THORIUM IN HIGH-TITANIA SLAG 7

IntroductionHeavy mineral deposits often contain relatively high levelsof radioactive elements (thorium and uranium in particular).It is difficult to obtain clean separation of ilmenite andmonazite (the main impurity mineral containing radioactiveelements), and physical intergrowths of the two mineralsare, in fact, not uncommon. As a result, ilmeniteconcentrates obtained from such deposits often contain highlevels of thorium and uranium. Because of the largenegative free energies of formation of ThO2 and U3O8 boththorium and uranium report to the slag during smelting andif an ilmenite concentrate contaminated with monazite isused to make high titania slag, the concentrations of theseelements in the slag end up about 40% higher than in theilmenite feed.

Previous work, patented by RGC Mineral Sands Limited,and subsequently confirmed by Mintek, showed thatradioactivity may be reduced by roasting an ilmeniteconcentrate with borax at 1000°C to 1100°C and leaching itwith hydrochloric acid1. It is evident that the cost of anilmenite roast/leach process will be high and there may beeconomic benefits in removing the radioactive elementsfrom the slag instead (the volume of slag to be treated issmaller than the volume of ilmenite feed).

To remove radioactive elements from the slag it isnecessary to know how these elements occur in the slag. Inother words, the distribution of uranium and thoriumbetween the various oxide, silicate and metallic phases inhigh titania slag must be determined. Once the deportmentof thorium and uranium between the various phases hasbeen determined, an appropriate process for slagpurification can be developed.

Test samplesHigh-titania slag was prepared in a pilot-scale DC furnacefrom ilmenite concentrates with different levels of uraniumand thorium. Subsequently, ten slag samples representing a

range of compositions were selected for bulk chemicalanalysis, scanning electron microscopy and micro-analysis.The samples were crushed to a maximum particle size of 1 mm and from each sample a 20-kg batch was riffled outfor analysis.

Bulk chemical compositionChemical compositions of the samples selected for testwork are given in Table I. TiO2, Al2O3, SiO2, MgO, MnO,FeO and V2O5 were analysed by ICP-OES, Na2O, CaO andCr2O3 by atomic absorption spectroscopy and U and Th byXRF (the XRF detection limit for U and Th was around 4 ppm). The ‘B’ series has a higher silica content than theothers as a result of small silica additions that were madeduring smelting. All samples have elevated alumina levelsas a result of furnace refractory contamination duringsmelting.

As expected, there is an inverse relationship between theconcentrations of TiO2 and FeO in the samples (Figure 1).The TiO2 and FeO levels are similar to the levels found inindustrial slags2. Extreme conditions of reduction underwhich reduction of more ‘refractory’ oxides (e.g., chromia,magnesia, silica, alumina and even urania and thoria) mighttake place were not explored in the study. Note that BB1and BBB1 contain more than 7% combined silica andalumina (SiO2 as a flux addition during smelting, and Al2O3from refractory contamination) and they do not fall on thegeneral trend defined by the other samples.

The U and Th concentrations of sample E2 is noteworthy.In the preparation of slag E2, monazite concentrate wasadded to the ilmenite to increase the concentrations ofradioactive elements in the slag. This was done primarily tofacilitate the detection of Th and U during the phasechemical characterization of the slag.

Radioactivity measurementsGamma ray emissions from the samples were measured

NELL, J. Thorium in high-titania slag. The 6th International Heavy Minerals Conference ‘Back to Basics’, The Southern African Institute of Mining andMetallurgy, 2007.

Thorium in high-titania slag

J. NELLMintek (now at Hatch)

Table IBulk chemical compositions of the slag samples selected for test purposes

HEAVY MINERALS 20078

with a gamma ray detector. Measurements were performedon 50-g samples, counting for 400 seconds. Repeatedmeasurements (between 3 and 6) were performed on eachsample to calculate mean values and standard deviations(Table II). There is good agreement between gamma rayemissions and the combined thorium and uraniumconcentrations of the samples (Figure 2). A regression ofthe data gives:

[1]

where NC is the normalized gamma-ray emission in unitsof counts.gram-1.seconds-1. It is important to note thatEquation [1] is strictly valid for the slag samples that wereanalysed and that it cannot be applied to other materials.

