STATE OF ILLINOIS HENRY HORNER, Governor DEPARTMENT OF REGISTRATION AND EDUCATION JOHN J. HALLIHAN. Director DIVISION OF THE STATE GEOLOGICAL SURVEY M. M. LEIGHTON, Chief URBANA REPORT OF INVESTIGATIONS — NO. 68 EFFECT OF FLUORSPAR ON SILICATE MELTS WITH SPECIAL REFERENCE TO MINERAL WOOL J. S. Machin and J. F. Vanecek PRINTED BY AUTHORITY OF THE STATE OF ILLINOIS URBANA, ILLINOIS 1940 ILLINOIS GEOLOGICAL SURVEY LIBRARY STATE OF ILLINOIS HON. HENRY HORNER, Governor DEPARTMENT OF REGISTRATION AND EDUCATION HON. JOHN J. HALLIHAN, Director BOARD OF NATURAL RESOURCES AND CONSERVATION HON. JOHN J. HALLIHAN, Chairman EDSON S. BASTIN. Ph.D., Geology HENRY C. COWLES. Ph.D., D.Sc., Forestry. WILLL\M A. NOYES, Ph.D.. LL.D., Chem.D., D.Sc, (Deceased) Chemistry ARTHUR CUTTS WILLARD, D.Engr., LL D, LOUIS R. HOWSON, C.E., Engineering President of the University of Illinois WILLIAM TRELEASE, D.Sc, LL.D., Biology STATE GEOLOGICAL SURVEY DIVISION JJrbana M. M. LEIGHTON, Ph.D., Chief ENID TOWNLEY, M.S., Assistant to the Chief JANE TITCOMB. M.A., Geological Assistant GEOLOGICAL RESOURCES Coal G. H. CADY, Ph.D., Senior Geologist and Head L. C. McCABE, Ph.D., Assoc. Geologist JAMES M. SCHOPF, Ph.D., Asst. Geologist J. NORMAN PAYNE, Ph.D., Asst. Geologist CHARLES C. BOLEY, M.S., Asst. Mining Eng. Industrial Minerals J. E. LAMAR, B.S., Geologist and Head H. B. WILLMAN, Ph.D., Assoc. Geologist DOUGLAS F. STEVENS, M.E., Research Associate ROBERT M. GROGAN, Ph.D., Asst. Geologist ROBERT R. REYNOLDS, B.S.. Research Assistant Oil and Gas A. H. BELL, Ph.D., Geologist and Head G. V. COHEE, Ph.D., Asst. Geologist FREDERICK SQUIRES, B.S., Assoc. Petr. Eng. CHARLES W. CARTER, Ph.D., Asst. Geologist WILLIAM H. EASTON, Ph.D., Asst. Geologist ROY B. RALSTON, B.A., Research Assistant WAYNE F. MEENTS, Research Assistant Areal and Engineering Geology GEORGE E. EKBLAW, Ph.D., Geologist and Head RICHARD F. FISHER, B.A., Research Assistant GEOCHEMISTRY FRANK H. REED, Ph.D., Chief Chemist W. F. BRADLEY, Ph.D., Assoc. Chemist G. C. FINGER, Ph.D., Assoc. Chemist ROBERTA M. LANGENSTEIN, B.S., Research Assistant Fuels G. R. YOHE, Ph.D., Assoc. Chemist in Charge CARL HARMAN, M.S., Research Assistant Industrial Minerals J. S. MACHIN. Ph.D., Chemist and Head JAMES F. VANECEK, M.S., Research Assistant Analytical O. W. REES, Ph.D., Chemist and Head L. D. McVICKER, B.S., Asst. Chemist GEORGE W. LAND, M.S., Research Assistant P. W. HENLINE, M.S., Research Assistant MATHEW KALINOWSKI, M.S., Research Assistant ARNOLD J. VERAGUTH, M.S., Research Assistant WILLIAM F. WAGNER, M.S., Research Assistant MINERAL ECONOMICS Subsurface Geology L. E. WORKMAN, M.S., Geologist and Head ELWOOD ATHERTON, Ph.D., Asst. Geologist MERLYN B. BUHLE, M.S., Asst. Geologist I. T. SCHWADE, M.S., Asst. Geologist FRANK E. TIPPIE, B.S., Research Assistant Stratigraphy and Paleontology J. MARVIN WELLER, Ph.D., Geologist and Head CHALMER L. COOPER, M.S., Assoc. Geologist Petrography RALPH E. GRIM, Ph.D., Petrographer RICHARDS A. ROWLAND, Ph.D., Asst. Geologist Physics R. J. PIERSOL, Ph.D., Physicist DONALD O. HOLLAND, M.S., Asst. Physicist PAUL F. ELARDE, B.S., Research Assistant W. H. VOSKUIL, Ph.D.. Mineral Economist GRACE N. OLIVER, A.B., Assistant in Mineral Eco- nomics EDUCATIONAL EXTENSION DON L. CARROLL, B.S., Assoc Geologist PUBLICATIONS AND RECORDS GEORGE E. EKBLAW, Ph.D., Geologic Editor CHALMER L. COOPER, M.S., Geologic Editor DOROTHY ROSE, B.S., Technical Editor KATHRYN K. DEDMAN, M.A., Asst. Technical Editor ALMA R. SWEENY, A.B., Technical Files Clerk FRANCES HARPER LEHDE, A.M., Asst. Technical Files Clerk MEREDITH M. CALKINS, Geologic Draftsman LESLIE D. VAUGHAN, Asst. Photographer DOLORES C. THOMAS, B.A., Geologic Clerk Consultants: Ceramics, CULLEN W. PARMELEE, M.S., D.Sc, and RALPH K. HURSH, B.S., University of Illinois: Pleistocene Invertebrate Paleontology, FRANK COLLINS BAKER, B.S., University of Illinois. Topographic Mapping in Cooperation with the United States Geological Survey. This Report is a Contribution of the Industrial Minerals Division of the Geochemistry Section. ,2 ILLINOIS STATE GEOLOGICAL SURVEY 3 3051 00005 7194 (A30846— IM— 9-40) September 1, 194C CONTENTS Page Introduction • 5 Fluorspar in silicate melts 5 Summary of literature 5 Experimental part 6 Apparatus and procedure 7 Reproducibility of results 7 Discussion of data . . .• 10 A mechanism for the effect of fluorspar on the viscosity of sihcate melts / 11 Some quantitative aspects of the network theory / . . . . 14 Summary 14 Bibliography 15 TABLES Page 1. Fiber and shot diameters 8 2. Average fiber diameter of rock wool containing 35 per cent Si02 and varying amounts of fluorspar and blown at temperatures of 1400 and 1500 °C. (see figure 1) 8 3. Average fiber diameter of rock wool containing 40 per cent Si02 and varying amounts of fluorspar and blown at temperatures of 1400 and 1500 °C, (see figure 2) 9 4. Average fiber diameter of rock wool containing 50 per cent Si02 and varying amounts of fluorspar and blown at temperatures of 1400 and 1500 °C. (see figure 3) 9 ILLUSTRATIONS Figure Page 1. Relationship between fiber diameter of rock wool and fluorspar content of the melt (see table 2) . . . . 8 2. Relationship between fiber diameter of rock wool and fluorspar content of the melt (see table 3) . . . . 