ANALYST, MAY 1991, VOL. 116 449 Rapid Method for the Determination of the Major Components of Magnesite, Dolomite and Related Materials by X-ray Spectrometry Michael H. Jones and 6. William Wilson" CSIRO Division of Mineral Products, P.O. Box 124, Port Melbourne, Victoria 3207, Australia A new flux, MAG 5743, has been developed, which contains 57% m/m lithium tetraborate and 43% m/m lithium metaborate for use in the analysis of raw magnesite, caustic and dead burnt magnesia, (i.e., magnesite heated to 700-1100 "C and ~ 1 4 0 0 "C, respectively) and dolomite. The flux can be used for the determination of loss on fusion followed by X-ray fluorescence analysis of the glass disc for its major component elements, as their oxides MgO, CaO, SiOz, A1203 and Fe203, on the one sample. As the flux is both fluid and reactive at 1000 "C, fusion times are short (10 min) regardless of the reactivity of the sample.The results obtained from the analysis of three certified reference materials using the new flux were statistically indistinguishable at the 95% confidence level from the certificate values. Keywords: X-ray fluorescence; magnesia, magnesite and dolomite analysis; loss on fusion The rapid and accurate determination of the principal components of magnesite, magnesia, dolomite and calcite (as the oxides MgO, CaO, SO2, A1203, Fe203) and the loss on ignition (LOI) is necessary for quality control in mining and processing. Current X-ray fluorescence (XRF) techniques require the sample to be calcined before fusion,lJ determine only one component3.4 and/or require the use of high melting-point fluxes with relatively long fusion times.Overall, the tech- niques currently used, although accurate, are time consuming and labour intensive. In addition, making a separate measure- ment of the LO1 has several disadvantages, these being: the time taken for measurement; the temperature of ignition being sometimes insufficient to break down any refractory constituents; and the possibility of re-adsorption of H20 and/or C02 on cooling. These last two disadvantages can contribute to a low LO1 measurement. Furthermore, the LO1 temperatures in recommended methods range from 1000- 1100 "(25-9 The flux used to prepare discs for XRF analysis in most methods is lithium tetrab0rate.1~2~~~~0 An 'acidic' flux such as lithium tetraborate or sodium metaphosphatelo is necessary for the dissolution of samples and production of stable discs.Lithium tetraborate at the usual fluxing temperature of 1100-1200 "C has such a high viscosity that even modern fluxing machines require dissolution times of 30 min or longer. Mixed lithium tetraborate-metaborate fluxes with melting- points of t l O O O "C and a high fluidity are widely available but they usually contain a high proportion of lithium metaborate, which produces unstable discs when used with samples that have high concentrations of Mg or Ca. This paper describes the development of a new flux, designated MAG 5743, with a composition of 57% m/m lithium tetraborate and 43% m/m lithium metaborate, which not only produces stable discs but is also sufficiently reactive so as to obtain the complete dissolution of magnesites, calcined magnesites and dolomites in 10 min at 1000 "C.Two methods are described. The magnesite method, in which the glass disc composition has a flux to sample (FTS) ratio of 5 : 1, is used with samples for which the LO1 is >40% (i, e., uncalcined raw materials). This method provides oxide analysis and a loss on fusion (LOF) value simultaneously. The general method, in which the glass disc composition has an FTS ratio of 10 : 1, is essentially for samples where the LO1 is <40%. ( i . e . , calcined products). Certified reference materials (CRMs)" and four magnesite samples are analysed using the new flux. * To whom correspondence should be addressed. Experimental Sample Preparation The British Chemical Standard (BCS) CRMs magnesites 319 and 389, which are magnesias, were re-ignited at 1025 "C and the dolomite (BCS CRM 368) was oven dried at 110 "C before use.Three discs were prepared from each CRM for analysis and the results obtained were compared with the certificate values. Four samples of magnesite (labelled magnesite A, B, C and D) from the Kunwarara deposit (Queensland, Australia) with a range of MgO, CaO