2.3.9. Mining Recipe

In this part of the software, particular minerals of interest and elements can be selected to perform advanced textural and chemical quantification. In the "Assay" part of the user interface, the user has the ability to select the elements of interest in this sample (Figure S7). The software will then perform an assay measurement, whereby the chemically quantified pixel data and the specific gravity (data that is added to each mineral classification) are used to calculate a mass. This is done for all pixel/elements selected across the sample and is used to calculate an assay. This can therefore be used as a rough "bulk rock assay" where the analysis is based on a single 2D plane throughout the entire sample.

Additional value can be gathered from this aspect of the software whereby "elemental distribution" data can be gathered. This is where the chemical distribution is quantified. For example, when the chemical distribution of Mn is of interest: the data will give the wt% amount of Mn in all specified minerals, and also a distribution percentage based the total percentage of Mn found in each phase. This data is provided from the directly chemical quantification during the automated analysis and not from idealized or pre-defined concentrations. This provides researchers with a more reliable AQM technique to understand and quantify chemical and mineralogical variations. However, in cases where element concentrations cannot be measured reliably (e.g., for Boron in tourmaline), a concentration can be assigned in the mineral list for that specific element.

The minerals to be analysed for the map can be selected, as Target minerals. Byproducts and Gangue minerals can also be defined. The example shows a sample with a mineral list with a morphochemical classification (Figure S7). The mutual interconnection between minerals can be described by defining the liberation parameters, which can be exported for all minerals after analysis, together with the association and interlocking data.

#### **3. Examples of Applications**

#### *3.1. Mapping Speed and Composition*

Sample 511923 was analyzed with the purpose of investigating how different types of Mineralogic analyses (line scan and mapping) and a variation in step size affect the analysis speed and results in a fairly homogenous sample. In total six different analyses at 235.2× magnification were run on the same 0.5 × 0.5 cm area of the sample (Figure 2) consisting of 20 frames. Two line-scan analyses with 10 and 20 μm step size, plus four mapping analyses with step sizes ranging from 40 to 5 μm were carried out on the same area of the sample. The parameters for the analyses were the same using 20 kV acceleration voltage, 120 μm aperture, 275 kcps throughput rate for the EDX detector, and 0.004 sec dwell time.

**Figure 2.** Same area of sample 511923 analyzed with different methods. (**a**) BSE image. (**b**) 20 μm step size line-scan. (**c**) 40 μm step size mapping. (**d**) 20 μm step size mapping. (**e**) 10 μm step size mapping. (**f**) 5 μm step size mapping. For analysis times see Table 3. The 10 μm step size map consists of 244606 analysed pixels.

The line-scan analyses both completed in just around 4 min, and the most time consuming of this type of analysis is to capture the (BSE) image prior to analysis. For the mapping analyses, there is a huge increase in analysis time, when decreasing the step size (see Table 3).

**Table 3.** Comparison between most abundant and selected other minerals between the different analysis methods in area % for di fferent step sizes in line scanning and mapping. STD = standard deviation. Sample 511923: 24 frames at 235.2× magnification (ca. 0.5 × 0.5 cm). Sample 521106: 220 frames at 235.2× magnification (1.5 × 2 cm).


As seen in Figure 3, the distribution of mineral phases between the runs are very similar. Also, when looking on the standard deviation for each mineral phase between the di fferent analysis, there are only small di fferences (Table 3). However, there are more minor phases picked up by the 5 μm mapping method than the other methods. With the line-scan methods no zircon or rutile were analyzed, however both phases were present in all of the mapping methods. Line scan does not find all minor phases in the sample, and in foliated samples detected small equidimensional phases may ge<sup>t</sup> overrepresented when scanning perpendicular to the foliation.

We use the more inhomogeneous sample 521106 to investigate how the di fferent analysis methods apply on a more complex sample. The setup is similar to that used for sample 519923, except for the area analyzed, which is 220 frames at 235.2× magnification (1.5 × 2 cm) (Figure 4), and only one line-scan, at 20 μm, was applied (Figure 5). The results are similar as well, with zircon not identified in the line-scan analysis. Otherwise, the di fferent step sizes in the mapping analysis produce results with minimal variations (Figure 5; Table 3).

The line-scan feature has proven to be very useful as a way of obtaining a quick and precise overview of the geochemistry and major element mineralogy of the sample. Also, a more detailed mineral list can be built from the line scan runs. Where the focus of investigation lies on the chemistry and mineralogy, line scan is a good alternative to running a full map. However, many textural features (e.g., zonations, reaction rims, included minerals) and accessory minerals are not picked up by line-scan analysis.

