The production, properties, and selection of ferrosilicon powders for ...

65D 1O0D 150D 270D

0,5 0 0 03,0 0,2 0 08,0 1,2 0,5 0

20,0 5,0 2,0 0,255 35 25 1045 65 75 9042-49 60-70 70-80 90+

The production, properties, and selection offerrosilicon powders for heavy-mediumseparation

SYNOPSIS

by B. COLLlNS*. B.Se. Eng. (Rand) (Fellow).T. J. NAPIER-MUNNt. B.Sc. (Eng.). A.R.S.M. (Visitor). andM. SCIARONEt. M.Se. Eng. (Delft) (Visitor).

There are basically two types of ferrosilicon powder available to operators of heavy-medium plants in SouthAfrica: spherical and milled grades. The production processes are described, and the properties of the two typesare examined and compared. Guidelines are given for the selection of the correct grade of ferrosiliconfor any particular application.

SAMEVATTING

Daar is basies twee so::>rte ferrosilikonpoeier beskikbaar vir diegene wat swaarmediumaanlegginge in Suid-Afrikabedryf: steriese en gemaalde grade. Die produksieproses word beskryf en die eienskappe van die twee soorte wordondersoek en vergelyk. Daar word leidrade aangegee vir die keuse van die korrekte graad ferrosilikon vir enigebepaalde gebruik.

INTRODUCTION

Ferrosilicon of 14 to 16 per centsilicon has become widely acceptedas the most suitable medium for theheavy-medium separation of oreshaving a specific gravity in therange of approximately 2,5 to 4,0.Ferrosilicon has many propertiesessential to a metal or alloy powderthat is to be usei as a heavy medium,some of the more important beingthe following:(a) resistance to abrasion,

(b) resistance to corrosion,(c) high specific gravity,

(d) magnetism, which allows easymagnetic recovery with subse-quent easy demagnetization,and

(e) low cest.Ferrosilicon contain:n~ between

14 and 16 per cent silicon is foun:!to have the optimum combinationof these propertiesl. If the siliconcontent is lower than 14 per cent,the specific gravity and magneticproperties are improved, but resist-ance to corrosion decreases rapidly.Above 16 per cent silicon, the corros-ion resistance of the alloy is notsignificantly improved, but the mag-netic properties and specific gravitydeteriorate.

*Hoechst South Africa (pty) Ltd.tDe Beers Industrial Diamond Division.:j:Metalloys Limited.

THE PRODUCTION OFFERROSILICON POWDERS

At present there are basicallytwo types of ferrosilicon availablein South Africa to the operators ofheavy-medium plants. These aremilled and spherical grades.

Milled Ferrosilicon

The standard method is to melt steelscrap, quartz, and a reductant in asubmerged-arc furnace. The moltenalloy is tapped into a sandbed,allowed to cool, and then broken intolumps, after which it is crushed intwo stages and milled to the requiredSlze range.

At the Kookfontein Works ofAmcor, a 7,5 MV A bubmerged-arcf..unace is used for the production

of milled ferrosilicon. The followingare the characeristics of this furnace:

Shell diameter 6400 mmHearth diameter 4000 mmElectrode diameter 890 mmSecondary voltage 90 to 120 vResistance 0,85 m 0Reactance 1,0 m o.

The power consumption (includingauxiliaries) for 1 tonne of ferrosilicon(15 per cent silicon) amounts to2100 kWh. The furnace is chargedon a semi-continuous basis and isusually tapped into a ladle every twohours. The temperature of the melton tapping is 1600°0.

The liquid ferrosilicon from theladle is granulated into small part-icles in a high-pressure water jet.This granulation process, developed

TABLE ISPECIFICATIONS FOR MILLED FERROSILICON 14 TO 16 PER CENT SILICON

Size Analysis

Tyler me3h 48D

+ 65+100+150+200+325-325 (typical)-325 (limits)

51530507525

20-30

Chemical and Physical Specification..

SiliconCarbonSulphurPhosphorusRust IndexNon-magneticsSpecific gravity

14-16 %1 % max.

0,05 % max.0,10% max.

1,0% max.0,75% max.6,7-6,95

JOURNAL OF THE SOUTH AFRICAN INSTITUTE OF MINING AND METALLURGY DECEMBER 1974 103

Size Analysis

Sishen ISCOR Cyclone CycloneTyler mesh Coarse Coarse Coarse Fine 60 * 40 t

+ 65 3 4 3 1 0 0+100 13 13 11 8 0 0

+150 30 31 28 18 2 0+200 49 44 40 33 7 2+325 74 67 62 55 27 10-325 (typical) 26 33 38 45 73 90-325 (limits) 23-30 30,1-35 35,1-40,5 40,6-50 65-75-400 (typical) 65 85

by Iscor, eliminates the breaking andcrushing stages and is believed tohave the added advantage of pre-venting segregation in the alloy oncooling. The granulated materialis recovered by spiral classifier anddried in a gas- fired rotary dryer.

The final powder is obtained bymilling in a ball mill in closed circuitwith an air classifier, which is setto yield a product of required sizeanalysis.

A list of the grades of milledferrosilicon available in South Africa,together with their chemical andsize specifications, is given in Table I.

Spherical Ferrosilicon

There are two methods for theproduction of spherical grades offerrosilicon.Atomized Ferrosilicon

Knapsack AG, a subsidiary ofHoechst AG, was the first to applythe atomizing process to the product-ion of ferrosilicon for heavy-mediumseparation2. This process is now alsoused in the Hymat plant at Amcor'sKookfontein works.

Ferrosilicon of 75 per cent siliconis diluted with high-grade steel scrapin an induction furnace. Character-istics of this furnace are as follows:Holding capacity 3 tonnesElectrical capacity 750 kWPower consumption 600 kWh/tType Mains frequency.

By means of a tilting device, thefurnace is tapped at approximately3-hourly intervals at a charge tem-perature of approximately 1550 °C.The molten material flows into aceramic tundish, which ensures thatthe alloy stream entering the atom-izing steam cone has a constantdiameter and velocity.

In the atomizing nozzle, thestream of molten ferrosilicon comesinto contact at its apex with a coneof steam, and the melt is immediatelybroken into fine particles, which arequenched in water.

The resultant ferrosilicon pulp isfiltered in a drum filter, dried in agas dryer, then screened for theremoval of oversize material. Thisoversize fraction (approximately 3to 10 per cent of the atomizedferrosilicon) is milled in the ball mill.

A list of the grades of atomizedferrosilicon available in South Africa,together with their chemical, physic-

104 DECEMBER 1~74

TABLE nSPECIFICATIONS FOR ATOMIZED FERROSILICON 14 TO 16 PER CENT SILICON

*Cyclone 60: + 100 mesh 0,02 % max.

+200 mesh 12% max.t Cyclone 40: Is not produced at this stage because there is no local demand.

Chemical Specification

Normal Grades Special GradesSi 14-18% 14-16%C :tI % 1 0,5 %S 0,05 % max. 0,05 % max.P 0,1 % max. 0,1 % max.A!. 10,8% 0,5% max.Ma 0,75% max. 0,75% max.Cu 0,8% max. 0,5% max.Cr. 0,5% max. 0,5% max.

Physical and Physiochemical Specification:

Content of spheres: Normal grades : 1 50 %Cyclone and special grades: 190%

Magnetic contont : 99%+Pyknometer density : 6,6-7,0 gfcm3Bulk density : 3,5-4,2 gfcm3

aI, and size specifications, is given inTable H.Spheroidized Ferrosilicon

A second method for the product-ion of spherical ferrosilicon wasdeveloped by IscorI. Test" haveshown that in many respects thismaterial is identical to that pro-duced by the Knapsack method.

Spheroidized ferrosilicon is pro-duced initially in the same inductionfurnace. However, the stream ofmolten alloy is broken up in a seriesof concentric cones of high-pressurewater. Clearly, the energy availablein this system is not as great as insteam atomizing, and consequentlythe water-granulated product iscoarser.

In fact, only a fraction of thismaterial (about 30 per cent) consti-tutes the final product. The granu-lated product is dried and screenedat 48 mesh, and the screen oversize(about 70 per cent) is milled to therequired size in a ball mill in closedcircuit with an air classifier. Themilled fraction is then spheroidized,and the two fractions are combinedin the desired proportions. Thespheroidized powder constitutes the

Cyclone Grades14-16%1,5 % max.0,0.5 % max.0,1 % max.0,5 % max.0,75% max.0,5 % max.0,5 % max.

fine fraction of the final product.In spheroidizing, the milled

powder is allowed to fall through aflame that has an oxidizing centreand a 'reducing outer zone, whichpermits each individual particle tobe melted without being oxidized.The surface tension and fluidity offerrosilicon are such that each part-icle contracts into a sphere. Thedesign of the burner preventsagglomeration, and, after spheroid-izing, the powder is quenched inwaste gas to prevent oxidation.