and Si02 contents, were ground to a particle size of t 1 5 0 pm (100%) for analysis by both the magnesite and general methods. Three glass discs were prepared from each of the magnesite samples using both methods, i.e., the LOFs were determined simultaneously. Each disc was analysed separately over a period of 4 d. The data were compared with those obtained from the same magnesites after calcination at 1300 "C; the LO1 was deter- mined separately.Preparation of MAG 5743 Flux The flux was a mixture of 57% m/m lithium tetraborate and 43% m/m lithium metaborate. This flux can be obtained custom blended (by special order from suppliers such as Sigma or Johnson Matthey) or can be blended in the laboratory by mixing the constituents, which are then dried at 500 "C in a muffle furnace for at least 4 h. After removal from the furnace, the flux was cooled in a desiccator charged with a suitable desiccant. The flux should be stored in a capped jar in a desiccator. In order to test whether the flux was in a proper condition for use, a 3 g portion was fused at 1000 "C for 10 min to determine the LOF.(If the LOF exceeds 0.2% the flux must be re-dried at 500 "C.) Preparation of Glass Discs A 0.3000 k 0.0005 g sample (general method) or, a 0.6000 k 0.0010 g sample (magnesite method) was transferred into a 95% Pt-5% Au alloy crucible which had been previously heated to 1000 "C and cooled to constant mass. The sample masses were chosen in order to prepare 30-32 mm diameter discs. Larger diameter discs required larger masses of both sample and flux. The sample was then intimately mixed (by stirring thoroughly with a clean platinum rod) with 3.000 k 0.001 g of dry flux and the total mass of the crucible and contents recorded. The crucible was placed in a muffle furnace450 ANALYST, MAY 1991, VOL. 116 (containing an agitator mechanism) and the sample fused at lo00 "C for 10 min.Alternatively, any comercially available fusion device fitted with an agitator could be used under equivalent conditions. The crucible was removed from the furnace, cooled rapidly on an aluminium block heat sink contained in a desiccator and then re-weighed. The crucible and glass were returned to the furnace at 1000 "C, allowed to melt, and then poured into a casting mould (95% Pt-5% Au alloy, 30 mm i.d.), which had been pre-heated to at least lo00 "C over an oxygen-liquified petroleum gas (LPG) flame. A muffle furnace of appropriate size (set at lo00 "C) is a suitable alternative for casting the disc. The casting mould containing the molten glass was removed from the heat source and cooled rapidly on a graphite block or an air-cooled aluminium block.The disc was then available for analysis in a suitable X-ray spectrometer. [N.B. When a very large number of fusions are made daily, it might be desirable to give the aluminium block a ceramic coating to minimize the possibility of any aluminium adsorption.] Measurement The instrument conditions for the determination of the major components MgO, CaO, SO2, A1203 and Fe2O3 are listed in Table 1. A Philips PW 1404 wavelength dispersive, sequential X-ray spectrometer equipped with a scandium-molybdenum dual anode side-window X-ray tube (operated at 40 kV), with a flow-proportional counter, was operated under vacuum to measure the Ka lines of each element. A combination of flow-proportional and scintillation counters in tandem was used to measure the intensity of the iron Ka line.Pulse-height selection or line-overlap corrections can be used to reduce interference from the fifth-order calcium Ka line on the first-order magnesium Ka line for samples with a high-calcium and low-magnesium content. Calibration measurements were carried out using high- purity oxides or carbonates. The calibration was not carried out utilizing CRMs because only three CRMs for this rock type were available, vit., BCS CRMs 319, 368 and 389. A series of fifteen calibrations on discs containing only single elements was used to calibrate the spectrometer. Multiple regression analysis was used to calibrate the concen- tration of the ignited oxide (the temperature of ignition was determined by the metal constituent) against intensity (count rate).The matrix interference correction for the element on itself was also calculated simultaneously utilizing the de Jongh equation,l2 which is provided in the Philips X40 software. A further ten discs were prepared with the elements paired as follows: MgO-CaO, MgO-Si02, MgO-Al203, MgO-Fe203, CaGSi02, Ca0-AI2O3, CaO-Fe203, SiO2-AI2O3, Si02- Fe2O3, and A1203-Fe203. The inter-element matrix correc- tion for the paired elements was calculated by regression analysis. 