▲83The Journal of The South African Institute of Mining and Metallurgy MARCH 2002

Introduction

General background

Dense medium separation is the process inwhich a heavy liquid of an intermediatedensity is used to separate minerals ofdifferent specific gravity. The heavy liquid istypically a suspension of dense powder suchas ferrosilicon or magnetite in water.Ferrosilicon is used for high-densityapplications (medium density: 3200–4200kg/m3).

A ferrosilicon suspension must have alarge fraction of solids in water to achieve highdensities. For example to achieve a mediumdensity of 4000 kg/m3, 7 kg of ferrosilicon(density approximately 7000 kg/m3) should beadded to 1 litre of water to make up 2 litres ofheavy medium.

The rheology of this thick and fast-settlingsuspension can be best described by itsviscosity and stability. Viscosity is a measureof the medium resistance to fluid flow, whilestability is a measure of the tendency of themedium to settle. These two properties arestrongly influenced by parameters such asmedium density, particle shape, particle sizedistribution, and the level of contaminationwith slimes. The viscous characteristics of thedense medium are generally non-Newtonian

(which means that its viscosity is a function ofthe shear rate), and the term apparentviscosity (at a defined shear rate) is preferred.An ideal medium has a low viscosity tomaximize separation and pump efficiency. Ahigh viscosity is undesirable because itreduces the velocity of mineral particles beingseparated, increasing the chance formisplacement and reducing the separationefficiency. A low viscosity is typically obtainedfor a low medium density, coarse particles,smooth rounded particles, and clean unconta-minated medium1.

The shape of ferrosilicon particles dependson the manufacturing process (milling oratomization). In this work, atomizedferrosilicon with a high degree of sphericity(Figure 1) was used.

Stability

Stability is a property of a suspensionconsidered as a non-homogeneous two-phasesystem; it is related to the rheology of the solidphase in an environment constituted by theliquid phase. The relative movement of thesolids in the liquid phase under mass andsurface forces determines the degree ofhomogeneity of the suspension, an importantmedium property in dense medium separation.The medium stability determines the densitygradient of the medium in the separation zoneand thus directly influences separationsharpness. Therefore, it is one of the mostimportant parameters to keep under control2.

An ideal medium has a high stability,which results from high medium densities, finemedium particles, irregularly shaped particlesand the presence of low-density contamination(ore slimes).

Influence of slimes

Slimes are the fines of the ore being treated

The stability of ferrosilicon densemedium suspensionsby J.D. Grobler*, R.F. Sandenbergh*, and P.C. Pistorius*†

†Paper written on project work carried out in partial fulfilment of B.Eng(Mining Engineering) degree

Synopsis

The stability of a ferrosilicon dense medium suspension is one ofthe most important parameters to keep under control since itdetermines the density gradient of the medium in the separationzone and thus directly influences separation sharpness. Thestability of ferrosilicon suspensions were characterized usingsettling tests employing a differential pressure method. Particlessmaller than 45 µm and especially the ore fines played an importantrole in the settling behaviour of the solids. High medium stability isfavoured by ore fines, fine medium particles, and high mediumdensities.

Keywords: dense media, ferrosilicon, dense medium stability,ore fines

* Department of Materials Science and MetallurgicalEngineering, University of Pretoria.

© The South African Institute of Mining andMetallurgy, 2002. SA ISSN 0038–223X/3.00 +0.00. This paper was first presented at the SAIMMStudent Colloquium in Oct. 2001.

The stability of ferrosilicon dense medium suspensions

and are usually smaller than 10 µm. When the ore is addedto the dense medium the slimes go into suspension, alteringthe properties of the dense medium. In order to limit thiseffect most plants make use of pre-washing screens toremove this material. However, during normal plantoperation it is inevitable that some percentage of slimes willbe present in the dense medium.

The presence of slimes is known to have a negative effecton viscosity and a positive effect on stability. One reason forthis is that the slimes or contaminants have a lower densitythan the ferrosilicon; a contaminated medium will thus needto contain a higher solids concentration to achieve the samedensity. A higher solids concentration will cause greaterresistance to medium flow1.

In recent work on the rheology of ferrosilicon densemedium suspensions, regarding viscosity1 and stability2, nocontaminant fines additions were used. Napier-Munn et al.3suggested that the effect of contamination is large, but noquantitative data is available. In the present work thestability of ferrosilicon suspensions containing fines wasinvestigated.

Experimental

Materials usedThe atomized ferrosilicon samples contained 14.5–14.6 percent (by mass) silicon. Fresh samples were used for eachtest. Samples of four different size distributions wereemployed; these are depicted in Figure 2. As the Figureshows, the four samples were characterized by havingdifferent percentages smaller than 45 µm, namely 23, 29, 35and 41 per cent.

