optimizing xrf calibration protocols for elemental ... - ICDD

220

OPTIMIZING XRF CALIBRATION PROTOCOLS FOR ELEMENTAL QUANTIFICATION OF MINERAL SOLIDS FROM ATHABASCA OIL SANDS B. Patarachao*, P.H.J. Mercier*, J. Kung*, J.R. Woods*, L.S. Kotlyar*, B.D. Sparks**, T. McCracken* *Institute for Chemical Process and Environmental Technology, National Research Council Canada, 1200 Montreal Road, Ottawa, ON, Canada K1A 0R6 **V. Bede Technical Associates, 614 Laverendrye Drive, Ottawa, ON, K1J 7C4 ABSTRACT As world reserves of conventional oil keep decreasing, there is greater incentive to further develop the Athabasca oil sands of Alberta (Canada). Oil sands being composed of coarse sand, silt and clay solids (8085%), bitumen (515%) and water (15%), complete extraction of bitumen from such heterogeneous mixtures is not easy. Studies have shown the adverse effects of some types of mineral solids on bitumen recovery. Quantitative analysis of the elemental composition of mineral solids in ores is then of great importance to the oil sands industry. Two calibration protocols were developed here and implemented for accurate determination of major and minor elements in oil-sand solids by wavelength dispersive X-ray fluorescence (WDXRF) analysis using a fusion-based procedure. Commercially available standards do not span the ranges of element concentrations found in the mineral solids from oil sands. As such, calibration standards for seventeen elements were then designed by mixing pure synthetic oxides or geological reference materials, in order to mimic the elemental concentrations of oil-sand solids fractions. Measurement conditions were optimized to ensure best signal-to-background ratio and minimum line overlap. The limit of detection, calibration ranges and uncertainty errors of the resulting calibration curves are reported, showing excellent precision and accuracy even without matrix-effects correction. Application to analyze a suite of oil sands samples showed that the elemental concentrations of Ti and Zr in problematic solids components correlated well with the concentration of these elements in the entire mineral solids content present in the ores. This observation might be relevant for the development of elemental compositions-based processability markers to identify problem ores yielding poor extraction performance in commercial operations. INTRODUCTION Oil sands are heterogeneous mixtures comprising 8085% mineral solids (coarse sand, silt and clay), 515% bitumen and 15% water. Complete recovery of bitumen from the ore is often difficult to achieve. Ideally, the bitumen is separated from the inorganic mineral matrix by a layer of interstitial water, allowing it to be easily   However, a water-wet mineral condition is not universal to all oil reservoirs (McCaffery and Bennion, 1974; Berkowitz and Speight, 1975; Clementz, 1976). In oil sands deposits, the lack of water wettability arises because of the presence of toluene insoluble organic matter (TIOM) (Greenland, 1971; Parfitt, 1978; Fysh et al., 1983; Cavallero and McBride, 1984). TIOM can exist as a surface layer strongly adsorbed on individual particles or as the binding agent in porous aggregates of clay and mineral particles. Inorganic cementing materials (carbonates, iron oxides, etc.) can also act as binding agents for the clays and other mineral components.

This document was presented at the Denver X-ray Conference (DXC) on Applications of X-ray Analysis. Sponsored by the International Centre for Diffraction Data (ICDD).

This document is provided by ICDD in cooperation with the authors and presenters of the DXC for the express purpose of educating the scientific community. All copyrights for the document are retained by ICDD. Usage is restricted for the purposes of education and scientific research. DXC Website – www.dxcicdd.com

