The production, properties, and selection of ferrosilicon powders for heavy-medium separation by B. COLLlNS*. B.Se. Eng. (Rand) (Fellow). T. J. NAPIER-MUNNt. B.Sc. (Eng.). A.R.S.M. (Visitor). and M. SCIARONEt. M.Se. Eng. (Delft) (Visitor).
SYNOPSIS There are basically two types of ferrosilicon powder available to operators of heavy-medium plants in South Africa: spherical and milled grades. The production processes are described, and the properties of the two types are examined and compared. Guidelines are given for the selection of the correct grade of ferrosilicon for any particular application.
SAMEVATTING Daar is basies twee so::>rte ferrosilikonpoeier beskikbaar vir diegene wat swaarmediumaanlegginge in Suid-Afrika bedryf: steriese en gemaalde grade. Die produksieproses word beskryf en die eienskappe van die twee soorte word ondersoek en vergelyk. Daar word leidrade aangegee vir die keuse van die korrekte graad ferrosilikon vir enige bepaalde gebruik.
INTRODUCTION THE PRODUCTION OF of milled ferrosilicon. The following FERROSILICON POWDERS Ferrosilicon of 14 to 16 per cent are the characeristics of this furnace: Shell diameter 6400 mm silicon has become widely accepted At present there are basically Hearth diameter 4000 mm as the most suitable medium for the two types of ferrosilicon available Electrode diameter 890 mm heavy-medium separation of ores in South Africa to the operators of Secondary voltage 90 to 120 v having a specific gravity in the heavy-medium plants. These are Resistance 0,85 m 0 range of approximately 2,5 to 4,0. milled and spherical grades. Ferrosilicon has many properties Reactance 1,0 m o. essential to a metal or alloy powder Milled Ferrosilicon The power consumption (including that is to be usei as a heavy medium, The standard method is to melt steel auxiliaries) for 1 tonne of ferrosilicon some of the more important being scrap, quartz, and a reductant in a (15 per cent silicon) amounts to the following: submerged-arc furnace. The molten 2100 kWh. The furnace is charged (a) resistance to abrasion, alloy is tapped into a sandbed, on a semi-continuous basis and is usually tapped into a ladle every two (b) resistance to corrosion, allowed to cool, and then broken into hours. The temperature of the melt (c) high specific gravity, lumps, after which it is crushed in two stages and milled to the required on tapping is 1600°0. (d) magnetism, which allows easy Slze range. The liquid ferrosilicon from the magnetic recovery with subse- At the Kookfontein Works of ladle is granulated into small part- quent easy demagnetization, Amcor, a 7,5 MVA bubmerged-arc icles in a high-pressure water jet. and f..unace is used for the production This granulation process, developed (e) low cest. Ferrosilicon contain:n~ between TABLE I 14 and 16 per cent silicon is foun:! SPECIFICATIONS FOR MILLED FERROSILICON 14 TO 16 PER CENT SILICON to have the optimum combination Size Analysis of these propertiesl. If the silicon content is lower than 14 per cent, Tyler me3h 48D 65D 1O0D 150D 270D the specific gravity and magnetic + 65 5 0,5 0 0 0 properties are improved, but resist- +100 15 3,0 0,2 0 0 ance to corrosion decreases rapidly. +150 30 8,0 1,2 0,5 0 +200 50 20,0 5,0 2,0 0,2 Above 16 per cent silicon, the corros- +325 75 55 35 25 10 ion resistance of the alloy is not -325 (typical) 25 45 65 75 90 significantly improved, but the mag- -325 (limits) 20-30 42-49 60-70 70-80 90+ netic properties and specific gravity deteriorate. Chemical and Physical Specification.. Silicon 14-16 % Carbon 1 % max. Sulphur 0,05 % max. Phosphorus 0,10% max. *Hoechst South Africa (pty) Ltd. Rust Index 1,0% max. t De Beers Industrial Diamond Division. Non-magnetics 0,75% max. :j:Metalloys Limited. Specific gravity 6,7-6,95
JOURNAL OF THE SOUTH AFRICAN INSTITUTE OF MINING AND METALLURGY DECEMBER 1974 103 by Iscor, eliminates the breaking and TABLE n crushing stages and is believed to SPECIFICATIONS FOR ATOMIZED FERROSILICON 14 TO 16 PER CENT SILICON have the added advantage of pre- Size Analysis venting segregation in the alloy on cooling. The granulated material Sishen ISCOR Cyclone Cyclone Tyler mesh Coarse Coarse Coarse Fine 60 40 is recovered by spiral classifier and * t dried in a gas- fired rotary dryer. + 65 3 4 3 1 0 0 The final powder is obtained by +100 13 13 11 8 0 0 milling in a ball mill in closed circuit +150 30 31 28 18 2 0 with an air classifier, which is set +200 49 44 40 33 7 2 to yield a product of required size +325 74 67 62 55 27 10 -325 (typical) 26 analysis. 33 38 45 73 90 -325 (limits) 23-30 30,1-35 35,1-40,5 40,6-50 65-75 A list of the grades of milled -400 (typical) 65 85 ferrosilicon available in South Africa, together with their chemical and Cyclone 60: * + 100 mesh 0,02 % max. size specifications, is given in Table I. +200 mesh 12% max. t Cyclone 40: Is not produced at this stage because there is no local demand. Spherical Ferrosilicon Chemical Specification There are two methods for the Normal Grades Special Grades Cyclone Grades Si 14-18% 14-16% 14-16% production of spherical grades of C :tI % 1 0,5 % 1,5 % max. ferrosilicon. S 0,05 % max. 0,05 % max. 0,0.5 % max. P 0,1 0,1 Atomized Ferrosilicon % max. 0,1 % max. % max. A!. 10,8% 0,5% max. 0,5 % max. Knapsack AG, a subsidiary of Ma 0,75% max. 0,75% max. 0,75% max. Cu 0,8% max. 0,5% max. 0,5 % max. Hoechst AG, was the first to apply Cr. 0,5% max. 0,5% max. 0,5 the atomizing process to the product- % max. ion of ferrosilicon for heavy-medium Physical and Physiochemical Specification: separation2. This process is now also Content of spheres: Normal grades : 1 50 % used in the Hymat plant at Amcor's Cyclone and special grades: 190% Magnetic contont : 99%+ Kookfontein works. Pyknometer density : 6,6-7,0 gfcm3 Ferrosilicon of 75 per cent silicon Bulk density : 3,5-4,2 gfcm3 is diluted with high-grade steel scrap in an induction furnace. Character- aI, and size specifications, is given in fine fraction of the final product. istics of this furnace are as follows: Table H. In spheroidizing, the milled Holding capacity 3 tonnes Spheroidized Ferrosilicon powder is allowed to fall through a Electrical capacity 750 kW A second method for the product- flame that has an oxidizing centre Power consumption 600 kWh/t ion of spherical ferrosilicon was and a 'reducing outer zone, which Type Mains frequency. developed by IscorI. Test" have permits each individual particle to By means of a tilting device, the shown that in many respects this be melted without being oxidized. furnace is tapped at approximately material is identical to that pro- The surface tension and fluidity of 3-hourly intervals at a charge tem- duced by the Knapsack method. ferrosilicon are such that each part- perature of approximately 1550 °C. Spheroidized ferrosilicon is pro- icle contracts into a sphere. The The molten material flows into a duced initially in the same induction design of the burner prevents ceramic tundish, which ensures that furnace. However, the stream of agglomeration, and, after spheroid- the alloy stream entering the atom- molten alloy is broken up in a series izing, the powder is quenched in izing steam cone has a constant of concentric cones of high-pressure waste gas to prevent oxidation. diameter and velocity. water. Clearly, the energy available It is claimed that the water- In the atomizing nozzle, the in this system is not as great as in granulated fraction of the final stream of molten ferrosilicon comes steam atomizing, and consequently product is superior to the correspond- into contact at its apex with a cone the water-granulated product is ing fraction of the atomized product, of steam, and the melt is immediately coarser. making it possible to operate at broken into fine particles, which are In fact, only a fraction of this higher pulp densities1. quenched in water. material (about 30 per cent) consti- THE PROPERTIES OF FERRO- The resultant ferrosilicon pulp is tutes the final product. The granu- SILICON 14 TO 16 PER CENT filtered in a drum filter, dried in a lated product is dried and screened SILICON gas dryer, then screened for the at 48 mesh, and the screen oversize removal of oversize material. This (about 70 per cent) is milled to the Premier Mine installed the first. oversize fraction (approximately 3 required size in a ball mill in closed heavy-medium separation plant in to 10 per cent of the atomized circuit with an air classifier. The South Africa in 1948, the complete ferrosilicon) is milled in the ball mill. milled fraction is then spheroidized, specification for the ferrosilicon used A list of the grades of atomized and the two fractions are combined in that plant being 25 per cent minus ferrosilicon available in South Africa, in the desired proportions. The 325 mesh. Since then, considerable together with their chemical, physic- spheroidized powder constitutes the progress has been made both in the
