Towards Automated Logging of Ore: Positive Identification of Sulphides in the Ores of Agnico Eagle Kittilä and New Boliden Mines

(1)

MASTER'S THESIS

Towards Automated Logging of Ore

Positive Identification of Sulphides in the Ores of Agnico Eagle Kittilä and New

Boliden Mines

Anna Arqué Armengol

2015

Master of Science (120 credits)

Natural Resources Engineering

Luleå University of Technology

(2)

TOWARDS AUTOMATED

LOGGING OF ORE

Positive identification of sulphides in the ores of

Agnico Eagle Kittilä and New Boliden mines

Anna Arqué Armengol

August 2015

Supervisors:

Cecilia Lund

Pertti Lamberg

(3)

ii

Abstract

Drill core logging in the mining industry has traditionally been done by geologists and technicians using basic tools such as a loupe, knife, and magnet, amongst others. The descriptions are usually time-consuming, qualitative, subjective and require the use of expensive complementary techniques, i.e. chemical assays, for acquisition of precise information. Mineralogical and textural relations valuable for the beneficiation stages in a geometallurgical approach are usually not considered; therefore, techniques that allow automated logging and a more detailed characterisation of the ores are necessary for a mineralogical-oriented logging system. Hyperspectral imaging has been recently used in the iron ore industry since the main iron minerals (i.e. hematite, goethite, jarosite) are easily distinguished with this technique. Hyperspectral imaging in the sulphide and gold ore industry is not as extended and is focused on an exploration level, as alteration minerals indicators of the mineralized areas are easily traceable. However, no direct detection of sulphide minerals is at the moment possible with the available commercial techniques. Possible sulphide discrimination and characterisation of their textural relations during logging would allow better knowledge of the ore and a better behaviour forecasting in the beneficiation process. Tests for sulphide detection have only been performed with microwave heating under infra-red thermography for ore gradation and sulphide detection purposes.

This Master Thesis presents the results of a mineralogical and textural characterisation of three different sulphide-bearing ores of both orogenic gold and polymetallic VMS deposits from the northern Fennoscandian Shield. A detailed mineralogical and textural characterisation using optical microscopy, SEM-EDS and XRD is included for the evaluation of newly tested techniques. Hyperspectral imaging in different bandwidths and resolutions is evaluated for mineral detection. For the orogenic gold ores, hyperspectral imaging could be used for exploration purposes as the proximal areas to the mineralization presented different mineral responses than the distal ones. For the polymetallic VMS ores, not all alteration minerals of interest for the characterisation of the main lithologies within the deposit were detectable in hyperspectral imaging. Moreover, the high-resolution hyperspectral scans did not provide additional mineralogical information for the technique to be implemented at a mine production scale.

The development of new techniques using IR thermography are promising and indicated the potential to detect and characterize sulphide minerals using illuminating sources, and the possibility to discriminate between sulphides using heat sources, i.e. laser heating. Sulphide minerals reflected slightly higher thermal radiation than the non-sulphide mineralogy in the blowtorch technique and the presented well defined contours for the medium- to coarse-sulphide grains. The finest sulphide grains were also detected as higher thermal responses were observed in within the fine silicate-matrix. In the laser technique, galena and pyrite grains could be differentiated from sphalerite and non-sulphide minerals. However, no distinct responses between pyrite and arsenopyrite could be obtained.

The tests performed indicate that the use of combined techniques is required for full automation of drill core logging in the mineralogically complex ores from orogenic and polymetallic VMS gold. Additionally, more research needs to be performed for reduction of the subjectivity, easiness of the techniques setups, and possibility to achieve quantitative mineralogy for early forecasting in a geometallurgical context.

(4)

iii

Acknowledgements

I would like to thank in first instance my supervisors at LTU. Dr. Cecilia Lund, I am greatly thankful for all your time and dedication, unconditional guidance throughout this project, as well as for the numerous written corrections. Prof. Pertti Lamberg, thanks for the supervision along the project and for the contribution of ground-breaking ideas. Many thanks also to Dr. Therese Bejgarn, for leading the ORESC program, for your expertise in the gold ores and for your written corrections.

Special thanks also go out to all people at Agnico Eagle Finland Oy and New Boliden. Leena Rajavuori and Jukka Välimaa, thank you for all the provided data and your time. Many thanks also to all the talks from the people in the geological, mining and metallurgy departments in Agnico Eagle. Lena Albrecht and Johan Magnusson, thanks for your time in Boliden. Also many thanks to the people providing the hyperspectral imaging data: Rainer Bärs from SpecIm Ltd., Phil Harris and Luisa Ashworth from GeoSpectral Imaging Ltd.

I would also like to thank Dr. Martin Simonsson, for your dedication and involvement in the project, and for sharing the knowledge in imaging and statistics. Snr Lec. Per Gren, thanks for the experimental setup and provision of the thermal camera. Lec. Nils Jansson, thanks for your expertise in the polymetallic ores. Additional thanks goes to my classmates Jara Fraile del Río, Mahamudul Hashan, Renato Contessotto and Dan Oliric Manaig for starting the sample preparation and mineral characterization as first stage of the ORESC program.

Many thanks to my family, for their unconditional support and for being always so close even if living thousands of kilometres away, moltes gràcies per tot! Also, I am very grateful to the time shared with my friends here up north: p i l h nk o J K n o n E oğ n, you made everything much easier! Ramon, for all the shared runs! Finally, thanks to the EMerald family, for these two years of living and studying together, I will miss you!

(5)

iv

1. Introduction

The concept of automation in drill core logging is raised in a geometallurgical context. Geometallurgy is the semi-discipline that combines geological and metallurgical information to create spatially-based models for forecasting in mineral processing plants (Lamberg 2011). The first stage of a geometallurgical program consists of geological data collection in exploration and production mining programs. Diamond drilling represents a significant part of the costs for data collection. The vast amounts of drill cores generated require detailed examination from the geologists, and in most of the cases, additional data acquisition from chemical analyses. These analyses are usually sent to external laboratories and therefore imply a significant turnaround time (Darling 2011; Kruse 1996; Ross et al. 2013). In addition, not all drilled cores are always logged, or they are only assayed for few critical elements and thus the drilling data generated is usually not exploited at its full potential.

A new logging methodology is therefore necessary to facilitate the daily tasks of the geologists and allow cost-effective, non-destructive and fast logging of large quantities of drill cores. The rock properties (i.e. petrophysical properties, geochemistry) and quantitative mineralogy (i.e. modal composition, mineral association, grain size, textural relations) should be considered for geometallurgical-oriented logging. This would provide more consistent, objective and statistically-robust data for the geological model, geotechnical characterization and grade control (Huntington et al. 2006; Lamberg 2011; Ross et al. 2013). Therefore, operational multi-sensor technology is urged for the automation of drill core logging in a digital format, ensuring the availability of quantitative information at the stages of ore evaluation and production planning (Huntington et al. 2006; Kruse 1996; Ross et al. 2013).

This Master Thesis is part of a larger Optimising Resource Characterisation program (ORESC). The project is a partnership within the RockTechCentre, Luleå University of Technology, Agnico Eagle Finland Oy, New Boliden and LKAB. The program integrates the utilization of imaging techniques for automated scanning of rock samples with respect to lithology, mineralogy and rock mechanical properties. This Master Thesis focuses on positive identification of the major sulphide minerals in the Agnico Eagle Kittilä orogenic gold ores and in the volcanogenic massive sulphide (VMS) ores of Boliden. These include minerals such as pyrite, arsenopyrite, galena, sphalerite and chalcopyrite. Drill core samples will be used for the study, previously scanned with scanning electron microscopy and hyperspectral imaging (HSI) during the first stage of the ORESC program, conducted for students. The ―ground truth‖ mineralogy of the scanned drill core samples will be used as the base for validation of the detected ―HSI-min l ph ‖. Fin lly n w m ho ology will b v lop for the sulphide detection over drill core samples using infra-red (IR) thermography.