12 The same series of discs was used for calibration using both the magnesite and general methods. The discs were originally used for the general method but, by using the following equations, the data were corrected for use with the magnesite method. Equation (1) is a general equation that, in this Table 1 Instrumental parameters for Philips PW 1404 Element Time/s Crystal Collimator 29/" Fe 40 LiF 200 Fine 57.57 Ca 40 LiF 200 Fine 113.19 Si 40 PET* Coarse 109.17 A1 40 PET* Coarse 145.08 Mg 200 PXlt Coarse 23.16 * PET = Pentaerythritol.t PX1 = Synthetic layered crystal, 2d = 4.9800 nm. instance, becomes eqn. (2), which in turn, simplifies to eqn. (3) CMAG = CGEN x 0.3000 - x - 3.600 0.6000 3.300 (3) where, CMAG is the calibration concentration for the mag- nesite method; CGEN is the calibration concentration for the general method; SGEN is the sample mass for the general method; SMAG is the sample mass for the magnesite method; TMAG is the sample plus flux mass for the magnesite method; and TGEN is the sample plus flux mass for the general method. Repeatability The repeatability of the results using the two XRF methods was tested over a period of 2-3 d by analysing each of the three individual discs from magnesite sample C in sequence.This measurement sequence was repeated twice at daily intervals and finally one disc was analysed in triplicate. The result for an individual disc reading was compared with that from the group. Effect of Initial Temperature on Loss on Fusion Raw magnesite sample B was analysed using the magnesite method. The crucible and contents were placed in the fusion furnace, operating over a range of initial fusion temperatures (500-1000 "C), and the LOF was determined. Results and Discussion Development of Flux Bennett and Oliver13 have detailed the relative merits and disadvantages of existing fluxes, such as lithium tetraborate (100%) and mixtures with lithium metaborate, for forming fused glass discs.The MAG 5743 flux was developed from Norrish 1222 flux (a mixture of 35% m/m lithium tetraborate and 65% m/m lithium metaborate). Norrish 1222 is an alkaline flux with a low melting-point and a low solubility for oxides such as MgO, CaO and A1203 (the so-called basic oxides). Stable discs could not be produced with samples of a high magnesium or calcium content with Norrish 1222. The solubility of the basic oxides in Norrish 1222 flux can be increased by the addition of another oxide such as SO2, Ti02 or B203 in order to decrease the basicity of the flux. Boron(II1) oxide is a non-interfering oxide in XRF spectrometry and, when added as a modifier, increases the acidity and also the viscosity of the flux.A mixture of 100 g of Norrish 1222 flux plus 12 g of boron(m) oxide was considered to be necessary in order to obtain a suitable compromise of viscosity, high oxide solubility and low temperature of fusion. The use of this composite flux, however, was limited to pre-ignited samples because the mass of flux lost on fusion is approximately 15 mg g-1 which precludes its use for accurate LOF determinations. The variable water content of boron(i1i) oxide is the cause of the mass loss and makes the flux difficult to stabilize by heating or drying. The Norrish 1222 flux has a LizO : B2O3 ratio of 20 : 80 and calculations showed that the same ratio could be achieved with a lithium tetraborate-lithium metaborate mixture of 57% m/m lithium tetraborate and 43% m/m lithium metaborate.This flux, which can be obtained custom mixed from a supplier or produced by the user, is stable long-term when dried by heating at 500 "C and stored in a desiccator. The MAG 5743 is a reactive flux, which is fluid at 10oO "C, in contrast to the lithium tetraborate flux usually used for fusionANALYST, MAY 1991, VOL. 116 45 1 of carbonate rocks. The fusion of raw magnesite samples with the MAG 5743 at 1000 "C requires fluxing times (10 min) similar to those required when lithium tetraborate is used at 1100-1200 "C. This is a result of mixing due to the loss of volatiles on fusion. In contrast, caustic calcined and fused magnesia samples require 20-30 min for complete reaction with lithium tetraborate, but only 10 min with the MAG 5743 flux.The MAG 5743 is a specialized flux, hence, matrix corrections were not possible using commercial calculation programs. Therefore, the corrections obtained for the ele- ments were determined empirically and compared with those calculated using the modified NRL-XRF program14 for lithium tetraborate and the Norrish 1222 flux, because the composition of the MAG 5743 flux is between these two fluxes. The values obtained empirically were within the range of the values of the program-calculated correction coefficients for Fe203, CaO, Si02, A1203 and MgO with the lithium tetraborate and Norrish 1222 flux. Table 2 Repeatability ( n = 4) of measurements on three glass discs prepared from sample C Relative Conccntration Standard standard Determinand (Yo 1 deviation (YO) deviation (YO) 44.83 0.20 0.5 2.10 0.04 1.9 MgO Si02 1.12 0.06 5.4 Fed& 0.05 0.01 20 A1203 0.04 0.01 25 CaO LOF 51.80 0.13 0.3 Less than 1% of the beads produced from MAG 5743 cracked; these beads were, nevertheless, still usable.Repeatability Table 2 gives details of the results obtained from three separate discs prepared from the magnesite sample C. The results obtained from the individual discs showed that reproducible, stable discs giving repeatable results could be prepared, for analysis by XRF using the rapid fusion method. The discs slowly absorb moisture (at a rate determined by the relative humidity), which decreases the intensity of the signal. Therefore, long-term storage of the discs in a desiccator is necessary.Determination of the five major components together with the LOF can be obtained within 35-40 min. Analysis of CRMs The values obtained using the two methods ( i e . , magnesite and general methods) are compared with the certificate values of a series of BCS CRMs in Table 3. As both magnesite standards had already been calcined only the general method could be used. The dolomite (BCS CRM 368) had not been calcined during preparation so it could be used as a basis for comparison of the two methods. Comparison of the results for pre-ignited dolomite (BCS CRM 368) by the general method was not considered valid because the recommended tempera- ture of ignition (1025 "C) does not decompose the sample completely. The measured LOF of the pre-ignited sample was 1.1 -t 0.1%, which together with the measured LO1 at 1025 "C of 46.5 k 0.1% gave a total loss of 47.6 k 0.2%.This value compares favourably with the LOF at 1030 "C of 47.5 rt 0.3% Table 3 Comparison of XRF results ( n = 12) with the certificate values for the CRMs (all values expressed as percentages) Certified reference material BCS 319* BCS 3681- General method General method Magnesite method Deter- Certificate Concen- Standard Certificate Concen- Standard Concen- Standard minand value tration deviation value tration deviation tration deviation MgO 90.46 90.37 0.05 20.9 20.57 0.09 20.47 0.12 CaO 2.28 2.28 0.03 30.8 30.67 0.33 30.58 0.13 Si02 1.55 1.58 0.05 0.92 0.92 0.05 0.89 0.01 A1203 0.97 0.94 0.02 0.17 0.12 0.02 0.14 0.01 Fe203 4.63 4.58 0.05 0.23 0.24 0.02 0.20 0.02 LOF NA$ 0.16 0.05 ND$ 47.27 0.15 47.50 0.28 LO1 ND ND ND 46.7 ND ND ND ND * Magnesite method not applicable to these samples; pre-treatment temperature, 1025 "C.1- Pre-treatment temperature. 110 "C. $ NA = not applicable, ND = not determined. BCS 389* General method Certificate value 96.7 1.66 0.89 0.23 0.29 NA ND Concen- tration 96.65 1.66 0.90 0.21 0.32 0.21 ND Standard deviation 0.11 0.02 0.05 0.02 0.03 0.10 ND Table 4 Comparison of XRF results (n = 3) for fused-disc methods (all values expressed as percentages) Sample A B C Deter- minand 1* 2-t 3$ 1* 21 33 1" 2t 3$ MgO 46.21 46.16 46.30 45.12 45.19 45.35 45.12 45.15 45.08 CaO 1.28 1.25 1.27 2.32 2.31 2.35 2.08 2.07 2.11 Si02 0.20 0.20 0.18 0.27 0.25 0.23 1.24 1.17 1.16 A1203 0.04 0.03 0.03 0.08 0.04 0.05 0.06 0.04 0.05 Fe203 0.04 0.05 0.05 0.04 0.07 0.04 0.05 0.06 0.05 LO13 52.23 NAfi NA 52.03 NA NA 51.55 NA NA LOFll NA 52.42 52.30 NA 52.07 52.05 NA 51.83 51.66 * Results for ignited sample (0.3 g, 1300 "C) with values adjusted for LOI.t Results for sample (0.3 g) with no adjustment, and LOF determined using the general method. $ Results for sample (0.6 g) with no adjustment, and LOF determined using the magnesite method. 