on the electrical conductivity of silicates at high tempera

tures.

Although the authentic data which have recently accumulated on the behaviour of the silicates at high temperatures possess a great interest and value from the standpoint of the mineralogist and the geophysicist, a knowledge of the physical melting-points and fields of stability of the silicates is not the most important factor for consideration in so far as application to the metallurgy of iron is concerned. In a study of the blast-furnace we are particularly interested in the behaviour of the slag from the time it enters the zone of fusion until it is fiushed from the cindernotch. In transit through this region where the smelting process occurs the most important physical property of the slag is its viscosity, while its most important chemical property is its desulphurising power, or ability to absorb the sulphur of the charge.

It was early apparent to furnacemen that blast-furnace slag in the molten condition was much more "viscous or "viscid" than molten iron and the fused salts of ordinary acids, and that the slag underwent a gradual softening on heating rather than a sudden change to a mobile liquid, as is characteristic of sodium sulphate, for instance. This particular characteristic was from the first rightly attributed to the silica content of the slag, rather than the lime or alumina content; for, while both lime and alumina combine with a number of different acids to form solids which upon melting exhibit no unusual degree of viscosity, the presence of silica in a chemical compound usually confers upon it a high viscosity in the liquid state. This peculiar property appears to be due to the nature of the molecule SiO2 rather than to the element silicon itself; for, as is well known, SiCl4 is a volatile colourless liquid, which boils at 59° C., while SiF4 is a gaseous substance, which has a boiling-point of -100° C.

In 1913 Dr. Laue, of the University of Zurich, conceived the idea of employing a crystal as a "space diffraction grating" for X-rays. This epoch-making discovery, in the hands of Messrs. W. H. and W. L. Bragg, has yielded extremely interesting facts concerning the structure of crystalline solids. To quote from the latter investigators, "The architecture of crystals has been laid open to examination; crystallography is no longer obliged to build only on the external forms of crystals, but on the much firmer basis of an exact knowledge of the arrangement of the atoms within." Hitherto the chemical molecule had been supposed to exist as such in the solid state. The X-ray spectrometer, however, has shown clearly that in the case of most crystals each atom is arranged in an ordered manner at definite points of a "space-lattice," and that in the case of a crystal of potassium chloride, for instance, there is no such thing as a molecule of KCI in the usual sense of the word, but that each potassium atom is equidistant from six chlorine atoms, while each chlorine atom is equidistant from six potassium atoms. In other words, the valence of each potassium atom and chlorine atom is divided between at least six complementary atoms. When W. H. and W. L. Bragg took up the examination of quartz (25), i.e. SiO2, by means of this method, they found that it presented a structure more complicated than that of any substance which they had at that time investigated.

Instead of finding that silicon and oxygen atoms were arranged separately at definite points of a space-lattice, Messrs. Bragg concluded that three silica molecules were associated with each point of the space-lattice. It is a matter of common knowledge that highly associated or polymerised liquids possess unusually high viscosity; and hence it seems plausible to argue that, since silica appears to be unusually complex in the solid state, in the liquid state this association or polymerisation tendency must be the fundamental cause of the extreme viscosity of silica itself and of the high viscosity of the silica compounds. This explanation, as based on X-ray analysis, does not seem to have been brought forward hitherto in dealing with the cause of the high viscosity of silicates.