1.1. State of the art

Several techniques and operational devices have been developed during the recent years for automated mineral identification in drill cores. These are typically based on the technology used for remote sensing, imaging spectrometers or hyperspectral sensors, which measure the reflectance of the Earth´s surface in hundreds of spectral bands (Kruse 1996; Ramanaidou and Wells 2012; Ross et al. 2013). The iron ore industry is one of the main user of hyperspectral imaging as detection of Fe-oxides of sedimentary origin (i.e. hematite, goethite, jarosite) are well-detected in the infrared wavelengths (Ramanaidou and Wells 2012). The use of these systems in exploration programs is also developed due to well-detectability of most alteration minerals (e.g. chlorites, clays, phyllosilicates, and sulphates). However, direct detection of metallic minerals such as sulphides, gold, magnetite and others is not

(8)

2 available in hyperspectral imaging (Harris 2014; SGS 2014). Logging of ore minerals is nowadays done indirectly by detection of the alteration mineralogy within the deposit (Ramanaidou and Wells 2012; SGS 2014, SPECIM 2014).

Even though no commercial techniques exist for sulphide mineral identification, few tests have been carried out recently under microwave heating. The tests conducted by Weert (2007) for ore sorting purposes determined that sulphide-bearing rocks responded readily to microwave radiation; however, no discrimination within sulphides was achieved. The need to study and develop techniques that allow full automation of drill core logging should therefore consider:

1) Full mineral identification, and special attention to ore minerals.

2) Textural characterisation at a drill core scale, including identification of fractures, veins and weak zones for geotechnical purposes. Characterisation at a mineralogical scale, providing particle relations that allow geometallurgical interpretation.

3) Improvements of the available scanning instruments in order to provide better operational conditions, and even portable equipment.

1.2. Objectives

The aim of this Master Thesis is to assess and propose state of the art techniques that allow automation of drill core logging for sulphide-bearing ores. For this, three types of ores (the Suurikuusikko and Rouravaara orogenic gold, the Kristineberg hydrothermally-altered VMS and the Garpenberg skarn-altered VMS) are selected. The report is divided into two parts:

Part I comprises an overall characterisation of the orogenic gold ores and focuses in:

1) Validate the hyperspectral imaging outcome for differentiation and detectability of lithologies, mineralogy and textural properties of the ores.

Part II includes a detailed mineral and sulphide characterisation of few drill core samples of the orogenic gold and polymetallic ores, with main objectives to:

2) Evaluate the (high-resolution) hyperspectral imaging results based on mineral detectability and textural characterisation of the sulphide-bearing samples for mine production automated logging.

3) Explore and develop techniques that use IR thermography to provide positive sulphide detection that complement already existing techniques for automated drill core logging. The development of a new technique is included in the third objective. The research relies on the known differences in the thermal and reflectivity responses that literature studies show for various sulphide minerals. Therefore, illuminating and heat sources will be used in combination with infra-red thermography, representing fully innovative testing that corresponds to the initial phases of the Technology Readiness Level (TRL) (Figure 1). According to the description of each TRL, the achieved level lies around 3 and 4 as the concept that sulphides can be detected is proved and active research in lu ing ―low fi li y ― l bo o y u i are conducted during this Master Thesis.

(9)

3

Figure 1. Technology Readiness Level scale. Each level characterizes the progress in the development of a

technology, from the idea (Level 1) to the full deployment of the product in the marketplace (Level 9).

2. Literature survey

This section includes: 1) the description and theoretical principles of traditional and semi-automated drill core logging techniques used for the characteristics of the mineral properties of ores; 2) the revision of previous research done for sulphide identification and the theoretical principles for development of new techniques; and 3) the geological context of the sulphide-bearing ores selected in this study.

2.1. Practices in drill core logging

Drilling and logging allow data acquisition regarding lithology, alteration, structure, and ground conditions of a studied area, and logging refers to the standardized system used for direct data recording of a drill core or drill cuttings, and indirect data recording of boreholes (Hartman and Mutmansky 2002; Darling 2011). Murphy and Campbell (2007) claim that logging procedures at site can be inconsistent or use vague nomenclatures, susceptible to bring up quality control issues. Riles and Cottrell (2012) indicated the low cost of the logging procedures in comparison to the drilling costs. Thus, suggested that drill core logging within a mining context should be multiple use of a single drill hole and include data regarding geological, geotechnical and metallurgical character altogether. Two logging approaches are presented in this section with their corresponding advantages and drawbacks. 2.1.1. Traditional drill core logging procedures

Current practices in drill core logging are classified according to three main systems (Marjoribanks 2010): prose logging, where description of the intervals of interest is done in words; analytical spread-sheet

logging, where the rock characteristics are described in categories (e.g. colour, grain size, mineral

content); and graphical scale logging, which is a combination of the latter types and includes a graphical log and the corresponding interval description (Table 1). Darling (2011) distinguishes within two logging systems: the graphic and narrative-style, and the data-entry style, where a code is associated to each logged feature. In some cases an intensity estimate (e.g. weak, moderate) is also included. Darling (2011) refers to different logging forms, i.e. paper-based, portable computer-based spreadsheets, database

(10)

4 software; and different recording data type, i.e. graphic, alpha, and numeric, which are subjected to the logging purpose and the software utilized.

Table 1. Classification of the basic logging systems for drill cores (Marjoribanks 2010). Logging

system Description Advantages Drawbacks

Prose logging Written description of the selected _{intervals in depth.} Includes argumentation _{and discussion.} Ineffective, subjective data _{records, not quantitative.} Analytical

spread-sheet logging

Description in columns of different categories i.e. colour, grain size,

mineralogy, type of veins, alteration, etc. Symbols, abbreviations and numbers used for semi-quantification.

Standardized and detailed. Ideal for direct computer entry by codes.

Limited to the categories of observation established and predefined intervals. Cannot show gradual changes.

Graphical scale logging

Pictorial log with columns for lithology, alteration, veining and structure. Allows description or commentaries.

Different features can be described at different depths. Indicates gradual changes.

Slow and tedious, not suitable for large drilling programs, not quantitative. *The classification of Darling (2011) combines the prose and graphical scale logging defined by Marjoribanks (2010) into one system, the graphic and narrative; the data-entry style system is equivalent to the analytical spread-sheet logging.

A drill core logging methodology is mostly dependent on the data precision needed at each level of study, i.e. exploratory, pre-feasibility, feasibility, or on-going mining stages. But it is also dependent of the mineralization context, i.e. orebody type, host rocks, alteration degree, influence of the geological structures (Murphy and Campbell 2007). The recommended logging stages based on different authors’ points of view are summarized in six stages as follows (Hartman and Mutmansky 2002; Moon et al. 2006; Murphy and Campbell 2007; Darling 2011):

1) Selection of wet or dry core logging, or a combination of both.

2) Overall drill core examination for detection of lithology contacts, structures, veins, and ore occurrences; evaluation of the core recovery (less than 85 % core recoveries cannot be considered as a representative sample).