It is claimed that the water-granulated fraction of the finalproduct is superior to the correspond-ing fraction of the atomized product,making it possible to operate athigher pulp densities1.

THE PROPERTIES OF FERRO-SILICON 14 TO 16 PER CENT

SILICON

Premier Mine installed the first.heavy-medium separation plant inSouth Africa in 1948, the completespecification for the ferrosilicon usedin that plant being 25 per cent minus325 mesh. Since then, considerableprogress has been made both in the

,",OURNAL OF THE SOUTH AFRICAN INSTITUTE OF MINING "NO MHALLURGY

production and quality control ofmiI1ed ferrosilicon.

Notwithstanding these advances,the uSe of milled ferrosilicon islimited. The production of atomizedferrosilicon was a welcome develop-ment because it made possible theheavy-medium separation of certainores that could not be treated withmilled ferrosilicon.

Particle Shape

There is a marked difference inboth the shape and nature of theparticles of ferrosilicon powders pro-duced by grinding and by atomizing.Particles produced by grinding arerough and angular in shape, whereasparticles produced by atomizing aresmooth and largely spherical orrounded. This difference in theindividual particle shape has markedeffects on the rheological propertiesof aqueous suspensions of the twotypes of ferrosilicon.

Rheological Properties

The rheological properties ofheavy-medium suspensions play animportant part both in the heavy-

..UI

~~UJI-4:0::

0::4:UJ:I:<J>

vYiel d Stress

medium separation itself and in thehandling of the medium with regardto pumping and storage. Therheology of fast-settling suspensionsof this kind can best be described bytwo properties of the suspension:viscosity and stability.

ViscosityThe viscosity of a ferrosilicon

suspension directly affects theseparating efficiency of the simplestform of heavy-medium separator-the static bath-by influencing theterminal velocity of ore particlessettling in the medium in a mannerakin to that described by Stokeslaw for Newtonian fluids3. One in-vestigator succeeded in deriving adirect relationship between plasticviscosity and the probable error ofthe separatIOn (defined by the TrompCurve) in tests on coal separations4.The influence of medium viscosityon dynamic separators such as thecyclone or Dyna Whirlpool is lesscertain. Many authorities maintainthat the effect is negligible in relationto the high dynamic forces exertedon the ore particles. One study5,however, has suggested that viscosity

VisCOsiTY. 1"/ ~~Pois,

SHEAR STRESS ~

i"'(dyn/cm2)

Fig I-Generalized flow curves for main rheological types

,",OURNAL OF THE SOUTH AFRICAN 'N~TITUH OF MINING AND METALLURGY

does affect the performance of aheavy-medium cyclone, and thisview has been supported recently bysome limited plant studies of cycloneoperations within the diamondindustry.

The measurement of medium vis-cosity is made difficult by thenecessity of maintaining the solidparticles in homogeneous suspensionswhile the measurement is beingmade. A number of investigatorshave approached this problem withvarying degrees of success, and anumber of ingenious instrumentshave been developed for the pur-pose6, including flow-rate orcapillary-type consistometers7.8, apressure-diaphragm viscometer9,concentric cylinder devices4.10,1l,andother rotational instruments suchas the modified Stormer visco-meter12-15, which is used by Knap-sack16 and is widely used in SouthMrica.

Unfortunately, most of theseinstruments are able to provide onlya relative measure of the viscosityof a sample based on calibration withliquids of known viscosity such asglycerine. Accordingly, the detailedshear-rate and shear-stress character-istics of ferrosilicon suspensions arestill not well understood, althoughsome attempts have been made toobtain this information by suitablecalibrations and interpolation12,

It is generally agreed, however,that the rheological characteristics offerrosilicon suspensions, as of mostunstable suspensions, are non-Newtonian3. The main types ofrheological fluid are well categorized,and their flow curves are shown inFig. 1. However, it is by no meansclear to which class ferrosiliconsuspensions belong. Some worklO onvarious types of heavy media usinga rotational viscometer, reported in1957, suggests that ferrosilicon sus-pensions show pseudoplastic charac-

DECEMBER 1974 105

rPulp 100

V !Scos, tyI cP I

75

50

25

2.1. 2,8

j

3,2 3,6 1.,0

Fig. 2-Pulp specific gravity versus viscosity for various grades of ferrosilicon

teristics at low shear rates (up toabout 300S-1) and dilatent character-istics at high shear rates. Thiscould have implications in heavy-medium cyclone operations whererelatively high shear rates are ex-perienced. A study5 of the influenceof medium viscosity on DriessenCone separations did suggest that anincrease in viscosity leads to anincrease in the proportion of orereporting to the underflow.

All investigators agree on thegeneral shape of the curves for thepulp density versus viscosity offerrosilicon suspensions, e.g. Eveson6Fig. 2 shows curves for selectedgrades manufactured by Amcor inSouth Africa. In each case, a criticalvalue of pulp density (or volumeproportion of solids) is reachedbeyond which the viscosity risessharply. In general, it is inadvisableto use any given grade in anyapplication in the region beyondthis critical point, since small in-creases in density or the presenceof small quantities of fine non-magnetic contaminants can cause

106 DECEMBER 1974

Pu I P sp. go.

large increases in viscosity, withdeleterious effects on separation andpumping characteristics.

The critical value of pulp densityfor the atomized grades is shown tobe much higher than for the milledgrades, and not as clearly defined.Unlike the viscosity of the atomizedgrades, that of milled ferrosilicon isshown to increase with fineness orgrade of the ferrosilicon. Fig. 3shows that the relationship betweenviscosity and percentage minus 325mesh is in fact almost linear.

At densities above 3,0, the vis-cosities of suspensions of atomizedferrosilicon are significantly lowerthan those of the milled grades owingto the greater friction of collisionof the irregularly shaped milledparticles, which requires a largerenergy input to achieve a givendeformation of the suspension18.

This superior property of atomizedferrosilicon allows very high pulpdensities to be obtained (up to 3,8),which has made possible the heavy-medium separation of certain oresthat were not previously amenable

to the treatment.In general, it is not usual to

employ milled ferrosilicon at a pulpdensity much in excess of 3,0.Stability

The stability of a suspension canbe defined as the reciprocal of thesettling rate.

As with viscosity, a number oftechniques have been developed forthe measurement of stability. Thesimplest method is to agitate asample of the medium, allow it tostand, and time directly the settlingrate of the clear water interface7.However, in many cases, particularlywhere excessive fines are present, noclear demarcation line is visible.Another method developed by theDiamond Research Laboratory, andused by Knapsack16, employs adifferential-pressure technique, inwhich air is bubbled through thesettling medium and the change inpressure at a point in the suspensionis timed by use of a manometer19.

A more sophisticated technique,an advanced version of which iscurrently under test at the Diamond

JOURNAL OF THE SOUTH AFRICAN INSTITUTE OF MINING AND METALLURGY

MediumV"co,,!y(oPI

50'Linear Correlation Coo";o;en! 0 0,97

.0

30

20

10

35 55 75 95

'i,-'5tW' (325 m.,hl-

Fig. 3-Apparentviscosity versus fines content of milled ferrosilicon-pulp density2,80 gJcm3 (after Wilson17)

Research Laboratory, was developedat the Fuel Research Institute. Thisinvol ves the measurement, by anoptical method, of the rate ofchange in specific gravity at a pointin the settling medium2°. The changein specific gravity over a timedinterval has also been measured bya gamma-ray attenuation method21.

Another simpler, though less pre-cise, method of timing the change inspecific gravity (slightly modifiedfrom that described by Geer et al.9)is as follows. A perspex tube, about1 cm in diameter and 30 cm high,closed at the lower end, with a holedrilled in the side about one-thirdof the distance from the top andfitted with a bung, is placedvertically. The medium is thenintroduced into the tube so as tofill it completely, the top is closed,and the whole is shaken vigorouslyso as to ensure homogeneity of thesuspension. The tube is then allowed

to stand for 60 seconds, after whichthe bung in the side is removed andthe top portion of the suspension isdrained off. The tube is weighedbefore and after this procedure, and,from the dimensions and dry massof the apparatus, the change indensity of the medium in the topportion of the tube rehtive to theinitial density is easily calculated.The method, though relative andarbitrary with respect to the dimen-sions of the apparatus and the timeallowed for settling, is particularlysuitable for rapid phnt measure-ments. In the present case, thestability index 8 is given by

8=1- ( ~ -=-~2) X 100 %1

( D2 -1 ) 100 0/= D1 -1X /0 ,

where D1= initial density of themedium (g/cm3), and

D2=final density in the topportion of the tube.