The slimes (obtained from an iron ore mine) containedmainly hematite (78 per cent) silica (12 per cent) andalumina (9 per cent) (as found by X-ray fluorescenceanalysis), and had a size distribution of 70 per cent smallerthan 10 µm, with 34 per cent smaller that 1 µm (see Figure 3).

Method

The simplest method for the measurement of stability is toagitate a sample of medium, and then to follow the settlingrate of the mud-line (water-solid interface) under staticconditions. However, in many cases where fines are present,no clear mud-line is visible. The settling of the larger mediumparticles would also be faster and this settling would not bevisible due to the clouding of the medium by the fines.Komorniczyk et al.4 developed a method based on pressuremeasurement within the medium: the change in the averagedensity of the medium above the measuring point causes thepressure to change. A similar method was used here, withthe difference that an electronic differential pressuretransducer (Motorola MPX 12DP) was employed, rather thanthe manometer used by Komorniczyk et al.4. This transducerproduces a voltage output (which is proportional to thepressure difference); the voltage output was recorded bycomputer.

A schematic of the experimental configuration used inthis work is shown in Figure 4, which shows that thepressure at the measurement depth in the medium (whichwas 110 mm below the medium surface, above the mud-line

▲

84 MARCH 2002 The Journal of The South African Institute of Mining and Metallurgy

Figure 1—Scanning electron micrograph of atomized ferrosilicon

Figure 3—Size distribution of ore slimes (measured with a Malvern sizeanalyser)

0.1 1 10 100 1000particle size (µm)

100

80

60

40

20

0cum

ulat

ive

mas

s pe

rcen

tage

Figure 2—Size distributions of the four types of atomized ferrosiliconused to make up media (distributions measured with a Malvern sizeanalyser). The vertical broken line is at 45 µm

1 10 100 1000particle size (µm)

100

80

60

40

20

0cum

ulat

ive

mas

s pe

rcen

tage

20µm

which formed if the medium had settled completely) wastransmitted to the sensor through an air column. Thepressure at the second inlet into the differential pressuretransducer was controlled with a second air column which ledto a container of water. Not shown in Figure 4 is a collectionof valves, which were used to control the volume of air in thetubes before measurements started (largely with the aim ofensuring that the bottom of the air column in the suspensioncontainer was at the sampling depth of 110 mm). The glasscontainer for the suspension had an inside diameter of 95mm and was 400 mm deep, and was kept in a temperature-controlled water bath (maintained at 28°C). The stirrer hadtwin flat blades, of total length ca. 60 mm.

In the experiments, the medium was first suspendedthrough stirring at 1500 rpm for five minutes (longer orstronger stirring was found not to affect the results, asdiscussed below). At the end of the stirring period differentialpressure measurements started once the stirrer had beenstopped. The differential pressure was recorded every 2seconds.

The influences of the medium density (values of 3400,3600, 3800 and 4000 kg/m3), slimes content (0, 2% and 5%by mass), pH values (8, 10 and 12) and ferrosilicon particlesize distributions (23, 29, 35, 41 per cent smaller than 45 µm, see Figure 2), were evaluated.

A typical set of results (for media of different densities)is shown in Figure 5, which also compares mud linemeasurements for this system (for the case where themedium contained no slimes). It is clear that the shapes ofthe curves for the pressure measurement and mud linedisplacement are quite different. This is as expected: thepressure measurement gives the average medium density inthe volume above the sampling depth (as stated earlier, thesampling depth was 110 mm below the surface), whereas themud line is a local measurement. However, it is clear fromthis comparison that medium density has a similar effect onthese two measures of stability (with improved stability athigher medium densities).

The pressure measurements showed a sharp change inslope after an initial period of little change in relativepressure (see the arrowed points in Figure 5). The origin ofthis effect is not clear at this stage. The size of the effect wasfound to be affected by the stirring procedure: there was amajor difference in the time of the slope change, forsuspensions stirred for respectively 30 seconds and 3minutes (the change occurring earlier with the shorterstirring period). However, no difference could be detectedbetween 3 and 10 minutes stirring; for this reason, aminimum of 5 minutes of stirring was employed before allthe settling measurements.

To characterize the stability of the medium, the inverseslope of the pressure curve beyond the sharp slope changewas used; hence the units of stability used here are s/kPa.(More stable media take longer to settle, and have shallowerslopes, and hence have larger values of this stability index.)Several repeat measurements were performed for each set ofconditions, and the reproducibility was found to be good.

Results and discussion

The results are summarized in Figure 6, which shows thestability index as a function of medium density, for the four

ferrosilicon size distributions considered. These figuresconfirm that the trend visible in Figure 5 is consistent for allthe size distributions and slimes contents considered: higher-density media have higher stability (of course, these are alsoexpected to have higher viscosities).