3) Detailed geological drill core examination to mark the exact limits of lithology, alteration, mineralization, structure, and other relevant features such as grain size, texture, colour, etc. Provision of systematic and quantitative descriptions or observations, such as estimation of ore mineral percentages. The use of handy specimens such as hand lens, knife, UV-lamp or others may be adequate for the optimal core examination.

4) Calculation of geotechnical parameters, such as rock quality designation (RQD), fracture frequency (FF), joint condition, rock stress classification, etc.

5) Provision of down-hole measurements with physical constants measurements and completion of the previously detected geological and geotechnical descriptions. In some cases, core orientation is considered for further rock mechanical calculations or 3D modelling purposes. 6) Marking of the sampling and assaying areas for metallurgical purpose, and / or possible areas

needed of more accurate microscopic studies – thin and polished sections.

7) Completion of the detailed log, photographic record box by box, and of any detailed area of interest, and summary provision in the drill-hole report.

8) Storage of the drill cores in correctly labelled boxes for further sampling, and posteriorly, for any future need of revision.

The mentioned logging procedure is established to systemize and minimize the logging subjectivity of the geologists and geotechnicians, and exemplify a multiple use of a single drill core logging. However,

(11)

5 parameters such as the lithological contacts may be diffuse, and quantification of the alteration degree or ore grades might be roughly estimated without the use of complementary techniques (e.g. chemical analyses). Also, depending on the stage of the study, the considered logging intervals may vary, as well as the precision of the descriptions (Marjoribanks 2010).

2.1.2. Mineralogical-oriented drill core logging techniques

Several techniques that provide a higher mineralogical characterisation during logging already exist. These constitute the basis of the geometallurgical approach, where the collected geological data has a quantitative character and is utilized for the creation of a geological model. The model should include additional mineralogical properties: rock properties, mineral associations, grain size and variations within the ore (e.g. grade, penalty elements) to the traditional core logging system. These properties can be introduced to the block model and are utilized to: increase the knowledge of the ore bodies, and better control the mining planning and metallurgical performance. Therefore, substantial improvements in the resource exploitation and overall recovery of the beneficiation process can be achieved (Lamberg 2011).

Optical microscopy is the most widely used method for mineral and textural characterisation; however, the quality is dependent on the skills of the mineralogist and the resolution limit is about 1 µm for 500x magnification (Table 2). Scanning Electron Microscope (SEM) provides high quality images (Table 2) and coupled with Energy Dispersive X-ray Spectroscopy (EDS) can provide quantitative analysis. This is though, expensive and very time consuming (Zhou et al. 2004). X-Ray Diffraction (XRD) can provide quantitative analyses if combined with the Rietveld method. Nevertheless, XRD is dependent on preferred mineral orientations, crystal size and crystallinity (Lund et al. 2013). Analytical assays can be used for mineral quantification (i.e. modal mineralogy) if the elemental composition is transformed with the mineral-to-element conversion tool. Yet this requires non-complex mineralogy (Lamberg 2011).

Table 2. Existing techniques for geological and mineralogical data collection (Haque 1999; Zhou et al. 2004;

Weert 2007; Marjoribanks 2010; Darling 2011; Harris 2014).

Technique Method Properties Advantages Drawbacks

Logging

(visual) Geological - geotechnical

logging

Lithology, alteration, grain size, colour, weak zones, RQD Easily adaptable to required descriptions Subjective, time consuming, qualitative, missing information Imaging Optical

microscope Mineralogy Inexpensive, reliable, fast Time consuming, qualitative, small amount

of samples Scanning electron

microscopy (SEM) Mineralogy High detection limits, reliable Time consuming, expensive, small amount

of samples Determinative

mineralogy X-ray diffraction (XRD) Mineralogy Relatively fast Expensive, small amount of samples,

qualitative Analytical

chemistry AAS, ICP, XRF Elements High detection limits, reliable No mineralogical information, expensive

Hyperspectral imaging (HSI)

SisuROCK,

SisuCHEMA Mineral phases (focused on alteration

mineralogy)

Fast, high amount

of samples Requires post-processing,

non-identification of ore mineralogy

(12)

6 Other semi-automated systems already commercialized, i.e. visible/near infrared spectrometry, allow data recording and semi-quantification. Spectrometers compose a non-destructive technique that can be used for mineral characterization at different scales: laboratory (powders, drill chips, drill cores) and field (face mapping) (Clark 1995). Spectroscopy is sensitive to both crystalline and amorphous materials unlike other techniques (e.g. X-ray diffraction) and is highly sensitive to small changes in the structure of the material. The interpretation of these changes provide more detailed information of the chemical composition of the minerals (Clark 1999). As not all the minerals can be detected by spectroscopy, additional detectors can be added to the technique to increase the number of detectable mineralogy. These consist of detectors capable to measure: volumetric magnetic susceptibility, density using gamma-ray attenuation, or chemical elements using energy-dispersive X-ray fluorescence spectrometry (Rothwell and Rack 2006; Ross et al. 2013).

2.2. Mineral identification in hyperspectral imaging

2.2.1. Spectral features of the minerals

Multispectral or hyperspectral techniques are able to discriminate among minerals based on the differences existing within their spectral properties. The spectral features occur in the form of absorption bands in the visible and near infrared and are caused by electronic and vibrational processes (e.g. crystal field effects, charge-transfers, colour centres, transitions to conduction bands, and overtone vibrational transitions) (Hunt 1977; Clark 1999).

Detectable minerals consist of clay minerals, OH-bearing minerals, Fe-oxides and hydroxides, carbonates, sulphates, olivines and pyroxenes. Different features are detected within a determined bandwidth and due to different processes. Fe-oxides and transition metals are detected in the visible/near infrared bandwidth due to electronic transitions. Hydroxyls (e.g. clays and phyllosilicates), H2O, carbonates, hydrated sulphates and OH-bearing minerals (e.g. amphiboles) are detected in the

short-wave infrared bandwidth as a result of vibrational transitions. Silicates (e.g. olivine, pyroxene, garnet, quartz, feldspars) are better distinguished in the long-wave infrared (Harris et al. 2012).

2.2.2. Principles of spectroscopy and factors of influence of the spectral features Spectroscopy is the study of light as a function of an emitted, reflected or scattered wavelength from a solid, liquid or gas (Clark 1999); and reflectance spectroscopy is the base of the hyperspectral imaging technique. Spectrometers can work on different spectral ranges, depending on the spectral absorptions that need to be covered (Table 3).

Table 3. Spectral range and respective wavelength lengths (Clark 1999).

Common use Remote sensing use

Spectral range Wavelength length Spectral range Wavelength length

Visible (RGB colour) 0.4 to 0.7 µm Visible-near-infrared (VNIR) 0.4 to 1.0 µm

Near-infrared (NIR) 0.7 to 3 µm Short-wave infrared (SWIR) 1.0 to 2.5 µm

Mid-infrared (MIR) 3.0 to 30 µm Mid-wave infrared (MWIR) _{Long-wave infrared (LWIR)} 3.0 to 6.0 µm _{6.0 to 14.0 µm}

The physical principles of interaction within photons and plate surfaces are summarized from Clark (1999): a group of photons are reflected or refracted when contacting a medium with a change in the refraction index. The index of refraction of each mineral is described by Equation 1:

(13)

7 where m = complex index of refraction; n = real part of the index; ( ) _{; and}_K_{= imaginary} part of the index of refraction. Absorption of the photons when entering a medium is described by Beers Law (Equation 2):

₍₂₎

where I = observed intensity; I0 = original light intensity; k = absorption coefficient; and x = distance travelled through the medium. And the absorption coefficient is related to the complex index of refraction (Equation 3):

(3)

where λ = wavelength of light.