. . . . ..This provides a relative scale from

100 per cent (no settling) to 0 percent (total settling of the ferrosiliconsolids to below the level of the tap-oflhole).

Stability data21 for certain gradesofferrosilicon manufactured in SouthAfrica are given in Fig. 4. It is clearthat, as in the case of viscosity,stability increases with both thefineness of the ferrosilicon and withpulp density.

For both static and dynamicseparations, the desired conditionsare thol'se of minimum viscositywith maximum stability. A lowstability, due for example to the useof too coarse a grade of ferrosilicon,will result in a large differentialbetween the specific gravities of theoverflow and underflow media fromthe separator, whether it be staticbath or cyclone. This in turn willlead to the artificial creation of a:arge body of middlings material,and consequently a poor separationwill be achieved.

As might be expected, grades ofmilled ferrosilicon show greater sta-bility than grades of atomized ferro-silicon of equivalent size analysis.Generally, it is necessary to use agrade of atomized ferrosilicon thatis finer than the milled ferrosiliconthat might be used under the sameconditions. This entails no disadvant-age, however, as the viscosities ofsuspensions of the atomized gradesare not significantly dependent ontheir fineness.

Modification of Rheology

The modification of the rheologicalproperties of ferrosilicon suspensionscan be achieved in three main ways:(1) by addition of polymeric com-

pounds or other reagents,(2) by addition of ore slimes or

clays, or(3) by demagn8tization of the circu-

lating medium.Considerable work has been carried

out on the stabilization of heavy-medium suspensions for static-bathseparators by the addition of poly-mers. It is claimed that significantimprovement in stability (with acorresponding increase in viscosity)can be achieved by the addition ofas little as 0,1 per cent by mass ofsolids of certain polymers. Thisallows the use of coarser (and

DECEMBER 1974 107

cheaper) grades of medium, easierstart-up due to reduced settling, andother advantagesll. Reductions ofviscosity by the addition of peptizingagents have also been reported22.This effect is found to be highlypH-dependent.

The addition of small quantities ofore slimes or clays has been foundto increase significantly the viscosityand stability of ferrosilicon suspen-sions. In some cases, the slimesrepresent an undesirable constituentof the ore being treated, and aproportion of the circulating mediumhas to be continuously removed andcleaned through magnetic separators.In other cases, natural slimes canbe used to stabilize a relativelycoarse grade of medium23.

The degree of magnetization of theferrosilicon particles, induced bytheir passage through magnetic Eepa-rators during normal recovery, has

Stabd;tys/em

601

50

.0

30

20

10

2,6 2,8

been found to influence significantlythe viscosity of the medium. In orderto reduce viscosity, most static-bathprocesses include a demagnetizationcoil on the pipe that returns mediumfrom the magnetic separators to themain circuit. Up to now this has notnormally been considered necessaryfor cyclone plants23.

The design of the coil with regardto attenuation characteristics andfield strength is of considerableimportance24. A recent exercise atthe Premier Mine25 established aninverse relationship between coilfield strength and the viscosity ofthe circulating medium in the heavy-medium cone circuits (Fig. 5).

In general, atomized ferrosiliconrequires a field strength for de-magnetization that is four timeshigher than that for milled ferro-sillicon.

3,0P,lp sp.gr-

3,2 3,'

108 DECEMBER 1974

Fig. 4--Stabilities of various grades offerrosi Iicon (data from Nesbitt and Loesch19)


The absolute density of a ferro-

Size Distribution

Apart from the primary classes ofatomized and milled ferrosilicon,the various manufactured grades aredistinguished and classified exclu-sively by their size distribution.Specifications for the size distribu-tions of the various grades manu-factured by Amcor are given inTables I and 11.

Sizing for quality control is carriedout by conventional screening downto 400 mesh (38 microns) and bysub-sieve methods therafter toabout 5 microns; sub-sieve tech-niques employed successfully in thiscountry and overseas include theBahco air elutriator, the Alpinesuction/sieve apparatus, and theCyclosizer.

The size distributions of bothmanufactured ferrosilicon and ferro-silicon taken from operating heavy-medium circuits have been foundto be well lepresented by theRosin-Rammler distribution, whichis given by

WR=lOO e(x/a)/),

where Wr= % mass retained,x=size (determined by

nominal sieve aper-ture), and

a,b = constants.This expression can be linearized

as follows:

100In 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-teristic size of the distribution isgenerally defined as the size atwhich x=a. This gives W R=lOO e -1

= 36,8 per cent.It has been found that, for

ferrosilicon distributions, the linearcorrelation coefficient for straightlines fitted in this way is normallyin excess of 0,95. Fig. 6 shows theRosin-Rammler plots for samplestaken from a number of Amcorgrades. The ability of ferrosiliconsize distributions to be representedin this way is extremely useful bothfor specification purposes and in theevaluation of research and planttests.

Absolute Density

silicon alloy is governed essentiallyby its metallurgical constituents,and varies within relatively narrowlimits depending on the exact pro-portions of the elements present. Thehigh density of the powder (6,7 to7,0) allows much higher pulp densi-ties to be achieved than with othercon v e n t ion a I heavy-mediummaterials.

The measurement of density isusually carried out with specificgravity bottles or pyknometers. Thedetermination must be made withgreat care because of the extremefineness of the powder. To wet theparticles completely, paraffin isusually preferred to water as thewetting medium, and the bottlesare usually evacuated, by use of awater vacuum pump, for 2 to 4hours, being gently tapped fromtime to time to remove all entrainedair. All weighing is done to 0,0001 g.The density of the paraffin is critical,since an error of only 0,005 gfcm3(about 0,6 per cent) in the measure-ment of the paraffin density willlead to an error of about 20 per centin the estimate of the ferrosilicondensity. The determinations havetherefore to be made at an accuratelycontrolled temperature. With care,a standard deviation of 0,02 gfcm3in the estimate of ferrosilicon densityis usually possible, based on threereplicate determinations.

Magnetic Properties

Being an iron alloy, ferrosilicon isinherently magnetic. This propertyof the medium permits easy recoveryand cleaning of the medium incircuit. A certain residual magnetismis induced in the medium by passagethrough a magnetic separator duringnormal operation. As noted earlier,excessive residual magnetism canhave a deleterious effect on theviscosity of the medium. In addition,if a magnetic ore is being treated,this residual magnetism in themedium will lead to excessive lossesof the medium by adhesion. Theagglomeration effects due to residualmagnetism can be observed directlyunder a binocular microscope; aninvestigation of residual magnetismwith a sedimentation photometer hasbeen reported26.

The study of inherent ferro-magnetism is more complex. In

Circulat;ngMediumViscosity

IcPI

50

Cone 4

40

30

20- 100 150 200

Oe-Mag. Coil Field Strength Igaussl-

Imeasured "ternally with medium flowing through coils 1

Cone 3

250 300 350 400 450

Fig. 5-Demagnetized coil field strength versus medium viscosity for Premier Mineheavy-medium cones (after Napier-Munn and Wilson2")-field strength: mean of

two parallel coils

most of the investigations reported(e.g. that by Schmeiser et al.27),some form of magnetic balance wasused. This type of device is used tomeasure the magnetic moment of aferromagnetic sample when satu-rated in a strong magnetic field. Acommercial instrument of this typeis the Satmagan balance, manu-factured by Outokumpu Oy ofFinland, and originally designed forthe determin'ttion of magnetite inores and slags.

Tests on the application of theSatmagan to ferrosilicon analysis atthe Diamond Research Laboratoryhave shown that, although theinstrument gives an indication ofthe proportion of magnetic ironpresent in the alloy, the reading isalso influenced by the metallurgicalspecies present and by size distribu-tion. It is probable, however, thatthe results obtained can be relateddirectly to the probability of thesample25 being recovered by a mag-


netic separator. From this point ofview it is interesting to note thatthe minus 400 mesh fraction of anyferrosilicon sample always gives alower reading than the plus 400mesh, and that atomized ferrosiliconalways gives a significantly higherreading than the milled material.This is in accordance with plantexperience that the fines in themedium are lost preferentially fromthe magnetic separators. It is alsousual to find that readings are lowerfrom dried samples obtained froman operating circuit, even after finenon-magnetic contamination hasbeen removed, probably owing to thepresence of an oxidized surface onthe ferrosilicon as a result of corros-ion.