9 3. Relationship between fiber diameter of rock wool and fluorspar content of the melt (see table 4) . . . . 10 4. Schematic diagram showing efi^ect of fluorspar additions on viscosity of slags: curve 1, no fluorspar- curves 2, 3, 4, and 5, increasing quantities of fluorspar 11 5. Hypothetical oxide A2O3, illustrating arrangement of atoms in crystalline modification (A) and in glassy modification (B) 12 6. Hypothetical glass networks, illustrating the effect of metallic fluorides in weakening such networks. 13 [3] Digitized by the Internet Arcliive in 2012 with funding from University of Illinois Urbana-Champaign http://archive.org/details/effectoffluorspa68mach EFFECT OF FLUORSPAR ON SILICATE MELTS WITH SPECIAL REFERENCE TO MINERAL WOOL J. S. Machin AND J. F. Vanecek INTRODUCTION THE QUESTION of the influence of fluor- spar on rock or slag wool melts has commanded the occasional interest of both producers of mineral wool and producers of fluorspar. The origin of interest in the material on the part of the mineral wool manufacturers is hard to trace but probably it arose from the fact that certain slags in which fluorspar was used have turned out to be good sources of raw material for the manufacture of slag wool. The attitude most frequently encountered toward the question of the value of fluorspar in min- eral wool cupola operation is one of doubt that it has any value except as a flux for use in improving melting conditions within the cupola. However the well known fact that it does have a fluxing or thinning action on the melt naturally suggests its use when troubles due to improper or incomplete fusion of the cupola charge are encountered. FLUORSPAR IN SILICATE MELTS Summary of Literature The use of fluorspar as a flux in basic open hearth slags dates back nearly fifty years. The suggestion of its use as an aid in the removal of sulfur from steel in the open hearth process is commonly ascribed to Saniter. The use of the mineral as a flux in foundry cupola slags and blast furnace slags is also fairly common although its corrosive action on the refractory linings has dis- couraged wide use of the material in such apparatus. Carsten^ says it is very useful in cupola operation when the quality of raw material is low. He indicates that its use is increasing in England due to the high pro- duction rates demanded by war. He says that excessive corrosion of refractory parts of the cupola can be controlled by proper proportioning of the fluorspar with the other components of the slag. There is a fairly voluminous literature that deals with the use of fluorspar in slags and that was abstracted and correlated by Schwerin-* in 1934. Since that time the subject has been further dealt with by Mat- sukawa^ by Rait, M'Millan and Hay^ by Lawrie^, by Hartmann*^- ^, and by EndelP. EiteP made a phase rule study on the effect of fluorspar as a mineralizer in cement clinker. Except for the work of Eitel these reports in cases where experimental work was done have dealt mainly with viscosity. There appears to be general agreement that the addition of small amounts of fluorspar (up to 10 per cent and in some situations much more) results in reduction of viscosity. In general when the viscosity of a molten slag is plotted as a function of the temperature, the shape of the curve obtained is roughly similar to that of the rectangular hyperbola. If additions of fluorspar are made to a given slag, the effect is to shift the curve along the axis in the direction of decreasing tempera- ture (see fig. 4). At least this shift is nearly always observed when the quantity of fluor- spar added does not exceed 8 to 10 per cent. (See data of Schwerin^*^, of Matsukawa^, of Hartmann'^', and of Herty^^). 'Superior numbers refer to bibliographj', p. 15. [51 FLUORSPAR IN SILICATE MELTS Various theories have been suggested to provide a mechanism to explain the decrease in the viscosity of molten slags which takes place when fluorspar is added. The theory most commonly encountered is due to Ham- ilton^-. According to this, fluorspar pro- motes the interaction of lime and silica to form calcium silicate with or without the help of an intermediate compound SiF^. The following equations illustrate the as- sumed type of reaction although the theory recognizes that the complete picture is in all likelihood more complex. (1) 3 SiO, + 2 CaF, -^ SiF4 + 2 CaSiOa (2) 3 CaO -f SiF4 -^ 2 CaFi + CaSiOs Against the "catalysis" hypothesis is the fact that synthetic slags which have been melted, crushed, and remelted two or more times (Herty, Schwerin, op. cit.) and in which presumably the interaction of lime and silica has already largely taken place, are still much reduced in viscosity after additions of fluorspar. Sisco^^ analyzed high-lime slags by a method which differentiated between "true silica" and "total silica." His "total silica" was always greater than his "true silica." He suggested that the difference between total silica and true silica was silica present as silicofluoride. He thought that the silico- fluoride might have been formed as repre- sented by the equation : (3) 3 CaF, + 3 SiOo -^ CaSiFe + 2 CaSiO, Granting the presence of silicofluoride in the slag, however, merely changes the prob- lem without offering