2O_3} &\text{—with m.p. } 694^\circ,\\ \mathrm{Na_2O\cdot 4B_2O_3} &\text{—” ” } 783^\circ. \end{aligned} \]

Between them, borax, and \(\mathrm{B_2O_3}\), there are three minima.

\[ \begin{aligned} \mathrm{B} &\text{—with m.p. } 580^\circ,\\ \mathrm{D} &\text{—” ” } 658^\circ,\\ \mathrm{F} &\text{—” ” } 582^\circ. \end{aligned} \]

The melting temperature of one component of this system—borax—was established quite accurately by the forced-crystallization method and on the basis of the cooling curve; it is \(=732^\circ\). It has not yet been possible to determine the melting temperature of \(\mathrm{B_2O_3}\), because mixtures with a high content of \(\mathrm{B_2O_3}\) required a large amount of time to reach the equilibrium boundary between the crystals and the liquid phase; for example, for the last mixture with 64.5% excess \(\mathrm{B_2O_3}\), continuous heating for days was required, and the appearance of the first nucleus was observed after 10 hours, whereas for borax the first nucleus appeared after 15 minutes, and after 50 minutes the boundary was established.

A mixture with 90% excess \(\mathrm{B_2O_3}\) was heated continuously for three weeks; crystals formed in the viscous part of the melt. It is extremely interesting to note that \(\mathrm{B_2O_3}\) had not hitherto been crystallized, but by this method it was possible to approach very closely the crystallization of \(\mathrm{B_2O_3}\), since the crystals obtained probably contained only about 3% \(\mathrm{Na_2O}\).

Thus, the former opinion that the anhydride of boric acid cannot be obtained in the crystalline state may be regarded as refuted. By the method of forced crystallization, any glass can be crystallized; the difference will be only in the time during which the glass must be subjected to heating.

The crystals formed were subjected to microscopic examination. Pure borax crystallized in star-shaped spherulites. Melts with 5 to 10% excess \(\mathrm{B_2O_3}\) crystallized in the form of plates with an attached pyramid (Fig. 11, a). Melts corresponding to the branch BCD on the equilibrium curve (Fig. 10) also crystallize in the form of plates, with sharpened edges (—b); melts corresponding to the branch DEF yielded crystals of coffin-like form—c.

Comparing these data, we conclude that in the system under discussion we have an uninterrupted continuous series of mixed crystals. On the equilibrium curve there are two clearly expressed maxima; the composition of the mixtures at these maxima obeys the law of simple multiple ratios between the components, so that these two types of crystals, в and с, may be regarded as belonging to two different compounds. At the minima the curve has a smooth transition, which corresponds to mixed crystals, and not the sharp bend characteristic of a eutectic.

2. System \(\mathrm{Na_2O\cdot 2B_2O_3—3CaO\cdot P_2O_5}\).

The first system was the beginning of a series of investigations in the field of melts that yield glasses upon cooling. The work was carried out in the laboratory of Prof. Tammann in Göttingen.

In Tomsk it proved possible to begin the investigation of the planned systems, with the aim of clarifying the question of what glass is.

The starting materials for this system were Kahlbaum preparations:

$$ \mathrm{Na_2O}\cdot 2\mathrm{B_2O_3}, $$

$$ \mathrm{CaHPO_4}\ \text{and} $$

$$ \mathrm{CaCO_3}. $$

From these, 4 mixtures were prepared which, according to stoichiometric calculation, after melting gave the following points for the system:

$$ \mathrm{Na_2O}\cdot 2\mathrm{B_2O_3}—3\mathrm{CaO}\cdot \mathrm{P_2O_5} $$

The mixtures were melted in a platinum crucible and poured into a platinum boat, which was placed in a furnace so that only one end of this boat was heated by the flame of a gas burner.

The order of mixing was usually as follows: $\mathrm{CaCO_3}$ was poured into the molten borax and, after the evolution of $\mathrm{CO_3}$ bubbles had ceased, $\mathrm{CaHPO_4}$ was added. The phosphate dissolves slowly; the mixtures were melted for 3–4 hours, until the last white grains of $\mathrm{CaHPO_4}$ had disappeared.