During the manufacturing processof milled ferrosilicon, a small propor-tion of non-magnetic material be-comes mixed with the ferrosilicon.This is mainly graphite, which canoften be seen as a black scum

DECEMBER 1974 1 Q9

floating off when fresh medium isadded to a heavy-medium circuit.

Manufacturing specifications statethat the proportion of non-magneticsshould not exceed 0,75 per cent bymass, but in practice it is found thatthe proportion rarely exceeds 0,1per cent.

Corrosion Resistance

The inclusion of 14 to 16 per centsilicon in the alloy results in arelatively high resistance to corrosionor rusting. This is important forthree reasons.1. Corrosion leads to loss of ferro-

silicon.2. The finely-divided products of

corrosion tend to increase theviscosity of the medium andthus impair the separatingefficiency of the process.

3. Corrosion in situ can lead tothe cementing of ferrosiliconparticles when they stand inwater, which results in difficul-ties in starting up a heavy-medium process after a longshut down.

Corrosion is a surface pheno-menon28 and results from the electro-chemical oxidation of the ferrosiliconsurface to produce non-magneticiron oxides. Under static conditions,a passive layer is rapidly built upon the surface, which effectivelyprevents the progress of furthercorrosion. Under plant conditions,however, this passive layer is con-tinually being removed by abrasion,which tends to accelerate the cor-rosion process.

The irregular-shaped particles ofmilled ferrosilicon have a greatersurface area than the roundedatomized particles, and are thereforemore susceptible to rusting. Inaddition, the sharp points andcrevices of the milled particles makeideal nucleation points for the f',or-rosion process. The resistance torusting of the atomized material isextremely high owing to a passivityimparted to it during the quenchingprocess2.

The provision of adequatecorrosion resistance becomes moreimportant in plants operating underabnormal conditions such as athigh density (or in other words atthe limit of acceptable viscosity),

110 DECEMBER 1974

high temperature, and acid Of salinewater.

Quality-control techniques for theassessment of corrosion resistanceare based on the rusting of ferro-silicon samples under controlledconditions, and measurement of theeffect of the rusting. In the case ofmilled ferrosilicon, the sample isrepeatedly wetted with distilledwater and dried for a fixed period at110 QC. The resultant increase inmass due to oxidation is then ex-pressed as a percentage, termed theRust Index (RI.).

In the past, the manufacturer'sspecification for the ground materialhas stated that the RI. should notexceed 0,4 per cent for a three-cyclewetting and drying procedure. Testsat the Diamond Research Labora-tory, however, have shown that, ifsufficient cycles are carried out, theR.I. always reaches a constant valuedue to the format,ion of the passiveoxide layer (Fig. 7). The number ofcycles required has been found tovary from about two to ten depend-ing on the sa'llple, although nocorrelation was found between thenumber of cycles required and thefineness of the ferrosilicon. A newspecification maximum of 1 per centh1S therefore been adopted, subjectto each analysis being carried to theend-point. It has also been proposedthat the rate of corrosion measuredin this way could be used as a morerealistic indicator of corrosion resist-ance, in view of the fact that undernormal conditions of operation thepassive layer is continually beingremoved.

The superior corrosion propertiesof the atomized material precludesthe use of the R.I. as a measure ofcorrosion resistance, because theR.I. for atomized ferrosilicon mea-sured in this way rarely exceeds0,01 per cent. An alternative methodused by Knapsack2,16 involves themeasurement of the volume ofhydrogen evolved over a period of100 hours from a fixed volume ofstirred medium at a given densityin an acid environment. The equa-tions for the reactions are as foJlows:

Si+2 H2O-+SiO2+H2 t2 Fe+H2O+O2-+Fe203+H2 tFe+H2O-+FeO+H2 t .The phenomenon of hydrogen

evolution is frequently encountered

on a large scale In heavy-medIumplants21, particularly after a settledmedium has stood for some time ina storage vessel such as a spiraldensifier.

Some investigators havequestioned the validity of thesemethods for the assessment of cor-rosion resistance23, and indeed it hasnot been possible to relate laboratorydata directly with the rusting be-haviour of the medium in an operat.ing plant.

One alternative that has beenadvocated is a plant test of themedium on pilot-plant scale inorder to fully simulate productionconditions, although clearly it wouldnot be practical to employ such amethod for the regular qualitycontrol of manufactured batches.

Adhesion Losses

One of the major sources of lossof ferrosilicon from a productionplant is by adhesion to the processedore due to inadequate washing. Theextent of adhesion is partly a func-tion of the surface characteristics ofboth the ore and the ferrosilicon,and it has been shown2 that adhesionloss is less with atomized ferrosiliconthan with the milled material. Asmentioned earlier, adhesion loss canbe significantly increased by thepresence of magnetic constituents inthe ore.

SELECTION OF THE MOSTSUITABLE FERROSILlCON

GRADE FOR A GIVENAPPLICATION

A wide range of grades of ferrosiliconis now marketed in South Africa, andone of the first problems involved indesigning a new heavy-medium plantis to select the correct grade ofmedium for the particular separatingconditions. The selection must bebased on a consideration of thefollowing in relation to the viscosityand stability characteristics of themedium:(1) density of separation required,(2) size of ore to be treated,(3) sharpness of separation re.

quired, and(4) costs.

The Tromp Curve

The distribution curve introducedby Tromp is generally accepted as a


1,0

CYCLONE '0

2,O

I>-

x = LN (SIZE) .- mm

Y = LN LN[ lOO 1EUM 'I. COARSER.]

-1,0

3.0 3,5 ',0 x ',5 5,0

Fig, 6-Rosin-Rammler plots of size distributions of various grades of ferrosilicon

Rust 1 00Index'('t.)

0,75

,CYCLE No. ~

0,50

0,25

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

'I, olsp.gr.in leadto underllow

DistributionNumber

100

90

eo

--------70

60

so -------

measure of the efficiency of separa-tion in any heavy-medium system.The distribution number, definedas the mass percentage of the feedore in a specific gravity intervalreporting to the concentrate, isplotted against specific gravity (Fig.8).

The specific gravity of separation(ys) is defined as the specific gravityat which the distribution number is50. The probable error (Ep) isdefined as half the specific gravityat a distribution number of 75minus the specific gravity at adistribution number of 25. In thisway a vertical line would indicateperfect separation at a particularspecific gravity, whereas a horizontalline would indicate no separationat all.

Specific Gravity- Density of Separation

In general, for actual separationsabove about 3,0 for static baths and3,2 for cyclones, the use of atomizedferrosilicon is necessary, becauseabove a true pulp density of about

"30

Fig. 8- The Tromp Curve

ys=specific gravity of separation

Ey75-y25

p-2

QI!.' - 6 . '/2 mm kimberlite.Pulp Density.- 2,60 9/em3

150 D,- -ys.= 2,98

E'p = 0,071../. ore to under flaw

=25,3%

medium sp-go diflerentia! = 0,85

30 32 33sp.gr_-

35

"

10

"1/5 ,/S "175

100

75CYCLONE 60:-

-Y' =2,89Ep = 0,033

31.31

112 DECEMBER 1974

Fig. 9- Tromp Curves for separation in a IOOmm cyclone using ISOD and Cyclone 60 ferrosilicon


'I, ore to underllow= 19,' %

medium sp.gr.differential = 1.1.5

50

25

2 9

3,0 the viscosity of the milledmaterial has been shown to increaserapidly to unmanageable proportions(Fig. 2). At higher densities it is aquestion merely of selecting a gradeof atomized ferrosilicon that exhibitsadequate stability in a suspensionof the required density.

Size of Ore

The size of the ore to be treatedgenerally determines the type ofseparating vessel to be used. Orepieces larger than 25 mm cannot behandled in a dynamic separator.This in turn affects the choice ofmedium because a more stablemedium is required in dynamicseparation. A finer medium is neces-sary for dynamic separations thanfor static separations, because thehigher forces involved in dynamicseparation increase the tendencyfor the medium to thicken andcreate high differentials.