a satisfying explana- tion of the effect of the fluorspar on the slag viscosity. Possibly one reason for the many attempts to explain the thinning action of fluorspar on molten slags by a catalytic mechanism is to be found in the frequently published statement that all or at least a large part of the fluorine is lost from the slag. Mat- sukawa^^ says that there seems to be general agreement that one reaction which occurs on addition of fluorspar to a basic open- hearth slag is represented by : (4) 2 CaF2 -f- SiOa -^ 2 CaO + SiF4 This, he says, was proved by experimental evidence adduced by various workers, to whose writings he refers, and confirmed by analysis and calculations on slags used in his own viscositv measurements. While there is no reason to doubt the reliability of any of this work except insofar as it depends in some instances on the unsatisfactory methods for determination of fluorine in the presence of silica which were available for some of the earlier work, it should be pointed out that the loss of fluorine as SiF^ is usually incomplete and slow. On this point Matsukawa concludes that in the open- hearth slags with which he worked, 15 to 65 per cent of the CaFo added is left in the slag. His data indicate that when larger amounts of CaF^ are added, larger percentages of the total fluorine are lost by volatilization, presumably as silicon tetrafluoride, and vice versa. Rait et al* found no variation with time up to 8 hours in the viscosity of blast fur- nace slags to which 3 per cent of fluorspar had been added. They concluded that there- fore the reaction represented by equation (4) had not taken place to any appreciable extent, since this would have caused an increase in the relative CaO content of the melt which would have produced an alter- ation in the viscosity at constant tempera- ture. The possibility that the thinning action of fluorspar on silicate melts may be due to the formation of low-melting eutectics must not be overlooked. Karandief^^ studied the phase relationships in the system CaF^- CaSiOg. He found an eutectic which melted at about 1130°. No published information bearing direct- ly on the effect of fluorspar on the mineral wool making process was found. EXPERIMENTAL PART This investigation was undertaken for the purpose of studying the effects produced by fluorspar upon the character of the fiber of mineral wool that contained small amounts of this mineral. The experimental procedure consisted, in brief, of preparation of syn- thetic mixtures of lime, magnesia, alumina, and silica, to which the fluorspar was added ; the mixtures were then melted and blown into mineral wool and the wool was studied. The materials used were the purest ob- tainable in quantity. The silica was ground quartz sand. The alumina was in the form of precipitated aluminum hydrate. The lime and magnesia were in the form of precipi- tated carbonates. All of these materials were WITH REFERENCE TO MINERAL WOOL analyzed. The total amount of impurities exclusive of ignition loss and lime, magnesia, alumina, and silica was in no case more than 0,2 per cent. The materials were weighed out in quan- tities calculated to give batches with the desired composition on the calcined basis and were thoroughly mixed by hand. This pro- cedure does not of course take into account variation with time in the free moisture content of the components of the batch. P'or this reason the figures given for the consti- tution of the various batches are to this extent approximate. However, since all of the materials in the batches of a given series of tests using the same composition were weighed out and mixed at one time, any error due this factor would apply equally to all members of that series. The mixed batch was then calcined at about 1000° C. The principal purpose of this calcination was to increase the useful life of the graphite crucibles used in melting. A secondary purpose was to prevent foam- ing of the molten charge which might other- wise take place under some conditions. The calcinate was weighed out in one-kilogram batches which were sealed in glass fruit jars to prevent reabsorption of moisture and carbon dioxide until such time as they could be melted and blown. The fluorspar (plus 98 per cent CaFo) used was the best acid grade. The fluorspar additions were made to the calcined batch just prior to melting. Care was taken to mix the spar thoroughly with the batch although this was probably a needless pre- caution since electrically conducting melts in induction furnaces are automatically stirred. Apparatus and Procedure The apparatus and testing procedure was in general similar to that used in earlier work in this laboratory (Bull. 61)^^ with the exception that a steam nozzle with a V-shaped orifice was used because it gave much better yields of wool. This orifice used more steam so that the pressure drop across the nozzle was in general 65 pounds ± 5 pounds, instead of 70 pounds as specified. One-kilogram charges were melted in an induction furnace, heated to 1500° C. and held at that temperature for a fifteen-minute fining period. If it was desired to blow at 1500° the charge was blown at the end of this fifteen-minute period. If it was desired to blow at a lower temperature the charge was allowed