ICDD Website - www.icdd.com

221

We have developed a procedure for the separation of such organic-mineral complexes from oil sands. We refer to this component as     ) (Kotlyar et al., 1988, 1998; Sparks et al., 2003). The presence of TIOM on the external surfaces and within the pores of ORS particles/aggregates provides potential sites for sorption and fixation of bitumen. During extraction processes for bitumen recovery, the ORS component of oil sands can therefore be responsible for significant bitumen losses with certain ore types. Characterization of the mineral solids in ORS is then of great importance in view of the detrimental effect that ORS can have on the processability of oil sands. However, the technique for separation of ORS is tedious, labor intensive and very time consuming. As the ORS content accounts for ~15 wt% of an oil sands sample, in practice treating 10 g of ore to separate its ORS provides only small amounts of material (~100500 mg). On the other hand, oil sands samples are analyzed routinely by industry for geological assessment of drilled cores to determine the bitumen, water and solids contents of the ore using the standard Soxhlet-Dean and Stark extraction method (Bulmer and Starr, 1974). The extracted mineral solids produced by this technique are referred to as bitumen free solids (BFS). The BFS represent the entire solids content of an oil sands ore. Unlike ORS, BFS are thus readily obtained in larger amounts and their separation using Soxhlet-Dean and Stark extraction is much less complicated than ORS separation. In this work, we develop X-ray fluorescence (XRF) calibration protocols for accurate elemental quantification of mineral solids in ORS and BFS. We then apply these results to examine the metal concentrations in a suite of oil sands ore samples. These same samples have been characterized in detail previously to assess the potential impact that mineralogy and size distribution of clays may have on bitumen recovery (Mercier et al., 2008). Our aim here is to find correlations between the elemental compositions of ORS and BFS. If such correlations were established, it may lead to the development of elemental compositions-based processability markers to identify problem ores yielding poor bitumen recovery ahead of extraction in commercial operations. EXPERIMENTAL Separation To separate the BFS, oil sands samples were extracted using the standard Soxhlet-Dean and Stark method with toluene as the solvent (Bulmer and Starr, 1974). The ORS fraction was separated from each oil sands using the following procedure. Toluene (~20 g) and distilled water (~60 g) were added to an ore sample (~10 g) in a glass jar (100 mL) with a well-sealed screw cap. The resulting mixtures were then mixed vigorously for 0.5 min using a Spex Mixer, followed by centrifugation for 20 min at 2100 rpm. Three layers were produced after centrifugation: (1) a densely packed sediment of residual solids at the bottom of the container; (2) an intermediate (middlings) layer, comprising solids in aqueous suspension; and (3) a top layer of bitumen-toluene solution and associated solids. After removal of the top layer by skimming with a spatula, the agitation and centrifugation treatments were repeated with successive aliquots of fresh distilled water and toluene until no more bitumen was separated. The combined bitumen-toluene top-layer fractions, the middlings and the bottom-layer sediments were then individually washed with toluene until free of soluble components. The three types of resulting bitumen-free materials were then separated into ORS and hydrophilic solids fractions

222

by manually shaking the samples in jars containing toluene (15 g) and distilled water (60 g). The ORS material forms a distinct  e layer at the solvent/water interface that can be easily separated with a spatula. If necessary, i.e. when a rag layer did not form after manual shaking of the samples, the mixtures were centrifuged for 5 min at 2100 rpm. Separated ORS was repeatedly washed with fresh toluene/water mixtures to remove any weakly attached or occluded hydrophilic solids. Sample preparation The lithium borate fusion method was chosen for its excellent repeatability to prepare both ORS and BFS specimens and corresponding standards for XRF analysis. For ORS, 100 mg of sample was mixed with 7 g of flux. The type of flux used was Claisse 66.67%Li2B4O732.83%LiBO2  0.5%LiBr. For BFS, 1 g of sample was mixed with 7 g of flux. In this case, the type of flux used was Claisse 49.25%Li2B4O749.25%LiBO20.5%LiBr. The mixtures were fused with a Claisse M4 three-position gas fluxer producing good quality 32-mm beads. Instrumentation An automated Bruker AXS S4 Pioneer was used to analyze all of our specimens. The S4 pioneer is a 4kW sequential wavelength-dispersive X-ray fluorescence (XRF) spectrometer using a Rhanode X-ray tube with 75 thin Be end window. All the measurements were done under vacuum mode using a 28 mm mask and a 0.23 degree collimator. All elements (Table 1) were analyzed using K1 analytical lines, except for Pb where the L1 line was used. RESULTS AND DISCUSSION Calibration for ORS Table 2 lists the certified reference materials selected to build a calibration to analyze ORS-type materials. These standards were purchased from USGS (United States Geological Survey), NIST (National Institute of Standards and Technology), and CNRS (Centre National de la Recherche Scientifique). For Fe, Ti and Zr, Table 1 shows that these standards do not span the ranges of element concentrations typically found in the ORS fraction of oil sands. As Fe, Ti and Zr are three major elements typically found in ORS, the concentration ranges of the calibration standards needed to be extended for an accurate quantification of these elements in ORS. To achieve this goal, a few certified reference materials were selected and then mixed with various concentrations of high purity reagent grade oxides of Fe, Ti and Zr (99.999%TiO2, 99.999%Fe2O3 and 99.99%ZrO2). Concentration ranges of TiO2, Fe2O3 and Zr were expanded to 30 wt%, 60 wt% and 148000 ppm (14.8 wt%), respectively. Blending geological references with pure oxides has been studied and shown to yield accurate results (Nakayama and Nakamura, 2008). The mixtures were homogenized with a high intensive shaker for 5 minutes. All 39 reference materials plus 14 mixed standards specimens were prepared by the fusion method described above. Calibration curves for selected elements in ORS are shown in Figure 1. Data points are narrowly distributed around the fitted linear-regression calibration lines, which were built by plotting net intensity count rates as function of certified compositions. Table 1 gives the ranges of validity of the resulting calibration curves and their associated 1 standard uncertainty (s.u.) measurement errors for each of the elements analyzed. The green arrows shown in Figure 1