3 LO1 determined by igniting sample at 1300 "C before fusion. 7 NA = not applicable. (1 LOF determined after fusion at 1000 "C. D 1" 46.69 1 .oo 0.12 0.02 0.03 52.15 NA 21- 46.63 0.98 0.12 0.03 0.04 NA 52.43 3$ 46.53 1 .oo 0.10 0.01 0.03 NA 52.44452 ANALYST, MAY 1991, VOL. 116 (Table 3). Loss on ignition results are always lower because of incomplete decomposition.The LO1 results are only compar- able to LOF results if the magnesite dolomite has been ignited at 1300 "C to constant mass. Certified reference materials (BCS CRMs 319, 368 and 389) can only be used for comparison (Table 3) because CRMs 319 and 389 are burnt magnesites and 368 is a dolomite. Analysis of Magnesites Originally, the flux was developed for the rapid analysis of calcined magnesia but, because the flux was so stable, it was decided that the method could be modified and applied to unignited samples, thus including LOF in the analysis. The LOF value was therefore included as a correction factor in the de Jongh equation.15 With unignited samples, the LOF can be as much as 52% for magnesite and 42% for calcite which means that, if a 10 : 1 FTS ratio is used, the line intensities are reduced to about half those produced using a disc prepared from ignited material. In order to compensate for this loss in signal, the mass of sample was doubled thus halving the FT'S ratio.The alternative method of maintaining the signal, i.e., by doubling the count time, was rejected because of the concomitant increase in analysis time. A further benefit was achieved by increased accuracy in weighing. Table 4 shows a comparison of the results obtained for the four magnesite samples, using the conventional, general and magnesite methods. As with the results for the CRMs, agreement was achieved between the conventional analysis and the two proposed methods. Effect of Initial Fusion Temperature on LOF Value Loss on fusion values for magnesite sample B of 52.03,52.07, 52.10, 52.16 and 52.11% were obtained for initial fusion temperatures of 500,600,700,800 and lo00 "C, respectively.The LOF value obtained using the magnesite method was The results show that there was virtually no sample loss during the fusion. If any loss of mass had occurred owing to decrepitation of the sample, as opposed to loss due to CO2 and H20 evolution, the LOF value would have been greater at higher initial fusion temperatures. 52.09 k 0.13%. Conclusion A rapid fusion technique has been developed for the determination by XRF of the major oxides in magnesites, magnesias and dolomites. The total analysis time is as short as 40 min. The flux used in the fusion (MAG 5743), if prepared properly, has no LO1 and produces stable discs. Thus it can be used for the accurate and rapid simultaneous determination of the major oxides and the LOF. The validity of the method was checked by comparison of the results with the certificate values of three CRMs. The work reported here was sponsored by Queensland Magnesia Pty. Ltd. References 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 King, B.-S., and Vivit, D., X-Ray Spectrom., 1988, 17, 85. King, B.-S., and Vivit, D., X-Ray Spectrom., 1988, 17, 145. Simonov, K. V., Zos'ka, A. V., Polovinkina, R. S., and Dremina, V. A., Znd. Lab. (Engl. Transl.), 1979, 44, 1026. Prager, M. F., and Graves, D., J. Am. Oil Chem. Soc., 1977,60, 1386. British Chemical Standard, Certificate of Analysis. Standards Association of Australia, AS2503.4, 1987. International Standards OrganizatiodDraught International Standard, 10058. American Society for Testing and Materials, C574, 1982. Deutsche Industrie Norm, 273, 1981. Pert], A., Lehmann, H., and Grubitsch, H., Radex Rundsch., 1976, 1, 639. Govindaraju, K., Geostand. Newsl., 1989, XI11 (Special Issue), Appendix 1, p. 27. de Jongh, W. K., X-Ray Spectrom., 1973,2, 151. Bennett, H., and Oliver, G. J., Analyst, 1976, 101, 803. Norrish, K., Commonwealth Scientific and Industrial Research Organization (CSIRO), Adelaide, South Australia, Australia, personal communication, 1990. de Jongh, W. K., X-Ray Spectrom., 1979, 8, 52. Paper 0104412A Received October lst, 1990 Accepted December 19th, 1990