Combined X-Ray Computed Tomography and X-Ray Fluorescence Drill Core Scanning for 3-D Rock and Ore Characterization: Implications for the Lovisa Stratiform Zn-Pb Deposit and Its Structural Setting, Bergslagen, Sweden

The globally increasing demand for metals requires the development of more time- and cost-efficient methods to collect and synthesize geologic data (e.g., Mudd et al., 2017; Nassar et al., 2020). In addition, current requirements in mining and exploration to exploit deeper-seated ore deposits and to reduce the environmental impacts of extraction demand integration of multiscale 3-D models based on geologic, geochemical, and geophysical data sets (e.g., Niiranen et al., 2019; Dentith et al., 2020). Key data for the generation of such 3-D models include ore and rock structures, textures, mineralogy, geochemistry, and petrophysical properties, most or all of which are primarily derived from drilling (e.g., Wood and Hedenquist, 2019). The subsequent sampling and analyses needed to extract such data from drill cores are typically time-consuming and sample destructive.

During the last decades, several fast and nondestructive techniques for extracting various data from rock samples and drill cores have emerged. In this context, geologic structures in drill core can be measured with the assistance of modern handheld devices, using, for example, laser alignment, although these typically exclude structures that occur within the core, such as lineations and fold hinges (e.g., Kramer-Bernhard et al., 2020). X-ray fluorescence (XRF) and hyperspectral infrared scanning are used to provide 2-D elemental and mineral maps of rock surfaces, respectively (e.g., Kruse et al., 2012; Wiedenbeck, 2013; Sjöqvist et al., 2015; Young et al., 2016; Mathieu et al., 2017; Ryan et al., 2018). X-ray computed tomography (XCT) is another relatively new scanning technique used in geologic applications for 3-D imaging and rock characterization (e.g., Williams et al., 2018). Up until recently, high-resolution XCT studies for mineral exploration have mainly been conducted on gold and platinum group metal ores to characterize their paragenesis and textures, thus contributing to genetic and economic evaluations and to more efficient ore processing (e.g., Godel et al., 2006; Kyle et al., 2008; Godel, 2013; Kyle and Ketcham, 2015). However, despite its potential for ore and rock characterization and the generation of structural data based on 3-D information, XCT is not frequently applied in mineral exploration and especially not for the assessment of long sections of drill core. This is mainly due to the lack of an instrument that can scan several meters of drill core at sufficient speed and resolution and that can also be transported easily to a mine or exploration site (Ketcham, 2005; Cnudde and Boone, 2013; Godel, 2013; Tonai et al., 2019).

Here, we present the results from an experimental study applying a high-resolution drill core scanner that simultaneously combines XCT, XRF, and density analyses to reveal 3-D rock textures, structures, and chemical compositions and their downhole variability. Scanning by the GeoCore X10 instrument was performed both on unoriented and oriented drill core from the actively mined Paleoproterozoic Lovisa stratiform Zn-Pb deposit in Bergslagen, south-central Sweden (Fig. 1). The Lovisa deposit was selected for this pilot study because of its variety of rock types and distinctive ore zones as well as for its complex structural history visible at a range of scales. The obtained results were compared and merged with structural, textural, and chemical data collected by conventional field mapping, drill core logging, and chemical analysis. The merged data sets and interpretations were then used to construct the first multiscale 3-D geologic model of the Lovisa Zn-Pb deposit, also including the nearby Håkansboda Cu-Co and Stråssa and Blanka Fe deposits, thereby providing an integrated 3-D geologic framework for future mineral exploration and new insights into the regional deformation history of the Bergslagen ore province.

The Lovisa Zn-Pb sulfide deposit is located in the Bergslagen ore province, south-central Sweden, one of the most diversely mineralized and extensively mined regions of the Fennoscandian Shield (Fig. 1). Mineralization at Lovisa was interpreted by Jansson et al. (2018) as an originally stratiform syngenetic-exhalative deposit that formed within a continental back-arc basin. The deposit was subsequently deformed and metamorphosed during the ca. 1.9–1.8 Ga Svecokarelian orogeny (e.g., Allen et al., 1996; Jansson et al., 2018; Stephens and Jansson, 2020). The deposit consists of two steeply SE-dipping ore horizons; the finely laminated “sphalerite ore” stratigraphically underlies the more massive, sphalerite-galena–bearing “main ore,” both with varying individual thicknesses from 0.1 to 3 m. The two ore horizons are separated by an ~1- to 3-m-wide zone of metamorphosed rhyolitic ash-siltstone, all occurring within a folded and faulted succession of interbedded felsic metavolcanic rocks, banded iron formation, and marble (Fig. 2). The metamorphic grade is characterized by low-pressure and high-temperature amphibolite facies conditions, and the mineralized succession has been affected by both ductile and brittle deformation, resulting in a complex combination of syn- and postgenetic ore textures that provide evidence of extensive sulfide deformation, remobilization, and recrystallization (Jansson et al., 2018; Sahlström et al., 2019a; Jonsson et al., 2020).