to drop to the lower figure and held there for ten or fifteen minutes or until it was certain that the temperature had become constant before blowing. All tem- perature measurements were made with plat- inum platinum-rhodium thermocouples and a Leeds and Northrup portable potenti- ometer No. 8659. The thermocouples were checked at intervals against a Bureau of Standards calibrated couple. After blowing, the wool was collected and weighed. The fiber and shot diameters were measured with the aid of a microscope with an eyepiece micrometer. This measure- ment was carried out as follows. Five small samples of wool were removed from the different parts of the mass sample and mounted on slides under cover glasses. Fibers were measured in each sample by focusing first on the top fibers and measuring all fibers in sharp focus which cross a certain limited area of the field of view. Without moving the slide the tube was then lowered so that a new collection of fibers came into sharp focus and the process was continued until fifteen to twenty fibers were measured. This process was repeated with each of the {\ve small samples. The result was that about eighty-five fibers were measured from each mass sample. The average of these measurements was recorded as the fiber di- ameter of the mass sample. The shot were measured in the same samples but the results of shot measurements of this kind are of doubtful value because the larger shot will fall out of a wool sample small enough to be mounted on a slide. Reproducibility of Results In order to test the reproducibility of work of this kind a series of eight charges were melted and blown. The composition of these charges was 40 per cent SiOo, 30 per cent CaO, 20 per cent AI2O3, and 10 per cent MgO. The results of the tests are shown in table 1. These were divided into two groups (A and B) in order to pro- vide a check on the efficiency of the mixing methods. The materials for tests SI 2-6 to SI 2-9 inclusive were mixed in one large batch, calcined and then four one-kilogram batches of calcined material were weighed out and tested. In tests S12-10 to S12-13 the materials for each one-kilogram batch were weighed out, calcined and all of the 8 FLUORSPAR IN SILICATE MELTS calcinate tested so that the premixing could have no effect on the over-all composition of the individual melt. The crucible temper- ature just prior to blowing was 1500° C. in all cases. Table 1.- — Fiber and Shot Diameters Test No. No. of fibers measured Average fiber diameter microns No. of shot measured Average shot diameter microns Group // S-12-6.... 83 1.97 50 67 5-12-7.... 79 2.23 50 66 S-12-8.... 81 2 19 50 57 S-12-9.... 89 1.94 50 63 Group B S-12-10... 83 2.03 50 65 S-12-11... 83 1.88 50 59 S-12-12... 84 2.01 50 64 S-12-13... 93 1.95 50 62 Average of Group A . . Average of Group B . . 2.08 1.97 63 62.5 Av. of Groups A and B 2.02 62.7 Av. deviation ±0.1 micron This average deviation of ±: 0.1 micron indicates a precision about the same as that with which an experienced worker can re- peat measurements on the same mass sample. The agreement within group B is slightly better than within Group A which indicates that our results might have been slightly improved by mixing each charge separately. Although the above tests indicate an ac- ceptable degree of precision, examination of the results indicates that in general such precision was not attained. Because of the number of variables involved it is difficult to determine a reason for this that will fit every case. Both the viscosity and the sur- face tension of the melt are almost certainly influential factors in the control of fiber diameter. Since both are dependent on temperature one might expect that in ranges where the temperature rate of change of both viscosity and surface tension is low (see fig. 4, Region B) it might be relatively easy to reproduce results. However, both viscosity and surface tension data are lack- ing for these glasses so nothing can be said with certainty on this point. It is also possible that the length of the fining period may have an effect on fiber diameter since it is not certain that equilib- rium is attained in any given case. Table 2.- — Average Fiber Diameter of Rock Wool Containing 35 Per Cent Si02 and Vary- ing Amounts of Fluorspar and Blown at Tem- peratures of 1400° and 1500° C. (See figure 1.) Melt No. Pouring temp. (°C) Percentage Fiber diam. CaF2 (microns) Series No. 3— Composition: 35% Si02, 30% CaO, 25% AI2O3, 10% MgO. S3-1 1400 3.7 S3-3 1400 1 3.1 S3-6 1400 2 3.8 S3-2 1500 2.6 S3-5 1500 1 30 Series No. 4— Composition: 35% Si02, 20% CaO, 35% AI2O3, 10% MgO S4-2 1400 16.4 S4-4 1400 1 7.2 S4-3 1400 2 7.8 S4-1 1500 4.4 S4-5 1500 1 4.0 Series No. 15 — Composition: 35% Si02, 40% CaO, 15% AI2O3, 10% MgO S15-2 1400 2.2 S15-4 1400 1 2.4 SI 5-3 1400 2 1.8 S15-1 1:00 1.6 SI 5-6 1500 1 1 4 POURING TEMPERATURE 1400° C. POURING \- TEMPERATURE 1500° C. AI2O3 35% CaO 20% S4 AI2O3 25% CaO 30% S3 AI2O3 15% CaO 40% SI5 2 % FLUORSPAR Fig. 1. — Relationship between fiber diameter of rock wool and fluorspar content of melt (see table 2). WITH REFERENCE TO MINERAL WOOL Table 3. — Average Fiber Diameter of Rock Wool Containing 40 Per Cent Si02 and Vary- ing Amounts of Fluorspar and Blown at Tem- peratures of 1400° AND 