REFLECTANCE SPECTROSCOPY AND IMAGING SPECTROSCOPY OF RARE EARTH ELEMENT-BEARING MINERAL AND ROCK SAMPLES by DAVID JAMES TURNER B.Sc., The University of Victoria, 2000 M.Sc., The University of British Columbia, 2003 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Geological Sciences) THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) April 2015 © David James Turner, 2015 ii Abstract A significant knowledge gap between the fields of reflectance spectroscopy and rare earth element (REE) mineralogy prompted this research effort. It narrows the knowledge gap through detailed study of thirty three samples representing three important mineral classes: REE fluorocarbonates (bastnaesite, synchysite, and parisite), REE phosphates (monazite, xenotime, and britholite), and REE-bearing silicates (cerite, mosandrite, kainosite, zircon and eudialyte). Reflectance spectroscopy was carried out in the visible to short wave infrared regions (500 nm to 2500 nm) and each sample was characterized using scanning electron microscopy and electron microprobe analysis. Spectral features of these minerals are primarily related to numerous 4f-4f intraconfigurational electronic transitions of trivalent lanthanides (Ln3+), as well as 5f-5f electronic transitions of uranium and vibrational overtones and combinations of CO32-, H2O, PO43- and OH- where applicable. In general, the respective spectra of these REE minerals are sufficiently distinct for spectral classifications, and compilation diagrams with representative spectra are given. Broadly speaking, the light REE-enriched minerals are dominated by sharp absorptions related to Nd3+, Sm3+, and Pr3+ with lesser input from Eu3+, whereas heavy REE-enriched minerals are dominated by sharp spectral features related to Er3+, Dy3+ and Yb3+ with lesser input from Nd3+, Tb3+, Ho3+ and Tm3+ depending on their specific concentrations. For those minerals that do not show strong preference for light or heavy REE, a mixed set of absorption patterns is seen. Spectral variability of specific 4f-4f absorptions were substantial between different minerals and these variations are interpreted to be driven by the specific anion coordination at the Ln3+site across various crystal structures. Shifts in wavelength position and relative strengths of related absorptions can be significant enough to be highly relevant for hyperspectral remote sensing. This is especially applicable for REE mineral identification in field-based settings and high spatial resolution imaging spectroscopy. Three themes of ongoing and potential research are discussed: additional REE mineral spectroscopy, exploitation of diagnostic features for REE mineral detection and identification, and REE ore grade estimation. Overall, the research presented in this dissertation sets the foundation for future interpretation and exploitation of reflectance spectra for the REE minerals. iii Preface This dissertation presents original and independent work by the author, David J. Turner, under the supervision of Prof. Lee A. Groat and Prof. Benoit Rivard. Some data collection was carried out by third parties, and is identified within the thesis at the appropriate location. The contents of Chapter 2 represent a published article reformatted to fit within this thesis: Turner, D. J., Rivard, B., & Groat, L. A. (2014). Visible and short-wave infrared reflectance spectroscopy of REE fluorocarbonates. American Mineralogist, 99 (7), 1335-1346. Chapters 3 and 4 form the basis of two articles that are in preparation for submission. Chapter 5 was presented at an international conference and published in the conference proceedings as an extended abstract: Turner, D., Rivard, B., & Groat, L. (2014, July). Rare earth element ore grade estimation of mineralized drill core from hyperspectral imaging spectroscopy. In: Geoscience and Remote Sensing Symposium (IGARSS), 2014 IEEE International (pp. 4612-4615). IEEE. iv Table of contents Abstract .................................................................................................................................................... ii Preface .................................................................................................................................................... iii Table of contents ..................................................................................................................................... iv List of tables ......................................................................................................................................... viii List of figures ........................................................................................................................................... x Acknowledgements .............................................................................................................................. xvii Chapter 1. Introduction and Literature Review ....................................................................................... 1 1.1 Introduction ............................................................................................................................ 1 1.2 The lanthanides (rare earth elements) .................................................................................... 2 1.3 Rare earth element mineralogy .............................................................................................. 6 1.3.1 Introduction ..................................................................................................................... 6 1.3.2 Coordination states and effective ionic radii ................................................................... 7 1.3.3 Important mineral classes ............................................................................................... 7 1.4 Principal geological settings for REE deposits ...................................................................... 8 1.4.1 Carbonatites .................................................................................................................... 8 1.4.2 Peralkaline intrusive complexes.................................................................................... 10 1.4.3 Other deposit types ....................................................................................................... 11 1.5 Reflectance spectroscopy ..................................................................................................... 11 1.5.1 Introduction and general principles ............................................................................... 11 1.5.2 Specular and diffuse reflection ..................................................................................... 12 1.5.3 Absorption bands .......................................................................................................... 12 1.5.4 Transmission to reflectance .......................................................................................... 14 1.5.5 Grain size ...................................................................................................................... 14 1.5.6 Background continua of reflectance spectra ................................................................. 14 1.5.7 Mixtures, spectral unmixing and target detection ......................................................... 15 1.6 Electronic processes and absorption bands .......................................................................... 17 1.6.1 Crystal field theory ....................................................................................................... 17 1.6.2 Coordination geometry and crystal field splitting ......................................................... 18 1.6.3 Intervalence and oxygen-metal charge transfers ........................................................... 20 1.6.4 Band gaps and colour centres ....................................................................................... 21 1.7 Vibrational processes and absorption bands ........................................................................ 22 1.8 Spectroscopy of the lanthanides ........................................................................................... 23 v 1.8.1 Introduction ................................................................................................................... 23 1.8.2 Judd-Ofelt theory .......................................................................................................... 24 1.8.3 Intraconfigurational 4f-4f transitions ............................................................................ 24 1.9 Reflectance spectroscopy of REE bearing rocks and minerals ............................................ 27 1.9.1 REE-bearing minerals ................................................................................................... 27 1.9.2 Remote sensing of REE related geological targets ....................................................... 29 1.10 Summary ............................................................................................................................ 31 Chapter 2. Visible to Short Wave Infrared Reflectance Spectroscopy of REE Fluorocarbonates......... 32 2.1 Chapter summary ................................................................................................................. 32 2.2 Introduction .......................................................................................................................... 32 2.3 General spectroscopy of the lanthanides .............................................................................. 33 2.4 Review of REE-related reflectance spectroscopy studies .................................................... 34 2.5 Crystal structure reviews of bastnaesite, parisite and synchysite ........................................ 35 2.6 Experimental methods ......................................................................................................... 40 2.6.1 Samples ......................................................................................................................... 40 2.6.2 Reflectance spectroscopy .............................................................................................. 41 2.6.3 Scanning electron microscopy and electron microprobe analysis ................................ 44 2.7 Results .................................................................................................................................. 44 2.7.1 Bastnaesite .................................................................................................................... 50 2.7.2 Parisite .......................................................................................................................... 51 2.7.3 Synchysite ..................................................................................................................... 51 2.8 Discussion on the spectral variability of REE-fluorocarbonate minerals ............................ 52 2.8.1 Intraconfigurational 4f-4f electronic transitions ........................................................... 54 2.8.2 Vibrational bands of the carbonate radical ................................................................... 55 2.8.3 Spectral effects of the actinides .................................................................................... 56 2.9 Implications ......................................................................................................................... 56 Chapter 3. Visible to Short Wave Infrared Reflectance Spectroscopy of REE Phosphates .................. 58 3.1 Chapter summary ................................................................................................................. 58 3.2 Introduction .......................................................................................................................... 58 3.3 General spectroscopy of the lanthanides .............................................................................. 59 3.4 Review of REE-related reflectance spectroscopy studies .................................................... 60 3.5 Crystal structure reviews ...................................................................................................... 61 3.6 Experimental methods ......................................................................................................... 63 3.6.1 Samples ......................................................................................................................... 63 vi 3.6.2 Scanning electron microscopy and electron microprobe analysis ................................ 64 3.6.3 Reflectance spectroscopy .............................................................................................. 64 3.7 Electron microprobe compositions ...................................................................................... 66 3.8 Spectra and spectral variability of REE-phosphate minerals ............................................... 71 3.8.1 Monazite reflectance spectra ......................................................................................... 73 3.8.2 Xenotime reflectance spectra ........................................................................................ 76 3.8.3 Britholite reflectance spectra ........................................................................................ 80 3.9 Discussion on spectra and spectral variations ...................................................................... 82 3.9.1 Absorption patterns between monazite and britholite ................................................... 85 3.9.2 The Yb-Er related absorption near 978 nm ................................................................... 85 3.9.3 Complexity of xenotime’s absorption bands ................................................................ 86 3.9.4 Hydroxyl and water bands of the REE phosphate minerals .......................................... 87 3.9.5 Spectral effects of the actinides .................................................................................... 88 3.9.6 Comparison to USGS Spectral Library Spectra ............................................................ 88 3.9.7 Absorptions from 2150 nm to 2530 nm ........................................................................ 90 3.10 Implications ....................................................................................................................... 92 Chapter 4. Visible to Short Wave Infrared Reflectance Spectroscopy of REE-Bearing Silicates ......... 93 4.1 Chapter summary ................................................................................................................. 93 4.2 Introduction .......................................................................................................................... 93 4.3 General spectroscopy of the lanthanides .............................................................................. 94 4.4 Review of REE-related reflectance spectroscopy studies .................................................... 95 4.4.1 Spectroscopy of eudialyte ............................................................................................. 97 4.4.2 Spectroscopy of zircon .................................................................................................. 97 4.5 Crystal structure reviews ...................................................................................................... 98 4.6 Experimental methods ....................................................................................................... 103 4.6.1 Samples ....................................................................................................................... 103 4.6.2 Scanning electron microscopy and electron microprobe analysis .............................. 103 4.6.3 Reflectance spectroscopy ............................................................................................ 104 4.7 Electron microprobe compositions .................................................................................... 107 4.7.1 Cerite ........................................................................................................................... 113 4.7.2 Mosandrite .................................................................................................................. 113 4.7.3 Kainosite ..................................................................................................................... 113 4.7.4 Zircon .......................................................................................................................... 113 4.7.5 Eudialyte ..................................................................................................................... 114 vii 4.8 Spectra and spectral variability of REE-bearing silicates .................................................. 115 4.8.1 Cerite reflectance spectrum ......................................................................................... 117 4.8.2 Mosandrite reflectance spectrum ................................................................................ 120 4.8.3 Kainosite reflectance spectrum ................................................................................... 123 4.8.4 Zircon reflectance spectra ........................................................................................... 126 4.8.5 Eudialyte reflectance spectra ...................................................................................... 131 4.9 Discussion on spectra and spectral variations .................................................................... 137 4.9.1 Compilation of REE-bearing silicate spectra .............................................................. 137 4.9.2 Spectral patterns of eudialyte ...................................................................................... 139 4.9.3 The Er-Yb related absorption near 978 nm ................................................................. 140 4.9.4 Absorption band variations amongst REE-bearing silicate spectra ............................ 143 4.10 Implications ..................................................................................................................... 144 Chapter 5. REE Ore Grade Estimation of Drill Core by Imaging Spectroscopy ................................. 146 5.1 Introduction ........................................................................................................................ 146 5.2 Rock descriptions, mineralogy and geochemistry ............................................................. 146 5.3 Imaging spectroscopy and image processing ..................................................................... 150 5.4 Image processing results .................................................................................................... 156 5.5 Discussion .......................................................................................................................... 159 5.6 Conclusion and implications .............................................................................................. 163 Chapter 6. General Conclusions and Research Potential ..................................................................... 164 6.1 Summary ............................................................................................................................ 164 6.2 Future research ................................................................................................................... 165 6.2.1 Additional REE mineral spectroscopy ........................................................................ 165 6.2.2 Exploitation of diagnostic features for REE mineral detection and identification ..... 166 6.3 Conclusion ......................................................................................................................... 167 References ............................................................................................................................................ 169 Appendix A. Sample Photographs and Hyperspectral Imagery .......................................................... 190 Appendix B. Scanning Electron Microscopy ....................................................................................... 208 Appendix C. Electron Microprobe Compositions ............................................................................... 229 Appendix D. Reflectance Spectra of REE Oxides ............................................................................... 263 Appendix E. REE Ore Grading Geochemical Results ......................................................................... 274 Appendix F. REE Ore Grading Rietveld Method Results ................................................................... 277 viii List of tables Table 1.1 The rare earth elements. ................................................................................................................ 3 Table 1.2 Electronic configuration of the lanthanides. ................................................................................. 4 Table 2.1 Some basic properties of the REE fluorocarbonates. .................................................................. 37 Table 2.2 Average EMPA compositions of REE fluorocarbonate samples in wt.%. .................................. 45 Table 2.3 Prominent absorption features of the REE fluorocarbonates in the VNIR range. ...................... 47 Table 2.4 Prominent absorption features of the REE fluorocarbonates in the SWIR range. ...................... 48 Table 2.5 Significant chemical and spectral differences between bastnaesite samples. ............................. 49 Table 2.6 Significant chemical and spectral differences between parisite samples. ................................... 49 Table 3.1 Pixel counts per sample used to produce an average spectrum. ................................................. 66 Table 3.2 Electron microprobe compositions for monazite. ....................................................................... 67 Table 3.3 Electron microprobe compositions for xenotime. ....................................................................... 68 Table 3.4 Electron microprobe compositions for britholite. ....................................................................... 69 Table 3.5 Prominent absorption features of monazite samples in the VNIR and SWIR ranges. ................ 74 Table 3.6 Prominent absorption features of xenotime samples in the VNIR range. ................................... 77 Table 3.7 Prominent absorption features of xenotime samples in the SWIR range. ................................... 78 Table 3.8 Prominent absorption features of britholite samples in the VNIR and SWIR ranges. ................ 81 Table 4.1 Summary of REE site coordination polyhedra for the various REE bearing silicate minerals. 101 Table 4.2 Pixel counts per sample used to produce average spectrum in VNIR and SWIR ranges. ........ 105 Table 4.3 Electron microprobe compositions for mosandrite, cerite, and kainosite. ................................ 108 Table 4.4 Electron microprobe compositions for zircon. .......................................................................... 109 Table 4.5 Electron microprobe compositions for eudialyte samples (LREE Group). .............................. 110 Table 4.6 Electron microprobe compositions for eudialyte samples (HREE Group). .............................. 111 Table 4.7 Chemical variation relevant to reflectance spectroscopy, as well as probable ages and geological settings of zircon samples. ................................................................................................................... 114 Table 4.8 Prominent absorption features of cerite in the VNIR and SWIR ranges. ................................. 118 Table 4.9 Prominent absorption features of mosandrite in the VNIR and SWIR ranges.......................... 121 Table 4.10 Prominent absorption features of kainosite in the VNIR and SWIR ranges. .......................... 124 Table 4.11 Prominent absorption features of zircon samples in the VNIR range. .................................... 128 Table 4.12 Prominent absorption features of zircon samples in the SWIR range. ................................... 129 Table 4.13 Prominent absorption features of HREE-enriched eudialyte samples in the VNIR and SWIR ranges. .................................................................................................................................................. 134 ix Table 4.14 Prominent absorption features of LREE-enriched eudialyte samples in the VNIR and SWIR ranges. .................................................................................................................................................. 135 Table 4.15 Cation site parameters for Yb3+ in REE-bearing silicates. ...................................................... 141 Table 5.1 Modeled modal mineralogy from Rietveld refinements. .......................................................... 150 Table 5.2 Best R2 values and associated thresholds for linear regressions between absorption proxy values and REE concentrations. ...................................................................................................................... 157 Table 5.3 Recalculation of thresholds and Best R2 values without sample Box104-5 (note that Box120 values are the same). ............................................................................................................................ 158 Table 5.4 Estimations of Nd, Sm, Pr and TREE+Y ore grades (all in ppm) through hyperspectral imaging and comparison with concentrations determined through traditional geochemical analyses. ............. 162 x List of figures Figure 1.1 Chondrite (CI) and REE-Ore values for the lanthanides in parts per million (CI data from McDonough and Sun 1995, REE Ore values from Ashram Zone, Eldor Carbonatite, Commerce Resources, Laferriere 2011). .................................................................................................................... 5 Figure 1.2 Chondrite-normalized plot for carbonatite REE-Ore grading ~1.6 wt.% TREO from Ashram Zone at Eldor Carbonatite Complex (normalization data from McDonough and Sun 1995). ................. 5 Figure 1.3 Ionic radii (Å) for REE3+ and selected elements with coordination of 6, 8 and 12 (Shannon 1976). ....................................................................................................................................................... 7 Figure 1.4 Carbonatite Complex Schematic from after Winter (2009) and ternary classification plot for carbonatites – rocks are from Mountain Pass carbonatite complex (data from Castor 2008a). ............... 9 Figure 1.5 Positions of diagnostic absorptions (from Hunt 1977). Widths of black bars indicate the relative widths of absorption bands. ...................................................................................................... 13 Figure 1.6 Basic orbital energy splitting in various isometric crystal fields: (a)-cubic, (b)-dodecahedral, (c)-tetragonal, (d)-barycentre, (e)-octahedral (from Burns 1993). ......................................................... 19 Figure 1.7 Splitting of energy levels of the five d orbitals by cubic (e, f) and octahedral crystal fields (a – d). The Jahn Teller effect describes the subsplitting of the orbital groups by term δ, due to asymmetric distortion of a cubic environment (from Burns 1993). .......................................................................... 20 Figure 1.8 Nomenclature of the energy levels for the lanthanide ions using Ho3+ as an example (from Walsh 2006). Splitting into sequential levels is via (1) electron repulsion, (2) spin-orbit interactions and (3) crystal field. The last division represents a single multiplet (5I5, near 890 nm / 11235 cm-1) on the Dieke Diagram. ................................................................................................................................ 25 Figure 1.9 “Dieke Diagram” of Ln3+ intraconfigurational transitions with values in X1,000 cm-1. Lines represent free ion energy levels with electron repulsion and spin-orbit interactions but before effects of a crystal field (from Dieke et al. 1968). ................................................................................................. 26 Figure 2.1 (a) Coordination polyhedron for the Ce1 site in parisite (Ni et al. 2000); also applicable to bastnaesite and synchysite. Ce1 (green) atom is coordinated with F1, F2 and F3 (lavender) atoms, whose plane is roughly perpendicular to c-axis, and 6 oxygen (red) atoms O11, O23, O32, O42, O53 and O61. Overall coordination number of 9 in a distorted tricapped trigonal prismatic arrangement. (b) Parisite crystal structure from Ni et al. (2000), (c) bastnaesite crystal structure from Ni et al. (1993), and (d) synchysite crystal structure from Wang et al. (1994). Atom colouring: red=oxygen, green=REE, lavender=F, brown=C. Polyhedra colouring: green=REEO6F3, dark blue=CaO8, brown=CO3. ........................................................................................................................................... 36 xi Figure 2.2 Bond characteristics for the CO3 radical polyhedra in REE fluorocarbonate. Data from Table 1 and references therein, Min=minimum, Mid=middle, Max=maximum. ............................................... 39 Figure 2.3 (a) Hyperspectral imagery of the synchysite-bearing sample from Narsarsuk in the short wave infrared (SWIR). (b) Mixture tuned matched filtering (MTMF) MF abundance results for synchysite in SWIR scene and (c) regions of interest (ROI) derived from thresholded MTMF results for generating an average spectrum. (d) Annotated digital photograph of the same sample from nominally the same perspective. Arrows in (c) and (d) point to the same patches of synchysite. Monomineralic patches of synchysite are on the order of 2 mm by 2 mm, large enough for successful imaging with a spatial resolution of 0.241 mm by 0.241 mm, to enable isolation of pure monomineralic pixels with confidence for averaging of their spectra. .............................................................................................. 43 Figure 2.4 Chondrite-normalized REE plot from microprobe results. Normalization values from McDonough and Sun (1995). Erbium (Er), Tm, Yb and Lu are below detection for all samples. ........ 46 Figure 2.5 Stacked spectra of bastnaesite in VNIR (left, 500 to 1000 nm) and SWIR (right, 975 to 2530 nm) from the two sisuROCK instrument cameras. Spectra from top to bottom: Sichuan (S), Burundi (B), Madagascar (M), Karonga (K). ...................................................................................................... 50 Figure 2.6 Stacked spectra of parisite in VNIR (left, 500 to 1000 nm) and SWIR (right, 975 to 2530 nm) from the two sisuROCK instrument cameras. Spectra from top to bottom: Muzo (M), Snowbird (SB). ............................................................................................................................................................... 51 Figure 2.7 Spectra of synchysite from Narsarsuk in VNIR (left, 500 to 1000 nm) and SWIR (right, 975 to 2530 nm) from the two sisuROCK instrument cameras. ....................................................................... 52 Figure 2.8 VNIR (500 to 1000 nm) spectra of bastnaesite (B, top), parisite (P, middle) and synchysite (S, bottom). Italic numbers denote groups with probable origin described in Table 3. Lines denote prominent absorption features with wavelength position, shaded boxes represent the approximate Full Width at Half Max for each absorption or absorption cluster, borderless box indicates narrow feature. Stacked spectra from sisuROCK instrument. ........................................................................................ 53 Figure 2.9 SWIR (975 to 2530 nm) spectra of bastnaesite (B, top), parisite (P, middle) and synchysite (S, bottom). Italic numbers denote groups with probable origin described in Table 4. Lines denote prominent absorption features with wavelength position, shaded boxes represent the approximate Full Width at Half Max for each absorption or absorption cluster, borderless box indicates narrow feature. Stacked spectra from sisuROCK instrument. ........................................................................................ 54 Figure 3.1 Coordination polyhedra for rare earth element cations in monazite, xenotime, and britholite. Data for polyhedra from Oberti et al. (2001) and Ni et al. (1995). ....................................................... 62 Figure 3.2 Bond distances for REE polyhedra in monazite, xenotime, and britholite, as described in the text. Multiple bond lengths are indicated where found (e.g., 4×O), the OH bond of REE2 in britholite xii is labeled, and the vertical bars illustrate the minimum, mean, and maximum bond lengths. Data for polyhedra lengths from Oberti et al. (2001) and Ni et al. (1995). ......................................................... 63 Figure 3.3 Example of false colour VNIR hyperspectral reflectance imagery – the sample is xenotime from Novo Horizonte “C”, looking perpendicular to the c-axis. For this particular scene, 2097 pixels were included in a ROI for calculating the averaged spectrum. Crystal measures 0.75 x 2.25 cm. ...... 66 Figure 3.4 Chondrite-normalized REE plots for monazite (top left), britholite (top right), xenotime (bottom left) and selected samples on combined plot (bottom right). ................................................... 71 Figure 3.5 Example spectrum showing types absorption band labels with data points indicated by diamonds. Spectrum is of monazite (UofA Unknown sample) – see Figure 3.6 and Table 3.5. ........... 72 Figure 3.6 Spectra of characterized monazite samples in VNIR (left, 500 to 1000 nm) and SWIR (right, 950 to 2530 nm) with absorption band clusters shown and prominent absorption lines of the UofA sample. From top down spectra are from Serra Verde (red, SV), UofA Unknown (blue), and Elk Mountain (black, EM). ........................................................................................................................... 73 Figure 3.7 Spectra of characterized xenotime samples in VNIR (left, 500 to 1000 nm) and SWIR (right, 950 to 2530 nm) with absorption band clusters shown and prominent absorption lines of the GQ sample. From top down spectra are from Serra Verde (purple, SV), Gunter Quarry (black, GQ), Novo Horizonte J (blue, NH-J), and Novo Horizonte C (red, NH-C). ............................................................ 76 Figure 3.8 Spectra of characterized britholite samples in VNIR (left, 500 to 1000 nm) and SWIR (right, 950 to 2530 nm) with absorption band clusters shown and prominent absorption lines of the CMNOC sample. Upper spectrum (blue) is from 001-Mariano sample and lower (green) is from CMNOC F90-8 sample. ................................................................................................................................................... 80 Figure 3.9 Representative spectra for xenotime (X, top, red), britholite (B, middle, green) and monazite (M, bottom, black) samples in the VNIR range (500 to 1000 nm). Absorption band clusters are shown with bars and prominent absorptions are labeled according to text descriptions. .................................. 84 Figure 3.10 Representative spectra for xenotime (X, top, red), britholite (B, middle, green) and monazite (M, bottom, black) samples in the SWIR range (950 to 2530 nm). Absorption band clusters are shown with bars and prominent absorptions are labeled according to text descriptions. .................................. 84 Figure 3.11 Reflectance spectra near 978 nm for samples showing absorptions at this wavelength. Spectra are labeled and Er and Yb contents are given in parentheses (Er2O3 wt.% / Yb2O3 wt.%). .................. 86 Figure 3.12 The 5I7 multiplet absorption characteristics in Ho3+ doped Spectralon (left, in reflectance) and Ho3+ doped YGG (right, transmission, plot from Gruber et al. 2009). Note the differences in wavelength ranges and overall absorption strength patterns. ................................................................ 87 xiii Figure 3.13 Reflectance spectra from the USGS Spectral Library (Version 06a). From uppermost to lowermost near 2500 nm: Chlorapatite (black), fluorapatite (red), hydroxyl-apatite (pink) and monazite (blue). ..................................................................................................................................................... 89 Figure 4.1 Eudialyte spectra from Red Wine Complex (Kerr et al. 2011). ................................................ 96 Figure 4.2 Coordination polyhedra for Ln3+ in various REE bearing silicates (see text for details). Top row includes polyhedra for zircon (Zr site, coordinated with 8 oxygen), kainosite (REE site, coordinated with 8 oxygen), and cerite (3 distinct but similar sites coordinated with 8 oxygen and 1 OH-). Middle row includes the two sites in eudialyte, M1 (left, coordinated with 6 oxygen) and Na4 (~9 coordinated, see text for comments). Bottom row includes polyhedra for mosandrite, the similar M4 and M5 sites (7-coordinated to 6 oxygen and 1 OH-) and the M3 site (coordinated to 2 oxygen and 6 mixed anion sites). Red spheres are oxygen, grey and green spheres are mixed anion sites and larger multicoloured spheres inside polyhedra are Ln3+. ....................................................................................................... 102 Figure 4.3 Imagery for Mount Malosa zircon hand sample in (A) SWIR False Colour, (B) SAM match strength to input spectrum (located at crosshairs), (C) SWIR false colour with overlapping ROIs based on tightening SAM thresholds (ROI colours match spectra in Figure 4.4), and (D) digital photograph. The box in (D) is placed around a prominent cluster of zircon crystals. ............................................. 106 Figure 4.4 Mean spectra in the SWIR of ROI based on different SAM thresholds, as labeled, for illustrative purposes of generating an average spectrum. The top black spectrum is the input spectrum from a single pixel, the purple spectrum is the mean from pixels included when the SAM threshold is set to >0.02, resulting in 40 pixels. Other spectra (colour coded to the corresponding ROI) are from less strict SAM thresholds as labeled and therefore larger ROIs. ........................................................ 107 Figure 4.5 Selected Chondrite-normalized REE plots for zircon, kainosite and cerite. Missing points due to analytes being below detection. ....................................................................................................... 112 Figure 4.6 Selected Chondrite-normalized REE plots for eudialyte and mosandrite. Eudialyte CMN72-24 (higher LREE) is from MSH while hand sample F92-23 with both eudialyte (higher HREE) and mosandrite originates from Kipawa. Missing points due to analytes being below detection. ............. 112 Figure 4.7 Relative REE–Ca–Mn–Th compositional trends of eudialyte samples. Values in wt.%. ....... 115 Figure 4.8 Example spectrum showing types absorption band labels with data points indicated by diamonds. Spectrum is of monazite (UofA Unknown sample). .......................................................... 116 Figure 4.9 Reflectance spectra of cerite in the VNIR (500 – 1000 nm) and SWIR (975 – 2530 nm) ranges. Clusters are indicated by labeled thick horizontal lines and prominent absorptions highlighted in the index tables are labeled with wavelength position. ............................................................................. 117 xiv Figure 4.10 Reflectance spectra of mosandrite in the VNIR (500 – 1000 nm) and SWIR (975 – 2530 nm) ranges. Clusters are indicated by labeled thick horizontal lines and prominent absorptions highlighted in the index tables are labeled with wavelength position. .................................................................... 120 Figure 4.11 Reflectance spectra of kainosite in the VNIR (550 – 1000 nm) and SWIR (975 – 2530 nm) ranges. Clusters are indicated by labeled thick horizontal lines and prominent absorptions highlighted in the index tables are labeled with wavelength position. .................................................................... 123 Figure 4.12 Stacked reflectance spectra of zircon samples in the VNIR (550 – 1000 nm) and SWIR (975 – 2530 nm) ranges. Clusters are indicated by labeled thick horizontal lines and prominent absorptions highlighted in the index tables are labeled with wavelength position. Uranium related features are distinguished by lettered clusters, yellow horizontal bars and italicized wavelength labels. From top down, Mt Malosa (pink, MM) Green River (blue, GR), North Burgess (green, NB), Mudtank (red, MT), St Peters Dome (black, SP). ....................................................................................................... 126 Figure 4.13 Reflectance spectra of zircon samples in the VNIR (550 – 1000 nm) and SWIR (975 – 2530 nm) ranges. Clusters are indicated by labeled thick horizontal lines and prominent absorptions highlighted in the index tables are labeled with wavelength position. Uranium related features are distinguished by lettered clusters, yellow horizontal bars and italicized wavelength labels. Colour schemes remain for the unstacked spectra: Mt Malosa (pink, MM), Green River (blue, GR), North Burgess (green, NB), Mudtank (red, MT), St Peters Dome (black, SP). ............................................. 127 Figure 4.14 Reflectance spectra of eudialyte samples in the VNIR (550 – 1000 nm) and SWIR (975 – 2530 nm) ranges. Clusters are indicated by labeled thick horizontal lines and prominent absorptions for sample “Kipawa-Mariano” (black spectrum) are labeled with wavelength position. .................... 132 Figure 4.15 Stacked reflectance spectra of eudialyte samples in the VNIR (550 – 1000 nm) and SWIR (975 – 2530 nm) ranges. Clusters are indicated by labeled thick horizontal lines and prominent absorptions highlighted in the index tables are labeled with wavelength position. The stacked VNIR spectra are ordered by LREE:HREE ratio – upper samples above CMNOC478 have HREE enrichment greater than 1. The stacked SWIR spectra are ordered by the position of the reflectance maximum between 1050 and 1400. Clusters are repeated for the different groupings of samples. ...................... 133 Figure 4.16 Stacked reflectance spectra from representative REE-bearing silicate minerals in the VNIR (500 to 1000 nm) and SWIR (975 - 2530 nm) ranges. Clusters as described in text are indicated by labeled thick horizontal lines with prominent absorptions identified by tick marks. .......................... 138 Figure 4.17 Selected samples showing the Er3+-Yb3+ related absorption band near 978 nm. Spectra are labeled with sample name and Er and Yb contents in parentheses (Er2O3 wt.% / Yb2O3 wt.%). ........ 142 Figure 4.18 Reflectance spectra (left) of selected REE bearing silicates and their continuum removed spectra (right) displaying relative intensity and positional differences for Nd3+ related absorptions xv centered at ~746 nm (4I9/2 4F7/2+4S3/2), ~803 nm (4I9/2 4F5/2+2H9/2) and 875 nm (4I9/2 4F3/2). Influence from Dy3+ in these samples is minimal but would be greatest in the ~803 nm cluster. Weight % of Nd2O3 for each sample is given in parentheses. .......................................................................... 143 Figure 5.1 Schematic of sawn core sample. Subsampling followed the red dashed line. Hyperspectral imaging included the original top surface of the large half and the newly exposed portion of the small half. The smaller portion was sent for destructive chemical and mineralogical testing. ..................... 147 Figure 5.2 Boxes 104 and 120 of pre-sampled drill core, PQ diameter (85 mm core diameter). Note lengthwise saw marks for subsampling during this study and assay tags indicating original 1 meter assays by the company. ........................................................................................................................ 148 Figure 5.3 Example of dry drill core as originally acquired (photograph, left), wet drill core during subsampling (photograph, middle), and hyperspectral image of subsampled drill core (VNIR false colour, right). Box 104 – Subsample 3. ............................................................................................... 148 Figure 5.4 Weighted assay values from subsampling of the REE (x-axis) plotted against the REE assay values for the same interval as provided by the company. .................................................................. 149 Figure 5.5 Example of absorption band depth calculation for a single pixel’s spectrum. Input bands for calculation are denoted with thin red lines on subset plot at right. Prominent Nd3+ related absorption locations are labeled. ............................................................................................................................ 151 Figure 5.6 Example Nd3+-related absorption depth image (left) and histogram (right) for the 796 nm Nd3+ related absorption band. Colour ramp for image relates to relative strength of absorption (dark=no absorption, blue-green=moderate absorption, red-white=strong absorption). Histogram at right shows the distribution of absorption strengths for each pixel in its associated “image”. ............................... 153 Figure 5.7 Annotated histogram for the 796 nm Nd3+ related absorption band from Box120-Sample1. Vertical line represents a Sum Value Threshold of 0.005, above which the pixel’s values are summed (represented by the red area) and normalized to the total number of pixels in the scene (e.g., 57,777 pixels in VNIR) to generate a single Proxy Value for Box120-Sample1. Note that the displayed graph covers only absorption depths from 0 to 0.025. ................................................................................... 154 Figure 5.8 Example of Proxy Values for three different Sum Value Thresholds (0.02, 0.0075 and 0.005) applied to the 796 nm absorption for Box 120 plotted against Nd geochemistry. R2 values are given for each linear regression. For example, using a threshold value of 0.02 for the 6 samples of Box120 establishes a weak linear regression between the resulting Proxy Values for each subsample and the subsample’s Nd content (R2=0.1362). ................................................................................................. 155 Figure 5.9 Different “Sum Value Thresholds” and their resulting R2 values for the 796 nm Nd3+ related absorption for Box 120 subsamples. Note the maximum R2 value of 0.7873 associated with the best xvi threshold value of 0.0075 (see Figure 5.8 and Table 5.2). Lower thresholds interpret noise as “signal” and higher thresholds exclude “true signals”. ...................................................................................... 155 Figure 5.10 Relationship between Nd Proxy Value (from SUM_Nd_741_796_864 images, SumValueThreshold = 0.002, see bolded and underlined values on Table 2) and Nd (ppm) across all subsamples for both Box104 and Box120. Note the outlying location of Box104-Sample5. ............. 160 Figure 5.11 Relationship between Nd Proxy Value (from SUM_Nd_741_796_864 image, SumValueThreshold = 0.002, see bolded and underlined values on Table 3) and Nd (ppm) across all subsamples for both Box104 and Box120 but without Box104-Sample5. .......................................... 160 Figure 5.12 Concentrations of Nd (ppm) vs. TREE+Y (ppm) showing high correlation (left). At right, is the relationship between the Nd Proxy Value “Sum_Nd_741_796_864” (SumValueThreshold = 0.002, see Table 5.3) and the TREE+Y contents (ppm) without Box104-Sample5. ...................................... 161 Figure 5.13 Example plot relating Proxy Values derived from absorptions within hyperspectral image pixels to Nd (ppm) content of the 13 subsamples (black squares, Boxes 104 and 120). The resulting linear regression is plotted (black line) along with the R2 value. “Whole-scene” Proxy Values were calculated for Boxes 104 and 120 and are plotted along the linear regression line, from which Nd (ppm) content can be estimated (Box 104–blue diamond, Box 120–red circle, see Table 5.4)........... 162 xvii Acknowledgements I am grateful for early inspirational discussions with A. Mariano & A. Mariano at the start of the PhD and for their gift of mineral samples that jump started data collection. Michel Picard with the Canadian Museum of Nature, Mackenzie Parker (formerly) with Pacific Museum of the Earth at the University of British Colombia and Andrew Locock with the Geology Museum at the University of Alberta are also thanked for their loans and gifts of samples. Jilu Feng (UofA), Mati Raudsepp (UBC) and Steve Creighton (SRC) were instrumental with data collection. I always looked forward to seeing fellow students and staff both at UofA and UBC – thanks for helping me out as I rushed around frantically trying to get twice as much work done in half as much time as I should have planned! Financial support in the form of a generous scholarship is appreciatively acknowledged from Natural Sciences and Engineering Research Council of Canada. The Mineralogical Association of Canada and the Department of Earth and Ocean Sciences also provided substantial support throughout the duration of the PhD, allowing me to focus on my research objectives and to participate in conferences while raising a family. The sincere and positive encouragement of two wonderful people, Benoit and Lee, were valued at every step. Discussions were always fruitful and not once did I ever feel less than 100% supported – thank you! Finally to my family, thank you with all my heart. Darcie, I truly couldn’t have taken on this project without you. Burwyn and Iris, thank you for all your patience while Dad “works and works and works” and of course all our rejuvenating fun times! To Baby #3, can’t wait to meet you! And to my parents and in-laws, thank you for always being there and for always being able to help us through our transitions. 1 Chapter 1. Introduction and Literature Review 1.1 Introduction The lanthanides are transition elements with electrons occupying the 4f orbitals, and range from lanthanum (chemical symbol La, atomic number 57) through lutetium (chemical symbol Lu, atomic number 71). Together with the element yttrium (chemical symbol Y, atomic number 39) and sometimes scandium (chemical symbol Sc, atomic number 21) they make up what is commonly known as the “rare earth elements” (REE). Lanthanide-bearing compounds are known to produce numerous sharp absorption features in the visible-near infrared to short wave infrared region (VNIR-SWIR, 400 to 2500 nm) due to 4f-4f orbital intraconfigurational electronic transitions, however, a significant knowledge gap exists between the fields of hyperspectral reflectance spectroscopy and REE mineralogy. Narrowing this knowledge gap has relevance in applied science because the mineralogy of REE deposits is critical in understanding their petrogenesis and also has significant implications for their economic viability. If the application of VNIR-SWIR reflectance spectroscopy is to be harnessed by the geological and mineral exploitation communities investigating REE-bearing rocks, a strong understanding of the fundamental spectral characteristics of REE minerals is necessary. Thus, the principal aim of this research was to address the glaring fundamental knowledge gap. Chapters 2, 3 and 4 of the thesis describe the reflectance spectroscopy in the visible to short wave infrared regions (500 nm to 2500 nm) of well-characterized mineral samples. Each of these chapters addresses a different class of minerals and contains literature reviews relevant to the particular minerals in question. These chapters also contain the specific analytical methods used and conclusions within the scope of the chapter. Chapter 5 describes REE ore grade estimation by imaging spectroscopy of drill core from a hydrothermally altered peralkaline ore deposit. Chapter 2 addresses the rare earth element fluorocarbonate minerals bastnaesite, parisite and synchysite. These three minerals are amongst the most important REE minerals and can carry the bulk of the total REE content in a given ore deposit, especially carbonatite-hosted deposits (e.g., Mountain Pass carbonatite, California). These minerals are characterized by a strong preference for the light rare earth elements (LREE: Sc and La through Gd). Chapter 3 addresses the rare earth element phosphate minerals monazite, xenotime and britholite. These are arguably the second most important group of minerals that carry REE. Monazite is enriched in the LREE while xenotime shows preference for the heavy rare earth elements (HREE: Y and Tb through Lu). Both of these minerals occur in a wide variety of rocks, including as important ore minerals in carbonatites (e.g., Mt Weld, Australia), peralkaline intrusive suites (Nechalacho, Canada) and heavy mineral sands (e.g., coastal Australia). 