Structurally, the Lovisa deposit is located on the western limb of a NE-striking regional fold referred to as the Guldsmedshyttan syncline (Fig. 2; Lundström, 1983; Carlon and Bleeker, 1988; Luth et al., 2019). At Lovisa, this overturned syncline dips steeply to the southeast with a fold axis plunging moderately to steeply to the southwest. At Håkansboda, northeast of Lovisa, the Guldsmedshyttan syncline has been refolded by steeply SE-plunging folds with subvertical NW-to NNW-trending axial planes. The Guldsmedshyttan syncline is part of the larger West Bergslagen boundary zone, an ~150-km-long and 2- to 6-km-wide high-strain deformation zone comprising several NE-striking, steeply dipping shear zones (Fig. 1; Beunk and Kuipers, 2012). Besides the Lovisa and Håkansboda deposits, the West Bergslagen boundary zone hosts various other base metal deposits, of which the Garpenberg Zn-Pb-Ag-(Cu-Au) deposit is the largest by tonnage, and numerous Fe oxide skarn deposits including rare earth element (REE)-rich Bastnäs-type mineralization, such as at Riddarhyttan and Norberg (Fig. 1; e.g., Geijer, 1961; Allen et al., 1996; Sahlström et al., 2019b).

Active mining within the study area occurs only at Lovisa, which comprises a relatively small scale underground operation that yields about 42,000 tonnes/year from a proven and probable reserve of 636,000 tonnes (t) of ore with an average grade of 9.5 wt % Zn and 3.9 wt % Pb (Lovisagruvan AB, 2020). The reserve has been defined by 90 drill holes for 1,100 m along strike and down to 425 m below ground level and is open to the south and at depth.

Structural data from outcrops were collected in an area within 4 km from the Lovisa Zn-Pb deposit, including also the Håkansboda Cu-Co and Stråssa and Blanka Fe oxide deposits (Fig. 2). Field mapping focused on outcrops located along fold hinges, faults, or geophysical lineaments, as displayed on available bedrock and mine maps at various scales ranging between 1:50,000 and 1:1,000 (Lundström, 1983; Bleeker, 1984; Carlon and Bleeker, 1988; Jansson et al., 2018) or on magnetic and gravity anomaly maps (Luth et al., 2019). Oriented thin sections were prepared from field-collected rock samples and characterized for microstructures using a polarized-light microscope.

Selected drill core for XCT-XRF scanning comprised 700 m of unoriented full or half core from 10 drill holes (LOV1501–1506 and LOV1802–1805), which collectively transect the Lovisa Zn-Pb ore zones. An additional 800 m of oriented drill core was selected from three exploration drill holes (LOV1901–1903) located between Lovisa and Håkansboda intersecting marble and metavolcanic rock down to 300 m below the surface.

Whole-rock samples for lithogeochemistry were taken by Lovisagruvan AB from the unoriented drill cores (LOV1501–06) before scanning. Sample lengths varied between 10 and 150 cm, with one sample measuring 3.9 m (see App. 1). The samples were prepared by MS Analytical, Vancouver, Canada, and element concentrations were quantified by inductively coupled plasma-mass spectrometry (ICP-MS) and inductively coupled plasma-atomic emission spectroscopy (ICP-AES) following aqua regia (e.g., Co) or four-acid digestion (e.g., Zn, Pb, Fe₂O₃). For Zn concentrations exceeding 10 wt %, a wet titration technique was used (STI-8Zn). The analytical precision and accuracy for each applied method is summarized in Appendix 1.

Combined XCT-XRF drill core scanning was performed using the transportable GeoCore X10 instrument developed by Orexplore, Stockholm (Fig. 3A; Bergqvist et al., 2019; Orexplore, 2021).

The instrument contains a tungsten solid-anode X-ray source with a 50-μm focal spot size, a transmission detector with a pixel size of 100 μm, and a spectrometer, all moving up and down during scanning while the drill core sections are rotating (spiral scanning). Batches of 4-m-long drill core sections with diameters from 36.65 (BQ) up to 50.5 mm (NQ2) are loaded vertically into the instrument after inserting the individual core pieces into 1-m-long plastic tubes while preserving their relative positions and orientations with respect to the marked core orientation line when present (Fig. 3A). Scanning time may vary between 1 and 6 m/h where low speed results in a higher signal-to-noise ratio. In this study, whole and half drill cores with an initial diameter of 39 mm (WL67) were scanned at rates between 1 and 3 m/h, with an applied X-ray energy of 120 kVp, producing results in 3-D images with a resolution of 200 μm per voxel side.