223

illustrate the extension of concentration ranges for Fe, Ti and Zr obtained by mixing geological standards with pure oxides to cover the element concentrations found in the ORS. The accuracy of the calibration for ORS was tested by selecting 10 certified reference materials that were not included to derive the linear-regression curves. We then evaluated the elemental concentrations of these standards using the newly developed calibration for ORS and compared the results against the certified compositions. The standard deviation (SD) values of differences between determined and certified concentrations for these 10 standards were evaluated (Table 1). For all elements where it was possible to do so, the accuracy test SD values were always smaller or equal to the 1 s.u. errors corresponding to the precision of element-concentration determination using our calibration for ORS. Although these calibration curves were built without matrix-effects correction, the accuracy test showed elemental analysis results that were accurate to within the precision of the determination using the calibration for ORS. Table 1. Concentration ranges and measurement errors of the calibration for ORS. Concentration Ranges of Certified Reference Materials

Na2O MgO K2O CaO Al2O3 SiO2 P 2O 5 TiO2 Fe2O3 MnO

Min 0.021 0.01 0.014 0.037 2.589 0.24 0.003 0.012 0.075 1.30

Max 10.59 35.21 12.81 67.87 70.04 75.8 1.145 2.73 38.47 0.10

Cr Ni Pb Sr Zn Zr

2 0 2 3 6 0.7

2300 2000 5532 5394 6952 800

Major Oxides (wt%) Concentration Concentration Ranges Found in Ranges of ORS samples ORS Calibration Min Max Min 0 0.35 0.021 0.5 1.90 0.01 0 2.08 0.014 1.22 6.18 0.037 1.9 9.33 2.589 4.9 37.74 0.24 0.11 0.92 0.003 1.17 21.6 0.012 12.13 48.75 0.075 0.82 0.1 1.30 Trace Elements (ppm) 144 7830 2 80 2860 0 0 0 2 0 1450 3 0 1700 6 0 54709 0.7

1 s.u. Measurement Error (Precision)

Accuracy Test Standard Deviation

Max 10.59 35.21 12.81 67.87 70.04 75.8 1.145 30 60 0.10

           !      !  "

           "  19  "1   -

2300 2000 5532 5394 6952 148060

"   "" " "

-

Table 2. List of certified reference materials used in this study. USGS NIST

CNRS

BCR-2, BHVO-2, BIR-1, DNC-1,GSP-2, QLO-1, SDC-1, W2-a SRM1880a, SRM1881a, SRM1882a, SRM1883a, SRM1884a, SRM1885a, SRM1886a,SRM1887a, SRM1888a, SRM2586, SRM2857, SRM2709, SRM2710, SRM2711, SRM2782 AC-E, AL-I, AN-G, BE-N, DR-N, DT-N, FK-N, GA, GH, GL-O, GS-N, Mica-Fe, Mica-Mg, PM-S, UB-N, WS-E

22 20 18 16 14 12 10 8 6 4 2 0

 !#$

220

 !wt%

215

Net Intensity

Net Intensity

224

210 205 200 195

n=53 0

10

20

30

40

50

60

70

190 185

80

n=69 90

92

 wt%

1.6

Net Intensity

1.2

Net Intensity

100

  wt%

0.8

1.4

1.0 0.8 0.6 0.4

0.6

0.4

0.2

0.2

n=53

0.0

0

10

20

30

40

50

60

70

n=51

0.0

80

0

1

Net Intensity

60 40 20 0

20 10

20

30

40

50

n=66

0

60

0

2000

 wt%

30

4000

6000

8000

10000

Fe (ppm)