The absorption coefficient as a function of wavelength is more relevant to spectroscopy as it is more variable depending on the materials, and especially at visible and NIR wavelengths; the refraction index is less variable. Therefore, the spectral reflectance of a mineral is mostly controlled by its absorption grade, which occurs due to two main processes: electronic and vibrational (Clark 1999). Electronic processes usually cause broad absorption features, whereas vibrational processes cause sharp ones (Raja et al. 2010).

Electronic processes exist if transition of an electron from one energy level to another occurs while the

element is hit by a photon. The most common absorption features due to electronic processes are crystal field effects such as unfilled electron shells of transition elements (e.g. Ni, Cr, Co, Fe); and charge transfer between ions (e.g. Fe2+_{and Fe}3+_{, which provide the typical absorption bands for iron}

oxides) (Figure 2). Other electronic processes are colour centres, occurring due to impurities in the mineral lattice, and conduction bands, corresponding to the band gap existing within high level of electrons moving freely, and low level of electrons attached to the atoms in the valence orbitals (Hunt 1977; Clark 1999).

Vibrational processes include bending and stretching vibrations of bands between radicals or molecules. A

molecule with N atoms has 2N-6 modes of vibrations, called fundamentals. Additionally, vibrations can include multiples of single fundamental mode, referred as overtones, and different modes of vibration, referred as combinations. Therefore, a molecule with fundamentals v1, v2, v3 can have overtones at 2v1,

3v1, 2v2 and combinations at v1+v2, v2+v3, v1+v2+v3, etc. Each overtone and combination of a higher

level provides a weaker absorbance response; however, those can be detected in spectroscopy even if the result is a rather complex spectrum of a mineral (Hunt 1977; Clark 1999). Vibrations are only observed in spectroscopy if a dipole moment exists. Typical absorption features due to vibration process of common minerals and molecules are (Clark 1999):

- Quartz: Si-O-Si asymmetric stretch fundamental at 9 µm and between 12 and 13 µm.

- H2O- and OH-bearing minerals: overtones of the OH stretches at around 1.4 µm (e.g. kaolinite)

and combinations of H-O-H bend with the OH stretches at around 1.9 µm (e.g. halloysite).

- Carbonates: the strongest overtone bands of the CO3 fundamentals occur around 2.50- 2.55 µm

(14)

8

Figure 2. Spectral signature diagram showing the mineral absorption features due to electronic and vibrational

processes; the black bars indicate the relative widths of absorption peaks (Hunt 1977).

Apart from the mentioned variable responses of single minerals, reflectance of a particulate surface is more complex as a certain percentage is absorbed while the photons encounter each grain. Several factors influence the reflectance of the particulate surfaces, which increment the difficulty of spectra identification (Clark 1999):

- Grain composition: different absorption rates due to presence of particular chemical elements or ions, ionic charge of certain elements, and the geometry of chemical bonds between elements (principle of the spectroscopy technique for mineral characterization).

- Grain colour: bright grains scatter most of the photons and create a random walk process that can develop hundreds of encounters; dark grains absorb most of the photons and few encounters occur.

- Grain size: in visible and NIR bandwidths, reflectance decreases as grain size increases; in MIR grain size effects are more complex as adsorption coefficients are much higher.

- Viewing geometry: the angle of incidence and the angle of reflection of the photons in the particulate surface can slightly affect the band depths.

- Crystallinity of the minerals, water and organic matter presence in the samples, sample roughness, and temperature are also influencing parameters in the reflectance of the particulate surfaces.

(15)

9 2.2.3. Hyperspectral imaging and mineral detectability

Hyperspectral imaging combines digital imaging and reflectance spectroscopy where each pixel in an image contains the light intensity of several hundred, narrow adjacent spectral bands (Ramanaidou and Wells 2012). Therefore, all pixels in an image contain a derived continuous spectrum (Figure 3). Models are created to correlate the different spectra with the mineralogy. Spectral libraries compiling the characteristic spectra of particulates are used to identify the mineralogy of hyperspectral images by matching each image spectrum individually to one of the reference reflectance spectra in a spectral library; however, this technique requires complete conversion of the spectra into reflectance (Clark 1999; Smith 2012). Since most hyperspectral images contain pixels representing spatial mixtures of different materials, composite spectra is derived. Therefore, the spectrum fitting the best within the observed spectra and the spectra from spectral libraries is assigned to the pixel. The resulting image is a map of dominant materials for the assigned image pixels (Smith 2012).

Minerals present characteristic positions, strength (or depth), and shapes of the absorption features; whereas rocks and soils may lack distinctive absorption features. Therefore, the spectra of rocks and soils must be characterized by the overall shape. Different matching methods exist: fitting through the minimum difference in reflectance, which uses the square root of the sum of the squared errors; if considering each spectrum as a vector in the spectral space fitting is done by finding the smallest angles within observed and reference spectrum (Smith 2012).

Figure 3. Schematic representation of acquisition of a continuous spectrum for each pixel processed in

a hyperspectral image (Smith 2012).

Table 4 shows that similar minerals provide results over different wavelengths ranges, referred to as duplication. Therefore, when mineral combinations complicate the identification of an area, the confirmation of a certain mineral presence can be determined by adjacent wavelength ranges (GeoSpectral Imaging, 2014).

(16)

10

Table 4. Mineral identification table using hyperspectral imaging based on different wavelengths (Baldridge et al.

2009; Clark et al. 1990; Harris 2014; SGS 2014).

Structure Mineral Group Example _VNIR Responses in different bandwidths* _SWIR _LWIR

Inosilicates Amphibole Actinolite Non-Diagnostic Good Moderate

Pyroxene Diopside Good Moderate Good

Cyclosilicates Cordierite Cordierite Non-Diagnostic Possibly Possibly

Tourmaline Elbaite Non-Diagnostic Good Moderate

Nesosilicates Andalusite Andalusite Non-Diagnostic Non-Diagnostic Possibly

Garnet Grossular Moderate Non-Diagnostic Good

Olivine Forsterite Good Non-Diagnostic Good

Sorosilicates Epidote Epidote Non-Diagnostic Good Moderate

Phyllosilicates Mica Muscovite Non-Diagnostic Good Moderate

Paragonite Non-Diagnostic Good Non-Diagnostic

Phlogopite Non-Diagnostic Good Possibly

Chlorite Clinochlore Non-Diagnostic Good Moderate

Clay Minerals Illite Non-Diagnostic Good Moderate

Kaolinite Non-Diagnostic Good Moderate

Talc Non-Diagnostic Good Possibly

Tectosilicates Feldspar Orthoclase Non-Diagnostic Non-Diagnostic Good

Albite Non-Diagnostic Non-Diagnostic Good

Silica Quartz Non-Diagnostic Non-Diagnostic Good

Carbonates Calcite Calcite Non-Diagnostic Moderate Good

Siderite Non-Diagnostic Moderate Possibly

Dolomite Ankerite Non-Diagnostic Moderate Possibly

Dolomite Non-Diagnostic Moderate Good

Hydroxides Gibbsite Non-Diagnostic Good Moderate

Sulphates Alunite Alunite Moderate Good Moderate

Gypsum Non-Diagnostic Good Good

Phosphates Apatite Apatite Moderate Non-Diagnostic Good

Monazite Monazite Moderate Non-Diagnostic Good

Oxides Hematite Hematite Good Non-Diagnostic Non-Diagnostic

Ilmenite Ilmenite Non-Diagnostic Non-Diagnostic Inferred

Rutile Rutile Non-Diagnostic Non-Diagnostic Possibly

Spinel Chromite Non-Diagnostic Non-Diagnostic Non-Diagnostic

Magnetite Non-Diagnostic Non-Diagnostic Non-Diagnostic

Sulphides Sulfarsenide Arsenopyrite Inferred Non-Diagnostic Non-Diagnostic

Sulphides Chalcopyrite Inferred Non-Diagnostic Non-Diagnostic

Galena Inferred Non-Diagnostic Non-Diagnostic

Pyrite Inferred Non-Diagnostic Non-Diagnostic

Pyrrhotite Non-Diagnostic Non-Diagnostic Non-Diagnostic

Sphalerite Non-Diagnostic Inferred Non-Diagnostic

Native element Copper Gold Non-Diagnostic Non-Diagnostic Non-Diagnostic

Graphite Graphite Non-Diagnostic Non-Diagnostic Non-Diagnostic

*Legend: Good, well-characterized minerals in IR; Moderate, identifiable minerals in IR but not easily distinguished in