2 Britholite carries both LREE and HREE, and is found in select ore deposits as a volumetrically important ore mineral (e.g., Kipawa deposit, Ontario). Chapter 4 address various rare earth element bearing silicates, including cerite, mosandrite, kainosite, zircon and eudialyte. Each of these minerals can be important to specific ore deposits, and with the exception of cerite can all be appreciably enriched in HREE. The REE bearing silicates are most relevant to peralkaline intrusive suites, a deposit type that Canada is particularly endowed with (e.g., Kipawa, Red Wine Complex, Nechalacho, Strange Lake, Ting, etc…). Rare earth element bearing silicates have not been historically as economically important as the REE fluorocarbonates and REE phosphates, however, they are becoming more important as demand increases for HREE. The secondary objectives of the research were to investigate the application of knowledge gained to assess rock samples from REE ore deposits and to evaluate the collective spectral variability of REE absorptions across different mineral hosts. The goal of REE ore grading via imaging spectroscopy was brought to a “proof of concept” stage and forms Chapter 5. The development of a decision tree for identifying REE minerals by their reflectance spectra is discussed in the final chapter (Chapter 6), along with a summary and conclusion of the work within this dissertation. Appendices can be found at the end of the thesis and include sample photographs and hyperspectral imagery of mineral and rock samples, scanning electron microscopy images and energy dispersive spectra, full electron microprobe compositions, reflectance spectra of REE oxides from various sources, geochemical results from REE-mineralized drill core, and Rietveld refinement results for the REE-mineralized drill core. Brief reviews on the nature of lanthanides (ie., rare earth elements, the “REE”), broad statements about REE mineralogy, geological settings important for REE ore deposits, reflectance spectroscopy and lanthanide spectroscopy are given here in the remainder of Chapter 1. These are not meant to be exhaustive analyses of the various topics, but rather to inform a non-specialist about the intersection of these diverse topics and to place this thesis in its intended context. More detailed mineralogical reviews specific to each group of minerals studied can be found in the appropriate chapter, however, because the thesis was built around the three base manuscripts (fluorocarbonates-published, phosphates-in prep, silicates-in prep) plus future papers (REE Ore Grade Estimate, Spectral Patterns of REE Minerals) there is inevitable repetition of some information. 1.2 The lanthanides (rare earth elements) The lanthanides (Ln) are a series of 15 elements from lanthanum (La, atomic number 57) to lutetium (Lu, atomic number 71). One of these, promethium (Pm, atomic number 61) is not naturally occurring, leaving only 14 for consideration in natural systems. The lanthanides are characterized by the presence of the 4f orbital block in 3 the conventional periodic table of the elements that can accommodate 14 electrons in the seven orbital configurations. Elements of the 5f orbital block are known as the actinides and include uranium (U) and thorium (Th). In geological environments the lanthanides are commonly found together in their trivalent state, with the exception of divalent europium (Eu) and in some cases tetravalent cerium (Ce) and terbium (Tb). The term “rare earth elements” (Table 1.1) comprises the lanthanide group of elements, commonly yttrium (Y, atomic number 39) and sometimes scandium (Sc, atomic number 21). These two latter elements show similar non-f-block electronic structure and are sometimes enriched in similar rocks (Table 1.2). The rare earth elements (REE) are often inconsistently subdivided by the mineral exploration industry and geoscientists into the light rare earth elements (LREE, often La through Gd) and heavy rare earth elements (HREE, often Tb through Lu + Y). Uncommonly a medium rare earth element (MREE) group is defined with variable identity. The rare earth elements are sometimes reported in oxide form, giving the abbreviation REO and sometimes TREO (a summation of the total rare earth oxides). Thorium and uranium often occur in similar geological settings and occur in many of the same minerals as the lanthanides, but are not considered part of the rare earth elements. This is relevant because airborne radiometric surveys are often used in mineral exploration for REE and provide an opportunity for integrated hyperspectral surveys, but also because in some cases the original minerals have become metamict and lose their crystallinity. Table 1.1 The rare earth elements. Atomic Number Element Name Element Symbol Number of Ln3+ f-electrons Effective ionic radius for VILn3+ (Å) Effective ionic radius for VIIILn3+ (Å) 21 Scandium Sc 0 0.745 0.87 39 Yttrium Y 0 0.900 1.019 57 Lanthanum La 0 1.032 1.16 58 Cerium Ce 1 1.010 1.143 59 Praseodymium Pr 2 0.990 0.97 60 Neodymium Nd 3 0.983 1.126 61 Promethium Pm 4 0.970 1.109 62 Samarium Sm 5 0.958 1.079 63 Europium Eu 6 0.947 1.066 64 Gadolinium Gd 7 0.938 1.053 65 Terbium Tb 8 0.923 1.04 66 Dysprosium Dy 9 0.912 1.027 67 Holmium Ho 10 0.901 1.015 68 Erbium Er 11 0.890 1.004 69 Thulium Tm 12 0.880 0.994 70 Ytterbium Yb 13 0.868 0.985 71 Lutetium Lu 14 0.861 0.977 *values from Shannon (1976) 4 Table 1.2 Electronic configuration of the lanthanides. Atomic Number Element Ln0 Ln3+ Spectroscopic term for Ln3+ ground state 57 La [Xe] 5d 1 6s 2 [Xe] 4f 0 1S0 58 Ce [Xe] 4f 1 5d 1 6s 2 [Xe] 4f 1 2F5/2 59 Pr [Xe] 4f 3 6s 2 [Xe] 4f 2 3H4 60 Nd [Xe] 4f 4 6s 2 [Xe] 4f 3 4I9/2 61 Pm [Xe] 4f 5 6s 2 [Xe] 4f 4 5I4 62 Sm [Xe] 4f 6 6s 2 [Xe] 4f 5 6H5/2 63 Eu [Xe] 4f 7 6s 2 [Xe] 4f 6 7F0 64 Gd [Xe] 4f 7 5d 1 6s 2 [Xe] 4f 7 8S7/2 65 Tb [Xe] 4f 9 6s 2 [Xe] 4f 8 7F6 66 Dy [Xe] 4f 10 6s 2 [Xe] 4f 9 6H15/2 67 Ho [Xe] 4f 11 6s 2 [Xe] 4f 10 5I8 68 Er [Xe] 4f 12 6s 2 [Xe] 4f 11 4I15/2 69 Tm [Xe] 4f 13 6s 2 [Xe] 4f 12 3H6 70 Yb [Xe] 4f 14 6s 2 [Xe] 4f 13 2F5/2 71 Lu [Xe] 4f 14 5d 1 6s 2 [Xe] 4f 14 1S0 In general, when moving from a neutral lanthanide atom with basic electronic configuration to a Ln3+ cation, electrons from the 4f, 5d and 6s shells are stripped leaving a clean configuration of a sequentially occupied 4f orbital. Lanthanum and lutetium have no unpaired 4f electrons (f0 and f14), with a sequential pattern of unpaired electrons (e.g., Ce and Yb as f1 and f13) towards gadolinium, which has seven unpaired electrons (f7). The outer radius of the 4f electron shells (~30 pm) for the lanthanides is much less than that of the filled 5s and 5p shells (~200 pm, ~105 pm). Consequently, first assumptions suggest that the local electronic environment of Ln3+ cations interacts primarily with those outer shells, leaving the 4f shell relatively ‘sheltered’ and non-participatory in bonding. Two sets of orbital wavefunctions are in common use depending on the local environment, cubic and non-cubic (“general”), with implications for ion-ligand interactions. The general abundance of rare earth elements is noteworthy in that their crustal concentration (e.g., Ce ~60 ppm, Gd ~5.4 ppm) is actually quite comparable relative to other important industrial metals, such as Cu (~55 ppm) and Sn (~2.5 ppm). The lanthanides display an even-odd abundance pattern that is also reflected in rocks due to the geochemical similarity of the element set (Figure 1.1). Normalization of the REE in a set of geochemical data to carbonaceous chondrite “CI” (assumed to be unfractionated since the formation of the solar system) smoothens this trend and simplifies interpretation (Figure 1.2). 5 Figure 1.1 Chondrite (CI) and REE-Ore values for the lanthanides in parts per million (CI data from McDonough and Sun 1995, REE Ore values from Ashram Zone, Eldor Carbonatite, Commerce Resources, Laferriere 2011). Figure 1.2 Chondrite-normalized plot for carbonatite REE-Ore grading ~1.6 wt.% TREO from Ashram Zone at Eldor Carbonatite Complex (normalization data from McDonough and Sun 1995). The rare earth elements have a variety of uses, especially in the high technology sector. Hatch (2012) lists a variety of uses for the REE and divides them into Process Enablers (e.g., catalytic converters, polishing media) and Technology Building Blocks (e.g., permanent magnets, energy storage). The highest demands are for permanent magnets (using Nd, Pr, Sm, Dy and Tb), catalysts (Ce) and metal alloys (all REE for various specific 0.010.1110100100010000La Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tm Yb LuLog of Concentration (ppm)M&S 1995Example Carbonatite HostedREE Ore(Bastnaesite+Monazite)110100100010000100000La Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tm Yb LuCI NormalizationCI Chondrite Normalized Values for REE Ore in Carbonatite (Bastnaesite+Monazite)Rock / CI (norm)6 properties). Total demand in 2011 was estimated at 105,000 tons of REE oxide and projections for 2016 were for 160,000 tons REE oxides. Their diverse use and critical importance in the tech sector was highlighted in 2010 when China, who produced greater than 95% of the world’s REE supply, decided to reduce international trade quotas. A substantial price increase followed and pushed what was known as the “rare earth boom” in the mineral exploration industry. Since then prices have steadily dropped but awareness of the REE market and its weak points have greatly increased. 1.3 Rare earth element mineralogy 1.3.1 Introduction The mineralogy of rare earth element deposits is particularly relevant to the economics of physical beneficiation and chemical treatment. For example, there are many circumstances where allanite, a REE-bearing epidote mineral of the sorosilicate class, and zircon, a nesosilicate with incorporation of REE, can be found in significant abundance and with significant overall grade in rock samples. Unfortunately, the processes required to extract REE from these phases are onerous and involve significant chemical treatment by sulphuric acid baking and caustic sodium hydroxide cracking (e.g., Great Western Minerals Group Ltd. 2007, Cox et al. 2011). This is overshadowed by the established, simple and lower cost treatment of ores comprising bastnaesite, a REE fluorocarbonate, and monazite, a REE phosphate. Recovery of REE from the mixed silicate-phosphate-fluorocarbonate-oxide ores in test plants from the Nechalacho deposit is ~75% (Cox et al. 2011, Grammatikopoulos et al. 2013) but reagent consumption is high, whereas recovery from monazite and bastnaesite ores from active mines such as Mountain Pass is ~98% (Gupta and Krishnamurthy, 2004). Recent positive developments, however, have been made at Matamec Explorations Inc’s Kipawa deposit comprising dominantly eudialyte, a zirconosilicate, where ~85% REE recovery was reported “using a low-cost proprietary leaching method” (Matamec Explorations Inc, 2011). Despite continued metallurgical studies of REE silicate ores, the reality is that non-bastnaesite or non-monazite deposits will always be benchmarked against established mine sites such as the Mountain Pass carbonatite (California) or the Bayan Obo carbonatite system (Mongolia). Broad concepts of the lanthanide coordination and REE-bearing minerals are briefly discussed here, while more thorough crystallographic-structural settings for REE cations in specific minerals are given in the related chapters. As a general statement, REE mineralogy is complex and subtle differences between the numerous species with similar crystal structures have probably led to many misidentifications in non-mineralogical studies. 7 1.3.2 Coordination states and effective ionic radii The REE predominantly populate roomy but distorted crystallographic sites with coordination numbers of eight or more. The contraction of ionic radius along the lanthanide sequence results in the heavy rare earth elements having a greater predisposition for smaller coordination numbers while the light rare earth elements prefer higher coordination numbers and longer ligand distances. This difference results in certain minerals having distinct enrichment patterns in either the light or heavy rare earths. For example, the allanite group minerals (11-coordinated, Ln-O to ~3.2 Å) are LREE-enriched whereas xenotime (8-coordinated, Ln-O ~2.3 Å) is defined by HREE+Y enrichment. Figure 1.3 shows ionic radii of the REE and other commonly associated elements with various coordination numbers. Figure 1.3 Ionic radii (Å) for REE3+ and selected elements with coordination of 6, 8 and 12 (Shannon 1976). 1.3.3 Important mineral classes The mineralogy of REE dominant phases can be characterized largely by four main broad mineral classes: carbonates, phosphates, silicates and oxides. Other rare REE minerals occupy other classes, however, the most important REE minerals are found in the classes briefly described here. Chapters 2, 3 and 4 provide greater detail for their respective themes. The REE carbonates are dominated by the fluorcarbonate minerals bastnaesite, synchysite, parisite and rontgenite which all share a common stacking arrangement of CeF-CO3 and Ca-CO3 layers in variable ratios. These minerals can incorporate OH and often show substitution of Ba and Sr. Cordylite, burbankite and ancylite are REE-Ba-Sr-bearing carbonates while sahamalite is an REE-Mg-Fe species. Coordination of REE in these minerals is usually either 9 or 10, however mixed anion cases (e.g., O2-, OH- and F-) are present for most. 0.50.70.91.18.104.22.168La3+Ce3+ Pr3+Nd3+Pm3+Sm3+Eu3+Gd3+Tb3+Dy3+Ho3+ Er3+Tm3+Yb3+ Lu3+ Y3+ Fe3+Mg2+ Zr4+Sc3+Fe2+U4+Th4+Ca2+ Sr2+Pb2+Ba2+Effective Ionic Radius in AngstromsCN=6CN=8CN=128 Consequently, bond lengths vary considerably (from ~2.37 to 2.77 Å) and distortion of the coordination polyhedra is significant. REO contents for these minerals are generally high, with bastnaesite in particular showing up to ~75%. The phosphate classes of REE-bearing minerals show a range of coordination states from seven to 12 with bond lengths as short as 2.25 and as long as ~3.2 Å. Cation and anion substitutions in these minerals can be substantial with complex crystal chemical implications, such as in britholite. On the other hand, the crystal structures of monazite and xenotime have been well studied and are relatively well understood. The REE-bearing silicates are a diverse set of minerals with a wide range of coordination and overall crystal chemistry and structure. Coordination numbers for cation polyhedra range from 6 (gittinsite) to 11 (allanite), with some cases showing mixed anion coordination (cerite, 8 oxygen + 1 OH-/F-) and other minerals having more than one distinct REE site (mosandrite). Documented bond lengths vary accordingly, ranging from ~2.2 to ~3.2 Å. Crystal structures of many of the REE silicates are still not fully understood. The oxide class comprises a number of different groupings based on how the oxide is structured, the number of distinct cation sites as well as the identity and charge of the cations. Most REE-bearing oxide minerals group together within Dana’s Class 08, “Multiple Oxides with Nb, Ta, and Ti”, with the REE3+ and similar cations, such as Ca2+, residing at an A site while the high field strength elements (HFSE) reside at a B site. These minerals are often U and Th bearing, and metamict samples seem to be the norm rather than the exception in the literature and in mineral museums. 1.4 Principal geological settings for REE deposits The following sections briefly describe the geological models of the more important deposit types with respect to rare earth element and rare metal deposits. It will not be an exhaustive discussion on the nature of these deposits, which are often still under debate and poorly understood, but rather a general overview with a focus on the types of ore and gangue mineral assemblages expected. The articles of Pell (1994), Richardson and Birkett (1996a, b), Castor (2008a), Berger et al. (2009), and references therein were used as primary sources. 1.4.1 Carbonatites Carbonatites are igneous rocks with modal carbonate mineralogy greater than 50%. This primarily includes calcite, dolomite, ankerite and siderite, which drive the classification of carbonatite by way of a CaO:MgO:FeO+Fe2O3+MnO ternary diagram, with calicocarbonatites predominating the global population (Figure 1.4; Woolley 1982, 1987, Bell 1989, Woolley and Kempe 1989, Mitchell 2005, Woolley and Kjarsgaard 2008). Carbonatites can be extrusive or intrusive, however, their rapid weathering precludes preservation in the rock record, although one natrocarbonatite volcano, Ol Doinyo Lengai, is active today (e.g., Dawson 1962). 9 Intrusive carbonatite can take on many geometries including sills, dykes, and plugs and are commonly associated with alkalic complexes and chemically diverse rock types. Later carbo-hydrothermal alteration and overprinting is common and brecciation is a textural feature in many carbonatite complexes (e.g., Mountain Pass, Wicheeda, Eldor, Niobec, Bayan Obo etc…). Dynamic recrystallization and structural “drawing out” during deformation of these rheologically weak rocks might overstate sill or dyke morphologies. Figure 1.4 Carbonatite Complex Schematic from after Winter (2009) and ternary classification plot for carbonatites – rocks are from Mountain Pass carbonatite complex (data from Castor 2008a). Carbonatites and their associated intrusive complexes can host a variety of rare elements and industrial minerals and in some cases precious and base metals. They have particular significance for supply of REE, Nb, fluorite, and phosphate. Classic and selected carbonatite-related ore deposits and intrusive complexes include Mountain Pass (California/USA, Castor 2008b), Araxa (Brazil, Traversa et al. 2001), Bayan Obo (Mongolia, Drew et al. 1990, Le Bas et al. 1992, Chao et al. 1997, Yang and Le Bas 2004, Wu 2008, Smith et al. 2015), Niobec (Quebec/Canada, Eby 1975), Mt Weld (Australia, Lottermoser 1990, Middlemost 1990), Palabora (South Africa, Groves and Vielreicher 2001) and Wicheeda (British Columbia/Canada, Pell 1987, Mader and Greenwood 1988, Dalsin et al. 2015). Bayan Obo and Palabora have also been argued to be more related to Iron-Oxide-Copper-Gold deposits than carbonatites, however, their unique characteristics and varied mineralization impedes obvious “categorization”. Modal mineralogy of carbonatite-related REE deposits is dominated by carbonate minerals as per the classification scheme, however the following are also common: aegirine, albite, arfvedsonite, biotite, celestite, chalcopyrite, diopside, feldspar, fluorite, galena, goethite, hematite, kaolinite, magnetite, montmorillonite, 10 nepheline, olivine, phlogopite, pyrite, quartz, riebeckite, richterite, strontianite, thorite, tremolite, vermiculite. Because of the unusual nature and geochemistry of carbonatites as well as pervasive alteration, ore mineralogy can be very diverse. The following are common REE-Nb ore minerals found in carbonatite deposits: allanite, ancylite, baddeleyite, bastnaesite, britholite, burbankite, cerite, columbite, cordylite, euxenite, fergusonite, fersmite, florencite, goyazite, monazite, parisite, pyrochlore, synchysite, sahamalite, tantalite, xenotime, zircon. Each deposit will have its own distinct mineral assemblages. 1.4.2 Peralkaline intrusive complexes Peralkaline intrusive rocks are high in alkalis (Na2O+K2O>Al2O3), variable in SiO2 content, and are found in large intrusive multiphase igneous complexes with varied rock types typically in anorogenic extensional settings. Rare earth element and rare metal (e.g., Be, Ta, Nb, Zr, Hf, U, Th) mineralization is typically associated with late phases, including pegmatites, veins, and hydrothermal overprinting/remobilization. Richardson and Berkitt (1996b) divide peralkaline rare metal deposits into primary magmatic and hydrothermal/metasomatic but themselves cede that “end members” are rare, if not absent. Example deposits include Strange Lake (Quebec-Labrador, Vasyukova and Williams-Jones 2014) as ‘magmatic’ and Nechalacho/Thor Lake (NWT, Trueman et al. 1988, Sinclair et al. 1994, Sheard et al. 2012) as ‘metasomatic’ while Kipawa (Quebec, Currie and Van Breeman 1996, Camus and Laferriere 2010) has been described as ‘uncertain’. Illimausaq (Greenland), Bokan Mountain (USA, Alaska, Thompson 1988, Robinson et al. 2011) and Lovozero of the Kola Peninsula (Russia, Kogarko et al. 2002) are included in the peralkaline category. Mineralogy of these deposits can be very diverse due to the high alkali and variable silica content ranging from undersaturated to quartz normative; however, alkali amphiboles and pyroxenes are diagnostic of peralkaline systems (Sorensen 1997). The collection of rare elements, including the high field strength elements (HFSE), hydrothermal overprinting and discrete zoning in many deposits adds to the complexity of mineral assemblages and unique minerals at specific localities. Illimausaq and Lovozero are both host to 300+ identified minerals, are type localities for over 50 mineral species and have mineralogical publications spanning back to the 19th century. Peralkaline deposits of magmatic origin tend to be large tonnage (>50 Mt) and later orthomagmatic-metasomatic-hydrothermal fluids can be effective at concentrating desirable commodities in more discrete areas (<5 Mt). Common gangue minerals of peralkaline-related rare metal mineralization include aegirine, albite, biotite, calcite, chlorite, diopside, dolomite, fluorite, K-feldspar, magnetite, muscovite, nepheline, phlogopite, quartz, richterite, riebeckite, scapolite, and titanite. Common rare metal ore minerals include aeschynite, allanite, bertrandite, bastnaesite, britholite, columbite-tantalite, eudialyte, fergusonite, fersmite, gadolinte, gittinsite, helvite, loparite, monazite, mosandrite, parisite, phenakite, polycrase, pyrochlore, samarskite, synchysite, thorite, vlasovite, xenotime, and zircon. 11 1.4.3 Other deposit types Other sources of REE include the South China ion-adsorped clays, various heavy mineral placer deposits and uranium processing residuals (raffinates). The South China Clays are actually fairly low grade, however, the REE are adsorped onto clay particles and easily leached using oxalic acids. The end product is an REE precipitate with even distribution across the element suite but environmental impact is such that mining similar global deposits would be unlikely to get government approvals. Heavy mineral sands (and analogous paleoplacers) in either alluvial or beach placer deposits will have modal compositions related to the source rocks, however, most recognized deposits are characterized by monazite, xenotime, and zircon. Because of the geochemical association between the REE and U/Th, the tailings and raffinates from active U and/or IOCG mines often have recoverable REE contents (e.g., Pea Ridge, Olympic Dam). 1.5 Reflectance spectroscopy 1.5.1 Introduction and general principles The aim of reflectance spectroscopy is to obtain and interpret diagnostic absorption, reflection and emission properties of a target (i.e., mineral) as compared to incident light (e.g., Clark 1999). By relying on reflected light, sample preparation is greatly reduced and the analytical environments and definitions of sample targets dramatically expand, as compared to transmitted light spectroscopy commonly used by chemists in a laboratory setting. In the remote sensing field this has allowed reflectance spectrometers to be mounted on satellites and aircraft with targets for incident sunlight including forest canopies, urban development and planetary landscapes. Where possible, remote sensing surveys are best complemented by ground and laboratory based studies to obtain ‘pure endmember data’ for comparison. Early remote sensing spectrometers had coarse spectral resolution and were termed multispectral (e.g., Landsat satellites) with a few strategically placed bands (Perry 2004). More recently, select satellites and aircraft have been deployed with high spectral resolution imaging spectrometers (e.g., Goetz et al. 1985, Vane and Goetz 1993 and more recent review by van der Meer 2012) . This increase in spectral (and also spatial) resolution results in the collection of very large amounts of data. New algorithms for working with such large dataset are constantly being developed and processing speeds of dedicated computers are continually being optimized. Ranges used in reflectance spectroscopy are roughly divided into ultraviolet (UV, 1 to 400 nm), visible (VIS, 400 nm to 700 nm), near infrared (NIR, 700 to 1000 nm), short wave infrared (SWIR, 1000 to 2500 nm), mid infrared (MIR, 2.5 to 30 μm) and thermal infrared (TIR, 30 μm to 1 mm). The term VNIR (visible to near infrared) is typically from 400 nm to 1000 nm. The advancement of spectrometers and their respective spectral ranges coupled with different researchers and different research objectives has led to differences in defining the various spectral ranges (e.g., Hunt 1977, Clark 1999, Van der Meer et al. 2012, Robles-Kelly and Huynh 2013). 12 The interaction between electromagnetic radiation and crystalline matter results in reflection, refraction, absorption and emission (e.g., Clark 1999). Together, reflection and refraction define the geometrical re-distribution of photons and can be collectively termed scattering. As light passes through crystalline material it can also be absorbed by electronic and vibrational processes. Relevant to reflectance spectroscopy in field based studies, emission occurs mainly from relaxation of excited electrons (photoluminescence) and from material above 0 K (grey body emission), however, their contributions in the visible to shortwave infrared are generally negligible in the context of the proposed studies. 1.5.2 Specular and diffuse reflection Specular reflectance is the case whereby mirror-like reflections occur off a polished surface with high reflectivity, as with a silver mirror or polished thin section (e.g., Burns 1993, Hapke 1993). In these cases light is assumed to have not significantly penetrated the material before being reflected back towards the spectrometer. Diffuse reflectance is the case whereby there are multiple reflections of light off rough surfaces and boundaries of small particles. In these cases it is assumed that the incident light has penetrated the material before being reflected back towards the spectrometer. Therefore, diffuse reflectance is more effective at allowing incident light to acquire diagnostic ‘signals’ of the target before being reflected back to the spectrometer for detection and analysis. The longer the ray path interacts with the target, the greater the absorption at diagnostic wavelengths from diagnostic features. 1.5.3 Absorption bands Absorption bands are local attenuations of reflected light at specific wavelengths, and are generated by the target absorbing energy. The location, intensity and shape of these features can be used qualitatively and quantitatively to discern chemical information about the target material (e.g., Mustard 1992, Salisbury 1993, Cloutis et al. 2006). In the VNIR-SWIR range, electronic and vibrational processes are the most relevant for reflectance spectroscopy of geological materials. Electronic processes (see Figure 1.5, from Hunt 1977) in minerals that generate absorption features in the VNIR-SWIR include crystal field effects, intervalence and metal-oxygen charge transfers, band gaps and color centers. Crystal field effects deal primarily with intraorbital transitions of unpaired electrons belonging to elements of the transition or lanthanide elements. Intervalence charge transfer involves the transfer of typically one electron between cations in a crystal with mixed valences whereas metal-oxygen charge transfers occur between the cation and anion. Colour centers are the result of crystal defects in which excitable electrons can reside in order to achieve charge balancing of the crystal. Band gaps occur in semi-conducting minerals with predominantly covalent bonding and absorbance features are generated from the excitation of electrons from a lower valence band into a higher energy conductance band. 13 Vibrational processes (see Figure 1.5, from Hunt 1977) that generate absorption features in the VNIR-SWIR and into the mid and longwave infrared are the result of resonance between incident electromagnetic radiation and chemical bonds in minerals. In particular, the presence of water, hydroxyl and carbonate generate absorption features in the 1000 to 2500 nm through combinations and overtones of fundamental vibrations (e.g., Metal-OH bending or OH stretching) while the bonds between oxygen and silicon, phosphorous and sulfur are more prominent at longer wavelengths beyond 2500 nm. Figure 1.5 Positions of diagnostic absorptions (from Hunt 1977). Widths of black bars indicate the relative widths of absorption bands. 14 1.5.4 Transmission to reflectance General relationships of absorption features between reflectance spectroscopy and transmission/absorption spectroscopy are that the locations of reflectance minima (troughs) correspond to absorption maxima (peaks). However, the two types of spectra are not always directly comparable with respect to relative intensities, widths and shapes of features. The Kebulka-Munk remission function is sometimes used to convert diffuse reflectance spectra into absorption/transmission spectra but has limitations (e.g., Hapke 1993). Consequently, transmission/absorption studies can be useful to guide interpretation of reflectance spectra but the two fields do not produce interchangeable datasets. This is important because most of the studies of lanthanide-doped synthetic material use transmission spectroscopy to determine absorption coefficients of specific absorption bands in the various crystal lattices and glasses. 1.5.5 Grain size Grain size is important because the further light can propagate through a medium the more the grain is able to absorb light and produce a clearer “signal” (e.g., Clark and Roush 1984, Clark 1999). When the particle diameter is greater than the wavelength of incident light, light partially reflects on “macroscopic” crystal faces and boundaries and partially penetrates into the sample undergoing refraction and internal reflections before emerging from the surface of grains in a diffuse manner. As grain size decreases, the probability of light continually being reflected instead of refracted is larger because the surface to volume ratio increases. Mustard and Hays (1997) showed in the VNIR-SWIR range that at smaller grain sizes, overall reflectance is generally higher (brighter) and larger grain sizes show decreased overall reflectance (darker). Hapke (1993) discussed the effect of grain size on spectral contrast for absorption bands and noted that small grain sizes show optimal spectral contrast, whereas very small and large grain sizes show lesser spectral contrast. In the context of this dissertation, all mineral samples can be considered coarse grained (large) and display relatively suppressed overall reflectance and moderate spectral contrast for absorption bands. 1.5.6 Background continua of reflectance spectra Clark (1999) notes that a reflectance spectrum can be described by two parts: a continuum and individual features. The continuum is generally accepted to be a background absorption that reduces overall albedo and upon which specific absorption bands are superimposed. Depending on the spectral range of investigation, contributions to the continuum can include broad absorptions within the range of investigation (e.g., ferrous iron), tails or wings centered outside the range of investigation (e.g., from absorption bands in the UV), grain size effects, illumination and influence from intimate mixtures. The continuum is also sometimes referred to as the convex hull of a reflectance spectrum. In theory, the continuum has a shape that can be described by a series of typically very broad absorptions with Gaussian shapes (e.g., Clark and Roush 1984, Kokaly 2000), however, in practice it is usually described by a series of local linear sections across the spectral range of interest (e.g., Swayze 15 2004). Importantly, the continuum may have considerable slope in certain parts of the spectrum. If a steep continuum slope is in the same region as a more narrow absorption feature, the apparent position of this feature can appear to shift “downslope” in the reflectance spectrum. The analysis of reflectance spectra by the remote sensing community often includes “Continuum Removal”, which is defined by establishing the continuum across a spectral range and dividing the spectrum by the continuum (e.g., Clark and Roush 1984). This results in a “Continuum Removed Spectrum” and is sometimes referred to as the “Hull Quotient Spectrum”. The new spectrum effectively gets ‘pulled up’ to a normalized value of 1 where the continuum (or “hull”) touch the reflectance spectrum, which is typically at the inflection points of the continuum if using linear sections. This largely removes the effect of the now-modeled continuum on specific absorption features and has been shown to greatly aid in mineral identification. The modeling approach taken and the points chosen for defining the continuum can vary considerably (e.g., Gaussian modelling vs. simple linear sections), which can lead to variations in the new continuum removed spectra. For example, consider the reflectance spectrum of bastnaesite in the VNIR range (see Figure 2.8, spectrum B) and the absorption cluster near 864 nm and 889 nm. If only the 864 nm absorption is investigated a linear section between local maxima near 850 nm and 875 nm would be sloping to longer wavelengths. Conversely, if both the absorptions at 864 and 889 nm are investigated a linear section between local maxima near 850 nm and 925 nm would be sloping to shorter wavelengths. These two opposite-sloping continua would therefore shift the absorption feature at ~864 nm (in reflectance) in opposite directions. The implications of this extend beyond the investigation of a single cluster of absorptions in a single mineral (as described above) when comparing different mineral species with different morphologies of absorptions in similar wavelength ranges (e.g., Figure 4.16 at ~875 nm in cerite and mosandrite). Because this thesis presents an extensive body of new spectral observations, the reflectance spectra and the resulting absorption features tabulated are presented without removal of the continuum. In doing so, the decision on how to remove the continuum and where to define the continuum is avoided and researchers can examine the unmodified data though it is recognized that some feature positions listed here are influenced by a sloping continuum. Future work that exploits these features for various applications will, however, need to address the continuum and its impact on absorptions. 1.5.7 Mixtures, spectral unmixing and target detection Spectra obtained through reflectance spectroscopy can be largely assumed to be the result of a systematic combination of absorption features from multiple phases within a field of view (i.e., a spectral mixture). The type of mixture (linear vs. non-linear) is influenced by the spatial resolution of the observation. In a geological context, airborne imaging spectrometers (i.e., hyperspectral sensors) with ~4 m pixels investigate lithology and mineralogy of rocks (often with key diagnostic mineral features, such as carbonate-related absorptions) whereas laboratory 16 imaging spectrometers with ~1 mm pixels can more reliably uniquely investigate mineralogy of crystals. Furthermore, if the grain size of a given mineral is substantially larger than the spatial resolution of the imaging spectrometer, pure spectra of that target can comprise entire pixels thus bypassing “mixtures”. Typically in geological airborne or spaceborne remote sensing studies, spectral mixing is dominantly a linear process (e.g., an areal mixture) where objects in the field of view are optically distinct, occupying areas that encompass a substantial portion of the field of view, and the resulting spectrum can be derived from fractional abundances of each spectral endmember (e.g.,Clark 1999). For targets that contain materials which are not optically separated in the field of view (e.g., an intimate mixture) and photons are scattered by more than one material, as in the case of fine grained rocks with respect to the field of view, spectral mixing can become non-linear (e.g., Singer 1981). This phenomena can be particularly significant for materials of highly contrasting absorption coefficient (e.g. opaque phases set in silicates in the VNIR). Hapke (1993) notes that the parameters affecting reflectance spectra of intimate mixtures are non-linear, and therefore any spectral unmixing for these types of pixels should be treated non-linearly. As a general statement, spectral unmixing aims to discriminate and quantify end member populations contributing to final spectra of a scene (e.g., Mustard and Pieters 1989) whereas target detection that makes use of absorption features aims to quantitatively match unknown spectra within a scene to known spectra (Kruse et al. 1993). Both spectral unmixing and target detection rely on the use of a spectral library that contains ‘end-member’ spectra either externally derived (e.g., Clark et al. 2007) or derived from within the imaged scene (e.g., Plaza et al. 2004, Rogge et al. 2007, Zhang et al. 2008). Spectral unmixing and target detection using hyperspectral data is an established yet fast growing field of research with many recent contributions from computing science (e.g., Robles-Kelly and Huynh 2013, Heylen et al. 2014). Within the context of this thesis, spectral unmixing and target detection were used within a single-sample scene to isolate a discrete number of pixels that best represented a single spectral end-member target (i.e., a REE mineral phase). The products derived from spectral unmixing and target detection were used only to better understand the scene and to generate reliable Regions of Interest (ROI) used for calculating average spectra when sufficiently large single crystals were not available for simple polygonal ROI. A ‘reliable’ ROI comprised both the inclusion of compositionally pure “good pixels” and the exclusion of compositionally impure “bad pixels”. Accordingly, high spatial resolution imaging spectroscopy allowed the discrimination of spectra from specific mineral targets that would have otherwise been too small to obtain a reliable spectrum from a conventional spot spectrometer. Two methods were used to ascertain the locations of representative pixels for spectral averaging: Mixture Tuned Matched Filtering (spectral unmixing) and Spectral Angle Mapper (target detection). Both of these 17 methods are embedded within the commonly used software package ENVI (Environment for Visualizing Images). Broadly speaking, mixture tuned matched filtering (MTMF, Boardman 1998, Boardman and Kruse 2011) employs an initial minimum noise fraction transformation (MNF, Green et al. 1988) followed by matched filtering (MF) for each input spectrum and mixture tuning (MT) to assess for false positives for each abundance image (“infeasibility”). High abundance values (“MF Score”) combined with low infeasibility values for a particular spectrum implies a good fit to an input “end-member” reference spectrum (i.e., the REE mineral phase in question). The MF Score is roughly correlative to end member abundance, however, the method is not constrained to calculate an abundance of “100%”. In this work, only the purest pixels that represented monomineralic pixels were considered for inclusion in the ROI for averaging, and no use of the fractional abundances from unmixing were made. It is important to note that MTMF employs linear unmixing principles. In this study the field of view is very large compared to the wavelength of light used for measurements. In addition for most samples of this study, the grain size is medium to coarse and thus individual mineral grains occupy a substantial portion of a pixel. Consequently photons in a field of view interact principally with a single mineral phase before being scattered back to the detector. This implies that linear mixing is the dominant means to model a mixed spectrum. Simplistically, the Spectral Angle Mapper algorithm compares input spectra to reference spectra and calculates how well pairs of spectra match (Kruse et al. 1993). The input and reference spectra are projected as vectors into n-dimensional space (where n is equal to the number of input spectral bands) and the angle between the vectors are calculated (hence the name Spectral “Angle” Mapper). Spectra that are similar will have similar vectors and will therefore have a small angle. Accordingly, a small output value from the SAM algorithm indicates a good match between the input spectrum and reference spectrum. A commonly cited advantage of the SAM methodology in conventional imaging spectroscopy is that the amplitude of the vectors compared does not impact the angle measured, which allows better comparison of bright (high illumination) and dark (low illumination) pixel spectra. Both MTMF and SAM algorithms are heavily reliant on an appropriate reference spectral library and for both algorithms the input wavelengths can be restricted so as to obtain results based on particular spectral features. 1.6 Electronic processes and absorption bands 1.6.1 Crystal field theory Many of the absorption features seen in VNIR-SWIR reflectance spectroscopy of geological materials can be explained by crystal field effects of transition metal-bearing compounds (e.g., Burns 1993, Wildner et al. 2004, Rossman 2014). The most common transition metal cations with unfilled d orbitals are those of the first series and include Fe, Mn, V, Cr, Ni, Zn, Cu, and Co. Unfilled d orbitals of an isolated ion have equal energy levels and 18 probabilities of being occupied, and are thus considered degenerate in such a spherical (isometric) electric field. Once this ion with unfilled orbitals is placed into a crystal field, the energy levels of the orbitals become split into lower and higher states, balanced about the original barycenter (Figure 6). This allows for electron transitions to occur from lower levels to higher levels upon the incidence of light of an appropriate wavelength. The relative energy levels of split orbitals depend primarily on the symmetry and coordination geometry of the locally bonding ions as well as their identity (i.e., the cation-anion polyhedron). Pressure and temperature can also influence the splitting. An important simplification of crystal field theory is that it treats the electric charges of the ions as point locations dominated by ionic bonding. This is not entirely true in most minerals, however, the simplification does a fairly good job at mathematically approximating energy level splitting. It also facilitates the integration of X-ray diffraction crystal structures that report point locations of atoms. 1.6.2 Coordination geometry and crystal field splitting The coordination state of the transition metal cation plays the most important role in crystal field splitting, and therefore the absorption features expected in reflectance spectroscopy. Ligand coordination of the transition elements is dominated by an octahedral environment (6-fold) in most minerals, however, other important coordination states in minerals include tetrahedral (4-fold), cubic (8-fold) and dodecahedral (12-fold). Less common coordination states include triangular (3-fold), square planar (4-fold), trigonal bipyramid (5-fold), square pyramid (5-fold), trigonal prism (6-fold), pentagonal bipyramid (7-fold), and square antiprism (8-fold). Each of these distinct coordination geometries will result in distinct first order crystal field splitting parameters, and therefore generate variations in absorption feature patterns. The energy levels initially split into t2 (dxy, dyz, dzx) and e (dx2-y2, dz2) groups with three and two of the d orbitals, respectively, as defined by the interaction between the geometry of the orbital lobes and the geometry of the polyhedron. The splitting must be balanced against the original barycenter of the d orbitals, which is why the offset between the groups is not symmetrical (Figure 1.6). The energy level splitting of the field is denoted by Δ, the crystal field stabilization energy (CFSE), with a subscript describing the type of coordination (e.g., Δo for an octahedral field). Magnitudes of the CFSE for common transition metal valences in octahedral environments cluster in the 10,000 to 18,000 cm-1 range (1000 nm to 555 nm), in accordance with the strong visible colouration of their compounds. 19 Figure 1.6 Basic orbital energy splitting in various isometric crystal fields: (a)-cubic, (b)-dodecahedral, (c)-tetragonal, (d)-barycentre, (e)-octahedral (from Burns 1993). The results of crystal field splitting are fairly predictable for sites with ideal symmetry. In reality, the actual symmetry of cation sites in minerals rarely have high symmetry due to various cation substitutions, crystal defects and strain. This distortion of the cation polyhedra drives further change of the CFSE and subsplitting between energy levels of the t2g and eg terms, which correspond to different orbitals (Figure 1.7). This reconfiguration of the crystal field is explained by the Jahn-Teller effect, which addresses distortion of the cation site to decrease electronic instability. Magnitude of the subsplitting (δ) can be negligible in higher symmetry distortions, but can reach values similar to those of the CFSE in highly distorted environments. 20 Figure 1.7 Splitting of energy levels of the five d orbitals by cubic (e, f) and octahedral crystal fields (a – d). The Jahn Teller effect describes the subsplitting of the orbital groups by term δ, due to asymmetric distortion of a cubic environment (from Burns 1993). 1.6.3 Intervalence and oxygen‐metal charge transfers Intervalence charge transfer (IVCT) absorption features are different from crystal field spectra in that electrons are transferred between two metal cations instead of being excited between split 3d energy levels of a single cation (e.g., Nassau 1978, Burns 1993, Nassau 2001). Common examples of IVCT include homonuclear Fe2+-Fe3+ in vivianite and heteronuclear Fe2+-Ti4+ in corundum (var. sapphire). Cases of IVCT typically show strong pleochroism with maximum absorbance of the IVCT band along the axis between the two interacting cation sites. IVCT has been well documented for the transition metals but has also been noted in some lanthanide compounds (e.g., Dorenbos et al. 2010). Charge transfers can also occur between a cation and oxygen (OMCT). Absorptions related to this process typically occur at shorter wavelengths (ultraviolet range) but wings of the absorption bands can impact the shorter regions of the VNIR range. Examples include Fe-O charge transfers in limonite or goethite and Cr-O charge transfers in crocoite. 21 1.6.4 Band gaps and colour centres Important in the optical spectroscopy of sulphide minerals, band gap electronic transitions can occur between the conduction and valence bands of materials. The band gap refers to the difference in energy between the “top” of the conduction band and the “bottom” of the valence band. In this scenario, photons with energies above the energy gap defined by the band gap will be absorbed. This is sometimes referred to as the “absorption edge”. In the silicate minerals this band gap is fairly large and therefore generally exists in the ultraviolet region, and so has little to no impact on VNIR spectra. In the sulphide minerals, the energy required to cause transitions between the bands (i.e., the absorption edge) is much lower (closer conduction and valence bands) and can occur in the visible range (e.g., Wood and Sterns 1979, Boldish and White 1998). Sulphide minerals such as pure sphalerite, ZnS, have a moderate band gap energy near 350 nm (~3.54 eV) that leads to absorption of UV light and shorter wavelengths (i.e., it appears fully transparent to the eye). Proustite, Ag3AsS3, has a small band gap energy that occurs near 650 nm (~1.9 eV) which leads to its absorption edge being near the middle of the VNIR range and its “ruby silver” colouration. Galena, PbS, has an even smaller band gap that affects lower energy photons, absorbing all photons with wavelengths shorter than 3350 nm (~0.37 eV). Band gaps have been studied in lanthanide sulphide and oxide crystals and occur as low as ~1.3 (VNIR range) and ~4 eV (UV range), respectively (e.g., Jin et al. 2007, Gillen et al. 2013), but so far these types of compounds have not been noted in nature. Colour centres are the result of ionizing radiation (e.g., decay of uranium and emission of gamma rays) that can displace electrons in a crystalline matrix, leading to areas deficient in charge or with trapped electrons. When this occurs, the modified ions or trapped electrons have their own electronic state that can give rise to absorption bands that are typically found in the UV-VNIR range with full widths at half max (FWHM) greater than 100 nm. A common example is irradiated topaz where the blue colour is a result of an electron displaced from oxygen, leading to an arrangement of Al-O--Al that drives an absorption band centered at 620 nm (e.g., Da Silva et al. 2005, Krambrock et al. 2007, Rossman 2014). The geochemical association of lanthanides with actinides would suggest that many REE minerals would have received radiation exposure and potentially show colour centres. Fluorite is an example where the incorporation of REE and Mn plus incidence of radiation has driven specific colour centers, leading to a variety of colours including yellow, purple, blue and green (Bill and Calas 1978, Kempe et al. 2002). Similarly, investigations of naturally irradiated zircon by Kempe et al. (2010) suggest that electron holes and lattice defects can give rise to green, red and yellow colours, however, they also note the many discrepancies in the literature between assignments and colour. Klinger et al. (2012) build on the work by Kempe et al. (2010) and supply additional constraints on the configuration of electron holes with respect to substitutional Y and Tb. 22 1.7 Vibrational processes and absorption bands Vibrational processes in minerals and molecules generate absorption features. The wavelength position and intensity of these absorption features are primarily a function of bond lengths, bond angles and identity of the bonding elements. When resonance occurs between the frequency of incident light and the fundamental vibrational mode an absorption feature is observed. Fundamental vibrations in minerals occur within the infrared regions and are denoted by symbols v (stretching vibrations), δ (planar bending), and γ (out-of-plane bending) with additional identifier subscripts. Vibrational overtones are when a single fundamental vibration is summed or subtracted with itself or its derivatives, and vibrational combinations are when multiple fundamental vibrations are involved. Combinations and overtones tend to be of lower intensity than the primary fundamental vibrations. Infrared spectroscopy between 4000 cm-1 (2.5 μm) and 400 cm-1 (25 μm) probes this region of fundamental vibrations and is commonly used for structure interpretation, compound identification and concentration determination in controlled laboratory environments. Characteristic positions have been established for the vibrational features of minerals most relevant reflectance spectroscopy in the VNIR-SWIR range (OH, H2O, CO3, PO4, SO4) and the exact locations of these features are strongly affected by mineral chemistry and crystal structure (e.g., Salisbury et al. 1991). For example, the common constituent OH- of many silicates has a characteristic “O-H stretch” near 2.75 μm with overtones at 1.4 μm and 0.95 μm, however, as the vibration near 2.75 μm changes due to crystal structure the overtones and other combinations will also change (e.g., Clark et al. 1990, Libowitzky and Beran 2004). The identity of the metal bonding to OH- in various crystal structures has systematic impacts on the fundamental and overtone absorptions and this has prompted a number of useful studies that exploit these changes (e.g., Duke 1994, Bishop et al. 2008, Laukamp et al. 2012). The studies generally focus on Mg, Al, Fe, Ca and Mn since these are the most common cations driving systematic variations in common rock forming minerals (e.g., amphiboles, micas, clays). Bishop et al. (2008) comment that for the tri-octahedral phyllosilicate minerals, the nature of the cation (i.e., identity and valence state) determines how strongly it can pull charge away from the OH bond, thereby affecting the vibrational energy required for resonance (i.e., the location of absorption bands). Their order of strongest to weakest is Al3+ > Fe3+ > Fe2+ > Mg2+, and each of the cations produces its own sets of absorption features. Laukamp et al. (2012) noted that in actinolite and talc, higher Mg# correlates to shorter overall wavelengths for overlapping overtones and combinations. Simplistically, this results in a general trend of M-OH related features being at longer wavelengths for Al3+ and shorter wavelengths for Mg2+ with Fe2+,3+ showing intermediate wavelengths for a particular mineral or related minerals. In the di-octahedral “white micas” (e.g., paragonite, muscovite, phengite) it has been noted that replacement of Al3+ by various divalent cations leads to changes in M-OH bond lengths, which in turn causes the ~2200 nm M-OH related combination absorption to be 23 shifted to longer wavelengths as Al content decreases (Duke 1994). Thus, it is clear that compositional trends affect the location of these features but the overriding factor is the mineralogical identity and its particular crystal chemical controls. The particular example of M-OH vibrational features was used because it is easy to consider a comparable occupation by lanthanides and variations relating to cation identity, whether that is considering REE3+ substituting into an M-OH site or the differences between La3+-OH and Lu3+-OH. How this might be expressed in the SWIR remains an open question and is complicated by the numerous 4f-4f electronic transitions. In fact, Rowan et al. (1986) made note of unexplainable absorption bands near 2.35 μm in reflectance spectra from reagent grade Nd and Pr material, which is in agreement with similar unexplained features recorded by White (1967) near 2335 nm. 1.8 Spectroscopy of the lanthanides 1.8.1 Introduction The spectroscopy of REE and their compounds requires investigation of additional electronic phenomena that are not normally considered in conventional reflectance spectroscopy; namely, the understanding of intraconfigurational 4f-4f transitions. A great deal of spectroscopic data is available for rare earth element compounds in the fields of analytical chemistry and solid state physics, however, these data are largely focused on simple compounds, “free ions” or an aqueous environment. Furthermore, most data collected has been generated through transmittance studies or theoretical modeling (e.g., Dieke et al. 1968, Carnall et al. 1989, Peijzel et al. 2005). Data available from the materials engineering field are largely focused on modification and lanthanide-doping of structures to tune/shift the wavelengths of absorption and emission. Accordingly, the amount of data available for specific compounds that are “relatively simple” compared to naturally occurring mineral species is quite abundant. This wealth of transmittance-absorbance and emission data provides a robust dataset to which reflectance spectra can be compared against, but which cannot be relied on for direct comparisons. Furthermore, although there has been considerable advancement in the modeling of lanthanides subjected to crystal fields in various compounds using various approaches (e.g., superposition model, angular overlap model, ligand field theory), simulation results are still unsatisfactory in many regards for even the simpler lanthanides (Gorller-Warland and Binnemans 1996). The objective of reviewing the fundamentals of lanthanides and their electronic structure is not to take a theoretical approach and predict absorption features in minerals, but rather to enable the discrimination of relevant features observed in reflectance spectra. For minerals, the field of luminescence and cathodoluminescence spectroscopy has been taking advantage of the particular nature of intraconfigurational transitions of the lanthanides to probe their structural environment in natural crystal lattices (e.g., Nasdala et al. 2004, Lenz et al. 2013, MacRae et al. 2013). In particular, this has 24 proven useful on the “micro” scale for investigating potential nuclear waste forms and for better understanding mineral growth and dissolution systematics, often as applied to geochronology. The outcomes of these studies provide reference spectra for indirect comparison to reflectance spectra, and emission lines are typically acquired in the visible range with few collecting light into the NIR. In particular, Lenz et al. (2013) make note of the significant impacts that coordination geometry of the REE3+ has on resulting emission spectra. 