Irving Langmuir (26), in a valuable contribution to the

theory of the constitution of solids and liquids, has very recently reviewed the work of the Braggs on crystal structure from the viewpoint of the chemist and the relation of this work to theories of chemical constitution such as those of Werner, Stark, J. J. Thomson, and Lewes. Langmuir considers the complex silicates as apparently built up of compounds of the first order, i.e., atomic groups held together by "primary" valence, which are in turn held together by secondary or residual valence, in much the same way that the metallic compounds are built up of the atoms of the metals. He further states (27): "From the fact that glasses, when heated, change to the liquid state by a continuous process, we are led to conclude that the structure of a liquid does not differ from that of a glass in any essential respect except that in the liquid there is a certain mobility (entirely analogous to tautomerism). Therefore, even in a liquid we look upon every atom (or group-molecule) as combined chemically (or adsorbed) to all the adjacent ones. The molecular weight is therefore a term that has very little significance in the case of a liquid." When, therefore, a silicate gradually softens with rising temperature and passes entirely over into the liquid state, it is probable that the increased fluidity is due to a weakening of the residual-valence attraction between the group-molecules, whereas the relatively high viscosity of the melt, as compared with that of molten metals and ordinary salts, is due to the preponderance of the groupmolecules of silica, alumina, and lime, and possibly to a particularly large degree upon a highly polymerised condition of the silica group-molecule.

Slag Viscosity as related to Fuel Economy. While it is theoretically possible to render any silicate mixture whatsoever sufficiently fluid to flow from the cinder-notch of a blast-furnace, it is necessary in practice that a slag attain this necessary fluidity at a temperature which is not beyond the working limit of the blast-furnace lining and which does not demand an unusually high fuel consumption.

It is obvious that if, for instance, a slag requires a minimum temperature of 1400° C. in order to attain a working fluidity, no iron will be produced in a furnace using this slag, regardless of the number of B.T. units developed within the furnace, unless the temperaturedistribution is such that the slag acquires the necessary temperature of 1400° C. Thus the fuel economy of a blastfurnace is to a great extent dependent upon the temperatureviscosity relations of the slag.

The maximum temperature to which it is theoretically possible to heat the slag, assuming an absence of heat loss by conduction or radiation, is the theoretical combustion temperature of the exothermic reaction which occurs near the tuyeres, CO CO, taking into account the fact that the oxygen represented in the equation represents in reality ordinary air. The old style heat-balance of Sir Lowthian Bell considered only quantity of heat and not its intensity, i.e., temperature. According to this early method of calculating tuel economy, five B.T.U. produced in the bosh had a definite significance, regardless of the temperature at which they were produced. It is evident, on the contrary, that any factor which operates so as to increase the combustion temperature in the region of the tuyeres will exert a relatively large effect so far as fuel economy is concerned when compared with a factor wbich increases the total heat in the furnace without appreciably affecting the temperature of the tuyere region.

When James Gayley (28) published the results of his experiments with the dry blast at the Isabella furnaces near Pittsburgh, it became evident that the gain in fuel economy was greater than that simply represented by the fraction of the total heat developed in the furnace which was contributed by the drying process. Many furnacemen openly questioned the results of the tests. Ledebur, Bell, and Le Chatelier offered in turn unsatisfactory explanations. A. Lodin (29), however, appears to have been the first to

publish an explanation which went to the root of the matter. He says, in part:

"La fusion du laitier par exemple nécessite un certain nombre de calories qui doivent être fournées au-dessus d'une température minimum, celle du point de fusion. Ces calories seront empruntées aux produits gazeux de la combustion de carbone: la seule partie du pouvoir calorifique de celui-ci utilisable pour cet usuage détermine sera celle correspondant au refroidissement de produits gczeux depuis la température de combustion jusqu'au point de fusion du laitier."

In September, 1904, J. E. Johnson presented a paper before the American Institute of Mining Engineers (30), entitled "Notes on the Physical Action of the Blast furnace," in which was brought forward the idea that the fuel economy of the blast-furnace depends upon the "available heat" furnished above the "critical temperature," this temperature corresponding with the free-flowing temperature of the slag. Johnson's theory differs from that of Lodin in that the former takes the free-flowing temperature of the slag as the "critical" temperature, while the latter assumes it to be the temperature of the fusion slag. Jahnson (31) has recently further expounded his theory as substantiated by the results of blast-furnace practice. He does not in any case point out clearly the physical signifi cance of the term "available heat" for any given "critical" temperature, although the detailed calculations are given. The author has shown elsewhere (32) that the "available" hearth heat of Johnson at a critical temperature T is equal to the heat of the isothermal reaction C+ O CO at a temperature T, minus the heat required to raise the temperature of the blast up to a temperature T, together with that required to decompose the moisture of the blast; and that this available hearth heat includes (a) the heat lost by radiation and conduction between those two zones in the furnace which are at the critical temperature, and (b) the heat required for those final steps of reduction of ore and carburisation] of iron, and other similar adjustments in composition of molten iron, molten slag, and furnace gases, which occur between the "critical zones."