In dynamic separations, the ..izedistribution of the medium must

'I, of sp.gr.infoodto 100

undorllow

f

75 270 D1-

"fS = 2 ,90Ep

= 0,018

'I, Of< 10 undorfIow= 15,3'!.

modium sp.gr.differonticl

= 0,54

so

25

2,8 2,9

also be considered in relation to thesize of the separating vessel itself;smaller vessels engender higherseparating forces owing to thesmaller radius of rotation, and thusincrease the tendency for the mediumto thicken. Finer medium is thereforerequired for smaller cyclones. Forexample, in the tests reported here,which were carried out on a 100 mmcyclone (the results are shown inFigs. 8 and 9), of all the grades ofmedium currently manufactured byAmcor, only 1O0D, 150D, 270D,Cyclone 40, and Cyclone 60 couldbe used owing to the excessivespecific-gravity differentials experi-enced with the coarser grades (oftengreater than 2,0), resulting in thebulk of the ore reporting to theunderflow product.

Sharpness of Separation

In the lower density ranges,either milled or atomized ferrosiliconmay be used. Milled ferrosilicon isgenerally preferred because it is

"IS = 2,98Ep

= 0,032

'!. Of< to underflow = 22,9'1.modiu", sp.9r.difhf<nllcl = 1,22

3,0 3,'spgr.-3,2 3,.

cheaper. However, where a particu-larly accurate separation is requiredor where severely corrosive condi-tions are encountered, the use ofatomized ferrosilicon may bejustified.

Fig. 9 illustrates the results oftreating a kimberlite sample in a100 mm cyclone using 150D milledand Cyclone 60 atomized ferrosilicon.The atomized ferrosilicon is shownto provide a sharper separation.The density of separation, the Epvalue, and the proportion of concen-trate produced are lower for theatomized than for the milledmaterial, although recoveries at thehigher densities are almost identical.

It should be noted that 150D andCyclone 60 grades have an almostidentical size distribution. From thespecific-gravity differentials it canbe seen that the milled grade ismore stable than the atomizedgrade. In practice, to achieve thesame stability of the medium, agrade of atomized ferrosilicon would

0"'- -6.hmm kimberlitepulp donsity:- 2,60 9/cm3

3,5

Fig, IO-Tromp Curves for separation in a IOOmm cyclone using 270D and IGOD ferrosilicon

3,3

~OURNAL OF THE SOUTH AFRICAN INSTITUTE OF MINING AND METALLURGY DECEMBER 1974 113

be selected that is finer than therequired grade of milled ferrosilicon.In this way an even better separationwould be expected because theviscosities of suspensions of atomizedferrosilicon have been shown to beindependent of size distribution inthe lower density ranges (Fig. 2).

In cases where it has been decidedthat milled ferrosilicon will providean adequate separation, it is essentialthat the medium employed shouldpossess adequate stability. Fig. 10shows the Tromp curves for thetreatment of an identical kimberliteore in a 100 mm cyclone using grades270D and lOOD at low density. The1O0D medium is shown to be un-stable at this density by the highspecific-gravity differential. As aresult, a sharper cut with a lowercut-point specific gravity and alower proportion of ore reportingto the underflow is obtained withthe finer, more stable 270D medium.

As noted earlier, it is oftenpossible to use extremely coarsegrades of ferrosilicon with static-bath separators in processes wherethe medium is stabilized by slimesgenerated by the feed or addedartificially. This practice is not tobe recommended, however, becauseit inevitably leads to an increase inmedium viscosity to the detrimentof separating efficiency.

Costs

At present, the price of atomizedferrosilicon is more than doublethat of milled ferrosilicon. Whena choice is being made between themilled and atomized material, it isnot enough simply to compareprices. The overall economic advant-ages of atomized ferrosilicon shouldbe taken into account. These includelower consumption due to lowercorrosion, adhesion, and magnetic-

.separator losses; greater efficiencyof separation; and a smaller bulkconcentrate, reducing subsequenttreatment costs. An accurate costcomparison can be made only byfull.scale plant tests, but un-fortunately this is not always con-venient.

A NOTE ON THE HANDLINGOF MEDIA

Finally, aspects of handling theselected grade of medium should

~14 DECEMBER 1974

also be considered. When the ferro-silicon is added to an opening circuit,care should be taken to ensure thatthe fines are not lost by flotation.Some plants include a mixing circuitfor wetting the particles thoroughlybefore they are added to the mainheavy-medium circuit: laboratorytests have shown that atomizedferrosilicon is less prone to loss inthis way, since it appears to bemore wettable. The settling charac-teristics of the selected grade arealso important in relation to staticdensifiers. For example, a plant-scale test of Cyclone 40 at Consoli-dated Diamond Mines of South WestAfrica Ltd had to be abandonedbecause it was found that theferrosilicon would not settle ade-quately in the spiral densifiers usedon the plant, even though prelimin-ary indications were that an im-proved separation was being ob-tained in the 600 mm heavy-mediumcyclones using this medium.

CONCLUSION

It is hoped that this paper hasshown that ferrosilicon is a simpleand effective material for heavy-medium separation, subject to theproviso that the correct grade isselected for the particular applica-tion; to assist potential users, someguidelines have been provided forthe correct choice of grade for anygiven duty.

The technology of ferrosiliconmanufacture is now well established,and it seems unlikely that any majoradvance in this field is either pos-sible or warranted at this stage.However, the literature reviewundertaken for the paper has clearlyshown that the field of research intothe rheological properties of ferro-silicon suspensions, although rela-tively old in years, is neverthelessin its infancy as regards progressand achievement. Important contri-butions have been made by a numberof investigators, particularly in thiscountry, but many fundamentalquestions remain unanswered. Thesolution to these problems is likelyto have an important bearing bothon grade selection and on theoptimization of the performance ofheavy-medium separators. In par-ticular, the rheological behaviourand characteristics of ferrosilicon

suspensions in dynamic separatorsare not well understood, and it maybe that significant improvementsin performance could be achievedeither by changes in operating con-ditions or by modification of thecharacteristics (particularly sizedistribution) of the ferrosiliconpowder itself.

To this end, further research isplanned at the Diamond ResearchLaboratory on the stability andviscometric properties of ferrosiliconsuspensions and their modification,in order to promote a greater under-standing of the fundamental prin-ciples of heavy-medium separationusing this material, and thus tooptimize performance.

REFERENCES

1. BEETON and Du PLESSIS. A novelprocess for the production of metalspherules for use as a heavy mediumpowder in ore beneficiation. (Unpub-lished).

2. Roms, F. International MineralDressing Congress. Goslar May 8-11,1955. Published as supplement toZ. fur Erzbergbou und Metallnuten-wesen.

3. CRAESCU, I., and DUMITRU, C. Therheological properties of suspensionsand their influence on benefication inheavy media. Revista Minealor,vol. 20, no. 1, 1969. (Polish text).

4. WHITMORE, R. L. Coal preparation:the separation efficiency of densemedium baths. J. Inst. Fuel, Oct.1958. pp. 422-428.

5. LILGE, E. 0., FREGREN, T. E., andPURDY, S. R. Apparent viscositiesof heavy media and the DriessenCone. Trans. Instn Min. Metall., vol.67, no. 6. 1957-58 pp. 229-249.

6. EVESON, G. F. A rheological ap-proach to certain features of dense-medium coal-cleaning plant operation.J. Oil. 001. Ohem. Assocn, Feb. 1959.pp. 146-179.

7. DE VANEY, F. D.,andSHELToN,S. M.Properties of suspension mediums forfloat-and-sink concentration. U.S.Bureau of Mines, R.I. 3469/R. May1940.

8. KIRCHBERG, VaN H. The propertiesof dense media and the quality of theraw ore and their influence on theseparating sharpness of a heavymedium separator. Bergakademie Frei-bury, no. 5. 1959. pp. 272-279.(German).

9. GEER, M. R., SOKASKI, M., WEST,J. M., and YANCEY, H. F. The role ofviscosity in dense-medium coalcleaning. U.S. Bureau of Mines, R.I.5354. Aug. 1957.

10. GOVIER, G. W., SHOOK, C. A., andLILGE, E. O. The rheological proper-ties of water suspensions of finelysubdivided magnetite, galena andferrosilicon. Oan. Min. Metall. Bull.May 1957. pp. 261-268.

11. VALENTYIK, L. The rheological pro-perties of heavy media suspemionsstabilised by polymers. Trans. AIMEvol. 252. Mar. 1972. pp. 99-105.