1.8.2 Judd‐Ofelt theory The Judd-Ofelt theory (Judd 1962, Ofelt 1962) states that induced electric dipole (ED) intraconfigurational transitions are only observed if the point group of the site is not centrosymmetric. In this case, the odd part of the crystal field potential will be non-zero and spectral intensity (absorption) can be achieved by mixing configurations of opposite parity into the 4f wave functions. Otherwise, the transitions are “forbidden” and have negligible probability of occurring (i.e., no intensity). Accordingly, the greater the degree of asymmetry in a site, the greater the probability of a transition and therefore intensity. Intraconfigurational 4f–4f features are also generated by magnetic dipole (MD) transitions in both centrosymmetric and non-centrosymmetric local point groups but are less numerous than ED transitions (Gorller-Walrand and Binnemans 1998). 1.8.3 Intraconfigurational 4f‐4f transitions Despite the efficient shielding of the 4f open orbitals by the 5s and 5p closed orbitals, the energy levels defining intraconfigurational 4f absorption features are not static. Electrostatic repulsion of the base ion generates a first level splitting of spectroscopic states, 2S+1L (e.g., 5I). Next, spin-orbit coupling splits these into 2S+1LJ multiplets, or J-levels (e.g., 5I8). Once placed into a crystal field the J-levels are then split depending on the specific local point group, with more asymmetric point groups allowing more transitions (Gorller-Walrand and Binnemans 1998). These are known as the Stark (sub)levels, 2S+1 LJ (MJ). This sequential splitting of energy levels is depicted in Figure 1.8. For an odd number of electrons (e.g., Nd3+, 4f3) in a noncubic field the number of theoretically possible Stark sublevels (also known as Stark manifolds) is equal to J + ½ components. For an even number of electrons there are a maximum of 2J+1 components. The number, location and intensity of Stark sublevels are dependent on the symmetry of the ion within the crystal field. All considered, the positions and intensities of energy levels for a given lanthanide ion are governed by electrostatic (Coulomb field), spin-orbit and crystal field interactions (e.g., Binnemans 1996). An early compilation of absorption features for trivalent lanthanide ions in LaCl3 was completed by Dieke et al. (1968) and is commonly known as the “Dieke Diagram” (Figure 1.9). This diagram includes energy levels for both absorption (solid lines) and fluorescence (halfcircles under solid lines) corresponding to 2S+1LJ multiplet levels (e.g., 4I13/2 of Er3+). The thicknesses of lines on Dieke Diagram relate to the relative magnitude of splitting of the Stark sublevels. This publication and diagram have been widely used in the continuing study and 25 understanding of lanthanide spectroscopy. Later compilations of lanthanide related data include observed absorption lines in various crystal lattices in tables and for aqueous solutions in the form of band registries (e.g., Sastri et al. 2003, Peijzel et al. 2005) and typically used in absorption spectroscopy by analytical chemists. Splitting of the Stark sublevels can be substantial. For example, the 3H6 multiplet for Pr3+ in the high symmetry host LaCl3 shows a splitting magnitude of ~300 cm-1 with a barycenter near 4300 cm-1 (175 nm split near 2325 nm), and the 4I9/2 multiplet of Er3+ in the high symmetry host LaCl3 shows a ~160 cm-1 Stark splitting with a center near 12350 cm-1 (9 nm split near 810 nm). It is worthwhile to note that a 100 cm-1 splitting amplitude near 500 nm is ~3 nm, and near 2000 nm is ~ 250 nm; thus an equivalent splitting at smaller cm-1 values (SWIR range) will result in a greater spread than at larger cm-1 values (VNIR range). Figure 1.8 Nomenclature of the energy levels for the lanthanide ions using Ho3+ as an example (from Walsh 2006). Splitting into sequential levels is via (1) electron repulsion, (2) spin-orbit interactions and (3) crystal field. The last division represents a single multiplet (5I5, near 890 nm / 11235 cm-1) on the Dieke Diagram. Stark LevelsVo = crystal field(Electric field of host ) 4f 104f 95d5S5F5I5I45I55I85I65I7ConfigurationsH o = central field(Electrons in fieldof the nucleus)Terms 2S+1LHc = Coulomb field(Mutual repulsionof electrons) Levels2S+1LJHso = spin orbit(Coupling between sp in andorbital angular momentum)Ho > Hc , Hso > VoHc >>HSO (LS-coupling)Hc <<HSO (jj-coupling)Hc = HSO (Intermediate couplin g)No degeneracy: (2J+1) Stark levelsirreducible representat ions:term degeneracy:(2S+1)(2L+1)4+2N!(4+2-N)!configuration degeneracy:for (4f 10): l =3, N=10degeneracy = 1001level degeneracy:(2J+1)for (5I): S= 2, L= 6degeneracy = 65for (5I5) : J = 5degeneracy = 11S P D F G H I K L M0 1 2 3 4 5 6 7 8 9L =26 Figure 1.9 “Dieke Diagram” of Ln3+ intraconfigurational transitions with values in X1,000 cm-1. Lines represent free ion energy levels with electron repulsion and spin-orbit interactions but before effects of a crystal field (from Dieke et al. 1968). 27 1.9 Reflectance spectroscopy of REE bearing rocks and minerals 1.9.1 REE‐bearing minerals There are a limited number of studies pertaining to the reflectance spectroscopy of REE-bearing minerals and there have been very few spectroscopy studies that attend to more than one REE mineral at a time. Detailed spectroscopy reviews, including reflectance spectroscopy, for each of the mineral classes attended to in this thesis are given in their appropriate chapter, however, three articles stand out with respect to understanding the spectroscopy of REE-bearing minerals and rocks: Adams (1965), White (1967) and Rowan et al. (1986). Adams (1965) investigated the optical absorption spectra from 400 to 700 nm of several REE minerals using a spindle stage and a handheld portable microspectroscope. The spectra of monazite, bastnaesite, parisite, rhabdophane, xenotime, cenosite (name rejected in favour of kainosite) and gadolinite were recorded and compared against aqueous solutions of trivalent lanthanides. Roughly three groups of absorption patterns were evident and correlated to minerals that were LREE-enriched, HREE-enriched, and non-selective to the REE. Adams also noted (somewhat pessimistically) that the absorption lines were not fixed from mineral to mineral but slightly variable. He followed this with a suggestion that perhaps systematic changes could aid in mineral identification with further research, and that field spectroscopy could be useful for REE mineral exploration and especially so for heavy mineral sands. Although Adams only looked at a relatively narrow range of light his conclusions and remaining questions are both echoed and confirmed through the reflectance spectroscopy contained within this thesis. White (1967) studied the diffuse reflectance of Pr, Nd, Sm, Eu, Gd, Tb, Dy, Ho, Tm, Er, and Yb oxides from 225 to 2700 nm. Although not REE minerals, these crystalline reagent grade oxide powders established a baseline for comparison. The absorption bands recorded match the broad positions of 4f-4f transitions on the “Dieke Diagram”, as expected, but subtle differences were noted and ascribed to the effects of the crystal field. The mono-lanthanide reflectance spectra collected during this thesis work are in rough agreement with those by White (1967) but differences are present in some of the lanthanides, thus indicating different coordination environments. Rowan et al. (1986) were the first to publish on the reflectance spectroscopy of carbonatites and their associated lithologies by conducting laboratory-based measures of representative rocks from four carbonatite hosted rare metal deposits (Mountain Pass, Iron Hill, Gem Park and Oka). Objectives were to assess the applicability of reflectance spectroscopy for geological mapping and mineral exploration of carbonatite complexes. Their data showed the distinct spectral properties of reagent grade lanthanide oxides (Eu2O3, Nd2O3, Sm2O3 and Pr2O3) and REE mineral-bearing samples, leading to strong conclusions regarding the potential use of reflectance spectroscopy in REE exploration. They attributed the bulk of the spectral features to Nd3+ with lesser 28 input from Sm3+ and also commented on the possibility of band shifting for the 675 nm absorption feature between rocks from Mountain Pass and reagent grade Nd2O3. The compilation of spectra from the four different carbonatite complexes was the most comprehensive review to date of REE bearing rocks and minerals, and has served as the base for interpreting rock types at carbonatite and alkaline intrusive igneous complexes via remote sensing. The highly cited and web-accessible benchmark chapter on reflectance spectroscopy by Clark (1999) acknowledges that the lanthanides can drive absorption features. Reflectance spectra of reagent grade trivalent Eu, Nd, Sm, and Pr are presented, although there are no mineral spectra of REE-bearing samples. The main reason this reference is notable, however, is that it states “these [reflectance spectra of rare-earth oxides] are due to crystal-field transitions involving deep-lying electrons of the rare-earth element and do not shift when the rare-earth ion is in another mineral.” Indeed it is true that the influence of the crystal field on the deeper lying orbitals is less than those of the “exposed” 3d and 4d orbitals, however, the statement is false as outlined in the previous section (Spectroscopy of the Lanthanides). Unfortunately, this inaccurate statement by Clark (1999) is the most accessible current reference on the state of reflectance spectroscopy applied to REE minerals. Wavelength calibration standards for reflectance spectroscopy sometimes use lanthanide-doped compounds. Accordingly, well documented reflectance spectra are available for a few select standard reference materials. Weidner et al. (1986) developed NIST standards based on mixed rare earth oxides for use in the near infrared and chose Dy2O3, Er2O3 and Ho2O3 so as to have a range of absorptions across the NIR range without significant overlap. A band registry is supplied for the pressed powder, and similar registries are available for Ho-doped glass (Allen 2007) and Ho in perchloric acid (Weidner et al. 1985) for transmission spectroscopy. Labsphere LLC produces single and multiple lanthanide-doped SpectralonTM wavelength calibration pucks for reflectance using Ho, Er and Dy. These pucks are described as being quite inert and stable over time and are also supplied with a band registry for select strong absorptions. It is notable that in each of these standards the REE3+ are hosted in different coordinations and therefore band shifts are seen between the registries. For example, a band occurs at 420 nm in Ho:Spectralon, at 418.5 nm in Ho:glass, and at 416.1 nm in Ho:perchloric acid, whereas in Ho2O3 powder it occurs as a doublet with minima at 424 and 427 nm. Consequently, apart from White (1967) and the little reflectance data of minerals that has been published, the most transferrable information from the literature for understanding the reflectance spectra of REE bearing minerals originates mainly from optical (transmission) absorption spectroscopy of REE-doped synthetic material and secondarily from transmission studies of natural materials. Although the REE-doped synthetic compounds are somewhat useful for elucidating general locations of specific absorption features, the crystals tend not to be very analogous to structures found in the more common REE minerals. Detailed spectra are usually only given for the 29 visible range as phosphors and lasers are the prime areas of research. Transmission studies of REE bearing minerals are generally uncommon and typically focused on single minerals. In a similar manner, optical spectroscopy and crystal field theory reviews such as Burns (1993), Wildner et al. (2004) and Rossman (2014) acknowledge that 4f-4f intraconfigurational absorptions are the source of the sharp features in REE-bearing minerals but discussions in those reviews are almost exclusively restricted to the first series of d-block transition elements. 1.9.2 Remote sensing of REE related geological targets Remote sensing surveys of carbonatites and alkaline intrusive complexes in the literature can be divided into (1) coarse spectral resolution spaceborne studies (e.g., Oppenheimer 1998, Rowan 1998, Rowan and Mars 2003, Mars and Rowan 2011), (2) fine spectral resolution airborne studies (e.g., Crowley et al. 1988, Rowan et al. 1995, Bowers and Rowan 1996, Rowan et al. 1996, Bedini 2009), and (3) fine resolution ground and lab-based studies (e.g., Rowan et al. 1986, McHugh et al. 2000). The studies using ASTER imagery (14 spectral bands from VNIR to TIR, 15 to 90 m spatial resolution) were focused on lithological mapping. The general results are that lithologies of carbonatite complexes can be identified with moderate to strong confidence using ASTER satellite data from areas with established geological maps (e.g., Khanneshin, Mountain Pass, Ol Doinyo Lengai). Because of their coarse spectral resolution no direct REE3+-related features were detected, however, their host carbonatite lithology can be spectrally distinct from other carbonatite phases (e.g., Rowan and Mars 2003). Watson et al. (1996) successfully conducted lithological mapping for the Iron Hill area using airborne thermal multispectral data (6 bands from 8.4 μm to 11.5 μm) at 8 m spatial resolution but this is incapable of identifying REE minerals. Remote sensing using the finer resolution AVIRIS airborne system (~16-20 m spatial resolution) allowed detection of REE3+ absorption bands at the Mountain Pass carbonatite mine (Crowley et al. 1988, Green and Vane 1989, Rowan and Mars 2003). Similar airborne imaging at Iron Hill carbonatite, which had supporting hand samples showing detectable Nd3+ signals, did not detect these absorptions in the airborne data (e.g., Rowan et al. 1995). Similar results again were achieved by Bowers and Rowan (1996) at the Ice River Complex using AVIRIS data where Nd3+ absorptions were noted in laboratory spectra but not the airborne scene. Bedini (2009, 2011) investigated the REE-mineralized Sarfartoq carbonatite complex and the Kap Simpson felsic syenite complex of Greenland using the HyMap airborne hyperspectral imager (4 m spatial resolution) focusing on the classification of rock type and predictive mapping. Laboratory spectra of sovite carbonatite from Sarfartoq included a sample (SOV1) with Nd3+ related absorptions, however, no REE-bearing phases were detected in the airborne data. Most recently, McDowell and Kruse (2014) used the Mountain Pass area as a test area for integrating VNIR-SWIR (AVIRIS, 224 bands, 15.5 m pixel size) and LWIR (MASTER, 10 bands, 34.4 m pixel size) airborne imaging 30 spectroscopy with an aim to optimize remote predictive mapping. Several distinct REE-bearing spectral endmembers were extracted from the VNIR-SWIR scene. The work by McHugh et al. (2000) looked at slope stability of mine walls using hyperspectral ground based imagery with 47 bands from 480 to 940 nm, and one of their test sites was the Mountain Pass mine. They show a relatively clean reference spectrum of bastnaesite and pixels isolated as being REE bearing, although the published spectra from the scene are somewhat noisy with amplitudes on the same order as absorption features. Through their image interpretations they were successfully able to map REE ore on the mine wall using SAM and matched filtering for abundance estimation. Boesche et al. (2014) presented work on hyperspectral imaging of an outcrop with known REE mineralization at the well-known and easily accessible Fen Carbonatite Complex in Norway. Their tripod-based HySpex spectrometer collected imagery from 450 to 1000 nm, but neither spatial nor spectral resolution information was provided (spatial resolution was likely on the cm-scale and the number of spectral bands was probably 108 or 160, according to the HySpex website). Their abstract implies that multiple hand samples were taken, but it seems as though only a single thin section was investigated, from which monazite was determined to be the main REE mineral with confirmation by microprobe. Making use of the 800 nm Nd3+-related absorption they mapped the distribution of LREE mineralization across the outcrop, thus building on the work by McHugh et al. (2000). Although not imaging-based, Dai et al. (2013) used reflectance spectroscopy in the laboratory to quantitatively estimate REE content of leachates and contaminated streams in the Jiangxi Province of China, an active processing area of ion-absorbed clays. Based on absorptions at 443, 520, 574, 736, 790 and 861 nm they were able to develop linear regressions to detect total REE content down to ~75 ppm in a lab setting with R2 values of up to 0.97. Dai et al. (2013) did not officially recognize that in fact their estimates were primarily probing the Nd3+ content as per the noted absorption wavelengths (e.g., see Dieke Diagram or band registry of Sastri et al. 2003), but the high geochemical correlation between Nd and the Total Rare Earth Elements is strong enough to make that extension. A few questionable discrepancies exist within their publication, however, there is obvious promise for monitoring of REE content via remote sensing. Several researchers affiliated with CSIRO and the Geological Survey of Western Australia are also conducting ongoing research into the spectroscopy of REE minerals, however, most of their work is restricted to conference presentations and non-peer reviewed government publications (e.g., Morin-Ka 2012, Hancock et al. 2012). Similarly, work is being conducted at the USGS but publications thus far are limited to conference abstracts (e.g., Swazye et al. 2013, Hoefen et al. 2013). 31 1.10 Summary Six main topics were reviewed in order to place this thesis in a broader context. A general description of the lanthanides and rare earth elements were followed by a short review of the more common REE mineral classes and the geological settings that host REE deposits. Next, general principles of reflectance spectroscopy were reviewed, as were the general spectroscopic characteristics of the lanthanides. Finally, a broad summary of the available literature on the reflectance spectroscopy of REE minerals and remote sensing of REE-related geological settings was given with the caveat that more detailed descriptions on the spectroscopy of each mineral group are given in the appropriate Chapters. 32 Chapter 2. Visible to Short Wave Infrared Reflectance Spectroscopy of REE Fluorocarbonates 2.1 Chapter summary The mineralogy of rare earth element deposits is critical in understanding their petrogenesis and has significant implications for their economic viability. Lanthanide-bearing compounds are known to produce sharp absorption features in the visible to short wave infrared region (VNIR-SWIR), however, a significant knowledge gap exists between the fields of reflectance spectroscopy and rare earth element mineralogy. Reflectance spectra were collected from four bastnaesite samples, two parisite samples and one synchysite sample from the visible into the shortwave infrared. These REE fluorocarbonate mineral samples were characterized via scanning electron microscopy and electron probe microanalysis. Sharp absorptions of REE-bearing minerals are mostly the result of 4f-4f intraconfigurational electron transitions and for the light REE-enriched fluorocarbonates, the bulk of the features can be ascribed to Nd3+, Pr3+, Sm3+ and Eu3+. The lanthanide-related spectral responses of the REE fluorocarbonates are consistent across the group, supporting the notion that the REE cation site is very similar in each of these minerals. Carbonate-related spectral responses differed between these minerals, supporting the notion that the crystallographic sites for the carbonate radical differ between bastnaesite, synchysite and parisite. Exploitable spectral differences include a distinct absorption band at 2243 nm that separates bastnaesite from synchysite and parisite. Similarly, for bastnaesite a dominantly Pr3+ related absorption band located is at 1968 nm while in synchysite and parisite it occurs at 1961 nm. 2.2 Introduction The mineralogy of rare earth element deposits is critical in understanding their petrogenesis and has significant implications for their economic viability. Rapid determination of ore modal mineralogy for these deposit types by reflectance spectroscopy would provide immediate feedback on the strength, type and relevance of mineralization. Lanthanide-bearing compounds are known to produce sharp absorption features in the visible to short wave infrared region (VNIR-SWIR) and have been conventionally viewed as unchanging features in the field of remote sensing (e.g., Clark 1999); however, a significant knowledge gap exists between the fields of reflectance spectroscopy and rare earth element mineralogy. The lanthanides (Ln) are a series of 15 elements belonging to the 4f block of the periodic table, and from lightest to heaviest are lanthanum (La), cerium (Ce), praseodymium (Pr), neodymium (Nd), promethium (Pm), samarium (Sm), europium (Eu), gadolinium (Gd), terbium (Tb), dysprosium (Dy), holmium (Ho), erbium (Er), thulium (Tm), ytterbium (Yb) and lutetium (Lu). In geological environments the lanthanides are commonly found 33 together in their trivalent state, with the exception of divalent europium (Eu2+) and sometimes tetravalent cerium (Ce4+). The term “rare earth elements” (REE) comprises the lanthanides, commonly yttrium (Y) and sometimes scandium (Sc) due to similar chemical characteristics. Coordination states of the REE cations in the assorted mineral structures can be quite variable (e.g., Miyawaki and Nakai 1996), ranging from a reasonably symmetrical 8-coordinated site in xenotime, to britholite with two distinct sites with coordination numbers of 7 and 9, and the REE fluorocarbonates with multiple bonding anions (oxygen and fluorine) with a coordination number of 9 (6O+3F). The mineralogy of REE ore deposits is important because the metallurgy of only certain REE-bearing phases is well established. Namely, most global REE production is derived from REE fluorocarbonate minerals (e.g., bastnaesite, parisite and synchysite), REE phosphate minerals (e.g., monazite and xenotime) and from ion-absorbed clays, however, recent advances have been made in the treatment of REE silicate mineral ore (e.g., eudialyte and britholite) (e.g., Mariano and Mariano 2012). Furthermore, REE deposits can show a complex mineral paragenesis with REE-mineral assemblages varying significantly across a single geological system. Consequently, there exists significant potential for reflectance spectroscopy to play an important role in the exploration and exploitation of REE ore deposits. The objectives of this study are to present mineralogical characterization and reflectance spectra in the visible to short wave infrared region of the main rare earth element fluorocarbonate minerals; bastnaesite, parisite and synchysite. Following mineralogical and spectroscopic background of REE bearing minerals we provide a band registry for these minerals as well as interpretations of spectral absorption features related to the lanthanides. The spectral features of the registry will be the focus of further study in the development of hyperspectral reflectance imaging spectroscopy to carry out REE mineral identification, REE mineral abundance estimates and rare earth element abundance estimates. 2.3 General spectroscopy of the lanthanides The outer radius of the 4f electron shells (~0.3 Å) for the lanthanides is much less than that of their filled 5s and 5p shells (~2 Å, ~1 Å). It can then be approximated that the local electronic environment of Ln3+ cations interacts primarily with those outer shells, leaving the 4f electrons ‘relatively sheltered’ but not completely non-participatory in bonding (e.g., Liu, 2005). Electrostatic repulsion of the base ion generates a first level splitting of spectroscopic states, 2S+1L (e.g., 5I). Next, spin-orbit coupling splits these into multiplets, or “J-levels” (e.g., 5I8), and once placed into a crystal field the J-levels are then split into “Stark Sublevels”. Crystal field interactions for the Ln cation include variables such as ligand type, coordination number and polyhedron asymmetry which all play a role in the location and intensity of energy levels and the associated absorption (Görller-Walrand and Binnemans, 1998). Each of the resulting sublevels provides the potential for promotion of a relaxed electron into 34 an excited state, giving rise to absorption of electromagnetic radiation (e.g., light) at a specific energy level (e.g., wavelength). The spectroscopy of REE-bearing phases is well established in the fields of physics and chemistry, however, the well understood principles and well studied doped compounds do not lend to direct translation into mineralogy and hyperspectral remote sensing. For example, the Dieke Diagram (Dieke et al. 1968) details intraconfigurational 4f-4f transitions for ‘free ions’ as deduced through studies of largely mono-lanthanide synthetic compounds, however, the transitions shown do not include splitting of energy levels due to a crystal field nor the complexities of naturally occurring REE minerals with variable REE distributions and other elemental substitutions. Consequently, this diagram and other band registries can only act as proxies to help identify origins of absorption features in reflectance spectra from minerals. In general, and excluding other physical parameters such as grain size, the strength of absorption features by the lanthanides will be primarily a function of the concentration of the ion as well as the specific absorptivity of that ion’s intraconfigurational transitions within a given crystal structure. The location of lanthanide-related absorption features will be primarily a function of the cation’s specific coordination and asymmetry in the host crystal structure. 2.4 Review of REE‐related reflectance spectroscopy studies Few studies have been published addressing rare earth element bearing minerals and rocks in the field of reflectance spectroscopy. Surveys of carbonatites and alkaline intrusive complexes in the literature can be divided into coarse spectral resolution spaceborne studies (e.g., Oppenheimer 1998, Rowan and Mars 2003, Mars and Rowan 2011), fine spectral resolution airborne studies (e.g., Crowley et al. 1988, Rowan et al. 1995, Bowers and Rowan 1996, Bedini 2009), and fine resolution ground and lab-based studies (e.g., Rowan et al. 1986, McHugh et al. 2000). The early work by Rowan, Crowley and Mars recognized the potential of exploiting REE absorption features with hyperspectral imaging, however, focus of their future research shifted to other geological systems, use of ASTER multispectral satellite imagery, and planetary remote sensing. The studies with coarse resolution used ASTER imagery and were focused on lithological mapping via endmember extraction and band ratios. The general results are that lithologies of carbonatite complexes can be identified with moderate confidence using ASTER satellite data from areas with well-established geological maps (e.g., Khanneshin, Mountain Pass, Ol Doinyo Lengai). Remote sensing using the finer resolution AVIRIS airborne system shows improved classification of rock types and even interpreted detection of REE from Mountain Pass (e.g., Crowley et al., 1988). Bedini (2009) investigated the REE-mineralized Sarfartoq carbonatite complex of Greenland using the HyMap airborne hyperspectral imager focusing on the classification of rock type and predictive mapping, however, there was no mention of any specific REE phases being detected from the 35 airborne platform. Similar results were achieved by Bowers and Rowan (1996) at the Ice River Complex using AVIRIS data. Ground and laboratory based studies by Rowan et al. (1986) of four carbonatite hosted rare metal deposits focused on rock type classification and were able to discriminate between several carbonatite phases, as well as identify the sharp absorption features caused by Nd3+ in field samples. This earliest work is referenced in all of the above literature and has served as a base for the spectral classification of rock types at carbonatite and alkaline intrusive igneous complexes. Early infrared spectroscopy investigations of carbonate minerals by Adler and Kerr (1963) included bastnaesite and parisite because of their distinct carbonate arrangements, but did not investigate the influence of the lanthanides and stopped short of the shortwave infrared region. The benchmark mineral spectroscopy paper of Hunt (1977) lists monazite (misspelled as monzonite) in his tabulation of common minerals and respective spectral signatures, however, La2+ is listed as the origin of absorptions but La would have 3+ valence in monazite and would therefore not have spectral features in the VNIR-SWIR (e.g., Liu 2005). Clark (1999) covers several REE oxides in his Spectroscopy of Rocks and Minerals review and recognizes that the patterns seen in REE minerals are a combination of several lanthanides but states that absorptions are independent of mineralogy. Reflectance spectra in the VNIR-SWIR of bastnaesite and parisite with limited discussions have been presented by Kerr et al. (2011) and Morin-Ka (2012), with both publications stating that REE signatures are present in these REE minerals and that more detailed documentation is warranted. 2.5 Crystal structure reviews of bastnaesite, parisite and synchysite Bastnaesite, parisite and synchysite are rare earth element fluorocarbonate minerals of economic significance in many REE deposits and occurrences. Bastnaesite, CeCO3F, is the most commonly reported of these minerals, however, it is also the most familiar of the group. Synchysite, CaCe(CO3)2F, is the most Ca-enriched of the REE fluorocarbonates, sometimes forms euhedral prismatic crystals and is the second most reported. Parisite, CaCe2(CO3)3F2, can form distinct doubly terminated pyramidal crystals and is the least reported of these rare earth fluorocarbonate minerals. Rontgenite, Ca2Ce3(CO3)5F3, is another rare REE fluorocarbonate mineral with a similar structure but was not addressed in this study. Total rare earth oxide (REE2O3) content increases from ~52 wt.% in Ca-rich synchysite, to ~60 wt.% in parisite and up to ~75 wt.% in Ca-absent bastnaesite. These minerals show preference for the light rare earth elements (LREE) and have similar slopes and trends in Chondrite-normalized diagrams. These three minerals are commonly found together, are structurally related, can show syntaxial intergrowth and are also commonly found together in secondary mineral mixtures. Structurally, each of the phases can be assembled using a common set of building blocks (Ni et al. 1993), which include layers of CeF, CO3, and Ca. On an atom basis, bastnaesite is Ca-absent, parisite shows REE:Ca=2:1 36 and synchysite shows REE:Ca=1:1. Accordingly, the stacking order of the blocks gives rise to the different compositional proportions and mineral species (Figure 2.1). Polytypism has been documented for these minerals (e.g., Meng et al. 2001). It is important to note that all of the REE will occur in the Ce site but only Ce will be listed for brevity. Table 2.1 documents some basic crystallographic data for the REE fluorocarbonates. Figure 2.1 (a) Coordination polyhedron for the Ce1 site in parisite (Ni et al. 2000); also applicable to bastnaesite and synchysite. Ce1 (green) atom is coordinated with F1, F2 and F3 (lavender) atoms, whose plane is roughly perpendicular to c-axis, and 6 oxygen (red) atoms O11, O23, O32, O42, O53 and O61. Overall coordination number of 9 in a distorted tricapped trigonal prismatic arrangement. (b) Parisite crystal structure from Ni et al. (2000), (c) bastnaesite crystal structure from Ni et al. (1993), and (d) synchysite crystal structure from Wang et al. (1994). Atom colouring: red=oxygen, green=REE, lavender=F, brown=C. Polyhedra colouring: green=REEO6F3, dark blue=CaO8, brown=CO3. 37 Table 2.1 Some basic properties of the REE fluorocarbonates. Mineral bastnaesite parisite synchysite Reference Ni et al. (1993) Ni et al. (2000) Wang et al. (1994) Formula REECO3F CaREE2(CO3)3F2 CaREE(CO3)2F Symmetry Hexagonal Monoclinic Monoclinic Space group P62c Cc C2/c a (Å) 7.1175 12.305 12.329 b (Å) 7.1053 7.110 c (Å) 9.7619 28.25 18.741 Β (°) 98.257 102.68 Structure REE coordination 9: 3×F, 6×O 9: 3×F, 6×O 9: 3×F, 6×O REE site shape tricapped trigonal tricapped trigonal tricapped trigonal Number of unique REE cation sites 1 6 2 REE-O (Min, Å) 2.542 2.466 2.497 REE-O (Mean, Å) 2.571 2.564 2.558 REE-O (Max, Å) 2.591 2.631 2.619 REE-F (Min, Å) 2.403 2.370 2.376 REE-F (Mean, Å) 2.407 2.397 2.399 REE-F (Max, Å) 2.416 2.445 2.417 CO3 influences Each CO3 layer influenced by REE only 3 distinct CO3 layers, one influenced by REE only, two influenced by REE and Ca Each CO3 layer influenced by Ca and REE Number of unique CO3 polyhedra 1 9 3 Average C-O bond lengths (Å) CO3 #1 1.287 1.2478 1.2817 CO3 #2 1.2728 1.2854 CO3 #3 1.2718 1.2752 CO3 #4 1.2636 a CO3 #5 1.2646 a CO3 #6 1.2736 a CO3 #7 1.3004 CO3 #8 1.2887 CO3 #9 1.2945 Average CO3 polyhedra volume (Å3) 0.0017 0.0223 0.0146 Average CO3 polyhedra distortion index 0.0094 0.0282 0.0159 General Chemistry REE2O3 wt% 73 59 51 CaO wt% 0 10.5 17 F wt% 5 4.5 2.6 CO3 wt% 20 24 28 a These CO3 polyhedra in parisite are coordinated only to REE 38 In bastnaesite Ni et al. (1993) described the hexagonal crystal structure as being built by layers of REE-F alternating with layers of CO3 in (0001) arrangement. There is one REE site, which is coordinated with 3 in-plane fluorine (F) atoms and two sets of 3 oxygen (O) atoms from bordering CO3 layers. Bond lengths of REE with O are reported at 2.591, 2.542 and 2.579 Å while lengths with F are shorter and between 2.403 and 2.416 Å. The resulting 9-coordinated tricapped trigonal prismatic REE site is therefore asymmetrical with two distinct sets of bonding ligands and lengths. Ni et al. (2000) studied the crystal structure of parisite and determined that, unlike bastnaesite, it is monoclinic. The structure was described as being built of two portions of bastnaesite layers connected by a Ca layer, stacked along the c-axis (i.e., CeF-CO3, CeF-CO3, Ca-CO3). Coordination for REE is similar to bastnaesite (3 × Ce-F, 6 × Ce-O) with Ce-O mean lengths between 2.55 and 2.57 Å and Ce-F means between 2.38 and 2.41 Å. Minimum and maximum Ce-O bond lengths are 2.47 and 2.63 Å, while for Ce-F they are 2.37 and 2.45 Å. The resulting 9-coordinated tricapped trigonal prismatic REE site is therefore asymmetrical with two distinct sets of bonding ligands and lengths. The unit cell for parisite contains three carbonate radical layers, two of which are influenced by both the REE-F and Ca layers and one of which is influenced only by REE-F layers. The crystal structure of synchysite, CaCe(CO3)2F, was investigated by Liben et al. (1994) and expanded upon by Ni et al. (2000). Synchysite is also monoclinic and is built by stacking layers of CeF-CO3 and Ca-CO3 in a one-to-one ratio along (001), making it the most Ca-rich member of the REE fluorocarbonate group. Coordination for REE is similar to bastnaesite and parisite (3 × Ce-F, 6 × Ce-O) with Ce-O mean lengths between 2.557 and 2.567 Å and Ce-F means between 2.394 and 2.407 Å. Minimum and maximum Ce-O bond lengths are 2.50 and 2.62 Å, while for Ce-F they are 2.38 and 2.41 Å. The resulting 9-coordinated tricapped trigonal prismatic REE site is therefore asymmetrical with two distinct sets of bonding ligands and lengths. In the case of synchysite, the carbonate layer is now always influenced by bonding with both Ca and REE-F layers. Each of the REE fluorocarbonate minerals hosts REE in 9 fold tricapped trigonal prismatic coordination with 6 oxygen atoms at the apices of the prism and 3 fluorine atoms in planar configuration through the faces of the prism (Figure 2.1). Minor differences in the bond lengths of REE-F and REE-O amongst the REE fluorocarbonates are reported in the literature, however, a significant difference between each of the minerals is the local environment surrounding the carbonate radical. In bastnaesite, the oxygen apices of the CO3 radical interact only with REE cations both below and above the CO3 plane. In parisite, two of the three repeating CO3 layers interact with REE and Ca above and below the plane while the third repeating CO3 layer interacts only with REE. For synchysite, the apices of the CO3 radical always interact with both REE and Ca on either side of the plane. Consequently, bastnaesite and synchysite show only one configuration for cation bonding with the CO3 39 radical whereas parisite shows two configurations that are unevenly populated. Furthermore, the geometry of the CO3 polyhedra is much more variable for parisite than for synchysite or bastnaesite (Figure 2.2, Table 2.1). Figure 2.2 Bond characteristics for the CO3 radical polyhedra in REE fluorocarbonate. Data from Table 1 and references therein, Min=minimum, Mid=middle, Max=maximum. The cation sites for REE in these minerals are very similar and one would therefore expect that the spectral features due to intraconfigurational electron transitions of the REE would also be very similar. It would then be expected that variations between spectra of the same mineral would be predominantly in the relative strengths of REE-specific absorption features while variations of REE-related spectral features between the different minerals might include slight shifts in the locations of REE-related absorption features as well as the strength of some of these features. The coordination environment of the carbonate radical is quite different for each of these minerals, relatively speaking, and one would expect to see a difference in the carbonate related absorption features between the minerals. Other variables that may influence spectra include the signal to noise ratio of the spectral data, crystallographic orientation, variations in other trace elements, and possibly syntaxial intergrowths with other REE fluorocarbonates. 40 The infrared spectra of bastnaesite and parisite were presented in Adler and Kerr (1963). Bastnaesite is characterized by carbonate related absorptions common to what is seen in calcite and dolomite, however, the strongest features for bastnaesite are the ν2(CO3) and ν3(CO3) modes at 11.52 μm (868 cm-1) and 6.93 μm (1443 cm-1), respectively, and are asymmetric. The ν2(CO3) and ν3(CO3) modes in parisite are the most prominent and are located at 11.49 μm (870 cm-1) and 6.90 μm (1449 cm-1), respectively, and are asymmetric. Both the ν2(CO3) and ν3(CO3) absorption features are broader for parisite than for bastnaesite. For parisite Adler and Kerr (1963) also made note of the multiple non-equivalent carbonate radicals and observed doublets for the ν1(CO3) (9.19 and 9.27 μm, or 1088 and 1079 cm-1) and ν4(CO3) (13.40 and 13.62 μm, or 746 and 734 cm-1) vibrational modes while the ν1(CO3) and ν4(CO3) modes for bastnaesite occur at 9.21 μm and 13.74 μm (1086 and 728 cm-1). Raman and infrared spectroscopy studies by Frost and Dickfos (2007) and Yang et al. (2008) revealed that OH stretching bands are present in most REE-fluorocarbonate samples studied, although the bands recorded are variable in number and position. 2.6 Experimental methods 2.6.1 Samples Three bastnaesite samples were borrowed from the Canadian Museum of Nature’s Mineral Collection (CMNMC) and were labeled as originating from Burundi (#39382), the Karonga Mine (Congo, #56255) and Madagascar (#50588). These samples were all single crystal fragments, honey brown in colour, and approximately 1 cm3. A fourth bastnaesite crystal sample from the Diao Lou Shan area in Sichuan (China) was obtained from A. Mariano, was light brown in colour and measured 3 × 3 × 2 cm. Two sample sets of parisite crystals from Snowbird (Montana, USA) and Muzo (Colombia) were obtained from A. Mariano. The Snowbird set comprised four euhedral elongate tapering prismatic crystal fragments, each approximately 1 × 1 × 2.5 cm and light grey-brown in colour. The Muzo set comprised two euhedral tapering prismatic crystal fragments, each approximately 0.5 × 0.5 × 0.5 cm and golden brown with high translucency. Scanning electron microscopy (SEM) investigations showed that all the bastnaesite and parisite samples were not compositionally zoned, did not show syntaxy, and did not show any other mineral inclusions. Three samples of synchysite were studied, originating from Narsarsuk (South-west Greenland, #UBC-3376), the White Cloud Mine (Colorado, USA) and Morris County (New Jersey, USA). Both the White Cloud Mine (from CMNMC, #37320) and Morris County (from CMNMC, #37321) samples were fine grained and SEM investigations showed that the complex mineralogy was not suitable for baseline spectroscopic characterization studies. The Narsarsuk hand sample (from UBC Museum Collection) included a cluster of ~25 light brown elongate prismatic euhedral synchysite grains up to ~1 mm long, hosted on a feldspar-biotite dominated matrix. SEM examination and electron microprobe analyses (EMPA) data also revealed very minor amounts of ancylite and an unknown REE-Ca-Sr-Ba phase. Reagent-grade lanthanide 41 oxide powders and REE-doped Spectralon wavelength calibration samples were also investigated in order to aid in band assignment. 2.6.2 Reflectance spectroscopy Reflectance spectroscopy was primarily carried out using the sisuROCK instrumentation (manufactured by SPECIM Spectral Imaging Ltd.) at the University of Alberta’s CoreSensing Facility, and data was handled using ENVI 4.4, a widely used and commercially available software package. Two imaging spectrometers (“cameras”) acquired reflectance spectra in the visible-near infrared (VNIR, 396 nm to 1003 nm over 784 channels for an average spectral resolution of 0.77 nm) and shortwave infrared (SWIR, 928 nm to 2530 nm over 256 channels for an average spectral resolution of 6.26 nm) portions of the electromagnetic spectrum in high spatial resolution mode. Spatial resolution of the cameras in this mode was approximately 0.079 mm / pixel in the VNIR and 0.241 mm / pixel in the SWIR. Noise was very prevalent in the shortest wavelength portion of the VNIR camera below ~550 nm and moderate from 550 nm to ~650 nm. In the high spatial resolution mode, averaging ~16 pixels resulted in reliable spectra in the noisier ranges that would be useable in spectral libraries. Spectra presented originate from single crystals for parisite and bastnaesite. The synchysite spectrum is an aggregate of multiple single crystals. The euhedral crystals of bastnaesite from Sichuan, parisite from Muzo, and parisite from Snowbird were also large enough to permit imaging both parallel and perpendicular to the grains’ c-axis. Spectra documented here are nominally an average of 2004 pixels for the VNIR camera and 1029 pixels for the SWIR camera. Samples were placed on a matte black surface that translates the samples under the camera and has very low reflectance across the sampled wavelength range. Some samples were propped up with foam blocks to ensure surfaces of interest faced the spectrometers. All samples were also substantially thick enough to assume the reflectance spectra are representative of the mineral target. A TerraSpec Pro point spectrometer manufactured by Analytical Spectral Devices Inc. (ASD) was used in the earliest studies and records 2151 channels from 350 to 2500 nm, for a spectral resolution of ~1 nm. Spot size of the point spectrometer is approximately 1 cm in diameter, which significantly restricted the use of this instrument. Methodology of acquiring spectra followed the manufacturer’s recommendation of ~30 second acquisition time with periodic darkfield and white reference Spectralon panel normalization and wavelength calibrations using Ho, Er and Dy doped Spectralon samples (manufactured by Labsphere). Though the ASD instrument provided limited spectra (e.g., for bastnaesite from Mountain Pass), they were valuable for confidence in band assignment below 600 nm because of lower noise levels than using the sisuROCK system. Averaged spectra from the imaging spectrometer were preferred over spectra from the spot spectrometer because some of the samples investigated had grain sizes smaller than were confidently resolvable by the point spectrometer. This allowed for non-destructive testing of mineral specimens and the ability to exclude spectral 42 effects from other minerals that would have otherwise been in the field of view of the spot spectrometer. For larger single crystal samples, a simple Region Of Interest (ROI) was used to select the target mineral’s pixels for averaging. For finer grained samples, a priori knowledge about the sample allowed several baseline spectra to be isolated from single pixels. These spectra were then used to run mixture tuned matched filtering within the ENVI software package on the entire scene, from which a strict qualitative threshold allowed a discrete selection of macroscopic unmixed pixels (i.e., only pixels with abundance estimates nearing 100%) to be averaged (see Figure 2.3). Reflectance spectra did not have the continuum removed so as to present the data unmodified and to facilitate comparison against other earlier publications. 43 Figure 2.3 (a) Hyperspectral imagery of the synchysite-bearing sample from Narsarsuk in the short wave infrared (SWIR). (b) Mixture tuned matched filtering (MTMF) MF abundance results for synchysite in SWIR scene and (c) regions of interest (ROI) derived from thresholded MTMF results for generating an average spectrum. (d) Annotated digital photograph of the same sample from nominally the same perspective. Arrows in (c) and (d) point to the same patches of synchysite. Monomineralic patches of synchysite are on the order of 2 mm by 2 mm, large enough for successful imaging with a spatial resolution of 0.241 mm by 0.241 mm, to enable isolation of pure monomineralic pixels with confidence for averaging of their spectra. 44 2.6.3 Scanning electron microscopy and electron microprobe analysis The Philips XL30 scanning electron microscope (SEM) at the University of British Columbia, which is equipped with an energy-dispersion X-ray spectrometer (EDS), was used for preliminary examination of mineral mounts of selected minerals and rock fragments studied by reflectance spectroscopy. Selected samples were then analyzed by electron microprobe at the Saskatchewan Research Council’s Advanced Microanalysis Centre using a Cameca SX-100 equipped with 5 tunable wavelength dispersive spectrometers. Operating conditions were: 40° takeoff angle, beam energy of 15 keV, beam current of 20 nA, beam diameter of 5 μm. The MAN background intensity data was calibrated and continuum absorption corrected. Elements were acquired using analyzing crystals LLIF for FeKα, TaLα, PrLα, EuLα, DyLα, TmLα, MnKα, LaLα, NdLα, GdLα, HoLα, YbLα, BaLα, CeLα, SmLα, TbLα, ErLα, LuLα, PET for CaKα, KKα, ClKα, TiKα, NbLα, YLα, SrLα, ZrLα, PKα, UMα, ThMα, and LTAP for MgKα, FKα, NaKα, SiKα, AlKα. Counting times were 10 seconds for Zr and P and 15 seconds for all other elements, with off peak count times of 10 seconds. The standards (with elements) were SPI-Barite (Ba), SPI-Celestite (Sr), SPI-YAG (Y, Al), Smithsonian Cr-augite (Mg, Ca), Smithsonian Ilmenite (Fe, Ti), Smithsonian Apatite (F, P), Smithsonian Microcline (K), Smithsonian Zircon (Zr), Harvard Albite (Si, Na), Cameca Mn (Mn), SPI2-TlCl (Cl), SPI2-Nb (Nb), SPI2-La (La), SPI2-Ce (Ce), SPI2-Pr (Pr), SPI2-Nd (Nd), SPI2-Sm (Sm), SPI2-Eu (Eu), SPI2-Gd (Gd), SPI2-Tb (Tb), SPI2-Dy (Dy), SPI2-Ho (Ho), SPI2-Er (Er), SPI2-Tm (Tm), SPI2-Yb (Yb), SPI2-Lu (Lu), SPI2-Ta (Ta), SPI2-Th (Th), and SPI2-U (U). Amounts of CO2 (*, as CO3) and OH- (**, as H2O) were determined by stoichiometry based on 4, 11 and 7 anions for bastnaesite, parisite and synchysite, and for full occupation of the F atomic site by F-, Cl- and OH-. Ta, Ti, Mn, K and Ba were not detected and are not included on Table 2.2. 2.7 Results Samples described here were selected from a larger set and were characterized by imaging reflectance spectroscopy, scanning electron microscopy and microprobe analysis. Several samples have spectra collected from a point spectrometer. Table 2.2 documents electron microprobe results for the REE fluorocarbonates in this study and Figure 2.4 shows Chondrite-normalized patterns of the samples. Tables 2.3 and 2.4 document prominent absorption features for the REE fluorocarbonates and includes probable origins of the features, as chosen through comparison with reflectance spectra from unpublished reagent grade lanthanide oxide spectra, REE-doped calibration standard spectra, and other REE-bearing mineral spectra for which compositional data exists, as well as REE spectroscopy literature. The position and shape of these features were recorded using reflectance spectra (i.e., not continuum removed spectra). Tables 2.5 and 2.6 compare significant chemical and spectral differences amongst the bastnaeiste and parisite sample sets, respectively, and include the number of pixels averaged from the sisuROCK imaging scenes by the VNIR and SWIR cameras. 45 Table 2.2 Average EMPA compositions of REE fluorocarbonate samples in wt.%. Mineral Sample ID bastnäsite Sichuan bastnäsite 39382 bastnäsite 50588 bastnäsite 56255 parisite Muzo parisite Snowbird synchysite UBC3376 Locality Diao Lou Shan, China Burundi Madagascar Karonga, Congo Colombia Montana, USA Narsarsuk, Greenland n 10 2σ 5 2σ 5 2σ 5 2σ 6 2σ 5 2σ 3 2σ Nb2O5 0.01 0.02 0.00 0.00 0.01 0.03 0.00 0.00 0.00 0.02 0.00 0.01 0.00 0.00 P2O5 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.05 0.17 SiO2 0.05 0.01 0.11 0.02 0.06 0.00 0.11 0.01 0.06 0.01 0.08 0.08 0.38 0.52 ZrO2 0.00 0.00 0.01 0.03 0.00 0.00 0.01 0.03 0.00 0.00 0.00 0.00 0.01 0.03 UO2 0.53 0.10 0.00 0.00 0.00 0.00 0.00 0.00 0.39 0.13 0.34 0.19 0.00 0.00 ThO2 0.14 0.13 0.00 0.00 0.37 0.48 0.00 0.00 0.83 0.64 1.62 0.26 0.54 0.12 Al2O3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.04 0.15 La2O3 27.77 1.88 19.44 1.00 25.22 0.64 20.76 0.75 14.03 0.53 12.87 0.37 12.17 1.33 Ce2O3 31.14 0.67 32.58 0.31 32.95 0.11 32.47 0.48 26.11 0.90 24.77 0.77 22.29 0.99 Pr2O3 2.50 0.32 3.07 0.33 2.38 0.19 2.98 0.15 2.90 0.19 2.71 0.19 2.17 0.29 Nd2O3 7.25 1.18 11.05 1.07 6.62 0.15 10.28 0.20 11.90 0.59 12.14 0.17 8.25 0.57 Sm2O3 0.47 0.18 1.18 0.14 0.29 0.07 1.02 0.01 1.85 0.14 2.64 0.23 1.42 0.05 Eu2O3 0.00 0.02 0.16 0.08 0.02 0.05 0.11 0.05 0.03 0.07 0.10 0.10 0.17 0.01 Gd2O3 0.20 0.18 0.53 0.14 0.08 0.10 0.40 0.10 1.06 0.19 1.70 0.25 1.03 0.08 Tb2O3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.03 0.08 0.04 0.02 Dy2O3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.01 0.35 0.16 0.38 0.08 Ho2O3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.001 0.00 Er2O3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Tm2O3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Yb2O3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Lu2O3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Y2O3 0.01 0.03 0.17 0.12 0.01 0.01 0.11 0.06 0.74 0.32 1.69 0.13 1.47 0.53 SrO 0.01 0.04 0.02 0.06 0.08 0.13 0.02 0.05 0.00 0.00 0.00 0.00 0.22 0.36 MgO 0.00 0.00 0.02 0.02 0.01 0.02 0.01 0.02 0.00 0.00 0.00 0.00 0.00 0.00 FeO 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.17 0.43 CaO 0.11 0.10 0.00 0.00 0.05 0.02 0.00 0.00 10.48 1.06 10.23 0.16 16.49 2.69 Na2O 0.03 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.01 0.02 0.01 0.01 0.10 0.04 Cl 0.01 0.02 0.08 0.02 0.06 0.02 0.08 0.04 0.02 0.01 0.02 0.01 0.12 0.05 F 5.22 0.42 4.42 0.22 4.56 0.15 4.27 0.23 4.48 0.20 4.12 0.57 2.87 1.04 CO2* 18.86 18.36 18.32 18.35 24.2 24.46 27.26 H2O** 1.38 1.64 1.57 1.17 1.17 1.38 1.4 -O≡Cl 0.00 -0.02 -0.01 -0.02 0.00 0.00 -0.03 -O≡F -2.20 -1.86 -1.92 -1.80 -1.89 -1.73 -1.21 Total 93.50 90.96 90.73 90.32 98.37 99.51 97.80 [REE+Y]2O3 69.35 68.19 67.57 68.12 58.61 58.99 49.38 46 Figure 2.4 Chondrite-normalized REE plot from microprobe results. Normalization values from McDonough and Sun (1995). Erbium (Er), Tm, Yb and Lu are below detection for all samples. 