A brief introduction to the FEI Mineral Liberation

A brief introduction to the FEI Mineral Liberation Analyzer™: the technique & results Michael Shaffer INCO Innovation Centre Memorial University St. John’s, Newfoundland mshaffer@mun. ca Advanced Techniques in EPMA Seminar August 7, 2010 University of Oregon Eugene, Oregon

MLA: points of interest n Particle analysis n n n Rocks crushed, sized and representative Most accurate E. G, iron ore from Labrador n “Large particle” analysis n n n e. g. , 25 x 45 mm section Questionably representative Large grain sizes textures E. G, Himalayan garnet shist 2

BEI: Fe-rich minerals 3

Fe-rich minerals of interest & spectral ambiguity n Hematite & magnetite [Fe 2 O 3 versus Fe 3 O 4] n n Generally not distinguishable with x-ray spectra Associations important to client n Titano-magnetite n n n Distinguishable with x-ray spectra BSE similar to Hm Titanium important to client n Goethite or limonite [Fe. O(OH) • (H 2 O)n] n n Generally with minor Al, Si, Mg, and usually distinguishable with x-ray spectra BSE darker than Hm (BSE classification would be helpful) n Siderite [Fe. CO 3] n n Generally with Ca, Mg, Mn, and usually distinguishable with x -ray spectra BSE darker than Hm (BSE classification would be helpful) 4

Mineral modes Mineral Hematite Magnetite Ti_magnetite Goethite Limonite Ilmenite Rutile Corundum Quartz Aluminosilicate Misc_silicates Siderite Siderit-Mn Rhodochrosite Rhodo-Fe. Mg Rhodo-Mg. Fe Siderit-Mg. Mn Siderit-Mg Ankerite Calcit-Mg. Mn Dolomit-Fe. Mn Magnesit-Fe. Mn Dolomite Calcite Unknown Wt% 4. 57 38. 54 0. 09 0. 17 0. 08 nd nd nd 35. 55 nd 0. 11 0. 06 0. 11 nd 0. 01 0. 00 7. 37 0. 96 0. 06 nd 11. 48 0. 22 0. 15 0. 08 0. 02 Mineral Wt% Pyrolusite 0. 00 Bixbyite_lo-Mn nd Bixbyite_hi-Mn nd Other_oxides 0. 00 Olivine 0. 00 Garnet 0. 00 Cpx 0. 01 Opx 0. 02 Amphibole 0. 00 Biotite 0. 03 Feldspar 0. 03 Muscovite 0. 04 Serpentine nd Chlorite 0. 14 Mn-rich_clay nd Calcit-REE nd Pyrite 0. 00 Pyrrhotite nd Chalcopyrite nd Sphalerite nd Misc_sulfides nd Apatite 0. 08 Miscellaneous 0. 00 Misc_metals 0. 01 Total 100. 0 Mineral Magnetite Hematite Hm_or_Mt Goethite Limonite Other_oxides Quartz Misc_silicates Carbonates Sulfides Misc Unknown Total Wt% 38. 54 4. 57 0. 00 0. 17 0. 08 0. 09 35. 55 0. 38 20. 50 0. 09 0. 02 100. 0 5

The particle table 4 k to 20 k particles 6

Properties of particles Density Wt% Area (microns) Area (pixels) Perimeter Max Span Length (MBR) Breadth (MBR) Hull Area Hull Perimeter EE Minor Axis Hull EE Minor Axis EE Major Axis (P&A) EE Minor Axis (P&A) EE Perimeter EC Diameter Angularity Enclosed Length Delta Form Factor All minerals (Wt%) e. g. , Hematite (Wt%) Magnetite (Wt%) Goethite (Wt%) Limonite (Wt%) Quartz (Wt%) … Misc (Wt%) Unknown (Wt%) Free Boundary, all minerals e. g. , Hematite (%) Magnetite (%) Goethite (%) Limonite (%) Quartz (%) … Misc (%) Unknown (%) All elements (Wt%) e. g. , Al (Wt%) Ca (Wt%) Cr (Wt%) Cu (Wt%) Fe (Wt%) H (Wt%) K (Wt%) La (Wt%) Mg (Wt%) Mn (Wt%) Na (Wt%) Ni (Wt%) P (Wt%) Si (Wt%) Ti (Wt%) … Zn (Wt%) 7

datamining the particle table Si content for particles of density greater than SG 0. 8 0. 7 0. 6 SF+100 0. 5 SF+200 Si %0. 4 0. 3 0. 2 0. 1 0. 0 3. 5 3. 7 3. 9 4. 1 4. 3 4. 5 4. 7 4. 9 5. 1 5. 3 specific gravity of particles 8