12. VAN DER W ALT, P. J., and FOURIE,A. M. Determination of the viscosityof unstable industrial suspensionswith the aid of a Stormer Viscosi-meter J. S. Afr. Inst. Min. Metall.,Jul. 1957.

13. VAN DER WALT, P. J., and FOURIEA. M. Calibration of a Stormerviscosimeter with special referenceto turbulent flow conditions. J. S. Afr.Chem. Inst., vol. 6, no. 2. 1953.pp. 36-48.

14. VAN DER W ALT, P. J., and FOURIE,A. M. The calibration of a Stormerviscosimeter modified for testingsuspensions. J. S. Afr. Chem. Inst.,vol. 7, no. 2. 1953. pp. 61-78.

15. SCHMEISER, K., and KRAMER, H.Viscosity of ferrosilicon suspensions.Z. Erzbergb. u. Metall., Stuttgart,vol. 14, no. 3. 1961. pp. 132-135. (InGerman).

16. KOMORNICZYK,V. K., GLEISBERG,D.,and REITZ, M. Viscosity, stability andcorrosion properties of ferromagneticheavy medium for sink-float separa-tion. Erzmetall., vol. 26, no. 6. Jun.1973. pp. 300-304. (In German).

17. WILSON, G. D. Investigation of heavymedia viscosity using a modified

Stormer viscometer. Diamond Re-search Laboratory, Report No. E3.5/1.01. 27th Apr., 1973. (Unpub-lished).

18. CLARKE, B. Rheology of coarsesettling suspensions. Trans. Inst.Chem. Engrs., vol. 45. 1967. pp.T251-T256.

19. NESBITT, A. C., and LoEscH, R. P.Heavy media properties: report No.1. Diamond Research Laboratory,Report No. N548/63. 15th May, 1963.(Unpublished).

20. VAN DER WALT, P. J., FOURIE, A. M.,and VAN DoORNUM, G. A. W. Amethod for determining the settlingrate of heavy medium suspensions.Pretoria, Fuel Research Institute.Jun. 1959.

21. SCHMEISER, K. Stability measure-ments of ferrosilicon suspensionsusing radioactive irradiation. Erz-metall., vol. 14, no. 5. 1961. pp.235-237. (In German).

22. KLASSEN, V. I., LITOVKO, V. 1., andMYASNIKOV, N. F. Improvement ofthe physico-mechanical properties offerrosilicon suspensions with the aidof reagents. Soviet J. non-ferr. Metalsvol. 36, no. 10. 1963. pp. 17-20. (InRussian).

23. NAPIER-MUNN, T. J. (ed.). Proceed-ings of an informal symposium onferrosilicon. Diamond Research Lab-oratory, Report No. E3.5/3.01. 18thJan., 1974. (Unpublished).

24. AYERS, P. G. Demagnetising coils: aliterature survey, and coil design.Diamond Research Laboratory, Re-port No. E3.6/1.01. 29th May, 1974.(Unpublished).

25. NAPIER-MuNN, T. J., and WILSON,G. D. An operational study of thePremier HM Cones. DiamondResearch Laboratory, Report No.E3.2/2.01. 12th Jul., 1973. (Unpub-lished).

26. SOHMEISER,K., and UHLE, K. Calcu-lation of the magnetic condition offerrosilicon powder in a heavymedium. Erzmetall. vol. 15. 1962.pp. 247-251. (In German).

27. SCHMEISER, K., UHLE, K., andFRANK, K. Magnetic properties offerrosilicon powder. Erzmetall. vol.13. 1960. pp. 477-484.

28. ROBINSON, F. P. A., and Du PLESSIS,D. J. Polarization and corrosion offerrosilicon alloys for iron ore benefi-ciation media. Corrosion, vol. 22, no.5. May. 1966. pp. 117-131.

T. B. BEETON*

Discussion of the above paper

Introduction

The authors are to be congratu-lated on a very fine paper. It isimmensely gratifying to see theobvious co-operation and consensusof opinion evidenced in this paperby the manufacturer, the seller,and a major user of ferrosilicon. Asa pioneer in the manufacture of,and research into, this interestingmaterial here in South Africa, Iearnestly wish that this co-operationwill continue in the future, as suchan almost unique situation can leadonly to the eventual benefit of allparties concerned.

I should now like to commentbriefly and expand a little on a fewselected aspects of this paper.

Fundamental Properties

The authors have listed a numberof fundamental properties of theferrosilicon alloy that are vital tothe success of the material as aheavy-medium powder. I should,however, like to draw your attentionto two further properties that areof great importance to the successfulproduction of atomized material.

*ISCOR

The Fluidity of the Melt Just Abovethe Freezing Point

The fluidity is important in itseffect on the shape of the resultingparticles. The alloy should freezestraight from the liquid into a single-phase component, and should notgo through a 'mushy' or 'mischkristal' stage. In this respect, ferro-silicon of 14 to 16 per cent silicon isvery suitable, but additions of othermetals may interfere with this rapidfreezing. Small quantities of fluid-izers such as sulphur and phosphoruscan be added with some advantage,but we must be careful of theireffect on the second and moreimportant property, which is dealtwith below.

Surface Tension Just Above theFreezing Point

The surface tension of the alloyjust above the freezing point, likethe fluidity, should be as high aspossible.

In theory, the surface tension ofthe solid alloy should be directlyproportional to the free electron-atom ratio of the alloy. Workcarried out at ISCOR many yearsago confirmed that, in most cases,this theory holds true for liquidalloys at a temperature close to thefreezing point. Therefore, the addi-


tion of small quantities of substancesthat will increase this electron-atomratio should yield an improvementin the particle shape.

It is interesting to note howrapidly the surface tension of ferro-silicon drops as the silicon contentrises, and then how it takes asudden plunge as the alloy reachesthe stage of the first intermetalliccompound and the electron-atomratio drops to a theoretical zero.

I am mentioning these facts be-cause I feel that, although a gooddeal of theoretical work has beendone along these lines, it has beenaccompanied by very little practicalexperimentation. I am sure there isscope for more basic research workin this field.

Properties of the Heavy-mediumMaterial

One of the properties of thematerial itself not mentioned by theauthors is the basic one of hollowparticles, the so-called 'hollow balls',with which I would now like todeal briefly.Hollow Particles

I was led into this subject by themention of the ISCOR process in thepaper.

One of the reasons for our claimthat the water-atomized product

DECEMBER 1974 11~

was superior to the steam-atomizedwas the virtual absence of hollowparticles in the former. Why hollowparticles are produced, or not pro-duced, under certain conditions re-mains an enigma.

We know from experimental workon the production of large granulesthat the hollowness is produced bythe liberation of gas from the moltenmaterial during the formation of asolid crust on the outside of theparticle. The percentage of hollowparticles produced under givenconditions usually increases withincreasing particle size, and appearsto depend on the carbon and man-ganese content of the alloy.

These hollow particles, apart fromlowering the specific gravity of thematerial, tend to break open duringusage and thereby cause severecorrosion and other problems. Theymust be taken seriously in theproduction of a superior-qualityheavy-medium, but we do not reallyknow enough about the detailedconditions under which they areproduced.

The material produced in aninduction furnace always has fewerhollow particles than that producedin an arc furnace, and I often wonderwhat kind of material would beproduced if we were to vacuumde-gas a ladle of ordinary materialfrom the arc furnace prior toatomization.

Corrosion PropertiesOnce again the dangers of

accepting any laboratory corrosionresult as a criterion of possibleplant performance cannot be over-stressed. But I should like to dealbriefly with one aspect of corrosionresistance not mentioned by theauthors, and this is the influence ofthe carbon content of the materialon corrosion. Our experience formany years past has indicated that,in general, the higher the carboncontent, the greater the risk ofcorrosion problems.

Corrosion inhibitors can be, andhave been, used in some of ouralloys. Materials such as copper,nickel, or chromium have beenadded in small quantities, and evensome of the more exotic metals suchas uranium. Once again an abund-ance of laboratory work has beendone in this field, but not much

116 DECEMBER 1974

practical field work. One thing,however, is sure, and that is thatan increase in carbon content willadversely affect the specific gravityand the corrosion resistance of thematerial, but this is the cheapermethod of producing ferrosilicon.Can we not find some inexpensiveadditions that will offset this effectof carbon?

Corrosion resistance during storageof the wet medium can be counteredby the addition of inhibitors to thestorage tank or, even if freelyavailable, by freezing of the bathwith liquid nitrogen just sufficientlyto keep the temperature down beforethe next start-up of the plant.Cooling of the medium is an effectivemethod of combating corrosion.