Fe2O3 (wt%) 35

5

30

n=53 10

4

%

40

80

0

3

50

 !wt%

100

2

Al2O3 (wt%)

Al2O3 (wt%)

10

%

8

Net Intensity

25

Net Intensity

98

1.0

1.8

20 15 10 5

n=53

0

0

5

10

15

20

25

6 4 2

n=54

0

0

30

2000 4000 6000 8000 10000

Ti (ppm)

TiO2 (wt%) 140

"%

120

400

"%

300

100

Net Intensity

Net Intensity

96

SiO2 (wt%)

SiO2 (wt%)

Net Intensity

94

80 60 40 20

n=36

0

0

40000

80000

120000

Zr (ppm)

160000

200

100

0

n=48 0

2000

4000

6000

8000 10000

Zr (ppm)

Figure 1. Calibration curves for elements of interest in ORS (left) and BFS (right).

225

Calibration for BFS BFS contain SiO2 and Al2O3 as major oxides in the range of 90100 wt% and 05 wt% respectively. Other minor elements such as Mg, Ca, Ti, Cr, Mn, Fe, Ni, Zr, Zn and K are found in levels ranging from 010,000 ppm. The authors could not find any certified reference materials that cover these ranges of concentrations found in BFS (Table 3). Consequently, a set of 23 calibration standards was created from mixtures of high purity reagent grade oxides (>99.99% purity). XRF calibrations based on synthetic standards made from pure compounds prepared by the borate fusion method have been studied in detail elsewhere and shown to yield accurate results (Staats, 1990; Sieber, 2002). In this study, each synthetic standard was prepared from pure oxide mixtures carefully weighed, mixed and then homogenized by wet grinding using a McCrone Micronizing Mill. 10 g of oxide mixtures were mixed with 25 mL of isopropanol and then micronized for 5 min. The ground mixtures were air dried for 24 hrs. Sample preparation for both synthetic standards and BFS samples was done by the fusion procedure described above. By micronizing, the standards mixtures and BFS samples were homogenized and the particle sizes were also reduced, which helped producing good quality fused beads. The importance of micronizing is shown in Figure 2, where example calibration curves built for Zr with unmicronized samples (Fig. 2a) are compared with the ones derived with micronized standards (Fig. 2b). Three beads were fused and analyzed for each of the standards to check the repeatability of the results. It is very clear that the scatter of data points for unmicronized standards is much greater. By micronizing, we were thus able to considerably reduce this scatter, hence demonstrating that the results were clearly more repeatable using careful sample preparation. The end-result benefit of micronizing was that the 1 s.u. measurement errors (Table 3) of the resulting calibration curves for BFS (Fig. 1) were significantly reduced for each element analyzed, yielding highly precise determination of elemental concentrations in BFS. Table 3. Concentration ranges and measurement errors of the calibration for BFS. Major Oxides (wt%) Concentration Ranges of BFS 1 s.u. Calibration Measurement Error Min Max SiO2 90 100  ! Al2O3 0 5   Trace Elements (ppm) Mg 0 10000 " K 0 10000  Ca 0 10000  Ti 0 10000  Cr 0 10000  Mn 0 10000 ! Fe 0 10000  Ni 0 10000  Zn 0 10000 !! Zr 0 10000 "

226

40

 

(b)

"

400

300

30

Net Intensity

Net Intensity

(a)

20

10

0

2000

4000

6000

100

0

n=27 0

200

n=48 0

8000 10000

2000

4000

6000

8000 10000

Zr (ppm)

Zr (ppm)