47 Table 2.3 Prominent absorption features of the REE fluorocarbonates in the VNIR range. Bastnaesite Parisite Synchysite Probable Origin Group λ (nm) Shape λ (nm) Shape λ (nm) Shape 500 MIN 501 MIN-n 503 MIN-n Nd3+ 1 511 MIN 510 MIN-n 512 MIN-n Nd3+ 523 MIN 523 MIN 525 MIN-n Nd3+ 533 MIN-w 533 MIN-n 534 MIN-n Nd3+>Eu3+ 545 SH 545 SH-n Nd3+ 557 MIN-w Sm3+ 2 575 SH 574 SH-n 576 SH-n Nd3+ 580 MIN-st 580 MIN-st 580 MIN-st Nd3+ 595 SH 595 SH 595 SH Nd3+>Pr3+ 3 625 MIN 625 MIN-w 625 MIN-w, n Nd3+ 636 MIN-w 637 MIN-w, n Nd3+ 642 SH-w Nd3+ 4 676 MIN 676 MIN 678 MIN Nd3+ 681 SH 681 MIN-w, n 680 SH Nd3+ 689 MIN-w 689 MIN-w 691 SH Nd3+ 693 SH Nd3+ 5 734 SH 732 SH 732 SH Nd3+ 738 SH 738 SH 738 SH Nd3+ 741 MIN-st 741 MIN-st 740 MIN-st Nd3+ 749 MIN-w 749 SH 749 SH Nd3+ 755 SH 755 SH 756 SH Nd3+ 761 SH 762 SH 762 SH Nd3+ 768 SH 770 SH 770 SH Nd3+ 6 781 SH 781 SH 780 SH Nd3+ 784 SH-w 784 SH-w 784 SH-w Nd3+ 792 MIN-db 792 MIN-db 792 MIN-db Nd3+ 797 MIN-db 798 MIN-db 799 MIN-db Nd3+ 803 SH 804 SH 804 SH Nd3+ 813 SH 813 SH 813 SH Nd3+ 821 SH 822 SH 822 SH Nd3+ 7 845 SH 845 SH Nd3+ 864 MIN-st 864 MIN-st 864 MIN-st Nd3+ 871 SH 871 SH 871 SH Nd3+ 889 4819 889 MIN 889 MIN Nd3+ 898 MIN-w 899 SH 899 SH Nd3+ 8 953 MIN 953 MIN 953 MIN-w Sm3+ 978 MIN-w 978 SH-w Nd3+>Pr3+ Bolded values are indexed on Figure 8. Shape and modifier abbreviations: MIN=local minimum, SH=shoulder, -n=noisy, -w=weak, -st=strong, -db=doublet. 48 Table 2.4 Prominent absorption features of the REE fluorocarbonates in the SWIR range. Bastnaesite Parisite Synchysite Probable Origin Group λ (nm) Shape λ (nm) Shape λ (nm) Shape 9 1004 MIN-w 1004 SH-w 998 SH-w Pr3+ 1023 MIN 1023 MIN-w 1023 SH Pr3+ 1093 MIN-st 1093 MIN-st 1093 MIN-st Sm3+ 10 1225 SH-w 1225 SH-w 1225 SH-w Sm3+ 1251 MIN-st 1251 MIN-st 1251 MIN-st Sm3+ 1314 SH-w 1314 SH-w 1301 SH-w Pr3+? 11 1408 SH 1408 MIN-w 1408 MIN-w H2O 1465 MIN 1465 MIN 1459 MIN Pr3+ 1496 MIN-w 1496 MIN-w 1496 MIN-w Pr3+>Sm3+ 1547 MIN-st 1547 MIN 1547 MIN-st Pr3+>Sm3+ 1578 MIN 1578 SH 1578 SH Pr3+>Sm3+>Nd3+ 1666 SH-w 1666 SH-w 1666 SH-w Nd3+ 1710 MIN 1710 MIN 1710 MIN Nd3+ 1742 SH-w 1742 SH-w 1742 SH-w Nd3+ 1761 SH-w 1767 SH-w 1767 SH-w Nd3+ 1817 SH-w 1817 SH-w 1817 SH-w Nd3+ 1886 SH-w 1880 SH-w Pr3+? 1943 SH-w Pr3+ 12 1968 MIN-st 1961 MIN-st 1961 MIN-st Pr3+>Sm3+ & H2O 2037 SH 2037 SH-w 2037 SH-w Pr3+>Eu3+ 2112 SH-w 2118 SH-w 2105 SH-w CO3? 2143 MIN-w 2143 SH-w Pr3+ 13 2243 MIN-st ABSENT ABSENT CO3 14 2312 MIN-db 2312 SH 2312 SH-w CO3>Pr3+>Nd3+>Sm3+ 2324 MIN-db 2324 MIN / SH 2337 MIN CO3>Pr3+>Nd3+>Sm3+ 2355 SH 2355 SH-w 2368 SH-w CO3>Pr3+>Nd3+>Sm3+ 2393 SH-w 2393 MIN-w CO3 15 2437 SH 2443 SH-w CO3>Pr3+>Nd3+>Sm3+ 2480 SH-w 2474 SH 2474 SH-w CO3 2499 SH-w 2499 MIN 2499 SH CO3 2518 MIN 2518 SH-w 2518 MIN CO3 Bolded values are indexed on Figure 2.9. Shape and modifier abbreviations: MIN=local minimum, SH=shoulder, -n=noisy, -w=weak, -st=strong, -db=doublet. 49 Table 2.5 Significant chemical and spectral differences between bastnaesite samples. Sample VNIR pixels SWIR pixels Notes Diao Loa Shan – Sichuan, China 5715 4379 Chemically, shows intermediate values for REE. Spectrally, shows the most resolved features in the VNIR. Burundi 2312 173 Chemically, shows the highest values for heavier REE, such as Gd. Spectrally, is very similar to the Sichuan sample. The 1093 nm and 1251 nm absorptions (Sm3+ related) are particularly pronounced in the suite, which is correlative with this sample having the highest Sm content. Accordingly, the Sm3+ related 1408 nm feature is represented as a local minimum rather than a shoulder Madagascar 613 387 Chemically, shows higher La and Ce but lower middle REE within the bastnaesite suite. Spectrally, it generally shows subdued REE absorptions as compared to other samples. However, a significant difference is that the carbonate related 2243 nm and 2312/2324 nm doublet features are very subdued. Karonga Mine, Congo 1341 135 Chemically, shows intermediate values for REE with a greater abundance of middle REE than the Sichuan sample. Spectrally, it is similar to the Burundi sample with respect to the 1093, 1251 and 1408 nm Sm3+ features, however, it displays less resolved Nd3+ related features at 741, 792/797 and 864 nm. Table 2.6 Significant chemical and spectral differences between parisite samples. Sample VNIR pixels SWIR pixels Notes Muzo, Colombia 265 162 Chemically, more enriched in La, Ce and Pr. Spectrally, tends to show better resolved and narrower features than for the Snowbird sample. Snowbird, USA 2130 1790 Chemically, shows greater values for heavier REE, from Nd to Lu, but is still LREE-enriched. Spectrally, it shows broadened features and finer structure is generally subdued (e.g., clusters near 580 nm and 1500 nm). The carbonate-realted features near 2324 and 2474 nm are also less resolved. The sample’s greater amount of Th will have subjected these crystals to greater radioactivity, which might contribute to broadening of spectral features as the crystal loses crystallinity. 50 2.7.1 Bastnaesite Chemical variability amongst the four samples is moderate, with the samples from Burundi (#39382) and Karonga (#56255) showing greater amounts of heavier REE, such as Nd, Sm and Gd (Table 2.2 and Figure 2.4). None of the crystals show unusual concentrations of other cations or notable compositional zoning. Published analyses of bastnaesite from other localities are comparable to the samples analyzed here (e.g., Zaitsev et al. 1998, Holtstam and Andersson 2007). Comparing the spectra (Figure 2.5) of the four bastnaesite samples in both the VNIR and SWIR ranges, differences are primarily the relative strengths of absorption features generated by lanthanides. The sample from Sichuan is taken as the baseline because it shows the cleanest spectrum, is derived from the most number of pixels, and has supporting spectra from a point spectrometer. The sample from Madgascar shows the largest differences in that, in general, its features are more subdued than its peers. Spectra were collected both perpendicular and parallel to the c-axis for the Sichuan sample, however, no differences were noted. Figure 2.5 Stacked spectra of bastnaesite in VNIR (left, 500 to 1000 nm) and SWIR (right, 975 to 2530 nm) from the two sisuROCK instrument cameras. Spectra from top to bottom: Sichuan (S), Burundi (B), Madagascar (M), Karonga (K). 51 2.7.2 Parisite Chemically, the Snowbird sample shows a greater proportion of heavy rare earth elements (Table 2.2 and Figure 2.4) and especially thorium than the sample from Muzo and neither shows notable compositional zoning. Published analyses of parisite from other localities are comparable to the samples analyzed here (e.g., Zaitsev et al. 1998, Ni et al. 2000). Spectra of the two samples show strong similarities (Figure 2.6) with variations predominantly in the strength of absorption features and relative depths between specific features. Spectra from Muzo samples also show narrower absorption features. Spectra were collected both perpendicular and parallel to the c-axis for both parisite samples, however, no differences were noted. Figure 2.6 Stacked spectra of parisite in VNIR (left, 500 to 1000 nm) and SWIR (right, 975 to 2530 nm) from the two sisuROCK instrument cameras. Spectra from top to bottom: Muzo (M), Snowbird (SB). 2.7.3 Synchysite Three samples of synchysite were studied, however, only the sample from Narsarsuk (Greenland) produced reliable and unambiguous spectra. Spectra indicative of REE-fluorocarbonates could be resolved from the White Cloud Mine (CO, USA) and Morris County (NJ, USA) samples, however, SEM investigations of both revealed complex REE mineralogy, including synchysite, and very fine grained 52 material. Consequently only the Narsarsuk sample is suitable for baseline characterization and discussed in detail. Microprobe results are in accordance with other published analyses (e.g., Forster 2000, 2001, Guastoni et al. 2009). The VNIR and SWIR spectra included 1651 pixels and 175 pixels, respectively. The spectrum (Figure 2.7) of the Narsarsuk sample is dominated by the high relative concentrations of spectrally active lanthanides and resembles spectra of bastnaesite and parisite. Synchysite shows carbonate related minima at 2337 and 2518 nm with a shoulder at 2499 nm. Figure 2.7 Spectra of synchysite from Narsarsuk in VNIR (left, 500 to 1000 nm) and SWIR (right, 975 to 2530 nm) from the two sisuROCK instrument cameras. 2.8 Discussion on the spectral variability of REE‐fluorocarbonate minerals The reflectance spectra from samples of bastnaesite, parisite and synchysite, all members of the REE-fluorocarbonate mineral group, are consistent amongst their respective mineral species. The location of absorption features related to the lanthanides will be a function of the specific REE coordination and asymmetry of crystal structure; because all three minerals have very similar cation sites for the REE consistent patterns were observed. As alluded to in the review of these minerals’ crystal structures, the differences in coordination of the CO3 radical result in the greatest distinction between the REE fluorocarbonates. 53 In general, 15 ‘clusters’ of spectral features are present between 500 nm and 2500 nm in bastnaesite, the REE-fluorocarbonate with the highest concentration of REE and simplest crystal structure. Spectra from parisite and synchysite show strong similarities to bastnaesite spectra, however, noise at shorter wavelengths reduces confidence in this region for these two minerals. Tables 2.3 and 2.4 document resolvable absorption features of the fluorocarbonates based primarily on spectra from the Sichuan (bastnaesite), Snowbird (parisite) and Narsarsuk (synchysite) localities. Figures 2.8 and 2.9 show representative VNIR and SWIR spectra of bastnaesite, parisite and synchysite with bolded index lines and approximate Full Width at Half Max regions for the dominant absorptions of each group. Probable origins for absorptions are described in Tables 2.3 and 2.4. Figure 2.8 VNIR (500 to 1000 nm) spectra of bastnaesite (B, top), parisite (P, middle) and synchysite (S, bottom). Italic numbers denote groups with probable origin described in Table 3. Lines denote prominent absorption features with wavelength position, shaded boxes represent the approximate Full Width at Half Max for each absorption or absorption cluster, borderless box indicates narrow feature. Stacked spectra from sisuROCK instrument. 54 Figure 2.9 SWIR (975 to 2530 nm) spectra of bastnaesite (B, top), parisite (P, middle) and synchysite (S, bottom). Italic numbers denote groups with probable origin described in Table 4. Lines denote prominent absorption features with wavelength position, shaded boxes represent the approximate Full Width at Half Max for each absorption or absorption cluster, borderless box indicates narrow feature. Stacked spectra from sisuROCK instrument. 2.8.1 Intraconfigurational 4f‐4f electronic transitions The REE-related absorptions in the VNIR-SWIR for these LREE-enriched minerals can be largely attributed to Pr3+, Nd3+ and Sm3+, with much lesser influence from Eu3+. Trivalent Y, La3+ and Lu3+ do not have free electrons in the f orbitals to generate absorptions, and absorptions related to Ce3+ do not fall within the VNIR-SWIR range. Similarly, Gd3+ shows absorptions at too high of energy levels, and Dy3+, Tb3+, Ho3+, Er3+, Tm3+ and Yb3+ are either too low in concentration and/or overlapping with other stronger absorption bands to have a noticeable effect in the spectra of the studied minerals. Most differences between the REE fluorocarbonate mineral spectra are small shifts in the location of the absorption minima or apparent strengths of overlapping absorptions represented as shoulders or weak local minima. Most of the differences in absorption minima locations are usually on the scale of one or two spectral channels suggesting that in some cases it may be the result of noise or a function of spectrum slope. Noise at shorter wavelengths in spectra from the sisuROCK imaging spectrometer precludes delineation of some features below ~650 nm, however, data from the ASD point spectrometer 55 with greater signal to noise performance supports subtle features from 650 nm down to 450 nm. Better resolution of local minima as opposed to overlapping absorptions or shoulders in spectra is primarily observed for bastnaesite over synchysite and parisite, likely because of the greater concentration of REE in the Ca-absent bastnaesite. For example, the local minima at 689 nm and 898 nm (Figure 2.8) are generally better defined in the bastnaesite samples, and two resolved absorptions are prevalent with minima at 1547 and 1578 nm (Figure 2.9). Parisite from Muzo and synchysite from Narsarsuk do, however, show more resolved features than parisite from Snowbird near 676 nm, 889 nm and 1547 nm (Figures 2.6 and 2.7). Bastnaesite consistently shows a deep and relatively broad absorption centered at 1968 nm, while parisite and synchysite both show the minimum of this absorption at 1961 nm. A water combination band occurs in this region and likely plays a role, however, Pr3+ also exhibits a strong absorption here. This wavelength shift is consistent with Görller -Walrand and Binnemans (1996) listing this intraconfigurational transition (excitation from ground to the 7F2 multiplet) as being particularly sensitive to the crystal field. 2.8.2 Vibrational bands of the carbonate radical The most important observation amongst the spectra of REE fluorocarbonates is that a distinct absorption at 2243 nm (4458 cm-1) distinguishes bastnaesite from parisite and synchysite, neither of which have this feature (Figure 2.8). Bastnaesite also shows a well resolved doublet with absorption bands at 2312 and 2324 nm (4325 and 4303 cm-1), a shoulder at 2355 nm (4246 cm-1) and an additional band at 2518 nm (3971 cm-1). Notable is that the bastnaesite sample from Madagascar only shows a weak shoulder at 2243 nm and at present the reason for this is uncertain (Figure 2.5). It is possible that structurally bound hydroxyl present in the bastnaesite samples may be the origin of the 2243 nm feature, however, the same substitution of fluorine could be said for both parisite and synchysite and neither show this absorption. Furthermore, EMPA composition of the Madagascar sample does not suggest it as being OH-absent and F-rich. Moreover, our unpublished spectra of uncharacterized bastnaesite from other localities do show the 2243 nm absorption feature. Both parisite and synchysite have broader absorption features that are likely related to the carbonate radical. Parisite shows a weak absorption band at 2324 nm (4303 cm-1) and another near 2499 nm (4002 cm-1) (Figure 2.9). Synchysite shows a more resolved absorption band at 2337 nm (4279 cm-1) and another at 2518 nm (3971 cm-1). 56 2.8.3 Spectral effects of the actinides Most REE minerals will also host the actinide elements Th4+ and U4+. The electronic structure of Th4+ is similar to La3+, and so has no spectral response related to f-orbital transitions. Tetravalent U has a similar structure to Pr3+ and Binnemans et al. (1999) show that although similar, U4+ shows stronger but broader absorption features than Pr3+ because the 5f electrons are less efficiently shielded in actinides than the 4f electrons in lanthanides. Dominant absorptions of U4+ have been recorded near ~1111 nm (~9,000 cm-1) and ~666 nm (15,000 cm-1). For the REE fluorocarbonate minerals studied here, Th and U contents were low enough that no actinide-related spectral features were observed. Unpublished spectra of Th-U-bearing zircon do, however, show spectral features in these regions. 2.9 Implications The REE are strategic elements in many high-tech industries due to their special properties in permanent magnets, metal alloys, catalysts and phosphors, among other applications. Recent volatility in supply and demand has increased exploration in North America for these critical elements but ore mineralogy can be significantly variable and can have significant implications for later beneficiation (e.g., Mariano and Mariano 2012). The reflectance spectra of bastnaesite, parisite and synchysite presented and discussed here form the first systematic spectral characterization and comparison of these three important REE minerals. Their spectra are feature-rich, distinct, and comprise many narrow overlapping absorption bands that can be mostly attributed to 4f-4f intraconfigurational electron transitions of Nd3+, Pr3+, and Sm3+ and possibly Eu3+. Consistent lanthanide-related spectral responses of the REE fluorocarbonates support the notion that their REE cation site is very similar. Variations in these consistent lanthanide-related responses suggest REE concentration information could be extrapolated from reflectance spectra. Variable carbonate-related spectral responses support the notion that the crystallographic sites for the carbonate radical differ between bastnaesite, synchysite and parisite. Exploitable differences for mineral identification using hyperspectral imaging include the carbonate-related absorption at 2243 nm for bastnaesite, as well as a resolved doublet with minima at 2312 nm and 2324 nm. Synchysite shows a carbonate-related minimum at 2337 nm and parisite shows a less resolved carbonate-related minimum near 2324 nm. A Pr3+ and water related absorption is exhibited at 1968 nm for bastnaesite while in synchysite and parisite it occurs at 1961 nm. Collectively, the spectral patterns of this mineral grouping are distinct from other REE minerals phases, such as monazite, xenotime, and eudialyte, among others. 57 This work starts to address the knowledge gap that exists between reflectance spectroscopy and rare earth element mineralogy and future research will include continued characterization and comparison of other important REE-bearing minerals. Accordingly, although hyperspectral remote sensing is becoming increasingly prevalent in the mining and mineral exploration industries, its adoption into REE-focused programs could be of particularly great benefit, such as in defining and assessing ore mineralogy or rapidly logging diamond drill core. Incorporating these spectra into larger spectral databases harnessed by a variety of users will also facilitate the identification of REE fluorocarbonate minerals that may have otherwise gone unnoticed. 58 Chapter 3. Visible to Short Wave Infrared Reflectance Spectroscopy of REE Phosphates 3.1 Chapter summary Reflectance spectroscopy in the visible to short wave infrared regions (500 nm to 2500 nm) was carried out using natural samples of the REE phosphate minerals monazite, xenotime and britholite. Samples were characterized by scanning electron microscopy and electron microprobe analysis. Absorption band positions were recorded with their probable origins and spectral variability amongst the samples is discussed. Spectral features of these minerals are driven primarily by 4f-4f intraconfigurational electronic transitions of trivalent lanthanides. The distinct REE distributions of monazite, xenotime and britholite drive their bulk spectral patterns, which in turn are sufficiently distinct to enable spectral classification. Spectral variability of some specific REE-related absorptions are interpreted to be driven by differences in the cation sites between the crystal structures. The work presented here sets the foundation for the interpretation of reflectance spectra of these REE phosphates and enables exploitation of the observed features by the remote sensing community for detection, identification and quantification of REE phosphates. This is especially relevant for hyperspectral imaging with high spatial resolutions where the spectral response of a pixel becomes increasingly dominated by mineralogy rather than lithology. 3.2 Introduction The rare earth element (REE) bearing phosphates monazite, xenotime and britholite are important mineral phases in certain REE deposits and can also be important for geochronology. Monazite and xenotime are found in a wide range of rocks, whereas britholite is more commonly restricted to alkaline igneous complexes although its distribution is comparatively less well documented. Monazite, LnPO4, is composed primarily of the light rare earth elements from La through to Gd. Xenotime, also LnPO4, is composed primarily of the heavy rare earth elements from Gd through to Lu and including Y. Despite their similar anhydrous compositions, monazite and xenotime crystallize in different space groups. Monazite is monoclinic (P21/n) and xenotime is tetragonal (I41/amd), in part due to the decreasing ionic radii of the lanthanides from La to Lu and the resulting effect on the coordination polyhedra of the Ln3+ and the crystal structure. Britholite-(Ce) has the general formula (Ce,Ca)5(SiO4, PO4)3(OH) and has been reported with both hexagonal and monoclinic symmetry, however, Oberti et al. (2001) suggest the best space group to use is P63. 59 Reflectance spectroscopy, a rapid non-destructive analytical technique requiring little sample preparation, has been used to study the reflectance of these minerals in the visible to short wave infrared regions. This research builds on the study by Turner et al. (2014) that documented the reflectance spectra of the REE fluorocarbonate minerals bastnaesite, parisite and synchysite. Mineralogical and spectroscopic background of REE bearing minerals is provided, followed by a band registry for these phosphate minerals as well as interpretations of spectral absorption features related to the lanthanides. The spectral features of the registry will be the focus of further study in the development of hyperspectral imaging to carry out REE mineral identification, REE mineral abundance estimates, and REE abundance estimates. 3.3 General spectroscopy of the lanthanides The outer radius of the 4f electron shells (~0.3 Å) for the lanthanides is much less than that of their filled 5s and 5p shells (~2 Å, ~1 Å). It can then be approximated that the local electronic environment of Ln3+ cations interacts primarily with those outer shells, leaving the 4f electrons ‘relatively sheltered’ but not completely non-participatory in bonding (e.g., Liu 2005). Electrostatic repulsion of the base ion generates a first level splitting of spectroscopic states, 2S+1L (e.g., 5I). Next, spin-orbit coupling splits these into multiplets, or “J-levels” (e.g., 5I8), and once placed into a crystal field the J-levels are then split into “Stark Sublevels”. Crystal field interactions for the Ln cation include variables such as ligand type, coordination number, and polyhedron asymmetry which all play a role in the location and intensity of energy levels and the associated absorptions (Görller-Walrand and Binnemans 1998). Each of the resulting sublevels provides the potential for promotion of a relaxed electron into an excited state, giving rise to absorption of electromagnetic radiation (light) at a specific energy level (wavelength or wavenumber). The spectroscopy of REE-bearing phases is well established in the fields of physics and chemistry, however, the well understood principles and well-studied doped compounds do not lend to direct translation into mineralogy and hyperspectral remote sensing. For example, the Dieke Diagram (Dieke et al. 1968) details intraconfigurational 4f-4f transitions for ‘free ions’ as deduced through studies of largely mono-lanthanide synthetic compounds, however, the transitions shown do not include splitting of energy levels due to a crystal field nor the complexities of naturally occurring REE minerals with variable REE distributions and other elemental substitutions. Consequently, this diagram and other band registries can only act as proxies to help identify origins of absorption features in reflectance spectra from minerals. In general, and excluding other physical parameters such as grain size, the strength of absorption features by the lanthanides will primarily be a function of the concentration of the ion as well as the 60 specific absorptivity of that ion’s intraconfigurational transitions within a given crystal structure. The location of lanthanide-related absorption features will be primarily a function of the cation’s specific coordination and asymmetry in the host crystal structure. Synthetic lanthanide and actinide orthophosphates and silicophosphates have also been studied in some detail in research fields such as ceramics, nuclear waste, phosphors and nanotechnology and some studies do include natural samples (e.g., Hayhurst et al. 1981, Bernstein 1982, Becker et al. 1984, Rapaport et al. 1999, Assaaoudi et al. 2001, Carpena et al. 2001, Boatner 2002, Silva et al. 2006, Zhang and Vance 2008, Hernadez and Martin 2007, Ardanova et al. 2010, Savchyn et al. 2012, Lenz et al. 2013, Heuser et al. 2014). These studies generally focus on infrared spectroscopy (400 to 4000 cm-1), Raman spectroscopy and UV-VIS luminescence spectroscopy, however, information can be gathered to help understand and assign absorption bands of natural samples in reflectance spectra. 3.4 Review of REE‐related reflectance spectroscopy studies The earliest report of reflectance data from REE phosphate minerals is for monazite from 350 to 2500 nm (Gerharz 1965), for which no conclusive origin was resolved. In his benchmark mineral spectroscopy paper, Hunt (1977) lists monazite (misspelled as monzonite) in his tabulation of common minerals and respective spectral signatures. However, La2+ is listed as the origin of absorptions but La would have a 3+ valence in monazite and would therefore not have spectral features in the VNIR-SWIR (e.g., Liu 2005). Amongst many REE-bearing rock sample spectra, Rowan et al. (1986) include a spectrum (0.4 to 2.5 μm) of britholite from Oka (Quebec) and ascribe bands in the VNIR to Nd3+ and suggest that a series of bands in the SWIR are related to Sm3+. It is unclear if this is a mineral sample or a britholite-dominant rock sample. Clark (1999) covers several REE oxides in his “Spectroscopy of Rocks and Minerals” review and recognizes that the patterns seen in REE minerals are a combination of several lanthanides but states that absorptions are independent of mineralogy. The widely used USGS spectral library contains spectra for fluorapatite, chlorapatite, hydroxylapatite and monazite. The F and Cl bearing apatite samples both contain spectral features in the VNIR suggesting minor Nd3+ content, whereas the monazite spectrum clearly contains REE3+ spectral features and shows many similarities to the monazite spectra studied here. No xenotime spectra are included in the USGS spectral library Version 06 (Clark et al. 2007). Morin Ka (2012) provided some partial spectra for monazite, xenotime and REE-bearing apatite, but no detailed investigations were carried out. Similarly, Kerr et al. (2011) present spectra of monazite, xenotime and REE-bearing apatite but provide no chemical data or interpretation of the spectra. 61 Abstracts by Swayze et al. (2013) and Hoefen et al. (2013) indicate that baseline and applied research is being conducted on REE minerals and deposits, however, no detailed information is available. The abstract by McDowell and Kruse (2014) is focused on the integration of visible to long wave infrared wavelengths for terrain mapping at Mountain Pass, a well-known REE bearing carbonatite. Turner et al. (2014) described the reflectance spectroscopy of the REE fluorocarbonates bastnaesite, parisite and synchysite. In the context of instrument calibrations various authors, such as Weidner et al. (1986), Allen (2007) and Mann et al. (2014), have characterized holmium-based wavelength standards, including the widely used Spectralon-based standards commonly used in reflectance spectroscopy. Thus, there remains an information gap for the interpretation of reflectance spectra for REE phosphate minerals, with respect to identifying and explaining spectral features as well as in supporting the discrimination of minerals based on their spectral characteristics. 3.5 Crystal structure reviews The phosphate classes of REE-bearing minerals show a range of coordination states from eight to 12 with bond lengths as short as 2.25 and as long as ~2.8 Å. Cation and anion substitutions in these minerals can be substantial with complex crystal chemical implications, such as in the alunite supergroup. On the other hand, the crystal structures of monazite and xenotime have been well studied and are well understood. Monazite, (Ce,La,Nd,Th)PO4, and xenotime, YPO4, are very common REE orthophosphate minerals that occur in a wide range of rocks. Despite their similar compositions they have different crystal structures and coordination geometry for the REE site, owing primarily to the size reduction of lanthanides towards Lu. Ni et al. (1995) and Clavier et al. (2011) carried out detailed studies on synthetic REEPO4 and showed that monazite hosts LREE at a distorted 9-coordinated site whereas xenotime hosts HREE at a higher symmetry 8-coordinated site. These studies also demonstrated that the REEPO4 crystals show a systematic decrease in REE-O bond distances when moving from LaPO4 through to LuPO4. Eight of the LREE-O bond lengths for synthetic monazite cluster into three groups between ~2.35 and 2.65 Å, with a ninth outlier having a considerably greater length at ~2.77 Å. Mullica et al. (1984) studied the ninefold coordination polyhedron of La in synthetic LaPO4 and determined that it most closely resembled a distorted pentagonal interpenetrating tetrahedral polyhedron. In xenotime, the bonds are organized into a tighter and more symmetrical structure where four short bonds of ~2.25 Å and four medium bonds of ~2.35 Å create an HREO8 dodecahedron. The total REO content of natural monazite and xenotime attain 70%. 62 Britholite-(Ce)’s more general formula, (Ce,Ca,Th,La,Nd)5(SiO4,PO4)3(OH,F), is chemically and structurally similar to apatite whereby Si replaces P and REE replaces Ca (Mariano 1989, Oberti et al. 2001). Studies of britholite have reported both monoclinic (P21) and hexagonal (P63 and P63/m) symmetry (e.g., Oberti et al. 2001, Pekov et al. 2007). According to refinements by Oberti et al. (2001), the REE can occupy three distinct sites with either 9 (REE1 and REE1a sites, distorted tricapped trigonal prism) or 8 (REE2 site, distorted dodecahedron) coordination. Bond lengths at the 9 coordinated sites cluster into 2.389, 2.510 and 2.700 Å sets of three each, and for the 8 coordinated site they range from 2.377 to 3.180 Å, averaging ~2.467 Å. The shortest of these 8 bonds is coordinated with an OH group. Although the REE1 and REE1a sites are not particularly symmetrical, the REE2 site is highly asymmetrical and is also host to a second ligand type. Total REO content is up to 55%. Figure 3.1 Coordination polyhedra for rare earth element cations in monazite, xenotime, and britholite. Data for polyhedra from Oberti et al. (2001) and Ni et al. (1995). 63 Figure 3.2 Bond distances for REE polyhedra in monazite, xenotime, and britholite, as described in the text. Multiple bond lengths are indicated where found (e.g., 4×O), the OH bond of REE2 in britholite is labeled, and the vertical bars illustrate the minimum, mean, and maximum bond lengths. Data for polyhedra lengths from Oberti et al. (2001) and Ni et al. (1995). 3.6 Experimental methods 3.6.1 Samples Three monazite, four xenotime, and two britholite samples from a larger suite were studied in detail with the scanning electron microscope and none showed compositional zoning. One honey-brown monazite crystal (1 cm × 1 cm × 1 cm) from an unknown location was obtained from the University of Alberta’s geology museum (“537-542 SE Quadrant”) and one similarly sized pinkish brown monazite fragment from Elk Mountain (Nebraska) was obtained from A. Mariano. The third monazite sample comprising multiple smaller grains is from Serra Verde, Brazil, and also obtained from A. Mariano. A xenotime sample from Serra Verde that comprises multiple smaller grains was also obtained. Two euhedral xenotime crystals from Novo Horizonte, Brazil, measured 2.25 × 0.75 × 0.75 cm (sample C) and 2.5 × 1.4 × 1 cm (sample J) and were dark purple in colour with patches of transparency. The xenotime sample from Gunter Quarry was a fragment, measures 1.75 × 2.5 × 2 cm, and is opaque and dark brown. Both britholite samples came from the Kipawa deposit in Ontario. One sample was obtained from A. Mariano and comprised several small grains while the second was borrowed from the Canadian Museum 64 of Nature (sample CMNOC F90-8). Hand sample CMNOC F90-8 contains britholite grains measuring on average ~1 mm × 1 mm × 1 mm. Reagent-grade lanthanide oxide powders, REE-doped Spectralon wavelength calibration samples, and other REE mineral phases were also investigated using EMPA in order to aid in band assignment. 3.6.2 Scanning electron microscopy and electron microprobe analysis The Philips XL30 scanning electron microscope (SEM) at the University of British Columbia, which is equipped with an energy-dispersion X-ray spectrometer (EDS), was used for preliminary examination of mineral mounts of selected minerals and rock fragments studied by reflectance spectroscopy. Selected samples were then analyzed by electron microprobe at the Saskatchewan Research Council’s Advanced Microanalysis Centre using a Cameca SX-100 equipped with five tunable wavelength dispersive spectrometers. Operating conditions were: 40° takeoff angle, beam energy of 15 keV, beam current of 20 nA, beam diameter of 5 μm. The MAN background intensity data was calibrated and continuum absorption corrected. Elements were acquired using analyzing crystals LLIF for FeKα, TaLα, PrLα, EuLα, DyLα, TmLα, MnKα, LaLα, NdLα, GdLα, HoLα, YbLα, BaLα, CeLα, SmLα, TbLα, ErLα, LuLα, PET for CaKα, KKα, ClKα, TiKα, NbLα, YLα, SrLα, ZrLα, PKα, UMα, ThMα, and LTAP for MgKα, FKα, NaKα, SiKα, AlKα. Counting times were 10 seconds for Zr and P and 15 seconds for all other elements, with off peak count times of 10 seconds. The standards (with elements) were SPI-barite (Ba), SPI-celestite (Sr), SPI-YAG (Y, Al), Smithsonian Cr-augite (Mg, Ca), Smithsonian ilmenite (Fe, Ti), Smithsonian apatite (F, P), Smithsonian microcline (K), Smithsonian zircon (Zr), Harvard albite (Si, Na), Cameca Mn (Mn), SPI2-TlCl (Cl), SPI2-Nb (Nb), SPI2-La (La), SPI2-Ce (Ce), SPI2-Pr (Pr), SPI2-Nd (Nd), SPI2-Sm (Sm), SPI2-Eu (Eu), SPI2-Gd (Gd), SPI2-Tb (Tb), SPI2-Dy (Dy), SPI2-Ho (Ho), SPI2-Er (Er), SPI2-Tm (Tm), SPI2-Yb (Yb), SPI2-Lu (Lu), SPI2-Ta (Ta), SPI2-Th (Th), and SPI2-U (U). Formulae were calculated based on four anions for monazite and xenotime, and 26 anions (O+F+Cl) for britholite. The amount of OH- (as H2O) was determined by stoichiometry based on full occupation of the “O4” atomic site by two atoms of F-, Cl- and OH- as per Oberti et al. (2001). 3.6.3 Reflectance spectroscopy Reflectance spectroscopy was primarily carried out using the sisuROCK instrumentation (manufactured by SPECIM Spectral Imaging Ltd.) at the University of Alberta’s CoreSensing Facility, and data was handled using ENVI 4.4, a widely used and commercially available software package. Two imaging spectrometers (“cameras”) acquired reflectance spectra in the visible-near infrared (VNIR, 396 nm to 1003 nm over 784 channels for an average spectral resolution of 0.77 nm) and shortwave infrared 65 (SWIR, 928 nm to 2530 nm over 256 channels for an average spectral resolution of 6.26 nm) portions of the electromagnetic spectrum in high spatial resolution mode. Spatial resolution of the cameras in this mode was approximately 0.079 mm / pixel in the VNIR and 0.241 mm / pixel in the SWIR (Figure 3). Noise was very prevalent in the shortest wavelength portion of the VNIR camera below ~550 nm and moderate from 550 nm to ~650 nm. In the high spatial resolution mode, averaging spectra ~16 neighboring pixels resulted in reliable spectra in the noisier ranges that would be useable in spectral libraries. Spectra presented here originate from single crystals, multiple crystals within a single rock sample, and from multiple loose single crystals. Spectra documented here are nominally an average of 3249 pixels for the VNIR camera and 1109 pixels for the SWIR camera. Samples were placed on a matte black surface that translates the samples under the camera and has very low reflectance across the sampled wavelength range. Some samples were propped up with foam blocks to ensure surfaces of interest faced the spectrometers and were in focus. All samples were thick enough that we could assume that the reflectance spectra are representative of the mineral target. Simple Regions Of Interest (ROI) were used on most samples to select target pixels for averaging. For the loose grains and crystals in hand samples, a priori knowledge about the sample allowed several baseline spectra to be isolated from single pixels. These spectra were then used to run mixture-tuned matched filtering within the ENVI software package on an entire hyperspectral image, from which a strict qualitative threshold allowed a discrete selection of pixels to be averaged. This process enabled the averaging of tens to hundreds of pixels per sample (Table 3.1) to produce a representative spectrum as stated above. Reflectance spectra did not have the continuum removed so as to present the data unmodified and to facilitate comparison against data from other earlier publications. 66 Table 3.1 Pixel counts per sample used to produce an average spectrum. Mineral Sample VNIR ROI Pixels SWIR ROI Pixels Monazite Serra Verde 1995 180 Elk Mountain 1283 1121 UofA Unknown SE Quad 6226 848 Xenotime Serra Verde 1848 215 Gunter Quarry 2826 419 Novo Horizonte C 2097 1931 Novo Horizonte J 5240 4955 Britholite Kipawa CMNOC F90-8 4001 186 Kipawa Mariano 001 3727 122 Figure 3.3 Example of false colour VNIR hyperspectral reflectance imagery – the sample is xenotime from Novo Horizonte “C”, looking perpendicular to the c-axis. For this particular scene, 2097 pixels were included in a ROI for calculating the averaged spectrum. Crystal measures 0.75 x 2.25 cm. 3.7 Electron microprobe compositions Samples described here were selected from a larger set and were characterized by imaging reflectance spectroscopy, scanning electron microscopy and microprobe analysis. Tables 3.2, 3.3 and 3.4 document electron microprobe compositions for the REE phosphate minerals in this study and Figure 3.4 shows Chondrite-normalized patterns for the samples. 67 Table 3.2 Electron microprobe compositions for monazite. Sample # 119 15-H-7 UofA 537-542 SE 119 15-H-7 UofA 537-542 SE Locality Serra Verde Elk Mountain Unknown Serra Verde Elk Mountain Unknown # Analyses 5 5 5 Nb2O5 (wt.%) 0.01 0.00 0.09 Nb5+ (apfu) 0.00 0.00 0.00 P2O5 25.78 22.62 27.77 P5+ 0.91 0.85 0.96 SiO2 1.08 1.90 0.79 Si4+ 0.05 0.08 0.03 TiO2 0.01 0.00 0.00 Ti4+ 0.00 0.00 0.00 ZrO2 0.04 0.02 0.03 Zr4+ 0.00 0.00 0.00 UO2 0.39 0.32 0.05 U4+ 0.00 0.00 0.00 ThO2 7.36 12.15 3.55 Th4+ 0.07 0.12 0.03 La2O3 13.56 10.60 14.33 La3+ 0.21 0.17 0.22 Ce2O3 25.91 22.98 33.16 Ce3+ 0.39 0.37 0.49 Pr2O3 3.35 3.03 3.84 Pr3+ 0.05 0.05 0.06 Nd2O3 12.72 11.86 11.42 Nd3+ 0.19 0.19 0.17 Sm2O3 3.64 4.94 1.90 Sm3+ 0.05 0.08 0.03 Eu2O3 0.01 0.12 0.02 Eu3+ 0.00 0.00 0.00 Gd2O3 2.01 2.83 0.42 Gd3+ 0.03 0.04 0.01 Tb2O3 0.28 0.48 0.00 Tb3+ 0.00 0.01 0.00 Dy2O3 0.61 1.64 0.08 Dy3+ 0.01 0.02 0.00 Ho2O3 0.00 0.00 0.00 Ho3+ 0.00 0.00 0.00 Er2O3 0.00 0.08 0.00 Er3+ 0.00 0.00 0.00 Tm2O3 0.00 0.00 0.00 Tm3+ 0.00 0.00 0.00 Yb2O3 0.00 0.00 0.00 Yb3+ 0.00 0.00 0.00 Lu2O3 0.00 0.00 0.00 Lu3+ 0.00 0.00 0.00 Y2O3 1.62 0.45 0.11 Y3+ 0.04 0.01 0.00 MnO 0.00 0.00 0.02 Mn2+ 0.00 0.00 0.00 CaO 0.46 0.64 0.03 Ca2+ 0.02 0.03 0.00 Na2O 0.01 0.02 0.00 Na+ 0.00 0.00 0.00 Cl 0.01 0.02 0.03 Cl- 0.00 0.00 0.00 F 0.39 0.36 0.36 F- 0.05 0.05 0.05 O≡Cl 0.00 0.00 -0.01 O2- 3.95 3.95 3.95 O≡F -0.16 -0.15 -0.15 TOTAL 99.08 96.90 97.84 CATSUM 2.01 2.03 1.99 [REE+Y]2O3 63.71 59.01 65.28 AN SUM 4.00 4.00 4.00 *Ta, Al, Sr, Mg, Fe, Ba and K were sought but not detected. *Formula contents on a basis of 4 anions pfu 68 Table 3.3 Electron microprobe compositions for xenotime. Sample # 1131 C 1131 J 119 CMN F92-15 1131 C 1131 J 119 CMN F92-15 Locality Novo Horizonte Novo Horizonte Serra Verde Gunter Quarry Novo Horizonte Novo Horizonte Serra Verde Gunter Quarry # Analyses 5 5 6 5 Nb2O5 (wt.%) 0.07 0.08 0.06 0.08 Nb5+ (apfu) 0.001 0.001 0.001 0.001 P2O5 29.12 30.83 32.18 34.16 P5+ 0.899 0.919 0.938 0.967 SiO2 0.2 0.23 0.46 0.53 Si4+ 0.007 0.008 0.016 0.018 TiO2 0 0.01 0 0 Ti4+ 0 0 0 0 ZrO2 0.13 0.11 0.16 0.17 Zr4+ 0.002 0.002 0.003 0.003 UO2 0.07 0.01 1.59 0.58 U4+ 0.001 0 0.012 0.004 ThO2 0.07 0.33 0.33 0.53 Th4+ 0.001 0.003 0.003 0.004 Al2O3 0 0 0 0.01 Al3+ 0 0 0 0 La2O3 0 0 0 0 La3+ 0 0 0 0 Ce2O3 0 0.02 0.04 0 Ce3+ 0 0 0.001 0 Pr2O3 0 0 0 0.01 Pr3+ 0 0 0 0 Nd2O3 0.05 0.08 0.29 0.15 Nd3+ 0.001 0.001 0.004 0.002 Sm2O3 1.05 0.7 0.77 0.34 Sm3+ 0.013 0.008 0.009 0.004 Eu2O3 0.35 0.12 0 0 Eu3+ 0.004 0.001 0 0 Gd2O3 5.71 3.88 2.74 1.46 Gd3+ 0.069 0.045 0.031 0.016 Tb2O3 1.48 1.02 0.87 0.33 Tb3+ 0.018 0.012 0.01 0.004 Dy2O3 9.75 7.73 6.93 3.58 Dy3+ 0.115 0.088 0.077 0.039 Ho2O3 1.72 1.67 1.51 1.02 Ho3+ 0.02 0.019 0.017 0.011 Er2O3 4.94 5.26 5.18 3.88 Er3+ 0.057 0.058 0.056 0.041 Tm2O3 0.51 0.54 0.66 0.61 Tm3+ 0.006 0.006 0.007 0.006 Yb2O3 2.14 2.51 4.09 5.15 Yb3+ 0.024 0.027 0.043 0.052 Lu2O3 0.11 0.23 0.6 0.82 Lu3+ 0.001 0.002 0.006 0.008 Y2O3 41.82 44.51 42.87 45.72 Y3+ 0.812 0.834 0.785 0.813 FeO 0.02 0 0 0 Fe2+ 0.001 0 0 0 BaO 0 0 0 0.01 Ba2+ 0 0 0 0 CaO 0 0 0.01 0.01 Ca2+ 0 0 0 0 Na2O 0.01 0 0 0 Na+ 0.001 0 0 0 Cl 0 0 0 0.02 Cl- 0 0 0 0.001 F 0.3 0.33 0.35 0.51 F- 0.035 0.037 0.038 0.054 O≡Cl 0 0 0 0 O2- 3.965 3.963 3.962 3.945 O≡F -0.13 -0.14 -0.15 -0.21 TOTAL 99.49 100.06 101.54 99.46 CATSUM 2.052 2.036 2.017 1.994 [REE+Y]2O3 69.63 68.27 66.55 63.07 AN SUM 4 4 4 4 *Ta, Sr, Mg, Mn, and K were sought but not detected *Formula contents on a basis of 4 anions pfu 69 Table 3.4 Electron microprobe compositions for britholite. Sample # Mariano 001 CMNOC F90-8 Mariano 001 CMNOC F90-8 Locality Kipawa Kipawa Kipawa Kipawa # Analyses 5 6 Nb2O5 (wt.%) 0.01 0.02 Nb5+ (apfu) 0.001 0.002 P2O5 3.03 3.99 P5+ 0.708 0.854 SiO2 17.23 20.12 Si4+ 4.758 5.088 ZrO2 0.01 0.02 Zr4+ 0.001 0.002 UO2 0.22 0 U4+ 0.014 0 ThO2 1.26 3.24 Th4+ 0.079 0.186 La2O3 9.01 6.09 La3+ 0.918 0.568 Ce2O3 17.29 15.4 Ce3+ 1.748 1.426 Pr2O3 2.04 1.91 Pr3+ 0.205 0.176 Nd2O3 7.75 7.22 Nd3+ 0.764 0.652 Sm2O3 1.91 1.63 Sm3+ 0.182 0.142 Eu2O3 0.07 0.16 Eu3+ 0.007 0.014 Gd2O3 1.53 1.52 Gd3+ 0.14 0.127 Tb2O3 0.33 0.26 Tb3+ 0.03 0.022 Dy2O3 1.93 1.92 Dy3+ 0.172 0.156 Ho2O3 0.22 0.39 Ho3+ 0.019 0.031 Er2O3 1.15 1.26 Er3+ 0.1 0.1 Tm2O3 0.02 0.1 Tm3+ 0.002 0.008 Yb2O3 0.52 0.96 Yb3+ 0.044 0.074 Lu2O3 0 0.01 Lu3+ 0 0.001 Y2O3 9.81 11.35 Y3+ 1.442 1.527 SrO 0.03 0.03 Sr2+ 0.005 0.004 MnO 0.03 0.02 Mn2+ 0.007 0.004 BaO 0 0.01 Ba2+ 0 0.001 CaO 16.33 17.57 Ca2+ 4.832 4.76 Na2O 0.08 0 Na+ 0.043 0 Cl 0.01 0.05 Cl- 0.005 0.021 F 1.59 1.28 F- 1.389 1.024 H2O * 0.33 0.57 OH- 0.607 0.955 O≡Cl 0 -0.01 O2- 24 24 O≡F -0.67 -0.54 TOTAL 93.07 96.55 Sum [Si] 5.466 5.942 [REE+Y]2O3 53.58 50.18 Sum [REE1, 2] 10.755 9.983 Sum (F, Cl, OH) 2.001 2 CATSUM 16.219 15.927 AN SUM 26 26 *Determined by stoichiometry, H2O calc. assuming 2 (OH-, Cl-, F-) pfu **Formula contents on a basis of 26 anions (O+F) pfu, ***Ta, Ti, Al, Mg, Fe and K were sought but not detected 70 Monazite microprobe compositions show similar chemical distributions of the REE for all samples (Table 3.2, Figure 3.4, top left), however, the Elk Mountain sample showed slightly higher rare earth elements from Sm onwards. Totals for Serra Verde and Unk-537-542 samples are satisfactory while for the analyses of the Elk Mountain sample are lower and show high Th and Si contents, indicating substitutions moreso related to huttonite, ThSiO4, than cheralite, CaTh(PO4)2 (Linthout 2007, Clavier et al. 2011). Xenotime microprobe compositions for the two Novo Horizonte samples show elevated Eu-Gd-Tb-Dy and lesser Yb, Lu and U as compared to the samples from Gunter Quarry and Serra Verde, resulting in two main groups (Table 3.3, Figure 3.4, bottom left). Neodymium and Sm concentrations are roughly equivalent in all four samples, and La, Ce, and Pr contents are either very low or below detection. Analytical totals are satisfactory. The two britholite samples originated from different collections but the same ore deposit (Kipawa) and show similar concentrations of rare earths (Table 3.4, Figure 3.4, top right). Britholite is commonly noted to be at least partially metamict (e.g., Pasero et al. 2010) and in general this is true for samples from Kipawa (e.g., Noe et al. 1993). The Mariano-001 sample displays a low analytical total while sample CMNOC F90-8 is satisfactory for a mineral species commonly reported as metamict. Microprobe compositions show that Ce is the most abundant REE, REE>Ca, Si>P and F>OH,Cl, therefore these samples are fluorbritholite-(Ce). Figure 3.4 (bottom right) shows the Chondrite-normalized EMPA results for all samples, as well as selected samples from each mineral to demonstrate relative total REE contents and patterns. Xenotime is strongly enriched in the heavy rare earth elements (HREE) and monazite is enriched in the light rare earth elements (LREE), while britholite contains moderate amounts of all REE. 71 Figure 3.4 Chondrite-normalized REE plots for monazite (top left), britholite (top right), xenotime (bottom left) and selected samples on combined plot (bottom right). 3.8 Spectra and spectral variability of REE‐phosphate minerals Each of the three REE-phosphate minerals exhibits its own characteristic distribution of rare earth elements and therefore spectral patterns derived from 4f-4f intraconfigurational electronic transitions. As seen from microprobe compositions, xenotime is populated by the heavy rare earths (Gd to Lu), monazite by the light rare earths (La to Sm), and britholite accommodates both lights and heavies yet has lower overall REE contents (Figure 3.4 and Tables 3.2 to 3.4). Accordingly, the intramineral variations and spectra for each mineral will be discussed and then overall spectral variability will be addressed. Figures 3.6, 3.7 and 3.8 document the spectra from samples of each mineral and show the location of clusters and prominent absorptions listed in Tables 3.5 through 3.8.These four “Band Index Tables” document prominent absorption features for the REE phosphates and include probable origins of the features, as chosen through comparison with reflectance spectra from unpublished reagent grade lanthanide oxide spectra, REE-doped calibration standard spectra, and other REE-bearing mineral spectra for which compositional data exists, as well as REE spectroscopy literature. Clusters are denoted by shading and important absorptions are bolded for each mineral. All documented clusters and main 72 absorption features are shown on “Index Figures” 3. 9 (VNIR range) and 3.10 (SWIR range) for representative spectra of monazite, britholite and xenotime. The position and shape of these features were recorded using reflectance spectra (i.e., not continuum removed spectra). Descriptors for absorption bands include MIN (minimum) with modifiers –st (strong), -w (weak), -n (noisy), –b (broad) and SH (shoulders) with modifiers –w (weak) and –n (noisy). Noisy and very weak absorption bands are typically restricted to scenarios where other spectra show reliable features near the same wavelength position. For example, the ~957 nm weak minimum and noisy shoulders of LREE-enriched monazite were included based on the presence of the 953 nm absorption in LREE-enriched bastnaesite. Divisions between VNIR and SWIR ranges on the tables are denoted by heavy horizontal line. Figure 3.5 Example spectrum showing types absorption band labels with data points indicated by diamonds. Spectrum is of monazite (UofA Unknown sample) – see Figure 3.6 and Table 3.5. 73 3.8.1 Monazite reflectance spectra The Serra Verde and UofA Unknown monazite samples have fairly similar elemental distributions, however, the sample from Elk Mountain shows higher medium REE values and considerably higher Th and Si and lower P content than the other two samples. Figure 3.6 Spectra of characterized monazite samples in VNIR (left, 500 to 1000 nm) and SWIR (right, 950 to 2530 nm) with absorption band clusters shown and prominent absorption lines of the UofA sample. From top down spectra are from Serra Verde (red, SV), UofA Unknown (blue), and Elk Mountain (black, EM). 74 Table 3.5 Prominent absorption features of monazite samples in the VNIR and SWIR ranges. MonaziteClusters wavelength shape wavelength shape wavelength shape1 574 SH 575 SH 574 SH Nd576 SH 576 MIN ‐ n 576 MIN Nd579 MIN ‐ st 579 MIN ‐ s t 579 SH Nd581 MIN ‐ st 581 MIN ‐ s t 581 MIN ‐ s t Nd585 MIN ‐ w, n 584 MIN ‐ w, n 585 SH Nd592 MIN ‐ w, n 592 MIN ‐ w, n 592 SH Nd, Pr2 622 MIN ‐ n 622 MIN ‐ n 622 MIN ‐ n Nd624 MIN ‐ n 624 MIN ‐ n 624 MIN ‐ n Nd625 MIN ‐ n 625 MIN ‐ n 626 MIN ‐ n Nd652 MIN 653 SH ‐ w, n, b 651 MIN ‐ w, n, b Ho, Er, U?672 MIN 673 MIN 672 SH Nd676 SH 676 SH 677 SH Nd678 MIN 679 MIN 679 MIN Nd689 SH 689 SH 688 SH Nd3 734 SH 734 SH Nd738 MIN 738 MIN 738 MIN Nd745 MIN ‐ st 745 MIN ‐ s t 745 MIN ‐ s t Nd747 SH 747 SH 748 SH Nd749 SH 749 SH 749 SH Nd757 SH 757 SH 758 SH Nd770 SH 770 SH 770 SH ‐ w Nd4 792 SH 792 MIN ‐ w 792 SH Nd796 SH 796 MIN ‐ w 795 SH Nd800 MIN ‐ st 800 MIN ‐ s t 800 MIN ‐ s t Nd804 SH 804 MIN ‐ w 804 SH Nd814 SH 814 SH 814 SH Nd830 SH 830 SH 830 SH Nd5 863 MIN 863 MIN 863 MIN Nd871 MIN ‐ st 871 MIN ‐ s t 872 MIN ‐ s t Nd877 SH 877 SH 877 SH Nd886 MIN 887 MIN 887 MIN Nd896 MIN ‐ w 896 SH 898 SH Nd906 MIN ‐ w 906 MIN ‐ w 907 SH Nd6 941 MIN 941 MIN 941 MIN Sm957 SH ‐ n 957 MIN ‐ w 959 SH ‐ n Sm7 978 MIN ‐ n 980 MIN ‐ n 978 MIN ‐ n Yb>Er8 1023 SH 1023 MIN 1017 SH Pr1080 MIN ‐ st 1074 MIN ‐ s t 1074 MIN ‐ s t Sm1105 MIN 1105 SH 1105 SH Sm1143 SH 1150 SH Pr9 1232 MIN ‐ st 1232 MIN ‐ s t 1232 MIN ‐ s t Sm1257 MIN 1257 MIN 1257 MIN Sm10 1377 MIN 1377 MIN 1377 MIN Sm1415 SH 1415 SH 1415 SH H2O1452 SH 1452 SH 1446 SH Pr1471 MIN 1471 MIN 1471 MIN Pr1553 MIN 1553 MIN 1547 MIN Sm?1578 SH 1578 SH 1578 SH Pr1685 SH 1691 MIN 1691 SH Dy, Nd1710 SH 1710 MIN 1710 MIN Nd1717 SH 1717 SH 1717 SH Nd1735 SH 1735 SH Nd11 1961 MIN ‐ st 1968 MIN ‐ s t 1955 MIN ‐ s t Pr, Sm, H2O2005 SH 2011 SH 2011 SH Pr, Sm12 2212 MIN 2212 SH 2212 SH REE/OH/PO42312 SH REE/OH/PO42362 SH 2368 MIN 2362 SH REE/OH/PO42399 SH 2393 SH 2393 SH REE/OH/PO42424 MIN 2424 MIN 2418 SH REE/OH/PO42474 MIN 2499 MIN 2499 MIN REE/OH/PO4Serra Verde UofA_UnkSE Elk Mtn Probable Origin75 VNIR For monazite, the VNIR spectral range has been divided into 7 main clusters of absorptions (Figure 3.6, Table 3.5). Cluster 1 comprises 1 main absorption band doublet with minima at 579 and 581 in addition to several shoulders. Noise complicates the identification of definitive shoulders, however, the features are persistent across the suite of 3 monazite samples with EMPA data as well as others without chemical data. Probable absorptions occur near 525 nm, however, noise reduces confidence in these features. Cluster 2 comprises 2 main bands at 625 and 679 nm, with a series of lesser shoulders and minima. The absorptions near 625 nm typically comprise three absorptions at 622, 624 and 625 nm, however, noise prevents definitive assignment of a triplet. Cluster 3 has a strong minimum at 745 nm, a small minimum at 738 nm and four subtle shoulders up to 770 nm. Cluster 4 shows a broad absorption band at 800 nm and shoulders at both shorter and longer wavelengths. Cluster 5 is characterized by a series of sharp absorptions. The deepest occurs at 871 nm, then at 863 nm, followed by a series of lesser absorptions minima and shoulders at 877, 887, 896 and 906 nm. Cluster 6 is a minimum at 941 nm with a shoulder or weak minimum at 957 nm, while Cluster 7 is a single shallow absorption at 978 nm. The most notable differences in the VNIR spectra across the 3 monazite samples are the strengths of the 652 and 671 nm absorptions of Cluster 2, which are strongest in the sample from Serra Verde. The absorption band at 978 nm also shows some variability in strength, with the Elk Mountain sample being most pronounced and the UofA Unknown being the weakest. SWIR For monazite, the SWIR spectral range has been divided into 5 main clusters of absorptions starting after Clusters 6 and 7, which are better resolved in the VNIR range (Figure 3.6, Table 3.5). Cluster 8 is characterized by a strong absorption band centered at 1074 to 1080 nm with pronounced shoulders at 1023 nm, 1105 nm and near 1150 nm. Cluster 9 shows a strong band at 1232 nm with a lesser absorption at 1257 nm. Cluster 10 is a broad region with numerous absorptions and shoulders. The three main absorption minima are located at 1377 nm, 1471 nm and 1547 to 1553 nm. Cluster 11 comprises a strong minima centered near 1961 nm with a shoulder near 2011 nm. Cluster 12 is characterized by a series of weaker bands ranging from 2212 nm to 2499 nm. Prominent minima in this range for monazite can occur near 2424 and 2499 nm. Spectra for the monazite samples in the SWIR range are very consistent. Subtle differences include a shift in the ~1960 nm feature and the relative intensities of absorption bands at ~2424 and ~2499 nm. 76 3.8.2 Xenotime reflectance spectra Chemical differences in xenotime samples separate the 2 from Novo Horizonte (higher Eu through Dy) from those from Serra Verde and Gunter Quarry (higher Yb, Lu, U). Bulk patterns are consistent across all xenotime samples, however, the notable chemical distributions give rise to spectral differences in both the VNIR and SWIR. The euhedral samples from Novo Horizonte allowed examination of spectra both parallel and perpendicular to the c-axis of these crystals. In both crystals the total reflectance in both VNIR and SWIR was lower when looking parallel to the c-axis but all the major absorption features were present and in roughly proportional relative strengths. However, the lower reflectance meant that noise was more prevalent and therefore more subtle features were harder to reliably identify and record (e.g., shoulders). Consequently, we will only discuss results of the spectra that were acquired across the c-axis of the Novo Horizonte samples. Figure 3.7 Spectra of characterized xenotime samples in VNIR (left, 500 to 1000 nm) and SWIR (right, 950 to 2530 nm) with absorption band clusters shown and prominent absorption lines of the GQ sample. From top down spectra are from Serra Verde (purple, SV), Gunter Quarry (black, GQ), Novo Horizonte J (blue, NH-J), and Novo Horizonte C (red, NH-C). 