1500° C. (See figure ^.) Melt No. Pouring temp. (°C) Percentage CaF2 Fiber diam. (microns) Series No. 10— Composition: 40% SiOa, 40% CaO, 10% AI2O3, 10% MgO SlO-1 1400 3.0 SlO-5 1400 1 3,2 SlO-4 1400 2 2,3 SlO-6 1400 4 1,5 SlO-3 1500 2.6 SlO-8 1500 1 1,8 Series No. 11— Composition: 40%, SiOa, 20%o CaO, 30% AI2O3, 10% MgO Sll-2 1400 9.2 SI 1-5 1400 1 9.3 SI 1-3 1400 2 8.3 Sll-6 1400 4 5,9 Sll-1 1500 4.4 SI 1-4 1500 1 4.2 Series No. 12— Composition: 40%o Si02, 30% CaO, 20% AI2O3, 10% MgO S12-2 1400 3.5 SI 2-4 1400 1 2.1 SI 2-5 1400 2 2.0 S12-19 1400 4 2.7 S12-1 1500 1.9 SI 2-3 1500 1 1.8 Series No. 16— Composition: 40% Si02, 23% CaO, 20% AI2O3, 17% MgO S16-2 1400 2.3 SI 6-3 1400 1 2.5 SI 6-4 1400 2 2.4 SI 6-6 1400 4 2.3 S16-1 1500 2,0 SI 6-5 1500 1 2,1 Series No. 17— Composition: 40% Si02, 37% CaO, 20% AI2O3, 3% MgO S17-1 1400 5,9 S17-3 1400 1 5,1 SI 7-4 1400 2 4,5 SI 7-6 1400 4 4,1 SI 7-2 1500 2,5 SI 7-5 1500 1 2,2 Tests were made at three silica levels as indicated in the diagrams (figs. 1, 2, and 3). The range of composition for the tests was chosen so as to cover approximately the more important portion of the quaternarv system (SiO.-AlA^-CaO-MgO) which Bulletin 61 (op. cit.) states is suitable for the production of rock wooF\ The results are listed in tables 2, 3, and 4 and pictured graphically in figures 1-3. POURING TEMPERATURE I400°C • AI2O3 20% A AI2O3 10% POURING TEMPERATURE I500°C. CaO 20% SI4 CaO 30% SI3 z o 12 12 % FLUORSPAR Fig. 2. — Relationship between fiber diameter of rock wool and fluorspar content of melt (see table -^. Table 4. — Average Fiber Diameter of Rock Wool Containing 50% Per Cent Si02 and Vary- ing Amounts of Fluorspar and Blown at Tem- peratures OF 1400° AND 1500° C. (See figure ^) Melt No. Pouring temp. (°C) Percentage CaF2 Fiber diam. (microns) Series No. 13— Composition: 50% Si02, 30% CaO, 10% AI2O3, 10% MgO S13-1 1400 2.4 S13-4 1400 1 3,1 S13-6 1400 2 2.9 S13-3 1500 2.2 S13-5 1500 1 2.1 Series No. 14— Composition: 50% Si02, 20%, CaO, 20% AI2O3, 10% MgO S14-2 1400 6 3 S14-3 1400 1 6.6 S14-4 1400 2 5.3 S14-1 1500 3.4 S14-5 1500 1 3.7 10 FLUORSPAR IN SILICATE MELTS (J) i.o POURING TEMPERATURE I400°C. POURING TEMPERATURE I500°C. MgO 17% CaO 23% SI6 MgO 10% CaO 30% SI2 MgO 3% CaO 37% SI7 MgO 10% • AI2O3 30% CaO 20% SI I ^ AI2O3 20% CaO 30% SI2 ■ AI3O3 10% CaO 40% SIO 3 4 % FLUORSPAR Fig. 3. — Relationship between fiber diameter of rock wool and fluorspar content of melt (see table ^. DISCUSSION OF DATA The effect of fluorspar on fiber diameter as revealed by these data may be summed up as follows. When the melt without fluorspar is high- ly viscous so that the fibers of the wool produced from it are coarse, the fiber diame- ter may be reduced considerably by adding fluorspar. An extreme example of this is to be found in series S4. With a pouring temperature of 1400° C. the fiber diameter is reduced more than 50 per cent by the addition of one per cent of fluorspar. When, however, this same melt is blown at 1500° C. the same fluorspar addition produces a comparatively insignificant reduction in fiber diameter. If the pouring of the melts is watched carefully it nearly always appears that melts which contain fluorspar are visibly thinner or more fluid than those which contain no fluorspar even if the amount of fluorspar is as little as one per cent. When the melt is sufficiently fluid to pro- duce fibers of the order of two to three microns in diameter, the addition of fluor- spar has little effect insofar as causing further reduction in fiber diameter is con- cerned. It is almost as though there were a lower limit to the fiber diameter. Consider for instance series S12 or S16. Addition of fluorspar affects the fiber diameter but little. This phenomenon is in line with the observation of Schwerin^*^ who found that the effect of fluorspar additions on viscosity of basic open-hearth slags was greater at low temperatures than at higher temperatures, and also with observations of Rait et al"^ who found that fluorspar lowered the vis- cosity of slags in every case but that the effect was less marked in the more basic melts and at the higher temperatures. Fig- ure 4 is a hypothetical representation of the relation of viscosity to temperature and fluorspar content of slags. Actual data of this character for certain slags may be found in the reports of Matsukawa, Schwerin, and various others. If, as is probable, the tem- perature viscosity curves for compositions such as are considered in this paper are similar to figure 4, it might be expected that when the blow is made at temperature K° indicated on the diagram, the first addi- tion of fluorspar would cause a large reduc- tion in fiber diameter and that succeeding additions would be proportionately much less effective. If the blow were made at temperature L°, only small effects could be expected from fluorspar additions. WITH REFERENCE TO MINERAL WOOL 11 TEMPERATURE Fig. 4. — Schematic diagram showing the effect of fluorspar additions on the viscosity of slags: curve 1, no fluorspar; curves 2, 3, 4, and 5, in- creasing quantities of fluorspar. A Mechanism for the Effect of Fluorspar on the Viscosity OF Silicate Melts It is interesting to consider the problem of the effect of fluorspar on silicate melts from the viewpoint