Reconciliation of the radioactivity of the slags with theconcentrations of radioactive elements was not undertaken.There can be no doubt that the radioactive decay series inthe slags are not in equilibrium owing to factors such as thevolatilization of some of the daughter elements duringsmelting.

Phase-chemical characterization of the slag

X-ray diffractometryPowder X-ray diffraction was used to identify the phasespresent in the samples. In all the samples, the mostabundant phase belongs to the M3O5 multi-component

solid-solution series (in high-titanium slags M may be Ti4+,Ti3+, Fe2+, Mg2+ and Al3+, appropriately combined tomaintain charge balance in the structure). All the slagscontain TiO2 (rutile) in trace amounts.

The shape of the background at small diffraction anglesindicates the presence of small amounts of silicate glass inthe slags. This feature was particularly prominent in the ‘B’series of samples and is consistent with the addition, duringsmelting, of SiO2 to these samples.

Micro-analysisThe slags, after tapping, consist mostly of iron-titaniumoxides (M3O5 solid solutions where M = Ti4+, Ti3+, Fe2+,Mg2+, Mn2+ and Al3+) and silicate phases. The aim of thispart of the characterization is to determine how theradioactive elements are distributed between these phases.This information is critical to develop methods for theremoval of impurities from the slag.

Proton induced X-ray emission spectroscopy (PIXE)Measurements on slag F 3, which was doped withmonazite, were performed by iThemba Laboratories. Theadvantage offered by PIXE is a relatively low detectionlimit for elements with high atomic mass, such as thoriumand uranium. However, there are also draw backs to thetechnique:

• The inability to analyse elements with an atomicnumber below about 20 (Ca) when these are containedin phases with a high mean atomic number

• The semi-quantitative nature of the analyses(measurements are standardless, although matrixcorrections are performed).

It is therefore not possible to perform comprehensive,quantitative chemical analyses of the phases present in theslag using this technique.

However, the instrument can be used to generateconcentration maps for areas of slag measuring almost 0.5 mm2. This provides a powerful visual image of thedistribution of trace elements between the coexisting phasesover a relatively large area of slag. Such maps for a grain ofslag consisting of silicate and oxide phases are shown inFigure 3 and Figure 4. In Figure 3 the silicate phases arecharacterized by high concentrations of iron and manganeseand a relatively low titanium content. The observations maybe summarized as follows:

Table IIGamma ray emissions from samples selected for test work.Results are expressed as (counts.second-1.gram-1), standard

deviations are given in parentheses

Figure 1. Relationship between the TiO2 and FeOconcentrations in the selected slags

Figure 2. Gamma ray emissions from the samples as a function ofthe combined uranium and thorium concentrations

THORIUM IN HIGH-TITANIA SLAG 9

• All the trace-elements detected by the instrument (Th,U, Y, Zr, Nb, Ta, Ba and Ce) partition preferentiallyinto silicate phases (Figure 4)

• Some of the elements, such as Th, Zr, Nb and Ta, areuniformly distributed throughout the silicate phases

while others such as U, La and Ba are concentrated inlocalized regions within silicate grains.

The partitioning of Th, U, Y, Zr, Nb, Ta, Ba and Ce intosilicate phases is noteworthy. These elements have largeatomic masses and large ionic radii. This is also acharacteristic of the radioactive daughter elementsgenerated by the decay series of thorium and uranium. Byimplication these elements should also concentrate in thesilicate phases, rather than in the oxide phases present in theslag.

The PIXE data for chromium are more ambiguous.Several grains of slag containing silicate and oxide phaseswere mapped; in one instance chromium showed a slightpreference for silicate phases but in other cases thedistribution was more uniform. Microprobe data (seebelow) are in agreement with the latter scenario, indicatinga slight preference of chromium for M3O5 oxide phases.