77 Table 3.6 Prominent absorption features of xenotime samples in the VNIR range. XenotimeClusters wavelength shape wavelength shape wavelength shape wavelength shape1 644 MIN 643 MIN 643 MIN 644 MIN Ho>Er647 MIN 647 MIN 648 MIN 648 SH Ho, Er651 SH 651 MIN 651 MIN Ho, Er653 MIN 653 MIN 653 MIN 653 MIN Ho, Er658 MIN 657 MIN 658 MIN 657 SH Ho, Er661 SH 661 SH 661 MIN 661 SH Ho, Er665 MIN ‐ w 665 MIN 665 SH 665 SH Ho, Er667 SH ‐ n 667 MIN ‐ n 667 SH 667 SH Ho, Er668 MIN ‐ w 668 SH 668 MIN 668 MIN Er>Ho, U?678 MIN ‐ n 678 SH ‐ n 678 MIN 678 MIN Er>Ho, Nd689 SH 689 SH 689 MIN ‐ w 689 MIN ‐ w Er>Ho, Nd692 MIN 692 MIN 692 SH 692 SH Tm > Nd2 734 MIN Nd?740 MIN 740 MIN 740 MIN 740 MIN Dy>Ho, Nd?742 SH 742 SH 742 SH 742 SH Dy>Ho, Nd?748 MIN 747 MIN 748 MIN 747 MIN Dy>Ho, Nd?753 MIN 753 MIN 754 MIN 752 MIN Dy>Ho, Nd?760 MIN 760 MIN 760 MIN 760 MIN Dy>Ho, Nd?767 SH ‐ w 767 SH ‐ w 767 SH ‐ w 767 SH ‐ w Nd?770 MIN ‐ w Nd?777 SH ‐ w Nd?3 781 MIN ‐ w 781 MIN ‐ w 781 MIN ‐ w 781 SH ‐ w Tm803 MIN 803 MIN 803 MIN 803 MIN Dy>Er, Nd?811 MIN ‐ s t 811 MIN ‐ s t 811 MIN ‐s t 811 MIN ‐ s t Dy>Er, Nd?817 SH 817 SH 817 SH 817 SH Dy, Er, Nd?822 MIN 822 SH 822 MIN 822 SH Dy, Er, Nd?823 MIN 823 MIN 823 SH 823 SH Dy, Er, Nd?826 SH 826 MIN 826 MIN Dy, Er, Nd?828 MIN 828 SH 829 MIN 829 SH Dy, Er, Nd?4 862 SH Nd865 MIN Nd872 MIN 872 MIN 872 MIN 872 MIN Nd876 SH 876 SH 876 MIN ‐ w 876 SH Nd877 SH 877 SH 877 MIN ‐ w 877 SH Nd5 896 MIN 896 SH 896 MIN 896 MIN Dy, Yb, Ho909 SH 909 SH 909 SH 908 SH Dy, Yb, Ho914 MIN ‐ s t 914 MIN ‐ s t 914 MIN ‐ s t 915 MIN ‐ s t Dy, Yb, Ho920 SH 920 SH 920 SH 920 SH Dy, Yb932 MIN 932 MIN 933 MIN 933 MIN Dy, Yb951 MIN ‐ n 951 MIN ‐ n 951 MIN ‐ n 951 MIN Dy, Yb, Sm953 MIN ‐ n 953 MIN ‐ n 953 MIN ‐ n 953 MIN Dy, Yb, Sm961 MIN ‐ n 961 MIN ‐ n 961 MIN ‐ n Dy, Er, Yb6 977 MIN ‐ s t 977 MIN ‐ s t Yb, Er978 MIN ‐ s t 978 MIN ‐ s t 978 MIN ‐ s t 978 MIN ‐ s t Yb, ErNovo C Novo J Gunter Quarry Serra Verde Probable Origin78 Table 3.7 Prominent absorption features of xenotime samples in the SWIR range. VNIR For xenotime, the VNIR spectral range has been divided into 6 main clusters of absorptions (Figure 3.7, Tables 3.6 and 3.7). Cluster 1 comprises 4 main absorption bands at 643 nm, 653 nm, 668 nm and 678 nm plus a 5th weaker feature at 692 nm. Numerous shoulders and weak absorptions also exist in this range. Cluster 2 comprises 4 main bands at 740, 748, 754 and 760 nm, typically with deepening absorption towards 760 nm. Cluster 2 also shows a series of shoulders and subtle absorptions. Cluster 3 has one very strong band at 811 nm and is flanked by prominent features at 803 nm and 828 nm, as well XenotimeClusters wavelength shape wavelength shape wavelength shape wavelength shape1004 SH 1004 SH Yb, Er7 1036 SH 1036 SH Yb, Er1080 SH 1080 SH Sm1099 MIN ‐ s t 1099 MIN ‐ s t 1099 SH 1099 SH Dy1112 MIN 1112 MIN 1118 SH 1118 MIN Dy, U?1143 SH 1143 SH 1143 MIN ‐ s t 1143 MIN ‐ s t Ho, Dy, U?1168 SH 1175 SH Dy>Ho1187 MIN 1187 MIN 1193 SH 1193 SH Ho, Dy8 1213 SH 1213 SH Dy, Tm1257 MIN 1257 SH 1257 SH 1257 SH Dy, Sm1276 MIN 1276 MIN 1276 MIN 1276 MIN Dy1301 MIN 1301 MIN 1307 MIN 1307 MIN Dy1320 SH 1320 SH 1326 SH 1326 SH Dy1358 SH 1358 SH 1358 MIN 1364 MIN Dy9 1408 MIN 1408 MIN 1415 MIN 1415 MIN H2O10 1440 MIN 1433 MIN 1452 SH 1452 SH Er1503 MIN 1503 MIN 1503 MIN ‐ s t 1503 MIN Er>Sm, U?1528 MIN 1528 MIN 1528 MIN 1534 MIN Er>Sm1559 SH 1559 SH 1559 SH 1559 SH Er1585 MIN 1585 MIN 1585 SH 1585 SH Er11 1616 MIN 1616 MIN 1616 SH ‐ w Dy>Sm, Tm1660 SH 1660 SH Dy1704 MIN 1704 MIN 1704 MIN ‐ s t 1704 MIN ‐ s t Dy1723 MIN 1723 MIN 1723 SH 1723 SH Dy, Tb1761 SH 1761 SH 1761 SH 1761 SH Dy, Tb1805 MIN 1805 MIN 1805 SH 1805 SH Dy12 1848 SH 1848 SH Tb, Ho1880 MIN 1880 MIN 1880 MIN Tb, Ho13 1936 MIN 1936 MIN 1936 MIN ‐ s t 1936 MIN Ho, Tb, H2O1961 MIN 1961 MIN 1961 MIN 1961 MIN Ho1980 MIN 1980 MIN 1980 SH 1980 SH Ho, Sm2005 MIN ‐ s t 2005 MIN ‐ s t 2005 MIN 2005 MIN ‐ s t Ho, Sm, Tb2030 MIN 2030 MIN 2030 SH 2030 SH Ho, Sm2043 MIN 2043 MIN 2049 SH 2049 SH Ho, Sm2130 SH 2130 SH Ho, Sm, Tb14 2212 MIN ‐ s t 2212 MIN ‐ s t 2218 MIN 2218 MIN ‐ s t REE/OH/PO42262 MIN ‐ s t 2262 MIN ‐ s t 2262 SH REE/OH/PO42312 MIN ‐ s t 2312 MIN ‐ s t 2312 MIN 2318 MIN REE/OH/PO42349 SH 2349 SH 2355 MIN REE/OH/PO42399 MIN 2399 MIN 2405 SH REE/OH/PO42462 MIN 2462 MIN 2462 SH 2462 MIN REE/OH/PO42474 SH 2480 SH REE/OH/PO42499 MIN 2499 SH 2487 MIN ‐ s t 2493 SH REE/OH/PO4Gunter Quarry Serra Verde Probable OriginNovo C Novo J79 as a number of other more subtle shoulders and local absorption minima. Cluster 4 is a single absorption band at 872 nm. Cluster 5 is characterized by a prominent absorption at 914 nm with a series of 2 or more weaker absorption bands up to ~965 nm including a consistent but noisy pair of absorptions at 951 and 953 nm. Cluster 6 is centered on the strong absorption at 978 nm and in some samples is expressed as a doublet. Differences in the VNIR spectra across the 2 groups of xenotime samples are typically slight, such as the presence or absence of a shoulder feature or the relative depth of a specific absorption. The Novo Horizonte samples show stronger absorptions at 643, 690, 760, 827, 932 and 961 nm while the Gunter Quarry and Serra Verde samples show stronger absorptions at 669, 678, and 803 nm. These two groups hold true for other unpublished xenotime VNIR spectra that show similar spectral patterns but that do not have supporting microprobe data. SWIR For xenotime the SWIR spectral range has been divided into 8 clusters beyond 978 nm, where Cluster 6 is better described in the VNIR range (Figure 3.7, Tables 3.6 and 3.7). Cluster 7 comprises a broad absorption centered at 1143 nm with shoulders at 1099, 1118 and 1193 nm for the Gunter Quarry and Serra Verde samples while the Novo Horizonte samples have doublet minima at 1099 and 1112 nm, a shoulder at 1143 nm and another absorption minimum at 1187 nm. Cluster 8 is a series of absorption bands with doublet minima at 1276 and 1301 to 1307 nm, and with shoulders on either side. The Serra Verde and Gunter Quarry xenotime samples also exhibit subdued minima at 1364 and 1358 nm, respectively. Cluster 9 is a strong band at 1408 nm for the Novo Horizonte samples while the Serra Verde and Gunter Quarry samples show a shallower absorption band located at 1415 nm. Cluster 10 comprises a deep absorption at 1503 nm with subsequent lesser minima and shoulders near 1440, 1528, 1559, and 1585 nm. Cluster 11 similarly has a first strong absorption at 1704 nm with subsequent lesser minima and shoulders near 1616, 1723, 1742, 1761, and 1805 nm. Cluster 12 is an absorption band at 1880 nm, and the Novo Horizonte samples also show a shoulder at 1848 nm while the Gunter Quarry sample is lacking the 1880 nm feature altogether. Cluster 13 is a consistent series of 6 resolved absorptions located at 1936, 1961, 1980, 2005, 2030 and 2043 nm. The absorptions at 1936 and 2005 nm are typically the strongest. For the Gunter Quarry and Serra Verde samples, Cluster 14 comprises two main absorption minima at 2218 nm and 2462 nm (2487 nm for Gunter Quarry). Several shoulders and minima exist intermediate to these more prominent absorptions. For the Novo Horizonte samples, 5 distinct absorption minima are located at 2212, 2262, 2313, 2399 and 2462 nm. 80 Differences in the SWIR spectra across the 2 groups of xenotime samples are listed above throughout the cluster descriptions, but the most prominent variations are in clusters 7, 8, 9, 12 and 14. 3.8.3 Britholite reflectance spectra The two britholite samples have very similar REE distributions (Figure 3.4), however, their bulk compositions differ. The CMNOC F90-8 sample shows totals closer to 100% and better cation populations after formula normalization by anion content. This sample shows lower La and Ce content, equivalent Pr through Lu and higher Y, as well as higher P, Si, Th, Ca and OH contents. Figure 3.8 Spectra of characterized britholite samples in VNIR (left, 500 to 1000 nm) and SWIR (right, 950 to 2530 nm) with absorption band clusters shown and prominent absorption lines of the CMNOC sample. Upper spectrum (blue) is from 001-Mariano sample and lower (green) is from CMNOC F90-8 sample. 81 Table 3.8 Prominent absorption features of britholite samples in the VNIR and SWIR ranges. VNIR For the 2 britholite samples, the VNIR spectral range has been divided into 6 main clusters of absorptions (Figure 3.8, Table 3.8). Cluster 1 comprises absorptions broadly centered at 584 nm, but is resolved into two bands at 579 nm and 591 in the “001” sample. Cluster 2 comprises 4 noticeable but weak shoulders at 621 and 634 nm and absorptions at 650 and 684 nm. Cluster 3 is a single strong and broad absorption at 749 nm with a slight shoulder at 743 nm. Similarly, Cluster 4 is another strong broad absorption located at 805 nm with slight shoulders at 797 and 799 nm. Cluster 5 comprises a strong absorption at 880 nm and a local absorption minimum at 867 nm. Finally, Cluster 6 is a sharp and strong absorption at 978 nm. BritholiteClusters wavelength shape wavelength shape1 579 MIN Nd584 MIN ‐ b, n Nd592 MIN Nd, Pr2 621 SH ‐ w 621 SH ‐ n Nd634 SH 635 SH ‐ n Nd650 MIN 653 MIN ‐ n Er, Ho684 MIN 685 MIN ‐ n Nd3 743 SH 743 SH Nd>Dy749 MIN ‐ s t, b 747 MIN ‐ s t, b Nd>Dy797 SH 797 SH Nd>Dy, Er799 SH 799 SH Nd>Dy, Er4 805 MIN ‐ s t, b 805 MIN ‐ s t, b Nd>Dy, Er5 867 MIN 867 MIN Nd880 MIN ‐ s t 879 MIN ‐ s t Nd6 978 MIN ‐ s t 978 MIN ‐ s t Yb>Er7 1074 MIN ‐ s t 1074 MIN ‐ s t Sm1105 SH 1105 SH Sm8 1225 MIN ‐ s t 1225 MIN ‐ s t Sm1257 SH 1257 SH Sm > Dy1288 SH 1288 SH Dy9 1377 MIN 1377 MIN Sm, Dy1415 SH 1415 SH H2O1478 SH 1478 SH Pr1528 MIN ‐ s t 1528 MIN ‐ s t Sm > Er1559 SH Pr1591 SH 1591 SH Sm, Pr10 1723 MIN 1723 MIN Nd, Dy1936 MIN ‐ s t 1943 MIN ‐ s t Pr, Sm, H2O1986 SH Pr, Sm2018 SH 2018 SH Pr, Sm, Ho11 2168 MIN REE/OH/PO412 2312 SH 2318 MIN REE/OH/PO42343 MIN REE/OH/PO42387 SH 2387 MIN REE/OH/PO42405 SH REE/OH/PO42437 SH REE/OH/PO42449 MIN ‐ s t 2462 MIN REE/OH/PO42480 SH 2480 MIN REE/OH/PO4Kipawa_001 Kipawa_CMNOC Probable Origin82 The only significant difference between the two spectra is how Cluster 1 is resolved into two features in sample “001”, which generally shows stronger absorptions elsewhere. SWIR For britholite, the SWIR spectral range has been divided into 6 clusters beyond the 978 nm absorption, which is ascribed to Cluster 6 in the VNIR range (Figure 3.8, Table 3.8). Cluster 7 is dominated by a strong minimum at 1074 nm with a weak shoulder at 1105 nm. Cluster 8 is a prominent absorption at 1225 nm with a noticeable shoulder at 1257 nm. Cluster 9 comprises a small absorption minimum at 1377 nm followed by a series of shoulders and band minimum at 1528 nm. Cluster 10 is dominated by a distinct absorption at 1723 nm and a broader deep absorption near 1936 nm, as well as numerous very subtle shoulders from 1880 through to 2043 nm. The 2168 nm absorption minimum of Cluster 11 is only evident on the CMNOC sample. Cluster 12 is more variable - the CMNOC sample shows a broader series of shoulders and minima with deepest bands at 2343 and 2480 nm, while the “001” sample has a single deepest minimum located at 2449 nm. Intramineral variations for the two britholite samples in the SWIR range are relatively minimal, however, the absorptions of Clusters 11 and 12 show variations. 3.9 Discussion on spectra and spectral variations The typical distribution of rare earth elements in monazite (LREE), xenotime (HREE) and britholite (LREE and HREE) directly affects their spectral signatures. The most important LREE that are spectrally active in the VNIR-SWIR range are Pr3+, Nd3+ and Sm3+. The most important spectrally active HREE are Dy3+, Er3+ and Yb3+ but Tb3+, Ho3+ and Tm3+ still need to be considered. Therefore, typical REE distribution for a given mineral means that it will also display a typical distribution of spectrally active REE3+. This allows for quick recognition of a mineral’s bulk REE distribution but more subtle spectral features then need to be considered when addressing mineral identification. Accordingly, the most notable differences between the mineral samples are driven by the relative distributions of light versus heavy rare earth elements. Monazite is LREE enriched and thus its spectral signature is driven primarily by Nd3+ with lesser influence by Pr+3 and Sm+3, while HREE-enriched xenotime has spectral features driven primarily by Yb+3, Er+3, and Dy+3. Britholite’s spectrum is dominated by Nd+3 with lesser influence by Pr+3 and Sm+3 but signals of Dy+3, Er+3 and Yb+3 are observed. Consequently, britholite spectra more closely resemble spectra of monazite than xenotime and a good understanding of britholite’s spectral characteristics can be derived by ‘modifying’ monazite spectra with minor input from Dy+3, Er+3 and Yb+3. 83 Figures 3.9 and 3.10 show representative spectra for xenotime, britholite and monazite in the VNIR and SWIR ranges with all absorption clusters and distinct absorptions labeled. Samples “UofA Unk SE” for monazite, “Gunter Quarry” for xenotime, and “Kipawa CMNOC F90-8” for britholite are used as the baselines for describing intermineral variations. The monazite sample shows the nearest chemical composition to ideal monazite, the xenotime sample represents one of the two spectral-chemical populations and the britholite sample shows the highest total from microprobe data and better atomic site assignments. 84 Figure 3.9 Representative spectra for xenotime (X, top, red), britholite (B, middle, green) and monazite (M, bottom, black) samples in the VNIR range (500 to 1000 nm). Absorption band clusters are shown with bars and prominent absorptions are labeled according to text descriptions. Figure 3.10 Representative spectra for xenotime (X, top, red), britholite (B, middle, green) and monazite (M, bottom, black) samples in the SWIR range (950 to 2530 nm). Absorption band clusters are shown with bars and prominent absorptions are labeled according to text descriptions. 85 3.9.1 Absorption patterns between monazite and britholite When comparing spectra of monazite to britholite, which have similar LREE concentration patterns, trends of shifting band minima and band broadening emerge. Many absorption minima related to Nd3+ undergo systematic shifts to longer wavelengths in britholite as compared to monazite. Shifted bands (from monazite to britholite) in the VNIR include 579/581 nm to 584 nm, 679 nm to 685 nm, 745 nm to 747 nm, 800 nm to 805 nm, 863 nm to 867 nm and 871 nm to 879 nm. The absorptions for britholite typically show less structure, most likely attributable to multiple REE sites and therefore multiple overlapping bands in close proximity. For example, the fine structure of the Nd3+-related 745 nm absorption of monazite is lost in britholite except for a weak shoulder at 743 nm. The many local minima from 871 to ~900 nm in monazite are also lost in britholite, likely due to REE site multiplicity and resulting fine band shifts and overlap. The small amounts of spectrally active Dy3+ in britholite but not monazite would contribute numerous weak absorptions in Clusters 3 to 5, as seen in xenotime, which may also impact the apparent broadening of features ascribed predominantly to Nd3+. Turner et al. (2014) described the reflectance spectra of REE fluorocarbonates and noted that Sm3+ related features at 1093 nm and 1251 nm are quite sharp. In monazite, these two features are each consistently split into two discrete absorptions at 1074 and 1105 nm, and at 1232 and 1257 nm (Figure 3.10). The barycenter of the absorptions in monazite remains close to that of the REE fluorocarbonates. In britholite, concentrations of Sm are lower, however, the splitting appears greater near 1251 nm with a strong band minima at 1225 and shoulder at 1257 nm (Figure 3.10). Near 1093, the absorption in britholite is broader with a minimum at 1074 nm and a weak shoulder at 1105 nm. The weaker Sm3+ related absorption at 953 nm in bastnaesite is similarly split to 941 nm and 957 nm in monazite (Figure 3.9). In britholite, however, this absorption is not seen due to again lower Sm contents. 3.9.2 The Yb‐Er related absorption near 978 nm Britholite and xenotime show an absorption feature at 978 nm that is related to both Yb3+ (2F7/2 2F5/2) and Er3+ (4I5/2 4I11/2). This particular Yb3+ transition is known to be hypersensitive and also absorb more strongly than Er3+ at this wavelength (e.g., Zou and Toratani 1995, Strohhofer and Polman 2003). In britholite, the strength of the 978 nm absorption is equal or stronger than in xenotime (Figure 3.11), however, the concentrations of Er and Yb in britholite are lower by a factor of ~4 (Tables 3.3 and 3.4, and Figure 3.4). Thus, the absorption coefficient for this particular Yb3+ hypersensitive transition is much stronger in britholite than in xenotime. This is most probably due to the REE site in britholite being more asymmetric than in xenotime, as well as being better suited to host REE3+ cations with larger ionic radii, such as Ce3+, rather than the smaller Yb3+. 86 Figure 3.11 Reflectance spectra near 978 nm for samples showing absorptions at this wavelength. Spectra are labeled and Er and Yb contents are given in parentheses (Er2O3 wt.% / Yb2O3 wt.%). 3.9.3 Complexity of xenotime’s absorption bands Spectral differences between xenotime samples are minor, usually only a variation in the strength of a feature. Unlike LREE minerals dominated by Pr3+, Nd3+ and Sm3+ spectral features, minerals with abundant HREE are harder to definitively assign individual features to individual lanthanides because many of the multiplet levels from Sm3+, Tb3+, Dy3+, Ho3+, Er3+, Tm3+ and Yb3+ overlap. Some of the multiplet levels can produce many absorptions of varying strength as deduced through mono-lanthanide doping studies of various compounds. For example, a study by Gruber et al. (2009) investigated energy levels in Ho-doped yttrium gallium garnet, which has a total of 252 theoretical Stark multiplet levels, leading to over 900 electronic transitions between 320 nm and 1950 nm. The single 5I7 multiplet of Ho3+ ranges from 1745 to 2145 nm in Ho3+ doped Spectralon and comprises 15 discrete bands. In Ho3+ doped yttrium gallium garnet (YGG) this same multiplet stretches only from 1825 to 1975 nm (Gruber et al. 2009) and relative absorption strengths for the 15 bands also differ. In Ho3+:YGG the strongest absorptions are on the flanks of the 87 multiplet while in Ho3+:Spectralon the central bands are strongest, as shown in the figure below (Figure 3.12). These differences are driven by crystallographic site characteristics for the lanthanides. Figure 3.12 The 5I7 multiplet absorption characteristics in Ho3+ doped Spectralon (left, in reflectance) and Ho3+ doped YGG (right, transmission, plot from Gruber et al. 2009). Note the differences in wavelength ranges and overall absorption strength patterns. Notable spectral differences, however, include the Serra Verde sample which shows a number of small minima and shoulders in the VNIR (Table 3.6, Figure 3.7) that can be attributable to Nd3+, as this sample contains the highest Nd of the xenotime suite. Cluster 7 (around 1100 nm) morphology in the SWIR range (Table 3.7, Figure 3.7) is slightly different due to the Novo Horizonte samples having higher Dy. Uranium also has absorptions in this region, and is highest in the Gunter Quarry and Serra Verde samples. Cluster 12 (around 1880 nm) of xenotime has a distinct absorption band located at 1880 nm attributable to Tb3+, being strongest in the Novo Horizonte samples (~1.5 wt.% Tb2O3) and absent in the Gunter Quarry sample (0.33 wt.% Tb2O3). Although in low concentration, the main expressions of Tm3+ content exist at 781, 1213 and 1616 nm in xenotime. 3.9.4 Hydroxyl and water bands of the REE phosphate minerals Xenotime and monazite can accommodate minor amounts of OH into their crystal structures, trap fluid inclusions during growth, and become metamict through radiation damage related to decay of U and Th which can then lead to hydration (e.g., Talla et al. 2011). The xenotime samples studied contained low amounts of charge balancing anions and were not considered to be appreciably metamict. Monazite showed generally higher calculated OH content and also considerably more Th. Britholite normally accommodates OH- and some samples from Kipawa studied by others have been shown to be partially 88 metamict. Consequently, the monazite and britholite samples show absorption bands at 1377 nm that are related to OH and near ~1950 nm that are related to water and REE. Xenotime spectra show bands from 1408 to 1415 nm that are related to an OH- overtone, which is consistent with Talla et al. (2011) who identified stretching vibrations near 3500 cm-1 in xenotime. The bands near 1936 nm are more likely tied Ho3+ and Tb3+ content but may also be influenced by water. 3.9.5 Spectral effects of the actinides Electronic absorption bands for the 5f energy levels of U4+, 5+ are strongest near ~2069, ~1500, ~1330 and ~1111 nm in the SWIR and ~666 nm in the VNIR (e.g., Binnemans et al. 1999, Zhang et al. 2002). All of these bands are in ranges of various other REE3+ absorptions. Xenotime from Serra Verde contains the most uranium of all the samples studied with 1.59 wt.% UO2 while those from Novo Horizonte have ~0.05 wt.% UO2. The most significant difference possibly attributable to uranium was near 1111 nm where the 1143 nm feature was fairly pronounced in the Serra Verde and Gunter Quarry xenotime samples (Figure 3.7), however, this absorption could also be attributable to Ho3+ and Dy3+ in xenotime. 3.9.6 Comparison to USGS Spectral Library Spectra The USGS Spectral Library Version 06a (Clark et al. 2007) contains spectra for monazite, chlorapatite, fluorapatite and hydroxylapatite (Figure 3.13). The sample descriptions contain basic descriptions of XRD results suggesting minor impurities for each spectrum but there is no complete chemical information. 89 Figure 3.13 Reflectance spectra from the USGS Spectral Library (Version 06a). From uppermost to lowermost near 2500 nm: Chlorapatite (black), fluorapatite (red), hydroxyl-apatite (pink) and monazite (blue). For monazite, all the main spectral features are present and band minima are at approximately comparable wavelengths to spectra in this study. Slight differences arise due to lower spectral resolution of the library’s data. Hydroxylapatite shows no absorptions related to REE content, but does show an OH related absorption with a minimum at 1925 nm and a single broad absorption from ~2300 out past the end of the spectrum, 2530 nm. The weak REE related absorptions in fluorapatite and chlorapatite are consistent between each other and can all be attributed to Nd3+. Definitive minima occur at 577, 738, 748, and 807 nm, slight absorptions near 883 nm, and both samples show a broad noisy absorption near 1960 nm. These minima locations of chlorapatite and fluorapatite are roughly equivalent to what is seen in monazite spectra from both the USGS Library and our studies, but band shifts might be present near 807 nm as compared to monazite. The chlorapatite sample shows several other noisy and broad absorptions between 900 and 1700 nm, however, none of these appear to correspond with likely REE absorptions. A fairly sharp absorption exists at 2315 nm for the chlorapatite sample, however, its origin is unclear. Reflectance is fairly strong for the fluorapatite and chlorapatite samples out to ~2500 nm. The lower spectral resolution USGS data compared to spectra from the sisuROCK instrument combined with the lack of chemical data in the library make it difficult to justify drawing conclusions 90 based on spectra from the three phosphate samples with REE3+related absorptions. However, that data does show at least rough consistency of spectral response in fluorapatite and chlorapatite to monazite and suggests variances in band positions for Nd3+. Dedicated research on REE-bearing apatite is warranted. 3.9.7 Absorptions from 2150 nm to 2530 nm The REE phosphate minerals display a number of overlapping absorptions expressed as minima and shoulders between 2150 nm (~4650 cm-1) and 2530 nm (~3953 cm-1). As a general statement, each mineral shows moderate consistency between the samples, but significant differences are observed in the two classes of xenotime samples (Figures 3.6, 3.7 and 3.8). The exact origin of these absorption bands is unclear. In their study of hydrated and hydroxylated phosphate minerals (e.g., childrenite – FeAlPO4(OH)2*H2O), Lane et al. (2011) show reflectance spectra with sharp absorptions in the 2100 to 2500 nm range, which they attribute to overtones and combinations of OH and PO4 vibrations. In a similarly themed study of hydrated hydroxylated phosphate minerals (e.g., wardite – NaAl3(PO4)2(OH)4*2H2O), Frost and Erickson (2005) ascribed bands from ~4600 to 4000 cm-1 (2174 to 2500 nm) to combinations of OH stretching and OH deformational vibrations. Assaaoudi et al. (2001) and Onac et al. (2005) reported that amongst different hydrated orthophosphates (e.g., churchite, YPO4*2H2O), small band shifts of the fundamental vibrations for P-O, O-H and H-O-H are potentially related to variable REE content. Talla et al. (2011) studied OH defects in xenotime and identified an OH stretching mode near 3500 cm-1. Combining that stretching with a fundamental PO4 vibration near 1000 cm-1 (e.g., Farmer 1974) results in a potential combination band near 4500 cm-1 (2222 nm). Small absorptions near 4500 cm-1 can been seen in one IR spectral plot of xenotime from Talla et al. (2011), but are not attended to in the article’s scope. Pekov et al. (2007) recorded overlapping Si-O and P-O bands for metamict fluorcalciobritholite at 930 and 1070 cm-1, respectively, and Oberti et al. (2001) recorded a broad OH stretching band at 3437 cm-1 in fluorbritholite but no spectrum is provided. Thus, theoretical calculations of combination bands for the REE phosphate minerals support the notion of their influence in this region between 2150 and 2530 nm, however, neither xenotime nor monazite have structural OH or H2O in any great amounts and reflectance spectra consistently show more than one resolved band. Trivalent lanthanides with electronic energy levels in the ~2150 to 2500 nm range (4650 cm-1 to 3953 cm-1) include Pr3+ (~2310 nm, 3H6) and Tb3+ (~2200 nm, 7F3) with potential influence from Nd3+ (~2500 nm, 4I13/2) and Eu3+ (~2100 nm, 7F6) when placed in a crystal field, according to energy levels recorded by Dieke et al. (1968) and Carnall et al. (1989). Reflectance spectra of REE2O3 reagents and the little reflectance and transmission data available from the literature in this range suggest that REE-related 91 absorptions could be responsible for the patterns seen in the REE phosphates. In particular, Talla et al. (2011) synthesized REE-doped xenotime with Pr, Nd, Sm, Dy, Ho and Er and collected polarized IR absorption spectra from 7000 to 2500 cm-1 (1429 – 4000 nm). Of these synthetic xenotime crystals, the Nd-doped sample exhibits 2 (possibly 3) bands roughly at 2400 and 2475 nm (the third near 2500 nm) and the Pr-doped sample exhibits only one band near 2375 nm. Unfortunately neither Eu2O3 nor Tb2O3 were included in their EMPA results and they did not synthesize Eu- and Tb-doped xenotime. It is unlikely that Eu would play a significant role due to relatively low contents, however, Tb2O3 concentrations in xenotime can be appreciable. Focusing just on xenotime, the two samples from Novo Horizonte have contain the highest Tb2O3 content and show the strongest absorptions near 2262 nm and 2312 nm, suggesting that at least these two bands are linked to Tb3+. Looking back at the one natural sample shown by Talla et al. (2011) in this range, two more small absorptions can be seen near 2247 and 2325 nm. Thus, with the spectra from Talla et al. (2011) but without spectra from Tb-doped xenotime it is still difficult to conclusively tie intramineral chemical variations to spectral variations. A spectrum from reagent grade La powder (qualitatively, La(OH)3 >> La2O2(CO3), see Appendix D) shows absorption bands in the 2300 to 2500 nm range (in addition to water related absorptions at ~1420 nm and ~1950) but La3+ should not show 4f-4f electronic transitions as it has no electrons in the f orbital. Conversely, reagent grade Gd2O3, Y2O3 and Yb2O3 powders show minor absorptions near 1420 and 1950 nm but no distinct absorptions between 2300 and 2500 nm. Consequently, it is apparent that at least some of the REE reagents are generating absorptions related to REE-OH bonding (with possible CO3 absorptions), thus making interpretations on other REE-related absorptions in this region difficult. If absorptions in this region were solely attributable to PO4 / OH vibrational combinations we would expect more coherency between the PO4 minerals, less variability in the xenotime spectra, and lesser absorptions where OH is minimal (ie, xenotime and monazite vs. britholite). If they were solely the result of REE3+ electronic transitions we would expect greater support from existing REE spectroscopy literature and not observe absorptions in La2O3. If they were solely the result of REE-OH vibrational bands we would expect to see more systematic variations within and across the phosphate groups. It is therefore most likely that each of these plays a role in absorptions from 2150 to 2530 nm, but without further investigations using additional techniques it is difficult to determine the exact origins of these absorption bands across monazite, britholite and xenotime. 92 3.10 Implications This work is the first published systematic study of the REE phosphate minerals monazite, xenotime and britholite using reflectance spectroscopy in the VNIR-SWIR range. Spectra coupled with microprobe data indicate that spectra of monazite and britholite are dominated by Nd3+, Pr3+, and Sm3+. Britholite’s multiple sites with mixed coordination for Ln3+ allows for easy incorporation of Tb, Dy, Ho, Er, Tm and Yb. The mixed coordination of cation sites REE1/REE1a (9-coordinated) and REE2 (8-coordinated) results in multiple sets of crystal field splitting for the Ln3+, thus potentially broadening each absorption band related to electronic transitions. Xenotime’s spectrum is dominated by the distribution of HREE, namely Tb3+, Dy3+, Ho3+, Er3+, Tm3+ and Yb3+. Many of the energy levels for these elements overlap, making definitive assignments of probable origin difficult; however, absorption clusters are consistent across samples. The sensitive nature of the Yb3+ and Er3+ transitions near 978 nm are well illustrated between xenotime and britholite. Despite xenotime having significantly higher concentrations of Yb and Er, britholite exhibits an equal absorption band at this wavelength because of greater asymmetry of the coordination polyhedra as compared to xenotime. This data provides the diverse geological remote sensing community with a benchmark to which comparison can be made regarding spectra with Ln3+ related electronic absorptions. Diagnostic variations between the REE phosphates can also be exploited in a number of circumstances, such as remote predictive mapping, petrology of thin sections and rocks, identifying potential mineral grains for geochronology, drill core and chip logging, mine wall imaging or ore sorting. Future work should include a systematic study of mono-lanthanide doped phosphate crystals using reflectance spectroscopy and FTIR spectroscopy, as well as other natural REE phosphate minerals with variable REE content. This will enhance the growing database of REE mineral reflectance spectra and help clarify ambiguity for assigning absorptions to specific lanthanides across the entire VNIR-SWIR range, but especially from 2150 to 2530 nm. 93 Chapter 4. Visible to Short Wave Infrared Reflectance Spectroscopy of REE‐Bearing Silicates 4.1 Chapter summary Reflectance spectroscopy in the visible to short wave infrared regions (500 nm to 2500 nm) was carried out on natural samples of the rare earth element (REE) bearing silicate minerals cerite, mosandrite, kainosite, zircon and eudialyte. Samples were characterized by scanning electron microscopy and electron microprobe analysis. Absorption band positions were recorded with their probable origins and spectral variability amongst the samples is discussed. Spectral features of these minerals are driven primarily by 4f-4f intraconfigurational electronic transitions of trivalent lanthanides, as well as 5f-5f electronic transitions of uranium and vibrational overtones and combinations of H2O and OH-. Spectra of eudialyte are also impacted by the relative amounts of IVFe2+ and VFe2+. The respective spectra of these REE bearing silicates are sufficiently distinct to enable spectral classification, and a compilation diagram with representative spectra is given. Spectral variability of some specific REE-related absorptions, such as an Er3+ and Yb3+ related absorption near 978 nm and Nd3+ related absorptions near ~746 nm, ~803 nm and 875 nm, are interpreted to be driven by differences in the cation sites between the crystal structures. The work presented here sets the foundation for the interpretation of reflectance spectra of these silicates and enables exploitation of the observed features by the remote sensing community for detection, identification and quantification of REE-bearing silicate minerals. This is especially relevant for hyperspectral imaging with high spatial resolutions where the spectral response of a pixel becomes increasingly dominated by mineralogy rather than lithology. 4.2 Introduction Rare earth element bearing silicates show a large diversity of overall crystal structures, chemical compositions and host sites for the lanthanides. They can show strong enrichment in light rare earth elements (LREE, e.g., cerite), heavy rare earth elements (HREE, e.g., kainosite), or display relatively elevated values of all REE (e.g., mosandrite). In these minerals the REE can form specific structural components (e.g., kainosite), be important constituents across multiple sites (e.g., eudialyte), or exist as trace to minor elements (e.g., zircon). The suite of minerals studied here (cerite, mosandrite, kainosite, zircon and eudialyte) covers a wide breadth of variability but by no means is entirely comprehensive. The REE-bearing silicates can be locally abundant and contain high amounts of REE but have been traditionally viewed in a negative light with respect to their economic significance. Recent mineral 94 exploration and metallurgical developments, however, are proving that some silicate phases are amenable to beneficiation (Mariano and Mariano 2012). Despite this, most REE-bearing silicate phases will still not prove to be worthwhile exploration targets when compared to other REE mineral sources, and their existence in a given ore deposit could be detrimental to beneficiation. Understanding differences in the spectral responses of REE bearing silicates is important if reflectance spectroscopy is to be used in the exploration and exploitation of these commodities. The use of REE bearing silicates as geochronometers, especially zircon, provides an additional motivation for understanding spectral responses of these minerals. Hancock et al. (2012) suggest that recognizing spectra of zircon in large spectral databases, such as hyperspectral core logs, could facilitate petrological studies by identifying areas with suitable (i.e., non-metamict) zircon. Reflectance spectroscopy, a rapid non-destructive analytical technique requiring little sample preparation, has been used to study the reflectance of these minerals in the visible to short wave infrared regions. This research builds on the study by Turner et al. (2014) that documented the reflectance spectra of the REE fluorocarbonate minerals bastnaesite, parisite and synchysite. As in that study, mineralogical and spectroscopic background of REE bearing minerals is provided, followed by a band registry for these REE bearing silicate minerals as well as interpretations of spectral absorption features related to the lanthanides. The spectral features of the registry will be the focus of further study in the development of hyperspectral imaging to carry out REE mineral identification, REE mineral abundance estimates and rare earth element abundance estimates. The work presented here sets the foundation for the interpretation of reflectance spectra of these REE bearing silicates. 4.3 General spectroscopy of the lanthanides The outer radius of the 4f electron shells (~0.3 Å) for the lanthanides is much less than that of their filled 5s and 5p shells (~2 Å, ~1 Å). It can then be approximated that the local electronic environment of Ln3+ cations interacts primarily with those outer shells, leaving the 4f electrons ‘relatively sheltered’ but not completely non-participatory in bonding (e.g., Liu, 2005). Electrostatic repulsion of the base ion generates a first level splitting of spectroscopic states, 2S+1L (e.g., 5I). Next, spin-orbit coupling splits these into multiplets, or “J-levels” (e.g., 5I8), and once placed into a crystal field the J-levels are then split into “Stark Sublevels”. Crystal field interactions for the Ln cation include variables such as ligand type, coordination number and polyhedron asymmetry which all play a role in the location and intensity of energy levels and the associated absorption (Görller-Walrand and Binnemans, 1998). Each of the resulting sublevels provides the potential for promotion of a relaxed electron into an excited state, giving rise to absorption of electromagnetic radiation (e.g., light) at a specific energy level (e.g., wavelength). 95 The spectroscopy of REE-bearing phases is well established in the fields of physics and chemistry, however, the well understood principles and spectral characteristics of well-studied doped compounds do not lend to direct translation into natural minerals and hyperspectral remote sensing. For example, the Dieke Diagram (Dieke et al. 1968) details intraconfigurational 4f-4f transitions for ‘free ions’ as deduced through studies of largely mono-lanthanide synthetic compounds, however, the transitions shown do not include splitting of energy levels due to a crystal field nor the complexities of naturally occurring REE minerals with variable REE distributions and other elemental substitutions. Consequently, this diagram and other band registries can only act as proxies to help identify origins of absorption features in reflectance spectra from minerals. In general, and excluding other physical parameters such as grain size, the strength of absorption features by the lanthanides will be primarily a function of the concentration of the ion as well as the specific absorptivity of that ion’s intraconfigurational transitions within a given crystal structure. The location of lanthanide-related absorption features will be primarily a function of the cation’s specific coordination and asymmetry in the host crystal structure. 4.4 Review of REE‐related reflectance spectroscopy studies The benchmark mineral spectroscopy paper of Hunt (1977) does not include any REE bearing silicates, but does list monazite (misspelled as monzonite) in his tabulation of common minerals and respective spectral signatures. However, La2+ is listed as the origin of absorptions but La would have 3+ valence in monazite and would therefore not have spectral features in the VNIR-SWIR (e.g., Liu 2005). Similarly, Clark (1999) does not attend to REE bearing silicates but does include several REE oxides in his Spectroscopy of Rocks and Minerals review. It is recognized there that the patterns seen in REE minerals are a combination of several lanthanides but it is stated that absorption positions are independent of mineralogy. The widely used USGS spectral library (Version 06, Clark et al. 2007) contains one spectrum for zircon from Brazil and one spectrum for metamict allanite from Ontario. Version 2 of the ASTER Library (Baldridge et al. 2009) contains one spectrum of zircon from Malawi. No chemical data are available for these samples. Kerr et al. (2011) present 5 excellent spectra of eudialyte from the Red Wine Intrusive Suite that show evidence of Ln3+ absorption bands but provide no chemical data or interpretation of the spectra (Figure 4.1). Hancock et al. (2012) presented on the potential of reflectance spectroscopy to identify zircon in drill core and hand samples using U-related absorptions. Abstracts by Swayze et al. (2013) and Hoefen et al. (2013) indicate that baseline and applied research is being conducted on REE minerals and deposits, however, no detailed information is available. Turner et al. (2014) described the 96 reflectance spectroscopy of the REE fluorocarbonates bastnaesite, parisite and synchysite. In the context of instrument calibrations various authors, such as Weidner et al. (1984, 1985, 1986), Allen (2007) and Mann et al. (2014), have characterized Ho3+, Dy3+ and Er3+-based wavelength standards. Thus, there remains an information gap for the interpretation of reflectance spectra from REE-bearing silicates, with respect to identifying and explaining spectral features as well as in supporting the discrimination of minerals based on their spectral characteristics. Figure 4.1 Eudialyte spectra from Red Wine Complex (Kerr et al. 2011). Considerable research has been completed on the optical and infrared spectroscopy of zircon and eudialyte, but comparatively little has been written about kainosite, cerite or mosandrite. The abundance of data relating to zircon and synthetic ZrSiO4 is due to its appearance and usefulness in diverse geological settings as well as its importance in a variety of industrial applications. Eudialyte, on the other hand, has received extra attention because of its striking pink-red colour and challenging crystal structure and crystal chemistry. Accordingly, brief reviews are warranted of factors specific to eudialyte and zircon that affect their reflectance spectra but that are not Ln3+-related 4f-4f electronic transitions. 97 4.4.1 Spectroscopy of eudialyte Polshin et al. (1991), Burns (1993) and Rossman and Taran (2001) discuss the particular four and five-fold coordination of Fe2+ in eudialyte at the M2 site. In four fold coordination, IVFe2+ assumes a square planar arrangement with O2- that leads to absorptions centered near 529 nm (18900 cm-1) and 1366 nm (7320 cm-1). Absorption bands from iron in this arrangement show weak pleochroism. In five-fold coordination, VFe2+ assumes a pyramidal arrangement with four O2- and one OH- at the apex of the pyramid, leading to absorptions centered near 917 nm (10900 cm-1) and 2500 nm (4000 cm-1). The result is that eudialyte with predominantly IVFe2+ shows pink to crimson red colour and is optically positive whereas eudialyte with VFe2+ results in brownish hues and is optically negative. These yellow-brown or red-brown optically negative varieties are also known by the unofficial name eucolite. Eudialyte is also host to Fe3+ but typically in lesser amounts. Manganese in eudialyte is assumed to be divalent and in either 5-fold (M2 site) or 6-fold (M1 site) coordination, according to Johnsen and Grice (1999) and Johnsen et al. (2003). Burns (1993) lists five main bands for Mn2+, located near 535, 440, 405, 355 and 345 nm depending on the mineral host. No systematic studies of REE related spectroscopy have been carried out on eudialyte. 4.4.2 Spectroscopy of zircon The spectroscopy of zircon has been studied in detail with a variety of techniques due to its importance in geochronology and its potential as a storage phase for nuclear waste, as well as other applications of materials with zircon-type structures (e.g., Zhang et al. 2002, 2003, Nasdala et al. 2003, Kempe et al. 2010). Richman et al. (1967) and Fielding (1970) provided early studies of the absorption spectrum of zircon in the visible and infrared. Vance and Mackey (1978) established energy levels for U4+ and U5+ in zircon, hafnon and thorite. Their strongest absorption lines for U4+ were located at 1682, 1119, 653, 636 and 537 nm. They also observed the U5+ absorption bands at 1107 and 1492 nm. Krupa and Carnall (1993) studied the energy levels of U in ThSiO4 in greater detail, providing a more thorough evaluation of specific levels but in a different host lattice. Kempe et al. (2010) investigated a suite of zircon samples using a variety of spectroscopic techniques, including optical absorption. Their sample from Mt Malosa, which was reported as metamict, showed a few bands in the VNIR range (up to ~900 nm) that they related to Nd3+ and Er3+. Kempe et al. (2010) also found evidence for minor amounts of U6+ in zircon in the form of uranyl (UO22+) through the use of time-resolved laser-induced photoluminescence. Zhang et al. (2002, 2003) studied infrared absorption spectroscopy of select features related to Si-O, OH, U4+ and U5+ in zircon. Notably, they observed that: U5+ in a crystalline site shows absorption at 6668 cm-1 (1500 nm) and with continued radiation damage is replaced by (not shifted to) U5+ in an amorphous site at 6650 cm-1 (1504 nm) 98 U5+ in a crystalline site shows absorption at 9030 cm-1 (1107 nm) and with continued radiation damage is replaced by (not shifted to) U5+ in an amorphous site at 8969 cm-1 (1115 nm) U4+ in a crystalline site shows absorption at 4833 cm-1 (2069 nm), which broadens upon radiation damage and the location of maximum absorption shifts towards 4800 cm-1 (2083 nm) Other prominent U4+ crystalline absorption bands occur at 10922 (916 nm), 6779 (1475 nm), 6508 (1537 nm), 6022 (1661 nm), 5861 (1706 nm) and 4567 cm-1 (2190 nm) Using electron paramagnetic resonance techniques (EPR) Klinger et al. (2012) investigated the role that electron holes and crystal defects play in generating colour in zircon. Their crystals ranged from colourless through yellow-brown to red and they identified two principal electron hole related bands that are centered at ~340-350 and ~510-515 nm with full width at half maximum (FWHM) values of ~175 nm. These electron holes were found to be dependent on the existence of lanthanides in zircon. Also via EPR studies or zircon, Laruhin et al (2002) noted the conversion of Dy3+ to Dy4+ and of Tb3+ to Tb4+ from incident radiation, however, it was unclear as to what proportion of the Dy and Tb were being converted during their laboratory experiments and what could be expected in nature. Ebendorf-Heidepriem and Ehrt (2000) note that a ligand-metal charge transfer band for Tb4+ is located near 370 nm in phosphate glasses and Hansen et al. (1996) describe a 4f-5d transition for Tb4+ in zircon also in the UV range near 280 nm. Similarly, Kempe et al.(2010) conclude that charge transfer bands of Pr4+ and perhaps Tb4+ in the UV show tails that extend into the VIS range, impacting colour as in synthetic Pr:Zircon yellow pigments (e.g., Badenes et al. 2002). The ability of zircon to host various cations and to display a multitude of defect centres (e.g., Lees 2001) makes the details of optical spectroscopy difficult to unravel, however, the majority of the spectroscopic impact and subsequent research is within the UV and VIS ranges. 4.5 Crystal structure reviews The REE-bearing silicates are a diverse set of minerals with a wide range of coordination and overall crystal chemistry and structure. Coordination numbers range from 6 (gittinsite) to 11 (allanite), with some cases showing 8+1 coordination (cerite) and other minerals having more than one distinct REE site (mosandrite). Documented bond lengths vary accordingly, ranging from ~2.2 to ~3.2 Å. Zircon, ZrSiO4, although not an “REE Mineral” is a common carrier of moderate amounts of REE, along with Th, U, Hf and other high field strength elements (HFSE). Collectively, the articles by Finch et al. (2001), Hanchar et al. (2001) and Finch and Hanchar (2003) cover the incorporation of REE into zircon in great detail. Synthetic REE-doped zircon showed the coupled xenotime-substitution “REE3+ + P5+ = Zr4+ + Si4+” whereby REE enter the fairly symmetric 8-coordinated Zr site (Figure 4.2) and are 99 charge-balanced by P5+ replacing Si4+. Bond lengths for (Zr,REE)-O range from 2.130 to 2.278 Å, and total REO content usually only achieves ~5% with a preference for HREE+Y and Ce4+. Friis et al. (2010) studied the photoluminescence of REE doped zircon from ~250 to 600 nm and among their conclusions they showed that REE in zircon are being hosted in two ZrO8 polyhedra that are slightly non-equivalent. Finch and Hanchar (2003) also note that interstitial sites are a possibility for hosting smaller high field strength elements. Little work has been done on this topic. Kainosite-(Y), ideally Ca2Y2(SiO3)4(CO3)·H2O, is typically found in alkalic pegmatites and contains up to ~40% REO with a strong preference for yttrium and the HREE. The REE are found in a single fairly symmetric site with 8-fold coordination and bond lengths between 2.24 and 2.52 Å (Figure 2). Kainosite is host to a CO3 radical with an average bond length of 1.2526 Å, but is asymmetric because two of the three edges of the CO3 plane share edges with adjacent YO8 polyhedra (Rumanova et al.1967). Cerite-(Ce), ideally Ce9Fe3+(SiO4)6(SiO3OH)(OH,F), contains three distinct but similar REE sites that are all 9-coordinated with average bond lengths of ~2.59 Å, respectively, but individual bond lengths from 2.41 to 3.00 Å (Moore and Chen 1983, Pakhomovsky et al. 2002). Each of these 3 sites is bonded to 8 oxygen and an (OH,F) group with REE-OH distances of 2.55, 2.59 and 2.49 Å (Figure 4.2). Total REO is ~ 73% and although cerite has not been identified at many localities, it is present in significant volumes at type locations such as Bastnas and the Mountain Pass mine. The eudialyte group comprises at least 25 approved mineral species and the crystal chemistry of the group allows for up to ~10% REO (Johnsen and Grice 1999, Johnsen et al. 2003). The minerals of this group are Na-rich zirconosilicates with three and nine membered rings of SiO4. The general formula of the mineral group is: Na15[M(l)]6[M(2)]3Zr3[M(3)] (Si25O73) (O,OH,H2O)3X2, while eudialyte itself is Na15Ca6(Fe2+,Mn2+)3Zr3Si[Si25O73](O,OH,H2O)3(OH,Cl)2. The REE typically occupy the M1 site (REE-O ~2.30 Å) in place of Ca with distorted 6-fold coordination as well as within the cavity-like Na(3) and Na(4) sites. The cavernous nature of the Na sites, however, forces the cations to occupy whatever space results from the various substitutions in this flexible formula. Johnsen and Grice (1999) note that coordination of the Na sites can range from 6 to 10 and anions include oxygen and the contents of the X site, although the Na(4) typically assumes a distorted 9-fold coordination polyhedron. The coordination polyhedra seen in Figure 4.2 shows the Na(4) site with possible oxygen atoms (red) and mixed anion site (green). This figure (modified from Johnsen and Grice 1999) also shows the linkages to Si tetrahedra (magenta), the M2 site (grey polyhedra) and M3 site (blue octahedra). With respect to site assignment of REE in eudialyte group minerals when only chemical data is available, Johnsen et al. (2003) and Johnsen and Grice (1999) suggest that all Ca be assigned to the M(1) 100 site, and any remaining space be filled with Mn and subsequently with the REE. Remaining REE then is assigned to the Na(4) site. With respect to the objectives of this work, we can then say that high Ca samples will encourage REE to predominantly occupy the Na(4) site over the M(1) site. However, with the aid of single crystal structure refinement data, nearly all the analyses in Johnsen and Grice (1999) have REE in both sites irrespective of Ca content. Consequently, we should also expect this to be the case when interpreting absorptions due to the REE3+. Mosandrite is a historically poorly understood complex titanium silicate with REO contents up to ~25% that has been recently studied in detail by Sokolova and Camara (2008) and Bellezza et al. (2009), but with slightly different outcomes. According to Sokolova and Camara (2008) the ideal formula is Na2Ca4REETi(Si2O7)2OF3. Using their refinement, REE in mosandrite occupy two 7-coordinated sites with bond lengths from 2.403 to 2.674 Å (AP site, average=2.484 Å) and 2.353 to 2.749 Å (MH site, average=2.452 Å). The AP site is coordinated to 6 oxygen atoms and a single F whereas the MH site is coordinated with 6 oxygen atoms and a mixed occupancy anion site. Bellezza et al. (2009) gave a more flexible formula of Ti(□,Ca,Na)3(Ca,REE)4(Si2O7)2[H2O,OH,F]4*1H2O, which notably includes structural H2O. Their refinements suggest REE being in primarily the very similar 7-coordinated M4 and M5 sites, which are comparable to the two described by Sokolova and Camara (2008), as well as allowing small amounts of REE (~10% of total REE content) at a third 6-coordinated site (M3) with slightly shorter bond lengths (Figure 4.2). This M3 site is bonded to two oxygen and 4 mixed anion sites. The mixed anion site OF(1) is also coordinated to the M1 site, which is host to Ti4+. Later studies by Camara et al. (2011) investigated the structurally and chemically similar mineral rinkite, gave its formula also as Na2Ca4REETi(Si2O7)2OF3 and stated that mosandrite is different but needs to be more deeply investigated. Our spectrum for mosandrite favours the H2O-bearing structure suggested by Bellezza et al. (2009). 