104 DECEMBER 1~74 ,",OURNAL OF THE SOUTH AFRICAN INSTITUTE OF MINING "NO MHALLURGY production and quality control of medium separation itself and in the does affect the performance of a miI1ed ferrosilicon. handling of the medium with regard heavy-medium cyclone, and this Notwithstanding these advances, to pumping and storage. The view has been supported recently by the uSe of milled ferrosilicon is rheology of fast-settling suspensions some limited plant studies of cyclone limited. The production of atomized of this kind can best be described by ferrosilicon was a welcome develop- two properties of the suspension: operations within the diamond ment because it made possible the viscosity and stability. industry. heavy-medium separation of certain ores that could not be treated with Viscosity The measurement of medium vis- milled ferrosilicon. The viscosity of a ferrosilicon cosity is made difficult by the suspension directly affects the necessity of maintaining the solid Particle Shape separating efficiency of the simplest form of heavy-medium separator- particles in homogeneous suspensions There is a marked difference in the static bath-by influencing the while the measurement is being both the shape and nature of the terminal velocity of ore particles made. A number of investigators particles of ferrosilicon powders pro- settling in the medium in a manner have approached this problem with duced by grinding and by atomizing. akin to that described by Stokes varying degrees of success, and a Particles produced by grinding are law for Newtonian fluids3. One in- rough and angular in shape, whereas vestigator succeeded in deriving a number of ingenious instruments particles produced by atomizing are direct relationship between plastic have been developed for the pur- smooth and largely spherical or viscosity and the probable error of pose6, including flow-rate or rounded. This difference in the the separatIOn (defined by the Tromp capillary-type consistometers7.8, a individual particle shape has marked Curve) in tests on coal separations4. pressure-diaphragm viscometer9, effects on the rheological properties The influence of medium viscosity concentric cylinder devices4.10,1l,and of aqueous suspensions of the two on dynamic separators such as the types of ferrosilicon. cyclone or Dyna Whirlpool is less other rotational instruments such certain. Many authorities maintain as the modified Stormer visco- Rheological Properties that the effect is negligible in relation meter12-15, which is used by Knap- The rheological properties of to the high dynamic forces exerted sack16 and is widely used in South heavy-medium suspensions play an on the ore particles. One study5, Mrica. important part both in the heavy- however, has suggested that viscosity Unfortunately, most of these instruments are able to provide only a relative measure of the viscosity of a sample based on calibration with VisCOsiTY. 1"/ Pois, ~~ liquids of known viscosity such as glycerine. Accordingly, the detailed shear-rate and shear-stress character- istics of ferrosilicon suspensions are still not well understood, although .. some attempts have been made to UI obtain this information by suitable ~~ calibrations and interpolation12,
UJ It is generally agreed, however, 4:I- 0:: that the rheological characteristics of 0:: 4: ferrosilicon suspensions, as of most UJ :I:
,",OURNAL OF THE SOUTH AFRICAN 'N~TITUH OF MINING AND METALLURGY DECEMBER 1974 105 r Pulp 100 V !Scos, ty I cP I j
75
50
25
2.1. 2,8 3,2 3,6 1.,0 Pu I P sp. go.
Fig. 2-Pulp specific gravity versus viscosity for various grades of ferrosilicon teristics at low shear rates (up to large increases in viscosity, with to the treatment. about 300S-1) and dilatent character- deleterious effects on separation and In general, it is not usual to istics at high shear rates. This pumping characteristics. employ milled ferrosilicon at a pulp could have implications in heavy- The critical value of pulp density density much in excess of 3,0. medium cyclone operations where for the atomized grades is shown to Stability relatively high shear rates are ex- be much higher than for the milled The stability of a suspension can perienced. A study5 of the influence grades, and not as clearly defined. be defined as the reciprocal of the of medium viscosity on Driessen Unlike the viscosity of the atomized settling rate. Cone separations did suggest that an grades, that of milled ferrosilicon is As with viscosity, a number of increase in viscosity leads to an shown to increase with fineness or techniques have been developed for increase in the proportion of ore grade of the ferrosilicon. Fig. 3 the measurement of stability. The reporting to the underflow. shows that the relationship between simplest method is to agitate a All investigators agree on the viscosity and percentage minus 325 sample of the medium, allow it to general shape of the curves for the mesh is in fact almost linear. stand, and time directly the settling pulp density versus viscosity of At densities above 3,0, the vis- rate of the clear water interface7. ferrosilicon suspensions, e.g. Eveson6 cosities of suspensions of atomized However, in many cases, particularly Fig. 2 shows curves for selected ferrosilicon are significantly lower where excessive fines are present, no grades manufactured by Amcor in than those of the milled grades owing clear demarcation line is visible. South Africa. In each case, a critical to the greater friction of collision Another method developed by the value of pulp density (or volume of the irregularly shaped milled Diamond Research Laboratory, and proportion of solids) is reached particles, which requires a larger used by Knapsack16, employs a beyond which the viscosity rises energy input to achieve a given differential-pressure technique, in sharply. In general, it is inadvisable deformation of the suspension18. which air is bubbled through the to use any given grade in any This superior property of atomized settling medium and the change in application in the region beyond ferrosilicon allows very high pulp pressure at a point in the suspension this critical point, since small in- densities to be obtained (up to 3,8), is timed by use of a manometer19. creases in density or the presence which has made possible the heavy- A more sophisticated technique, of small quantities of fine non- medium separation of certain ores an advanced version of which is magnetic contaminants can cause that were not previously amenable currently under test at the Diamond
106 DECEMBER 1974 JOURNAL OF THE SOUTH AFRICAN INSTITUTE OF MINING AND METALLURGY ...... This provides a relative scale from 100 per cent (no settling) to 0 per cent (total settling of the ferrosilicon solids to below the level of the tap-ofl hole). Medium V"co,,!y Stability data21 for certain grades (oPI offerrosilicon manufactured in South Africa are given in Fig. 4. It is clear 50 that, as in the case of viscosity, 'Linear Correlation Coo";o;en! 0 0,97 stability increases with both the fineness of the ferrosilicon and with pulp density. For both static and dynamic .0 separations, the desired conditions are thol'se of minimum viscosity with maximum stability. A low stability, due for example to the use of too coarse a grade of ferrosilicon,
30 will result in a large differential between the specific gravities of the overflow and underflow media from the separator, whether it be static bath or cyclone. This in turn will
20 lead to the artificial creation of a :arge body of middlings material, and consequently a poor separation will be achieved. As might be expected, grades of milled ferrosilicon show greater sta- 10 35 55 75 95 bility than grades of atomized ferro-