Magnetic PropertiesTime will allow comment on only

one more property of the material,namely the magnetic property. Wedid a great deal of work on themagnetic properties of various ferro-silicon alloys some years ago atISCOR, and eventually came to theconclusion that any ferrosiliconalloy of 14 to 16 per cent silicon hasmagnetic properties that are sosuperior to other materials, such asmagnetite, that virtually any reput-able make of magnetic separator ordemagnetizer would be adequatelyeffective (providing, of course, thatit was in good working condition).

We never measured magneticattractability, but did measure thehysteresis loop under high fieldstrengths. We found very littledifference between atomized andmilled ferrosilicon.

In this regard, I should like to askthe authors some questions.(1) At what field strength is ferro-

silicon saturated?(2) Are the differences in attract-

ability not related to particlesize and shape, and to packingdensity in the sample?

(3) Is the loss of fines in anypractical magnetic separator notdue to the fact that, whendealing with very small part-icles, the relative drag forces inthe fluid medium become greaterthan the magnetic attractionforce and mechanically preventthe small particles from beingpulled out of the liquid flow?If this is the case, then the

loss of fines would not be dueto any inherent magnetic pro-perty of the material itself.

Conclusion

In conclusion, I should once morelike to thank the authors for amost informative and well set-outpaper, which has more than ade-quately covered the three aspectsreferred to in the title. However,before closing I should like to stressthe point made in this paper regard-ing the selection of the correct gradeof ferrosilicon for the requirementsof the job. This is most importantif the maximum benefits of the useof ferrosilicon, which is after all nota cheap material, are always to beobtained.

J. B. SEE*

In his discussion, Dr Beetonindicated that the high surfacetensiont of ferrosilicon 14 to 16 percent silicon was important in en-suring that spherical powders could beproduced by steam atomization ofthe liquid alloy. The aim of thiscontribution is to discuss in moredetail the influence of surface tensionon the morphology of particlesproduced during the atomization ofa liquid metal by a gas.

The final shape of a metal particleformed by the disintegration of aliquid-metal stream by a gas jetis determined by the relative magni-tudes of the time required forsolidification and the time requiredfor surface tension forces to restorethe particle to a spherical shape.Thus, all other atomization variablesbeing constant, a higher value ofsurface tension means that there ismore likelihood that spheres willbe formed before sufficient heattransfer from the particle hasoccurred to 'freeze in' an intermedi-ate shape between that of the primaryparticles and a sphere. Such inter-mediate shapes have been observedfor the atomization of iron andiron-nickel alloys by air and nitro-gen2,3 as well as for the atomization

of ferrosilicon by air4.

*National Institute for Metallurgy.tAt 1550 °C, the surface tension of thisalloy should be closer to that of iron!(about 1750 to 1870 dynjcm) than thatof silicon! (720 dynjcm).


Density ofVoid ferrosilicon

Type of bed fraction suspension

Dense random bed of equal sphores 0,4 4,6

25 % expanded bed of equal spheres 0,52 3,9

Dense random bimodal distribution of spheres 0,2 5,8

25 o/.,ex anded bed of bimodal distribution of s heres 0,36 4,8

investigations of the atomizationof lead and tin by nitrogen jets5,6indicate that these intermediateshapes are formed by a particulartype of disintegration mechanism.This mechanism requires filming ofthe liquid stream by the gas toform a sheet of liquid that breaksup into unstable ligaments fromwhich drops are formed.

The studies cited above indicatethat, under some circumstances,sufficient heat transfer can occurfrom liquid-metal particles for themto solidify during a time (tsol) thatis less than the time required forthe particle to spheroidize (tsPh)'Since spherical particles of ferro-silicon are produced by the Knap-sack process described in the paper,the atomization conditions must besuch that tSPh is less than tsol.

Formulae7 are available for thecalculation of both tsol and ts ph.However, before these formulae canbe used, it is necessary to determineexperimental details such as therelative velocity between the part-icles and the gas, and the preciseshape of the particles immediatelyafter their formation from the bulkliquid.

References

1. WEAST, R. C. (ed.). Handbook ofchemistry and physics. Cleveland (Ohio),Chemical Rubber Corporation, 1973.pp. F 22-23, F 29.

2. PUTIMTSEV, B. N. Soviet powder metal-lurgy and metal ceramics, vo!. 10, no. 5.1971. pp. 351-355.

3. PUTIMTSEV, B. N. Ibid., vo!. 11, no. 2.1972. pp. 85-88.

4. SNEZHTO, P. F., et al. Ibid., vo!. 10,no. 4. 1971. pp. 321-324.

5. SEE, J. B., RuNKLE, J. C. and KING,T. B. Metall. Trans., vo!. 4. 1973.pp. 2669-2673.

6. JOHNSTON, G. H., and SEE, J. B.Process engineering of pyrometallurgy.London, Institution of Mining andMetallurgy. 1974. pp. 1-7.

7. NICHIPORENKO, O. S. Soviet powdermetallurgy and metal ceramics, vo!. 7,no. 12. 1968. pp. 931-933.

A. W. BRYSON*

In this paper, the authors indicatethat the particle size of manu-factured ferrosilicon can be wellrepresented by the Rosin-Rammlerdistribution, which has a unimodaldensity function. It would be ofinterest to investigate propertiessuch as slurry density, viscosity,

*University of the Witwatersrand.

p p

lI)3Cl)....Cl)

L:-a.lI)

2~dElI)

...H............o.o...-...-...o.....~.. .

0 .: ~! :: .

~0

lI)lI)

d

~

0

15 Cl)Cl)....Cl)

L:-a.U')

10 ....Cl)0'1....d

5 e!-ll)lI)d

~

2 3RADIUS

5

and stability of a bimodal distribu-tion of spheroidized ferrosilicon. Thefigure above shows a log-normaldistribution of spheres (solid line),with a superimposed distribu-tion of smaller spheres that just fitinto the interstices of the largerones (dotted line). The table sum-marizes the theoretical void fractionsand slurry densities that would beexpected in beds of equal spheresand in beds with the proposed bi-modal distribution of spheres. Itwill be noted that the theoreticalslurry densities of 25 per centexpanded beds increases from 3,9to 4,8 when the bimodal distributionis used.

J. DE KOK*

Mr Collins and his co-authors havegiven an excellent survey of theproduction and properties of ferro-silicon powders and their use inheavy-medium separation.

It may be useful at this stage tolook more closely at some of thefundamentals of the rheology ofsuspensions and to discuss theirapplication to these operations.


*University of Cape Town.

,Particle-sizeStability

Distribution and

As described in the paper, thesettling velocity is a very convenientparameter for expressing the sta-bility of a heavy-medium suspension.Accurate correlations have beenestablished for the hindered settlingof uniformly sized rigid particles,and a particularly useful equationis that of Richardson and ZakiI:

U=UtEn,where U is the settling velocity of aparticle in a suspension of voidagefraction E, and Ut is the velocitythat the same particle would havein free fall. n is a function of Ret,the Reynolds number based on Ut.When Ret is small, n is equal to4,7, and it decreases to 2,4 for largevalues of Ret. For our purposes, theequation can be simplified to

u-gd2 (Ps-Pt) E4,7.

18fL

where d and Ps are the particlediameter and density, and fL and

Pt the fluid viscosity and density.This equation explains in a quali-

tative way why the stability of asuspension is improved by smallerparticle sizes and lower voidage

DECEMBER 1974 117

fractIons. Unfortunately, the quantI-tative agreement is not good whenapplied to the data in Fig. 4 of thepaper.

The calculated exponents for thev3rious grades are as follows:

Special coarse 7,7Special fine 8,3Cyclone 60 9,165D 12,6100D 17,5.These high values of n indicate a

more rapid change in stability withpulp specific gravity than would beexpected from the Richardson andZaki equation. The discrepancy isdue to the wide range of particlesizes in each grade. It is well knownthat size segregation occurs duringhindered settling, and that densitygradients are established in suchcases2. Because of this, the zoneof medium of the original startingspecific gravity can thus fall appreci-ably faster than indicated by thesedimentation velocity, and it couldbe argued that the medium is lessstable than indicated by this test.The segregation effect is more pro-nounced at higher voidage fractions,and the gradients are greatest at thetop of the bed, which is the regionin which sedimentation velocity isusually measured.