The glass obtained from the first two mixtures was fairly fluid—it poured easily from the crucible. The glass was milky white in color, and the color became denser with increasing $\mathrm{CaHPO_4}$ content.

The melting temperature was measured with a thermoelement after a sharply formed boundary between the crystals and the liquid phase had been reached.

Centers of crystallization appeared fairly quickly, after $1/2$–$3/4$ hour, in the first two mixtures:

$$ \text{with }95\%\text{ borax and} $$

$$ 90\%\quad » $$

and the boundary between the liquid phase and the crystals was reached within 4–5 hours. For mixtures with a higher content of $3\mathrm{CaO}\cdot \mathrm{P_2O_5}$, 8–10 hours of heating were required, and the melts of these mixtures had a much higher viscosity.

The fusibility curve of the system $\mathrm{Na_2O}\cdot 2\mathrm{B_2O_3}—3\mathrm{CaO}\cdot \mathrm{P_2O_5}$ gives a sharp and very interesting minimum (Fig. 12). When melting 90% borax, which has a melting temperature of 732°, with 10% tricalcium phosphate, whose melting temperature was not determined but is probably high, a mixture with a melting temperature of 475° was formed; the melting temperature was lowered by 275° upon addition of a high-melting component.

The coefficient of expansion of the glasses obtained is not high, because these glasses, while cooling on a platinum lid in the form of a thin layer, gave almost no cracks even upon rapid cooling. This is in full agreement with the data of Winkelmann and Schott, who expressed in definite numbers the dependence between the coefficient of expansion and the chemical composition of glass:

$$ \begin{aligned} \mathrm{B_2O_3} &= 0.1 \qquad & \mathrm{Na_2O} &= 10.0 \\ \mathrm{P_2O_5} &= 2.0 \qquad & \mathrm{CaO} &= 5.0\ \text{etc.} \end{aligned} $$

The results obtained are of great interest for glazes and especially for resolving the question of leadless glazes. Lead glazes are poisonous, especially in the form in which they are used in the manufacture of pottery articles, which are usually kitchenware. A lead glaze from

under the action of various organic acids it dissolves; lead, passing into food, produces slow poisoning of the organism, which is not detected soon. A large literature in ceramics is devoted to the lead question. In some countries laws have been worked out regulating the use of lead in glazes.

Small amounts of $\mathrm{SiO_2}$ were added to the mixtures studied. This addition considerably increased the viscosity of the melt and the melting temperature.

3. System $\mathrm{Na_2O \cdot 2B_2O_3--Al_2O_3}$.

Mixtures of borax and alumina (Kahlbaum preparations) were fused in a platinum crucible until completely dissolved. The liquid melt obtained was poured into a platinum boat about 12 sant. long; the boat was placed in a furnace and subjected to heating by the method described above.

In melts with a small content of $\mathrm{Al_2O_3}$ crystallization began rather quickly—after 2–3 hours; the formation of a sharp boundary between the molten part and the crystallized part was achieved after 6–8 hours of heating; after this the temperature was measured with a thermoelement in the usual way, as described in the method of forced crystallization of glasses.

The data obtained are given in the table and expressed by the melting curve (Fig. 13), where the measured equilibrium temperatures between the crystallized part and the molten part are plotted on the ordinate axis, and on the abscissa axis—the molecular percent of the components alumina and borax.

On the curve obtained we have 1 minimum and 1 maximum. Of especial interest is the first minimum at 5% $\mathrm{Al_2O_3}$ with a melting temperature of 400°, 332° below the melting temperature of the solvent—borax—from the addition of such a refractory substance as alumina.

Besides these melts, a melt with 50% (molecular) $\mathrm{Al_2O_3}$ was also prepared, therefore of a composition corresponding to the formula: $\mathrm{Na_2O \cdot 2B_2O_3 \cdot Al_2O_3}$.

To obtain a homogeneous melt, very prolonged heating was required—more than 20 hours. After cooling, a transparent, lustrous glass was obtained, which had such a small coefficient of expansion that, upon repeated melting on a blowpipe burner and very rapid cooling in a platinum crucible, the glass did not give cracks, as though possessing the same coefficient of expansion as platinum.