It is evident therefore that the viscosity-temperature relations of the slag do more than simply determine the facility with which the slag is handled at flush and the extent of desulphurisation accomplished within the hearth; these relations are fundamental in determining the fuel economy which can be realised in furnace opera tion.

The Work of the Bureau of Mines.

The Bureau of Mines is investigating the problem of slag viscosity, its variation with the temperature and with the composition of the slag, and its effect upon the distribution of the sulphur between molten iron and slag. These investigations are being conducted in the laboratories of the Pittsburgh station of the Bureau of Mines, and will represent, when published, an introduction to a series of contributions to the theory of the metallurgical

processes.

Acknowledgments.

In reporting the results of the experimental work described in the present article, the author wishes to make acknowledgment of the active interest of Mr. F. H. Willcox, metallurgical engineer, at whose suggestion the work on slag viscosity was undertaken. Aoknowledgment is also made of the encouragement and support received from Mr. Van H. Manning, director of the Bureau, at whose authorisation the research was initiated; and of the active co-operation and interest of Dr. F. G. Cottrell, chief metallurgist, and Mr. D. A. Lyon, metallurgist, during the progress of the investigations. The chemical analyses of slags were made by Mr, F. D. Osgood, junior chemist, under the direction of Mr. A. C. Fieldner, chemist.

Appreciation is also due to the numerous iron and steel companies from whom slag samples were obtained; and to the Jones and Laughlin Steel Company and the Clinton

Iron and Steel Company of Pittsburgh in particular for their courtesy in permitting measurements to be made at the furnace."

The Application of Viscosity Data to Metallurgical Operations.

Before taking up the description of the viscosity apparatus, furnace, and accessories, and tabulation of the experimental results obtained, it is advisable to consider briefly the application of accurate viscosity determinatious on metallurgical slags to manufacturing processes, and also to related research in this field. Particular attention must be given in the present paper to a consideration of the principles of the metallurgy of iron, although in many other metallurgical operations, such as the smelting of copper, for instance, a knowledge of the temperatureviscosity relations of different types of slag is of great importance.

Apart from the question of mining cost and freightage, the value of an iron ore sufficiently rich in iron to be considered marketable is largely dependent upon whether it can be made to yield economically a slag of desirable viscosity and desulphurising power. A casual glance at the table of slag analyses given elsewhere in this paper shows at once the comparatively wide range of slag composition that has been found practicable by different manufacturers. In each case the particular slag composition was undoubtedly determined in a large measure by the composition of the ore mixture and fuel which it was deemed expedient to use, and also by the grade of iron produced. However, it is entirely probable that in certain cases the slag composition was not the optimum one from the standpoint of economy and excellence of product. It is one of the purposes of these investigations to determine what are the optimum conditions.

The experience of furnacemen with practically the same operating conditions differs quite widely in many instances. In fact, each furnace seems to have its own peculiarities. Successful operation is realised by a careful study of the past records of operation and production. Radical changes in operating methods are sometimes made, with results which may or may not be beneficial, and which cannot be predicted beforehand with certainty. This is, in a few words, the general situation in regard to iron metallurgy. Of course there are furnacemen, with a ready gift of intuition, to whom nothing is impossible; but these men are rare, and impart their knowledge with difficulty because of the fact that it is intuitive.