low-resolution scans; Inferred, inferred-characterization if combined with RGB or other bands; Non-diagnostic, not detectable at present time in IR; Possibly, possible characterization according to observed spectra in (Baldridge et al. 2009).

2.3. Mineral identification in infra-red thermography

2.3.1. Thermal properties of the minerals

The ability of a mineral to transfer heat is referred to thermal conductivity (Equation 4), measured as the heat in calories which passes through 1 cm2_{of surface 1 cm thick (Clark 1966). The thermal}

(17)

11 Metals and minerals with substantial contributions of metallic bonding, e.g. graphite, where heat is transferred largely through the flow of free electrons present relatively high thermal conductivities. Ionic and covalent crystals present much lower thermal conductivities and strong anisotropy, which is greatest in the direction of closer atomic packing (Wenk and Bulakh 2004).

(4)

Where q is the flux, is the thermal conductivity, and T is the temperature gradient. The negative

value indicates that heat is transferred in the direction of the decreasing temperature. Table 5 indicates the conductivity values of some minerals.

Table 5. Thermal conductivity (K) of single cubic and non-cubic crystals measured at different temperatures (T).

K c is the conductivity parallel to axis c; K c, perpendicular to axis c (Clark 1966; Wenk and Bulakh 2004).

Cubic

crystals (°C) T (10-3_{cal/cm sec °C)}K

Non-cubic

crystals System (°C) T (10K c K c -3_{cal/cm sec °C)}

Hematite 30 28.9 Calcite Trigonal 30 10.0 8.4

Magnetite 22 11.9 300 4.18 3.52*

Chalcopyrite - - Quartz Trigonal 30 28.9 35.1

Galena - - 300 11.3 6.5*

Pyrite 0 216.4 Graphite Hexagonal 300 89 355*

Sphalerite 0 151.9 Rutile Tetragonal 36 30 -

* indicates that the values correspond to K a, instead of K c. 2.3.2. Principles of thermography and heat transfer

Thermography, or thermal imaging, refers to the method used to determine the spatial distribution of heat on an object, and its time dependence (Gaussorgues 1994). IR thermography is a non-destructive testing technique consisting of an IR camera that converts IR radiation, corresponding to the electromagnetic spectrum between 0.9 and 14 µm, to a visual range by displaying thermal variations across an object (Ibarra-Castanedo et al. 2007; Flir Systems Inc. 2013). IR thermography can be: passive, if the object is at a higher or lower temperature than the background; or active, if an energy source is required to produce thermal contrast within the object and the background (Ibarra-Castanedo et al. 2007).

Heat transfer refers to the energy exchange due to differences in temperature of a medium (fluid or solid) and the surroundings. Conduction, convection and radiation are the three processes of heat transfer, being the latter related to thermography. All objects above the absolute zero temperature emit IR radiation, and the quantity of radiation increases with the temperature (Incropera and DeWitt 1999; Ibarra-Castanedo et al. 2007). Radiation is attributed to changes in the electronic configuration of the atoms, and the energy of the radiation field is transported by electromagnetic waves, or photons. Contrary to conduction and convection, radiation does not require the presence of a medium (solid or liquid) for the transfer to occur and behaves more efficiently in the vacuum (Incropera and DeWitt 1999).

Objects do not only emit radiation but also react to incident radiation from the surroundings by absorbing and reflecting some of it. This is described by the Kirchoff´s Law (Equation 5), where the coefficients E, T and R correspond to emission, transmission and reflection.

(18)

12 A perfect emitter (E = 1) and perfect absorber of radiant energy is referred as a blackbody, not existing in the nature but fundamental for the IR thermography theory. Also, Kirchoff´s Law states that emittance (E) equals to absorbance, as both values vary with the radiation wavelength. Planck´s Law relates the radiative properties of a blackbody in function of temperature, T, in Kelvins, and w v l ng h λ (Incropera and DeWitt 1999). Planck´s Law is represented as a series of curves that show the radiation per wavelength unit and area unit (or spectral radiant emittance of the blackbody); as seen in Figure 4, where the higher the temperature, the higher the emitted radiation. Therefore, an object at 30°C can be observed at wavelength around 10 µm, and an object at 1000°C, at 2.3 µm, wavelengths comprised in the IR spectrum (Flir Systems Inc. 2013).

Figure 4. Planck´s Law curves: relation within the spectral radiant emittance of a blackbody – y-axis, and the

spectral wavelength, in µm – x-axis. The curves represent the temperature that correspond to each spectral radiant emittance at a certain wavelength; maximum temperatures for each curve are pointed out (Flir Systems Inc. 2013).

The emissivity of a material is evaluated as the ratio of the energy radiated by the material at temperature T and the energy radiated by a black body at the same temperature; therefore, emissivity ranges within 0 to 1. Generally, the spectral emissivity of gases and liquids varies greatly with wavelength, whereas it fluctuates little for solids. Metals usually present high emissivity that increases steadily with temperature until oxidation occurs on the surface of the heated metal. If the emissivity of an object is low, the temperature can easily be calculated by Planck´s equation; however, if the reflectance of an object is high and thus the emissivity low, it is difficult to characterize the radiation due to self-emission and due to reflection (Gaussorgues 1994).

Radiometry is the measurement of radiant electromagnetic energy, associated to the IR spectrum, and with typical unit of radiance (expressed as Watts/steradian-cm2_{). Wh h mog phy ―how}

ho ‖ n obj i iom y ―how mu h n gy‖ h obj l . Th m l m m u irradiance, not temperature; those are related according to Planck´s Law. Therefore, radiance measurements are easier to process than temperature measurements (Flir Systems Inc. 2013).

2.3.3. Thermal imaging

Modern thermal cameras allow the detection of small thermal variations at high spatial resolution, where each pixel corresponds to the sum of all contributions over a fixed spectral range from the objects in the scene. Image processing strategies are required to recover the information from the imaging sequences recorded at different spectral bands or over a period of time (Gagnon et al. 2014).

(19)

13 According to Flir Systems Inc. (2013), several factors influence the IR camera temperature detection of an object, of which the most important are: the atmosphere, located between the object and the thermal camera, attenuating the radiation due to gas absorption and particle scattering; the reflections from the surroundings of the object, which interfere with the self-radiation of the object. The working principles of a thermal camera can be summarized in the following five steps (Flir Systems Inc. 2013):

1) Specific energy signature collected by the IR camera depending on the target object.

2) Depending on the detector used, the energy signature is collected as photons (photon detector) or as heat energy (thermal detector).