A precalculated data set of minerals common for Lovisa was used to correlate measured X-ray attenuations to known mineral attenuation values using cross-section data from PENELOPE-2006 (Fig. 3B; App. 2; Salvat et al., 2006). The X-ray imaging system was calibrated using stair-shaped aluminum phantoms, geometric objects used in tomographic imaging to calibrate pixel-to-pixel variations and to correct for beam-hardening effects. The GeoCore X10 utilizes a proprietary technique that combines XRF spectra, attenuation, and density values derived from tomography imaging and sample weight to calculate for mineral compositions and final elemental abundances in weight percent or parts per million (Fig. 2B; Bergqvist et al., 2019). This best-fit approach mitigates analytical challenges associated with heterogeneities in rock samples. Calculations of XRF responses of minerals and tomographic reconstructions were performed on a high-performance computing-class computer equipped with graphics coprocessors (GPUs) capable of teraflop performance.

Detection limits and precision: As with conventional XRF instruments, the detection limits of the GeoCore X10 depend on the applied X-ray energy, scanning rate, bulk composition, density, and particle size distribution of the sample (e.g., Laperche and Lemière, 2021). Detection limits for the GeoCore X10 are approximately 100 ppm for elements with atomic numbers between 25 and 92. Elements lighter than Cl are not measured but estimated based on stochiometric calculations, using theoretical half-value layers for XRF spectral lines, which correspond to the material thickness after which the intensity of the radiation has dropped to 50% (App. 3; Salvat et al., 2006). Estimations of the precision of the analyses are derived from repeated measurements on eight drill core samples from various rock and ore types from three different deposits including Lovisa Zn-Pb (see App. 4, Fig. A2). Each core sample measured 1 m in length and was scanned 12 times at a speed of 3 m/h. The obtained standard deviations are relative to the mean value of all 12 measurements and depend strongly on concentration and atomic number of the respective element.

The XCT data from the GeoCore X10 are systematically negatively affected by physical effects such as beam hardening, photon starvation, and scattered radiation (e.g., Joseph and Spital, 1982). Due to these effects, X-rays that pass through massive metallic ore particularly are highly attenuated, obscuring internal rock fabrics by artifacts that cause dark and bright streaks on corresponding images (e.g., Boas and Fleischmann, 2011, and references therein). To partly overcome this, the XCT camera in the GeoCore X10 is a photon-counting cathodoluminescence (CT) type, which offers a higher contrast-to-noise ratio, improved spatial resolution, and optimized spectral imaging (e.g., Willemink et al., 2018).

3-D visualization and data analysis: The obtained Lovisa scanning data sets were imported into the Orexplore Insight software (Orexplore, 2021), which visualizes the scanned drill core as 3-D, rotatable imagery next to a list of estimated and calculated elemental concentrations. The elemental concentrations were averaged for manually selected segments down to 8 mm in thickness. For oriented drill core (LOV1901–03), the orientation line marked on the core was transposed on the rotatable 3-D imagery, and planar and linear features were observed and measured manually on screen while changing the core viewing angle, rendering, and transparency settings. Survey data (azimuth and dip angle of drilling) were imported for each core to allow for the automatic conversion of structural measurements into true azimuth/trend and dip/plunge values. A brief assessment of the repeatability of the structural measurements on the 3-D tomography is presented in Appendix 4. The structural data and elemental concentrations were exported as csv files and used for further analysis and 3-D geologic modeling.

3-D geologic modeling was conducted in three steps, based on scales of the study and used data types. At first, a semiregional 3-D geologic model, measuring 7.5 × 7.5 × 2 km, was constructed using the implicit modeling software Leapfrog Geo v. 6.0, utilizing the geologic maps and profiles presented in Jansson et al. (2018) as well as the new lithological and structural data from this study. Secondly, the Lovisa Zn-Pb ore zones were modeled as a single orebody (ore shell) constrained by underground mapping and 15 drill logs from exploration drilling conducted during the last decade and an additional 75 historical drill logs dating back to the 1980s (e.g., Carlon, 1987; Lovisagruvan AB, 2020). Additional orebody models were constructed for the nearby Håkansboda, Stråssa, and Blanka deposits (Fig. 2) using historical underground mine maps and profiles (Tegengren, 1924; Geijer, 1927; Geijer and Magnusson, 1944; Koark, 1960; Carlon, 1986). The resulting 3-D orebody models were integrated with the semiregional model and their geologic context interpreted. Thirdly, XCT-XRF scanning data was utilized to refine local and regional modeled lithological volumes by adding textural, chemical, and structural constraints from the Lovisa drill cores and from the oriented cores drilled north of the Lovisa mine.