The glass is fairly resistant with respect to water and acids and presents great difficulties when removed from the walls of the platinum vessel, even if the glass has a thin layer—for example, when a drop of melt that has flowed down the wall solidifies.

The preparation of the melts and the measurements of the temperatures were carried out by K. B. Pospalskii.

VI. Devitrification, or Crystallization, of Glass.

On slow cooling of molten glass, crystals separate out from it. This phenomenon in technology is called devitrification of glass.

Devitrification of glass has been the subject of investigation by many scholars. Berthelot defines it as a process of physical change in the state of glass, similar to the way in which arsenious-acid anhydride passes from a glassy mass into ...

over time into an opaque white. Pelouze holds the same opinion on the basis of his analyses, which showed that there is no great difference between the crystals separated out from the glass and the remaining mass.

Dumas (1830) admitted the formation in glass of definite silicates, whose melting temperature is higher than that of the surrounding amorphous mass of glass. In composition these silicates differ from the rest of the mass. On the basis of his analyses Dumas found that the crystals of devitrified glass contain less alkali than the undevitrified part. Therefore Dumas tried to explain the separation of crystals by the volatilization of alkalis during the melting of glass—seeing in this a condition favorable to the rearrangement of the glassy melt.

Further studies of devitrification were based on the study of the fusibility of silicate systems, applying the phase rule. But investigations in this direction are considerably hampered by the fact that in glass we are dealing with a multi-component system; whereas in most of the systems studied the number of components rarely exceeds three.

Moreover, the chief obstacle is presented by the basic properties of glass:—1) a sharply expressed ability to supercool, 2) a weakly expressed capacity for spontaneous crystallization, and 3) a very small magnitude of the rate of crystallization. Therefore crystallization of glass usually occurs only with very slow cooling, which is what takes place when large masses of glass cool.

If, during manufacture or during reheating, glass separates out crystals, we have to do with bad glass. In the short time during which the making of glass and its tempering, cooling, take place, visible centers of crystallization must not appear. Optical glass, in particular, must satisfy this requirement; it is sometimes subjected to very slow cooling over the course of a week. Other kinds of glass devitrify during this time, which also occurs when the furnace is extinguished and glass is let out of the tank or from pots. With such slow cooling, together with supercooling, the rate of crystallization increases, as is expressed by the curve (Fig. 6). The temperature of maximum crystallizing ability in glasses lies lower than the range of temperatures with the greatest rate of crystallization,—as shown in drawing 7.

Passing gradually from high \(t^\circ\) to low ones, we first enter the region with a small number of spontaneously formed centers of crystallization, from which crystals begin to grow while the rate of crystallization has a sufficient magnitude. At low \(t^\circ\) new centers of crystallization are formed, but the rate of crystallization is so small that the crystals do not grow—they are not visible. Therefore, in devitrification, the following picture is observed: from individual, few places in the glass, crystals grow ray-like; these aggregates in the form of spheres are called spherulites; they sometimes reach a large size—up to 6 cm. Therefore the upper layers of the discharged glass solidify into transparent glass, while the deeper down, the more spherulites there are, and the lowest layers solidify into a continuous mass.

When glass passes into the crystalline state, the specific gravity increases. For example, according to determinations:

Geologists believed that, upon the melting of silicates, a decrease in volume occurs. This evidently does not correspond to reality¹. Crystals have greater hardness than glass, and they are more resistant to chemical reagents than glass. Crystals soften at \(t^\circ\) considerably

higher than glass. Silicate systems most often yield mixed crystals, which in composition differ little from the noncrystalline part. This is confirmed by the following analyses:

I — Terreil—Dinglers Polytechn. Journ. 148. 58.
II — Schwarz. — ” ” 205. 422.
III — Groger. — ” ” 142. 297.
IV — Peligot.—Bulletin de la Société d’Encouragement. 1874. p. 342.