(325 m.,hl- silicon of equivalent size analysis. 'i,-'5tW' Generally, it is necessary to use a grade of atomized ferrosilicon that is finer than the milled ferrosilicon Fig. 3-Apparentviscosity versus fines content of milled ferrosilicon-pulp density 2,80 gJcm3 (after Wilson17) that might be used under the same conditions. This entails no disadvant- age, however, as the viscosities of Research Laboratory, was developed to stand for 60 seconds, after which suspensions of the atomized grades at the Fuel Research Institute. This the bung in the side is removed and are not significantly dependent on invol ves the measurement, by an the top portion of the suspension is their fineness. optical method, of the rate of drained off. The tube is weighed change in specific gravity at a point before and after this procedure, and, Modification of Rheology in the settling medium2°. The change from the dimensions and dry mass The modification of the rheological in specific gravity over a timed of the apparatus, the change in properties of ferrosilicon suspensions interval has also been measured by density of the medium in the top can be achieved in three main ways: a gamma-ray attenuation method21. portion of the tube rehtive to the (1) by addition of polymeric com- Another simpler, though less pre- initial density is easily calculated. pounds or other reagents, cise, method of timing the change in The method, though relative and (2) by addition of ore slimes or specific gravity (slightly modified arbitrary with respect to the dimen- clays, or from that described by Geer et al.9) sions of the apparatus and the time (3) by demagn8tization of the circu- is as follows. A perspex tube, about allowed for settling, is particularly lating medium. 1 cm in diameter and 30 cm high, suitable for rapid phnt measure- Considerable work has been carried closed at the lower end, with a hole ments. In the present case, the out on the stabilization of heavy- drilled in the side about one-third stability index 8 is given by medium suspensions for static-bath of the distance from the top and separators by the addition of poly- 8=1- ( -=-~2) X 100 % fitted with a bung, is placed ~ 1 mers. It is claimed that significant vertically. The medium is then D2 -1 improvement in stability (with a X 100 0//0 , introduced into the tube so as to = ( D1 -1 ) corresponding increase in viscosity) fill it completely, the top is closed, where D1= initial density of the can be achieved by the addition of and the whole is shaken vigorously medium (g/cm3), and as little as 0,1 per cent by mass of so as to ensure homogeneity of the D2=final density in the top solids of certain polymers. This suspension. The tube is then allowed portion of the tube. allows the use of coarser (and
JOURNAL OF THE SOUTH AFRICAN INSTITUTE OF MINING AND METALLURGY DECEMBER 1974 107 cheaper) grades of medium, easier been found to influence significantly Size Distribution start-up due to reduced settling, and the viscosity of the medium. In order Apart from the primary classes of other advantagesll. Reductions of to reduce viscosity, most static-bath atomized and milled ferrosilicon, viscosity by the addition of peptizing processes include a demagnetization the various manufactured grades are agents have also been reported22. coil on the pipe that returns medium distinguished and classified exclu- This effect is found to be highly from the magnetic separators to the sively by their size distribution. pH-dependent. main circuit. Up to now this has not Specifications for the size distribu- The addition of small quantities of normally been considered necessary tions of the various grades manu- ore slimes or clays has been found for cyclone plants23. factured by Amcor are given in to increase significantly the viscosity The design of the coil with regard Tables I and 11. and stability of ferrosilicon suspen- to attenuation characteristics and Sizing for quality control is carried sions. In some cases, the slimes field strength is of considerable out by conventional screening down represent an undesirable constituent importance24. A recent exercise at to 400 mesh (38 microns) and by of the ore being treated, and a the Premier Mine25 established an sub-sieve methods therafter to proportion of the circulating medium inverse relationship between coil about 5 microns; sub-sieve tech- has to be continuously removed and field strength and the viscosity of niques employed successfully in this cleaned through magnetic separators. the circulating medium in the heavy- country and overseas include the In other cases, natural slimes can medium cone circuits (Fig. 5). Bahco air elutriator, the Alpine be used to stabilize a relatively suction/sieve apparatus, and the coarse grade of medium23. In general, atomized ferrosilicon Cyclosizer. The degree of magnetization of the requires a field strength for de- The size distributions of both ferrosilicon particles, induced by magnetization that is four times manufactured ferrosilicon and ferro- their passage through magnetic Eepa- higher than that for milled ferro- silicon taken from operating heavy- rators during normal recovery, has sillicon. medium circuits have been found to be well lepresented by the Rosin-Rammler distribution, which Stabd;ty s/em is given by 601 (x/a)/), WR=lOO e where Wr= % mass retained, x=size (determined by nominal sieve aper- 50 ture), and a,b = constants. This expression can be linearized as follows: 100 .0 In In - = b(ln x - In a). WR By the plotting of In In (lOO/W R) versus In x, a straight line is ob- tained with gradient b. The charac-
30 teristic size of the distribution is generally defined as the size at which x=a. This gives W R=lOO e -1 = 36,8 per cent. It has been found that, for ferrosilicon distributions, the linear 20 correlation coefficient for straight lines fitted in this way is normally in excess of 0,95. Fig. 6 shows the Rosin-Rammler plots for samples taken from a number of Amcor 10 grades. The ability of ferrosilicon size distributions to be represented in this way is extremely useful both for specification purposes and in the evaluation of research and plant tests. 2,6 2,8 3,0 3,2 3,' P,lp sp.gr- Absolute Density Fig. 4--Stabilities of various grades offerrosi Iicon (data from Nesbitt and Loesch19) The absolute density of a ferro-
108 DECEMBER 1974 JOURNAL OF THE SOUTH AFRICAN INSTITUTE OF MINING AND METALLURGY silicon alloy is governed essentially by its metallurgical constituents, and varies within relatively narrow limits depending on the exact pro- portions of the elements present. The high density of the powder (6,7 to 7,0) allows much higher pulp densi- ties to be achieved than with other Circulat;ng Medium con v e n t ion a I heavy-medium Viscosity materials. IcPI The measurement of density is usually carried out with specific 50 gravity bottles or pyknometers. The Cone 4 determination must be made with great care because of the extreme fineness of the powder. To wet the particles completely, paraffin is 40 usually preferred to water as the wetting medium, and the bottles are usually evacuated, by use of a water vacuum pump, for 2 to 4 hours, being gently tapped from 30 time to time to remove all entrained air. All weighing is done to 0,0001 g. The density of the paraffin is critical, Cone 3 since an error of only 0,005 gfcm3 (about 0,6 per cent) in the measure- 20 150 200 250 300 350 400 450 ment of the paraffin density will - 100 Oe-Mag. Coil Field Strength Igaussl- lead to an error of about 20 per cent Imeasured "ternally with medium flowing through coils 1 in the estimate of the ferrosilicon density. The determinations have therefore to be made at an accurately controlled temperature. With care, Fig. 5-Demagnetized coil field strength versus medium viscosity for Premier Mine a standard deviation of 0,02 gfcm3 heavy-medium cones (after Napier-Munn and Wilson2")-field strength: mean of in the estimate of ferrosilicon density two parallel coils is usually possible, based on three replicate determinations. most of the investigations reported netic separator. From this point of (e.g. that by Schmeiser et al.27), view it is interesting to note that Magnetic Properties some form of magnetic balance was the minus 400 mesh fraction of any Being an iron alloy, ferrosilicon is used. This type of device is used to ferrosilicon sample always gives a inherently magnetic. This property measure the magnetic moment of a lower reading than the plus 400 ferromagnetic sample when satu- of the medium permits easy recovery mesh, and that atomized ferrosilicon rated in a strong magnetic field. A always gives a significantly higher and cleaning of the medium in circuit. A certain residual magnetism commercial instrument of this type reading than the milled material. is the Satmagan balance, manu- is induced in the medium by passage This is in accordance with plant through a magnetic separator during factured by Outokumpu Oy of experience that the fines in the Finland, and originally designed for normal operation. As noted earlier, medium are lost