The main structures observed in outcrops around Lovisa are S₀ and S₁ planar fabrics that are generally parallel (S_0,1), asymmetric folds, stretching and intersection lineations, C-S shear fabrics, boudinage, and brittle faults and fractures (Figs. 4, 5). Planar and linear grain-shape fabrics are well developed in mica-rich metavolcanic rocks forming SL tectonites (Fig. 5), whereas marble- and quartz-rich metamorphosed volcanic and intrusive rocks were deformed into L tectonites with lineations plunging steeply to the southeast (Fig. 4A). The bulk of the observed mesoscopic folds have subvertical N- to NE-striking axial planes and are interpreted as F₂ shear folds that are parasitic to the Guldsmedshyttan syncline with predominantly Z asymmetry on its northern limb and S asymmetry on its southern limb (Fig. 4B). Many F₂ folds show isolated fold hinges and strongly vertically and horizontally stretched fold limbs without the development of an axial planar cleavage (Fig. 4B). In contrast, vertical NW-striking folds (F3) locally refold the F₂ folds in outcrop (Fig. 4C) and are portrayed on the map as the hook-like structure in the Håkansboda area (Fig. 2; Carlon and Bleeker, 1988). These younger F₃ folds are associated with a spaced vertical axial planar cleavage striking northwest along which both minor dextral and sinistral displacements were observed (Figs. 4C, F, 5D-F). Southwest of the Lovisa deposit, subvertical F₂ shear folds with variable amplitudes and wavelengths are developed within a well-layered sequence that is located stratigraphically in the hanging wall of the Lovisa deposit (Fig. 4D-F). Here, local refolding of asymmetric N-striking vertical folds (F_2a) resulted in folds with similar asymmetry and orientation (F_2b), suggesting that deformation was progressive (Fig. 4D, E). In addition, steeply S- to SE-plunging F₂ sheath and oblique folds with hinges oriented parallel to a penetrative L₂ stretching lineation further reinforce the interpretation that mesoscopic F₂ folding resulted from noncoaxial subvertical shearing (Fig. 4D). Ductile kinematic indicators, such as sigmoids, rotated clasts, and intrafolial folds, observed parallel to the penetrative stretching lineations all imply a reverse, southeast-side-up sense of shear with a sinistral horizontal component (Fig. 5B, C).

From the 3-D tomographic images (XCT) of the oriented drill cores, a large number of fabrics and kinematic indicators were identified and measured, including penetrative linear and planar fabrics, deformed clasts and mineral blasts, folds, and C-S fabrics, as well as fractures and veins (Fig. 6; App. 5). Structures were observed in all drilled lithologies, but the most useful examples are from metavolcanic rhyolitic silt/sandstone and calc-silicate (skarn) rocks in which compositional banding caused clear contrasts in spectral attenuation.

Within the rhyolitic silt/sandstones, the compositional banding is portrayed by slightly different attenuation values, where the lower values correspond to higher quartz contents (Fig. 6A). Whether this banding represents primary bedding/laminae (S₀) or is caused by subsequent deformation (S₁) is not directly evident from the tomography. A linear fabric is defined by elongated and aligned feldspar and mica grains that exhibit higher attenuation than quartz and represent an L₂ stretching lineation (Fig. 6A) or a mineral lineation (Fig. 6B). The lineations are oriented parallel to the compositional banding (Fig. 6A) or locally at moderate angles to it (Fig. 6B). When viewed normal to the lineation, the sigmoidal shape of some distinctive high-attenuation grains/clasts indicates reverse shearing. A reverse sense of shear with top-to-the-northwest movement is also inferred from C-S fabrics with moderately dipping C surfaces containing elongated high-attenuation grains/clasts and steeply dipping S surfaces in which the high-attenuation grains/clasts appear less elongated (Fig. 6C).

Various folds become visible while rotating the tomographic core images in 3-D (Fig. 6D-G). Most folds are Z and S asymmetric and are only visible when viewed normal to the L₂ stretching lineation and parallel to the axial plane (Fig. 6D). These folds appear to fold earlier-formed planar and linear fabrics and are interpreted as F₂ oblique or shear folds that formed by reverse ductile shearing with top-to-the-northwest movement, which is consistent with kinematic indicators observed in outcrops (see previous section). F₃ folds fold the planar (S_0,1) and linear fabrics, which was observed on the XCT image when viewed both normal and parallel to the L₂ stretching lineation (Fig. 6E-G). The F₃ folds are typically associated with fractures and veinlets oriented parallel to the fold axial planes. The bulk of the observed fractures and veins, however, are not associated with folding and clearly overprint the ductile fabric (Fig. 6H, I). Fractures with a fracture fill other than quartz show mostly high attenuation and are easily recognized in the tomographic images (Fig. 6H), whereas fractures filled with quartz or having no fill are determined indirectly from a disrupted preexisting fabric. In a few cases, a displacement vector could be measured by restoring dislocated and distinctive lithoclasts along a fracture plane (Fig. 6I).

Three examples of representative ore textures and associated XCT-XRF scanning data, obtained from a 20-m-long drilled section through the Lovisa Zn-Pb orebody, are shown in Figure 7. At the top of the section, the main ore is characterized by a distinctive ball ore texture (e.g., Geijer, 1971; Vokes, 1973), comprising variably sized silicate-rich fragments that occur within a deformed and recrystallized matrix dominated by fine-grained sphalerite and galena (Fig. 7B). The texture reflects folding (F₂ and F₃) of the metavolcanic host rock, followed by progressive breakup, rotation, and kneading of mostly competent host-rock fragments within a ductile sulfide matrix (e.g., Vokes, 1969, 1973; Fig. 7B). Corresponding concentrations and 3-D distribution of Pb and Zn suggest locally variable degrees of remobilization of sphalerite and galena in the ball ore, with galena preferentially occurring in fold hinges and along contacts with the adjacent metavolcanic rock (Fig. 7C).