101 Table 4.1 Summary of REE site coordination polyhedra for the various REE bearing silicate minerals. Zircon Kainosite Cerite Eudialyte Mosandrite REE Site 1 Coordination 8 8 9 6 (M1) 7 (M4) Distortion Index 0.032 0.044 0.037 0.015 0.040 Volume (Å3) 19.13 23.62 33.71 15.53 20.30 Comments Replaces Zr, 8×O 8×O 8×O, 1×OH Ca>Mn>REE, 6×O 6×O, 1×OH REE Site 2 Coordination 9 9 (Na4) 7 (M5) Distortion Index 0.032 0.079 0.025 Volume (Å3) 32.85 40.27 21.26 Comments 8×O, 1×OH Can be 6 to 10 coordinated, variable anion bonding 6×O, 1×OH REE Site 3 Coordination 9 6 (M3) Distortion Index 0.051 0.023 Volume (Å3) 34.64 16.22 Comments 8×O, 1×OH 4×Mixed Anion, 2×O *Data for polyhedra from Rumanova et al. (1967), Johnsen and Grice (1999), Finch et al. (2001), Pakhomovsky et al. (2002), and Bellezza et al. (2009). Distortion index based on Baur (1974) via Momma and Izumi (2011) 102 Figure 4.2 Coordination polyhedra for Ln3+ in various REE bearing silicates (see text for details). Top row includes polyhedra for zircon (Zr site, coordinated with 8 oxygen), kainosite (REE site, coordinated with 8 oxygen), and cerite (3 distinct but similar sites coordinated with 8 oxygen and 1 OH-). Middle row includes the two sites in eudialyte, M1 (left, coordinated with 6 oxygen) and Na4 (~9 coordinated, see text for comments). Bottom row includes polyhedra for mosandrite, the similar M4 and M5 sites (7-coordinated to 6 oxygen and 1 OH-) and the M3 site (coordinated to 2 oxygen and 6 mixed anion sites). Red spheres are oxygen, grey and green spheres are mixed anion sites and larger multicoloured spheres inside polyhedra are Ln3+. 103 4.6 Experimental methods 4.6.1 Samples One cerite, one mosandrite, one kainosite, five zircon, and nine eudialyte samples from a larger suite were studied in detail and none showed significant compositional zoning using the scanning electron microscope. The zircon crystals originate from Green River (several crystals up to 0.5 cm long), Mudtank (crystal ~ 2 cm long) North Burgess (crystal ~1.5 cm long), Mt Malosa (rock with abundant clusters of small crystals ~1 mm long each), and St Peters Dome (single crystal 0.75 cm across in hand sample). The kainosite specimen was three small patches (~3 × 3 mm) on a small hand sample. Cerite was found in dark pink-brown to grey massive aggregates within a ~10 × 10 cm hand sample. Hand sample F92-23 from Kipawa contained numerous crystals of both green-brown mosandrite (up to ~3 cm long, 0.5 cm wide) and patches of pink to red eudialyte up to ~2 × 1 cm. Two other samples of eudialyte from Kipawa, “Mariano Suite” and “UofA” were coarse monomineralic hand samples up to 5 cm across. The remaining 6 eudialyte samples all originate from Mont St. Hilaire and comprise either euhedral single crystals or coarse polycrystalline aggregates from ~1 × 1 cm to ~3 × 3 cm. Of these, CMNOC476 and 2045 are red in hue while the rest are brown to brick red. Reagent-grade lanthanide oxide powders, REE-doped Spectralon wavelength calibration samples, and other REE mineral phases with EMPA data were also investigated in order to aid in band assignment. 4.6.2 Scanning electron microscopy and electron microprobe analysis The Philips XL30 scanning electron microscope (SEM) at the University of British Columbia, which is equipped with an energy-dispersion X-ray spectrometer (EDS), was used for preliminary examination of mineral mounts of selected minerals and rock fragments studied by reflectance spectroscopy. Samples were then analyzed by electron microprobe at the Saskatchewan Research Council’s Advanced Microanalysis Centre using a Cameca SX-100 equipped with 5 tunable wavelength dispersive spectrometers. Operating conditions were: 40° takeoff angle, beam energy of 15 keV, beam current of 20 nA, beam diameter of 5 μm. The MAN background intensity data was calibrated and continuum absorption corrected. Elements were acquired using analyzing crystals LLIF for FeKα, TaLα, PrLα, EuLα, DyLα, TmLα, MnKα, LaLα, NdLα, GdLα, HoLα, YbLα, BaLα, CeLα, SmLα, TbLα, ErLα, LuLα, PET for CaKα, KKα, ClKα, TiKα, NbLα, YLα, SrLα, ZrLα, PKα, UMα, ThMα, and LTAP for MgKα, FKα, NaKα, SiKα, AlKα. Counting times were 10 seconds for Zr and P and 15 seconds for all other elements, with off peak count times of 10 seconds. The standards (with elements) were SPI-barite (Ba), SPI-celestite (Sr), SPI-YAG (Y, Al), Smithsonian Cr-augite (Mg, Ca), Smithsonian ilmenite (Fe, Ti), 104 Smithsonian apatite (F, P), Smithsonian microcline (K), Smithsonian zircon (Zr), Harvard albite (Si, Na), Cameca Mn (Mn), SPI2-TlCl (Cl), SPI2-Nb (Nb), SPI2-La (La), SPI2-Ce (Ce), SPI2-Pr (Pr), SPI2-Nd (Nd), SPI2-Sm (Sm), SPI2-Eu (Eu), SPI2-Gd (Gd), SPI2-Tb (Tb), SPI2-Dy (Dy), SPI2-Ho (Ho), SPI2-Er (Er), SPI2-Tm (Tm), SPI2-Yb (Yb), SPI2-Lu (Lu), SPI2-Ta (Ta), SPI2-Th (Th), and SPI2-U (U). Formula calculations for each mineral are given in their respective sections. 4.6.3 Reflectance spectroscopy Reflectance spectroscopy was primarily carried out using the sisuROCK instrumentation (manufactured by SPECIM Spectral Imaging Ltd.) at the University of Alberta’s CoreSensing Facility, and data was handled using ENVI 4.4, a widely used and commercially available software package. Two imaging spectrometers (“cameras”) acquired reflectance spectra in the visible-near infrared (VNIR, 396 nm to 1003 nm over 784 channels for an average spectral resolution of 0.77 nm) and shortwave infrared (SWIR, 928 nm to 2530 nm over 256 channels for an average spectral resolution of 6.26 nm) portions of the electromagnetic spectrum in high spatial resolution mode. Spatial resolution of the cameras in this mode was approximately 0.079 mm / pixel in the VNIR and 0.241 mm / pixel in the SWIR. Noise was very prevalent in the shortest wavelength portion of the VNIR camera below ~550 nm and moderate from 550 nm to ~650 nm. In the high spatial resolution mode, averaging ~16 pixels resulted in reliable spectra in the noisier ranges that would be useable in spectral libraries. Spectra presented originate from single crystals, multiple crystals within a single rock sample and from multiple loose single crystals. Spectra documented here are nominally an average of 6343 pixels for the VNIR camera and 1087 pixels for the SWIR camera (Table 4.2). Samples were placed on a matte black surface that translates the samples under the camera and has very low reflectance across the sampled wavelength range. Some samples were propped up with foam blocks to ensure surfaces of interest faced the spectrometers and were in focus. All samples were also substantially thick enough to assume the reflectance spectra are representative of the mineral target. Reflectance spectra did not have the continuum removed so as to present the data unmodified and to facilitate comparison against other earlier publications. Simple Regions Of Interest (ROI) were used on samples with larger crystals to select target pixels for averaging. For smaller crystals in hand samples, a priori knowledge about the sample allowed one or several pixels of the target mineral to be isolated. These isolated spectra were averaged and used as an input spectrum for the Spectral Angle Mapper (SAM) algorithm (Kruse et al. 1993) to re-evaluate the entire scene. Strict thresholds on SAM output rule images (goodness of fit to the input spectrum) allowed discrete selections of pure ‘end member’ pixels that were averaged to generate a single representative spectrum. For example, in the SWIR imagery for the Mt Malosa zircon sample a single end member pixel with a strong absorption at 1250 nm was chosen to perform spectral angle mapping (Figures 4.3 and 4.4). 105 A strict threshold of 0.02 was chosen and applied to the scene (51,490 pixels in total) to generate a Region of Interest comprising only 40 pixels. The intent is to isolate only as many pixels as needed to extract a representative, low-noise, average spectrum from the target mineral without introducing influence from other minerals/materials. Too loose of a threshold inevitably includes non-representative pixels and discrete absorption features become subdued (e.g., absorption at 1389 nm). These 40 pixels were averaged to obtain a single spectrum for zircon from Mt Malosa in the SWIR range. Figure 4.3 illustrates how sequentially less strict thresholds (most strict @ 0.02 to least strict @ 0.1) result in sequentially larger ROIs, and Figure 4.4 illustrates how a averaging spectra from different ROIs impacts the resulting average spectrum from this scene. For VNIR imagery (425,320 pixels), a similar process was undertaken but was based on the absorption depth at 808 nm, thresholded to values greater than 0.09 and resulting in 701 pixels. These 701 pixels were averaged to obtain a single spectrum for zircon from Mt Malosa in the VNIR range. Although these two classification methods were based on different approaches, they both produced very similar distributions of “best representative pixels” in the VNIR and SWIR imagery. Table 4.2 Pixel counts per sample used to produce average spectrum in VNIR and SWIR ranges. Mineral Sample VNIR ROI PixelsSWIR ROI PixelsCerite Bastnas 9459* 1853Mosandrite Kipawa CMN F92-23 2977 1228Kainosite LongLake 175* 22**Zircon Green River 4391 497 Mudtank 7424 2365 North Burgess 12172 2110 Mt Malosa 701* 40** St Peters Dome 948 195Eudialyte MSH CMN 72-24 9414 222 MSH CMN 88-79 (Pinch Collection) 5890 476 MSH CMNOC 2045 11171 730 MSH CMNOC 37104 7141 2867 MSH CMNOC 478 11324 986 MSH CMNOC 476 5856 178 Kipawa UofA 2416 227 Kipawa Mariano 4070 3186 Kipawa CMN F92-23 12305 1313*Used absorption depth near 745 or 808 nm to isolate pixels **Used Spectral Angle Mapper to isolate pixels 106 Figure 4.3 Imagery for Mount Malosa zircon hand sample in (A) SWIR False Colour, (B) SAM match strength to input spectrum (located at crosshairs), (C) SWIR false colour with overlapping ROIs based on tightening SAM thresholds (ROI colours match spectra in Figure 4.4), and (D) digital photograph. The box in (D) is placed around a prominent cluster of zircon crystals. 107 Figure 4.4 Mean spectra in the SWIR of ROI based on different SAM thresholds, as labeled, for illustrative purposes of generating an average spectrum. The top black spectrum is the input spectrum from a single pixel, the purple spectrum is the mean from pixels included when the SAM threshold is set to >0.02, resulting in 40 pixels. Other spectra (colour coded to the corresponding ROI) are from less strict SAM thresholds as labeled and therefore larger ROIs. 4.7 Electron microprobe compositions Samples described here were selected from a larger set and were characterized by imaging reflectance spectroscopy, scanning electron microscopy and microprobe analysis. Tables 4.3 to 4.6 document electron microprobe results for the REE-bearing silicates in this study and Figures 4.5 and 4.6 show Chondrite-normalized patterns of selected samples. Figure 4.5 shows kainosite and cerite, as well as the only zircon sample (Mt Malosa) with substantially high REE content. Figure 4.6 emphasizes eudialyte samples that are heavy rare earth element enriched (HREE) from Kipawa vs light rare earth element enriched (LREE) from Mont Saint Hilaire (MSH), as well as including mosandrite from Kipawa. 108 Table 4.3 Electron microprobe compositions for mosandrite, cerite, and kainosite. *H2O and CO2 determined by stoichiometry and formulae calculations can be found in text. K and Ba sought but not detected. Sample CMN F92‐23 Bastnas Long Lake CMN F92‐23 Bastnas Long LakeMineral Mosandrite Cerite Kainosite Mosandrite Cerite Kainosite# Analyses 5 4 5Nb2O5 (wt.%) 1.61 0.06 0.03Nb5+ (apfu) 0.10 0.01 0.00Ta2O5 0.05 0.00 0.00 Ta5+ 0.00 0.00 0.00P2O5 0.02 0.01 0.05 P5+ 0.00 0.00 0.01SiO2 29.76 21.20 34.45 Si4+ 4.00 6.81 4.07TiO2 8.29 0.00 0.00 Ti4+ 0.84 0.00 0.00ZrO2 0.44 0.00 0.01 Zr4+ 0.03 0.00 0.00UO2 0.00 0.00 0.00 U4+ 0.00 0.00 0.00ThO2 0.10 0.00 0.00 Th4+ 0.00 0.00 0.00Al2O3 0.06 0.01 0.00 Al3+ 0.01 0.00 0.00La2O3 1.47 13.78 0.00 La3+ 0.07 1.63 0.00Ce2O3 4.02 32.69 0.51 Ce3+ 0.20 3.84 0.02Pr2O3 0.54 4.23 0.17 Pr3+ 0.03 0.50 0.01Nd2O3 2.51 14.74 2.49 Nd3+ 0.12 1.69 0.11Sm2O3 0.74 1.94 2.48 Sm3+ 0.03 0.22 0.10Eu2O3 0.09 0.07 0.22 Eu3+ 0.00 0.01 0.01Gd2O3 0.82 0.77 3.44 Gd3+ 0.04 0.08 0.14Tb2O3 0.14 0.00 0.45 Tb3+ 0.01 0.00 0.02Dy2O3 1.05 0.08 3.38 Dy3+ 0.05 0.01 0.13Ho2O3 0.21 0.00 0.51 Ho3+ 0.01 0.00 0.02Er2O3 0.57 0.00 1.56 Er3+ 0.02 0.00 0.06Tm2O3 0.05 0.00 0.11 Tm3+ 0.00 0.00 0.00Yb2O3 0.32 0.00 1.55 Yb3+ 0.01 0.00 0.06Lu2O3 0.00 0.00 0.09 Lu3+ 0.00 0.00 0.00Y2O3 5.96 0.91 19.97 Y3+ 0.43 0.16 1.26SrO 0.09 0.00 0.00 Sr2+ 0.01 0.00 0.00MgO 0.03 1.46 0.00 Mg2+ 0.01 0.70 0.00FeO 0.05 1.00 0.00 Fe2+ 0.01 0.27 0.00MnO 0.05 0.06 0.00 Mn2+ 0.01 0.02 0.00CaO 26.68 1.94 15.50 Ca2+ 3.84 0.67 1.96Na2O 6.39 0.00 0.00 Na+ 1.67 0.00 0.00Cl 0.02 0.08 0.00 Cl ‐ 0.01 0.04 0.00F 4.63 1.00 0.00 F‐ 1.97 1.02 0.00H2O* 4.49 2.77 2.54 H+ 4.03 5.94 2.00CO2* 6.20 C4+ NA NA 1.00O=CL 0.00 ‐0.02 0.00 O2‐ 17.26 29.94 16.00O=F ‐1.95 ‐0.42 0.00TOTAL 99.30 98.36 95.71REE2O3 18.49 69.21 36.93109 Table 4.4 Electron microprobe compositions for zircon. *Formula contents on the basis of 4 anions. Ta, Ca, K and Ba sought but not detected. Sample Green River MudtankNorth BurgessMt MalosaSt Peters DomeGreen River MudtankNorth BurgessMt MalosaSt Peters Dome# Analyses 7 5 5 5 5Nb2O5 (wt.%) 0.18 0.16 0.18 1.53 0.19Nb5+ (apfu) 0.00 0.00 0.00 0.02 0.00P2O5 0.04 0.04 0.04 0.24 0.10 P5+ 0.00 0.00 0.00 0.01 0.00SiO2 31.93 32.11 32.21 30.65 31.70 Si4+ 1.00 1.00 1.00 0.99 1.01TiO2 0.00 0.00 0.00 0.03 0.00 Ti4+ 0.00 0.00 0.00 0.00 0.00ZrO2 65.10 65.69 65.95 58.01 63.70 Zr4+ 0.99 1.00 1.00 0.92 0.98UO2 0.08 0.00 0.02 0.00 0.00 U4+ 0.00 0.00 0.00 0.00 0.00ThO2 0.56 0.02 0.00 0.29 0.03 Th4+ 0.00 0.00 0.00 0.00 0.00Al2O3 0.00 0.00 0.00 0.00 0.00 Al3+ 0.00 0.00 0.00 0.00 0.00La2O3 0.00 0.00 0.01 0.02 0.00 La3+ 0.00 0.00 0.00 0.00 0.00Ce2O3 0.01 0.00 0.00 0.24 0.00 Ce3+ 0.00 0.00 0.00 0.00 0.00Pr2O3 0.00 0.00 0.01 0.05 0.00 Pr3+ 0.00 0.00 0.00 0.00 0.00Nd2O3 0.00 0.00 0.00 0.48 0.00 Nd3+ 0.00 0.00 0.00 0.01 0.00Sm2O3 0.00 0.00 0.00 0.42 0.00 Sm3+ 0.00 0.00 0.00 0.01 0.00Eu2O3 0.01 0.00 0.00 0.04 0.00 Eu3+ 0.00 0.00 0.00 0.00 0.00Gd2O3 0.00 0.00 0.00 0.37 0.00 Gd3+ 0.00 0.00 0.00 0.00 0.00Tb2O3 0.00 0.00 0.00 0.00 0.00 Tb3+ 0.00 0.00 0.00 0.00 0.00Dy2O3 0.01 0.00 0.01 0.29 0.01 Dy3+ 0.00 0.00 0.00 0.00 0.00Ho2O3 0.00 0.00 0.00 0.05 0.01 Ho3+ 0.00 0.00 0.00 0.00 0.00Er2O3 0.01 0.00 0.00 0.16 0.01 Er3+ 0.00 0.00 0.00 0.00 0.00Tm2O3 0.00 0.01 0.00 0.02 0.00 Tm3+ 0.00 0.00 0.00 0.00 0.00Yb2O3 0.00 0.01 0.00 0.30 0.04 Yb3+ 0.00 0.00 0.00 0.00 0.00Lu2O3 0.01 0.00 0.00 0.00 0.00 Lu3+ 0.00 0.00 0.00 0.00 0.00Y2O3 0.27 0.02 0.01 2.39 0.03 Y3+ 0.00 0.00 0.00 0.04 0.00SrO 0.15 0.15 0.15 0.12 0.13 Sr2+ 0.00 0.00 0.00 0.00 0.00MgO 0.00 0.00 0.00 0.00 0.01 Mg2+ 0.00 0.00 0.00 0.00 0.00FeO 0.00 0.00 0.00 0.03 0.00 Fe2+ 0.00 0.00 0.00 0.00 0.00MnO 0.00 0.01 0.01 0.05 0.02 Mn2+ 0.00 0.00 0.00 0.00 0.00Na2O 0.00 0.00 0.00 0.01 0.00 Na+ 0.00 0.00 0.00 0.00 0.00Cl 0.00 0.00 0.00 0.00 0.01 Cl ‐ 0.00 0.00 0.00 0.00 0.00F 0.01 0.03 0.05 0.04 0.03 F‐ 0.00 0.00 0.01 0.00 0.00O=CL 0.00 0.00 0.00 0.00 0.00 O2‐ 4.00 4.00 4.00 4.00 4.00O=F 0.00 ‐0.01 ‐0.02 ‐0.02 ‐0.01TOTAL 98.37 98.24 98.63 95.82 96.04REE2O3 0.32 0.04 0.04 4.83 0.10110 Table 4.5 Electron microprobe compositions for eudialyte samples (LREE Group). *H2O determined by stoichiometry based on 5 apfu in the two X sites. Formula contents based on 29 apfu in the Zr and Si(7) sites (Si, Al, Zr, Ti, Nb, Ta). Sample CMNOC 2045CMNOC 476 CMN 88‐79 CMN 72‐24CMNOC 37104CMNOC 2045CMNOC 476 CMN 88‐79CMN 72‐24CMNOC 37104# Analyses 5 8 5 5 5Nb2O5 (wt.%) 1.80 1.95 2.24 2.48 1.96Nb5+ (apfu) 0.43 0.43 0.51 0.56 0.43Ta2O5 0.14 0.03 0.01 0.00 0.00 Ta5+ 0.02 0.00 0.00 0.00 0.00P2O5 0.02 0.02 0.00 0.01 0.01 P5+ 0.01 0.01 0.00 0.00 0.00SiO2 48.03 51.97 49.98 51.31 53.21 Si4+ 25.60 25.57 25.34 25.57 25.67TiO2 0.03 0.07 0.10 0.32 0.29 Ti4+ 0.01 0.03 0.04 0.12 0.11ZrO2 10.86 11.96 12.27 11.19 11.73 Zr4+ 2.82 2.87 3.03 2.72 2.76UO2 0.00 0.00 0.00 0.03 0.07 U4+ 0.00 0.00 0.00 0.00 0.01ThO2 0.05 0.03 0.14 0.11 0.09 Th4+ 0.01 0.00 0.02 0.01 0.01Al2O3 0.17 0.17 0.13 0.06 0.06 Al3+ 0.11 0.10 0.08 0.04 0.03La2O3 1.16 0.76 1.48 1.01 1.05 La3+ 0.23 0.14 0.28 0.19 0.19Ce2O3 2.04 1.32 2.93 1.93 1.88 Ce3+ 0.40 0.24 0.54 0.35 0.33Pr2O3 0.19 0.12 0.30 0.17 0.15 Pr3+ 0.04 0.02 0.06 0.03 0.03Nd2O3 0.58 0.39 0.92 0.59 0.52 Nd3+ 0.11 0.07 0.17 0.11 0.09Sm2O3 0.10 0.06 0.16 0.08 0.07 Sm3+ 0.02 0.01 0.03 0.01 0.01Eu2O3 0.01 0.00 0.00 0.00 0.00 Eu3+ 0.00 0.00 0.00 0.00 0.00Gd2O3 0.07 0.05 0.12 0.05 0.05 Gd3+ 0.01 0.01 0.02 0.01 0.01Tb2O3 0.02 0.00 0.00 0.00 0.00 Tb3+ 0.00 0.00 0.00 0.00 0.00Dy2O3 0.20 0.05 0.04 0.02 0.00 Dy3+ 0.03 0.01 0.01 0.00 0.00Ho2O3 0.02 0.01 0.01 0.01 0.00 Ho3+ 0.00 0.00 0.00 0.00 0.00Er2O3 0.08 0.02 0.02 0.01 0.01 Er3+ 0.01 0.00 0.00 0.00 0.00Tm2O3 0.00 0.00 0.00 0.00 0.00 Tm3+ 0.00 0.00 0.00 0.00 0.00Yb2O3 0.04 0.01 0.01 0.01 0.00 Yb3+ 0.01 0.00 0.00 0.00 0.00Lu2O3 0.00 0.00 0.00 0.00 0.00 Lu3+ 0.00 0.00 0.00 0.00 0.00Y2O3 0.55 0.35 0.62 0.35 0.34 Y3+ 0.16 0.09 0.17 0.09 0.09SrO 0.08 0.27 0.29 0.32 0.26 Sr2+ 0.03 0.08 0.09 0.09 0.07MgO 0.00 0.03 0.03 0.02 0.02 Mg2+ 0.00 0.02 0.02 0.02 0.01FeO 6.55 3.88 2.93 2.62 2.47 Fe2+ 2.92 1.60 1.24 1.09 1.00MnO 2.67 3.87 7.90 9.14 8.60 Mn2+ 1.21 1.61 3.39 3.86 3.52BaO 0.01 0.03 0.01 0.02 0.06 Ba2+ 0.00 0.01 0.00 0.00 0.01CaO 7.95 9.20 3.78 5.53 5.75 Ca2+ 4.54 4.85 2.05 2.95 2.97Na2O 11.79 7.63 6.21 5.73 4.46 Na+ 12.19 7.28 6.10 5.54 4.17K2O 0.38 0.42 0.36 0.30 0.18 K+ 0.26 0.26 0.23 0.19 0.11Cl 1.03 0.79 0.53 0.39 0.45 Cl ‐ 0.93 0.66 0.46 0.33 0.37F 0.18 0.09 0.19 0.18 0.08 F‐ 0.30 0.14 0.31 0.28 0.12H2O* 1.06 1.28 1.25 1.32 1.40 H+ 3.77 4.20 4.24 4.39 4.51O=CL ‐0.23 ‐0.18 ‐0.12 ‐0.09 ‐0.10 O2‐ 75.93 72.72 71.86 72.26 71.09O=F ‐0.08 ‐0.04 ‐0.08 ‐0.08 ‐0.03TOTAL 97.55 96.61 94.76 95.15 95.09REE2O3 5.06 3.14 6.61 4.23 4.07L / H 4.56 6.14 8.44 9.58 10.63111 Table 4.6 Electron microprobe compositions for eudialyte samples (HREE Group). *H2O determined by stoichiometry based on 5 apfu in the two X sites. Formula contents based on 29 apfu in the Zr and Si(7) sites (Si, Al, Zr, Ti, Nb, Ta). Sample Kipawa ‐MKipawa ‐UofACMN F92‐23CMNOC 478Kipawa ‐MKipawa ‐UofACMN F92‐23CMNOC 478# Analyses 11 5 5 15Nb2O5 (wt.%) 0.73 0.97 0.77 3.19Nb5+ (apfu) 0.18 0.22 0.18 0.80Ta2O5 0.01 0.08 0.05 0.00 Ta5+ 0.00 0.01 0.01 0.00P2O5 0.00 0.00 0.02 0.02 P5+ 0.00 0.00 0.01 0.01SiO2 47.58 51.30 50.90 45.54 Si4+ 25.32 25.63 25.81 25.28TiO2 0.35 0.30 0.29 0.10 Ti4+ 0.14 0.11 0.11 0.04ZrO2 12.75 12.21 11.48 10.44 Zr4+ 3.31 2.97 2.84 2.83UO2 0.00 0.02 0.00 0.00 U4+ 0.00 0.00 0.00 0.00ThO2 0.01 0.00 0.04 0.08 Th4+ 0.00 0.00 0.01 0.01Al2O3 0.09 0.10 0.10 0.08 Al3+ 0.06 0.06 0.06 0.05La2O3 0.33 0.38 0.42 1.28 La3+ 0.07 0.07 0.08 0.26Ce2O3 0.56 0.61 0.65 2.45 Ce3+ 0.11 0.11 0.12 0.50Pr2O3 0.05 0.05 0.04 0.22 Pr3+ 0.01 0.01 0.01 0.04Nd2O3 0.24 0.25 0.22 0.60 Nd3+ 0.05 0.05 0.04 0.12Sm2O3 0.12 0.06 0.07 0.10 Sm3+ 0.02 0.01 0.01 0.02Eu2O3 0.00 0.02 0.01 0.01 Eu3+ 0.00 0.00 0.00 0.00Gd2O3 0.15 0.08 0.08 0.08 Gd3+ 0.03 0.01 0.01 0.02Tb2O3 0.04 0.00 0.04 0.03 Tb3+ 0.01 0.00 0.01 0.01Dy2O3 0.49 0.40 0.27 0.32 Dy3+ 0.08 0.06 0.04 0.06Ho2O3 0.05 0.09 0.09 0.04 Ho3+ 0.01 0.01 0.02 0.01Er2O3 0.48 0.30 0.29 0.14 Er3+ 0.08 0.05 0.05 0.02Tm2O3 0.01 0.04 0.03 0.01 Tm3+ 0.00 0.01 0.01 0.00Yb2O3 0.41 0.40 0.36 0.20 Yb3+ 0.07 0.06 0.06 0.03Lu2O3 0.02 0.03 0.02 0.00 Lu3+ 0.00 0.01 0.00 0.00Y2O3 2.67 2.54 2.11 0.99 Y3+ 0.76 0.68 0.57 0.29SrO 0.19 0.18 0.22 0.12 Sr2+ 0.06 0.05 0.07 0.04MgO 0.09 0.10 0.10 0.01 Mg2+ 0.07 0.07 0.08 0.01FeO 2.44 2.28 2.68 3.18 Fe2+ 1.09 0.95 1.14 1.48MnO 1.33 1.37 1.46 6.68 Mn2+ 0.60 0.58 0.63 3.14BaO 0.09 0.21 0.15 0.01 Ba2+ 0.02 0.04 0.03 0.00CaO 11.91 12.44 12.71 8.49 Ca2+ 6.79 6.66 6.91 5.05Na2O 4.11 10.46 11.08 11.12 Na+ 4.24 10.13 10.89 11.97K2O 0.57 0.49 0.51 0.29 K+ 0.39 0.31 0.33 0.21Cl 1.08 1.13 1.20 0.92 Cl ‐ 0.97 0.96 1.03 0.87F 0.17 0.12 0.16 0.13 F‐ 0.29 0.19 0.26 0.23H2O* 1.05 1.16 1.10 1.05 H+ 3.74 3.85 3.71 3.91O=CL ‐0.24 ‐0.25 ‐0.27 ‐0.21 O2‐ 72.17 74.73 75.28 77.70O=F ‐0.07 ‐0.05 ‐0.07 ‐0.05TOTAL 89.86 99.86 99.38 97.66REE2O3 5.62 5.25 4.70 6.47L / H 0.35 0.38 0.46 2.74112 Figure 4.5 Selected Chondrite-normalized REE plots for zircon, kainosite and cerite. Missing points due to analytes being below detection. Figure 4.6 Selected Chondrite-normalized REE plots for eudialyte and mosandrite. Eudialyte CMN72-24 (higher LREE) is from MSH while hand sample F92-23 with both eudialyte (higher HREE) and mosandrite originates from Kipawa. Missing points due to analytes being below detection. 113 4.7.1 Cerite The single cerite sample shows a satisfactory total with strongly elevated LREE, totalling 69.21 wt.% Total Rare Earth Oxide (TREO). Formula contents were calculated based on 31 anions and are consistent with site occupancies reported by Moore and Chen (1983) and Pakhomovsky et al. (2002). 4.7.2 Mosandrite The single mosandrite sample shows a satisfactory total with elevated LREE and HREE, totalling 18.49 wt.% TREO. Formula contents were calculated based on 4 Si apfu, and are consistent with occupancies in Bellezza et al. (2009) including the presence of structural H2O and OH-. 4.7.3 Kainosite The single kainosite sample shows a satisfactory total with a preference for Y + HREE, totalling 36.93 wt.% TREO. Formula contents were calculated based on 16 anions and are consistent with site occupancies reported by Rumanova et al. (1967). 4.7.4 Zircon The five zircon samples studied all have satisfactory totals, although the St Peters Dome and Mt Malosa samples are marginal, and they show a good diversity of chemical variation to investigate their reflectance characteristics. The Mudtank sample shows no detectable U and the lowest REE content of all samples. The St Peters Dome sample has U below detection and a slight enrichment in HREE, notably with Er2O3 of 0.01 wt.% and Yb2O3 of 0.04 wt.%. The Mt Malosa samples shows U below detection but the highest REE (TREO of 4.83 wt.%). The North Burgess sample contains 0.02 wt.% UO2 and very low REE values. The Green River samples show the highest UO2 value of 0.08 wt.% (~700 ppm) and moderate REE (TREO of 0.32 wt.%). High Th content correlates with high U content. Table 4.7 summarizes select data for the zircon samples and provides geological and age context, relevant to radiation dosage through time. The average detection limit of the microprobe for UO2 in zircon is 0.075 % UO2. The Green River sample was the only one that was consistently above detection, while the North Burgess sample had only one analytical point above detection and the remainder of the sample’s points were all below detection but unlikely to be free of uranium. For example, the range of U in Mudtank zircon from the literature is ~5 ppm to 40 ppm (e.g., Currie et al. 1992, Jaeger et al. 2006). The U concentration range of zircon from the St Peter’s dome area is from 100 to 300 ppm (Smith et al. 1999), however, there is little context for the sample and the geological setting of this region is varied. 114 Zircon formulae were normalized to 4 anions following Breiter et al. (2006) who studied highly substituted zircon. They noted that this method sometimes resulted in high cation values for the A site, which hosts REE, but that the overpopulation of the A site in their highly substituted samples suggests that some of these cations would be sitting in interstitial sites. The high REE Mount Malosa zircon was the only sample of our suite that resulted in high A site population but is in an acceptable range. Kempe et al. (2010) noted that their zircon sample from Mt Malosa was metamict. Table 4.7 Chemical variation relevant to reflectance spectroscopy, as well as probable ages and geological settings of zircon samples. Sample U and Th REE Analytical Total (wt.%) Age Geological Setting Reference Mudtank Below detection Very Low 98.24 732 Ma Carbonatite Currie et al. 1992 St Peters Dome Below detection Low 96.04 ~1 Ga Pegmatites of A-type granite suite Smith et al. 1999 Mt Malosa Below detection High 95.82 ~113 Ma Pegmatites of A-type granite suite Eby et al. 1995, Guastoni et al. 2009 North Burgess Moderate Very Low 98.63 1 Ga? Pegmatite related metasomatic skarn? Currie 1951 Green River High Moderate 98.37 329 Ma Syenitic pegmatite Braun et al. 2009 *Analytical total is given as a rough proxy for degree of metamictization 4.7.5 Eudialyte Content of (REE+Y)2O3 for the eudialyte samples range from 3.12 to 6.62 wt.% and two populations exist based on relative REE content. The Kipawa samples show enrichment of HREE (Tb to Lu, Y) whereas the remainder of the eudialyte samples from Mont Saint Hilaire (MSH) show a greater occupation by LREE (La to Gd). One exception, Sample CMNOC478 from MSH shows an intermediate distribution of REE. Notably, sample CMNOC478 also showed some compositional zoning when using high contrast settings under the SEM, and 15 spots were analyzed on two grains. Three of the analytical points significantly impact the range of standard deviation values, and within the points from CMNOC478 show generally elevated REE2O3 and MnO with lower CaO and SiO2. Despite this sample showing some compositional variations, it has been kept within the suite of eudialyte samples primarily because it shows distinct “intermediate” chemistry and higher Nb2O5. This distinct character is not related to averaging of compositionally zoned areas. Total REE content is not appreciably correlated with any other cations, however, relative REE content (LREE / HREE) correlates positively with Mn and negatively with Ca contents (Figure 4.7). The higher Mn samples also show higher Th and Nb. Thus, the HREE population (Kipawa) shows higher Ca 115 while the LREE population (MSH) shows higher Mn, Th and Nb. These trends are likely the result of both the compositional environment of formation and crystal chemical controls. Figure 4.7 Relative REE–Ca–Mn–Th compositional trends of eudialyte samples. Values in wt.%. Eudialyte formulae were calculated based on the recommendations of Johnsen et al. (2001) of normalizing to 29 cations occupying the M3 and Si(7) sites (Si, Al, Zr, Ti, Nb, Ta) since no structural information is available. This is satisfactory for most of our samples with REE occupying both Na4 and M1 sites, however, those samples with higher Ca (i.e., from Kipawa) leave no room in the M1 site for the REE, pushing them all to the Na4 site. For the high Ca analyses of Johnsen and Grice (1999) that do have accompanying structural data, including samples from Kipawa, it is shown that Ca and REE both populate the M1 and Na4 sites and it is therefore likely the case with our high Ca samples. We can then conclude that all eudialyte samples analyzed for this study will show REE in both the M1 and Na4 sites. 4.8 Spectra and spectral variability of REE‐bearing silicates Spectroscopic descriptions of the REE-bearing silicates begin with strongly LREE-enriched cerite, followed by mosandrite which is LREE-enriched but also hosts HREE. The HREE-rich mineral kainosite is then described, followed by zircon with generally low amounts of U and REE. Finally, a suite of eudialyte samples is presented with examples of both LREE- and HREE-enrichment. A brief introduction to each mineral’s spectroscopic features is given, followed by spectra in the VNIR and SWIR ranges (Figures 4.9 to 4.15) and a “Band Index Table” or series of tables (Tables 4.8 to 4.14). Regions of spectral features are divided into numbered Clusters, which are outlined and shaded on the index tables. In the case of zircon, the numerous U-related absorptions are divided into Clusters labeled by sequential letters (A, B, C…). Prominent absorption bands are emphasized by shading on the tables. The tables also provide an interpretation on the origin of each spectral feature as chosen through comparison with reflectance spectra from unpublished reagent grade lanthanide oxide spectra, REE-doped 116 calibration standard spectra, and other REE-bearing mineral spectra for which compositional data exists, as well as REE spectroscopy literature. In most cases confidence in assignments is strong, however, ambiguity is present in other cases and denoted with “?”. Some interpretations of absorption bands include multiple causes, typically multiple lanthanides, because of the large number of overlapping multiplets present. The position and shape of the absorption features were recorded using reflectance spectra (i.e., not continuum removed spectra). Descriptors for absorption bands include MIN (minimum) with modifiers –st (strong), -w (weak), -n (noisy), –b (broad) and SH (shoulders) with modifiers –w (weak) and –n (noisy). Noisy modifiers are typically restricted to the short and long wavelengths of the VNIR spectrometer (<600 nm, >950 nm). Noisy and weak absorption bands are typically restricted to scenarios where other spectra show reliable features near the same wavelength position. For example, the ~627 nm noisy shoulder of LREE-enriched cerite was confidently included based on the presence of the 625 nm absorption in LREE-enriched bastnaesite. Divisions between VNIR and SWIR ranges on the tables are denoted by heavy horizontal line. Figure 4.8 Example spectrum showing types absorption band labels with data points indicated by diamonds. Spectrum is of monazite (UofA Unknown sample). 117 4.8.1 Cerite reflectance spectrum Cerite, Ce9Fe3+(SiO4)6(SiO3OH)(OH,F), is strongly enriched in the LREE, which means its spectral characteristics will be driven mostly by the spectrally active lanthanides Pr3+, Nd3+, and Sm3+. As a silicate with structural OH- one would also expect vibrational combination and overtone bands. During inspection of the image cube from the cerite hand sample several patches of bastnaesite were identified based on the distinct absorption at 2243 nm, the morphology of the 1080 and 1232 nm Sm3+ related absorptions, and the morphology of the Nd3+ related absorption near 870 nm (see Chapter 2, Turner et al. 2014). The prominent patches were excluded from the VNIR and SWIR pixels used to generate the average spectrum, however, it is likely that some bastnaesite exists with cerite below the spatial resolution of the imaging spectrometers (e.g., see Appendix B, Figure B.17). Figure 4.9 Reflectance spectra of cerite in the VNIR (500 – 1000 nm) and SWIR (975 – 2530 nm) ranges. Clusters are indicated by labeled thick horizontal lines and prominent absorptions highlighted in the index tables are labeled with wavelength position. 118 Table 4.8 Prominent absorption features of cerite in the VNIR and SWIR ranges. Note: Absorption bands marked with * share wavelength positions with bastnaesite. VNIR The spectrum for cerite in the VNIR range is divided into 6 clusters (Table 4.8, Figure 4.9). Cluster 1 is an absorption band with its minimum centered at 523 nm. Cluster 2 has one prominent band with its minimum at 583 nm and shoulders near 627, 642 and 661 nm. Cluster 3 is another single and absorption at 681 nm. Cluster 4 is characterized by a prominent absorption minimum at 746 nm with a Ceri te Cluster Abs Shape Probable OriginVNIR Range 1 523 MIN Nd2 583 MIN ‐ s t Nd, Pr627 SH ‐ n Nd642 SH ‐ n Nd661 SH ‐ n Nd3 681 MIN Nd4 737 MIN Nd746 MIN ‐ s t Nd753 SH Nd5 797 MIN Nd803 MIN ‐ s t Nd811 SH Nd825 SH Nd6 864 MIN ‐ s t Nd876 MIN ‐ s t Nd888 SH Nd898 SH Nd945 MIN ‐ w, n Sm961 MIN ‐ w, n SmSWIR Range 7 1010 SH Pr1080 MIN ‐ s t Sm1112 SH Sm8 1232 MIN ‐ s t Sm1263 SH Sm9 1383 MIN Sm1408 SH H2O / OH1452 MIN Pr1540 MIN ‐ s t Pr>Sm1578 SH Pr, Sm Nd1622 SH Pr, Sm Nd1710 MIN Nd10 1968 MIN ‐ s t Pr, Sm, H2O2030 SH Pr>Eu11 2193 MIN OH/REE/Mg‐Fe* 2243 MIN ‐ w OH/REE/Mg‐Fe* 2312 MIN OH/REE/Mg‐Fe* 2330 MIN OH/REE/Mg‐Fe2355 SH OH/REE/Mg‐Fe2380 SH OH/REE/Mg‐Fe2424 MIN ‐ s t OH/REE/Mg‐Fe2487 MIN OH/REE/Mg‐Fe2518 SH OH/REE/Mg‐Fe119 weaker absorption at 737 nm and a shoulder 753 nm. Cluster 5 is similar, showing a prominent absorption minimum at 803 nm with a weaker minimum at 797 nm and shoulders at 811 and 825 nm. Cluster 6 is a prominent doublet with equally strong minima at 864 and 876 nm. A number of weaker absorptions are observed at 888, 898, 945 and 961 nm. All significant features in the VNIR are attributable to Nd3+. SWIR The spectrum for cerite in the SWIR range is divided into 5 clusters. Cluster 7 shows a shoulder at 1010 nm, followed by a sharp absorption at 1080 nm and accompanying shoulder at 1112 nm. Cluster 8 is another sharp absorption at 1232 nm followed a shoulder at 1263. Cluster 9 is a broader collection of absorption features. A narrow absorption is located at 1383 nm, followed by a shoulder at 1408 and local minimum at 1452 nm. The deepest absorption band occurs at 1540 nm, followed by shoulders at 1578 and 1622 nm. A small but distinct absorption is located at 1710 nm atop a local reflectance high. Cluster 10 is a strong absorption minimum at 1968 nm followed by a shoulder at 2030 nm. Cluster 11 extends from ~2150 out to the end of the spectrometer’s range of 2530 nm. The strongest absorptions bands occur at 2193, 2312 and 2424 nm. Shoulders and other weak minima are located at 2243, 2330, 2355, 2380, 2487 and 2518 nm. The weak minimum at 2243 nm is particularly diagnostic of the bastnaesite spectrum, and absorption bands at 2312 and 2330 nm also coincide with bastnaesite. These features are marked on Table 4.8 and Figure 4.9 with an asterix and suggest a weak mixed response for the average spectrum from a sub-pixel level. 120 4.8.2 Mosandrite reflectance spectrum Mosandrite, Ti(□,Ca,Na)3(Ca,REE)4(Si2O7)2[H2O,OH,F]4*1H2O, shows a preference for the LREE but also accommodates moderate amounts of HREE and Y. This leads to the spectral signature of mosandrite being dominated by Nd3+, Sm3+ and Pr3+ but with influence from Dy3+, Er3+ and Yb3+, its three most abundant HREE. Mosandrite is also host to structural water and hydroxyl according to Bellezza et al. (2009) which will results in vibrational absorption bands. Figure 4.10 Reflectance spectra of mosandrite in the VNIR (500 – 1000 nm) and SWIR (975 – 2530 nm) ranges. Clusters are indicated by labeled thick horizontal lines and prominent absorptions highlighted in the index tables are labeled with wavelength position. 121 Table 4.9 Prominent absorption features of mosandrite in the VNIR and SWIR ranges. VNIR The spectrum for mosandrite in the VNIR range is divided into 7 clusters (Table 4.9, Figure 4.10). Cluster 1 is a main absorption at 527 nm with a shoulder near 547 nm. Cluster 2 is a pronounced pair of strong absorption minima centered at 574 and 586 nm, followed by a series of shoulders at 615, 627 and 651 nm. Cluster 3 is a moderate absorption at 681 nm. Cluster 4 is three strong overlapping absorptions with the strongest minima at 740 nm and flanking absorptions at 736 and 745 nm. Cluster 5 is a strong minimum located at 804 nm with flanking shoulders at 772, 795 and 811 nm. Cluster 6 is an Mosandri te Cluster Abs Shape Probable OriginVNIR Range 1 527 MIN Nd547 SH ‐ n Nd2 574 MIN ‐ st Nd, Pr586 MIN ‐ st Nd, Pr615 SH Nd627 SH ‐ n Nd651 SH ‐ b, n Er3 681 MIN Nd4 736 MIN Nd740 MIN ‐ st Nd>Dy745 MIN Nd > Dy5 772 SH Nd > Dy795 SH Nd > Dy804 MIN ‐ st Nd > Dy811 SH Nd > Dy6 864 SH Nd874 MIN ‐ st Nd880 SH Nd919 SH ‐ n Dy945 MIN ‐ w, n Sm > Dy7 976 MIN ‐ st Er, YbSWIR Range 8 1074 MIN ‐ st Sm1093 SH Sm9 1232 SH Sm1257 MIN ‐ st Sm10 1377 SH Sm1440 SH H2O / OH1471 MIN ‐ st Pr1528 MIN Sm>Pr, Er1585 SH Sm>Pr, Nd11 1729 SH Nd, Dy1817 SH Nd, Dy1930 MIN ‐ st H2O12 2318 MIN OH/REE/Ti2393 SH OH/REE/Ti2418 SH OH/REE/Ti2462 MIN OH/REE/Ti122 absorption minimum at 874 nm with shoulders at 864 and 880 nm and two subtle absorptions near 919 and 945 nm. Cluster 7 is a single prominent absorption band at 976 nm. SWIR The spectrum for mosandrite in the SWIR range is divided into 5 clusters (Table 4.9, Figure 4.10). Cluster 8 is characterized by a strong absorption with a minimum at 1074 nm and a subtle shoulder at 1093 nm. Cluster 9, similarly, has a prominent absorption band at 1257 but a shoulder at the shorter wavelength of 1232 nm. Cluster 10 is a collection of absorptions, starting with shoulders at 1377 and 1440 nm, a strong minimum at 1471 nm, a minimum at 1528 nm and a shoulder at 1585 nm. Cluster 11 has two shoulders at 1729 and 1817 nm, followed by a strong absorption at 1930 nm. Cluster 12 includes two minima at 2318 and 2462 nm and intermediate shoulders at 2392 and 2418 nm. These absorptions in Cluster 12 are assumed to be combinations and overtones related to bonding amongst H2O, OH, Ti and the REE. 123 4.8.3 Kainosite reflectance spectrum Kainosite-(Y), Ca2Y2(SiO3)4(CO3)·H2O shows preference for the HREE and Y, however, the sample studied from Long Lake does contain appreciable Nd. Its spectrum is thus driven by absorptions related to Nd3+, Sm3+, Dy3+ and Er3+ with lesser input from other spectrally active lanthanides. Since this mineral contains both water and the CO32- radical, absorptions related to vibrational features are also expected. Figure 4.11 Reflectance spectra of kainosite in the VNIR (550 – 1000 nm) and SWIR (975 – 2530 nm) ranges. Clusters are indicated by labeled thick horizontal lines and prominent absorptions highlighted in the index tables are labeled with wavelength position. 124 Table 4.10 Prominent absorption features of kainosite in the VNIR and SWIR ranges. VNIR In the VNIR range, there are 6 main clusters in the spectrum for kainosite (Table 4.10, Figure 4.11). Cluster 1 occurs near 585 nm and comprises three weak absorptions, however, there is considerable noise in this region given the relatively smaller number of pixel comprised in the ROI. Cluster 2 is a single and relatively broad absorption at 651 nm. Cluster 3 is a main absorption at 750 nm flanked by absorption minima at 737 and 754 nm along with several shoulders. Cluster 4 is a strong absorption Kainosite Cluster Abs Shape Probable OriginVNIR Range 1 576 MIN ‐ n Nd585 MIN ‐ n Nd596 SH ‐ n Nd2 651 MIN ‐ n Ho, Er3 737 MIN Nd, Dy743 SH ‐ n Nd, Dy745 SH ‐ n Nd, Dy750 MIN ‐ st Nd, Dy754 MIN Nd, Dy4 782 SH ‐ n Nd, Dy, Er795 MIN Nd, Dy, Er805 MIN ‐ st Nd, Dy, Er814 MIN Nd, Dy, Er820 SH Nd, Dy, Er5 865 MIN Nd876 MIN Nd887 MIN ‐ n Nd, Ho, Dy896 MIN ‐ n Dy, Nd6 978 MIN Er, YbSWIR Range 7 1080 MIN Sm1105 SH Dy, U?1156 SH Dy, Ho8 1232 MIN Dy, Sm1263 MIN Dy, Sm1288 SH Dy9 1377 MIN Sm10 1415 SH H2O / OH1484 MIN ‐ st Er, Pr1528 MIN Er>Sm11 1653 SH ‐ w Dy, Nd1723 MIN ‐ w Dy, Nd, Tb12 1961 MIN ‐ st H2O, Eu, Pr, Ho2055 MIN ‐ st U?, Sm, Pr, Ho, Tb2105 SH Sm, Pr, Ho13 2199 MIN ‐ w CO3 / OH / REE2243 SH CO3 / OH / REE2268 SH CO3 / OH / REE2318 SH CO3 / OH / REE2387 MIN ‐ st CO3 / OH / REE2474 MIN CO3 / OH / REE2505 SH CO3 / OH / REE125 located at 805 nm with notable flanking local minima. Cluster 5 is two main absorptions located at 865 and 876 nm, followed by two other weaker and noisy absorptions at 887 and 896 nm. Cluster 6 is a single absorption at 978 nm. SWIR In the SWIR range, there are 7 main clusters in the spectrum for kainosite (Table 4.10, Figure 4.11). Cluster 7 is a prominent absorption minimum at 1080 nm followed by two shoulders at 1105 and 1156 nm. Cluster 8 is two sharp absorptions with minima at 1232 and 1263 nm followed by a weaker shoulder at 1288 nm. Cluster 9 is a single band at 1377 nm. Cluster 10 is one main absorption at 1484 nm with flanking absorptions at 1415 and 1528 nm. Cluster 11 is a series of weak minima and shoulders stretching from ~1560 to ~1865 nm with the most prominent at 1653 and 1723 nm. The absorptions are nearing the level of noise present in the spectrum (22 pixels were averaged), but we report them as candidate bands since lanthanides are responsible for numerous fine bands in this wavelength region. Cluster 12 comprises two strong overlapping absorptions centered at 1961 and 2055 nm with a shoulder at 2105 nm. Cluster 13 consists of a series of absorption minima and shoulders out to 2530 nm with the strongest at 2387 and 2474 nm. These absorptions are assumed to be combinations and overtones related to bonding amongst CO3, OH and the REE. 126 4.8.4 Zircon reflectance spectra The main variables driving the spectrum of zircon (ZrSiO4) are U4+, U5+, a crystalline vs amorphous matrix, REE3+ content and OH/H2O bands. From this perspective, three spectral classes of zircon are evident that correlate with their chemistry: U bearing (North Burgess, Green River, Mudtank), high REE with U (Mt Malosa) and metamict (St Peters Dome). Band index tables have been split into the VNIR and SWIR to accommodate the multiple samples and the inclusion of uranium-related spectral features. Figure 4.12 Stacked reflectance spectra of zircon samples in the VNIR (550 – 1000 nm) and SWIR (975 – 2530 nm) ranges. Clusters are indicated by labeled thick horizontal lines and prominent absorptions highlighted in the index tables are labeled with wavelength position. Uranium related features are distinguished by lettered clusters, yellow horizontal bars and italicized wavelength labels. From top down, Mt Malosa (pink, MM) Green River (blue, GR), North Burgess (green, NB), Mudtank (red, MT), St Peters Dome (black, SP). 127 Figure 4.13 Reflectance spectra of zircon samples in the VNIR (550 – 1000 nm) and SWIR (975 – 2530 nm) ranges. Clusters are indicated by labeled thick horizontal lines and prominent absorptions highlighted in the index tables are labeled with wavelength position. Uranium related features are distinguished by lettered clusters, yellow horizontal bars and italicized wavelength labels. Colour schemes remain for the unstacked spectra: Mt Malosa (pink, MM), Green River (blue, GR), North Burgess (green, NB), Mudtank (red, MT), St Peters Dome (black, SP). 128 Table 4.11 Prominent absorption features of zircon samples in the VNIR range. * REE related absorption clusters labeled with numbers and shaded in grey, while U related absorptions are labeled with letters and shaded in yellow ZirconCluster U Abs Shape Abs Shape Abs Shape Abs Shape Abs Shape1 Nd 575 SHNd 576 MINNd 582 MINNd 589 SH ‐ w, n 585 SHNd 594 MIN2 A U4+ and Er 653 MIN ‐ w 654 MIN 651 MIN ‐ nEr 661 SH 660 SH ‐ nNd 683 SH ‐ w 681 MIN ‐ n, wB U4+ 691 SH ‐ w 690 MIN3 Nd>Dy 738 MIN ‐ w 739 MIN ‐ stNd>Dy 750 MIN 750 MIN ‐ stNd 756 SH ‐ w 757 MINNd 760 SH ‐ w 769 SHNd 774 SH4 Nd>Er,Dy 781 MIN ‐ w 781 MINNd>Er,Dy 797 SH 796 SHNd>Er,Dy 803 MIN 802 SHNd>Er,Dy 808 MIN ‐ st 808 MIN ‐ stNd>Er,Dy 819 SHEr 835 MIN ‐ nEr 844 MIN ‐ nC U4+? 849 MIN ‐ n b5 Nd 870 MIN ‐ stNd 882 SH 880 MIN ‐ stDy 892 MIN ‐ w 892 MIN ‐ w 894 MIN 893 MIND U4+ 916 MIN ‐ n, w 916 MIN ‐ n, w 916 MIN ‐ st 915 MIN ‐ wE U4+ 961 MIN ‐ n, w 961 SH ‐ b6 Er, Yb 978 MIN ‐ st 978 MIN St Peters Dome Mudtank North Burgess Green River Mt Malosa129 Table 4.12 Prominent absorption features of zircon samples in the SWIR range. VNIR Absorption spectra for zircon are primarily described using the Green River sample for high U and the Mt Malosa sample for high REE as they both display the prominent features (Table 4.11, Figures 4.12 and 4.13). The remaining samples (North Burgess, Mudtank and St Peters Dome) show subdued absorptions in the same regions for U, and only slight absorptions for REE. Note that weak REE features are still present in the Green River samples and conversely, weak U features are still seen in the Mt Malosa sample. ZirconCluster U Abs Shape Abs Shape Abs Shape Abs Shape Abs ShapeF U4+ 1010 SH ‐ w 1010 SH 1010 MIN ‐ w 1010 SHF U4+ 1061 SH ‐ w 1055 SH 1055 MIN 1061 MIN7 Sm 1086 SH ‐ w 1086 MINSm 1105 MING U5+ 1118 SH ‐ w 1112 MIN 1112 MIN 1118 MIN 1112 SHG U4+ 1137 SH ‐ w 1149 SH 1149 SH 1143 SH ‐ w 1143 SH8 Sm? 1200 SHSm 1244 MINSm, Dy 1263 SH 1263 MINDy 1288 MIN ‐ w 1288 SHH U4+ 1326 MIN 1326 MIN 1326 MINH U4+ 1345 MIN ‐ w 1345 SH 1345 SH 1332 SH9 Sm 1389 SH 1389 MINH2O / OH 1415 MIN 1415 MIN ‐ w 1415 MIN ‐ w10 I U4+, Er 1478 MIN ‐ w 1478 SH 1478 SH 1478 SH 1427 SH ‐ wI U5+ 1503 MIN ‐ w 1503 MIN ‐ st 1503 MIN ‐ st 1503 MIN ‐ st 1503 MIN ‐ stI U4+, Er, Sm 1522 SH ‐ w 1534 SH 1534 SH 1534 SH 1528 SHEr, Sm 1560 MINSm 1616 SH ‐ w 1616 MIN ‐ wJ U4+ 1654 MIN ‐ w 1660 MIN 1660 MIN 1660 MIN 1660 MIN11 Dy 1691 MIN ‐ wJ U4+ 1704 MIN 1704 MIN 1704 MIN ‐ wNd? 1729 MIN ‐ w, b 1742 SH ‐ w 1742 SH ‐ w 1729 MIN ‐ w, b 1729 MINK U4+ 1792 SH ‐ w 1792 SH ‐ w 1792 SH ‐ w12 H2O 1924 MIN ‐ st 1924 SH 1917 MIN ‐ b 1930 MIN 1917 SHSm 1943 MINL U4+ 2068 MIN ‐ w 2068 MIN ‐ st 2068 MIN ‐ st 2074 MIN ‐ st 2074 SH ‐ wM U4+ 2187 SH 2187 SH13 Comb & Overt 2206 MIN ‐ st 2206 SH 2206 MINComb & Overt 2256 SH 2268 MIN ‐ w 2262 SH 2262 SHComb & Overt 2312 SH 2306 SH 2306 MIN 2305 MIN 2312 MINComb & Overt 2355 MIN 2362 SH 2349 SH 2355 MIN 2349 MINComb & Overt 2387 MIN 2393 SH 2393 MIN 2474 MINComb & Overt 2443 SH 2462 MINComb & Overt 2499 MIN 2499 MINComb & Overt 2511 MIN ‐ wSt Peters Dome Mudtank North Burgess Green River Mt Malosa130 There are 5 main absorptions seen in the Green River sample that are related to uranium and labeled A through E (Figure 4.12 and 4.13). The strongest of these are at 654 (A) and 916 (D) nm, while weaker bands occur at 690 (B), 849 (C), and 961 (E) nm. The absorption near 654 nm also overlaps with a band ascribed to Er3+. There are 6 main REE related absorption clusters for the Mt Malosa sample (Figure 4.12 and 4.13). Cluster 1 shows moderate absorption bands at 576 and 582 nm. Cluster 2 is a series of three weak bands at 651, 660 and 681 nm. Cluster 3, the first of three prominent absorption clusters, and has strongest band minima at 739 and 750 nm with a weak minimum at 757 nm and two shoulders. Cluster 4 has a prominent band at 808 nm, exhibits shoulders at shorter and longer wavelengths, and also shows two weak and noisy absorptions at 835 and 844 nm. Cluster 5 is characterized by three absorption bands with minima at 870, 880 and 893 nm. Cluster 6 is a single band located at 978 nm. SWIR As in the VNIR range, three spectral classes of zircon are evident: U bearing (North Burgess, Green River, Mudtank), high REE with U (Mt Malosa) and metamict (St Peters Dome). All absorption features in the Mudtank and North Burgess spectra can be related to U4+, U5+ and OH/H2O (Table 4.12, Figures 4.12 and 4.13). Similarly, the Green River sample has a nearly identical set of absorptions related to U4+ and U5+, but also shows a few subtle REE3+-related absorptions. The Mt Malosa sample shows a collection of REE-related bands but also shows strong influence from the strongest U-related absorption clusters, G and I. The St Peter’s Dome zircon is the most featureless, however, weak absorption bands occur at all the appropriate wavelengths for U and sometimes the REE, suggesting low concentrations of REE, U4+ and U5+ in a host matrix that is poorly crystalline. Its strongest bands occur at 1415, 1924 and 2206 nm, all related to H2O / OH. There are 8 clusters clearly seen in the Green River, Mudtank and North Burgess samples that are associated with uranium (labeled F through M). Cluster F is a band at 1061 nm with a shoulder at 1010 nm. Cluster G shows an absorption minimum at ~1118 nm related to U5+ and another band near 1143 nm related to U4+ that is expressed as a shoulder. Cluster H shows two distinct and overlapping bands at 1326 and 1345 nm. Cluster I is similar and shows a strong U5+ related absorption minimum at 1503 nm with adjacent weaker U4+ related bands at 1478 and 1534 nm that are influenced by Er3+ and Sm3+ absorptions. Cluster J is a distinct minimum at 1660 nm and a second at 1704 nm. Cluster K is a weak shoulder near 1792 nm and Cluster L is a prominent single band located near 2074 nm. Cluster M is a weak shoulder at 2187 nm, however, it is not apparent in the Green River spectrum. 131 Using the Mt Malosa sample as reference, another 7 clusters are distinguished for zircon in the SWIR range. Cluster 7 is a pair of absorption minima at 1086 and 1105 nm. Cluster 8 is a strong absorption feature with two minima at 1244 and 1263 nm flanked by shoulders at 1200 and 1288 nm. Cluster 9 includes two absorption minima located at 1389 nm and 1415 nm. Cluster 10 overlaps with the uranium-related “Cluster I” and includes shoulders at 1427 and 1528 nm, plus additional absorption minima at 1560 and 1616 nm. Cluster 11 is a weak minimum at 1691 nm and a more moderate minimum located at 1729 nm. Cluster 12 comprises two overlapping bands: low REE samples have minima near 1924 nm while the high REE sample (Mt Malosa) has a shoulder at 1917 nm and a minimum pushed out to 1943 nm. Absorption features related to H2O / OH are best exhibited by the St Peters Dome sample. Strong bands are located at 1415 nm, 1924 nm and 2206 nm. Finally, for Cluster 13 a number of absorption minima and shoulders are recorded from ~2205 nm out to 2530 nm, however, at this point they are tentatively ascribed to combinations and overtones related to H2O, OH and the variably metamict host lattice (i.e., Zr, Si, U, REE). 4.8.5 Eudialyte reflectance spectra Eudialyte, Na15Ca6(Fe2+,Mn2+)3Zr3Si[Si25O73](O,OH,H2O)3(OH,Cl)2, has a flexible and complicated crystal structure and can show varied chemistry. The samples from Kipawa and Mt St Hilaire (MSH) can be split into two groups based on their ratio of LREE to HREE (“L/H”) but generally share most spectral features. The Kipawa samples show enrichment of HREE (Tb to Lu, Y) whereas the remainder of the eudialyte samples from MSH show a greater occupation by LREE (La to Gd). One exception, Sample CMNOC478 from MSH shows an intermediate distribution of REE. Consequently, the high LREE samples will be dominated by spectral features of Nd3+, Pr3+ and Sm3+, while the high HREE samples will be dominated by spectral features of Dy3+, Er3+, and Yb3+ but influenced by Nd3+. In the LREE enriched group sample CMN88-79 (Pinch) shows the highest concentration of LREE and in the HREE group sample Kipawa-Mariano shows the greatest concentration of HREE (and smallest LREE/HREE value). Notable, however, is that sample CMNOC37104 shows the largest LREE/HREE value but only carries 4.07 wt.% REE2O3. The high HREE samples are also bright red to pink in colour indicating a strong presence of IVFe2+, however, all samples have approximately the same amount of iron with the exception of CMNOC2045 (6.55 wt.% FeO). IVFe2+ will show absorptions near 530 and 1365 nm while VFe2+ will show absorptions near 917 and 2500 nm. The LREE samples also show higher Mn content, however, the electronic transitions will only affect the region from ~345 to 535 nm. 132 Band index tables for eudialyte in the VNIR and SWIR ranges have been split into the two groupings as defined by the LREE/HREE ratio. This ratio is given in each table. The maximum “peak” reflectance in the region between 1050 and 1400 nm is also given and described in the text. Note that the wavelength range of each cluster is the same for each group, but the presence and location of the individual features varies. Figure 4.14 Reflectance spectra of eudialyte samples in the VNIR (550 – 1000 nm) and SWIR (975 – 2530 nm) ranges. Clusters are indicated by labeled thick horizontal lines and prominent absorptions for sample “Kipawa-Mariano” (black spectrum) are labeled with wavelength position. 133 Figure 4.15 Stacked reflectance spectra of eudialyte samples in the VNIR (550 – 1000 nm) and SWIR (975 – 2530 nm) ranges. Clusters are indicated by labeled thick horizontal lines and prominent absorptions highlighted in the index tables are labeled with wavelength position. The stacked VNIR spectra are ordered by LREE:HREE ratio – upper samples above CMNOC478 have HREE enrichment greater than 1. The stacked SWIR spectra are ordered by the position of the reflectance maximum between 1050 and 1400. Clusters are repeated for the different groupings of samples. 