preferentially from excessive residual magnetism can the determin'ttion of magnetite in the magnetic separators. It is also ores and slags. have a deleterious effect on the usual to find that readings are lower viscosity of the medium. In addition, Tests on the application of the from dried samples obtained from if a magnetic ore is being treated, Satmagan to ferrosilicon analysis at an operating circuit, even after fine this residual magnetism in the the Diamond Research Laboratory non-magnetic contamination has medium will lead to excessive losses have shown that, although the been removed, probably owing to the of the medium by adhesion. The instrument gives an indication of presence of an oxidized surface on agglomeration effects due to residual the proportion of magnetic iron the ferrosilicon as a result of corros- magnetism can be observed directly present in the alloy, the reading is ion. under a binocular microscope; an also influenced by the metallurgical During the manufacturing process investigation of residual magnetism species present and by size distribu- of milled ferrosilicon, a small propor- with a sedimentation photometer has tion. It is probable, however, that tion of non-magnetic material be- been reported26. the results obtained can be related comes mixed with the ferrosilicon. The study of inherent ferro- directly to the probability of the This is mainly graphite, which can magnetism is more complex. In sample25 being recovered by a mag- often be seen as a black scum
JOURNAL OF THE SOUTH AFRICAN INSTITUTE OF MINING AND METALLURGY DECEMBER 1974 1 Q9 floating off when fresh medium is high temperature, and acid Of saline on a large scale In heavy-medIum added to a heavy-medium circuit. water. plants21, particularly after a settled Manufacturing specifications state Quality-control techniques for the medium has stood for some time in that the proportion of non-magnetics assessment of corrosion resistance a storage vessel such as a spiral should not exceed 0,75 per cent by are based on the rusting of ferro- densifier. mass, but in practice it is found that silicon samples under controlled Some investigators have the proportion rarely exceeds 0,1 conditions, and measurement of the questioned the validity of these per cent. effect of the rusting. In the case of methods for the assessment of cor- milled ferrosilicon, the sample is rosion resistance23, and indeed it has repeatedly wetted with distilled not been possible to relate laboratory Corrosion Resistance water and dried for a fixed period at data directly with the rusting be- QC. The inclusion of 14 to 16 per cent 110 The resultant increase in haviour of the medium in an operat. silicon in the alloy results in a mass due to oxidation is then ex- ing plant. relatively high resistance to corrosion pressed as a percentage, termed the One alternative that has been or rusting. This is important for Rust Index (RI.). advocated is a plant test of the three reasons. In the past, the manufacturer's medium on pilot-plant scale in 1. Corrosion leads to loss of ferro- specification for the ground material order to fully simulate production silicon. has stated that the RI. should not conditions, although clearly it would 2. The finely-divided products of exceed 0,4 per cent for a three-cycle not be practical to employ such a corrosion tend to increase the wetting and drying procedure. Tests method for the regular quality viscosity of the medium and at the Diamond Research Labora- control of manufactured batches. tory, however, have shown that, if thus impair the separating Adhesion Losses efficiency of the process. sufficient cycles are carried out, the 3. Corrosion in situ can lead to R.I. always reaches a constant value One of the major sources of loss the cementing of ferrosilicon due to the format,ion of the passive of ferrosilicon from a production particles when they stand in oxide layer (Fig. 7). The number of plant is by adhesion to the processed water, which results in difficul- cycles required has been found to ore due to inadequate washing. The ties in starting up a heavy- vary from about two to ten depend- extent of adhesion is partly a func- medium process after a long ing on the sa'llple, although no tion of the surface characteristics of shut down. correlation was found between the both the ore and the ferrosilicon, Corrosion is a surface pheno- number of cycles required and the and it has been shown2 that adhesion menon28 and results from the electro- fineness of the ferrosilicon. A new loss is less with atomized ferrosilicon chemical oxidation of the ferrosilicon specification maximum of 1 per cent than with the milled material. As surface to produce non-magnetic h1S therefore been adopted, subject mentioned earlier, adhesion loss can iron oxides. Under static conditions, to each analysis being carried to the be significantly increased by the a passive layer is rapidly built up end-point. It has also been proposed presence of magnetic constituents in on the surface, which effectively that the rate of corrosion measured the ore. in this way could be used as a more prevents the progress of further SELECTION OF THE MOST realistic indicator of corrosion resist- corrosion. Under plant conditions, SUITABLE FERROSILlCON however, this passive layer is con- ance, in view of the fact that under GRADE FOR A GIVEN normal conditions of operation the tinually being removed by abrasion, APPLICATION which tends to accelerate the cor- passive layer is continually being rosion process. removed. A wide range of grades of ferrosilicon The irregular-shaped particles of The superior corrosion properties is now marketed in South Africa, and milled ferrosilicon have a greater of the atomized material precludes one of the first problems involved in surface area than the rounded the use of the R.I. as a measure of designing a new heavy-medium plant atomized particles, and are therefore corrosion resistance, because the is to select the correct grade of more susceptible to rusting. In R.I. for atomized ferrosilicon mea- medium for the particular separating addition, the sharp points and sured in this way rarely exceeds conditions. The selection must be crevices of the milled particles make 0,01 per cent. An alternative method based on a consideration of the ideal nucleation points for the f',or- used by Knapsack2,16 involves the following in relation to the viscosity rosion process. The resistance to measurement of the volume of and stability characteristics of the rusting of the atomized material is hydrogen evolved over a period of medium: extremely high owing to a passivity 100 hours from a fixed volume of (1) density of separation required, imparted to it during the quenching stirred medium at a given density (2) size of ore to be treated, process2. in an acid environment. The equa- (3) sharpness of separation re. The provision of adequate tions for the reactions are as foJlows: quired, and corrosion resistance becomes more Si+2 H2O-+SiO2+H2 t (4) costs. important in plants operating under 2 Fe+H2O+O2-+Fe203+H2 t The Tromp Curve abnormal conditions such as at Fe+H2O-+FeO+H2 t . high density (or in other words at The phenomenon of hydrogen The distribution curve introduced the limit of acceptable viscosity), evolution is frequently encountered by Tromp is generally accepted as a
110 DECEMBER 1974 JOURNAL OF THE SOUTH AFRICAN INSTITUTE OF MINING AND METALLURGY 2,O
1,0 CYCLONE '0
I >-
LN (SIZE) mm x = .-
Y = LN LN[ lOO 1 EUM 'I. COARSER.]
-1,0
',5 5,0 3.0 3,5 ',0 x Fig, 6-Rosin-Rammler plots of size distributions of various grades of ferrosilicon
Rust 1 00 Index ('t.) '
0,75
0,50
0,25
, CYCLE No. ~
Fig, 7-Rust Index versus number of cycles for various samples of milled ferrosilicon JOURNAL OF THE SOUTH AFRICAN INSTITUTE OF MINING AND METALLURGY DECEMBER 1974 111 Distribution measure of the efficiency of separa- Number tion in any heavy-medium system. 100 The distribution number, defined as the mass percentage of the feed 90 ore in a specific gravity interval reporting to the concentrate, is eo plotted against specific gravity (Fig. ------8). 70 The specific gravity of separation (ys) is defined as the specific gravity 60 at which the distribution number is 50. The probable error (Ep) is so ------defined as half the specific gravity at a distribution number of 75 " minus the specific gravity at a 30 distribution number of 25. In this way a vertical line would indicate perfect separation at a particular " specific gravity, whereas a horizontal 10 line would indicate no separation at all.