No quantitative information ap-pears to be available on this pheno-menon, and it seems that it would beworth some further study, as thesimple sedimentation velocity maynot always be an adequate measureof stability.

The Effect of Shear Rate onViscosity

As pointed out in the paper, muchwork has been done on the viscosityof unstable suspensions. It has beenfairly well established that suchsuspensions are usually pseudo-plastic or shear-thinning, i.e., theapparent viscosity increases withdecreasing rates of deformation.Now, we are particularly interestedin the behaviour of slowly movingnear-gravity particles, and the vis-cous resistance of the medium tothese. It can readily be shown3 thatthe maximum shear rate in thevicinity of a falling spherical particle

is equal to 3;, and the effective

value would be somewhat lower. This

118 DECEMBER 1974

would mean, for Instance, that theeffective shear rate for a particle of1 cm diameter travelling at 1 cmfs isless than 3 S-I. Unfortunately,because of experimental difficulties,most measurements of viscosity havebeen done at shear rates in therange 30 to 300 S-I, and these wouldtend to underestimate the viscosity.

A test on a mixed ferrosiliconsample with a modified Brookfieldviscometer has, for instance, giventhe following result4:

Shear Rate(s-l)32,016,0

6,53,1

Apparent Viscosity(Cp)365298

141

More information on viscosities atlow shear rates would therefore berequired before we could hope tomake predictions of the behaviourof static-bath separators frommedium properties.

References

1. RICHARDSON, J. F., and ZAKI, W. N.Trans. Instn Chem. Engrs, vo!. 32.1954. p. 35.

2. WHITMORE, R. L. Brit. J. Appl. Phys.,vo!. 6. 1955. p. 239.

3. VAN WAZER, J. R., et al. Viscosity andflow measurement. Interscience Pub-lishers, 1963.

4. LEVY, C. D. (University of CapeTown). Private communication.

K. KOMORNICZYK*

As a contribution to the discussionof atomized versus milled ferro-silicon, I submit the findings of oneof our clients, who was last year inthe position to make a one-yearcomparison of two parallel-workingWemco drums. The feed of iron oreto each drum was 400 tfh, and thefeed per month to each drum was103 OOOt. The pulp density was3,0 to 3,2 gfmI. The average con-sumption of atomized ferrosiliconwas 5,8 tonnes per month, and ofmilled ferrosilicon 10,1 tonnes permonth.

AUTHORS' REPLY

The authors would like to thankthe contributors for their construct-ive and authoritative discussion ofthis paper. The points submitted arethought-provoking and worthy of

*Hoechst AG, Werk Knapsack.

careful consIaeratIon, ana In somecases the necessity for further re-search in this field is clearlyindicated.

Dr Bryson's suggestion that sig-nificantly higher pulp densities couldbe achieved by use of a ferrosiliconsize distribution having a bimodal,instead of a unimodal, densityfunction is of particular interest.The possibility of modifying ferro-silicon size distributions so as toimprove the performance of themedium has been considered in thepast; for example, it has beenproposed that the exclusive use ofthat size of ferrosilicon having themaximum residence time in a heavy-medium cyclone could improve per-formance by increasing mediumstability without unduly increasingviscosity. All such proposals, how-ever, would undoubtedly be difficultto achieve in practice since themanufacturing process does not allowthe easy modification of the widthof the size distribution in each case.

Dr Bryson's suggestion could nodoubt be accommodated by selectiveblending procedures, although thehigh level of control required toachieve the exact fitting of the twoaccurately produced unimodal distri-butions would probably result inprohibitive manufacturing costs. Inaddition, it seems likely that thereduction of the void fraction bythis procedure would lead to anincrease in viscosity for a givenpulp density. Nevertheless, boththe production of the desired distri-butions and the investigation of theproperties of the resultant mediumwould be relatively easy on alaboratory scale, and Dr Rryson'scomments are of sufficient interest towarrant further investigation.

Dr De Kok rightly points out thatZaki's equation has been founduseful in describing the hinderedsettling of uniformly sized particles.The equation can also be expressedas

U-=(l-c)n,Ut

where u=settling velocity of aparticle in a medium ofvolume concentration C,

Ut = equivalent Stokesiansettling velocity, and


'h=a'b. exponent, whIch onlyapproaches 4,7 for Rey-nold's numbers of lessthan 10-2.

However, the quantitative validityof Zaki's equation is seriously testedunder the conditions quoted in thepaper, for the following reasons.(1) The equation refers only to

perfect spheres; for irregular-shaped particles, the value of nshould be increased by 50 percent.

(2) The equation is not really validfor volume concentrations aboveabout 10 per cent. A ferrosiliconsuspension of density 3,0 g/cm3has a volume concentration ofabout 35 per cent.

(3) As Dr De Kok points out, theequation refers to uniformlysized particles, whereas a ferro-silicon suspension contains part-icles of a very wide range ofsizes.

(4) The substitution in the equationof settling velocities calculatedfrom the differential-pressurestability-measuring methodquoted in the paper must befraught with danger, since in noway can the method be assumedto provide direct quantitativedata on particle-settling velo-city, but rather on the rate ofchange of pulp density at anarbitrary level in the pulp.

In fact, at high volume concentra-tions, the properties of polydispersesystems cannot be predicted withany confidence. Dr De Kok's remarksconcerning size segregation and theestablishment of density gradientsare valid at low volume concentra-tions (less than 10 per cent) butcannot really be substantiated athigh volume concentrations. Anyattempt to quantitatively correlatesedimentation velocity with such an

arbItrary definitIon of stability would.seem to be a pointless exercise,since, as Dr De Kok rightly em-phasizes, the important variablefrom a practical viewpoint is pulp-density change rather than settlingrate.

As noted in the paper, work iscurrently being planned at theDiamond Research Laboratory onmedium 'stability' as characterizedby the rate of change of density,and it is hoped that reliable quantita-tive data, which Dr De Kok suggestswould be worth acquiring, will resultfrom this work.

Dr De Kok's comments on theeffect of shear rate on viscosity areof interest in view of the dearth ofreliable data on this subject. Therecertainly is evidence to indicate thatunstable suspensions in general arepseudoplastic, although there arefew direct data available on ferro-silicon suspensions specifically. Tothis characteristic must be addedthe indirect indication of thepresence of a yield stress in some suchsuspensions, a characteristic of theBingham Plastic class of materials.

A knowledge of the relationshipbetween shear rate and viscosity fora particle of known dimensions anda medium of known viscosity can beused to calculate the smallest sizeof particle that can be effectivelytreated in a static-bath separator ofknown geometry and dimensions.This was in fact done some years agoat the Diamond Research Labora-tory, where it was estimated thatthe Premier Mine 16 ft diameterheavy-medium cones were likely tolose any diamonds smaller than 7mesh (2,8 mm) in size owing to thehigh viscosity and resultant lowsettling rates prevalent for thatsize of stone (these units are fed with


a material larger than :3 mm in size).By use of a specially designed

Haake viscometer recently acquiredby the Diamond Research Labora-tory, it is hoped to carry out studiesof the viscometric characteristicsof ferrosilicon suspensions at shearrates as low as 2 S-1.

As noted in the paper, thereremains significant lack of under-standing of the rheological behaviourof ferrosilicon suspensions underdynamic flow conditions (e.g. in acyclone), in which the behaviour ofthe particle being separated iscomplicated by the shear of themedium itself caused by itsrotational flow in the vessel. Thisrepresents almost virgin territoryfor any potential investigator.

With regard to Dr Beeton's queriesconcerning the magnetic propertiesof ferrosilicon, Schmeiser et al.(reference 25 of the paper) quotes afigure of 4000 A turns/cm for the fieldintensity required to saturate ferro-silicon. Although the manufacturersof the Satmagan magnetic balancespecifically state that differences inpacking density of the sample areautomatically compensated for, thereis evidence to suggest that, over awide range of size distributions, thismay not entirely be so. The signifi-cant drop in magnetic susceptibilityconsistently noted for particles below38 um in size (compared with rela-tively consistent readings above thissize) is nevertheless interesting inrelation to the probable size of themagnetic domains present in alloysof this kind. This suggests that,although there is little doubt thatmechanical forces are indeed partlyresponsible for the observed prefer-ential loss of fines from magneticseparators, a low magnetic suscept-ibility may also be a cause.

DECEMBER 1974 119

The production, properties, and selection of ferrosilicon powders for ...

Documents

Transcript of The production, properties, and selection of ferrosilicon powders for ...