3) A signal voltage is detected corresponding to the collected energy. The signal is converted into digital count through A/D converter; the more IR energy incident on the detector, the higher is the digital count.

4) Transformation of the digital counts into radiance values.

5) Conversion of the radiance values to temperature using the measured (or known) emissivity (E) of the target object.

Most rocks absorb the infrared radiation at different energies; thus, the thermal camera contrasts will provide a characteristic spectral signature that is captured by different detectors of the thermal camera (Gagnon et al. 2014). However, as ores and minerals are in equilibrium with the surrounding atmosphere and do not reflect much heat, an active approach where uniform heating is applied to the samples should be adopted for better thermal contrast acquisition (Ghosh et al. 2014). Microwave heating in combination to thermal imaging was tested by Yixin and Chunpeng (1996), Weert (2007), Djordjevic et al. (2010), Koleini and Barani (2012), Ghosh et al. (2014) and Gagnon et al. (2014) for different purposes. Djordjevic et al. (2010) and Gagnon et al. (2014) tested thermal imaging in combination with spectral imaging for acquisition of information within several wavelengths.

2.3.4. Microwave heating

Microwaves create molecular vibrations by migration of ionic species or rotation of dipolar species (Haque 1999). Microwave heating does not transmit the heat from the surface of a material to its interior as conventional heating conduction methods do; instead, it generates heat from the interior of the materials and dissipates it (Yixin and Chunpeng 1996). Rocks are heated or cooled depending on their composition and optical surface properties. Most sulphides, arsenides, carbon compounds, graphite, and some oxides, respond to microwave radiation, whereas most silicates, carbonates and sulphates are transparent to microwave energy (Yixin and Chunpeng 1996; Weert 2007; Koleini and Barani 2012). According to Yixin and Chunpeng (1996), sulphide minerals present faster heating rates than oxide minerals, darker minerals tend to heat faster than lighter minerals, and elements with impurities or defects in the crystal lattice are heated differently than pure minerals. Also, rocks with higher average weight seem to heat up faster and cool slower (Weert 2007).

According to Weert (2007) advantages of microwave heating include detection of sulphide minerals in the IR, and the little sample preparation required since similar heating responses have been obtained for washed, dusty and wetted samples. A main disadvantage of this technique is related to the sinusoidal waveform emitted by microwaves: non-homogeneous heating responses of the rocks are dependent on the sinusoidal waveforms, and on their location and orientation in the microwave. The load of the microwave also influences on the thermal responses. A list of the non-detectable minerals in microwave radiation according to Haque (1999) is shown in Table 6. Table 7 includes the heating temperatures

(20)

14 reached under microwave radiation for different minerals over long periods of time. Those experiments were mainly focused on detection of maximum temperatures for mineral processing purposes. Shorter time exposure of the samples towards heat is enough for mineral identification, and in order to avoid effects such as fumes, sparkling, or mineral transformations. However, the variations in temperature observed within sulphides in IR thermography in contrast to other gangue minerals indicate possible positive identification of those.

Table 6. Minerals and compounds transparent to microwave radiation, tested for 5 minutes at 2450 MHz and

150 W (Haque 1999).

Mineral class Minerals

Carbonates Aragonite and calcite (CaCO3), dolomite (Ca,Mg(CO3)2), siderite (FeCO3)

Silicates Almandine (Fe3Al2(SiO4)3), allanite (Ce-, La-, Nd-, Y-hydrated silicate), anorthite

(CaAl2Si2O8), gadolinite (Ce,La,Nd,Y)2FeBe2Si2O10

Sulfates Barite (BaSO4), gypsum (CaSO4∙H2O)

Others Fergusonite ((Ce,La,Nd)NbO4), monazite ((Ce,La,Nd,Th)PO4), low-Fe sphalerite

((Zn,Fe)S), stibnite (Sb2S3)

2.3.5. Laser heating

Another source of infrared radiation are lasers (Gaussorgues 1994) that in combination with IR thermography may provide positive identification of minerals. Lasers can transmit large amounts of energy at near unlimited rates, which induce high heating rates (103_{– 10}10_{K /s) within an effective}

absorption depth (Elhadj et al. 2014). In order to understand its working principles, the three existing modes of excitation are presented: lock-in thermography and pulsed thermography, corresponding to external optical techniques; and vibrothermography, a non-optical technique that uses ultrasonic waves emitted by mechanical means that excite internal features.

The lock-in thermography mode illuminates periodically a surface using an intensity modulated laser beam, e.g. a halogen lamp, which emits thermal waves into the specimen. The thermal waves are typically sinusoidal, which may provide non-homogeneous surface heating. Usually, long acquisition times are required to obtain thermal responses and energy required is less than for other techniques, which may be interesting for low power source requirements and preservation of the materials.

Pulsed thermography consists of a high power heating source, e.g. photographic flashes and lasers, which

combine several periodic waves at different frequencies and amplitudes. The thermal response travels from the surface through the specimen, and as the time elapses, the surface temperature decreases uniformly. Discontinuities within the subsurface i.e. porosity, inclusions, etc. can be then detected with the IR camera. This technique is fast but provides non-uniform heating and is affected by emissivity variations, external reflections and surface geometry. However, advanced processing algorithms can be used to correct non-uniformity and the affecting parameters (Ibarra-Castanedo et al. 2007).

The use of lasers in combination with IR thermography is effective because parameters such as wavelength, beam size, exposure time, power, temporal and spatial shape, and pulse repetition frequency can be easily controlled. Furthermore, this technique is non-invasive and non-obstructive and it produces high temporal resolution temperatures, limited to the detector bandwidth (Elhadj et al. 2014). However, Elhadj et al. (2014) points out that temperature measurements in laser heating applications are lacking in literature studies, and that most of the applications of laser heating are usually related to specimen excitation to a photonic level, e.g. LIBS (Laser-Induced Breakdown Spectroscopy) and Raman spectroscopy.

(21)

15

Table 7. Mineral heating responses towards microwave radiation under long exposure times. Studies mainly focused on mineral processing purposes i.e. breakage of the mineral grains,

or for research purposes i.e. determination of the limit temperatures reachable for certain minerals.

Mineral

group Mineral

[1] [2] [3] [4] [5]

T (°C) t (min) T (°C) t (min) T (°C) Heating rate _(C/min) Power _(W) Heating response Product examination (long time _{exposures, high T reached)}

Sulphides

Arsenopyrite 80 Heats, some sparkling S and As fumes, some fusion. Pyrrhotite, As, _{Fe-arsenide and arsenopyrite} Chalcopyrite 707 1 920 1 >400 920 15 Heats readily with emission of S _fumes Two Cu-Fe sulfides or pyrite and Cu-Fe _sulfide

Galena 737 2 956 7 >650 130 30 Heats readily with much arcing Sintered glass of galena Pyrite 527 1.67 1019 6.76 150 30 Heats readily; emission of S fumes Pyrrhotite and S fumes Pyrrhotite 682 0.67 886 1.75 >380 500 50 Heats readily with arcing at high T Some fused; most unaffected Sphalerite 159 2.5 87 7 >160 12 100 Difficult to heat if cold ( Fe-rich) Converted to wurtzite

Oxides

Hematite 182 7 182 7 >980 50 Heats readily; arcing at high T No change

Ilmenite 987 2.5 1258 2.75

Magnetite 753 2.5 >700 460 50 Heats readily No change

Rutile 50 4 Room T Carbonates Ankerite Transparent Calcite 225 2.5 Transparent Dolomite Transparent Siderite Transparent Feldspars Albite 82 7 12 Orthoclase 67 7 10 Silicates Quartz 73 2.5 79 7 11

Phosphates Monazite > 150 Does not heat No change

[1] Microwave heating at 2450 MHz and 650 W. Samples of pure minerals (greater than 95 % by weight) with particle size within meshed -100 and +180 and sample weight of 30 g. Samples were placed in a quartz container, and argon was blown to achieve an inert atmosphere (Yixin and Chunpeng 1996).