The thorium content of the silicate phases was found tobe approximately 4 500 ppm. This means that the thoriumcontent of the titanium oxide phases must be below 220 ppm (the bulk thorium content of the sample). This

Figure 3. Backscattered electron image (a) and distribution mapsof titanium (b), iron (c) and manganese (d), to illustrate the

location of silicate and oxide phases in a fragment of slag fromsample F3. In the backscattered electron image the silicate phases

are black in colour and the titanium oxides are grey. Elementalmaps were composed using the PIXE facility at iThemba

Laboratories

Figure 4. Elemental maps of thorium (a), uranium (b), zirconium(c) and barium (d) in silicate and titanium oxide phases in the

grains shown in Figure 3. Maps were composed using the PIXEfacility at iThemba Laboratories

Figure 4. continued. Elemental maps of tantalum (a), strontium(b), yttrium (c) and niobium (d) in silicate- and titanium oxide-

phases in the grains shown in Figure 3

Figure 4. (continued). Elemental maps of cerium (a) and lanthanum (b) in silicate and titanium oxide phases in

the grains shown in Figure 3

90000084000078000072000065000060000054000048000042000036000030000024000018000012000060000

0

90000084000078000072000065000060000054000048000042000036000030000024000018000012000060000

0

100 μm

100 μm

Ti

60056052048044040036032028024020016012080400

100 μm

Thi

300280260240220200180160140120100806040200

100 μm

Tal150014001300120011001000900800700600500400300200100

0

100 μm

Sr

150014001300120011001000900800700600500400300200100

0

100 μm

Y

6000560052004800440040003600320028002400200016001200800400

0

100 μm

Ce 30002800260024002200200018001600140012001000800600400200

0

100 μm

La

90008400780072006600600054004900420036003000240018001200800

0

100 μm

Nb

60565248444036322824201612840

100 μm

Ui

6000560052004800440040003600320028002400200016001200800400

0

100 μm

Zr120011201040960880800720640560480400320240160900

100 μm

Ba

Mn1800001680001560001440001320001200001080009600084000720006000048000360002400012000

0

100 μm

Fe

HEAVY MINERALS 200710

constrains the distribution coefficient for thorium betweensilicate and oxide phases (DTh

Silicate–oxide = wt % Th insilicate/wt % Th in oxide) to be greater than 20. Theapproximate uranium content of the silicate phases is about100 ppm but the distribution coefficient could unfortunatelynot be estimated because the uranium content of the bulksample was below the detection limit of the measurementtechnique.

Electron microprobe analysisElectron microprobe analysis using wavelength dispersiveX-ray spectroscopy (WDS) was performed to quantify thethorium content of the silicate phases present in each of thesamples (the uranium concentration in the silicate phaseswas similar to the detection limit and it was not measured).The detection limit for thorium is about 500 ppm (for acounting time of 300 seconds); well below theconcentration level in the silicate phases, but above theconcentration in M3O5 titanium-oxide phases. The thoriumcontent of the latter phases could therefore not bemeasured.

With rare exceptions, between 5 and 10 points of eachphase were analysed per sample. Mean values and standarddeviations for silicate and oxide phases are reported inTable III and Table IV respectively. Several factors maycontribute to the low analytical totals obtained for some ofthe analyses:

• The presence of sulphur (up to about 4% by weight inextreme cases), potassium (maximum of about 1% byweight), and other trace elements such as niobium inthe silicate phases. To include these elements wouldhave extended counting times well beyond 5 minutes

per analysis; this was considered to be counterproductive in terms of the objectives of the work

• The volatiliszation of sodium during analysis• Complicated interelement correction procedures due to

the high thorium and zirconium concentrations in thephases.

In most slags more than one type of silicate could beidentified on the basis of the concentrations of SiO2, MnOand FeO. The silicate phases are characterized by complexchemical compositions and a lack of crystalline textures,i.e., they have the chemical and morphologicalcharacteristics of glass. The presence of silicate glass in thesamples is consistent with the presence of amorphousmaterial detected by XRD. The glass represents silicatemelt that coexists with the titanium oxide melt at furnaceoperating temperatures.