Large sections

Spectral discrimination ~ garnet

grain boundaries resolved with BEI

grain boundaries not resolved with BEI

Grain associations Mineral Qtz Biot Plag Ksp Gt_Mg Qtz - 30 20 7. 3 1. 3 Biot 35 - 24 7. 3 1. 7 Plag 32 32 - 8. 9 0. 9 Ksp 29 25 23 - 0. 3 Gt_Mg 14 17 6. 7 0. 8 - 13

The grain table More than 52, 000 grains 14

Properties of grains Density Center X Center Y Wt% Area (microns) Area (pixels) Perimeter Max Span Angle Wt% (Particle) Area% (Particle) Wt% (Mineral) Area% (Mineral) Particle Max Span Particle Perimeter Length (MBR) Breadth (MBR) Angle Length (MBR) Hull Area Hull Perimeter EE Minor Axis Hull EE Perimeter EE Major Axis (P&A) EE Minor Axis (P&A) EC Diameter Aspect Ratio Angularity Enclosed Length Delta Form Factor Boundaries with other minerals e. g. , Quartz (%) Orthoclase (%) Garnet (%) Biotite (%) … free surface (%) 15

datamining the grain table: mineral textures plagioclase orientation 1. 4 1. 2 % plagioclase 1. 0 0. 8 0. 6 0. 4 0. 2 0. 0 0 30 60 90 120 150 180 angle for MBR 16

Applications at MUN n Mineral modes & associations n Mineral locking & liberation n Mineral searching (e. g. , zircon, baddeleyite, monazite) Includes x-y coordinate export n Precious mineral searching (e. g. , Au, PGM) n Includes associations with host minerals n Provenance determinations n Sourcing continental river & till sediments (mineral prospecting) n Sourcing offshore sediments with onshore (oil & gas) n Lateral correlation of offshore sediments (oil & gas) n n Some thought toward … n n Accurate determination of trace minerals (e. g. , apatite, corundum) Invisible gold with a FEG MLA Long-count EDX 17 Auxillary inputs …, e. g. , WDX, μXRF

Acknowledgements The MUN MLA team: David Grant Alan Maximchuk Dylan Goudie & thank you for your interest! 18

A typical frame, BSE relative to Ni metal 19

Is it possible with XBSE & MLA spectra? Difference is only 24 counts (2σ ~ 34) 28 Sensitive wt% O 20 to versus absorption 30% 15 counts (2σ ~ 58) 72 Sensitive wt% Feto versus charging 70%

The spectral-classification result Red implies mineral grain is either hematite or magnetite 21

BSE classification Qtz Hm “reliable” or Cumulative “full” histogram Mt Other silicates, carbonates and hydroxides 22

BSE-classification results – good & bad Magnetite Hematite “Darks” 23

$MLA BSE mode results – good & bad the smallest size fraction: -200 mesh$

MLA BSE mode results – good & bad the smallest size fraction: -200 mesh 24

Before “Merge Overlay” Mode BSE data acquisition Processed via gray level segmentation OR Mode XBSE data acquisition Processed via Spectral matching Classified data, modes, … Merge Overlay Classified data, modes, …

MLA “merge overlay” tool 26

Results from Merge Overlay n Spectrally classified “Hm-or-Mt” becomes: n Hematite, or n Magnetite, or n “Fe-ox_no-ID” n Which can generally be justified and grouped with limonite or goethite (… although pure siderite is also a possibility) n Smaller size fractions evaluated independently n Hm: Mt modal ratio might be assumed from larger SFs or their trends 27

Reproducibility: mineral modes same samples – 6 months between Samples A, B, C & D Size +100 M mineral modes 45 2008 40 2009 35 25 20 15 10 5 -ID no Mt Hm z Qt -ID no Mt Hm z 0 Qt Wt% 30 28

Reproducibility: mineral modes same samples – 6 months between Samples A, B, C & D Size +200 M mineral modes 45 2008 40 2009 35 25 20 15 10 5 -ID no Mt Hm z Qt -ID no Mt Hm z 0 Qt Wt% 30 29

ith Mt w ith th wi z Qt Hm z Qt Mt z Qt th wi ith Mt w Hm Hm Mt w ith th wi z Hm th wi ith Mt w Hm Hm Mt w ith Mt z Qt th th wi Mt w Hm wi Qt Hm z Qt Mt 20 Hm ith Mt w ith th wi Mt w Hm Hm Percentage of grain boundaries Reproducibility: mineral associations same samples – 6 months between Samples A, B, C & D Size +100 M mineral associations 2008 2009 15 10 5 0 30

ith Mt w ith z Qt Hm z Mt Qt th th wi Mt w Hm z Qt Hm 35 wi ith z Qt Mt 40 Hm Mt w th wi z Qt Hm z Mt Qt th wi ith Mt w Hm Hm Mt w ith th wi z Qt Hm z Mt Qt th wi Mt w Hm Hm ith Mt w ith th wi Mt w Hm Hm Percentage of grain boundaries Reproducibility: mineral associations same samples – 6 months between Samples A, B, C & D Size +35 M mineral associations 2008 2009 30 25 20 15 10 5 0 31