From the analyses cited it is evident that it is not a simple compound that separates out, but that all the components of the glass enter into the devitrification product. These products of devitrification of technical glasses are mixed crystals, in composition almost indistinguishable from the glass. Taking into account the property of mixed crystals—as solid solutions—of uniform mixing of the constituent parts, one may suppose that the best glasses are those which, upon devitrification, give crystals that do not differ in composition from the remaining mass.

Very interesting are the crystalline separations during the devitrification of glass from the tank at the Znamensky works. In the glass there form prismatic hexagonal plates with regular and sharp outlines of the contours (Fig. 14). At other works in European Russia I have had occasion to observe only spherulites, formed from needle-shaped crystals. In its composition the glass of the Znamensky works belongs to the category of ordinary glasses expressed by the formula Na₂O · CaO · 6SiO₂.

On March 20, 1921, I applied the method of forced crystallization of glass on a factory scale, setting glass from the tank to crystallize. After 24 hours the appearance of crystals was already observed. After 48 hours the experiment was finished. The glass was cooled in the annealing furnace. In the glass there formed layers of the same hexagonal prisms as during the cooling of swollen glass, only less sharply expressed and smaller. The wavelike distribution of these crystals depends, as it were, on the fact that the glass taken from the tank, being viscous, was distributed in the boat in the form of waves. The glass of the Znamensky works differs from others in that for melting not chalk but marble is used. It is possible that the sand also has a considerable influence, since it contains impurities. Since no chemical analysis of the raw materials is carried out at the Znamensky works, below is given an approximate calculation of the composition of the glass from the tank furnace of the aforementioned works.

Composition: sand 25 poods . . . . . . . . . . . . \(100 — \mathrm{SiO_2}\)
Marble 7 ” . . . . . . . . . . . . . . . . . \(28 — \mathrm{CaCO_3}\)
Sulfate \(9\frac{1}{2}\) ” . . . . . . . . . . . . . \(38 — \mathrm{Na_2SO_4}\)
Sawdust \(1\frac{1}{2}\) ” . . . . . . . . . . . .
\[ \frac{166}{} \]

\[ \mathrm{SiO_2}=100=100=75.6\% \quad \text{by weight} \]

\[ \mathrm{CaO}=28\cdot \frac{56}{100}=15.7=11.8\% \quad » \]

\[ \mathrm{Na_2O}=38\cdot \frac{43.7}{100}=\frac{16.6}{132.3}=\frac{12.6\%}{100.0\%}\quad » \]

\[ \begin{aligned} \mathrm{SiO_2}&=75.6/60=1.26\cdot 4.75=6,\\ \mathrm{CaO}&=11.8/56=0.212\cdot 4.75=1,\\ \mathrm{Na_2O}&=12.6/62=0.203\cdot 4.75=0.965,\quad \text{therefore} \end{aligned} \]

\[ 0.965\,\mathrm{Na_2O}\cdot \mathrm{CaO}\cdot 6\mathrm{SiO_2}. \]

For the investigation of the conditions of crystal formation it is necessary:

I. To carry out a chemical analysis of the glass and of the crystals. In order to separate the crystals from the glass one may use HF, which will dissolve the glass more rapidly—the crystals will remain;

II. To determine the specific gravity of the glass and of the crystals.

III. To pay attention to the intergrown branches of crystals, which have a definite direction.

Le Chatelier¹) indicates a curious case of devitrification of crystal glass, which occurred during the German invasion of Baccarat. At that time, in the furnaces there had been left at \(1350^\circ\) a sodium mixture for preparing crystal glass, called “quinquet,” intended for lighting.

During the 20-day occupation by the Germans the \(t^\circ\) was lowered to \(800^\circ\). But after the Germans had left, it was noticed that the furnace had to be stopped for repair and allowed to cool completely. The crystal glass proved to have devitrified. Crystals of \(\mathrm{SiO_2}\) of low specific gravity—tridymite—had formed, in the form of hexagonal plates of unusual size.

This is the first observed devitrification of crystal glass. Previously devitrification had been observed only in ordinary glasses, with the assumption that wollastonite separates out.

Attempts to reproduce this experiment were unsuccessful; it requires special conditions of slow cooling at low \(t^\circ\).