Silica-poor glass is characterized by high concentrationsof FeO and MnO. There is also a positive correlationbetween the concentrations of MnO and FeO in the silica-poor glass. Texturally, silica-rich glass occurrs as roundeddroplets within the silica-poor glass (Figure 5), suggestingimmiscibility of two silicate melts at high temperatures.The ThO2 and ZrO2 concentrations in the different silicatephases in a particular slag are inversely proportional to thesilica content, i.e., thorium and zirconium partitionpreferentially to the silica-poor glass. These relationshipsare illustrated by the data for samples E2 and H4 in Table IIII (owing to the poor optical imaging on theinstrument used for the WDS analysis it was not possible todistinguish between different silicate phases in the slag.Consequently, no analyses of the silica-poor phase wereobtained for B4, F3 and G3).

Table IIIElectron microprobe analyses of silicate phases in the slag samples. (standard deviations are given in parentheses)

*Analyses of the silica-poor phase in E2 are incomplete due to instrumental difficulties at the time of analysis

THORIUM IN HIGH-TITANIA SLAG 11

The high thorium content of the silicate phases is one ofthe most important observations of this study. It isimportant to note that the concentration levels of thorium insilica-poor glasses are still well below the values at whichsilicate melts (glasses) become saturated with thorium.High alkali-borosilicate glasses quenched from 1250°Ccontain between 2% and 5% Th (2.3%–5.7% ThO2 byweight) in solution3; data cited by Farges 4 suggest similarsolubility levels in alumino-silicate glasses quenched fromthe same temperature. He also noted an increase in thesolubility of thorium with increasing temperature. Thesilicate glasses in the high-titanium slags are therefore notexpected to be saturated in thorium (the reason for thedifference in thorium content of the different silicateglasses will be discussed in a later section).

The distribution of thorium between silicate and oxide-phases will be reported later; as an initial observation it issufficient to note that thorium has a strong preference forsilicate phases.

The oxide phases contain MgO in quantities similar tothose present in the coexisting silicate phase but virtuallyno CaO. To the extent that this signifies a tendency foralkali earth metals with large ionic radii to partition intosilicate phases, this is highly significant (such a tendency isin agreement with the PIXE results that show preferentialpartitioning of barium to the silicate phases). Radium, analkali earth metal considered to be an important source ofradioactivity in slag, has an ionic radius of 1.40 Å, muchlarger than the radii of Ba2+ and Ca2+ (1.35 Å and 0.99 Årespectively). The observed partitioning behaviour of MgO,CaO and BaO suggests that RaO also will partition to thesilicate phases present in the slag, but even more so than thelighter alkali earth metals.

Distribution of elements between silicate andoxide phases

The distribution of elements between silicate and oxide-phases is a reflection of the thermodynamic behaviour ofthe system. The distribution coefficient for element i, Di, isdefined as:

[2]

The value of Di, depends on variables such astemperature, oxygen partial pressure (i.e., slag-grade) andmelt composition. If there is a sufficient dependence onmelt composition it would, in principle, be possible tomanipulate the behaviour of impurities during smelting. Inthe context of this study it would be highly desirable tocalculate DTh for each of the slags. Due to the lowconcentration of thorium in the oxide phases it wasunfortunately not possible to measure DTh directly, althoughminimum values were estimated from mass-balanceconstraints.

Compositional data for silicate- and oxide-phases wereused to calculate DMn, DMg, DAl, DFe, DTi and DCr for eachof the slags (Table V). Di values for Ca and Si were notcalculated since the concentrations of these elements in theM3O5 solid solutions were not accurately measured. Datafor silicate phases with different compositions in the sameslag were used to calculate average values for the Di.Where data for silica-poor glass were not available,inequalities were used to indicate the bias (samples B4, E2,F3 and G3).

Figure 5. Backscattered electron image illustrating texturalrelationships between different silicate phases in slag H4.