Figure 2. Examples of BFS calibration curves for unmicronized (a) and micronized (b) standards. Application to analysis of mineral solids in oil sands ore The trends in changes of Zr and Ti contents in BFS and ORS were examined specifically for a suite of oil sands samples (Mercier et al., 2008). As shown in Figure 3, our main observation indicated a somewhat linear correspondence of the elemental concentrations of Ti and Zr in ORS with those in BFS. The correlation is stronger for Zr than it is for Ti, but this is only preliminary. More experimental data on a larger number of oil sands is required to further confirm and more firmly establish this finding. If these trends hold true, then either Ti or Zr could be used as markers for the ORS component of oil sands. As explained above, the ORS fraction can have a detrimental effect on many aspects of extraction performance in processing oil sands for bitumen recovery. However, separating the ORS from an ore sample is difficult and time-consuming, so this procedure cannot possibly become a practical processability marker to implement in industrial settings. The merit of the correlations noted for Zr and Ti concentrations between ORS and BFS fractions, is that it might lead to the development of  &%'s to evaluate the content and properties of ORS without having to perform time-consuming wet-chemistry separation of this component. Indeed, the BFS fraction representing the entire mineral solids content of an ore is already routinely separated in commercial operations using Soxhlet-Dean and Stark extraction. We have shown that highly precise and accurate elemental compositions of both BFS and ORS samples can be achieved even without matrix-effects correction. We hope that the XRF calibration protocols developed and tested in the present work will be adopted by oil sands researchers and industry operators to improve ore characterization methodologies and perhaps develop elemental compositions-based markers to predict extraction performance. 4000

700

Ti in BFS (ppm) TS_Ti(ppm)

BFS_Zr(ppm) Zr in BFS (ppm)

800 600 500 400 300 200 100 0

-1

0

1

2

3

4

5

6

ORS_Zr(wt%) Zr in ORS (wt%)

7

8

3000 2000 1000 0

0

2

4

6

8

10

12

ORS_Ti(wt%) Ti in ORS ORS(wt%) Ti

Figure 3. Correspondence of Zr and Ti concentrations between ORS and BFS.

227

ACKNOWLEDGEMENTS Part of the financial support for this work was provided by the EcoEnergy Technology Initiative program of the Canadian government.

REFERENCES Berkowitz, N. and Speight, J.G. (1975). The oil sands of Alberta,()54, 138-149. Bulmer, J.T. and Starr, J. (1974). Syncrude Analytical Procedures for Oilsands and Bitumen Processing,ublished by the Alberta Oil Sands Technology Research Authority. Cavallaro, N. and McBride, M.B. (1984). Effect of Selective Dissolution on Charge and Surface Properties of an Acid Soil Clay, Clays and Clay Minerals 32, 283-290. Clementz, D.M. (1976). Interaction of petroleum heavy ends with montmorillonite,*  Clay Minerals 24, 312-319. Fysh, S.A., Cashion, J.D. and Clark, P.E. (1983). Mossbauer effect studies of iron in kaolin. I. structural iron,* *+31, 285-292. Greenland, D.J. (1971). Interactions between humic and fulvic acids and clays, Soil Sci. 111, 34-41. Kotlyar, L.S., Ripmeester, J.A., Sparks, B.D., and Montgomery, D.S. (1988) Characterization of organic-rich solids fractions isolated from oil sands using a cold water agitation test,Fuel 67, 221-226. Kotlyar, L.S., Sparks, B.D., and Chung, K.H. (1998)    , %  - properties and role in processing,.#/ ess Chemistry and Engineering 1, 81-110. Mercier, P.H.J., Patarachao, B., Kung, J., Kingston, D.M., Woods, J.R., Sparks, B.D., Kotlyar, L.S., Ng, S., Moran, K., and McCracken, T. (2008). 0-ray diffraction (XRD)-derived processability markers for oil sands based on clay mineralogy and crystallite thickness Distributions,1 2()22, 3174-3193. McCaffery, F.G. and Bennion, D.W. (1974). The Effect of Wettability on Two-Phase Relative Permeabilities, J. Can. Pet. Tech. (Oct.-Dec.), 42-53. Nakayama, K. and Nakamura, T. (2008). *   ) % ,  glass bead x-ray fluorescence analyses of geochemical samples,0-ray Spectrom. 37, 204209. Parfitt, R.L. (1979). Anion adsorption by soils and soil materials,3 .3   %30, 1-50. Sieber, J. R. (2002). +&-independent XRF methods for certification of standard reference materials,3 .0-ray Analysis 45, 493-504. Sparks, B.D., Kotlyar, L.S., O* 45 6 4 *) 47 8 2003). Athabasca oil sands: effect of organic coated solids on bitumen recovery and quality,5 ) ,/ )%  Technology 39, 417-430. Staats, G. (1990). Synthetic macro-reference samples for the calibration of instruments in inorganic bulk analysis, Fresenius Z. Anal. Chem. 334, 326-330.