Mineral mapping at a large step size (here 40 μm in 8 min for 20 frames and 83 min for 220 frames) also is a very fast analysis and gives a good overview of the entire sample (see Figures 2 and 4 to compare image quality). Afterwards ares can be chosen for a more detailed analysis with the mapping feature at a small step size. This two-step procedure is faster than mapping the entire sample at a

small step-size. The difference in step sizes generally has little influence on the mineralochemical data (Figures 3 and 5), however details like zonations or reaction rims are often not detected with a larger step size (Figure 4).

**Figure 3.** Logarithmic diagrams comparing the different analysis methods (y-axis) against the most time consuming and most detailed 5 μm step size mapping method (x-axis). Data is shown as area %. (**a**) 20 μm step size line-scan. (**b**) 10 μm step size line-scan. (**c**) 40 μm step size mapping. (**d**) 20 μm step size mapping. (**e**) 10 μm step size mapping. Dateable minerals are zircon, apatite, rutile and sphene.

**Figure 4.** BSE and Mineral maps of sample 521106 and a selected area for comparing the map quality between different step sizes. (**a**) BSE image consisting of 220 frames. (**b**) 5 μm step size mineral map (ca. 9 million pixels). The square indicates the position of details in (**<sup>c</sup>**,**d**). (**c**) Selected area mapped with 40 μm step size. (**d**) Selected area mapped with 5 μm step size.

**Figure 5.** Logarithmic diagrams showing the correlation between different analysis methods (y-axis) and 5 μm step size mapping (x-axis). Data is shown as area %. (**a**) 20 μm step size line-scan. (**b**) 40 μm step size mapping. (**c**) 20 μm step size mapping. (**d**) 10 μm step size mapping.

#### *3.2. Garnet with Quartz Inclusions*

In sample 508607 large garne<sup>t</sup> grains were observed, which were investigated for their chemistry and mineralogy in order to gain more information on the metamorphic history of the volcanoclastic rocks. The sample was collected from the same unit of volcanoclastic sediments as sample 508599 discussed below. However due to lithological variation, its bulk composition is different (less Na, less K, less Si, more Ca, more Fe, more Al, see Table 2), which is expressed in a different mineralogy: quartz, anorthite, biotite, pale amphibole, garnet, apatite and pyrite. In the most biotite-rich amphibole-poor layers, large garnets have grown with the foliation bending around them (Figure 6). In the garnets, inclusions of apatite, rutile, ilmenite and especially quartz are present. The quartz is irregularly shaped with lobes intruding into the garnet. All inclusions are only found in the center of the garnets, but not in the rim. Pyrite is partially oxidized and is found in association with biotite.

The sample was imaged with the BSE and CL detectors, as well as analyzed for its mineralogy and chemistry, all with the Mineralogic software. In the SEM recipe of the Mineralogic software, the detector for imaging can be selected and changed to CL in order to make a stitchable series of images automatically. After creating a Mineral map (Figure 6), the Mineral recipe was applied to generate element concentration maps for the main elements in garne<sup>t</sup> (in Figure 7 Mg, Fe, and Ca are shown). The element concentrations displayed are true wt% concentrations, not relative intensities. Furthermore, the quality of the EDS systems was monitored with point analyses on silicate standards and yields an error < 1 wt% compared to electron microprobe analyses, with slightly too low values for light elements and too high on the heavier elements (see Supplementary Table S1).

The mineral map (Figures 6 and 7a) reveal that the inclusions in garne<sup>t</sup> are mainly observed in the core and inner rim of the garnet. The core of the garne<sup>t</sup> mainly yields quartz and biotite (plus pyrite) inclusions, while ilmenite, rutile, amphibole and apatite mainly, but not exclusively, are observed in the inner rim. The inner rim does not show biotite inclusions.

CL investigations show that the quartz inclusions in the garne<sup>t</sup> consist of smaller grains healed into larger inclusions (black arrows in Figure 7b). Anorthite adjacent to garne<sup>t</sup> reveals growth rims in CL (white arrows in Figure 7b), which are most strongly on the left and right sights of the grains.

**Figure 6.** Mineral map (**a**) and BSE micrograph (**b**) for the volcanoclastic sediment with garne<sup>t</sup> porphyroblasts. The mineral map contains 281714 pixels.