Droplets of a silica-rich phase are surrounded by asilica-poor phase

Table IVElectron microprobe analyses of iron-titanium oxide phases in the slag samples. (standard deviations are given in parentheses)

Silica-rich phase

Silica-poor phase

Metal

Fe-Ti-oxide

HEAVY MINERALS 200712

Elements may be classified according to their partitioncoefficients:

• Silicon, calcium (and thorium) have large partitioncoefficients which means that they are stronglyconcentrated in the silicate phases. Manganese alsoprefers silicate phases but its partition coefficient isconsiderably smaller (about 10)

• Iron, magnesium, chromium and aluminium havedistribution coefficients that are around unity, whichmeans that they are evenly distributed between silicateand oxide phases

• Titanium has a very small distribution coefficient (~ 1/100), which means that it is strongly concentratedin oxide phases.

The distribution coefficients indicate how the removal ofsilicate phases from the slag will affect the composition ofthe oxide residue. Elements that preferentially partition tosilicate phases (Di > 1) will be depleted in the residue (i.e.,the composition of the residual oxide slag will be upgradedthrough the removal of these elements), while elements forwhich Di < 1 will be enriched in the residue (providinginsoluble oxides do not precipitate during leaching).Leaching of silicate phases should therefore result in adecrease in the concentrations of SiO2, CaO and Th (to alesser extent also MnO), and an increase in theconcentration of TiO2 in the oxide residues. Theconcentrations of FeO, MgO, Al2O3 and Cr2O3 in the oxideresidues will not be significantly affected by the removal ofsilicate phases.

The chemical behaviour of thorium in high-titania slag

Distribution of thorium between silicate and oxide-phasesAlthough the distribution coefficient could not be measureddirectly, it is possible to constrain the distribution ofthorium through calculation.

First, the weight fractions of silicate and oxide phases ineach slag are calculated. The average concentrations ofMnO, SiO2, TiO2 and CaO in silicate and oxide phases areused for this purpose because of the magnitudes of thesilicate-oxide distribution coefficients of these elements.Next, the amount of thorium reporting to silicate phases (inμg/g slag) is calculated from the weight percentage silicatephases in the slag and the average thorium concentration inthe silicate phases. Standard error-propagation techniquesare used to calculate the uncertainties in these quantities.Within the uncertainty of the data, the amount of thoriumreporting to the silicate phases is enough to account for themean total thorium content of the slags and, as a result, onlya maximum value for the amount of thorium reporting tooxide phases can be calculated (this being the differencebetween the total thorium content of the slag and theamount of thorium present in silicate phases—taking theuncertainties in the data into account). Following thecalculation of a maximum value for the amount of thoriumpresent in oxide phase, a minimum value for the thoriumdistribution coefficient can be calculated. Results of thesecalculations are reported in Table VI.

Table VICalculated mass-percentage of silicate and oxide phases in each slag and the calculated thorium distribution

Table VAverage distribution coefficients (silicate oxide) for Mn, Mg, Al, Fe, Ti and Cr for each of the slags

(the range over which Di varies is reflected as an uncertainty)

Distribution coefficients for B4, E2, F3 and G3 are not well constrained due to a lack of analytical data. Symbols are used to indicate the bias resulting fromthe lack of data.

‘Bulk Th’ is the bulk Th content of the slag (from Table I)‘Th concentration in silicates’ is the average Th concentration in silicate phases determined by microprobe‘Th in slag contained in silicates’ is the amount of thorium contained in the silicate phases in the slag‘Th in slag contained in oxides’ is the amount of thorium contained in the oxide phases in the slag‘Th conc. in oxides’ indicates the maximum Th concentration in oxide phases calculated from the Th distributionDTh is the minimum value of the silicate-oxide distribution coefficient for ThThe Th distribution in E2 and H4 are subject to large uncertainties due to the large difference in the Th content of the silicate phasesCalculations are based on semi-quantitative PIXE data

THORIUM IN HIGH-TITANIA SLAG 13

The calculation is complicated by the large difference inthorium content of the silica-rich and silica-poor glasses. Inthe case of slags B4 and G3, calculation of the distributioncoefficient was not attempted because of a lack of data forthe thorium-rich, silica-poor, glass. Although WDS data arenot available for F3 either, semi-quantitative PIXE analysesfor this sample are consistent with a value for the thoriumdistribution coefficient of at least 20. Consistently largedistribution coefficients for thorium lead to an importantconclusion:

• The removal of silicate phases will also remove thebulk of the thorium and cause a significant decrease inthe thorium content of the oxide residue.