Directly below the main ore, networks of veins and fracture fillings consisting of remobilized sulfides cross-cut a laminated metavolcanic rock (Fig. 7E). Two sets of sulfide-bearing veins and fractures are present, oriented subparallel to primary layering and oriented at high angles to primary layering, respectively. The combined XRF and XCT scan data reveal a transition from Pb > Zn to Zn > Pb from top to bottom of the sample, indicating variable proportions of galena and sphalerite (Fig. 7G). A few meters downhole, the sphalerite ore preserves what has been interpreted as folded (F₂ and F₃) syndepositional layering (S_0,1) of finegrained sulfides (sphalerite with minor intergrown galena and pyrite) alternating with (meta-)rhyolitic interbeds (Fig. 7H; Jansson et al., 2018; Sahlström et al., 2019a). Locally, sulfide-rich veins or fracture fillings protrude from the fold hinges. Unlike in hand specimen (Fig. 7H), the thin sulfide layers can be readily distinguished from rhyolitic interbeds in the XCT scan, thereby enhancing visualization of delicate pinching-out and folding structures (Fig. 7I). Overall, galena was extensively remobilized from the preexisting layering into veins parallel to fold hinges during ductile deformation, while sphalerite largely remained within the layering. This differential sulfide remobilization has caused significant centimeter-scale variability in Zn (0.5–9%) and Pb grades (0.1–3%; Fig. 8J).

The obtained concentrations of Zn, Pb, Fe, and Co from XCT-XRF scanning have been compared with the results from 18 lithogeochemistry samples from a 20-m, ore-bearing drill core section (Fig. 8A, B; App. 1). The concentrations derived from XCT-XRF scanning were averaged over the same length intervals as the samples used for lithogeochemistry analysis. Additionally, XCT-XRF data for 8-mm-thick intervals, the minimum selectable thickness of core segments in Insight, were included in the comparison (Fig. 8A). Note that scanning was conducted after sampling for lithogeochemistry analysis, meaning that the comparison is based on two analyses of different material derived from opposing pieces of half core.

The resulting downhole logs (Fig. 8A) reveal broadly similar compositional trends with elevated concentrations of Zn, Pb, Fe, and Co in the main ore and sphalerite ore. A correlation plot for XCT-XRF scanning and lithogeochemistry data from 29 samples (Fig. 8B) shows significant variation for Zn (avg relative difference ~50%), whereas Pb and Fe reveal less variation (avg relative difference of ~37 and 21%, respectively). Due to analysis of opposing core pieces, this variation may be both geologic and analytical. No correlation exists between variation and sample core length. For Co, despite an overall similar relative compositional trend (Fig. 8A), concentrations from XCT-XRF scanning are one to two orders of magnitude higher than those obtained from lithogeochemistry analysis on whole-rock samples (Fig. 8B). This systematic overestimation by XCT-XRF for Co is likely due to overlap between emission lines Co-K_α and Fe-K_β (e.g., Potts and Webb, 1992), which is also supported by the observation that both elements follow the same general trend within the succession (Fig. 8A, C). Such a spectral interference between Co and Fe has also been noted in portable XRF analysis (e.g., Gallhofer and Lottermosser, 2018). Highly localized variations for all displayed elements are suggested by the XCT-XRF 8-mm interval plots (Fig. 8A). For example, locally elevated concentrations of Pb (at 192 and 194 m) and Zn (at 178 and 192 m) are missed by the sampled interval plots and can be explained by very localized concentrations of galena or sphalerite, respectively.

Structural measurements from outcrop were plotted in stereonets to determine regional trends and for a comparison with orientations interpreted from 3-D tomographic imagery (XCT) of oriented drill cores (Fig. 9). From outcrop measurements, poles to the S_0,1 planes plot along a great circle, reflecting regional-scale F₂ folding around a steeply plunging, SE-trending fold axis (Fig. 9A). The fold hinges of mesoscopic F₂ and F₃ folds in outcrop plunge between 45° and 85° to the south and southeast (Fig. 9B). Likewise, the bulk of the L₂ stretching lineations plunge between 40° and 75° to the south-southeast. Some folds and lineations, however, trend toward the east-northeast, which combined in stereonet give rise to a regional doubly plunging fold with an axial planar surface dipping ~50° to the east-southeast. Furthermore, brittle structures including fractures, spaced cleavages, and faults (Fig. 9D) are predominantly dipping between 75° and 90° toward the northeast and southwest. The bulk of the slickenlines on fractures, cleavages, and fault surfaces plunge between 5° and 30° toward the southeast and are consistent with sinistral movements.

The 3-D models of the Håkansboda Cu-Co and Blanka Fe deposits reveal rod-shaped orebodies that plunge between 50° and 60° to the southeast and east, respectively (Figs. 9C, 10). Similarly, the dominant folds of the modeled Lovisa Zn-Pb and Stråssa Fe deposits plunge steeply (~55°) to the south-southeast. Hence, the modeled ore shoots and folds have orientations comparable to those of the F₂ fold hinges and stretching lineations measured in outcrops (Fig. 9B, C).