"1/5 ,/S "175 Specific Gravity- Density of Separation In general, for actual separations Fig. 8- The Tromp Curve above about 3,0 for static baths and ys=specific gravity of separation 3,2 for cyclones, the use of atomized y75-y25 ferrosilicon is necessary, because E p- 2 above a true pulp density of about
100
'I, olsp.gr. in lead to underllow
CYCLONE 60:- 75 2,89 -Y' = Ep = 0,033 QI!.' - 6 . '/2 mm kimberlite. 'I, ore to underllow Pulp Density.- 2,60 9/em3 = 19,' % medium sp.gr. differential = 1.1.5
50 150 D,- -ys.= 2,98 E'p = 0,071. ./. ore to under flaw 25,3% = medium sp-go diflerentia! =0,85
25
2 9 30 31 32 33 31. 35 sp.gr_-
Fig. 9- Tromp Curves for separation in a IOOmm cyclone using ISOD and Cyclone 60 ferrosilicon
112 DECEMBER 1974 JOURNAL OF THE SOUTH AFRICAN INSTITUTE OF MINING AND METALLURGY 3,0 the viscosity of the milled also be considered in relation to the cheaper. However, where a particu- material has been shown to increase size of the separating vessel itself; larly accurate separation is required rapidly to unmanageable proportions smaller vessels engender higher or where severely corrosive condi- (Fig. 2). At higher densities it is a separating forces owing to the tions are encountered, the use of question merely of selecting a grade smaller radius of rotation, and thus atomized ferrosilicon may be of atomized ferrosilicon that exhibits increase the tendency for the medium justified. adequate stability in a suspension to thicken. Finer medium is therefore Fig. 9 illustrates the results of of the required density. required for smaller cyclones. For treating a kimberlite sample in a example, in the tests reported here, 100 mm cyclone using 150D milled Size of Ore which were carried out on a 100 mm and Cyclone 60 atomized ferrosilicon. cyclone (the results are shown in The size of the ore to be treated The atomized ferrosilicon is shown Figs. 8 and 9), of all the grades of to provide a sharper separation. generally determines the type of medium currently manufactured by separating vessel to be used. Ore The density of separation, the Ep Amcor, only 1O0D, 150D, 270D, value, and the proportion of concen- pieces larger than 25 mm cannot be Cyclone 40, and Cyclone 60 could handled in a dynamic separator. trate produced are lower for the be used owing to the excessive atomized than for the milled This in turn affects the choice of specific-gravity differentials experi- medium because a more stable material, although recoveries at the enced with the coarser grades (often higher densities are almost identical. medium is required in dynamic greater than 2,0), resulting in the separation. A finer medium is neces- bulk of the ore reporting to the It should be noted that 150D and sary for dynamic separations than underflow product. Cyclone 60 grades have an almost for static separations, because the identical size distribution. From the higher forces involved in dynamic Sharpness of Separation specific-gravity differentials it can separation increase the tendency be seen that the milled grade is for the medium to thicken and In the lower density ranges, more stable than the atomized create high differentials. either milled or atomized ferrosilicon grade. In practice, to achieve the In dynamic separations, the ..ize may be used. Milled ferrosilicon is same stability of the medium, a distribution of the medium must generally preferred because it is grade of atomized ferrosilicon would
'I, of sp.gr.in food to 100 undorllow
f
75 270 D1-
"fS = 2 ,90 -6.hmm kimberlite Ep pulp0"'- donsity:- 2,60 9/cm3 = 0,018 'I, Of< 10 undorfIow = 15,3'!. modium sp.gr. differonticl 0,54 = so 2,98 "IS = Ep 0,032 = '!. Of< to underflow = 22,9'1. 1,22 modiu", sp.9r.difhf 25 2,8 2,9 3,0 3,2 3,3 3,. 3,5 spgr.- 3,' Fig, IO-Tromp Curves for separation in a IOOmm cyclone using 270D and IGOD ferrosilicon ~OURNAL OF THE SOUTH AFRICAN INSTITUTE OF MINING AND METALLURGY DECEMBER 1974 113 be selected that is finer than the also be considered. When the ferro- suspensions in dynamic separators required grade of milled ferrosilicon. silicon is added to an opening circuit, are not well understood, and it may In this way an even better separation care should be taken to ensure that be that significant improvements would be expected because the the fines are not lost by flotation. in performance could be achieved viscosities of suspensions of atomized Some plants include a mixing circuit either by changes in operating con- ferrosilicon have been shown to be for wetting the particles thoroughly ditions or by modification of the independent of size distribution in before they are added to the main characteristics (particularly size the lower density ranges (Fig. 2). heavy-medium circuit: laboratory distribution) of the ferrosilicon In cases where it has been decided tests have shown that atomized powder itself. that milled ferrosilicon will provide ferrosilicon is less prone to loss in To this end, further research is an adequate separation, it is essential this way, since it appears to be planned at the Diamond Research that the medium employed should more wettable. The settling charac- Laboratory on the stability and possess adequate stability. Fig. 10 teristics of the selected grade are viscometric properties of ferrosilicon shows the Tromp curves for the also important in relation to static suspensions and their modification, treatment of an identical kimberlite densifiers. For example, a plant- in order to promote a greater under- ore in a 100 mm cyclone using grades scale test of Cyclone 40 at Consoli- standing of the fundamental prin- 270D and lOOD at low density. The dated Diamond Mines of South West ciples of heavy-medium separation 1O0D medium is shown to be un- Africa Ltd had to be abandoned using this material, and thus to stable at this density by the high because it was found that the optimize performance. specific-gravity differential. As a ferrosilicon would not settle ade- result, a sharper cut with a lower quately in the spiral densifiers used REFERENCES cut-point specific gravity and a on the plant, even though prelimin- 1. BEETON and Du PLESSIS. A novel lower proportion of ore reporting ary indications were that an im- process for the production of metal to the underflow is obtained with proved separation was being ob- spherules for use as a heavy medium powder in ore beneficiation. (Unpub- the finer, more stable 270D medium. tained in the 600 mm heavy-medium lished). As noted earlier, it is often cyclones using this medium. 2. Roms, F. International Mineral possible to use extremely coarse Dressing Congress. Goslar May 8-11, CONCLUSION 1955. Published as supplement to grades of ferrosilicon with static- Z. fur Erzbergbou und Metallnuten- bath separators in processes where It is hoped that this paper has wesen. 3. CRAESCU, I., and DUMITRU, C. The the medium is stabilized by slimes shown that ferrosilicon is a simple rheological properties of suspensions generated by the feed or added and effective material for heavy- and their influence on benefication in artificially. This practice is not to heavy media. Revista Minealor, medium separation, subject to the vol. 20, no. 1, 1969. (Polish text). be recommended, however, because proviso that the correct grade is 4. WHITMORE, R. L. Coal preparation: it inevitably leads to an increase in selected for the particular applica- the separation efficiency of dense medium viscosity to the detriment medium baths. J. Inst. Fuel, Oct. tion; to assist potential users, some 1958. pp. 422-428. of separating efficiency. guidelines have been provided for 5. LILGE, E. 0., FREGREN, T. E., and the correct choice of grade for any PURDY, S. R. Apparent viscosities Costs of heavy media and the Driessen given duty. Cone. Trans. Instn Min. Metall., vol. At present, the price of atomized The technology of ferrosilicon 67, no. 6. 1957-58 pp. 229-249. 6. EVESON, G. F. A rheological ap- ferrosilicon is more than double manufacture is now well established, proach to certain features of dense- that of milled ferrosilicon. When and it seems unlikely that any major medium coal-cleaning plant operation. J. Oil. 001. Ohem. Assocn, Feb. 1959. a choice is being made between the advance in this field is either pos- pp. 146-179. milled and atomized material, it is sible or warranted at this stage. 7. DE VANEY, F. D.,andSHELToN,S. M. not enough simply to compare However, the literature review Properties of suspension mediums for float-and-sink concentration. U.S. prices. The overall economic advant- undertaken for the paper has clearly Bureau of Mines, R.I. 3469/R. May ages of atomized ferrosilicon should shown that the field of research into 1940. be taken into account. These include the rheological properties of ferro- 8. KIRCHBERG, VaN H. The properties of dense media and the quality of the lower consumption due to lower silicon suspensions, although rela- raw ore and their influence on the corrosion, adhesion, and magnetic- tively old in years, is nevertheless separating sharpness of a heavy . separator losses; greater efficiency medium separator. Bergakademie Frei- in its infancy as regards progress bury, no. 5. 1959. pp. 272-279. of separation; and a smaller bulk and achievement. Important contri- (German). concentrate, reducing subsequent butions have been made by a number 9. GEER, M. R., SOKASKI, M., WEST, J. M., and YANCEY, H. F. The role of treatment costs. An accurate cost of investigators, particularly in this viscosity in dense-medium coal comparison can be made only by country, but many fundamental cleaning. U.S. Bureau of Mines, R.I. full.scale plant tests, but un- questions remain unanswered. The 5354. Aug. 1957. 