[2] Microwave heating at 2450 MHz. Samples consisting of powdered natural minerals, weight of 30 g. Table indicates the maximum temperature recorded in the indicated time (Haque 1999).

[3] Effect of microwave heating on the temperature of various minerals at 2450 MHz, 400 W and 4 min exposure time. Samples consisted of powder -200 mesh, 50 g weight (Koleini and Barani 2012).

[4] Rate of microwave induced mineral heating (Djordjevic et al. 2010).

(22)

16 2.3.6. Mineral solubility and reflectivity

Different solubility reactions occur to sulphide minerals when mixed with acids, which could translate into different heating responses in IR thermography. According to Jones (1987), sulphide minerals are reactive to hydrochloric acid (HCl) and nitric acid (HNO3), and in powdered conditions the solubility

reactions occur rapidly. Arsenopyrite, chalcopyrite and pyrite are decomposed by HNO3 and the

formation of a yellowish precipitate might occur. Galena is decomposed in an HCl dilution and a white or pale-yellow precipitate might form, and pyrrhotite and sphalerite present slow solubility towards HCl (Table 8). Generally, sulphides and oxides mixed with HNO3 might release NO2 in the form of dark

brown fumes, and sulphides mixed with HCl might release H2S together with its characteristic smell.

Effervescent reactions of carbonates and solubility of some oxides towards HCl have to be accounted while testing on mixed-mineralogy samples.

Minerals present different reflectivity depending on the wavelength and therefore could present different responses in IR thermography. In the visible light bandwidth, reflectivity is semi-linear for common sulphides (Table 8). In the infrared bandwidth, reflectivity is not as influent as absorption bands (Clark 1999). Thermal cameras in the infrared do not work with spectral features but instead capture the scattered light radiated and converts it to temperature values. Therefore, the use of a passive heating mode (e.g. using illumination sources) can result into temperature variations due to difference in reflectance of the sulphide grains.

Table 8. Solubility reactions and reflectivity indexes of common sulphide minerals (Jones 1987; Webmineral

2014). Solubility legend: I = unchanged; E = soluble / effervescence; S = slowly soluble; D = decomposed.

Mineral Formula _HClSolubility _HNO Reflectivity* (%)

3 Red (700 nm) Green (520 nm) Violet (400 nm)

Arsenopyrite FeAsS I D 50.5 52.4 53.0 Chalcopyrite CuFeS2 I D 41.4 34.3 12.6 Galena PbS D - 42.2 43.0 52.8 Pyrite FeS2 I D 57.0 53.6 38.2 Pyrrhotite Fe1-xS S - 31.0 38.6 44.1 Sphalerite (Zn, Fe)S S - 15.9 16.8 18.4

* Standardized intensity of the reflection spectra of various minerals in air, in the visible light.

The literature review demonstrates that several techniques exist for automated mineralogy, but still no technique is available for sulphide mineral identification. Few studies have been published regarding sulphide detection in IR thermography. These have focused in utilizing microwave heating for ore sorting, ore gradation and mineral processing purposes. Therefore, this study aims to fulfill the lack by utilizing other heat and illuminating sources with focus to sulphide mineral identification in drill core samples. The utilized samples belong to four different ore deposits. The geological settings, main host rocks and mineralogy of these ore deposits are described in the following section.

2.4. Kittilä, Kristineberg and Garpenberg mines

The mineral and sulphide identification in this project is performed over the ores of the Kittilä, Kristineberg and Garpenberg mines. The orogenic gold deposits of the Kittilä mine are located in the Lapland region of Finland (Figure 5) and represent the largest gold deposits in northern Europe. The volcanic-hosted massive sulphide (VMS) deposits of the Kristineberg mine are situated in the western part of the Skellefteå mining district, northern Sweden. These correspond to the largest VMS deposits

(23)

17

Figure 5. Location of the Kittilä, Kristineberg and Garpenberg deposits within Scandinavia.

in the district and produce Cu, Zn, Pb, Au and Ag. Finally, the VMS deposits of Garpenberg mine, located in the Bergslagen mining district, central Sweden, produce Zn, Pb, Ag and smaller amounts of Cu and Ag.

Geology and mineralogy of the Kittilä mine

The Kittilä mine comprises orogenic gold deposits occurring within the Palaeoproterozoic Central Lapland Greenstone Belt (CLGB). The deposits are located in the Kiistala Shear Zone, KiSZ (Eilu et al. 2007, and references therein), which strikes N to NE and dips steeply to sub-vertical to the west (Patison et al. 2007). The KiSZ extends for about 30 km and is consistently anomalous in gold for at least 15 km (Eilu et al. 2007 and references therein). Three main ore deposits are included within Kittilä mine: Suurikuusikko, Rouravaara and Rimpi; two minor, Ketola and Etelä. The deposits occur within the Kittilä Group, subdivided into two major formations: the Kautoselkä Formation with Fe-tholeiite igneous rocks, and the Vesmajärvi Formation with Mg-tholeiite igneous rocks. The Porkonen Formation divides the latter-mentioned formations and consists of metamorphosed and graphitic shallow sediments, cherts, and BIF. All Kittilä Group rocks are metamorphosed to greenschist-facies (Patison et al. 2007) and are characterized by intense alteration, mostly albitisation, sericitisation and carbonatisation (Eilu et al. 2007; Patison et al. 2007).

The mineralization occurs within mafic to intermediate rocks characterized by abundant pumaceous material, mafic lava breccias, pyroclastic textures and reworked volcano-sedimentary material (Figure 6). Ultramafic and mafic units presenting almost no deformation are not mineralized (Patison et al. 2007). Also, intense shearing and mineralization is accompanied by graphitic alteration, and albite and carbonate alterations (Patison 2007). Thus, it is likely that the Na-rich epigenetic fluids that caused albitisation of the host rocks after the first major stage of deformation were the mineralizing fluids (GTK, 2008).

Gold occurs mostly as refractory gold associated with arsenopyrite (49 – 2700 ppm Au content in arsenopyrite), pyrite (1 – 585 ppm Au content in pyrite) and occasionally gersdorffite. The main part of the gold (73.2 %) is bound in the lattice of arsenopyrite or exists as tiny inclusions, 22.7 % in pyrite and 4.1 % as free gold, both native and electrum (Kojonen and Johanson 1999). The gold-bearing arsenopyrite and pyrite occur disseminated in microfractures, shear fabrics and stylolitic features

(24)

18 (Patison 2007). However, there is a large variability between the different deposits, and Suurikuusikko is richer in arsenopyrite, whereas Rouravaara is richer in pyrite (AE, 2015). The strong alteration of the ore has resulted in intense albitisation that increased the hardness of the ore. The carbonate alteration, including calcite, and dolomite or ankerite veins, is not mineralized with gold but infills the brecciated mineralized and albitized rock. Graphite is common, and amorphous carbon occurring within the shearing zones is abundant and may be originated from carbon-rich sediments within the host sequence (Patison et al. 2007). Other alteration phases include leucoxene, chlorite, sericite, rutile, tetrahedrite, chalcopyrite, gersdorffite, chalcocite, sphalerite, pyrrhotite, bornite, chromite, galena, and Fe-hydroxides in different quantities (Patison et al. 2007).