From the XCT imagery, the orientations of the compositional banding in the (meta-)rhyolitic silt/sandstone are comparable to the orientation of the S_0,1 planes measured in outcrop, and their clustering reflects folding of S_0,1 around a steeply SE-plunging F₂ fold axis (Fig. 9E). The orientations of F₂ fold hinges and L₂ stretching lineations in the scan are clustered around 60° to the east-southeast (Fig. 9F, G), which is comparable with measurements in outcrop, albeit the scanning data shows a wider spread toward more gently plunging and NE-trending fold hinges and stretching lineations. Fracture orientations derived from XCT dip between 10° and 60° and therefore differ from the overall steeply dipping planes measured in outcrop (Fig. 9G, H). This contrast may be due to a sampling bias underlain by the underrepresentation of vertical sections in most outcrops as well as by the direction of drilling (perpendicular to S_0,1). Apart from varying dips ranging between 0° and 90° a general northwest-southeast strike of fractures is apparent from both data sets.

The structural and geochemical data derived from XCT-XRF drill core scanning were used to constrain and refine the 3-D geologic model at Lovisa (Fig. 11). The scanned oriented drill cores do not intersect the Zn-Pb ore but do provide valuable geologic constraints along a 2-km NW-SE–oriented section between the Lovisa (Zn-Pb) and Håkansboda (Cu-Co) deposits (Fig. 11A). Firstly, distinctive Fe-rich skarn layers occurring within the metarhyolitic silt/sandstones identified by XCT-XRF and conventional logging were used as stratigraphic marker horizons to correlate the drill cores. The resulting pattern outlines an overturned, SE-dipping syncline (Fig. 11B), the Guldsmedshyttan syncline, and is consistent with earlier interpretations based on lithology and way-up indicators from outcrop observations by Carlon and Bleeker (1988). Secondly, the structural measurements from outcrop and drill core 3-D tomography were manually interpolated, revealing a structural pattern dominated by moderately SE-dipping S folds and steeply dipping faults (Fig. 11C). At deeper levels, beyond the extent of drilling, the overall fold geometry of the Guldsmedshyttan syncline was interpreted from 2-D forward modeling of magnetic and gravity data presented in Luth et al. (2019).

At the scale of the Lovisa deposit, an initial Zn-Pb ore shell was modeled in 3-D using lithological and geochemical data from historical drilling and subsurface mapping (Fig. 12A). From the scanned unoriented drill cores intersecting the Zn-Pb ore zones, the orientations of planar and linear features were digitally measured with respect to the core axis and provided useful structural constraints when referenced to the regional trend of the dominant S_0,1 fabric (e.g., Scott and Berry, 2004). As a result, the boundaries of the main ore, sphalerite ore, and barren interlayers were identified in multiple drill cores, which through drill hole correlation resulted in a refined 3-D geologic model (Fig. 12B). While these major ore zones were already possible to log using conventional methods, the enhanced structural and mineralogical-chemical detail provided by XCT-XRF was particularly useful to distinguish and delineate the finely laminated sphalerite ore, which was previously difficult (cf. Fig. 7H-J).

The 3-D model of Lovisa displays a polyphase folding pattern in which NE-striking F₂ folds have been refolded by NW-striking vertical F₃ folds (Fig. 13). This folding pattern is consistent with earlier observations from Carlon and Bleeker (1988), although they interpreted this pattern as due to F₁-F₂ fold interferences. The difference between the two interpretations arises primarily from the S₀-S₁ relationship, which in this study was unequivocally determined as consistently parallel planar structures that have been jointly folded by F₂ folds at all scales and locally by F₃ folds.

Deformation history: We propose a fourfold structural history for the Lovisa area comprising three regional deformation phases, commencing with the development of a D₁ planar and linear grain-shape fabric related to tight to isoclinal folding of bedding (F₁; e.g., Stålhös, 1981). At Lovisa, determination of D₁-related fabrics is difficult due to transposition of primary S₀ structures into S₁ followed by recrystallization and finally the transposition of these structures into D₂-D₃ fabrics.