In spite of the lack of scientific research on the physical and chemical properties of slags at high temperature, it is quite well understood what functions the slag must perform in the blast-furnace. In the first place, it must be sufficiently fluid to flow from the cinder-notch at the temperature which exists in the hearth. In the case of charcoal practice, where desulphurisation is a minor item, the viscosity of the slag at flush is a primary consideration. If, for instance, it is found that a silicious charcoal slag possesses a viscosity at 800 at the temperature of the hearth, the question arises whether a limy slag would perform its functions properly if it possessed an equal viscosity. The limy slag may be prevented from performing these functions for one or both of the following reasons. In the first place, the limy slag with a viscosity of 800 might have an extremely high rate of change of viscosity with temperature, i.e., this particular point on the temperature-viscosity curve might occur at a temperature where the slag underwent rapid softening or hardening with small changes of temperature. In the second place, the limy slag might not be at a sufficiently high temperature to properly desulphurise the pig iron.

Notes.

1. Henry Le Chatelier, "Le Silice et les Silicates," Paris, 1914.

2. See footnote p. 108 "Mésure des Températures Elévées," by II. Le Chatelier and O. Boudouard,

Paris, 1900; also p. 6, Publication No. 31, Carnegie
Institute of Washington, 1905.

3. Hermann August Seger, Tonindustrie Zeitung, 1886, P. 135.

4. H. L. Callendar, "On the Practical Measurement of Temperature," Proc. Roy. Soc., 1886, xli., 231; for complete description and bibliography of platinum resistance thermometry see Bull. Bureau of Standards, vi., 149, Platinum Resistance Thermometry at High Temperatures," by C. W. Waidner and G. K. Burgess.

5. J. Joly, Proc. Roy. Irish Acad., 1891, iii., [2], 38. 6. H. Le Chatelier, "Pyrometric Optique," Comptes Rendus, 1892, cxiv., 214; also Journ. de Phys., 1892, [3], i.

7. E. Becquerel, "Mésure Optique des Temperatures," Comptes Rendus, 1863, lv., 826.

.

8. C. W. Waidner and G. K. Burgess, “Optical Pyrometry," Bull. Bureau of Standards, i., 189. This paper contains complete description of construction and method of calibration of all types of optical pyrometers, together with discussion of laws of radiation governing their operation.

9. Wanner, Phys. Z. S., 1902, iii., 12; Iron Age, Feb. 18, 1904, p. 24; Stahl u. Eisen, 1902, xxii., 207. 10. Féry, Journ. de Phys.. 1904, [4], iii., 32. 11. Holborn and Kurlbaum, Ber. d. K. Akad. d. Wiss., Berlin, 1901, p. 712; Ann. d. Phys., 1903, X., 225. 12. Morse, American Machinist, 1903; U.S. Patent of 1902, 696878, 696916.

13. Loc. cit., p. 242.

14. Féry, Comptes Rendus, 1902, cxxxiv., 977: See article by Burgess and Foote on "Characteristics of Radiation Pyrometers," Bull. Bureau of Standards, xii., 91. 15. Loc. cit.

16. C. Doelter, Tschermak Min. u. Petr. Mitth., 1901, xx., 210; 1902, xxi., 23; Sitzungsber. d. Wien. Akad., May, 1906, cxv., t. See also his monumental work, "Handbuch der Mineralchemie," 1912, i.; 1914, ii., Part 1.

17. A. L. Day and E. T. Allen, Publication No. 41, Carnegie Institution of Washington, 1905.

16

18. Sir Lowthian Bell, Principles of the Manufacture of

Iron and Steel," 1884.

19. Akermann, "Om Varmbehofven for Olika Masugneslaggers Smaltning," Stockholm, 1886. Translated in Stahl und Eisen, 1886, vi., 281. See also H. M. Howe," Use of Tri-axial Diagram and Triangular Pyramid for Graphical Illustration," Trans. Am. Inst. Min. Eug., 1899, xxviii., 346.

FROM the values of (F/c)o we can calculate λ the loss in potential energy when a grm. molecule of the solute passes from the interior to the surface of the solution. We have previously shown from Szyszkowski's data how the increase in λ for each CH2 could be determined. To estimate the approximate absolute value of a we may proceed as follows:

It is shown in works on the kinetic theory (for example Jean's, "Dynamical Theory of Gases, p. 78) that for a state of equilibrium the distribution of a gas between two regions in which the potential energy is different, is given by the relation

C c'

-X/RT

(24).