10. GOVIER, G. W., SHOOK, C. A., and fortunately this is not always con- solution to these problems is likely LILGE, E. O. The rheological proper- venient. to have an important bearing both ties of water suspensions of finely on grade selection and on the subdivided magnetite, galena and A NOTE ON THE HANDLING ferrosilicon. Oan. Min. Metall. Bull. OF MEDIA optimization of the performance of May 1957. pp. 261-268. heavy-medium separators. In par- 11. VALENTYIK, L. The rheological pro- perties of heavy media suspemions Finally, aspects of handling the ticular, the rheological behaviour stabilised by polymers. Trans. AIME selected grade of medium should and characteristics of ferrosilicon vol. 252. Mar. 1972. pp. 99-105. ~14 DECEMBER 1974 JOURNAL OF THE SOUTH AFRICAN INSTITUTE OF MINING AND METALLURGY 12. VAN DER W ALT, P. J., and FOURIE, Stormer viscometer. Diamond Re- 23. NAPIER-MUNN, T. J. (ed.). Proceed- A. M. Determination of the viscosity search Laboratory, Report No. E3. ings of an informal symposium on of unstable industrial suspensions 5/1.01. 27th Apr., 1973. (Unpub- ferrosilicon. Diamond Research Lab- with the aid of a Stormer Viscosi- lished). oratory, Report No. E3.5/3.01. 18th meter J. S. Afr. Inst. Min. Metall., 18. CLARKE, B. Rheology of coarse Jan., 1974. (Unpublished). Jul. 1957. settling suspensions. Trans. Inst. 24. AYERS, P. G. Demagnetising coils: a 13. VAN DER WALT, P. J., and FOURIE Chem. Engrs., vol. 45. 1967. pp. literature survey, and coil design. A. M. Calibration of a Stormer T251-T256. Diamond Research Laboratory, Re- 19. NESBITT, A. C., and LoEscH, R. P. port No. E3.6/1.01. 29th May, 1974. viscosimeter with special reference Heavy media properties: report No. to turbulent flow conditions. J. S. Afr. (Unpublished). 1. Diamond Research Laboratory, Chem. Inst., vol. 6, no. 2. 1953. 25. NAPIER-MuNN, T. J., and WILSON, pp. 36-48. Report No. N548/63. 15th May, 1963. (Unpublished). G. D. An operational study of the 14. VAN DER W ALT, P. J., and FOURIE, 20. VAN DER WALT, P. J., FOURIE, A. M., Premier HM Cones. Diamond A. M. The calibration of a Stormer and VAN DoORNUM, G. A. W. A Research Laboratory, Report No. viscosimeter modified for testing method for determining the settling E3.2/2.01. 12th Jul., 1973. (Unpub- suspensions. J. S. Afr. Chem. Inst., rate of heavy medium suspensions. lished). vol. 7, no. 2. 1953. pp. 61-78. Pretoria, Fuel Research Institute. 26. SOHMEISER,K., and UHLE, K. Calcu- 15. SCHMEISER, K., and KRAMER, H. Jun. 1959. lation of the magnetic condition of Viscosity of ferrosilicon suspensions. 21. SCHMEISER, K. Stability measure- ferrosilicon powder in a heavy Z. Erzbergb. u. Metall., Stuttgart, ments of ferrosilicon suspensions medium. Erzmetall. vol. 15. 1962. vol. 14, no. 3. 1961. pp. 132-135. (In using radioactive irradiation. Erz- pp. 247-251. (In German). German). metall., vol. 14, no. 5. 1961. pp. 27. SCHMEISER, K., UHLE, K., and 16. KOMORNICZYK,V. K., GLEISBERG,D., 235-237. (In German). FRANK, K. Magnetic properties of and REITZ, M. Viscosity, stability and 22. KLASSEN, V. I., LITOVKO, V. 1., and ferrosilicon powder. Erzmetall. vol. corrosion properties of ferromagnetic MYASNIKOV, N. F. Improvement of 13. 1960. pp. 477-484. heavy medium for sink-float separa- the physico-mechanical properties of 28. ROBINSON, F. P. A., and Du PLESSIS, tion. Erzmetall., vol. 26, no. 6. Jun. ferrosilicon suspensions with the aid D. J. Polarization and corrosion of 1973. pp. 300-304. (In German). of reagents. Soviet J. non-ferr. Metals ferrosilicon alloys for iron ore benefi- 17. WILSON, G. D. Investigation of heavy vol. 36, no. 10. 1963. pp. 17-20. (In ciation media. Corrosion, vol. 22, no. media viscosity using a modified Russian). 5. May. 1966. pp. 117-131. Discussion of the above paper T. B. BEETON* The Fluidity of the Melt Just Above tion of small quantities of substances the Freezing Point that will increase this electron-atom Introduction ratio should yield an improvement The fluidity is important in its The authors are to be congratu- in the particle shape. effect on the shape of the resulting lated on a very fine paper. It is It is interesting to note how particles. The alloy should freeze immensely gratifying to see the rapidly the surface tension of ferro- straight from the liquid into a single- obvious co-operation and consensus silicon drops as the silicon content phase component, and should not of opinion evidenced in this paper rises, and then how it takes a go through a 'mushy' or 'misch by the manufacturer, the seller, sudden plunge as the alloy reaches kristal' stage. In this respect, ferro- and a major user of ferrosilicon. As the stage of the first intermetallic silicon of 14 to 16 per cent silicon is a pioneer in the manufacture of, compound and the electron-atom very suitable, but additions of other and research into, this interesting ratio drops to a theoretical zero. metals may interfere with this rapid material here in South Africa, I I am mentioning these facts be- freezing. Small quantities of fluid- earnestly wish that this co-operation cause I feel that, although a good izers such as sulphur and phosphorus will continue in the future, as such deal of theoretical work has been can be added with some advantage, an almost unique situation can lead done along these lines, it has been but we must be careful of their only to the eventual benefit of all accompanied by very little practical effect on the second and more parties concerned. experimentation. I am sure there is important property, which is dealt I should now like to comment scope for more basic research work with below. briefly and expand a little on a few in this field. selected aspects of this paper. Surface Tension Just Above the Freezing Point Properties of the Heavy-medium Fundamental Properties The surface tension of the alloy Material The authors have listed a number just above the freezing point, like One of the properties of the of fundamental properties of the the fluidity, should be as high as material itself not mentioned by the ferrosilicon alloy that are vital to possible. authors is the basic one of hollow the success of the material as a In theory, the surface tension of particles, the so-called 'hollow balls', heavy-medium powder. I should, the solid alloy should be directly with which I would now like to however, like to draw your attention proportional to the free electron- deal briefly. to two further properties that are atom ratio of the alloy. Work Hollow Particles of great importance to the successful carried out at ISCOR many years I was led into this subject by the production of atomized material. ago confirmed that, in most cases, mention of the ISCOR process in the this theory holds true for liquid paper. alloys at a temperature close to the One of the reasons for our claim *ISCOR freezing point. Therefore, the addi- that the water-atomized product JOURNAL OF THE SOUTH AFRICAN INSTITUTE OF MINING AND METALLURGY DECEMBER 1974 11~ was superior to the steam-atomized practical field work. One thing, loss of fines would not be due was the virtual absence of hollow however, is sure, and that is that to any inherent magnetic pro- particles in the former. Why hollow an increase in carbon content will perty of the material itself. particles are produced, or not pro- adversely affect the specific gravity duced, under certain conditions re- and the corrosion resistance of the Conclusion mains an enigma. material, but this is the cheaper In conclusion, I should once more We know from experimental work method of producing ferrosilicon. like to thank the authors for a on the production of large granules Can we not find some inexpensive most informative and well set-out that the hollowness is produced by additions that will offset this effect paper, which has more than ade- the liberation of gas from the molten of carbon? quately covered the three aspects material during the formation of a Corrosion resistance during storage referred to in the title. However, solid crust on the outside of the of the wet medium can be countered before closing I should like to stress particle. The percentage of hollow by the addition of inhibitors to the the point made in this paper regard- particles produced under given storage tank or, even if freely ing the selection of the correct grade conditions usually increases with available, by freezing of the bath of ferrosilicon for the requirements increasing particle size, and appears with liquid nitrogen just sufficiently of the job. This is most important to depend on the carbon and man- to keep the temperature down before if the maximum benefits of the use ganese content of the alloy. the next start-up of the plant. of ferrosilicon, which is after all not These hollow particles, apart from Cooling of the medium is an effective a cheap material, are always to be lowering the specific gravity of the method of combating corrosion. obtained. material, tend to break open during Magnetic Properties usage and thereby cause severe Time will allow comment on only J. B. SEE* corrosion and other problems. They one more property of the material, must be taken seriously in the namely the magnetic property. We In his discussion, Dr Beeton production of