**Figure 7.** Detailed maps and image of the central garne<sup>t</sup> in Figure 6. (**a**) Mineral map. See Figure 6 for a legend of the colours applied. (**b**) Cathodoluminescence image mosaic of the same garnet. White and black arrows point to growth features in anorthite and quartz, respectively. (**c**) Mg element concentration map. (**d**) Ca element concentration map with black contours indicating the weak zonation in garnet. (**e**) Fe element concentration map. Arrows in (**<sup>c</sup>**,**<sup>e</sup>**) point to an example of an included biotite grain with different Mg–Fe ratios.

EDS analyses of the garne<sup>t</sup> showed ca. 19% Si, 13% Al, 2% Ca, 5% Mg and ca. 26% Fe. There is some Mn present as well (~0.5 wt%), but no Cr (less than 0.5 wt%, usually not detectable). There is a minor compositional variation between the core and the two rims of the garnet, with no difference in Si and Al (less than 1 wt%), but slightly lower Ca and higher Mg and Fe in the inner rim and slightly higher Ca, lower Mg in the outer rim (the difference in concentrations between the core and rims is always less than 2 wt%). These zonations are just visible in Figure 7c–e. With image manipulation (inversion of Ca-image, followed by adding the images together in Adobe Photoshop©), these differences can be enhanced and mapped. Figure 7d shows the contours for the weak zonation, based on Fe, Mg and Ca concentrations. The small variations in the chemical composition are visible in the garne<sup>t</sup> ternary diagram (Figure 8), where the XMg–XFeMn–XCa composition of the garnets in Figure 6 are plotted based on pixel-by-pixel data exported from Mineralogic analyses of the garnet. The garne<sup>t</sup> is an almandine, which are typical for aluminous rocks deformed at amphibolite facies conditions.

**Figure 8.** Ternary diagram showing the garne<sup>t</sup> porphyroblast minor element composition expressed as XMg–XFeMn–XCa. (**a**) Data for 6400 representative pixels; (**b**) Contours showing the data intensity of (a). The Figures are plotted applying WxTernary [42].

The garne<sup>t</sup> in this samples has probably grown from a quartz consuming reaction, possibly chlorite + quartz + muscovite = garne<sup>t</sup> + biotite + H2O, which can explain for the biotite and quartz inclusions in the core of the garne<sup>t</sup> (Figures 6 and 7). From the element concentration plots (Figure 7c,e) can be seen that the composition of the biotite in the inclusions is different from the biotite in the foliation: the included biotite is Mg-rich, while the biotite in the foliation is Fe-rich. The inner rim of the garne<sup>t</sup> showed a continuation of the same reaction, probably at a slightly higher temperature, thus slightly changing the Fe–Mg concentration in biotite and garne<sup>t</sup> [43] (Figure 7). Ti is largely incompatible in aluminous garne<sup>t</sup> and remains as ilmenite and rutile inclusions. The foliation of the rock bends around the previously formed garne<sup>t</sup> and some biotite is formed in the pressure shadows demonstrating a weak dextral sense of shear (Figures 6 and 7). The fractured and healed quartz and anorthite visible in the CL image show that deformation might have been fast and intense, as quartz and anorthite typically are deforming plastically at garnet-forming temperatures [44,45]. Large rims around anorthite show that anorthite precipitated under a differential stress from fluids during metamorphism (darker rims of anorthite visible in the CL image (Figure 7b) are thicker along the horizontal grain axis than the vertical). Ca in garne<sup>t</sup> may also originate from these fluids.

#### *3.3. Reaction Rims Around Rubies*

The ruby-bearing sample 521111 from the Fiskenæsset complex contains the reaction products of the interaction of a tonalitic sheet intruding into an ultramafic rock in contact with anorthosite. As not all of the ruby in the thin section is of gem-quality, we refer to the ruby as corundum. The reaction occurred at amphibolite-facies metamorphic conditions, i.e., post peak-metamorphism (which was at granulite facies conditions) [37,41]. It has been discussed previously whether the alumina of corundum and other aluminous minerals (spinel, kornerupine, cordierite, and sapphirine) are primary or secondary minerals [37,38,41,46]. Reaction temperatures are estimated to be ca. 600 ◦C, for the southern part of the complex [37,41], but may have been higher nearer to the village of Fiskenæsset where the current sample was collected. with a gradient from lower granulite facies conditions in the western part of the Fiskenæsset complex. The corundum-forming reaction occurred during the main folding phase in the region [37].