of recent theories con- cerning the structure of glasses. Eitel'' has called attention to the fact that such an hypothesis offers an explanation of the action of fluorides as mineralizers, that is to say, substances which, if present in liquid phases, act in such manner as to aid in the estab- lishment of equilibrium and promote the formation of crystalline solid phases. These theories apply to glass in the solid state but it would seem logical to assume that, since the transformation from the solid to the liquid state is gradual in glasses, the break-up of the solid structure is probably also gradual. In other words, the structure of softened glass must still bear some rela- tion to that of the solid up to an uncertain point at least, although it must be admitted that our knowledge of the structure of the liquid state is meagre indeed. Zachariasen^" developed a theory of the structure of glasses around the idea that glasses are built up of extensive three- dimensional atomic networks which lack the periodicity and symmetry that characterize crystals but which are in most other respects similar to crystal lattices. The forces that hold the atoms together are assumed to be essentially the same as those in crystals. It is difficult to draw a picture in two dimen- sions which represents the unsymmetrical three-dimensional network. Figure 5B rep- resents the manner in which atoms of a hypothetical oxide A2O3 might be joined together in a glass network in two dimen- sions. The filled circles represent the atoms A, the open circles the oxygen atoms. If this idea is applied to a glass which is domi- nated by a silicon oxygen network the pic- ture might be extended into three dimensions by adding one additional linkage bond to each filled circle. This bond would extend cither before or behind the plane of the paper and each filled circle representing a silicon atom would be joined to four open circles representing oxygen atoms. Each oxygen atom would thus act as a bridge connecting two silicon atoms. The whole mass would consist of a sort of polymer of SiOo with extension in all three dimensions limited only by the boundaries of the mass. Each silicon atom would be at the center of a tetrahedron with an oxygen atom at each of the four corners. These tetrahedra would be joined at their corners through the oxygen atoms a majority of which are shared by two tetrahedra. An arrangement such as described would be a crystal if the tetrahedra were arranged with mathematical order and symmetry and if the arrangement were periodic, that is, if groups of atoms recurred at regular inter- vals in space so as to permit a unit of infinite size with uniform constitution to be built up. A two-dimensional arrangement of this type appears in figure 5A. It relates to figure 5B in a manner analogous to that in which the hypothetical three-dimensional crystal is assumed to relate to the hypothetical SiO^, glass. Next be it assumed that the glass is in a liquid state at high temperature and that in addition to SiOo it contains one or more varieties of metallic oxides. Due to the more energetic vibration of the atoms at elevated temperatures the number of SiO, molecules joined together in networks would be rela- tively small. As the melt is allowed to cool these comparatively small groups might be 12 FLUORSPAR IN SILICATE MELTS LEGEND A ATOM OXYGEN ATOM Fig. 5. — Hypothetical oxide A2O3, illustrating arrangement of atoms in crystalline modification (A) and in glassy modification (B), expected to increase in size much as crystals grow. The units grow with drop in tem- perature and finally the melt takes on the characteristics of a solid glass in which the network is assumed to be more or less con- tinuous. It must however, be kept in mind that the metallic oxide was present within the melt and that our network was built up in such manner as to include these oxides within the space occupied by the more or less continuous silicon oxygen network. In other words the network grows around the molecules of metallic oxide which may possi- bly be ionized, if the term may be considered to have a meaning under such circumstances. These cavities or open spaces within the silicon oxygen net which contain the metal atoms or metal oxide molecules are assumed to be statistically distributed and their size and number controlled by the size, charge and number of the metallic cations. War- ren^'^ thinks that for the soda-silica glasses there is a definite scheme of coordination with each silicon atom tetrahedrally sur- rounded by four oxygen atoms, part of which are bound to two silicon atoms and part to only one silicon. But all oxygen atoms are assumed to be attached to at least one silicon atom. It is obvious that this last condition limits the value which the ratio of the number of metallic oxide molecules to the number of silica molecules may have and still leave the possibility of forming a continuous network in three dimensions, be- cause in order to form such a network a considerable proportion of the silicon atoms must be linked through the oxygen bridges to at least three other silicon atoms. Each valence on the metallic cations must be balanced by one on an oxygen attached to only one silicon atom. This is obvious from stoichiometric