It is informative to consider the significance of a largedistribution coefficient for thorium between silicate andoxide phases. Consider a slag containing 5 wt% silicatephases and a bulk Th content of 200 ppm. AssumingDTh

Silicate–oxide = 100, it follows that the Th concentration inthe silicate and oxide phases will be 3360 ppm and 33.6ppm respectively. Therefore, if the silicate phases could befully leached from the slag, it should be possible to reducethe thorium content by at least 84%.

A large value for DTh is consistent with the generalincompatible behaviour of thorium during thecrystallization of magma with different compositions5 (anelement is said to be ‘incompatible’ if it is concentrated inthe residual melt once different oxide and silicate mineralsstart to crystallize during cooling). Unfortunately, there areno data for DTh between silicate melt and M3O5 oxidephases in the geological literature that could be used forcomparison with the results of this study.

DiscussionThe silicate phases in the slags were previously identifiedas quenched silicate melt (i.e., glass). It is therefore relevantto consider the chemistry of thorium in silicate melts as itmay suggest options for optimizing the amount of thoriumreporting to the silicate melt phases during smelting. Themost stable oxidation state for thorium is +4; with respectto thorium metal, it is also one of the most stable oxides.The stability of ThO2 relative to Th plus oxygen is evidentfrom the Gibbs free energy of formation of ThO2;approximately -870 KJ.mol-1 at 1900 K (1627°C). Bycomparison, the Gibbs free energy of formation of TiO2(relative to Ti and O2) at 1900 K is about -603 KJ.mol-1. Interms of oxygen partial pressure (PO2) it means that ThO2will form when the PO2 exceeds 10-23.8 bar, while TiO2 will(form from Ti and O2) when the PO2 exceeds 10–16.5 bar. Apractical implication of this result is that metallic thoriumwill not form during smelting and the Th content of the pigiron will be negligible.

It is not known what the effective oxygen partial pressureduring smelting is; it will also be different for the differentslags (hence the differences in Fe/Ti ratio of the slags).Phase-chemically, the composition of the M3O5 phases thatcrystallize from iron-titanium oxide slag may effectively bedescribed in terms of the FeTi2O5-Ti3O5 solid solutionseries. The activity of Ti3O5 in the slag increases withincreasing grade, and when all the Fe2+ is reduced tometallic iron (Fe0) the activity of Ti3O5 is almost unity. Anindication of the minimum oxygen partial pressure duringsmelting may therefore be obtained from the equilibriumbetween solid Ti3O5 and TiO2 (rutile):

[3]

Using thermodynamic data of Barin6, the equilibriumoxygen partial pressure for reaction 1 at 1900 K is 10-9.3 bar(this is about 1.3 log10-units more reducing than the Fe-FeObuffer). The oxidation of solid Ti4O7 to TiO2 (rutile) at1900 K takes place at a very similar oxygen partial pressure(10-9.2 bar), indicating that solid Ti3+-bearing oxide phasesare stabilized at oxygen partial pressures approximately 7log10-units more oxidizing than the Ti-TiO2 buffer.Assuming similar behaviour in the thorium-oxygen system,the formation of Th3+-bearing oxide phases at 1900 Kmight be expected at oxygen partial pressures of about 10-17bar. This is about 8 log10-units more reducing than theminimum oxygen partial pressure likely to be attainedduring smelting (10-9.3 bar). Although these calculations arenot directly applicable to the oxidation state of multi valentcations in melt phases, the data suggest that the ratio ofTh3+ to Th4+ is likely to be very low, even under the mostreducing conditions expected during smelting.