Figure 6. Simplified geological cross section along the W-E direction of the Suurikuusikko gold deposit (Patison

et al. 2007).

Geology and mineralogy of the Kristineberg mine

The Skellefteå mining district is an east-west-trending belt of the Early Proterozoic that hosts more than 85 VMS occurrences. The Kristineberg area is a deformed and metamorphosed volcanic domain within the Skellefteå district (Hannington et al. 2003). The Kristineberg are itself host six different VMS deposits: Kristineberg, Kimheden, Rävlidmyran, Rävliden and Hornträsk (Figure 7). The area comprises three main ore zones hosted by metamorphosed volcanic and subvolcanic rocks of the Paleoproterozoic Skellefteå Group, which consist mainly of juvenile felsic volcanic, porphyritic

(25)

19 intrusions and lavas, and minor intercalated sediments (black mudstone, siltstone, and volcanic-derived sandstone and conglomerate) (Åreback et al. 2005). Figure 7 shows the presence of the volcanic sequences belonging to the Skellefteå Group in two large anticlinal structures plunging gently to the west and separated by a syncline of the Vargfors Group within the Kristineberg area (Årebäck et al. 2005).

The Kristineberg deposit is flanked to the south by the plagioclase porphyritic trondhjemite of the Viterliden intrusion, and the massive feldspar porphyritic rhyolite A of the Revsund granite. This forms the hanging wall and limits to the north (Årebäck et al. 2005). The ore zone consists partly of two main massive sulphide horizons of massive pyrite and sphalerite with minor chalcopyrite, the A and B-ores, also the Einarsson zone, which is a complex Cu-Au rich sulphide stockwork and small massive sulphide lenses (Barrett et al. 2005). The ore zones strike E-W for more than 1000 m deep, dip toward the south and their geometry varies considerably due to folding and shearing (Hannington et al. 2003). Both A- and B-ores are pyrite-dominated with subordinate sphalerite, chalcopyrite and galena. However, the A-ores are richer in sphalerite and the matrix is chlorite-rich, while B-A-ores are richer in chalcopyrite and the matrix is quartz-rich (Årebäck et al. 2005). Cordierite-bearing schists of andesitic to dacitic composition are found between these two zones and present a weaker alteration than the other rocks of the deposit. Moreover, rhyolite A lies south of the B-ore and the Viterliden mine porphyry lies north of the A-ore (Barrett et al. 2005). The Einarsson zone is located close to the 1000 m level (close to the B-ore) and consists of Au-Cu-rich veins of pyrite-chalcopyrite-pyrrhotite, strongly disseminated sulphides and massive sulphide lenses of pyrite-sphalerite-chalcopyrite. This zone includes the Einarsson and Einarsson W lenses (Au-Cu), the J-lens (Cu-Zn-Au) and the K-lens (Zn-Ag-Pb) (Årebäck et al. 2005).

(26)

20 The alteration history of the Kristineberg area is complicated due to the synvolcanic hydrothermal and subsequent regional and contact metamorphism and is the cause of the difficulties in recognizing the primary host-rock types (Hannington et al., 2002). The minerals present previous to metamorphism would have been the ones characteristic of strongly hydrothermally altered rocks including quartz, sericite, chlorite and pyrite (Barret et al., 2005). However, metamorphism caused the recrystallization of sericite and chlorite and the general alteration assemblage in the altered volcanic rocks hosting the Kristineberg deposit therefore consists of quartz, muscovite, chlorite, cordierite, phlogopite, andalusite, pyrite and in some cases, talc (Åreback et al, 2005).

Geology and mineralogy of the Garpenberg mine

The Garpenberg volcanogenic massive sulphide depost is located within the Bergslagen region, a major Paleoproterozoic igneous province formed and subsequently modified by deformation and metamorphism during the Svecokarelian orogeny between 1.9 and 1.8 Ga (Jansson 2011). The geology of this region is interpreted as an intra-continental rift or a back-arc extensional basin developed on continental crust dominated by multiple generations of synvolcanic to early orogenic plutons which enclose inliers of metasedimentary and felsic metavolcanic rocks (Allen et al. 2008).

The Garpenberg deposit was extensively intruded by synorogenicgranites between 1.95 and 1.86 Ga (Rasmussen 2007). The magmatism observed in the region is calc-alkaline and most of the volcanic activity occurred as explosive submarine volcanism (Vivallo 1985). Therefore, the ore is described as a metamorphosed syn-volcanic, stratabound subsea-floor replacement ore formed mainly by carbonate replacement in the vent of a large marine rhyolite-dacite-volcano. The ore formation was pre-tectonic or early syn-tectonic with respect to the earliest Svecokarelian ductile deformation (Jansson 2011). The characteristic host rocks of the deposit are mainly proximal rhyolitic pumice-breccia, rhyolitic ash-siltstones, dacite intrusions, mafic volcanic and a stromatolitic limestone-marble horizon (Rasmussen 2007). The mineralized system is emplaced at the contact between metasedimentary and metavolcanic rocks associated with altered limestone or is located in the uppermost part of the stratigraphic footwall volcanic rocks, see Figure 8 (Vivallo 1985). Characteristics of this district are the association of the polymetallic sulphide deposits and many of the iron ore deposits with skarns in marble units. The magnetite-rich Fe-oxide deposits present a close spatial association with the Zn-Pb-Ag-(Cu-Au)-sulphide deposits.

The ore bodies mainly comprise: 1) irregular Mg-Mn-rich skarns within and at the base of a prominent marble unit; 2) pre-cleavage, vein and disseminated mineralization within silicified and/or Mg-rich zones in the "footwall" volcanic rocks; and 3) post-cleavage, tectonic remobilised, vein ores in faults adjacent to the main marble (Allen et al. 2008). The polymetallic sulphide ore bodies generally show strong K-Mg-Fe-Si alteration, K-Si alteration and silicification in proximal zones and the stratiform Zn-Pb-Cu mineralization is underlain by Cu-bearing stockwork ore with an extensive alteration zone of quartz-phlogopite rocks with minor chlorite, cordierite, garnet, staurolite, and andalusite (Vivallo 1985). The alteration of the stratigraphic footwall and the carbonate host occurred synchronously in the hydrothermal system that formed the polymetallic sulphide ores (Jansson 2011). The sulphide ore minerals are chalcopyrite, galena and sphalerite with small amounts of pyrrhotite and pyrite. Silver is found mainly in freibergite, in tetrahedrite or as native silver and gold appears in amalgam together with some silver (Bolin et al. 2003).

Towards Automated Logging of Ore: Positive Identification of Sulphides in the Ores of Agnico Eagle Kittilä and New Boliden Mines

MASTER'S THESIS

Towards Automated Logging of Ore

Positive Identification of Sulphides in the Ores of Agnico Eagle Kittilä and New

Boliden Mines

Anna Arqué Armengol

2015

Master of Science (120 credits)

Natural Resources Engineering

TOWARDS AUTOMATED

LOGGING OF ORE

Positive identification of sulphides in the ores of

Agnico Eagle Kittilä and New Boliden mines

Anna Arqué Armengol

August 2015

Supervisors:

Cecilia Lund

Pertti Lamberg

Abstract

Acknowledgements

Table of Contents

1. Introduction

1.1. State of the art

1.2. Objectives

2. Literature survey

2.1. Practices in drill core logging

2.2. Mineral identification in hyperspectral imaging

2.3. Mineral identification in infra-red thermography

2.4. Kittilä, Kristineberg and Garpenberg mines