considerations and is illustrat- ed in figure 6 which is a schematic repre- sentation in two dimensions of the arrange- ment we are attempting to picture. One valence bond is omitted from each silicon atom on account of the difficulty of repre- senting it on a two-dimensional diagram. Figure 6a represents the condition assumed to exist in the solid state. As was pointed out in Zachariasen's paper^'^ a solid con- structed as postulated should not be expected to have a definite melting point since the thermal energy required to detach individual atoms from the network would vary from atom to atom. As the temperature rises the more loosely bound atoms would be detached first and those more strongly held would be freed later as the vibrational energy reached higher values. Probably the network would be broken up into clumps of atoms, rela- tively large at the lower temperatures but WITH REFERENCE TO MINERAL WOOL 13 LEGEND METALLIC CATION OXYGEN ATOM CHARGED TO BALANCE NEIGHBORING CATION O- OXYGEN ATOM )^- SILICON ATOM ® FLUORINE ATOM Fig. 6. — Hypothetical glass networks, Illustrating the effect of metallic fluorides in weakening such networks. A^ — Glass composed of metal oxide and silica. Note more open structure than in figure 5 B. B — Same as A except that fluoride has been introduced. Note more open structure than in A. becoming smaller as the increasing thermal energy reaches values sufficiently high to overcome the forces which hold the units of the clump together. Due to changes in the systems of balanced forces the network would undergo modifications as this process continued so that eventually a point would be reached where a resemblance to the orig- inal network could no longer be traced in the residual fragments. Those readers who desire a comprehensive and critical treatment of the theory of glass structure outlined above are referred to Zachariasen's paper^-' and also to papers by Hagg^"* and by Zachariasen^'\ For present purposes the outline given should suffice. Using as starting points the well known facts that fluorine forms stable compounds with silicon and that silicon and fluorine are slowly lost from silicate melts by volatili- zation as SiF^, it is possible to explain on the basis of the above theory of glass struc- ture the fact that fluorspar exerts a pro- found influence on the viscosity of silicate melts. Consider a silicate glass of given compo- sition at a given temperature above the softening point of the glass and possessing the viscosity imposed by these conditions. This viscosity would presumably be related to the size of the structural aggregates present. These aggregates we assume to be held together by the remnants of the net- work which according to the theory char- acterized the solid glass at lower tempera- ture. Bonds between oxygen and silicon are being broken and new ones are being formed according to the ordinary conceptions of chemical kinetics. If fluorspar is present bonds will be formed between silicon and fluorine. This will result in a reduction of the number of valence bonds on silicon atoms available for oxygen bridge formation. The average number of oxygen bridges holding the aggregates together will be re- duced and the aggregates consequently might be expected to break up at lower tempera- ture. In other words the viscosity should attain given values at lower temperatures when fluorspar or any substance capable of yielding fluoride ions is present. Compari- son of figure 6b with 6a illustrates the weak- ening of the structure brought about by the introduction of fluoride ions as pictured by the structure theory. This picture also appears to harmonize with the observed fact that elimination of fluorine as SiF^ from glassy melts is very slow except when the melts are thin or of low viscosity. To test this a batch containing approxi- mately 20 per cent ALO^, 37 per cent CaO, 14 FLUORSPAR IN SILICATE MELTS 3 per cent MgO and 40 per cent SiO. was combined with approximately 1 1 per cent of its weight of fluorspar. The mixture was then melted and held at 1500° C. for a period of 195 minutes. Samples were re- moved from this melt at intervals and were analyzed for fluorine and for silica. The results showed that there was no loss of fluorine or silicon from this melt during the period named which was large enough to be detected in the ordinary analytical pro- cedure. The migration of a fluorine atom or ion through a network such as we are consider- ing would be slow and consequently the number of collisions with active silicon at- oms relatively few. Moreover a silicon atom with three fluorine atoms attached might still be bound to the network through its fourth valence. When the last oxygen bridge is broken and the fourth fluorine atom has reacted to produce the molecule SiF^, the newly formed molecule must still remain intact and make its way to a free surface before it can be eliminated from the melt. Some Quantitative Aspects of the Network Theory Consider a glass made up by combining a metal oxide A^O, a metal fluoride AF, and SiO^. The composition of such a glass may be expressed by a formula A,„FpSinO where m, p and n represent the number of atoms of A, F and Si respectively, per oxy- gen atom. Each molecule of A2O or of AF will prevent the