134 Table 4.13 Prominent absorption features of HREE-enriched eudialyte samples in the VNIR and SWIR ranges. Eudia lyteL/HCluster Abs Shape Abs Shape Abs Shape Abs Shape1 576 SH ‐ n 576 MIN ‐ n Nd584 SH ‐ n 584 MIN ‐ n Nd2 651 MIN ‐ w 651 MIN ‐ w 651 MIN ‐ w Er, U?661 MIN ‐ w, n 661 MIN ‐ w, n 661 MIN ‐ w, n Er680 SH 680 SH 680 SH Er3 733 SH 735 MIN 734 MIN 735 MIN ‐ w Dy, Nd741 MIN ‐ s t 745 MIN ‐ s t 745 MIN ‐ s t 745 MIN ‐ w Dy, Nd750 MIN 752 MIN 752 MIN 754 MIN ‐ w Dy, Nd760 SH Dy4 800 MIN ‐ s t 800 MIN ‐ s t 800 MIN ‐ s t 800 MIN Nd, Dy, Er808 MIN 809 SH 809 SH 804 MIN ‐ w Nd, Dy, Er824 SH Nd, Dy, Er5 865 SH 865 SH 865 SH 865 SH ‐ w Nd, Dy873 MIN 872 MIN 872 MIN 872 SH ‐ w Nd, Dy880 MIN Nd, Dy888 SH ‐ b 888 SH ‐ b 888 SH ‐ b Dy910 MIN 910 MIN ‐ b 910 MIN ‐ b Dy6 974 MIN ‐ s t 974 MIN 975 MIN Er, YbVFe2+ ‐ ‐ 914 MIN ‐ b 914 MIN ‐ b 914 MIN ‐ b VFe2+7 1061 MIN 1067 SH 1067 SH Dy, Sm, U?1099 SH 1105 SH ‐ w Dy8 1168 SH 1168 SH 1168 SH Dy, U?1213 SH 1213 SH 1213 SH 1200 SH ‐ w Dy, Sm1276 MIN 1276 MIN 1276 MIN 1263 SH Dy, Sm1358 SH 1358 SH 1358 SH Dy, U4+?1408 MIN 1408 MIN 1408 MIN H2O / OH, Sm1433 MIN ‐ s t 1440 MIN ‐ s t 1433 MIN ‐ s t 1433 MIN ‐ s t H2O / OH1478 SH 1478 SH 1478 SH 1471 MIN Pr, Er, U4+?1534 MIN 1534 MIN 1534 MIN 1540 SH Er, Sm, U4+?1566 SH ‐ w 1566 SH ‐ w 1566 SH ‐ w Er9 Nd1811 SH 1811 SH 1811 SH 1792 SH U4+?1930 MIN ‐ s t 1930 MIN ‐ s t 1930 MIN ‐ s t 1924 MIN ‐ s t Pr, Sm, H2O2074 SH ‐ w, b 2093 SH ‐ w, b 2080 SH ‐ w, b U4+?10 2193 SH 2193 SH 2193 SH 2193 SH OH/REE/Mn‐FeOH/REE/Mn‐FeOH/REE/Mn‐Fe2312 SH 2312 SH 2318 MIN ‐ w 2312 MIN ‐ w OH/REE/Mn‐FeOH/REE/Mn‐Fe2437 MIN 2443 MIN 2437 MIN 2437 SH OH/REE/Mn‐FeOH/REE/Mn‐Fe2493 SH 2493 SH 2493 SH 2474 MIN OH/REE/Mn‐FeRefl . Peak 1118 1156 1143 1320Kipawa ‐Mariano Kipawa ‐ UofA Kipawa F92‐23 CMNOC 478 Probable Origin2.70.4 0.4 0.5135 Table 4.14 Prominent absorption features of LREE-enriched eudialyte samples in the VNIR and SWIR ranges. Eudia lyteL/HCluster Abs Shape Abs Shape Abs Shape Abs Shape Abs Shape1 576 MIN ‐ n, w Nd582 MIN ‐ w, n 581 MIN ‐ w, n 593 SH ‐ n Nd2 656 MIN ‐ w, n Er, U?ErEr3 735 SH Dy, Nd737 SH ‐ v w 739 MIN ‐ s t 740 MIN ‐ s t 740 MIN 740 MIN Dy, Nd753 SH ‐ v w 751 MIN 752 MIN 750 MIN 751 MIN Dy, NdDy4 801 MIN 798 SH ‐ n 796 MIN 795 MIN Nd, Dy, Er810 MIN ‐ s t 807 MIN ‐ s t 808 MIN 807 MIN Nd, Dy, Er826 SH 824 SH ‐ n Nd, Dy, Er5 869 SH 866 MIN 864 MIN 863 MIN Nd, DyNd, Dy880 MIN 880 MIN 880 MIN 879 MIN Nd, DyDyDy6 Er, YbVFe2+ 920 MIN ‐ b 916 MIN ‐ b 886 MIN ‐ b ‐ ‐ ‐ ‐ VFe2+7 1061 SH ‐ w 1061 SH ‐ w 1061 SH ‐ w 1061 SH ‐ w Dy, Sm, U?Dy8 1162 SH 1162 SH 1162 SH 1162 SH Dy, U?1200 SH ‐ w 1213 MIN ‐ w 1206 MIN ‐ w 1200 SH 1200 SH Dy, Sm1269 SH 1263 SH 1251 SH 1251 SH Dy, Sm1351 SH ‐ w 1351 SH 1351 SH 1351 SH 1351 SH Dy, U4+?1408 SH 1408 SH 1408 SH H2O / OH, Sm1440 MIN 1440 MIN ‐ s t 1440 MIN ‐ s t 1427 MIN ‐ s t 1433 MIN ‐ s t H2O / OH1478 MIN ‐ w 1478 SH ‐w 1478 SH 1478 SH 1478 SH Pr, Er, U4+?1553 SH Er, Sm, U4+?Er9 1729 SH Nd1817 SH 1817 SH 1805 SH 1805 SH U4+?1936 MIN ‐ s t 1930 MIN ‐ s t 1930 MIN ‐ s t 1930 MIN ‐ s t 1930 MIN ‐ s t Pr, Sm, H2O2087 MIN ‐ w, b 2093 SH ‐ w, b 2087 SH ‐ b U4+?10 2193 MIN ‐ w, b 2193 SH OH/REE/Mn‐Fe2230 MIN ‐ w 2237 SH 2249 MIN ‐ s t 2256 MIN ‐ s t OH/REE/Mn‐Fe2274 SH ‐ w 2268 SH 2274 MIN ‐ w OH/REE/Mn‐Fe2324 MIN ‐ w 2312 SH 2312 MIN 2312 MIN ‐ w 2305 SH ‐w OH/REE/Mn‐Fe2349 SH 2330 SH ‐ w OH/REE/Mn‐Fe2437 MIN ‐ s t 2443 MIN ‐ s t OH/REE/Mn‐Fe2462 MIN OH/REE/Mn‐Fe2487 SH 2487 SH 2480 MIN ‐ s t 2480 MIN ‐ s t OH/REE/Mn‐FeRefl . Peak 1307 1232 1307 1307 13076.2CMNOC 476CMNOC 20454.6CMN 88‐79 Pinch CMNOC 37104 Probable Origin10.69.58.5CMN 72‐24136 For describing variability between clusters, two groups are considered. The “LREE Group” comprises samples CMNOC2045, CMNOC476, CMN88-79 (Pinch), CMN72-24 and CMNOC37104, while the “HREE Group” comprises all three Kipawa samples (-Mariano, UofA and F92-23) as well as CMNOC478 because of its moderate Dy content despite having LREE/HREE>1. Band index tables were split accordingly. VNIR The spectra for eudialyte in the VNIR range (Tables 4.13 and 4.14, and Figures 4.14 and 4.15) are divided into 6 clusters for all samples, however, not all samples display all spectral features. Cluster 1 comprises two absorptions near 576 and 584 nm in a region that is typically steep sloped with minor noise. It is only seen in higher LREE samples. Cluster 2, prominent in the HREE group, is also located in a region that is typically steep sloped and consists of a main absorption band at 651 nm with a weak minimum at 661 and shoulder at 680 nm. Cluster 3 in the HREE Group comprises a central absorption minimum near 745 nm with flanking local minima near 735 and 752 nm. In the LREE Group this cluster is a doublet characterized by a moderate minimum near 739 nm followed by another minimum near 751 nm. Cluster 4 in the HREE Group is a strong absorption minimum at 800 nm followed by a weaker minimum near 808 nm. In the LREE Group this cluster is expressed as another doublet with a typically weaker band near 798 nm and a stronger band near 807 nm. Cluster 5 is most prominent in the HREE group and consists of a principal absorption minimum near 873 nm and a broad absorption at 910 nm with several other weak intermediate shoulders. In the LREE group it is characterized by two minima near 866 and 880. Cluster 6 is only seen in the HREE Group and consists of a single strong absorption at 974 nm. In addition to the 6 clusters, the approximate band center is given for samples that show evidence of VFe2+ in the form of a very broad absorption centered near 915 nm. SWIR The spectra for eudialyte in the SWIR range are divided into 4 clusters (Tables 4.13 and 4.14, and Figures 4.14 and 4.15). Cluster 7 includes a weak absorption near 1061 nm typically expressed as a shoulder, followed by a weaker shoulder near 1099 nm. Cluster 8 is the most prominent absorption with the most complexity and stretches from ~1115 to 1615 nm. The deepest absorption for all samples occurs near 1433 nm. The LREE Group samples are then characterized by a series of shoulders at both shorter and longer wavelengths. Notably, CMNOC2045, CMNOC476 and CMN88-79 (Pinch) each exhibit an absorption at 1408 nm expressed as a shoulder while CMNOC37104 and CMN72-24 do not show this absorption band at all. In the HREE Group additional distinct absorption minima occur at 1276, 1408 and 1534 nm as well as more subtle shoulders across the whole cluster. This region is also the general location of a documented IVFe2+ related band centered near 1366 nm (Polshin et al. 1991), however, this is hard to 137 objectively observe in our spectra. Cluster 9 is a sharp absorption band at 1930 nm along with several shoulders. Cluster 10 ranges from 2150 nm out to the spectrometer’s full range (2530 nm). In the LREE Group the strongest absorptions occur near 2437 and 2480 nm while the HREE Group has its strongest band near 2437 nm with a consistent shoulder at 2493 nm. Within the LREE Group, samples CMN72-24 and CMNOC37104 also have an additional distinct absorption at 2249 nm. In addition to the 5 clusters in the SWIR, each eudialyte also shows a maximum reflectance between 1050 and 1400 nm. 4.9 Discussion on spectra and spectral variations 4.9.1 Compilation of REE‐bearing silicate spectra The REE-bearing silicates studied here show a large diversity of overall crystal structures, chemical compositions and host sites for Ln3+. The mineral suite included samples that are LREE-dominant (e.g., cerite), HREE-dominant (e.g., kainosite), and less selective but still with high REE content (e.g., mosandrite). The REE were incorporated as trace elements (e.g., zircon), major substitutions (e.g., eudialyte) and structural components (e.g., kainosite). The mineral suite also included samples with structural H2O, structural OH, structural CO3, unusually coordinated Fe, and probable extensive radiation damage. Figure 16 is a compilation of representative stacked reflectance spectra from the suite of minerals documented and described and displays the wide range of expected spectral variability. Similar to the REE phosphates, a mineral’s preference for LREE or HREE and resulting distribution in a given sample will define which spectrally active lanthanides are present, and therefore what the overall 4f-4f transition-related spectral pattern will be. Of the REE silicates, cerite is a good representation of LREE-enrichment, kainosite represents HREE-enrichment, and mosandrite displays a good ‘mixed’ signal. The two zircon samples shown in Figure 16 represent high U and high REE examples and display sharp and diagnostic spectral features, however, it is notable that the overall contents of U and REE in these samples is actually quite low (up to 0.08 wt.% UO2 and up to 4.83 wt.% REE2O3). This is in contrast to kainosite (36.93 wt.% REE2O3) and eudialyte (sample 37104, 0.07 wt.% UO2) with significantly weaker to absent spectral features. The range of eudialyte compositions also provides an opportunity to investigate the effects of LREE vs HREE enrichment in different samples of the same mineral species. 138 Figure 4.16 Stacked reflectance spectra from representative REE-bearing silicate minerals in the VNIR (500 to 1000 nm) and SWIR (975 - 2530 nm) ranges. Clusters as described in text are indicated by labeled thick horizontal lines with prominent absorptions identified by tick marks. 139 4.9.2 Spectral patterns of eudialyte The eudialyte samples roughly form 2 groups based on the LREE/HREE value. Sample CMNOC478 is LREE enriched but also shows the most Dy content of that group. The larger number of samples for this mineral allows for greater intramineral comparisons, however, its flexible and complex crystal structure also allows for significant variations in reflectance spectroscopy. The following comparisons are made mostly based on the two LREE/HREE groups (Figures 4.14 and 4.15). The morphology of the absorption clusters near 745 and 800 nm allow LREE (Nd) rich samples to be distinguished from HREE (Dy, Er, Ho) rich samples. Although there are overlapping absorptions for these elements, the Nd3+ signal in eudialyte is expressed as two ‘doublets’. Near 745 nm, the shorter of the two wavelength absorptions (741 and 752 nm) is usually stronger, and near 800 nm the longer of the two wavelength absorption (800 and 810 nm) is usually stronger. For the HREE enriched samples, Dy3+, Er3+ and Ho3+ absorptions combine so that near 745 nm there is a central absorption at or close to 745 nm flanked by two absorption minima. At 800 nm, one observes an asymmetrical cluster with a central strong absorption band followed by several weaker absorptions at longer wavelengths that are typically expressed as shoulders or local minima out to ~825 nm. The HREE group also show a Dy3+ related absorption at 910 nm and an Er3+-Yb3+ related absorption at near 974 nm. Sample CMNOC2045 shows the least resolved LREE features in the VNIR, however, this is not due to lower Nd but rather its high proportion of VFe2+ (6.55 wt.%) that causes a very strong absorption band centered near 920 nm and which stretches from ~700 nm out past 1000 nm (see Figure 14, unstacked reflectance). This sample also shows lower reflectance in the SWIR, particularly at longer wavelengths. The lowest Nd-bearing sample in the LREE group is sample CMNOC476, which actually shows very nice Nd3+-related features owing to iron predominantly in square planar coordination (IVFe2+) and therefore less absorption (greater reflectance) from ~700 to 1000 nm that allows for more contrast in the regions where Nd3+ has its absorption bands. Unsurprisingly, this sample shows red colouration due to its high IVFe2+. These two spectra emphasize the impact that iron’s coordination state has on reflectance spectra. For all eudialyte samples there is a reflectance maximum in the SWIR somewhere between 1050 and 1400 nm (Figure 4.15). Qualitatively, this is related to the influence of Dy3+ (and to a lesser degree Sm3+) which has its strongest absorption near 1290 nm and a series of absorptions at shorter wavelengths back out to ~1085 nm. It is also due to the influence of square planar IVFe2+, which has a prominent broad absorption centered at 1366 nm (Polshin et al. 1991). In our sample suite it conveniently divides the high HREE and high IVFe2+ samples from the high LREE and high VFe2+ samples based on the location of the 140 reflectance maximum – the samples from Kipawa (HREE, IVFe2+) show maxima below 1200 nm while the MSH (LREE, VFe2+) samples show maxima beyond 1200 nm and typically beyond 1300 nm. The absorption at ~1433 nm (6978 cm-1) within Cluster 8 is attributed to H2O and/or OH in the crystal structure as it is present across all samples in a similar manner as the ~1930 nm absorption. The absorption at 1408 nm (7102 cm-1) is present as a local minimum in the samples from Kipawa, a shoulder in samples CMN88-79, CMNOC476 and CMNOC 2045, and is absent in CMN72-24, CMNOC37104 and CMONC478. In this region there is the possibility of a narrow Sm3+-related absorption, however, the Kipawa samples have lower Sm than the MSH samples, which do not display this resolved absorption band. It is possible the absorption at 1408 nm suggests that (1) an additional and distinct H2O or OH-related absorption is present in the high HREE Kipawa samples, (2) an H2O or OH-related absorption near 1408 nm in the MSH samples is broadened to the point that it cannot be resolved, or (3) that the H2O or OH-related absorption near 1408 nm is shifted in the MSH samples to a longer wavelength that overlaps with the 1433 nm absorption. The simplest conclusion would be either (2) or (3), however, additional infrared spectroscopy and crystal structure studies would be needed to satisfactorily resolve this question. Notably, this pattern also parallels the shift in reflectance maximum (Figure 4.15). A notable variation within these groups is a subclass of high Mn / low Fe samples (CMN72-24 and CMNOC37104) within the LREE enriched samples. Samples 37104 and CMN72-24 have district variations to their spectra. They do not show strong absorptions related to IVFe2+ or VFe2+, which is consistent with their low Fe contents. They also show the highest LREE/HREE values, highest Mn and highest U. In the SWIR range, the absorption band at 1408 nm is missing, and they possess strong bands at 2249 nm that are unlike any other samples. If one applies the results of this study to the five eudialyte spectra in Figure 1 from the Red Wine Complex in Kerr et al. (2011), their samples would likely be LREE enriched and IVFe2+ dominant. In a separate report, Kerr (2011) reviews the mineralization at the Red Wine Complex and provides a chondrite-normalized plot of eudialyte ore that supports this interpretation, as well as photographs of pink eudialyte. 4.9.3 The Er‐Yb related absorption near 978 nm The ~978 nm absorption due to Er3+ and Yb3+ is proving to be an important factor for mineral identification of HREE bearing samples because of its sensitivity to the mineral host. High Er and Yb alone will not drive the absorption near 978 nm – “fit” of Er3+ and Yb3+ in the substitutional site appears to play a strong role, and asymmetry and ligand identity are likely factors as well. In Figure 4.17 eudialyte and mosandrite spectra show comparable absorption strengths and Er+Yb concentrations. Zircon samples 141 for the displayed spectra host considerably lower Er and Yb contents yet still provide discernable absorptions. Kainosite, however, contains a much greater amount of Er and Yb than all of these samples but shows only a poorly resolved and weak absorption band. To get an idea of how well Yb3+ and Er3+ fit into the various sites, we can refer to Table 4.1 and Figure 4.2 for cation coordination environments, Shannon (1976) for ionic radii, and the various microprobe results for ‘true’ site occupancies (see Table 4.15). For mosandrite, REE3+ are hosted in three sites with mixed cation populations. The M4 and M5 sites are in 7-fold coordination with 6 oxygen and 1 OH- and can be thought of as being dominated by Ce3+ and Ca2+, while the third site, M3, is in 6-fold coordination with 2 oxygen and 4 mixed anion sites and in our sample is dominated by Na+. In eudialyte REE3+ are hosted in two sites with mixed cation populations, M1 and Na4. The M1 site is in 6-fold coordination with oxygen and is dominated by Ca2+ while the second cavity-like site (Na4) is generally in 9-fold coordination with mixed anions and occupied by Na+. In zircon REE3+ are hosted in the ZrO8 dodecahedron. In kainosite REE are hosted in a designated REE site in 8-fold coordination with oxygen. If we compare the valence charge and ionic radius of the “normal” cation site occupants (e.g., ionic radius for VIIIZr4+ is 0.84 Å, see Chapter 1) against the character of Yb3+, which has the second smallest ionic radius of the REE, a first approximation of “fit” is made (Table 4.15). Note that this does not take into consideration the arrangement of the 4f orbitals within the distorted crystal field, which if accounted for would provide a greater assessment of “fit”. Table 4.15 Cation site parameters for Yb3+ in REE-bearing silicates. Kainosite Zircon Mosandrite Eudialyte Cation Site Y Zr M4, M5 M3 M1 Na4 Normal Occupant Y≈HREE Zr Ca≈Ce Na Ca Na Normal Valence Charge +3 +4 ~+2.5 +1 +2 +1 Coordination # 8 (VIII) 8 (VIII) 7 (VII) 6 (VI) 6 (VI) 9 (IX) Ionic Radius(Å) ~1.002 0.84 ~1.065 1.02 1.00 1.24 Coordinated Anions 8×O 8×O 6×O, 1×OH 2×O, 4×Mixed 6×O "Cavity" Yb3+ Substitution Valence Charge 3 3 3 3 3 3 Ionic Radius (Å) 0.985 0.985 0.925 0.868 0.868 1.042 Charge Imbalance (Yb3+ - MX+) 0 -1 0.5 2 1 2 Yb3+ radius / MX+ Radius 0.98 1.17 0.87 0.85 0.87 0.84 *Ionic radii from Shannon (1976) 142 From Table 4.15 it can be seen that Yb3+ is a particularly poor fit for the Zr site in zircon, as well as in the M3 site in mosandrite and the Na4 site in eudialyte. Kainosite, as expected, shows a very good fit for Yb3+ in the Y3+ site. Descriptively for zircon, VIIIYb3+ is forced into a polyhedron where it has a larger ionic radius than the “displaced” and more strongly charged VIIIZr4+. The resulting absorption strength relative to Er+Yb concentration (i.e., the absorption coefficient) is thus relatively high because of the “misfit”. Modeling of 4f orbitals in the true shapes of the coordination polyhedra with consideration for ligand type would provide a better understanding of this behavior and perhaps provide a more quantitative relationship. For minerals that can be either LREE or HREE enriched, this Er3+-Yb3+ related absorption can be quite useful. For example, it has been shown that the Er3+-Yb3+ absorption in eudialyte can be used to discriminate high LREE vs high HREE samples. This absorption also exhibits wavelength shifts that may be exploitable for mineral identification (974 nm in eudialyte, 976 nm in mosandrite, ~978 nm in kainosite and 978/979 nm in zircon) although additional spectra of these minerals would provide greater confidence. Similar patterns are evident in the reflectance spectra of the REE-phosphate minerals monazite, xenotime and britholite. Figure 4.17 Selected samples showing the Er3+-Yb3+ related absorption band near 978 nm. Spectra are labeled with sample name and Er and Yb contents in parentheses (Er2O3 wt.% / Yb2O3 wt.%). 143 4.9.4 Absorption band variations amongst REE‐bearing silicate spectra Variations are observed for several of the REE3+-related absorption bands amongst the various minerals. Some of these variations are nearing the spectral resolution of the spectrometer making them hard to distinguish from noise and others are located on the flanks of steeper slopes of the spectra making their ‘shift’ perhaps just an artifact of the overall continuum. Nevertheless, some clusters of absorptions show strong changes in both relative strengths between related absorptions and band center positions. Figure 18 shows an example of this in the VNIR, but greater investigations are warranted with an expanded set of REE bearing minerals. Figure 4.18 Reflectance spectra (left) of selected REE bearing silicates and their continuum removed spectra (right) displaying relative intensity and positional differences for Nd3+ related absorptions centered at ~746 nm (4I9/2 4F7/2+4S3/2), ~803 nm (4I9/2 4F5/2+2H9/2) and 875 nm (4I9/2 4F3/2). Influence from Dy3+ in these samples is minimal but would be greatest in the ~803 nm cluster. Weight % of Nd2O3 for each sample is given in parentheses. 144 4.10 Implications We report the first reflectance spectra for cerite, mosandrite, and kainosite. These spectra are accompanied by microanalytical characterization. We also report the first set of reflectance spectra for eudialyte with supporting microanalytical characterization. Although reflectance spectra of zircon are available in the literature, we report the first systematic investigation of U-enriched and REE-enriched samples with accompanying microanalytical characterization. The REE bearing silicate minerals of this study display spectral characteristics in the VNIR-SWIR range that enable their discrimination and would thus allow for their automated detection and recognition. This includes the notable observation that some specific REE3+-related absorption bands can undergo wavelength shifts and changes in relative intensities between different minerals. A compilation of representative reflectance spectra and their absorption clusters is given in Figure 4.16. The reflectance spectrum for cerite is dominated by spectrally active LREE, which is consistent with microprobe compositions. The LREE-enriched fluorocarbonate mineral bastnaesite was identified in the imagery of the cerite hand sample, providing confirmation that two minerals with similar REE concentrations can be distinguished from one another. The reflectance spectrum for mosandrite is dominated by spectrally active LREE with influence from Er3+ and Yb3+, which is consistent with microprobe compositions. The reflectance spectrum for kainosite is dominated by spectrally active HREE, which is consistent with microprobe data, however, the 2055 nm absorption is unusually strong. This band is aligned for potential input from Pr3+, Eu3+, Tb3+, and Ho3+ electronic transitions but none of these elements are present in any great amount. Another possibility is U4+ despite EMPA results below the detection limit of 0.075 wt.% UO2, however, the zircon samples with U below detection still show strong U absorptions. Finally, the formula for kainosite includes structural water, SiO4 tetrahedra and carbonate radicals which could provide the appropriate vibrational combinations or overtones in this region. Zircon is known for hosting uranium and the reflectance spectra from various samples exhibit strong absorption bands for U4+ and U5+ despite generally low concentrations (<750 ppm). The high REE and low U sample from Mt Malosa and the low REE and high U sample from Green River provide excellent complementary baseline spectra for zircon. The eudialyte sample suite showed a range of compositions and comprised nine samples from two localities. Reflectance spectroscopy in the VNIR-SWIR range was able to distinguish LREE-enriched from HREE-enriched samples despite overlapping VNIR absorption clusters related to Nd3+ (LREE Group) and Dy3+ (HREE Group). Some variability was seen in absorption minima wavelength positions, 145 which is possibly tied to the complex structure of this diverse mineral group. Evidence of IVFe2+ and VFe2+ was observed in the VNIR and SWIR ranges, and in concert with Dy3+-related absorptions in the SWIR has a large influence on the location of maximum reflectance between ~1050 and 1400 nm. A larger and more diverse sample set would provide greater confidence for using this metric as a screening method for reflectance spectroscopy in the field. 146 Chapter 5. REE Ore Grade Estimation of Drill Core by Imaging Spectroscopy 5.1 Introduction Early work confirmed that the strength of REE-related absorptions in a target pixel’s spectrum was at least roughly correlative with the concentration of specific lanthanides. An opportunity arose to obtain several boxes of mineralized drill core from a REE ore deposit in the advanced stage of development. The samples originated from various areas of the deposit, including multiple ore zones with differing mineralogy, which is characterized as a peralkaline intrusion that underwent pervasive ore-forming hydrothermal alteration. Despite the deposit being geologically complex, the most important carriers for REE are generally reported to be monazite, allanite, synchysite and bastnaesite for the LREE and zircon and fergusonite for the HREE. Mineralization is generally fine grained (<1 mm) with zircon as small as ~30 μm in diameter, well below the size of the spectrometer’s pixels but well above the wavelengths used for imaging. These samples formed the basis from which to study whether REE ore grade could be estimated from hyperspectral imagery. . References to previous geological and mineralogical studies of the ore deposit have been removed due to confidentiality at this point in time, however, academic literature was only used to guide interpretations and research conducted is not compromised by this contextual omission. 5.2 Rock descriptions, mineralogy and geochemistry Understanding the rocks being studied was key to having confidence in the results. Two boxes from a single drillhole (~1/3 sawn, PQ diameter) at different depths were selected for detailed investigation: Box 104 (Upper Zone, 151 m depth) and Box 120 (Lower Zone, 173 m depth). The previously sampled drill cores were divided into 1 m sections by the company and were accompanied with geochemical assays (see Appendix E) and core log descriptions (pseudomorphs refer to zircon after likely eudialyte): “Box 104 (151-152m), 17021.9 ppm TREE+Y: precursor changes to coarse grained foyaite with randomly oriented kspar laths decreasing in size downhole. Pseudomorphs are mostly well preserved with some slightly stretched to form wavy/elongated stringers (eg. 150.2m). Pseudomorphs start to become mosaic downhole” 147 “Box 120 (173-174m), 21908.2 ppm TREE+Y: coarse, mosaic/poikilitic pseudomorphs with abundant aegirine needles inclusion, aegirine altered to magnetite and hematite” The one (1) meter sections were subsampled lengthwise at the University of Alberta into 7 (Box 104) and 6 (Box 120) paired pieces using a diamond saw, weighed and imaged using the sisuROCK instrumentation at the University of Alberta’s CoreSensing Facility (Figures 5.1, 5.2 and 5.3). After imaging, the smaller halves of the subsample pairs (averaging 76 grams each) were sent to ALS Chemex in North Vancouver, a commercial assay facility, for robust 41 element whole rock geochemistry (See Appendix E for analytical codes and full dataset). The remaining pulps were then sent to the Microbeam Lab at the University of British Columbia for X-ray powder diffraction studies and Rietveld refinement to determine modal mineralogy (see Appendix F). In short, the dataset comprised: Two 1-meter drill core lengths with geochemistry and geology determined by the company that were subsequently subsampled into 6 and 7 paired pieces Each subsampled piece was imaged with high spatial-spectral resolution imaging spectroscopy The small halves of each subsampled piece were destructively tested for geochemistry Each pulp from the geochemical testing was run for quantitative modal mineralogy Figure 5.1 Schematic of sawn core sample. Subsampling followed the red dashed line. Hyperspectral imaging included the original top surface of the large half and the newly exposed portion of the small half. The smaller portion was sent for destructive chemical and mineralogical testing. 148 Figure 5.2 Boxes 104 and 120 of pre-sampled drill core, PQ diameter (85 mm core diameter). Note lengthwise saw marks for subsampling during this study and assay tags indicating original 1 meter assays by the company. Figure 5.3 Example of dry drill core as originally acquired (photograph, left), wet drill core during subsampling (photograph, middle), and hyperspectral image of subsampled drill core (VNIR false colour, right). Box 104 – Subsample 3. 149 From the new geochemistry, subsamples in both boxes ranged from unmineralized to strongly mineralized with total rare earth element content (TREE+Y) from 380 ppm to 3.7 %. Strong positive correlations between TREE+Y, Nb, Ta and Zr were seen, indicating that these elements occur in tandem. Weighted averages for the REE from the subsamples matched the REE assays from the company with an R2 value of 0.9999 (Figure 5.4), an acceptable fit, and indicating that the volume of the smaller halves from subsampling are a very good geochemical approximation of the overall volume of core previously sampled by the company (~2/3 of the core). Interestingly, the weighted averages from Box 104 showed a linear regression less than 1:1 whereas Box 120 showed a linear regression greater than 1:1. The reason for this is not known, but the answer is likely tied to which laboratory techniques were used for sample preparation and analysis, and which minerals host REE, especially refractory minerals. Figure 5.4 Weighted assay values from subsampling of the REE (x-axis) plotted against the REE assay values for the same interval as provided by the company. Rietveld refinement of the X-ray powder diffraction data from the 13 specially prepared pulps confirmed that the two boxes of drill core originated from different ore zones with similar but different modal mineralogy (Table 5.1). With respect to understanding the petrogenesis of the deposit and exploiting spectral features of the minerals in the VNIR-SWIR, notable differences include the absence/presence of chlorite, relative abundance of magnetite to hematite, and the type of carbonate (as well as the type of alkali feldspar, relevant to the longwave infrared region). With respect to working 150 towards REE ore grading, it is important to note that (1) the REE fluorocarbonate minerals (REEFCO3) are the dominant carrier of LREE, that (2) zircon (and possibly undetected but academically reported fergusonite) is expected to carry the HREE, and that (3) allanite is only present in subordinate amounts in 2 samples in Box 104. Very notably, no monazite (or xenotime) was modeled to be present. An accurate and detailed framework was thus created through the determination of both the modal mineralogy and the geochemistry of the 13 subsamples. Table 5.1 Modeled modal mineralogy from Rietveld refinements. Modeled Mineral Phases Box104 Average Box 120 Average Quartz low 25.2 40.1Microcline (ordered) 24.2 5.0Albite low 14.1 12.3Aegirine/Augite? 0.6 0.0Biotite 1M 8.1 9.5Clinochlore II 4.3 0.0Hematite 3.9 8.7Magnetite 7.6 2.8Ankerite 6.1 0.0Calcite, magnesian 0.0 8.6Dolomite 0.0 1.1Zircon 3.7 6.6Hydroxylbastnaesite-(Ce) 1.0 1.9Synchysite-(Ce,Y?) 0.3 0.0Parisite-(Ce) 0.2 0.0Allanite (Ce) 0.6 0.0Nanoscale clay 0.0 0.0Fluorite 0.0 3.5Modal Total (modeled) 100.0 100.0 5.3 Imaging spectroscopy and image processing Hyperspectral imaging spectroscopy of drill core was carried out using the sisuROCK instrumentation (manufactured by SPECIM Spectral Imaging Ltd.) at the University of Alberta’s CoreSensing Facility, and data was handled using ENVI 4.4. Two imaging spectrometers (“cameras”) acquired reflectance spectra in the visible-near infrared (VNIR, 396 nm to 1003 nm over 784 channels for an average spectral resolution of 0.77 nm, later subset to 600 – 1000 nm) and shortwave infrared (SWIR, 928 nm to 2530 nm over 256 channels for an average spectral resolution of 6.26 nm) portions of the electromagnetic spectrum in high spatial resolution mode. Spatial resolution of the cameras in this mode 151 was approximately 0.079 mm / pixel in the VNIR and 0.241 mm / pixel in the SWIR. Image cubes in the VNIR were resized via pixel averaging using a 3 by 3 grid, resulting in a resolution of 0.273 mm / pixel. Image cubes were cropped to size, ROI accurately hand drawn to isolate the subsamples, and regions outside the drill core were masked. Each 1 m section of drill core was represented by a 2.5 Gb file for the VNIR range and a 1.5 Gb file for the SWIR range. Absorption depths were calculated using simple arithmetic on reflectance values of spectra (not continuum removed spectra) by averaging three values of adjacent wavelengths at the bottom of the absorption (wavelengths at bands “b4”, “b5” and “b6”) and three values outside and at each side of the absorption (b1, b2 and b3, & b7, b8 and b9) (Figure 5.5). This was done to minimize the effect of noise on absorption depths. ( (b1+b2+b3)/3 + (b7+b8+b9)/3 ) / 2 - (b4+b5+b6)/3 For example, in Figure 5.5, spectral band wavelengths used to calculate the depth of the 741 nm Nd3+-related feature were at 707 (b1), 708 (b2), 709 (b3), 740 (b4), 741 (b5), 742 (b6), 769 (b7), 770 (b8), and 771 nm (b9). Figure 5.5 Example of absorption band depth calculation for a single pixel’s spectrum. Input bands for calculation are denoted with thin red lines on subset plot at right. Prominent Nd3+ related absorption locations are labeled. 152 Ten (10) absorptions and absorption combinations were evaluated in the VNIR range for Nd3+ and Er3+ and ten (10) in the SWIR range for Er3+, Sm3+, Pr3+ and Nd3+ with the assumption that absorptions for Ln3+ would be predominantly related to bastnaesite, parisite or synchysite (according to the Rietveld results). Absorption feature morphologies, such as that in Figure 5.5, are most consistent with the REE fluorocarbonates (see Chapters 2, 3 and 4), which is consistent with Rietveld refinement results.The calculations were carried out on the hyperspectral image cubes and generated new ‘output’ images, each representing an ‘absorption map’ for the specific absorptions. For the “Combinations” of absorption features (e.g., SUM_Nd_741_796_864), previously calculated absorption depth ‘output images’ were simply added together to produce a new image. “Image band” (not absorption band) statistics for each of these images were exported and the resulting histograms for each subsample then represent the distribution of absorption strength across all pixels (Figure 5.6). From the histogram (Figure 5.7), it can be recognized that strong absorptions lie far out on the X-axis, “absorption-less” pixels are represented by values less than ~0, and the transition between noise-induced irrelevant absorptions and weak relevant absorptions lies somewhere near 0.005. Thus, according to the histogram the bulk of the pixels do not show absorption at 796 nm. Concerning ourselves with only “significant absorptions”, an arbitrary threshold value can be set along the X-axis (absorption strength) above which we will consider all pixels to carry REE, and below which we will discard. The arithmetic sum of all the pixel values above the threshold value (“Sum Value Threshold”) normalized to the total number of pixels in the scene then results in a value termed here as the “Proxy Value”. An assumption is made that this Proxy Value should be representative of the total content of the lanthanides driving the particular absorption feature examined in that scene. This is done for each subsample (i.e., each scene) at the same threshold value (“Sum Value Threshold”) and a single linear regression is drawn between the Proxy Values and the geochemical results for the lanthanide in question (Figure 5.8). An R2 value is calculated from the linear regression and represents how well the Proxy Values agree with the geochemical values (Figure 5.8). This is repeated at various Sum Value Thresholds until a “best threshold” is determined with highest R2 value (Figures 5.8 and 5.9). 153 Figure 5.6 Example Nd3+-related absorption depth image (left) and histogram (right) for the 796 nm Nd3+ related absorption band. Colour ramp for image relates to relative strength of absorption (dark=no absorption, blue-green=moderate absorption, red-white=strong absorption). Histogram at right shows the distribution of absorption strengths for each pixel in its associated “image”. 154 Figure 5.7 Annotated histogram for the 796 nm Nd3+ related absorption band from Box120-Sample1. Vertical line represents a Sum Value Threshold of 0.005, above which the pixel’s values are summed (represented by the red area) and normalized to the total number of pixels in the scene (e.g., 57,777 pixels in VNIR) to generate a single Proxy Value for Box120-Sample1. Note that the displayed graph covers only absorption depths from 0 to 0.025. 155 Figure 5.8 Example of Proxy Values for three different Sum Value Thresholds (0.02, 0.0075 and 0.005) applied to the 796 nm absorption for Box 120 plotted against Nd geochemistry. R2 values are given for each linear regression. For example, using a threshold value of 0.02 for the 6 samples of Box120 establishes a weak linear regression between the resulting Proxy Values for each subsample and the subsample’s Nd content (R2=0.1362). Figure 5.9 Different “Sum Value Thresholds” and their resulting R2 values for the 796 nm Nd3+ related absorption for Box 120 subsamples. Note the maximum R2 value of 0.7873 associated with the best threshold value of 0.0075 (see Figure 5.8 and Table 5.2). Lower thresholds interpret noise as “signal” and higher thresholds exclude “true signals”. 156 5.4 Image processing results The process of generating proxy values and calculating linear regressions with the geochemical data for various thresholds was carried out for each of the absorptions and combinations on each set of subsamples (i.e., Box 104 and Box 120). This was also done on subsamples of both boxes together. The best R2 values and associated thresholds for absorptions and absorption combinations are given in Table 5.2. At least 10 absorption depth thresholds were tested for each absorption and combinations listed in Table 5.2 in order to locate the best R2 value. Subsample 5 of Box 104 was an outlier for most absorption calculations, consistently having a higher Proxy Value than it “should have” according to the geochemical results. There is nothing particularly notable about this sample’s modal mineralogy or geochemistry and the deviation is likely due to a relative overrepresentation of REE minerals on the cut and imaged surface. With that outlier removed (justified or not) the R2 values for both the Box 104 regressions and the regressions for both boxes together improves significantly. Table 5.3 tabulates the R2 values without the inclusion and influence of Box104-Sample5. Tables 5.2 and 5.3 have the cell bolded for the example below of the 796 nm Nd3+-related absorption in Box 120. 157 Table 5.2 Best R2 values and associated thresholds for linear regressions between absorption proxy values and REE concentrations. Box 104 (7 samples) Box 120 (6 samples) Both Boxes (13 samples) R2 on linear regression Abs Depth Threshold R2 on linear regression Abs Depth Threshold R2 on linear regression Abs Depth Threshold VNIR Range 626 (Nd3+) 0.5316 0.003254 0.7238 0.005 0.5268 0.001155 676 (Nd3+) 0.7491 0.005 0.4644 0.005 0.3965 0.005 741 (Nd3+) 0.7674 0.0015 0.9253 0.005 0.7454 0.003 796 (Nd3+) 0.8532 0.0005 0.7873 0.0075 0.7038 0.004 864 (Nd3+) 0.8001 0.003 0.9641 0.002722 0.8008 0.004261 889 (Nd3+) 0.7985 0.004 0.8663 0.003 0.3085 0.005 953 (Sm3+) 0.2091 0.004 0.5089 0.004 0.0713 0.003 978 (Er3+) 0.3020 0.004 0.5456 0.004 0.0227 0.003 SUM_Nd_676_741_796_864_889 0.8466 0.004 0.9729 0.005330 0.7219 0.004 SUM_Nd_741_796_864 0.8282 0.005434 0.9739 0.005 0.8408 0.002 SWIR Range 978 (Er3+) 0.1700 0.00125 0.3044 0.005 No correlation 1093-V003 (Sm3+) 0.8432 0.00075 0.7811 0.0005 0.6860 0.00125 1251 (Sm3+) 0.8722 0.002 0.8677 0.003228 0.5729 0.002114 SmSUM1093_1251 0.9241 0.00143 0.8866 0.003719 0.6927 0.001645 1402 (Pr3+) 0.6710 0.002291 0.7382 0.002 0.5081 0.001457 1459 (Pr3+) 0.6555 0.000313 0.6936 0.001714 0.5323 0.002 1553-V002 (Pr3+) 0.6481 0.000502 0.7501 0.00075 0.4180 0.001503 Pr Full Triple 0.5921 0.002 0.6794 0.005 0.3289 0.004 PrSUM1402_1459_1553_V002 0.7431 0.002212 0.8216 0.002 0.5988 0.001388 1710-V002 (Nd3+) 0.5057 0.001 0.7668 0.001 0.1941 0.0015 *Bolded “Sum Value Threshold” cells relate to the example used in previous figures, 796 nm Nd3+-related absorption *Underlined and bolded values relate to Figure 5.10 158 Table 5.3 Recalculation of thresholds and Best R2 values without sample Box104-5 (note that Box120 values are the same). Box 104 (6 samples) Box 120 (6 samples) Both Boxes (12 samples) R2 on linear regression Abs Depth Threshold R2 on linear regression Abs Depth Threshold R2 on linear regression Abs Depth Threshold VNIR Range 626 (Nd3+) 0.5920 0.003254 0.7238 0.005 0.5779 0.001155 676 (Nd3+) 0.7636 0.005 0.4644 0.005 0.3949 0.001518 741 (Nd3+) 0.9456 0.003 0.9253 0.005 0.8946 0.003 796 (Nd3+) 0.9685 0.00075 0.7873 0.0075 0.8594 0.005 864 (Nd3+) 0.9049 0.005918 0.9641 0.002722 0.8791 0.004261 889 (Nd3+) 0.8450 0.004 0.8663 0.003 0.3593 0.005 953 (Sm3+) 0.2849 0.004 0.5089 0.004 0.1633 0.001238 978 (Er3+) 0.4789 0.004 0.5456 0.004 0.0347 0.003 SUM_Nd_676_741_796_864_889 0.9571 0.006077 0.9729 0.005330 0.8618 0.002 SUM_Nd_741_796_864 0.9620 0.002717 0.9739 0.005 0.9412 0.004195 SWIR Range 978 (Er3+) 0.1312 0.00125 0.3044 0.005 0.5048 0.00125 1093-V003 (Sm3+) 0.9182 0.0005 0.7811 0.0005 0.7177 0.0005 1251 (Sm3+) 0.8730 0.002 0.8677 0.003228 0.5605 0.002114 SmSUM1093_1251 0.9282 0.00143 0.8866 0.003719 0.6874 0.001645 1402 (Pr3+) 0.6903 0.002291 0.7382 0.002 0.5060 0.001457 1459 (Pr3+) 0.8731 0.002 0.6936 0.001714 0.7082 0.002 1553-V002 (Pr3+) 0.9693 0.000502 0.7501 0.00075 0.7911 0.002 Pr Full Triple 0.8895 0.004 0.6794 0.005 0.6471 0.004 PrSUM1402_1459_1553_V002 0.9344 0.002212 0.8216 0.002 0.8191 0.003 1710-V002 (Nd3+) 0.8888 0.002 0.7668 0.001 0.5467 0.002 *Bolded “Sum Value Threshold” cells relate to the example used in previous figures, 796 nm Nd3+-related absorption *Underlined and bolded values relate to Figure 5.11 ***Note that Sample Box104-5 is NOT included for these calculations 159 5.5 Discussion Tables 5.2 and 5.3 listed all of the absorptions evaluated and the most reliable show the highest R2 values with their associated Absorption Depth Threshold (i.e., the Sum Value Threshold). Some attempts failed completely, such as the estimation of Sm3+ in the VNIR using the absorption band at 953 nm, however, others are remarkably strong for such a simple approach with basic band depth calculations. In general, the best regressions for estimating Nd are the summation of absorption depths at 741 nm, 796 nm and 864 nm for both individual sample sets (i.e., Box104, Box120) and combined sample sets (Box104 and Box120). This is especially true when Box104-Sample5 is removed (Table 5.3, Figures 5.10 and 5.11). It also seems that the Nd estimation in the VNIR is rather insensitive to the other gangue minerals present (i.e., the regression holds across both boxes) and allows for robust estimation with differing modal mineralogy when using the sum of multiple bands. Similarly, the best regression for estimating Sm for each ore type is the combination of the absorptions at 1093 nm and 1251 nm. Improvements to the R2 values for Box104 and for all subsamples are seen for Sm estimation in the SWIR when removing Box104-Sample5, however, not to the extent that is seen in the VNIR with Nd estimations. The best regressions for Pr are also via combinations of absorptions at 1402, 1459 and 1553 nm. When Box104-Sample5 is removed the regressions become stronger in the case of Box 104 and all subsamples combined, and the regression for the single absorption at 1553 nm for Box 104 becomes quite strong. 160 Figure 5.10 Relationship between Nd Proxy Value (from SUM_Nd_741_796_864 images, SumValueThreshold = 0.002, see bolded and underlined values on Table 2) and Nd (ppm) across all subsamples for both Box104 and Box120. Note the outlying location of Box104-Sample5. Figure 5.11 Relationship between Nd Proxy Value (from SUM_Nd_741_796_864 image, SumValueThreshold = 0.002, see bolded and underlined values on Table 3) and Nd (ppm) across all subsamples for both Box104 and Box120 but without Box104-Sample5. 161 Finally, there is a strong geochemical relationship between Nd and the Total Rare Earth Element (TREE+Y) content of the samples. By extension, we can use the same estimations of Nd via Proxy Values to try and estimate the TREE+Y. Figure 5.12 shows the relationships between Nd and TREE+Y at left and at right shows the relationship between the best Nd Proxy Value (Sum_Nd_741_796_864, without Box104-Sample5) and the TREE+Y contents (ppm). As long as the relationship between Nd and TREE+Y is relatively stable across the deposit this TREE+Y estimate via Nd3+-related absorptions should be reliable. Figure 5.12 Concentrations of Nd (ppm) vs. TREE+Y (ppm) showing high correlation (left). At right, is the relationship between the Nd Proxy Value “Sum_Nd_741_796_864” (SumValueThreshold = 0.002, see Table 5.3) and the TREE+Y contents (ppm) without Box104-Sample5. Table 5.4 was built using the best absorption depth cut offs to calculate Proxy Values for the strongest linear regressions (highest R2 values) that included Box104-Sample5 (i.e., R2 values and thresholds from Table 5.2). For each of the spectral grade estimations, the total scene for each box was used (Samples 1 - 7 for Box 104, Samples 1 - 6 for Box 120). Estimations of Nd and TREE+Y in the VNIR use the regressions that take into consideration subsamples from both boxes (i.e., column 3, “Both Boxes”). Estimations of Sm, Pr and Nd in the SWIR use Box-specific regressions and thresholds (i.e., columns 1 and 2). Figure 5.13 depicts the regression and spectral estimation calculation of the Nd grade estimation for both boxes in the VNIR. 162 Figure 5.13 Example plot relating Proxy Values derived from absorptions within hyperspectral image pixels to Nd (ppm) content of the 13 subsamples (black squares, Boxes 104 and 120). The resulting linear regression is plotted (black line) along with the R2 value. “Whole-scene” Proxy Values were calculated for Boxes 104 and 120 and are plotted along the linear regression line, from which Nd (ppm) content can be estimated (Box 104–blue diamond, Box 120–red circle, see Table 5.4). Table 5.4 Estimations of Nd, Sm, Pr and TREE+Y ore grades (all in ppm) through hyperspectral imaging and comparison with concentrations determined through traditional geochemical analyses. Box 104 (~1 m sample) 7 subsamples: 675,755 pixels in VNIR and 759,812 pixels in SWIR Box 120 (~1 m sample) 6 subsamples: 520,476 pixels in VNIR and 566,451 pixels in SWIR REE Grade Estimations (Proxy Used) Spectral Estimation New Weighted Assay Corporate Assay Spectral Estimation New Weighted Assay Corporate Assay Nd via VNIR imagery (SUM_Nd_741_796_864) 2915 3138 2960 3312 3301 3430 TREE+Y (via Nd in VNIR) 17141 17504 17022 19384 20274 21908 Sm via SWIR imagery (SmSUM1093_1251) 667 714 709 793 768 788 Pr via SWIR imagery (PrSUM1402_1459_1553) 579 782 799 756 814 879 Nd via SWIR imagery (1710) 2590 3138 2960 2837 3301 3430 163 5.6 Conclusion and implications In summary, hyperspectral imaging spectroscopy can be used to directly estimate ore grades of several REEs from drill core in hydrothermally altered peralkaline rocks where REE fluorocarbonates are the major ore mineral. It was shown that (1) estimation of Nd ore grade across ore types in the VNIR appears robust, (2) estimation of Sm and Pr ore grades within an ore suite in the SWIR is fairly accurate, and (3) estimation of TREE+Y contents in the VNIR, when Nd:TREE+Y is stable, is reliable. An obvious limitation to this method is the fact that a 2D surface is being used to approximate the geochemistry of a volume. Refinements to the methodology could include testing the use of absorption area instead of the simple absorption depth, attempting absorption-related calculations on continuum removed spectra, deconvolving spectra into either wavelets (Rivard et al. 2008) or modified Gaussian shaped curves (Sunshine et al. 1990), and integrating pre-classification of pixels based on mineral identification and/or spectral unmixing abundances. Despite the limitations noted and possible refinements it is encouraging that even simple band depth calculations can provide direct ore grade estimates of REE with fairly strong confidence. 164 Chapter 6. General Conclusions and Research Potential 6.1 Summary Research conducted in the scope of the PhD successfully attended to the knowledge gap concerning the reflectance spectroscopy of REE-bearing minerals. Thirty three (33) samples representing eleven (11) mineral species were investigated in detail: bastnaesite, synchysite, parisite, monazite, xenotime, britholite, mosandrite, cerite, kainosite, eudialyte, and zircon. Thematic summaries and conclusions are given in each chapter regarding the spectra and spectral variations of these REE fluorocarbonates, REE phosphates and REE-bearing silicates. The locations of prominent absorption features are listed in various “Band Index” tables, and compilation figures of representative spectra have been presented for each mineral. Probable origins for the 4f-4f electronic transition-related absorption features have been ascribed through comparison with published transmission and emission spectra of mono-lanthanide synthetic materials and natural minerals, and unpublished reflectance spectra of minerals and mono-lanthanide materials. Significant effort was made to choose appropriate lanthanides (i.e., Ln3+) based on electron microprobe compositions and expected locations of the often overlapping absorption bands. Strength of the research and interpretations lie in the diversity of mineral species studied. However, the ongoing detailed investigations revealed that its greatest weakness was perhaps the limited size of the sample suite considering the vast diversity of REE minerals and their chemical variability. With respect to relevance for economic geology, a suite of apatite samples with variable anion and REE contents is the most significant omission. This is also an opportunity and is discussed below. The use of imaging spectroscopy (sisuROCK) over point spectroscopy (TerraSpec Pro) was critical. For single crystal samples imaging spectroscopy provided great confidence in knowing that the field of view was restricted to only the sample of interest and also allowed investigation of spatial-spectral variations. Hand samples with smaller grains are not amenable to collecting single-mineral reflectance spectra from conventional point spectrometers. High-resolution imaging spectroscopy, on the other hand, allowed sufficient spatial resolution in both the VNIR and SWIR ranges to confidently allow the acquisition of representative mineral spectra. On the theme of methodology, use of Fourier transform infrared (FTIR) spectroscopy from 6000 to 200 cm-1 would provide data about the fundamental vibrations of the mineral samples and hopefully allow stronger interpretations of absorption features beyond ~2150 nm. 165 The use of scanning electron microscopy was an important step to ensure that compositional variations for each mineral sample were negligible and that reflectance spectra could be confidently used. This step resulted in the exclusion of several samples for baseline research. The use of electron microprobe analysis was also important for properly describing the chemistry of each sample and for confirming mineral identity. This step also resulted in the exclusion of several samples which showed cation results that were incompatible for the anticipated mineral species. However, it must be noted that the averaged microprobe results reflect the chemistry of a very small sample area whereas average reflectance spectra represent a much greater area. Therefore a significant assumption must be made that these two different analytical methods generate corresponding representative data. 6.2 Future research The baseline research presented in this thesis sets the foundation for further work in the realm of both pure and applied REE mineral spectroscopy, of which there is plenty to do. Although research within the context of the PhD project has closed, initial research projects have commenced and additional research themes have been identified. Two main themes are described: Additional REE Mineral Spectroscopy and Exploitation of Diagnostic Features for REE Mineral Detection and Identification. 6.2.1 Additional REE mineral spectroscopy Research conducted during the course of the PhD project successfully established spectral characteristics of a number of well characterized REE bearing minerals under three main themes: REE fluorocarbonates, REE phosphates and REE-bearing silicates. It addressed the wide knowledge gap that existed between reflectance spectroscopy and rare earth element mineralogy but, as research often does, it also exposed topics worthy of further study. Additional spectra from well characterized samples would provide more input data to understand variability amongst the REE bearing minerals and a larger spectral library would allow other users greater confidence for mineral identification using reflectance spectroscopy. During the latter portion of the thesis research, spectra and electron microprobe data were collected from rare mineral specimens “labeled” donnayite-(Y) (NaCaSr3Y(CO3)6 • 3H2O), ancylite-(Ce) (SrCe(CO3)2(OH)• H2O) and remondite-(Ce) (Na3(Ca,Ce,Na)3(CO3)5) and await detailed analysis that will require X-ray diffraction methods to resolve unexpected microprobe results (e.g., anomalous totals and relative cation proportions). Reflectance spectra have also been collected from other uncharacterized mineral specimens of minerals studied in detail (e.g., zircon and xenotime) and although they lack electron microprobe data these spectra would still be useful in a larger scale study. A number of REE oxide and allanite samples were also the target of 166 imaging spectroscopy, however, their lack of spectral features coupled with dark and glassy lustres suggest they are metamict. An expanded sample suite that included characterized (and non-metamict) high-REE apatite (Ca5(PO4)3(Cl,F,OH)), gadolinite-(Ce) (Ce2FeBe2Si2O10) and florencite-(Ce) (CeAl3(PO4)2(OH)6) would address most of the remaining important naturally occurring REE phases, other than the REE-bearing oxides such as loparite-(Ce) ((Na,Ce,Ca)(Ti,Nb)O3), brannerite ((U4+,Ca)(Ti,Fe3+)2O6), and fergusonite-(Y) (YNbO4). Finally, the synthesis and study of mono-lanthanide bearing crystals would also be useful to help elucidate the origins of specific absorption bands. Additional investigations would also benefit from FTIR spectroscopy to probe the fundamental vibrations of these minerals. This information would help resolve impacts of combination and overtone absorption bands in the visible to shortwave infrared region. In particular, it would help resolve the origin of the absorption bands from 2150 nm to 2530 nm seen in all of the REE minerals studied thus far. Determining absorption coefficients for specific lanthanides in specific host minerals would be another avenue worthy of study, especially in applied research settings such as REE grade estimation. 6.2.2 Exploitation of diagnostic features for REE mineral detection and identification The main research chapters of this thesis document the reflectance spectroscopy of well characterized REE minerals. Summary figures were given for each mineral grouping (REE fluorocarbonates, REE phosphates, REE-bearing silicates) depicting representative spectra, absorption cluster locations and dominant absorption band wavelength positions. From these figures it is evident that sufficient spectral variability exists between the various minerals to allow mineral identification based on the morphology of select absorption clusters. This variability of certain absorption clusters is directly related to the crystal structures of each mineral and the nature of the cation coordination polyhedra. The exact reasoning on how and why these patterns differ is worthy of study and could follow guidelines used in Ln-doping studies (e.g., Gorller-Walrand and Binnemans 1996) and be carried out in conjunction with photo- or cathodoluminescence techniques (e.g., Lenz et al. 2013). Further analysis and deconvolution of spectra (e.g., via continuum removal and/or wavelet transformation) will be required in order to establish the wavelength variations for the numerous overlapping absorption bands in each of the representative mineral spectra. Absorptions clusters showing the prominent differences in morphology are the Nd3+-related “absorption triplet” in the VNIR at ~745 nm (4I9/2 4F7/2+4S3/2), ~800 nm (4I9/2 4F5/2+2H9/2) and ~875 nm (4I9/2 4F3/2), and the Sm3+-related “absorption pair” at ~1080 nm (6H5/2 6F9/2) and ~1250 nm (6H5/2 6F7/2). Other useful observations for mineral identification and detection include the Er-Yb related absorption near 978 nm, the general patterns for HREE vs LREE enriched minerals, and characteristics 167 specific to individual minerals or mineral groups (e.g., the 2243 nm absorption band in bastnaesite). Establishment of a formal “decision tree” for mineral identification and discrimination via reflectance spectroscopy would be a valuable tool, especially in field settings. A “decision tree” would also be very useful for imaging spectroscopy, and there are already several examples of mineral detection and identification contained within this thesis. These include (1) the identification of mineral spectra for improved ROI in order to generate more representative average spectra when the target pixels are contained within a hand sample (e.g., synchysite sample UBC3376, britholite sample CMNOC F90-8), (2) differentiation of mosandrite from eudialyte in hand sample CMNOC 92-23, and (3) the exclusion of bastnaesite from the ROI used to generate the spectrum for cerite. A short comment is warranted about reflectance wavelength standards used to calibrate spectrometers, which commonly use REE-doped matrices (e.g., Weidner et al, 1986). The spectral variability of Ln 3+ related absorptions across various minerals proves that the nature of the coordination polyhedron is important in determining wavelength locations of 4f-4f intraconfigurational electronic transitions. Consequently, different REE-doped wavelength standards with different matrices are not intercomparable (e.g., standards from different manufacturers with the same dopant could show significantly different absorption positions) and the long term stability of the host matrix is relevant. Mann et al. (2014) have started to address the calibration of different instruments at separate laboratories against a single set of REE-doped Spectralon-based Labsphere standards, however, it would be warranted to evaluate different standards across these instruments, especially for heavily used wavelength standards that have been subjected to various environmental conditions. Interestingly, the crystal structures into which REE dopants are placed for the Spectralon calibration pucks varies between the Ho, Dy and Er products (see Table D.1 in Appendix D). 