A second phase of folding and shearing occurred during D₂ under roughly north-south crustal shortening. D₂ deformation along the West Bergslagen boundary zone is characterized by sinistral transpression producing two distinctive folding patterns, which we interpret as the result of spatial and temporal variability in the degree of strain partitioning. Thus, during early D₂, N-S–directed crustal shortening was efficiently partitioned along the West Bergslagen boundary zone into contractional domains undergoing isoclinal folding, and a transcurrent component was localized along narrow mylonite zones (D_2a in Fig. 13). This stage resulted in the formation of the Guldsmedshyttan syncline with its included pretectonic mineralization, which was bounded along its limbs by sinistral mylonite zones to the northwest and southeast. During late D₂, enhanced mechanical coupling along the West Bergslagen boundary zone and less efficient strain partitioning led to an increased transcurrent component in the contractional domain causing refolding of the isoclinal upright folds along steeply SE-plunging S folds and the development of an L₂ stretching lineation (apparent in most of the XCT-XRF scans; Fig. 6), while the bounding mylonite zones accommodated sinistral-reverse top-to-the-northwest shearing (D_2b, Fig. 13). Locally, even the F₂ folds associated with this late D₂ event were refolded as a result of progressive shearing. Finally, crustal shortening became EW-directed during D₃, causing dextral transpression along the West Bergslagen boundary zone, which mainly resulted in NW-SE– to N-S–striking vertical F₃ folds with predominantly Z asymmetries, such as at Håkansboda, and dextral shearing along discrete NNW- to NNE-striking mylonite zones.

A consequence of the proposed deformation history outlined above is that there is no need to infer a D₁ deformation phase to have preconditioned vertical F₂ folding by tilting the stratigraphy into a vertical orientation, as proposed by Stålhös (1981) and adapted by Beunk and Kuipers (2012). Instead, all D₂-related deformation could have been achieved by regional north-south crustal shortening during which the degree of strain partitioning varied with time along the West Bergslagen boundary zone.

Controls on strain partitioning: The degree of strain partitioning depends in part on the rheological weakening along a shear zone (e.g., Mount and Suppe, 1987), which for Lovisa and the West Bergslagen boundary zone in general must have been strongly influenced by the well-stratified supracrustal rocks including sulfide-rich zones and interbedded carbonate, skarn, and chert layers, which strongly contrast with the homogeneous and highly competent intrusive rocks present on either side of the West Bergslagen boundary zone (Fig. 2). Another controlling factor on the degree of strain partitioning is the orientation of the convergence vector relative to the shear zone (e.g., Fossen et al., 1994; Tikoff and Teyssier, 1994), which for the West Bergslagen boundary zone was north-south during D₂ but turned to east-west during D₃ (e.g., Stephens and Jansson, 2020). As a result, D₃ deformation at Lovisa was most likely accommodated by right-lateral displacements and drag folding along the NW-SE–striking shear zones as well as by vertical folding along steeply dipping NNW-SSE–striking axial planes at Håkansboda and Stråssa (Fig. 13). It is plausible that during late D₂, strain partitioning finally became less efficient as the region evolved from moderately high temperatures to a brittle-ductile regime, leading to enhanced mechanical coupling along the West Bergslagen boundary zone.

Timing of deformation: It is suggested that the proposed D₁-D₃ progression roughly reflects three tectonic cycles at 1.9–1.76 Ga, which were governed by subduction along a convergent continental margin active in Fennoscandia at that time (e.g., Stephens et al., 2009; Stephens and Jansson, 2020). D₁-related deformation was preceded by sedimentation, synvolcanic mineralization and magmatism during an early phase of the first tectonic cycle marked by extension caused by a retreating subducting slab at ca. 1.89–1.87 Ga. A switch toward an advancing arc initiated D₁ shortening and M₁ metamorphism during the end of the first and the onset of a second tectonic cycle. Subsequent D₂ shortening and M₂ metamorphism at 1.84–1.82 Ga was probably preceded by another phase of regional extension due to renewed slab retreat (D₃ in Beunk and Kuipers, 2012) causing further magmatism and sedimentation. Beunk and Kuipers (2012) divided D₂ into two stages in which peak M₂ conditions relate to the closure of a back-arc basin, followed by accretion and collision resulting in the formation of the Bergslagen orocline, crustal-scale F₂ folding, and localization of shearing in high-strain zones such as the West Bergslagen boundary zone. Whether the continued magmatism that postdates D₂ marks another phase of crustal extension (e.g., Wahlgren and Stephens, 2020) or resulted from crustal thickening by oroclinal bending (Beunk and Kuipers, 2012) is uncertain, but the absence of penetrative tectonic fabrics in most of the 1.8–1.76 Ga intrusions, in combination with D₃ kink-like structures and spaced cleavages in the metasupracrustal rocks, suggests that D₃ deformation was local and occurred at relatively low temperatures. The timing of D₃, however, remains poorly constrained.

Relationships between mineralization and deformation: At Lovisa, the stratiform Zn-Pb sulfide deposit predates D₁-D₃ deformation and has clearly been affected by folding and shearing. As such, we suggest that the ball ore texture of the main ore developed during D₁-M₁ and was most likely accentuated during D₂-M₂ based on the presence of a strong penetrative fabric enclosing internally folded clasts. In line with the interpretation of Jansson et al. (2018), D₂-D₃ deformation also produced many of the observed parasitic and smaller fold structures affecting the current orebody geometry at Lovisa. Likewise, the overall shape and orientation of the steeply plunging ore shoots at Håkansboda reflect D₂-D₃ folding and stretching along the SE-dipping shear planes. Within the Håkansboda ore zones, primary depositional and D₁-related features have been transposed into parallelism, producing pinch-and-swell sulfide bands, breccias, and boudins of banded calc-silicate rocks, all intersected by sulfide vein networks, as noted by Carlon and Bleeker (1988). On a regional scale, steeply plunging ore shoots that developed from originally stratabound massive sulfides are a common feature in Bergslagen—for example, at the Falun Zn-Pb-Cu-(Au-Ag) sulfide deposit (e.g., Kampmann et al., 2016) or at the Garpenberg Zn-Pb-Ag-(Cu-Au) sulfide deposit (Allen et al., 2003; Jansson and Allen, 2011). In general, the high abundance of sulfide mineralization within highstrain zones suggests a strong control by D₂/D₃ deformation on ore remobilization and vice versa; the presence of sulfide mineralization has likely played a key role in localizing D₂/D₃ folding and shearing.