formation of one oxygen bridge between two silicon atoms. "T^ A^O + AF -o-i-o A"*" A^ o-i-^ A^ In order to have a continuous network in three dimensions, at least three corners of each tetrahedron should be linked to other tetrahedra through oxygen atoms common- ly held by two tetrahedra. This condition may be met if the combined number of molecules of A2O and AF is not more than half of the number of silicon atoms; that IS to say, it 4- p / — , or if the sum 2 ^ < 2 of the mole percentages of A2O and AF is not greater than 33.3-|-. The suggestion is that the glass with more than this content of A^O plus AF should begin to exhibit a tendency toward instability and weakness. It might also be expected that the vis- cosity at a given temperature would decrease as the combined mole percentages of AgO and AF increased. One mole of AgO should be about as effective as one of AF in reduc- ing viscosity. When a divalent cation such as calcium is under consideration the situ- ation is slightly different. In this case one mole of CaFo should be about as effective as two moles of CaO in weakening the net- work. As between metallic cations with the same valence some difference in effect should be expected because the repulsive force between the metallic cations and the silicon atoms is governed by the distance between the centers of their nuclei as well as by their respective charges. Thus magnesium might be expected to weaken the network slightly more than would calcium. It is proposed to investigate some of these points in the near future. SUMMARY The effect of fluorspar additions upon the fiber diameter of mineral wool produced from certain lime-magnesia-alumina-silica melts has been studied. Fluorspar additions cause considerable re- duction in fiber diameter of certain wools which are coarse textured when no fluorspar is present. As a rule when small additions of fluorspar cause considerable reduction in fiber diameter, larger additions will not cause proportionate reductions. If the wool is fine fibered with no fluorspar present, addition of fluorspar will cause little or no further reduction in fiber diameter. The mechanism of the action of fluorspar in lowering the viscosity of molten slags and glasses is discussed, and an explanation based on Zachariasen's network theory of the structure of glasses is offered. WITH REFERENCE TO MINERAL WOOL 15 BIBLIOGRAPHY 1. Carsten, Ad, Use of fluorspar in cupola oper- ation: Iron and Steel vol. 13, p. 139, 1940. 2. Schwerin, Lenher, Action of fluorspar on open hearth basic slags: Metals and Alloys, vol. 5, part I, p. 61; part II, p. 83, 1934. 3. Matsukawa, Tatsuo, On the viscosity of acid and basic open hearth and cupola furnace slags in molten state: Pamphlet, 65 pages pubHshed by the Tanaguchi Foundation for the Promotion of Industrial Progress, Osaka, Japan, March, 1935. 4. Rait, J. R., M'MiUan, Q. C, and Hay, R., Viscosity determinations of slag systems: The Journal of the Royal Technical College, vol. 4, p. 449, Glasgow, 1939. 5. Lawrie, W. B., The refining of metal in the basic open-hearth furnace. The influence of fluorspar on the process: Jour, of the Iron and Steel Institute, vol. 139 (Proc.) p. 257, 1939. 6. Hartmann, F., The Influence of different slag forming materials on the viscosity of blast furnace slag: Stahl u. Eisen, vol. 58, p. 1029, 1938. 7. Hartmann, F., Studies on the viscosity of Siemens-Martin slags: Arch. Eisenhiittenw., vol. 10, p. 45, 1936-37. 8. Endell, K., Heidkamp, G., and Hax, L., Con- cerning the degree of fluidity of lime silicates, lime ferrites and basic Siemens-Martin slags up to 1625 °C.: Arch. Eisenhiittenw., vol. 10, p. 85, 1936-37. 9. Eitel, von Wilhelm, The action of fluorides as mineralizers in cement clinker burning: Zement, vol. 27, part 2, p. 455, 1938. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. Schwerin, Lenher, Effect of fluorspar on the viscosity of basic open -hearth slags: Metals and Alloys, vol. 5, p. 118, 1934. Herty, C. H,, Hartgren, F. A., and Jones, G. T,, Temperature viscosity relations in the system CaO-Si02-CaF2: Carnegie Inst. Inst. Technology. Coop. Bull. 56, 32 pages, 1931. Hamilton, W. S., The action of fluorspar on basic open-hearth slags: Metallurgical and Chemical Engineering vol. 13, p. 8, 1915. Sisco, Frank T., Fluorine in the deoxidizing slag, and its influence on the refractories in basic electric furnace practice: Trans. Am. Electrochem. Soc. vol. 46, p. 531, 1924. Karandief, B., On the binary systems of cal- cium metasilicate with calcium fluoride and calcium chloride: Zeitschr. Anorg. Chem. vol. 68, p. 188, 1910. Lamar, J. E., Willman, H. B., Fryhng, C. F., and Voskuil, W. H., Rock wool from Ilhnois mineral resources: Illinois Geol. Survey Bull. 61, part III and appendix III, 1934. Zachariasen, W. H., The atomic arrangement in glass: Jour. Am. Chem. Soc, vol. 54, p. 3841, 1932. Warren, B. E., and Biscoe, J,_, Fourier analysis of X-ray patterns of soda-silica glass: Jour. Am. Cer. Soc. vol. 21, p. 259, 1938. Hagg, Gunnar, The vitreous state: Jour, of Chem. Physics, vol. 3, p. 42, 1935. Zachariasen, W. H., The vitreous state (com- ments on the article by Gunnar Hagg): Jour, of Chem. Physics, vol. 3, p. 162, 1935. ILLINOIS STATE GEOLOGICAL SURVEY Report of Investigations No. 68, 1940