6.3 Conclusion This research narrowed the knowledge gap between REE mineralogy and reflectance spectroscopy in the VNIR-SWIR range. The detailed characterization of 11 REE-bearing mineral species confidently resolves the notion that some absorption features related to 4f-4f electronic transitions have significant variations between different minerals at spectral resolutions relevant to remote sensing. The numerous Ln3+-related absorption features (i.e., the Stark multiplets) always occur in consistent wavelength ranges, and the specific positions of specific absorption bands are consistent across multiple samples of a single mineral species (e.g., monazite) and sometimes even across multiple minerals that host the REE in a similar coordination environment (e.g., bastnaesite, parisite, synchysite). Furthermore, 168 resolvable and consistent variations between minerals are seen for some absorption clusters with respect to both positions of absorption bands and their relative strengths, such as the Nd3+ related bands near 875 nm or Sm3+ related bands near 1250 nm. Therefore, a relationship exists between the coordination environment of the Ln3+ within the structure of its host mineral and the resulting Ln 3+-related spectral patterns. Not all lanthanides are spectrally active in the VNIR-SWIR range with respect to 4f-4f electronic transitions. Trivalent La, Lu3+ and Y3+ have no electrons in their f-orbitals to excite, and Gd3+ only shows absorption features in the UV range. This leaves Pr3+, Nd3+, Sm3+, Eu3+, Tb3+, Dy3+, Ho3+, Er3+, Tm3+ and Yb3+ as the main spectrally active lanthanides in minerals. Of these, Eu, Tb, Ho and Tm tend to be of lower concentration than their adjacent lanthanides due to natural abundance patterns. Thus, spectral patterns relating to the Ln 3+ in REE-bearing minerals can be thought of as being driven primarily by Pr3+, Nd3+, Sm3+, Dy3+, Er3+ and Yb3+. For LREE-enriched minerals (e.g., cerite), Nd3+, Sm3+ and Pr3+ will drive the bulk of the REE-related spectral features. For HREE-enriched minerals (e.g., kainosite), Er3+, Dy3+ and Yb3+ will be the dominant spectrally active lanthanides with lesser input from Nd3+, Tb3+, Ho3+ and Tm3+ depending on their specific concentrations. For those minerals that do not show strong selectivity for LREE or HREE (e.g., mosandrite and britholite), a mixed character of absorption patterns will be seen. Some mineral species can also show significant variations in their relative REE content, which will distinctly impact the Ln 3+ related absorption patterns (e.g., LREE vs HREE eudialyte). Therefore, a relationship exists between the REE distribution within REE-bearing minerals and the resulting Ln 3+ related spectral patterns. The existence of these spectral-structural and spectral-chemical patterns for the REE minerals allows for their exploitation via reflectance spectroscopy. Using conventional processing techniques (e.g., SAM) pixels of REE mineral species were able to be identified from hyperspectral images of hand samples (e.g., synchysite). Furthermore, distinction between chemically similar but structurally different REE minerals in a single hand sample was also shown to be possible (e.g., cerite and bastnaesite). Finally, direct ore grading of REE-mineralized drill core from hyperspectral imagery was proven to be reliable using a simple processing approach based on absorption band depths in both the VNIR and SWIR ranges. The narrowing knowledge gap and resulting promise for exploiting spectral patterns in a variety of applications validates the need for more research, both pure and applied, into the reflectance spectroscopy and hyperspectral imaging of REE-bearing minerals and rocks. 169 References Assaaoudi, H., Ennaciri, A., & Rulmont, A. (2001). Vibrational spectra of hydrated rare earth orthophosphates. Vibrational Spectroscopy 25, 81-90. Adams, J. W. (1965). The visible region absorption spectra of rare earth minerals. Am. Mineral 50, 356-366. Adler, H., & Kerr, P.F. (1963). Infrared spectra, symmetry and structure relations of some carbonate minerals. American Mineralogist 48, 839-853. Allen, D. W. (2007). Holmium oxide glass wavelength standards. Journal of Research of the National Institute of Standards and Technology 112, 303. Ardanova, L. I., Get’man, E. I., Loboda, S. N., Prisedsky, V. V., Tkachenko, T. V., Marchenko, V. I., ... & Lyashenko, A. S. (2010). Isomorphous substitutions of rare earth elements for calcium in synthetic hydroxyapatites. Inorganic chemistry 49, 10687-10693. Badenes, J. A., Vicent, J. B., Llusar, M., Tena, M. A., & Monr, G. (2002). The nature of Pr-ZrSiO4 yellow ceramic pigment. Journal of materials science 37, 1413-1420. Baldridge, A. M., Hook, S. J., Grove, C. I., & Rivera, G. (2009). The ASTER spectral library version 2.0. Remote Sensing of Environment 113, 711-715. Baur, W. H. (1974). The geometry of polyhedral distortions. Predictive relationships for the phosphate group. Acta Crystallographica Section B: Structural Crystallography and Crystal Chemistry 30, 1195-1215. Becker, P. C., Hayhurst, T., Shalimoff, G., Conway, J. G., Edelstein, N., Boatner, L. A., & Abraham, M. M. (1984). Crystal field analysis of Tm3+ and Yb3+ in YPO4 and LuPO4. The Journal of chemical physics 81, 2872-2878. Bedini, E. (2009). Mapping lithology of the Sarfartoq carbonatite complex, southern West Greenland, using HyMap imaging spectrometer data. Remote Sens. Environ. 113, 1208–1219. Bedini, E. (2011). Mineral mapping in the Kap Simpson complex, central East Greenland, using HyMap and ASTER remote sensing data. Advances in Space Research 47, 60-73. Bell, K. (1989). Carbonatites—Genesis and evolution: Boston, Mass., Unwin Hyman Ltd., 618 p. 170 Bellezza, M., Merlino, S., & Perchiazzi, N. (2009). Mosandrite: Structural and crystal-chemical relationships with rinkite, The Canadian Mineralogist 47, 897-908. Berger, V.I., Singer, D.A., & Orris, G.J. (2009). Carbonatites of the world, explored deposits of Nb and REE—Database and grade and tonnage models. U.S. Geological Survey Open-File Report 2009-1139 17 p. Bernstein, L. R. (1982). Monazite from North Carolina having the alexandrite effect. American Mineralogist 67, 356-359. Bill, H., & Calas, G. (1978). Color centers, associated rare-earth ions and the origin of coloration in natural fluorites. Physics and Chemistry of Minerals 3, 117-131. Binnemans, K. (1996). A comparative spectroscopic study of Eu3+ in crystalline host matrices. Bulletin des Sociétés Chimiques Belges 105, 793-798. Binnemans, K., Couwenberg, I., De Leebeeck, H., Görller-Walrand, C., & Adam, J. L. (1999). Spectroscopic properties of tetravalent uranium in glasses. Journal of Alloys and Compounds 285, 105-111. Bishop, J. L., Lane, M. D., Dyar, M. D., & Brown, A. J. (2008). Reflectance and emission spectroscopy study of four groups of phyllosilicates: Smectites, kaolinite-serpentines, chlorites and micas. Clay Minerals 43, 35-54. Boardman, J.W., (1998). Leveraging the high dimensionality of AVIRIS data for improved sub-pixel target unmixing and rejection of false positives: mixture tuned matched filtering. In: Summaries of the Seventh Annual JPL Airborne Geoscience Workshop 55. Boardman, J. W., & Kruse, F. A. (2011). Analysis of Imaging Spectrometer Data Using-Dimensional Geometry and a Mixture-Tuned Matched Filtering Approach. Geoscience and Remote Sensing, IEEE Transactions on 49, 4138-4152. Boatner, L. A. (2002). Synthesis, structure, and properties of monazite, pretulite, and xenotime. Reviews in mineralogy and geochemistry 48, 87-121. Boesche, N. K., Rogaß, C., Mielke, C., & Kaufmann, H. (2014). Hyperspectral digital image analysis and geochemical analysis of a rare earth elements mineralized intrusive complex (fen carbonatite complex in Telemark region, Norway). EARSeL 34th Symposium Proceedings 13, 1–6. 171 Boldish, S. I., & White, W. B. (1998). Optical band gaps of selected ternary sulfide minerals. American Mineralogist 83, 865-871. Bowers, T.L., & Rowan, L.C. (1996). Lithologic mapping of the Ice River Alkaline Complex, British Columbia, Canada, using airborne visible/ infrared imaging spectrometer data: Photogrammetric Engineering and Remote Sensing 62, 1379–1385. Braun, S. A., Bream, B. R., & Gualda, G. A. (2009). AGE AND CHEMISTRY OF MEGACRYSTIC ZIRCONS FROM ZIRCONIA, NORTH CAROLINA. In: 2009 Portland GSA Annual Meeting Conference Proceedings. Breiter, K., Förster, H. J., & Škoda, R. (2006). Extreme P-, Bi-, Nb-, Sc-, U-and F-rich zircon from fractionated perphosphorous granites: The peraluminous Podlesí granite system, Czech Republic. Lithos 88, 15-34. Burns, R. (1993). Mineralogical Applications of Crystal Field Theory, Second Edition, Cambridge University Press, Cambridge, 551 p. Cámara, F., Sokolova, E., & Hawthorne, F. C. (2011). From structure topology to chemical composition. XII. Titanium silicates: the crystal chemistry of rinkite, Na2Ca4REETi(Si2O7)2OF3, Mineralogical Magazine 75, 2755-2774. Camus, Y., & Laferriere, Y. (2010). NI 43101 Technical Report on Mineral Resource Estimation of the Kipawa Deposit, Zeus Project, Quebec, Matamec Explorations Inc., 95 p. Carnall, W. T., Goodman, G. L., Rajnak, K., & Rana, R. S. (1989). A systematic analysis of the spectra of the lanthanides doped into single crystal LaF3. The Journal of Chemical Physics 90, 3443-3457. Carpéna, J., Boyer, L., Fialin, M., Kiénast, J. R., & Lacout, J. L. (2001). Ca 2+, PO43−⇌ Ln3+, SiO44− coupled substitution in the apatitic structure: stability of the mono-silicated fluor-britholite. Comptes Rendus de l'Académie des Sciences-Series IIA-Earth and Planetary Science 333, 373-379. Castor, S. B. (2008a). Rare Earth Deposits of North America. Resource Geology 58, 337–347. Castor, S.B. (2008b). The Mountain Pass rare earth carbonatite and associated ultrapotassic rocks, California. The Canadian Mineralogist 46, 779–806. 172 Chao, E.C.T., Back, J.M., Minkin, J.A., Tatsumoto, M., Wang, J., Conrad, J.E., McKee, E.H., Zonglin, H., Qingrun, M., & Shengguan, H. (1997). The Sedimentary Carbonate- Hosted Giant Bayan Obo Ree–Fe–Nb Ore Deposit of Inner Mongolia, China: a Corner Stone Example for Giant Polymetallic Ore Deposits of Hydrothermal Origin USGS Bulletin 2143, 65p. Clark, R. N. (1999). Chapter 1: Spectroscopy of Rocks and Minerals, and Principles of Spectroscopy, In: Manual of Remote Sensing, Volume 3, Remote Sensing for the Earth Sciences 3-58. Clark, R. N., & Roush, T. L. (1984). Reflectance spectroscopy: Quantitative analysis techniques for remote sensing applications. Journal of Geophysical Research: Solid Earth 89, 6329-6340. Clark, R.N., King, T.V.V., Klejwa, M., Swayze, G., & Vergo, N. (1990). High Spectral Resolution Reflectance Spectroscopy of Minerals, J. Geophys Res. 95, 12653-12680. Clark, R. N., Swayze, G. A., Livo, K. E., Kokaly, R. F., Sutley, S. J., Dalton, J. B., McDougal, R. R., & Gent, C. A. (2003). Imaging spectroscopy: Earth and planetary remote sensing with the USGS Tetracorder and expert systems, J. Geophys. Res. 108(E12). Clark, R. N., Swayze, G. A., Wise, R. A., Livo, K. E., Hoefen, T. M., Kokaly, R. F., & Sutley, S. J. (2007). USGS digital spectral library splib06a, U.S. Geol. Surv. Digital Data Ser. 231p. Clavier, N., Podor, R., & Dacheux, N. (2011). Crystal chemistry of the monazite structure, Journal of the European Ceramic Society 31, 941-976. Cloutis, E.A., Hawthorne, F.C., Mertzman, S.A., Krenn, K., Craig, M.A., Marcino, D., Methot, M., Strong, J., Mustard, J.F., Blaney, D.L., Bell III, J.F., & Vilas, F. (2006). Detection and discrimination of sulfate minerals using reflectance spectroscopy. Icarus 184, 121-157. Cox, J., Goode, J., & Hains, D.H. (2011). Technical Report on the Thor Lake Project, Northwest Territories, Canada. NI 43-101 Company Report. Crowley, J., Rowan, L., Podwysocki, M., & Meyer, D. (1988). Evaluation of Airborne Visible/Infrared Imaging Spectrometer Data of the Mountain Pass, California carbonatite complex. Proceedings of the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) Performance Evaluation Workshop. 155-161. Currie, J. (1951). The occurrence and relationships of some mica and apatite deposits in southeastern Ontario. Econ. Geology 46, 765-778 173 Currie, K. L., Knutson, J., & Temby, P. A. (1992). The Mud Tank carbonatite complex, central Australia—an example of metasomatism at mid-crustal levels. Contributions to Mineralogy and Petrology 109, 326-339. Currie, K.L., & van Breemen, O. (1996). The origin of rare minerals in the Kipawa syenite complex, western Quebec. The Canadian Mineralogist 34, 435-451. Dalsin, M. L., Groat, L. A., Creighton, S., & Evans, R. J. (2015). The mineralogy and geochemistry of the Wicheeda Carbonatite Complex, British Columbia, Canada. Ore Geology Reviews, 64, 523-542. Da Silva, D. N., Guedes, K. J., Pinheiro, M. V. B., Spaeth, J. M., & Krambrock, K. (2005). The microscopic structure of the oxygen–aluminium hole center in natural and neutron irradiated blue topaz. Physics and chemistry of minerals, 32, 436-441. Dai, J., Wang, D., Wang, R., & Chen, Z. (2013). Quantitative estimation of concentrations of dissolved rare earth elements using reflectance spectroscopy. Journal of Applied Remote Sensing, 7, 073513, 1-14. Dawson, J. B. (1962). The geology of Oldoinyo Lengai. Bull Volc 24, 349-387. Dieke, G. H., Crosswhite, H, & Crosswhite H.M., (1968). Spectra and energy levels of the rare earth ions in crystals. New York: Wiley Interscience, 193 p. Dorenbos, P., Krumpel, A.H., van der Kolk, E., Boutinaud, P., Bettinelli, M., & Cavalli, E. (2010). Lanthanide level location in transition metal complex compounds. Optical Materials 32, 1681-1685. Drew, L.J., Meng, Q., & Sun, W. (1990). The Bayan Obo iron–rare-earth–niobium deposit, Inner Mongolia, China. Lithos 26, 46–65. Duke, E. F. (1994). Near infrared spectra of muscovite, Tschermak substitution, and metamorphic reaction progress: implications for remote sensing. Geology 22, 621-624. Ebendorff-Heidepriem, H., & Ehrt, D. (2000). Formation and UV absorption of cerium, europium and terbium ions in different valencies in glasses. Optical Materials 15, 7-25. Eby, G.N. (1975). Abundance and distribution of the rare-earth elements and yttrium in the rocks and minerals of the Oka carbonatite complex, Quebec. Geochemica et Cosmochemica Acta 39, 597–620. 174 Eby, G.N., Roden-Tice, M., Krueger, H.L., Ewing, W., Faxon, E.H. and Woollley, A.R. (1995). Geochronology and cooling history of the northern part of the Chulwa Alkaline Province, Malawi. Journal of African Earth Sciences 20, 275-288. Farmer, V. C. (1974). Infrared spectra of minerals. Mineralogical society. 539 p. Fielding, P.E. (1970). The distribution of uranium, rare earths, and color centers in a crystal of natural zircon. American Mineralogist 55, 428-440. Finch, R. J., & Hanchar, J. M. (2003). Structure and chemistry of zircon and zircon-group minerals. Reviews in mineralogy and geochemistry 53, 1-25. Finch, R. J., Hanchar, J. M., Hoskin, P. W. O., & Burns, P. C. (2001). Rare-earth elements in synthetic zircon: Part 2. A single-crystal X-ray study of xenotime substitution, American Mineralogist 86, 681-689. Förster, H. J. (2000). Cerite-(Ce) and thorian synchysite-(Ce) from the Niederbobritzsch granite, Erzgebirge, Germany: implications for the differential mobility of the LREE and Th during alteration, The Canadian Mineralogist 38, 67-79 Förster, H. J. (2001). Synchysite-(Y)–synchysite-(Ce) solid solutions from Markersbach, Erzgebirge, Germany: REE and Th mobility during high-T alteration of highly fractionated aluminous A-type granites. Mineralogy and Petrology, 72, 259-280. Friis, H., Finch, A. A., Williams, C. T., & Hanchar, J. M. (2010). Photoluminescence of zircon (ZrSiO4) doped with REE3+ (REE= Pr, Sm, Eu, Gd, Dy, Ho, Er). Physics and Chemistry of Minerals 37, 333-342. Frost, R.L., & Dickfos, M. J. (2007). Raman spectroscopy of halogencontaining carbonates. Journal of Raman Spectroscopy 38, 1516-1522. Frost, R. L., & Erickson, K. L. (2005). Near-infrared spectroscopic study of selected hydrated hydroxylated phosphates. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 61, 45-50. Gerharz, R. (1965). Absorption structures in the visible reflection spectrum of minerals. Economic Geology 60, 1721-1725. 175 Gillen, R., Clark, S. J., & Robertson, J. (2013). Nature of the electronic band gap in lanthanide oxides. Physical Review B, 87, 125116. Goetz, A. F., Vane, G., Solomon, J. E., & Rock, B. N. (1985). Imaging spectrometry for earth remote sensing. Science 228,1147-1153. Görller-Walrand, C., & Binnemans, K. (1996). Rationalization of Crystal Field Parameterization, In: Handbook on the Physics and Chemistry of Rare Earths, Elsevier, Amsterdam, Vol. 23, Chapter 155, editors: K.A. Gschneidner and L. Eyring (North-Holland Publishers, Amsterdam, 1998), 121 - 283. Görller-Walrand, C., & Binnemans, K. (1998). Spectral intensities of f-f transitions, In: Handbook on the Physics and Chemistry of Rare Earths, Vol. 25, Chapter 167, editors: K.A. Gschneidner and L. Eyring (North-Holland Publishers, Amsterdam, 1998), 101-264. Grammatikopoulos, T., Mercer, W., & Gunning, C. (2013). Mineralogical characterisation using QEMSCAN of the Nechalacho heavy rare earth metal deposit, Northwest Territories, Canada. Canadian Metallurgical Quarterly 52, 265-277. Great Western Minerals Group Ltd. (2007). Great Western Minerals Group Ltd. receives draft preliminary assessment—Announces breakthrough in metallurgical processing: Saskatoon, Saskatchewan, Canada, Great Western Minerals Group Ltd. media release, November 14, 2 p. Green, A.A., Berman, M., Switzer, P., Craig, M.D., (1988). A transformation for ordering multispectral data in terms of image quality with implications for noise removal. IEEE Trans. Geosci. Remote Sens. 26, 65–74. Green, R. O., & Vane, G. (1989). Lithology and structure within the basement terrain adjacent to Clark Mountains, California, mapped with calibrated data from the airborne visible/infrared imaging spectrometer. Quantitative Remote Sensing: An Economic Tool for the Nineties 1, 957-960. Groves, D.I., & Vielreicher, N.M. (2001). The Phalaborwa (Palabora) carbonatite-hosted magnetite-copper sulfide deposit, South Africa—An end-member of the iron-oxide copper-gold-rare earth element deposit? Mineralium Deposita 36, 189–194. Gruber, J. B., Burdick, G. W., Valiev, U. V., Nash, K. L., Rakhimov, S. A., & Sardar, D. K. (2009). Energy levels and symmetry assignments for Stark components of Ho3+ (4f10) in yttrium gallium garnet (Y3Ga5O12). Journal of Applied Physics 106, 113110. 176 Guastoni, A., Nestola, F., & Giaretta, A. (2009). Mineral chemistry and alteration of rare earth element (REE) carbonates from alkaline pegmatites of Mount Malosa, Malawi. American Mineralogist 94, 1216-1222. Gupta, C.K., & Krishnamurthy, N. (2004). Extractive Metallurgy of Rare Earths, CRC Press. 484 p. Hanchar, J. M., Finch, R. J., Hoskin, P. W., Watson, E. B., Cherniak, D. J., & Mariano, A. N. (2001). Rare earth elements in synthetic zircon: Part 1. Synthesis, and rare earth element and phosphorus doping. American Mineralogist 86, 667-680. Hancock, L., Kirkland, C.L., & Huntington, J. (2012). Hyperspectral data – a tool for identification of geological chronometers. In 34th International Geological Conference (IGC). Australian Geosciences Council. Hansen, S., Mosel, B. D., Müller-Warmuth, W., & Fielding, P. E. (1996). EPR studies of Tb4+ in single crystals of zircon and scheelite structure silicates and germanates. Zeitschrift fur Naturforschung A-Journal of Physical Sciences 51, 885-894. Hapke, B. (1993). Theory of reflectance and Emittance Spectroscopy, Cambridge University Press, New York. p. 469. Hatch, G.P. (2012). Dynamics in the Global Market for Rare Earths. Elements 8, 341-346. Hayhurst, T., Shalimoff, G., Edelstein, N., Boatner, L. A., & Abraham, M. M. (1981). Optical spectra and Zeeman effect for Er3+ in LuPO4 and HfSiO4. The Journal of Chemical Physics 74, 5449-5452. Hernández, T., & Martin, P. (2007). Microstructural and optical features of a Eu-monazite. Journal of the European Ceramic Society 27, 109-114. Heuser, J., Bukaemskiy, A. A., Neumeier, S., Neumann, A., & Bosbach, D. (2014). Raman and infrared spectroscopy of monazite-type ceramics used for nuclear waste conditioning. Progress in Nuclear Energy 72, 149-155. Heylen, R., Parente, M. & Gader, P., (2014). A Review of Nonlinear Hyperspectral Unmixing Methods, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 7, 1844-1868. 177 Hoefen, T. M., Swayze, G., Giles, S.A., Van Gosen, B.S., Emsbo, P., & Karst, A. (2013). SPECTROSCOPIC INVESTIGATIONS OF REE HOSTED HEAVY MINERAL SANDS AND PHOSPHATES. In Geological Society of America Abstracts with Programs 45, 278. Holtstam, D., & Andersson, U.B. (2007). The REE minerals of the Bastnas-type deposits, south-central Sweden. Canadian Mineralogist 45, 1073-1114. Hunt, G.R. (1977). Spectral signatures of particulate minerals, in the visible and near-infrared, Geophysics 42, 501-513. Jaeger, H., Pletzke, K. and Hanchar, J. M. (2006). Hyperfine Interaction Study of Short Range Order in Zircon. Environmental Issues and Waste Management Technologies in the Ceramic and Nuclear Industries IX, 155, 31 – 40. Jin, G. B., Choi, E. S., Guertin, R. P., Brooks, J. S., Bray, T. H., Booth, C. H., & Albrecht-Schmitt, T. E. (2007). Syntheses, Structure, Magnetism, and Optical Properties of the Ordered Mixed-Lanthanide Sulfides γ-LnLn'S3 (Ln= La, Ce; Ln'= Er, Tm, Yb). Chemistry of materials 19, 567-574. Johnsen, O., & Grice, J. D. (1999). The crystal chemistry of the eudialyte group, The Canadian Mineralogist 37, 865-891. Johnsen, O., Ferraris, G., Gault, R. A., Grice, J. D., Kampf, A. R., & Pekov, I. V. (2003). The nomenclature of eudialyte-group minerals, The Canadian Mineralogist 41, 785-794. Judd, B.R. (1962). Optical absorption intensities of rare-earth ions, Phys. Rev. 127, 750-761. Kempe, U., Plötze, M., Brachmann, A., & Böttcher, R. (2002). Stabilisation of divalent rare earth elements in natural fluorite. Mineralogy and Petrology, 76, 213-234. Kempe, U., Thomas, S. M., Geipel, G., Thomas, R., Plötze, M., Böttcher, R., & Trinkler, M. (2010). Optical absorption, luminescence, and electron paramagnetic resonance (EPR) spectroscopy of crystalline to metamict zircon: Evidence for formation of uranyl, manganese, and other optically active centers. American Mineralogist 95, 335-347. Kerr, A. (2011). Rare-earth element (REE) mineralization in Labrador: A review of known environments and the geological context of current exploration activity. In: Current Research Newfoundland and Labrador Department of Natural Resources, Geological Survey Report 11-1, 109-143. 178 Kerr, A., Rafuse, H., Sparkes, G., Hinchey, J., & Sandeman, H. (2011). Visible/Infrared spectroscopy (VIRS) as a research tool in economic geology: background and pilot studies from Newfoundland and Labrador. In: Current Research, Newfoundland and Labrador Department of Natural Resources Geological Survey Report 11-1, 145-166. Klinger, M., Kempe, U., Pöppl, A., Böttcher, R., & Trinkler, M. (2012). Paramagnetic hole centres in natural zircon and zircon colouration. European Journal of Mineralogy 24, 1005-1016. Kogarko, L.N., Williams, C.T. and Woolley, A.R. (2002). Chemical evolution and petrogenetic implications of loparite in layered, agpaitic Lovozero complex, Kola Peninsula, Russia. Mineral. and Petrol. 74, 1-24. Kokaly, R. F. (2001). Investigating a physical basis for spectroscopic estimates of leaf nitrogen concentration. Remote Sensing of Environment 75, 153-161. Krambrock, K., Ribeiro, L. G. M., Pinheiro, M. V. B., Leal, A. S., de BC Menezes, M. Â., & Spaeth, J. M. (2007). Color centers in topaz: comparison between neutron and gamma irradiation. Physics and Chemistry of Minerals 34, 437-444. Krupa, J. C., & Carnall, W. T. (1993). Electronic structure of U4+, Np4+, and Pu4+ doped into ThSiO4 single crystal. The Journal of chemical physics 99, 8577-8584. Kruse, F. A., Lefkoff, A. B., Boardman, J. W., Heidebrecht, K. B., Shapiro, A. T., Barloon, P. J., & Goetz, A. F. H. (1993). The spectral image processing system (SIPS)—interactive visualization and analysis of imaging spectrometer data. Remote sensing of environment 44, 145-163. Laferriere, A. (2011). Technical Report, Mineral Resource Estimation, Eldor Property – Ashram Deposit, Nunavik, Quebec, Commerce Resources Corporation. 100 p. Lane, M. D., Mertzman, S. A., Dyar, M. D., & Bishop, J. L. (2011). Phosphate Minerals Measured in the Visible-Near Infrared and Thermal Infrared: Spectra and XRD Analyses. In: 42nd Lunar and Planetary Science Conference Proceedings, The Woodlands, Texas. p.1013 Laruhin, M. A., Van Es, H. J., Bulka, G. R., Turkin, A. A., Vainshtein, D. I., & Den Hartog, H. W. (2002). EPR study of radiation-induced defects in the thermoluminescence dating medium zircon (ZrSiO4). Journal of Physics: Condensed Matter 14, 3813. 179 Laukamp, C., Termin, K. A., Pejcic, B., Haest, M., & Cudahy, T. (2012). Vibrational spectroscopy of calcic amphiboles–applications for exploration and mining. European Journal of Mineralogy 24, 863-878. Le Bas, M.J., Keller, J., Tao, K., Wall, F., Williams, C.T., & Zhang, P. (1992). Carbonatite dykes at Bayan Obo, Inner Mongolia, China. Mineralogy and Petrology 46, 195–228. Lees, N. S. (2001). EPR crystallography: studies of defect centres in single crystals. PhD Thesis, University of Canterbury, Christchurch, New Zealand. Lenz, C., Talla, D., Ruschel, K., Škoda, R., Götze, J., & Nasdala, L. (2013). Factors affecting the Nd3+ (REE3+) luminescence of minerals. Mineralogy and Petrology 107, 415-428. Liben, W., Ni, Y., Hughes, J.M., Bayliss, P., & Drexler, J.W. (1994). The atomic arrangement of synchysite-(Ce), CeCaF(CO3)2. Canadian Mineralogist 32, 865-871. Libowitzky, E., & Beran, A. (2004). IR spectroscopic characterisation of hydrous species in minerals. In: Spectroscopic Methods in Mineralogy, EMU Notes in Mineralogy 6, 227 – 279. Linthout, K. (2007). Tripartite division of the system 2REEPO4–CaTh (PO4) 2–2ThSiO4, discreditation of brabantite, and recognition of cheralite as the name for members dominated by CaTh (PO4)2. The Canadian Mineralogist 45, 503-508. Liu, G. (2005). Electronic Energy Level Structure. In: Spectroscopic Properties of Rare Earths in Optical Materials 83, 1 – 94. Lottermoser, B. (1990). Rare-earth mineralisation within the Mt. Weld carbonatite laterite, Western Australia. Lithos 24, 151–167. MacRae, C. M., Wilson, N. C., & Torpy, A. (2013). Hyperspectral cathodoluminescence. Mineralogy and Petrology 107, 429-440. Mader, U.K., & Greenwood, H.J. (1988). Carbonatites and Related Rocks of the Prince and George Claims, Northern Rocky Mountains (93J, 93I); In: Geological Fieldwork 1987 - BC Ministry of Energy, Mines and Petroleum Resources 1988-1, 375-380. Mann, P., Cloutis, E. A., Greenberger, R. N., Millikin, R. E., Hiroi, T., Mustard, J. F., Klima, R. L., Hibbitts, C. A., Plescia, J. B., Bell, J. F. (III), Roush, T. L., Bishop, J. L., & Elhmann, B. L. (2014). An Interlaboratory UV/VIS/NIR Wavelength Calibration Study. LPSC XLV, Abstract # 2392. 180 Mariano, A. N. (1989). Economic geology of rare earth minerals. Reviews in Mineralogy and Geochemistry 21, 309–338. Mariano, A.N., & Mariano, A. (2012). Rare earth mining and exploration in North America. Elements 8, 369-376. Mars, J., & Rowan, L.C. (2011). ASTER spectral analysis and lithologic mapping of the Khanneshin carbonatite volcano, Afghanistan. Geosphere 7, 276 – 289. Matamec Explorations Inc. (2011). Matamec Outlines a Simple Metallurgical Process for the Kipawa Deposit, News Release, August 4, 2011, 2 p. McDonough, W. F., & Sun, S. (1995). The composition of the Earth. Chemical Geology 120, 223-254. McDowell, M.L. and Kruse, F.A. (2014). TERRAIN MAPPING USING INTEGRATED VISIBLE TO NEAR INFRARED, SHORT WAVE INFRARED, AND LONG WAVE INFRARED SPECTROSCOPY. In Geological Society of America Abstracts with Programs 46, 320. McHugh, E.L., Girard, J.M., Denes, L.J., Metes, P., & Sabine C. (2000). Current Research on Slope Movement in Mines: Use of Hyperspectral Imagery. Proceedings of the 14th International Conference on Applied Geologic Remote Sensing 584-591. Meng, D., Wu, X., Mou, T., & Li, D. (2001). Microstructural investigation of new polytypes of parisite-(Ce) by high-resolution transmission electron microscopy, The Canadian Mineralogist 39, 1713-1724. Middlemost, E. (1990). Mineralogy and petrology of the rauhaugites of the Mt Weld Carbonatite Complex of Western Australia. Mineral. Petrol. 41, 145–161. Mitchell, R.H. (2005). Carbonatites and carbonatites and carbonatites: The Canadian Mineralogist 43, 2049–2068. Miyawaki, R. and Nakai, I. (1996). Crystal chemical aspects of rare earth minerals. In: Rare Earth Minerals (A.P. Jones, F. Wall & C.T. Williams, eds). Chapman & Hall, London, U.K., 21-40. Morin Ka, S. (2012) Hyperspectral characterisation of rare earth minerals. Geological Survey of Western Australia, Record 2012/12, 50 p. Moore, P. B., & Chen, J. (1983). Cerite, RE9(Fe3+,Mg)(SiO4)6(SiO3OH)(OH)3: its crystal structure and relation to whitlockite American Mineralogist 68, 996-1003. 181 Momma, K., & Izumi, F. (2011). VESTA 3 for three-dimensional visualization of crystal, volumetric and morphology data. Journal of Applied Crystallography 44, 1272-1276. Mullica, D. F., Milligan, W. O., Grossie, D. A., Beall, G. W., & Boatner, L. A. (1984). Ninefold coordination LaPO4: Pentagonal interpenetrating tetrahedral polyhedron. Inorganica chimica acta 95, 231-236. Mustard, J. F. (1992). Chemical composition of actinolite from reflectance spectra American Mineralogist 77, 345-358. Mustard, J. F., & Pieters, C. M. (1989). Photometric phase functions of common geologic minerals and applications to quantitative analysis of mineral mixture reflectance spectra. Journal of Geophysical Research 94, 13619-13634. Mustard, J. F., & Hayes, J. E. (1997). Effects of hyperfine particles on reflectance spectra from 0.3 to 25 μm. Icarus 125, 145−163. Nasdala, L., Zhang, M., Kempe, U., Panczer, G., Gaft, M., Andrut, M., & Plötze, M. (2003). Spectroscopic methods applied to zircon. Reviews in mineralogy and geochemistry 53, 427-467. Nasdala, L., Gotze, J., Hancar, J., Gaft, M., & Krbetschek, M. (2004). Luminescence techniques in earth sciences. Spectroscopic methods in mineralogy 6, 43-91. Nassau, K. (1978). The Origins of Color in Minerals. American Mineralogist 63, 219-229. Nassau, K. (2001). The Physics and Chemistry of Color: The Fifteen Causes of Color. 2nd edition Wiley Interscience. 496 p. Ni Y., Hughes J. M., & Mariano, A. N. (1993). The atomic arrangement of bastnasite-(Ce), Ce(CO3)F, and structural elements of synchysite-(Ce), röntgenite-(Ce), and parisite-(Ce). American Mineralogist 78, 415-418. Ni Y., Hughes, J. M., & Mariano, A. N. (1995). Crystal chemistry of the monazite and xenotime structures. American Mineralogist 80, 21-26. Ni, Y., Post, J.E., & Hughes, J.M. (2000). The crystal structure of parisite-(Ce), Ce2CaF2(CO3)3. American Mineralogist 85, 251–258. 182 Noe, D. C., Hughes, J. M., Mariano, A. N., Drexler, J. W., & Kato, A. (1993). The crystal structure of monoclinic britholite-(Ce) and britholite-(Y). Zeitschrift für Kristallographie-Crystalline Materials 206, 233-246. Oberti, R., Ottolini, L., della Ventura, G., & Parodi, G. C (2001). On the symmetry and crystal chemistry of britholite: new structural and microanalytical data, American Mineralogist 86, 1066-1075. Ofelt, G.S. (1962). Intensities of crystal spectra of rare-earth ions. J. Chem. Phys. 37, 511-520. Onac, B. P., Ettinger, K., Kearns, J., & Balasz, I. I. (2005). A modern, guano-related occurrence of foggite, CaAl (PO4)(OH) 2· H2O and churchite-(Y), YPO4· 2H2O in Cioclovina Cave, Romania. Mineralogy and Petrology 85, 291-302. Oppenheimer, C. (1998). Satellite observation of active carbonatite volcanism at Ol Doinyo Lengai, Tanzania. International Journal of Remote Sensing 19, 55-64. Pakhomovsky Y A, Men’shikov Y P, Yakovenchuk V N, Ivanyuk G Y, Krivovichev S V, Burns P C (2002). Cerite-(La), (La,Ce,Ca)9(Fe,Ca,Mg)(SiO4)3[SiO3(OH)]4(OH)3, a new mineral species from the Khibina alkaline massif: Occurrence and crystal structure. The Canadian Mineralogist 40, 1177-1184. Pasero, M., Kampf, A. R., Ferraris, C., Pekov, I. V., Rakovan, J., & White, T. J. (2010). Nomenclature of the apatite supergroup minerals. European Journal of Mineralogy 22, 163-179. Peijzel, P., Meijerink, A., Wegh, R., Reid, M., & Burdick, G. (2005). A complete energy level diagram for all trivalent lanthanide ions. Journal of Solid State Chemistry 178, 448-453. Pell, J. (1987). Alkaline Ultrabasic Rocks in British Columbia: Carbonatites, Nepheline Syenites, Kimberlites, Ultramafic Lamprophyres, and Related Rocks. BC Ministry of Energy, Mines and Petroleum Resources, Open File 1987-17, 109 p. Pell, J. (1994). Carbonatites, Nepheline Syenites, Kimberlites and Related Rocks in British Columbia; British Columbia Ministry of Energy and Mines, Bulletin 88, 133 p. Pekov, I. V., Pasero, M., Yaskovskaya, A. N., Chukanov, N. V., Pushcharovsky, D. Y., Merlino, S., & Zadov, A. E. (2007). Fluorcalciobritholite, (Ca,REE)5[(Si,P)O4] 3F, a new mineral: description and crystal chemistry. European Journal of Mineralogy 1, 95-103. 183 Perry, S.L. (2004). Review of airborne and orbiting remote sensing systems applied to worldwide mineral exploration – case histories focused on infrared spectroscopy. In Infrared Spectroscopy in Geochemistry, Exploration Geochemistry, and Remote Sensing (P.L. King, M.S. Ramsey and G.A. Swayze, eds.). Mineralogical Association of Canada. 33:227-236. Plaza, A., Martinez, P., Perez, R., & Plaza, J. (2004). “A quantitative and comparative analysis of endmember extraction algorithms from hyperspectral data,” IEEE Trans. Geosci. Remote Sens. 42, 650–663. Polshin, E. V., Platonov, A. N., Borutzky, B. E., Taran, M. N., & Rastsvetaeva, R. K. (1991). Optical and Mössbauer study of minerals of the eudialyte group. Physics and Chemistry of Minerals 18, 117-125. Rapaport, A., David, V., Bass, M., Deka, C., & Boatner, L. A. (1999). Optical spectroscopy of erbium-doped lutetium orthophosphate. Journal of luminescence 85, 155-161. Richardson, D.G., & Birkett, T.C. (1996a). Carbonatite associated deposits; In: Geology of Canadian Mineral Deposit Types, O.R.Ecstrand, W.D. Sinclair and R.I. Thorpe, Editors, Geological Survey of Canada, 541-558. Richardson, D.G., & Birkett, T.C. (1996b). Peralkaline rock associated rare metals. In: Geology of Canadian Mineral Deposit Types, O.R.Ecstrand, W.D. Sinclair and R.I. Thorpe, Editors, Geological Survey of Canada, 523–540. Richman, I., Kisliuk, P., & Wong, E. Y. (1967). Absorption Spectrum of U4+ in Zircon (ZrSiO4). Physical Review 155, 262-263. Rivard, B., Feng, J., Gallie, A., & Sanchez-Azofeifa, A. (2008). Continuous wavelets for the improved use of spectral libraries and hyperspectral data. Remote Sensing of Environment 112, 2850-2862. Robinson, R.J., Power, M., & Barker, J.C. (2011). Technical Report on the Exploration Program and Mineral Resource Estimate for the Bokan Mountain Property. Ucore Rare Metals Inc., Company Report, 190 p. Robles-Kelly, A., & Huynh, C. P. (2013). Imaging spectroscopy for scene analysis. Springer Science & Business Media. 270 pp. 184 Rogge, D. M., Rivard, B., Zhang, J., Sanchez, A., Harris, J., & Feng, J. (2007). Integration of spatial–spectral information for the improved extraction of endmembers. Remote Sensing of Environment 110, 287-303. Rossman, G. R. (2014). Optical spectroscopy. Reviews in Mineralogy and Geochemistry 78, 371-398. Rossman, G. R., & Taran, M. N. (2001). Spectroscopic standards for four-and fivefold-coordinated Fe2+ in oxygen-based minerals. American Mineralogist 86, 896-903. Rowan, L. C. (1998). Analysis of simulated advanced spaceborne thermal emission and reflection (ASTER) radiometer data of the Iron Hill, Colorado, study area for mapping lithologies. Journal of Geophysical Research: Atmospheres (1984–2012) 103, 32291-32306. Rowan, L. C., Kingston, M. J., & Crowley, J. K. (1986). Spectral reflectance of carbonatite and related alkalic igneous rocks from four North American localities. Economic Geology 81, 857–871. Rowan, L. C., Bowers, T. L., Crowley, J. K., Anton-Pacheco, C., Gumiel, P., & Kingston, M. J. (1995). Analysis of Airborne Visible-Infrared Imaging Spectrometer (AVIRIS) data of the Iron Hill, CO, carbonatite–alkalic igneous rock complex. Economic Geology 90, 1966–1982. Rowan, L. C., Clark, R. N., & Green, R. O. (1996). Mapping minerals in the Mountain Pass, California area using Airborne Visible–Infrared Imaging Spectrometer (AVIRIS) data. Proceedings of the 11th Conference on Geologic Remote Sensing 1, I177–I178. Rowan, L.C., & Mars, J.C. (2003). Lithologic mapping in the Mountain Pass, California area using Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) data. Remote Sensing of Environment 84, 350–366. Rumanova I M, Volodina G F Belov N V (1967). The crystal structure of the rare earth ring silicate kainosite Ca2(Y,Tr)2[Si4O12] CO3·H2O. Soviet Physics - Crystallography 11, 485-491 Salisbury, J.W. (1993). Mid-infrared spectroscopy: Laboratory data, In: Remote Geochemical Analysis: Elemental and Mineralogical Composition (C. M. Pieters, & P.A.J. Englert, eds.), Cambridge University Press, Cambridge, 79-98. Salisbury, J.W., Walter, L.S., Vergo, N., & D'Aria, D.M. (1991). Infrared (2.1-25 µm) spectra of minerals, The Johns Hopkins University Press, Baltimore, 267p. 185 Salisbury, J. W., & Wald, A. E. (1992). The role of volume scattering in reducing spectral contrast of restrahlen bands in spectra of powdered minerals. Icarus 96, 121-128. Sastri, V.S., Bünzli, C.G., Rao, V.R., Rayudu, G.V.S., & Perumareddi, J.R. (2003). Modern Aspects of Rare Earths and Their Complexes, Elsevier, Amsterdam. 1006 p. Savchyn, P., Karbovnyk, I., Vistovskyy, V., Voloshinovskii, A., Pankratov, V., Guidi, M. C., & Popov, A. I. (2012). Vibrational properties of LaPO4 nanoparticles in mid-and far-infrared domain. Journal of Applied Physics 112, 124309. Shannon, R.D. (1976). Revised effective ionic radii and systematic studies of interatomic distances in halides and chalcogenides. Acta Cryst A32, 751–767. Sheard, E. R., Williams-Jones, A. E., Heiligmann, M., Pederson, C., & Trueman, D. L. (2012). Controls on the concentration of zirconium, niobium, and the rare earth elements in the Thor Lake rare metal deposit, Northwest Territories, Canada. Economic Geology 107, 81-104. Silva, E. N., Ayala, A. P., Guedes, I., Paschoal, C. W. A., Moreira, R. L., Loong, C. K., & Boatner, L. A. (2006). Vibrational spectra of monazite-type rare-earth orthophosphates. Optical Materials 29, 224-230. Sinclair, W. D., Hunt, P.A., & Birkett, T.C. (1994). U-Pb Zircon and Monazite Ages of the Grace Lake Granite, Blachford Lake Intrusive Suite, Slave Province, Northwest Territories, In: Radiogenic Age and Isotopic Studies - Report 8; Geol. Surv. Can., Current Research 1994-F, 15 - 20. Singer, R. B. (1981). Near‐infrared spectral reflectance of mineral mixtures: Systematic Combinations of pyroxenes, olivine, and iron oxides. Journal of Geophysical Research: Solid Earth 86(B9), 7967-7982. Smith, D. R., Noblett, J., Wobus, R. A., Unruh, D., Douglass, J., Beane, R., Davis, C., Goldman, S., Kay, G., Gustavson, B., Saltoun, B., & Stewart, J., (1999). Petrology and geochemistry of late-stage intrusions of the A-type, Mid-Proterozoic Pikes Peak batholith (central Colorado, USA): implications for petrogenetic models. Precambrian Research 98, 271-305 Smith, M. P., Campbell, L. S., & Kynicky, J. (2015). A review of the genesis of the world class Bayan Obo Fe–REE–Nb deposits, Inner Mongolia, China: multistage processes and outstanding questions. Ore Geology Reviews, 64, 459-476. 186 Sokolova, E., & Cámara, F. (2008). From structure topology to chemical composition.VIII. Titanium silicates: the crystal chemistry of mosandrite from type locality of Låven (Skådön), Langesundsfjorden, Larvik, Vestfold, Norway. Mineralogical Magazine 72, 887-897. Sorensen, H. (1997). The agpaitic rocks; an overview. Mineralogical Magazine 61, 485-498. Strohhöfer, C., & Polman, A. (2003). Absorption and emission spectroscopy in Er 3+–Yb 3+ doped aluminum oxide waveguides. Optical Materials 21, 705-712. Swayze, G.A. (2004). Using reflectance spectroscopy to evaluate minerals of environmental concern. In Infrared Spectroscopy in Geochemistry, Exploration Geochemistry, and Remote Sensing (P.L. King, M.S. Ramsey and G.A. Swayze, eds.). Mineralogical Association of Canada. 33:181-191. Swayze, G.A., Pearson, N., Wilson, S., Benzel, W.M., Clark, R., Hoefen, T.M., Van Gosen, B., Adams, M., & Reitman, J. (2013). Spectrally distinguishing between REE-bearing minerals based on differences in their crystal field f-f transition absorptions. In Geological Society of America Abstracts with Programs 45, 278. Talla, D., Beran, A., Škoda, R., & Losos, Z. (2011). On the presence of OH defects in the zircon-type phosphate mineral xenotime, (Y,REE)PO4. American Mineralogist 96, 1799-1808. Thompson, T.B. (1988). Geology and uranium-thorium mineral deposits of the Bokan Mountain granite complex, south- eastern Alaska. Ore Geology Reviews 3, 193–210. Traversa, G., Gomes, C.B., Brotzu, P., Baraglini, P., Morbidelli, L., Principato, S.M., Ronca, S., & Ruberti, E. (2001). Petrography and mineral chemistry of carbonatites and mica-rich rocks from the Araxá complex (Alta Paranaíba province, Brazil). Anais da Academia Brasileira de Ciências 73, 1–32. Trueman, D.L., Pederson, J.C., de St. Jorre, L., & Smith, D.G.W. (1988). The Thor Lake rare-metal deposits, Northwest Territories, In: Recent Advances in the Geology of Granite-Related Mineral Deposits, Canadian Institute of Mining and Metallurgy, 39, 280-290 Turner, D. J., Rivard, B., & Groat, L. A. (2014). Visible and short-wave infrared reflectance spectroscopy of REE fluorocarbonates. American Mineralogist 99, 1335-1346. Turner, D., Rivard, B., & Groat, L. (2014, July). Rare earth element ore grade estimation of mineralized drill core from hyperspectral imaging spectroscopy. In: Geoscience and Remote Sensing Symposium (IGARSS), 2014 IEEE International 4612-4615. 187 Vance, E. R., & Mackey, D. J. (1978). Optical spectra of U4+ and U5+ in zircon, hafnon, and thorite. Physical Review B 18, 185. Van der Meer, F. D., Van der Werff, H. M., van Ruitenbeek, F. J., Hecker, C. A., Bakker, W. H., Noomen, M. F., & Woldai, T. (2012). Multi-and hyperspectral geologic remote sensing: A review. International Journal of Applied Earth Observation and Geoinformation 14, 112-128. Vane, G., & Goetz, A. F. (1993). Terrestrial imaging spectrometry: current status, future trends. Remote Sensing of Environment 44, 117-126. Vasyukova, O., & Williams-Jones, A. E. (2014). Fluoride–silicate melt immiscibility and its role in REE ore formation: Evidence from the Strange Lake rare metal deposit, Québec-Labrador, Canada. Geochimica et Cosmochimica Acta 139, 110-130. Walsh, B. (2006). Judd-Ofelt theory: principles and practices, In: Advances in Spectroscopy for Lasers and Sensing 403 – 433. Wang, L., Ni, Y., Hughes, J. M., Bayliss, P., & Drexler, J. W. (1994). The atomic arrangement of synchysite-(Ce), CeCaF(CO3)2. The Canadian Mineralogist 32, 865-871. Watson, K., Rowan, L. C., Bowers, T. L., Anton-Pacheco, C., Gumiel, P., & Miller, S. H. (1996). Lithologic analysis from multispectral thermal infrared data of the alkalic rock complex at Iron Hill, Colorado. Geophysics 61, 706-721. Weidner, V. R., Mavrodineanu, R., Eckerle, K. L., & Adams, B. (1984). Holmium Oxide Solution Wavelength Standard From 240 to 640 nm – SRM2034. NBS Special Publication 260-102. 69 p. Weidner, V. R., Mavrodineanu, R., Mielenz, K. D., Velapoldi, R. A., Eckerle, K. L., & Adams, B. (1985). Spectral Transmittance Characteristics of Holmium Oxide. Journal of Research of the National Bureau of Standards 90, 243-253. Weidner, V. R., Barnes, P. Y., & Eckerle, K. L. (1986). A wavelength standard for the near infrared based on the reflectance of rare-earth oxides. J. Res. Natl. Bur. Stand 91, 243. White, W. B. (1967). Diffuse-Reflectance Spectra of Rare-Earth Oxides. Applied Spectroscopy, 21, 167-171. 188 Wildner M., Andrut, M., & Rudowica, C.Z. (2004). Optical absorption spectroscopy in geosciences, Part I: Basic concepts of crystal field theory. In: Spectroscopic Methods in Mineralogy. European Mineralogical Union Notes in Mineralogy, Vol 6. 93 – 144. Winter, J.D. (2009). Principles of Igneous and Metamorphic Petrology, 2nd Ed., Prentice Hall, 702 p. Wood, B. J., & Strens, R. G. J. (1979). Diffuse reflectance spectra and optical properties of some sulphides and related minerals. Mineral. Mag. 43, 509-518. Woolley, A.R. (1982). A discussion of carbonatite evolution and nomenclature, and the generation of sodic and potassic fenites. Mineral. Mag. 46, 13-17. Woolley, A.R. (1987). Alkaline rocks and carbonatites of the world – Part 1: North and South America: London, British Museum (Natural History), 216 p. Woolley, A.R. and Kempe, D.R.C. (1989). Carbonatites: nomenclature, average chemical compositions, and element distribution. In: Carbonatites: Genesis and Evolution (K. Bell, Ed.). Unwin Hyman, London, 1-14. Woolley, A.R., & Kjarsgaard, B.A. (2008). Paragenetic types of carbonatite as indicated by the diversity and relative abundances of associated silicate rocks—evidence from global database. The Canadian Mineralogist 46, 741–752. Wu, C. (2008). Bayan Obo controversy: carbonatites versus iron oxide-Cu-Au-(REE-U). Resource Geology 58, 348–354. Yang, H., Dembowski, R. F., Conrad, P. G., & Downs, R. T. (2008). Crystal structure and Raman spectrum of hydroxyl-bästnaesite-(Ce), CeCO3(OH), American Mineralogist 93, 698-701. Yang, X.M., & Le Bas, M.J. (2004). Chemical compositions of carbonate minerals from Bayan Obo, Inner Mongolia, China: implications for petrogenesis. Lithos 72, 97-116. Zaitsev, A.N, Wall, F., & Bas, M.J.L. (1998). REE-Sr-Ba minerals from the Khibina carbonatites, Kola Peninsula, Russia: their mineralogy, paragenesis and evolution. Mineralogical Magazine 62, 225-225. Zhang, J., Rivard, B., & Rogge, D.M. (2008). The Successive Projection Algorithm (SPA), an Algorithm with a Spatial Constraint for the Automatic Search of Endmembers in Hyperspectral Data. Sensors 8, 1321-1342. 189 Zhang, M., Salje, E. K., & Ewing, R. C. (2002). Infrared spectra of Si-O overtones, hydrous species, and U ions in metamict zircon: radiation damage and recrystallization. Journal of Physics: Condensed Matter 14, 3333. Zhang, M., Salje, E. K., & Ewing, R. C. (2003). Oxidation state of uranium in metamict and annealed zircon: near-infrared spectroscopic quantitative analysis. Journal of Physics: Condensed Matter 15, 3445. Zhang, Y., & Vance, E. R. (2008). Plutonium in monazite and brabantite: Diffuse reflectance spectroscopy study. Journal of Nuclear Materials 375, 311-314. Zou, X., & Toratani, H. (1995). Evaluation of spectroscopic properties of Yb 3+-doped glasses. Physical Review B 52, 15889. 190 Appendix A. Sample Photographs and Hyperspectral Imagery Appendix A contains photographs of each sample as well as false colour hyperspectral imagery from the SWIR camera (0.241 mm/pixel resolution). Bands displayed for RGB were 197 (2162 nm), 203 (2199 nm), and 227 (2349 nm) --- these are default bands and were not chosen to display any particular feature across the suite. Some SWIR images have been rotated to better match the photograph, however, a few samples show a different orientation in the photograph from the image cube. 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 Appendix B. Scanning Electron Microscopy Appendix B contains scanning electron microscope (SEM) images of mineral grains and fragments that were mounted in epoxy pucks, as well as scans and drawings of the sample pucks themselves. Each SEM image is displayed in backscattered electron (BSE) mode with high contrast settings, which will emphasize any chemical zoning present since greyscale values are based on average atomic mass of each pixel. Conditions were generally carried out using a 15 kV beam. Some images are labeled with “SE”, however, these are not secondary electron images showing topographic relief. “PDBSE” stands for Plate Detector Back Scatter Electron image. Not all samples mounted and examined were probed (e.g., synchysite from Whitecloud), and not all samples probed were included in this thesis (e.g., donnayte, ancylite, remondite, fergusonite). 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 Appendix C. Electron Microprobe Compositions Samples were analyzed by electron microprobe at the Saskatchewan Research Council’s Advanced Microanalysis Centre using a Cameca SX-100 equipped with 5 tunable wavelength dispersive spectrometers. Operating conditions were: 40° takeoff angle, beam energy of 15 keV, beam current of 20 nA, beam diameter of 5 μm. The MAN background intensity data was calibrated and continuum absorption corrected. Elements were acquired using analyzing crystals LLIF for FeKα, TaLα, PrLα, EuLα, DyLα, TmLα, MnKα, LaLα, NdLα, GdLα, HoLα, YbLα, BaLα, CeLα, SmLα, TbLα, ErLα, LuLα, PET for CaKα, KKα, ClKα, TiKα, NbLα, YLα, SrLα, ZrLα, PKα, UMα, ThMα, and LTAP for MgKα, FKα, NaKα, SiKα, AlKα. Counting times were 10 seconds for Zr and P and 15 seconds for all other elements, with off peak count times of 10 seconds. The standards (with elements) were SPI-Barite (Ba), SPI-Celestite (Sr), SPI-YAG (Y, Al), Smithsonian Cr-augite (Mg, Ca), Smithsonian Ilmenite (Fe, Ti), Smithsonian Apatite (F, P), Smithsonian Microcline (K), Smithsonian Zircon (Zr), Harvard Albite (Si, Na), Cameca Mn (Mn), SPI2-TlCl (Cl), SPI2-Nb (Nb), SPI2-La (La), SPI2-Ce (Ce), SPI2-Pr (Pr), SPI2-Nd (Nd), SPI2-Sm (Sm), SPI2-Eu (Eu), SPI2-Gd (Gd), SPI2-Tb (Tb), SPI2-Dy (Dy), SPI2-Ho (Ho), SPI2-Er (Er), SPI2-Tm (Tm), SPI2-Yb (Yb), SPI2-Lu (Lu), SPI2-Ta (Ta), SPI2-Th (Th), and SPI2-U (U). Analyses were conducted in three batches at different times but using the same analytical program and conditions. Detection limits vary between analytical points due to changing target matrix (i.e., mineralogy) and average detection limits (in wt.%) are reported as: F (0.015), K (0.007), P (0.010), U (0.060), Y (0.029), Al (0.004), Ba (0.023), Ca (0.006), Ce (0.021), Cl (0.008), Dy (0.021), Er (0.020), Eu (0.018), Fe (0.008), Gd (0.020), Ho (0.021), La (0.023), Lu (0.021), Mg (0.004), Mn (0.008), Na (0.007), Nb (0.027), Nd (0.021), Pr (0.020), Si (0.003), Sm (0.018), Sr (0.025), Ta (0.026), Tb (0.019), Th (0.048), Ti (0.008), Tm (0.020), Yb (0.022), and Zr (0.033). All compositional values are in wt.%. Formulae are calculated on the average composition for each mineral and results given in atoms per formula unit (apfu). Accordingly, stoichiometrically calculated H2O and CO2 only appear for the average compositions. 230 Table C.1 Bastnaesite - Burundi Point 1 2 3 4 5 Average 2σ Atom apfuNb2O5 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Nb5+ 0.00P2O5 0.00 0.00 0.00 0.00 0.00 0.00 0.00 P5+ 0.00SiO2 0.12 0.10 0.11 0.12 0.10 0.11 0.02 Si4+ 0.00ZrO2 0.00 0.00 0.03 0.00 0.00 0.01 0.03 Zr4+ 0.00UO2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 U4+ 0.00ThO2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Th4+ 0.00Al2O3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Al3+ 0.00La2O3 18.97 19.88 19.56 19.92 18.86 19.44 1.00 La3+ 0.29Ce2O3 32.77 32.63 32.42 32.43 32.68 32.58 0.31 Ce3+ 0.48Pr2O3 3.16 2.93 2.95 3.01 3.32 3.07 0.33 Pr3+ 0.05Nd2O3 11.63 10.59 10.65 10.74 11.63 11.05 1.07 Nd3+ 0.16Sm2O3 1.24 1.14 1.08 1.18 1.25 1.18 0.14 Sm3+ 0.02Eu2O3 0.14 0.11 0.18 0.19 0.19 0.16 0.08 Eu3+ 0.00Gd2O3 0.55 0.45 0.48 0.63 0.54 0.53 0.14 Gd3+ 0.01Tb2O3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Tb3+ 0.00Dy2O3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Dy3+ 0.00Ho2O3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Ho3+ 0.00Er2O3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Er3+ 0.00Tm2O3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Tm3+ 0.00Yb2O3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Yb3+ 0.00L
UBC Theses and Dissertations
Reflectance spectroscopy and imaging spectroscopy of rare earth element-bearing mineral and rock samples Turner, David James 2015
Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.
- 24-ubc_2015_september_turner_david.pdf [ 15.1MB ]
- JSON: 24-1.0167182.json
- JSON-LD: 24-1.0167182-ld.json
- RDF/XML (Pretty): 24-1.0167182-rdf.xml
- RDF/JSON: 24-1.0167182-rdf.json
- Turtle: 24-1.0167182-turtle.txt
- N-Triples: 24-1.0167182-rdf-ntriples.txt
- Original Record: 24-1.0167182-source.json
- Full Text