Like for the garne<sup>t</sup> example above, the sample was mapped for its mineralogy, and afterwards recalculated with new mineral lists to create mineral association maps and an element ratio map (Figures 9 and 10). The mineral association maps can be made in the Mineral recipe tab by redefining the colours of the minerals, highlighting two or three phases, while all other mineral phases are displayed in white. As for the element concentration maps, the mineral list can also be redefined to show steps in element ratios in order to display e.g., changes in Si/Al or Fe/Mg ratios in the sample.

**Figure 9.** Corundum (Ruby-)bearing rock showing the mineral assembly after the formation of corundum, sapphirine, cordierite, anorthite, pale amphibole and biotite (peak metamorphic reaction products) and the retrograde reaction products magnesite, chlorite, and quartz. (**a**) BSE image of a selected part of the sample. (**b**) Mineral map of nearly the entire thin section of the sample (1.63 million pixels). (**c**) Mineral association map of biotite and ruby. All other minerals are displayed in white.

**Figure 10.** Detail of the sample displayed in Figure 9. (**a**) Corundum-sapphire-cordierite association map. (**b**) Corundum-sapphirine-anorthite association map. All other minerals are displayed in white. (**c**) Element ratio map showing the Si/Al ratios in the sample. wt% indicates that wt% data were used to calculate the ratios. Legend for (**<sup>a</sup>**,**b**) in Figure 9.

Desilification of the ultramafic rock results in the formation of pale amphibole (mainly gedrite), biotite, and one or more aluminous phases (cordierite, sapphirine and corundum), e.g., by the reactions: olivine + K-feldspar (tonalite) + H2O = biotite + anthophyllite or olivine + SiO2(aqueous) (tonalite) = anthophyllite [41] (Figure 9). The intruding tonalite is not peraluminous and can therefore not be the source of aluminium, however anorthite is able to react in a balanced reaction: olivine + anorthite + H2O (tonalite) = Ca-bearing amphibole + corundum [41]. In the case of sample 521111, where reaction temperatures may have been slightly higher, the corundum mineral is surrounded by two rims of reaction products (Figures 9b and 10a,b). Corundum in the centre, sapphirine around the corundum, and cordierite or anorthite around the sapphirine. The reaction rims are very well developed with the same width all around the corundum minerals, no symplectites or other fine-grained minerals are observed in the reaction rims, which are therefore interpreted to have developed near-simultaneously with the corundum-forming reaction at peak or near-peak metamorphic conditions. Figure 10c indicates that the minerals formed in the reaction rims are increasingly rich in Si compared to Al, ranging from no Si in ruby to more Si than Al in cordierite. This shows that the degree of silica-desaturation of the rock may have fluctuated during the ruby-forming reaction, or that the Al from the original anorthite was consumed towards the end of the reaction series. Cordierite or anorthite as the outer reaction rim, occur systematic (Figures 9b and 10a,b) and may reflect the original hornblenditic and anorthitic layers, respectively, in the anorthosite before intrusion of the tonalite. Biotite is not associated with corundum (Figure 9c), but cordierite seems more strongly associated with biotite, than anorthite (Figure 9b). Here, sapphirine and cordierite are peak metamorphic minerals, ruby started growing just before peak metamorphic conditions, sapphirine and ruby are part of the peak metamorphic assemblage, and cordierite and anorthite grow immediately after initial decompression [42], as was also observed in other parts of the Fiskenæsset complex [37,41]. However, a decrease in temperature could additionally create retrograde sapphirine or plagioclase, as also was described previously for other parts of the Fiskenæsset complex [46].

#### *3.4. Grain Size Distribution of Chromite in Leucogabro*

Within the Fiskenæsset gabbro and leucogabbro thin layers of chromitite have been observed, which previously have been investigated for their platinum group elements content [34,47]. Within the Fiskenæsset complex both primary and secondary chromite occur [48]. The chromite in the layer is homogeneous in composition but shows a wide variation in grain size. The chromite grains are situated in the Mg-rich hornblende layers of the metamorphosed gabbro (Figure 11a), while only a few grains are associated with anorthite. The chromitite is highly dominated by chromite (Figure 11b), thus individual chromite grains are touching each other, which makes automated grain size analysis more complex.

In order to measure the grain sizes a modified version of the watershed method, which the image processing tap of the software provides, was applied. After thresholding to select the chromite grains from the rock, the image was eroded in three iterations, followed by dilation, as a modification of the opening function that is routinely applied—this gives better separation of individual grains. The watershedding was set to 40 units difference in order to separate touching grains. After image processing the chromite was selected by mineral classification (minerals tab) and characterized by morphology and chemistry in the same tab.