a superior-quality did a great deal of work on the indicated that the high surface heavy-medium, but we do not really magnetic properties of various ferro- tensiont of ferrosilicon 14 to 16 per know enough about the detailed silicon alloys some years ago at cent silicon was important in en- conditions under which they are ISCOR, and eventually came to the suring that spherical powders could be produced. conclusion that any ferrosilicon produced by steam atomization of The material produced in an alloy of 14 to 16 per cent silicon has the liquid alloy. The aim of this induction furnace always has fewer magnetic properties that are so contribution is to discuss in more hollow particles than that produced superior to other materials, such as detail the influence of surface tension in an arc furnace, and I often wonder magnetite, that virtually any reput- on the morphology of particles what kind of material would be able make of magnetic separator or produced during the atomization of produced if we were to vacuum demagnetizer would be adequately a liquid metal by a gas. de-gas a ladle of ordinary material effective (providing, of course, that The final shape of a metal particle from the arc furnace prior to it was in good working condition). formed by the disintegration of a atomization. We never measured magnetic liquid-metal stream by a gas jet Corrosion Properties attractability, but did measure the is determined by the relative magni- Once again the dangers of hysteresis loop under high field tudes of the time required for accepting any laboratory corrosion strengths. We found very little solidification and the time required result as a criterion of possible difference between atomized and for surface tension forces to restore plant performance cannot be over- milled ferrosilicon. the particle to a spherical shape. stressed. But I should like to deal In this regard, I should like to ask Thus, all other atomization variables briefly with one aspect of corrosion the authors some questions. being constant, a higher value of resistance not mentioned by the (1) At what field strength is ferro- surface tension means that there is authors, and this is the influence of silicon saturated? more likelihood that spheres will the carbon content of the material (2) Are the differences in attract- be formed before sufficient heat on corrosion. Our experience for ability not related to particle transfer from the particle has many years past has indicated that, size and shape, and to packing occurred to 'freeze in' an intermedi- in general, the higher the carbon density in the sample? ate shape between that of the primary content, the greater the risk of (3) Is the loss of fines in any particles and a sphere. Such inter- corrosion problems. practical magnetic separator not mediate shapes have been observed Corrosion inhibitors can be, and due to the fact that, when for the atomization of iron and have been, used in some of our dealing with very small part- iron-nickel alloys by air and nitro- alloys. Materials such as copper, icles, the relative drag forces in gen2,3 as well as for the atomization nickel, or chromium have been the fluid medium become greater of ferrosilicon by air4. added in small quantities, and even than the magnetic attraction some of the more exotic metals such force and mechanically prevent *National Institute for Metallurgy. as uranium. Once again an abund- the small particles from being tAt 1550 °C, the surface tension of this alloy should be closer to that of iron! ance of laboratory work has been pulled out of the liquid flow? (about 1750 to 1870 dynjcm) than that done in this field, but not much If this is the case, then the of silicon! (720 dynjcm). 116 DECEMBER 1974 JOURNAL OF THE SOUTH AFRICAN INSTITUTE OF MINING AND METALLURGY investigations of the atomization of lead and tin by nitrogen jets5,6 Density of Void ferrosilicon indicate that these intermediate Type of bed fraction suspension shapes are formed by a particular type of disintegration mechanism. Dense random bed of equal sphores 0,4 4,6 This mechanism requires filming of 25 % expanded bed of equal spheres 0,52 3,9 the liquid stream by the gas to form a sheet of liquid that breaks Dense random bimodal distribution of spheres 0,2 5,8 up into unstable ligaments from 25 o/.,ex panded bed of bimodal distribution of spheres 0,36 4,8 which drops are formed. The studies cited above indicate that, under some circumstances, sufficient heat transfer can occur from liquid-metal particles for them lI)3 15 Cl) Cl) Cl) to solidify during a time (tsol) that ...... Cl) Cl) is less than the time required for L:- L:- a. a. the particle to spheroidize (tsPh)' lI) ... U') Since spherical particles of ferro- 2 H.. 10 .... ~ .. Cl) silicon are produced by the Knap- .. 0'1 d ...... sack process described in the paper, E o. d lI) ..o. the atomization conditions must be ..-. such that tSPh is less than tsol. ~ -. 0 ..o. 5 e!- Formulae7 are available for the .. lI) ~... ll) lI) lI) calculation of both tsol and ts ph. .0 . d : ~ d However, before these formulae can ~ ! : ~ be used, it is necessary to determine : . experimental details such as the 0 2 3 , 5 relative velocity between the part- RADIUS icles and the gas, and the precise shape of the particles immediately and stability of a bimodal distribu- Particle-size Distribution and after their formation from the bulk tion of spheroidized ferrosilicon. The Stability liquid. figure above shows a log-normal References distribution of spheres (solid line), As described in the paper, the with a superimposed distribu- settling velocity is a very convenient 1. WEAST, R. C. (ed.). Handbook of tion of smaller spheres that just fit chemistry and physics. Cleveland (Ohio), parameter for expressing the sta- Chemical Rubber Corporation, 1973. into the interstices of the larger bility of a heavy-medium suspension. pp. F 22-23, F 29. ones (dotted line). The table sum- Accurate correlations have been 2. PUTIMTSEV, B. N. Soviet powder metal- lurgy and metal ceramics, vo!. 10, no. 5. marizes the theoretical void fractions established for the hindered settling 1971. pp. 351-355. and slurry densities that would be of uniformly sized rigid particles, 3. PUTIMTSEV, B. N. Ibid., vo!. 11, no. 2. expected in beds of equal spheres and a particularly useful equation 1972. pp. 85-88. 4. SNEZHTO, P. F., et al. Ibid., vo!. 10, and in beds with the proposed bi- is that of Richardson and ZakiI: no. 4. 1971. pp. 321-324. modal distribution of spheres. It U=UtEn, 5. SEE, J. B., RuNKLE, J. C. and KING, will be noted that the theoretical T. B. Metall. Trans., vo!. 4. 1973. where U is the settling velocity of a pp. 2669-2673. slurry densities of 25 per cent particle in a suspension of voidage 6. JOHNSTON, G. H., and SEE, J. B. expanded beds increases from 3,9 fraction E, and Ut is the velocity Process engineering of pyrometallurgy. London, Institution of Mining and to 4,8 when the bimodal distribution that the same particle would have Metallurgy. 1974. pp. 1-7. is used. in free fall. n is a function of Ret, 7. NICHIPORENKO, O. S. Soviet powder metallurgy and metal ceramics, vo!. 7, the Reynolds number based on Ut. no. 12. 1968. pp. 931-933. J. DE KOK* When Ret is small, n is equal to 4,7, and it decreases to 2,4 for large Mr Collins and his co-authors have values of Ret. For our purposes, the A. W. BRYSON* given an excellent survey of the equation can be simplified to In this paper, the authors indicate production and properties of ferro- gd2 (Ps-Pt) that the particle size of manu- silicon powders and their use in u- E4,7. factured ferrosilicon can be well heavy-medium separation. 18fL represented by the Rosin-Rammler It may be useful at this stage to where d and Ps are the particle distribution, which has a unimodal look more closely at some of the diameter and density, and fL and density function. It would be of fundamentals of the rheology of Pt the fluid viscosity and density. interest to investigate properties suspensions and to discuss their This equation explains in a quali- such as slurry density, viscosity, application to these operations. tative way why the stability of a suspension is improved by smaller *University of the Witwatersrand. *University of Cape Town. particle sizes and lower voidage JOURNAL OF THE SOUTH AFRICAN INSTITUTE OF MINING AND METALLURGY DECEMBER 1974 117 fractIons. Unfortunately, the quantI- would mean, for Instance, that the careful consIaeratIon, ana In some tative agreement is not good when effective shear rate for a particle of cases the necessity for further re- applied to the data in Fig. 4 of the 1 cm diameter travelling at 1 cmfs is search in this field is clearly paper. less than 3 S-I. Unfortunately, indicated. The calculated exponents for the because of experimental difficulties, Dr Bryson's suggestion that sig- v3rious grades are as follows: most measurements of viscosity have nificantly higher pulp densities could Special coarse 7,7 been done at shear rates in the be achieved by use of a ferrosilicon Special fine 8,3 range 30 to 300 S-I, and these would size distribution having a bimodal, Cyclone 60 9,1 tend to underestimate the viscosity. instead of a unimodal, density 65D 12,6 A test on a mixed ferrosilicon function is of particular interest. 