Here c and c are the respective concentrations in the two regions and λ is the difference in potential energy per grm. molecule (The above equation is closely related to Boltzmann's conception, according to which entropy is equal to the logarithm of the probability. The ratio between the probability that a molecule will be in the interior and the probability that it will be in the surface is thus proportional to e- where is the entropy. The quantity A/RT corresponds to the entropy).

For solutions so dilute that Fc is constant, we may expect Equation 24 to be applicable, and we may thus use it to calculate the concentration c in the surface layer in terms of λ or vice versa.

The equation may also be derived from purely thermodynamical principles.

If A is expressed in calories, Equation 24 may be written

Determination of Mineral and Rock Densities at High Temperatures," Am. Fourn. Sci., 1914, xxxvii., I.

21. C. Barus, U.S. Geol. Survey, 1893, Bull. 103, p. 25. 22. C. Doelter and Sirk, Sitzb. Wien. Akad., 1911, cxx., [1].659.

23. W. P. White, "Specific Heats of Silicates and Platinum," Am. Journ. Sci., 1909, xxviii., 334. 24. C. Doelter, Sitzb. Wien. Akad., 1908, cxvii., 862; 1910, cxix., 73.

25. W. H. and W. L. Bragg, "X-Rays and Crystal Structure," p. 160 (London, 1915).

26. Irving Langmuir, "The Constitution and Funda

mental Properties of Solids and Liquids" (Part I., Solids), Journ. Am. Chem. Soc., November, 1916, xxxviii., 2221; CHEMICAL NEWS, cxvi., 42 et seq. 27. Loc. cit., p. 2244, lines 31-39. 28. James Gayley, "Application of Dry Air Blast to Manufacture of Iron," Trans. Am. Inst. Min. Eng., 1904, xxxiv., 746.

29. A. Lodin, Comptes Rendus, cxxxix., 922, Nov., 1914. 30. J. E. Johnson, jun., "Notes on the Physical Action of the Blast Furnace," Trans. Am. Inst. Min. Eng., 1906, xxxvi., 454.

to Nernst's equation of the electromotive force of rcversible cells.

The amount of solute q absorbed in the surface layer per sq. cm. may be roughly taken to be—

T(C1 - C)

q =

(26), where is the thickness of the surface layer in which adsorption takes place.

Combining Gibbs' Equation I with (12) we obtain

This equation was used for calculating the values of Aobs. given in Table III. The choice of the particular value of r to use in (28) is somewhat arbitrary, but since the molecules adsorbed in the surfaces of these dilute solutions lie flat in the surface it is probable that is of the same order of magnitude as the values found for ricinoleic acid or triricinolein, namely, 4'7 x 10-8.

The last column of Table III. gives the values of A calculated from B by an equation similar to (9) except that the numerical constant was 12.5 instead of 12.8, since the temperature in Traube's experiments were 15° instead of 20°.

It is seen that Aobs. increases on the average by about 625 calories for each CH2 added to the molecule. This means that can be expressed in general by an equation of the form

cal. = λο + 625 n. . .. (30), where n is the number of carbon atoms in the molecule. The quantity Ao has a different value for each type of substance, as is shown in Table IV.

When a double bond is present in the molecule the value of A is decreased by about 400. The addition of one or more hydroxyl groups to a mono- or dibasic acid decreases λo by about 800 for each hydroxyl group.

The fifth column of Table III. gives values of Acal, which have been calculated by Equation 30, using the values of Ao given in Table IV.