For this study, a resolution of 200 μm obtained from scanning 3 m/h appeared sufficient for the identification, characterization, and textural-structural interpretations of ore and host-rock assemblages. However, a challenge for the presented XCT-XRF drill core scanning technique may include the improvement of the tomographic resolution to be able to resolve small grains or mineral inclusions. In theory, a tomographic resolution down to 45 μm/pixel can already be reached with the GeoCore X10, but this will increase the scanning time significantly. Another challenge is to overcome the problem of overlapping XRF spectral lines, which causes the systematic overestimation of the concentrations of some elements (e.g., Co). A possible solution may be to use other spectral lines, such as Co-K_β, which is only 13.5% of the intrinsic intensity of Co-K_α and demands for separate simulations, calibrations, and measurement routines. Furthermore, a key future development of the XCT-XRF technology and associated processing software should be to allow direct visualization of absolute chemical concentrations and mineralogical composition in 3-D, rather than indirectly by attenuation.

In addition to the above challenges, XCT-XRF scanning adds another step to the core characterization procedure of an exploration or mining operation (see also Cioacă et al., 2020). However, scanning (1,500 m at 3 m/h) and subsequent digital logging (~25 m/h) of a moderate amount of core at Lovisa has provided a significantly enhanced level of structural and mineralogical-geochemical detail not previously obtainable via conventional methods, thereby allowing for refinement of 3-D geologic models at multiple scales. Some core logging at Lovisa, particularly structural measurements, kinematic indicators, and delineation of the sphalerite ore, was found to be not only more detailed but also considerably faster using the digital XCT-XRF core compared to conventional logging. With continued development, the capabilities and the efficiency of the XCT-XRF scanning technique may increase enough that it could fully substitute some traditional methods. We conclude, however, that the scanning technique at present is best utilized as a complement to conventional methods, where a priori knowledge and validation from the physical core is still needed together with a rigid quality control, enabling detailed, high-quality structural and compositional core logging in 3-D, preferably on well-preserved and oriented drill core (e.g., Kramer Bernhard et al., 2020; Orpen and Orpen, 2020).

It has been demonstrated that combined XCT-XRF drill core scanning is a powerful tool for 3-D rock and ore characterization, generating continuous downhole data sets on structure, texture, mineralogy, chemistry, and density from a single scan. The resulting rotatable 3-D imagery can be shared directly after scanning to allow for joint structural-mineralogical-chemical core logging by (remote) geologists and decision makers, thereby accelerating high-quality 3-D interpretations and multiscale geologic modeling. In its application to the Lovisa stratiform Zn-Pb deposit, XCT-XRF drill core scanning revealed 3-D ore and host-rock textures indicating remobilization of ore minerals and systematic Z and S folding and faulting. With the XCT-XRF–derived information, and in combination with additional geologic and geophysical data, 3-D geologic modeling at Lovisa was refined on multiple scales by adding new structural and chemical constraints to the Zn-Pb ore zones, host-rock units, and regional structural framework. The results indicate a fourfold deformation history at 1.9–1.76 Ga with a strong control exerted by transpressive D₂/D₃ deformation on ore remobilization and orebody geometries at Lovisa and surrounding deposits hosted by the West Bergslagen boundary zone.

This study is part of the X-MINE project funded by the European Union’s Horizon 2020 research and innovation program under grant agreement 730270. We gratefully acknowledge Ronald Arvidsson, Johan Jönberger, Edine Bakker, Elvis Stansvik, Peter Hedin, Karin Högdahl, and Nils Jansson for their input on various aspect of this study. We thank Michael Gazley, Tobias Bauer, Iain Dalrymple, and the Economic Geology Editorial Board for their extensive and constructive reviews of earlier versions of this manuscript.

Stefan Luth is a structural geologist at the Geological Survey of Sweden, where he combines field work with 3-D geologic and geophysical modeling to construct multiscale 4-D geologic frameworks for ore deposits throughout Sweden and other European countries. In 2011 he obtained his Ph.D. degree at Vrije Universiteit Amsterdam on lithosphere-scale analogue modeling of continental subduction below the Alps. More recently, his research interest focuses on the demonstration and implementation of new geologic and geophysical tools for mapping, 3-D modeling, and mineral exploration.