100D 17,5. sample with a modified Brookfield The possibility of modifying ferro- These high values of n indicate a viscometer has, for instance, given silicon size distributions so as to more rapid change in stability with the following result4: improve the performance of the pulp specific gravity than would be Shear Rate Apparent Viscosity medium has been considered in the expected from the Richardson and (s-l) (Cp) past; for example, it has been Zaki equation. The discrepancy is 32,0 36 16,0 52 proposed that the exclusive use of due to the wide range of particle 6,5 98 that size of ferrosilicon having the sizes in each grade. It is well known 3,1 141 maximum residence time in a heavy- that size segregation occurs during More information on viscosities at medium cyclone could improve per- hindered settling, and that density low shear rates would therefore be formance by increasing medium gradients are established in such required before we could hope to stability without unduly increasing cases2. Because of this, the zone make predictions of the behaviour viscosity. All such proposals, how- of medium of the original starting of static-bath separators from ever, would undoubtedly be difficult specific gravity can thus fall appreci- medium properties. to achieve in practice since the ably faster than indicated by the manufacturing process does not allow sedimentation velocity, and it could References the easy modification of the width be argued that the medium is less 1. RICHARDSON, J. F., and ZAKI, W. N. of the size distribution in each case. stable than indicated by this test. Trans. Instn Chem. Engrs, vo!. 32. Dr Bryson's suggestion could no The segregation effect is more pro- 1954. p. 35. nounced at higher voidage fractions, 2. WHITMORE, R. L. Brit. J. Appl. Phys., doubt be accommodated by selective vo!. 6. 1955. p. 239. blending procedures, although the and the gradients are greatest at the 3. VAN WAZER, J. R., et al. Viscosity and top of the bed, which is the region flow measurement. Interscience Pub- high level of control required to in which sedimentation velocity is lishers, 1963. achieve the exact fitting of the two 4. LEVY, C. D. (University of Cape accurately produced unimodal distri- usually measured. Town). Private communication. No quantitative information ap- butions would probably result in prohibitive manufacturing costs. In pears to be available on this pheno- K. KOMORNICZYK* menon, and it seems that it would be addition, it seems likely that the worth some further study, as the As a contribution to the discussion reduction of the void fraction by simple sedimentation velocity may of atomized versus milled ferro- this procedure would lead to an not always be an adequate measure silicon, I submit the findings of one increase in viscosity for a given pulp density. Nevertheless, both of stability. of our clients, who was last year in the position to make a one-year the production of the desired distri- The Effect of Shear Rate on comparison of two parallel-working butions and the investigation of the Viscosity Wemco drums. The feed of iron ore properties of the resultant medium would be relatively easy on a As pointed out in the paper, much to each drum was 400 tfh, and the laboratory scale, and Dr Rryson's work has been done on the viscosity feed per month to each drum was comments are of sufficient interest to of unstable suspensions. It has been 103 OOOt. The pulp density was fairly well established that such 3,0 to 3,2 gfmI. The average con- warrant further investigation. suspensions are usually pseudo- sumption of atomized ferrosilicon Dr De Kok rightly points out that plastic or shear-thinning, i.e., the was 5,8 tonnes per month, and of Zaki's equation has been found apparent viscosity increases with milled ferrosilicon 10,1 tonnes per useful in describing the hindered decreasing rates of deformation. month. settling of uniformly sized particles. Now, we are particularly interested The equation can also be expressed in the behaviour of slowly moving AUTHORS' REPLY as near-gravity particles, and the vis- U cous resistance of the medium to The authors would like to thank -=(l-c)n, these. It can readily be shown3 that the contributors for their construct- Ut the maximum shear rate in the ive and authoritative discussion of where u=settling velocity of a vicinity of a falling spherical particle this paper. The points submitted are particle in a medium of thought-provoking and worthy of volume concentration C, is equal to 3;, and the effective Ut = equivalent Stokesian value would be somewhat lower. This *Hoechst AG, Werk Knapsack. settling velocity, and 118 DECEMBER 1974 JOURNAL OF THE SOUTH AFRICAN INSTITUTE OF MINING AND METALLURGY 'h=a'b. exponent, whIch only arbItrary definitIon of stability would. a material larger than :3 mm in size). approaches 4,7 for Rey- seem to be a pointless exercise, By use of a specially designed nold's numbers of less since, as Dr De Kok rightly em- Haake viscometer recently acquired than 10-2. phasizes, the important variable by the Diamond Research Labora- However, the quantitative validity from a practical viewpoint is pulp- tory, it is hoped to carry out studies of Zaki's equation is seriously tested density change rather than settling of the viscometric characteristics under the conditions quoted in the rate. of ferrosilicon suspensions at shear paper, for the following reasons. As noted in the paper, work is rates as low as 2 S-1. (1) The equation refers only to currently being planned at the As noted in the paper, there perfect spheres; for irregular- Diamond Research Laboratory on remains significant lack of under- shaped particles, the value of n medium 'stability' as characterized standing of the rheological behaviour should be increased by 50 per by the rate of change of density, of ferrosilicon suspensions under cent. and it is hoped that reliable quantita- dynamic flow conditions (e.g. in a (2) The equation is not really valid tive data, which Dr De Kok suggests cyclone), in which the behaviour of for volume concentrations above would be worth acquiring, will result the particle being separated is about 10 per cent. A ferrosilicon from this work. complicated by the shear of the suspension of density 3,0 g/cm3 Dr De Kok's comments on the medium itself caused by its has a volume concentration of effect of shear rate on viscosity are rotational flow in the vessel. This about 35 per cent. represents almost virgin territory (3) As Dr De Kok points out, the of interest in view of the dearth of reliable data on this subject. There for any potential investigator. equation refers to uniformly certainly is evidence to indicate that sized particles, whereas a ferro- With regard to Dr Beeton's queries unstable suspensions in general are silicon suspension contains part- concerning the magnetic properties pseudoplastic, although there are icles of a very wide range of of ferrosilicon, Schmeiser et al. few direct data available on ferro- sizes. (reference 25 of the paper) quotes a (4) The substitution in the equation silicon suspensions specifically. To figure of 4000 A turns/cm for the field this characteristic must be added of settling velocities calculated intensity required to saturate ferro- the indirect indication of the from the differential-pressure silicon. Although the manufacturers presence of a yield stress in some such stability-measuring method of the Satmagan magnetic balance suspensions, a characteristic of the quoted in the paper must be specifically state that differences in Bingham Plastic class of materials. fraught with danger, since in no packing density of the sample are way can the method be assumed A knowledge of the relationship automatically compensated for, there to provide direct quantitative between shear rate and viscosity for is evidence to suggest that, over a data on particle-settling velo- a particle of known dimensions and wide range of size distributions, this city, but rather on the rate of a medium of known viscosity can be may not entirely be so. The signifi- change of pulp density at an used to calculate the smallest size cant drop in magnetic susceptibility arbitrary level in the pulp. of particle that can be effectively consistently noted for particles below In fact, at high volume concentra- treated in a static-bath separator of 38 um in size (compared with rela- tions, the properties of polydisperse known geometry and dimensions. tively consistent readings above this systems cannot be predicted with This was in fact done some years ago size) is nevertheless interesting in any confidence. Dr De Kok's remarks at the Diamond Research Labora- relation to the probable size of the concerning size segregation and the tory, where it was estimated that magnetic domains present in alloys establishment of density gradients the Premier Mine 16 ft diameter of this kind. This suggests that, are valid at low volume concentra- heavy-medium cones were likely to although there is little doubt that tions (less than 10 per cent) but lose any diamonds smaller than 7 mechanical forces are indeed partly cannot really be substantiated at mesh (2,8 mm) in size owing to the responsible for the observed prefer- high volume concentrations. Any high viscosity and resultant low ential loss of fines from magnetic attempt to quantitatively correlate settling rates prevalent for that separators, a low magnetic suscept- sedimentation velocity with such an size of stone (these units are fed with ibility may also be a cause. JOURNAL OF THE SOUTH AFRICAN INSTITUTE OF MINING AND METALLURGY DECEMBER 1974 119