An examination of Table IV. sho'vs that Ao becomes smaller as the active groups in the molecule become more polar in character. It also shows that the polar character is not additive. Two active groups attached to adjacent carbon atoms, as in glycol, caused a very great decrease in o. Such effects, with which chemists are familiar, are undoubtedly due to forces transmitted from atom to atom in the group molecule. It is probable that these forces are caused by a displacement in the relative positions of the electrons and positive nuclei. Thus if an oxygen atom is combined with a carbon atom at one end of a hydrocarbon chain the electrons in the carbon atom are probably displaced towards the oxygen atom and the positive nucleus of the carbon atom displaced away from the oxygen atom. This displacement causes a similar but smaller displacement of the electrons of the next carbon atom and so on. (Effects of this kind are probably of importance not only in organic chemistry, but in the study of the structure of the surfaces of crystals and of liquids. In some cases such phenomena may cause adsorbed layers to be more than one group molecule in thickness).

The close agreement between the observed and calculated values of A in Table III. shows that, except in the case of active groups in close proximity within the molecule, the change in potential energy a between the interior and surface of the liquid is an additive property. The results for as in Table III. are in substantial agreement with the results previously discussed. For the saturated acids, alcohols, and esters the values are approximately the same (30x 10-16 sq. cm.) and are independent of the number of carbon atoms.

The presence of double bonds (allyl alcohol and acetate) does not seem to cause any increase in as, probably because these are forced away from the water before the surface becomes saturated. With oxy-butyric acid as is much greater (48 × 10-16), indicating that the

hydroxyl group is in contact with the water even when the surface is saturated. The large value for acetone (425) and the small value for butyl aldehyde (248) are noteworthy, but until verified by other data it would hardly be safe to draw conclusions from these differences. There are many other data available from which the arrangements of group molecules in surface layers are determinable. Some of these will be briefly mentioned. Morgan and Egloff (Journ. Am. Chem. Soc., 1916, xxxviii., 844) give the surface tensions of solutions of phenol and water at three temperatures. From these data by Equation I the amounts of phenol adsorbed per sq. cm. (9) may be calculated.

It is found that with increasing concentration q increases, rapidly at first, then more slowly until it reaches a maximum of about 48 x 10-11 grm. molecules per sq. cm., showing that the surface becomes saturated with phenol molecules. This maximum value of q is approximately the same at all three temperatures (0°, 35°, and 65°), but at the higher temperatures it requires a greater concentration of phenol in the solution to give a saturated surface than at lower temperatures. This fact is a natural result of the kinetic agitation which tends to equalise the concentration in the surface and in the solution. From the above value of q it can be readily calculated that the area per molecule of adsorbed phenol is a。 = 34 × 1016 and the thickness of the film is 4'3 x 10-8 cm. These results would seem to indicate that the phenol molecules in the surface of an aqueous solution lie flat on the surface, and that the diameter of the disk-shaped (assumed) molecule is about one and a half times its thickness.

T

It is probable from these results that the three ethyl groups lie spread out upon the surface while the nitrogen atom is below the surface and is surrounded by water molecules combined with it by secondary valence.

Worley (Journ. Chem. Soc., 1914, cv., 260) gives data for aqueous solutions of aniline at several temperatures from 15° to 75°. The values of 9 calculated from (1) show a variation quite similar to that found from the data on phenol-water solutions. The surface becomes saturated more easily at low than at high temperatures. The data are apparently not as accurate as those of Morgan and Egloff, but they indicate a fairly constant saturation value of qs = 45 X 10-11. This corresponds to

ao = 37 X 10-16 sq. cm.

T= 40 X 10-8 cm.

Evidently the arrangement of the aniline molecules in the saturated surface is about the same as that of the phenol molecules.

We have thus far considered cases where the solute is adsorbed in the surface layer, so that the surface tension is less than that of the pure solvent. When inorganic salts are dissolved in water or alcohol the surface tension increases. The surface of the solution thus contains an excess of the solvent. In general we should expect the surface of any liquid to consist of molecules or atoms arranged in a rather definite manner different from that in the interior. If the molecules or atoms of the dissolved substance are surrounded by fields of force strong compared to those of the solvent, then it is improbable that the solute molecules will be able to displace solvent molecules in the surface. Hence the surface layer should consist of a single layer of molecules of the solvent from which solute molecules are excluded.

From Gibbs' Equation I we see that if the surface tension increases linearly with the concentration, there is a deficiency of the solute in the surface layer which is pro