Results comparison: MLA v. Rietveld XRD 50 Rietveld 1 45 Rietveld 2 40 MLA 35 30 25 20 15 10 5 0 Qtz Mt Hm Sample A SF+100 M Qtz Mt Hm A SF+200 M Qtz Mt Hm Sample B SF+100 M Qtz Mt Hm B SF+200 M 32

Results comparison: MLA v. Rietveld XRD Average absolute errors 20 18 16 14 12 Quartz 10 8 Magnetite 6 Hematite 4 2 0 XRD sampling XRD-v-MLA

Sources of data processing error 34

Sources of instrumental error: electron beam illumination 195 = Hm 198 = Mt 192 = Hm 195 = Mt 35

Sources of instrumental error: varying e-beam current 3 rd frame 143 rd frame 2 hours Later … 192 = Hm 195 = Mt 195 = Hm 198 = Mt 36

Remedying BSE problems n Non-uniform illumination n No remedy if the SEM manufacturer did not anticipate applications in quantitative BSE n Except to use high magnification Difficult to remedy if the SEM manufacturer did not provide alignment tools for uniformity n FEI Quanta SEMs: n Centering the illumination provided by e-gun tilt n Tetrode & gun alignment should be accurate n Illumination gradients worse for large spot sizes n 37

Remedying BEI problems n Varying beam current n Very common depending on age of filament … n Stability generally monotonic, i. e. , not erratic n … allows for breaking the BSE JKF file into 2 to 4 files, thereby creating more reliable histograms that represent time periods during analysis. n Note also that this method is quite dependent on a significant amount of Hm-Mt in the sample, which builds a more accurate reliable histogram 38

Anticipating problems we haven’t yet encountered, and possible improvements n MUN IIC has not yet applied this method to mineral assemblages other than the minerals discussed here n I. E. , a severe complication would arise for significant amounts of titano-magnetite, thereby blurring the distinction of Hm in the reliable histogram n A very helpful improvement, which would allow the same tools to be applied to other applications, would be for the spectra-classified result to mask the minerals of interest to be classified with BSE 39

MLA Mode BSE conclusions n Hm – Mt BSE discrimination works … And Hm-Mt associations are possible n … but not specifically with other minerals n and, by itself, cannot discriminate most other minerals because of average atomic number (i. e. , BSE ambiguity) n However, it presents a suitable solution for augmenting spectral classification (mode XBSE) n How to augment with spectral classification? … n 40

Summary n Hm–Mt BEI discrimination is possible … n n Hm-Mt associations are possible, and with all minerals Mineral modes and associations can be reproduced with acceptable accuracy A comparison with quantitative XRD is within errors associated with the difficulty associated with representative downsampling (XRD sampling independent of MLA sampling) However, a well-aligned and stable SEM is necessary … n n n Electron beam illumination must be uniform over 1 – 2 mm Beam current must be stable over the 2 – 3 hr analytical time (although data processing can accommodate a monotonic variation) This technique is more generally applicable, even to more complex mineral assemblages when chemistry (x-ray spectra) aids in masking the minerals of interest 41

Consider an independent approach … 42

Exported BEI frames into 3 rd-party software 43

The masked & cleaned frames 44

A clean histogram allows for automatic thresholding 45

Independent software results fortunate & unfortunate 46

Independent BEI conclusions n Hm – Mt discrimination works … n Associations Hm-Mt are not possible n Minerals of similar atomic number, identified by XBSE, do not affect calculated Hm: Mt n However, results can be biased if: n n n one mineral does not polish as well, or if one mineral’s grain size is typically smaller Not the best solution, but should be in the analyst’s toolbox 47

The results for the client n Primary modes and associations come from mode XBSE. n Whereas we had been providing Hm: Mt via the independent method … n Because titano-magnetite and pyrite are minimal and correctable, we do not augment XBSE with additional BSE results. n The good news is that Hm-Mt associations are provided but the bad news is that Hm-Mt-Qtz associations are not. n What is needed … 48

Results comparison: MLA v. Rietveld XRD Sample 1 SFs +100 & +200 45 40 35 30 25 Quartz Magnetite 20 Hematite 15 10 sampling error 5 0 Rietveld 1 Rietveld 2 MLA 49

Results comparison: MLA v. Rietveld XRD Sample 2 SFs +100 & +200 50 45 40 35 30 Quartz 25 Magnetite Hematite 20 15 10 5 0 Rietveld 1 Rietveld 2 MLA 50