It should be noted that tridymite formed in Na—crystal glass. In K—crystal glass it did not form. With Li crystallization proceeds still more readily. Here the dependence of the crystallizing ability on atomic weight is evident. \( \mathrm{Li}=7.03,\ \mathrm{Na}=23.05,\ \mathrm{K}=39.15.\)

This experiment also indicates that above \(800^\circ\) tridymite is stable, below \(800^\circ\)—quartz, while cristobalite has only temporary stability.

VII. Theory of Glass.

On the basis of the physico-chemical definitions set forth, I consider that glass should be called such a substance, irrespective of chemical composition, which, crystallizing from a molten magma, gives crystals of the same chemical composition as the magma. If, upon the separation of crystals, as for example in metals and many salts, there occurs a division of the substance into crystals of one composition and magma of another composition, then we are dealing not with

¹) Academic des Sciences. Séance du 5 juin 1916

To the article by I. F. Ponomarev,
“Glassy State and Devitrification”

Isotherm

Fig. 1.

Isobar

Fig. 2.

Crystalline substances  Cooling curves.

Fig. 3.

To I. F. Ponomarev’s article
“Glassy State and Devitrification”

Curves of Glass Viscosity

Fig. 4.

Axis labels:

• \(\eta\) — viscosity, friction
• Temperature
• \(a\)
• \(b\)
• \(t_1\), \(t_2\), \(t_3\), \(t_4\)

Curve labels:

• 1
• 2
• 3

Legend:

1—viscosity
2—rate of crystallization
3 and 4—number of crystallization centers.

Spontaneous Crystallization Capacity

Fig. 5.

Axis labels:

• Number of crystallization centers per unit volume
• Temperature

Rate of Crystallization

Fig. 6.

Axis labels:

• cm/min
• Temperature

Fig. 7.

Axis labels:

• Temperature

Curve labels:

• 1
• 2
• 3
• 4

Legend:

1—viscosity
2—rate of crystallization
3 and 4—number of crystallization centers.

Fig. 8.

Labels in the apparatus diagram:

• \(m\)
• \(A\)
• \(a\)
• \(b\)
• \(c\)
• \(d\)

Furnace for the crystallization of glass.

Fig. 9.

a.

Axis labels:

• Temperature
• Liquid melt
• Crystals
• Glass

b.

Axis labels:

• Number of centers
• Temperature

c.

Axis labels:

• Rate of crystallization
• Temperature

Na₂O·2B₂O₃ — B₂O₃

Fig. 10.

Vertical scale: 500, 600, 700, 800.

Horizontal scale: 10, 20, 30, 40, 50, 60, 70.

Horizontal-axis label: Weight % of excess B₂O₃

Marked points and compositions: B, C, D, E, G; Na₂O·3B₂O₃, Na₂O·4B₂O₃.

Crystals of the system

Na₂O·2B₂O₃ — B₂O₃

a

b

Fig. 11

c

Na₂O·2B₂O₃ — 3CaO·P₂O₅

Fig. 12.

Vertical scale: 400, 500, 600, 700.

Horizontal scale: 5, 10, 15, 20%.

Horizontal-axis label: Molecular % of 3CaO·P₂O₅

Na₂O·2B₂O₃ — Al₂O₃

Fig. 13.

Vertical scale: 300, 400, 500, 600, 700.

Horizontal scale: 5, 10, 15%.

Horizontal-axis label: Molecular % of Al₂O₃

Fig. 14.

glass, even if its external features resembled glass. Such substances are more properly called glass-like.

I arrived at this theoretical definition on the basis mainly of the following properties:

1. The composition of the crystals that separate out in glass is the same as that of the remaining molten mass.

2. The melting temperature of the separating crystals changes with a change in the composition of the mass; if the compositions of the crystals and of the glass were different and the crystals, as in metallic alloys, separated out with one and the same composition from different mixtures belonging to one branch of the melting curve, then the melting temperature of the crystals of these mixtures should have been the same.

Considering this theoretical definition to be a first approximation to an exact scientific characterization of glass, which has hitherto been lacking, I hope that further work will bring still greater definiteness to this very interesting and technically important field.

1. Tammann. Kristallisieren u Schmelzen. p. 48. ↩↩↩↩