The size distribution of the ferrosilicon particles affectsthe medium stability in the expected direction—higherstability with finer particles (again, with expected higherviscosity). High slimes contents have a striking effect onmedium stability, especially for finer medium particles andhigh medium densities. This favourable effect of slimes needto be offset against their expected deleterious effect onmedium viscosity; this is to be addressed in future work. Ashigher viscosities adversely affect the dense mediumseparation it is clear that an optimum medium stability(favoured a combination of by fine medium particles, a highmedium density and the presence of slimes) would exist for aparticular separation operation.

A change in pH had very little effect on the stability ofthe suspension (Figure 7). It should, however, be noted thatall experiments were done on freshly prepared samples which

The stability of ferrosilicon dense medium suspensions

▲85The Journal of The South African Institute of Mining and Metallurgy MARCH 2002

stirrer

open-ended

pressuresensor

tubes

air

air

suspension

water

Figure 4—Schematic of experimental set-up

200

100

0

0

-1

-2

-3

-40 200 400 600 800

t (s)

Mud

line

posi

tion

(mm

)

Cha

nge

in r

elat

ive

pres

sure

(kP

aP)

Figure 5—Correlation between mud-line movement and the change inrelative pressure at a depth of 110 mm below the medium surface.Measured for ferrosilicon media of different relative densities, as listedon the curves (atomized ferrosilicon of size 35 per cent smaller than 45 µm; medium contained no slimes; ph = 10.2; temperature of 28°C).Arrows indicate the points where the pressure curves show sharpchanges in slope

The stability of ferrosilicon dense medium suspensions

allowed little time for corrosion to take place; corrosion of theferrosilicon (which will change particle shape, and introducesolid corrosion products) is expected to be pH-dependent.

Conclusions

The following may be concluded from this work:➤ The pressure differential method is a good indicator of

the settling properties of ferrosilicon under staticconditions

➤ The use of the pressure transducer simplified datarecording and allowed for the accurate recording ofmultiple data points

➤ The settling behaviour of the ferrosilicon showed aslow initial rate followed by a faster but constantsettling rate. This second rate was used to characterizethe settling characteristics

➤ The pH of the medium did not influence the settlingcharacteristics of the ferrosilicon during short tests.This may change if more time is allowed for thecorrosion of the ferrosilicon

➤ The settling rate decreased with an increase in mediumdensity and fineness of medium. Slimes (very fine ore)had a very strong effect of decreasing the settling rate,especially at high medium densities and with finerferrosilicon particles.

It is recommended that future work include viscositymeasurements to find media where an appropriate balancebetween stability and viscosity is obtained.

Acknowledgements

The financial support of Kumba Resources and the assistanceof Johan Grobler and Jaco Vermaak are gratefullyacknowledged.

References

1. NAPIER-MUNN, T.J., DUNGLISON, M., and SHI, F. Rheology of ferrosilicondense medium suspensions. Mineral Processing and Extractive MetallurgyReview, vol. 20, no. 1–3, 1999. pp. 183–196.

2. BOZATTO, P., BEVILACQUA, P., and FERRARA, G. Static and dynamic stabilityin dense medium separation processes. Mineral Processing and ExtractiveMetallurgy Review, vol. 20, no. 1–3, 1999. pp. 197–214.

3. NAPIER-MUNN, T.J. and SCOTT, I.A. The effect of demagnetisation and orecontamination on the viscosity of the medium in a dense medium cycloneplant. Minerals Engineering, vol. 3, no. 6, 1990. pp. 607–613.

4. KOMORNICZYK, V.K., GLEISBERG, D., and REITZ, M. Viscosity, stability andcorrosion of ferromagnetic media used in sink-float separation processes.Erzmetall, vol. 26, no. 6, 1973. pp. 300–304. ◆

▲

86 MARCH 2002 The Journal of The South African Institute of Mining and Metallurgy

150

100

80

0

50

Sta

bilit

y (s

/kP

a)

0

50

050

03200 3400 3600 3800 4000 4200

Medium density (kg/m3)

(d)

(c)

(b)

(a)5% slimes2% slimes

0% slimes

Figure 6—Effect of medium density, slimes, and size distribution of theferrosilicon particles on medium stability. Size distributions, given inpercentages smaller than 45 µm are; a) 41 per cent, b) 35 per cent, c) 29 per cent, and d) 23 per cent. Measured at a temperature of 28°C and pH = 10.2

7 8 9 10 11 12 13pH

Sta

bilit

y (s

/kP

a)

150

100

50

0

5% slimes

2% slimes

0% slimes

Figure 7—Influence of pH on medium stability, for media containingdifferent fractions of slimes (atomized ferrosilicon particles, of size41 per cent smaller than 45 µm, medium density 4000 kg/m3, andtemperature of 28°C). Note that pH has little influence on the stability