The structural environment around Th4+ in silicate glassescontaining between 1% and 3% (by weight) Th4+ wasinvestigated by Farges4 using extended X-ray absorptionfine structure (EXAFS) spectroscopy. Several conclusionsfrom the work by Farges4 are directly applicable to currentproblem:

• Th4+ is mainly six co-ordinated by oxygen with a meanbond length, 〈VITh-O.〉 of about 2.3 Å. Eight-co-ordinated Th4+ (〈VIIITh-O〉 ~ 2.4 Å) is present at highthorium concentration levels (3%) and/or whenpolymerization increases. Mean metal-oxygen bondlengths for the two six-co-ordinated cation sites (M1and M2) in orthorhombic Ti3O5-rich solid solutions〈VIM1-O〉 and 〈VIM2-O〉 are approximately 2.04 Å and2.01 Å, respectively4. The much shorter metal-oxygenbond lengths in Ti3O5-rich solid solutions mean thatthorium is unlikely to partition into oxide phasescrystallizing from the slag during cooling

• The amount of six-co-ordinated Th4+ (VITh4+) increasesin the presence of major quantities of alkali cationssuch as Na+ and K+. Should the addition of such fluxesbe practical, it will facilitate the collection of thoriumby silicate melt phases during smelting

• The bond strength around six-co-ordinated Th4+ insilicate melt or glass is high relative to that in Th-bearing minerals such as ThO2 and ThSiO4. Thisexplains in part why these phases are not observed inthe slags, even at high bulk thorium concentrationlevels (> 300 ppm)

• Increasing polymerization of the silicate glass or melt(high concentrations of SiO2), will favour higher co-ordinations around Th4+ due to a lack of non-bridgingoxygens (NBOs) to which thorium preferentiallybonds. This explains why the silica-rich glass phasecontains less thorium than the co existing, low silica,glass phase in the slags.

To summarize: data on the chemical behaviour ofthorium in silicate melts and glasses support the mass-balance calculations according to which the amount ofthorium in silicate glass phases is sufficient to account forthe total thorium content of the slags.

SummaryElements with a large ionic radius preferentially partition tothe immiscible silicate melt that forms during smelting.These elements include thorium (which is present in suchlarge quantities that its concentration in silicate phases

HEAVY MINERALS 200714

could be measured by electron microprobe), zirconium,barium, lanthanum, niobium, cerium (the distribution ofthese elements was determined by PIXE), and possibly alsothe radioactive daughter elements of the thorium- anduranium-decay series.

Mass balance calculations suggest that the thoriumcontent of the silicate phases is high enough to account forthe total thorium content of the slags. This is also reflectedby the silicate-oxide distribution coefficients for Th that areoverwhelmingly in favour of the silicate phases. Dependingon the total silica content, silicate phases constituteapproximately 4 to 8 weight per cent of the slags that wereselected for test work.

Spectroscopic data in the literature are consistent with thepartitioning of thorium into depolymerized (low-silica)glasses and melts relative to crystalline phases and morepolymerized glasses and melts. These results are inagreement with the observations of this study and confirmthat silicate phases contain most of the thorium present inthe slag.

In principle, the radioactivity levels of high titania slagcan be lowered by the removal (physical or chemical) ofsilicate phases from the slag. Disposal of the radioactive,silicate fraction will be onerous and the economics of sucha step are questionable.

References1. RGC Mineral Sands Ltd., S.A. Patent no. 93/5474, 29

July 1993.

2. BESSINGER, D., DU PLOOY, H., PISTORIUS,P.C., and VISSER, C. Characteristics of some hightitania slags. Heavy Minerals. R.E. Robinson. (ed.)Johannesburg, South African Institute of Mining andMetallurgy, 1997. pp. 151–156.

3. ELLER, P.G., JARVINEN, G.D., PURSON, R.A.,PENNEMAN, R.R., LYTLE, F.W., and GREEGOR,R.B. Actinide abundances in borosilicate glass.Radiochimica Acta, vol. 39, 1985. pp. 17–22

4. FARGES, F. Structural environment around Th4+ insilicate glasses: Implications for the geochemistry ofincompatible M4+ elements. Geochimica etCosmochimica Acta, vol. 55, 1991. pp. 3303–3319.

5. MAHOOD, G.A. and HILDRETH, W. Large partitioncoefficients for trace elements in high silica rhyolites.Geochimica et Cosmochimica Acta, vol. 47, 1983. pp. 11–30.

6. BARIN, I. Thermochemical Data of Pure Substances,Parts I and II. VCH Verlagsgesellschaft. 1989.