The association data for sample show that 54% of the chromite is in contact association with hornblende and only 17.1% with anorthite, while 23% is associated with the background (mainly holes and cracks between chromite grains) and the remaining fraction of the association is made up by minor phases including rutile and chrome-spinel. The grain size distribution of the grains is visualised in Figure 11. It shows that individual chromite grains in this layer range in size (Feret mean diameter) from 100 to 1000 micrometer. Part of these grains are clusters of several grains, despite the watershed procedure during the image processing. Chromite settled together with hornblende during the cumulation of the igneous rock. Individual chromite grains were already inter-grown during their crystallization and their grain size and texture have not been affected by metamorphism. The chromite grains in this sample are primary minerals. An analysis of the roundness and orientation of the individual grains (Feret angle) shows that the chromite grains are not systematically flattened during the three tectonometamorphic events that affected the area, showing that the chromite grains are very rigid under those tectonometamorphic conditions (ca. 600 ◦C, 5 kbar [37,41]).

**Figure 11.** Leucogabbro with bands of chromite, sample 510152. (**a**) BSE micrograph of part of sample, showing chromite grains in white. (**b**) Mineral map of a larger part of the same sample. Chromite is associated with hornblende. 1.63 million pixels. (**c**) Grain size map showing the same part of sample as in (**a**), with colour-coding of the chromite grain sizes. (**<sup>a</sup>**,**<sup>c</sup>**) partially overlap with the upper left part of (**b**), indicated with squares in (**b**,**<sup>c</sup>**).

#### *3.5. Maximum Feret Angle Determination*

Sample 508599 was investigated for the size and orientation of the biotite minerals in the thin section, which define the foliation and the isoclinal fold in the sample. The orientation of the biotite minerals can be described with the maximum Feret angle. The maximum Feret angle is the angle between the maximum Feret diameter of a particle and the horizontal axis. The maximum Feret diameter is the longest diameter of irregular shaped particles.

In order to measure these morphological features for biotite in the volcanoclastic sediment, in the image processing recipe the BSE image of the sample (Figure 12A) was thresholded by grey scale value to only investigate the minerals with a bright BSE contrast; these include mainly biotite and a few accessory phases. The biotite grains are now isolated features of single grains or clusters of grains that lie in a darker matrix. The biotite grains can thus be investigated with a spot centroid or feature scan analysis. The mineral list was adapted to show all none-biotite grains in black, while a morphochemical criterion was added to the biotite classification in order to classify the biotite grains in colours according to their Feret angle (Figure 12B). The results are shown in Figure 12, together with the BSE image and the mineral map of the entire thin section.

**Figure 12.** Isoclinal fold in a volcanoclastic sediment containing biotite, amphibole, anorthite and quartz. (**A**) BSE micrograph of the sample. The three main phases (biotite, anorthite and quartz) are light, intermediate and dark grey respectively. (**B**) Maximum Feret angle map over the thin section. All none-biotite minerals are black, while biotite is coloured by its maximum Feret angle. Positive and negative angles are indicated in the same colours. (**C**) Mineral map showing the distribution of the mineral phases in the sample. This mineral map contains 2.04 million pixels. Squares define three smaller areas within the sample, with their own data for surface area and association data.

The BSE image and the Mineral map show a different behavior for the individual minerals in the sample across the isoclinal fold. In the flanks of the fold, biotite occurs in clusters of grains. Individual grains are occasionally oriented randomly, but the larger clusters of grains are oriented into elongated clusters with low Feret angles (preferentially 20–40◦; see Figure 13), while most individual biotite grains and smaller clusters have lower Feret angles (mainly 0–20◦) and are stacked stair-case-wise to follow the local foliation. Thus, the individual grains follow the main foliation in the area (oriented ca. 0–10◦ with respect to the horizontal axis), while they form stacked arrays that follow the local foliation (20–40◦) in the individual fold flank. In the core of the fold, biotite occurs more scattered, and all Feret angle orientations with an large elongation are observed between −30 and 40◦ (Figures 12 and 13), while the grains with the largest area have orientations around 40 and −10◦ (Figure 13), fitting with the orientation of the two flanks mapped in Area 2 (Figure 12). In the least deformed part of the fold (Area 3) biotite grains are smaller and show no preferred orientation for the most elongated grains (Figure 13).

**Figure 13.** Plots of the maximum Feret angle frequency, and this angle compared to the elongation and area of the biotite grains and clusters of grains. Figures represent the entire map in Figure 12, as well as the three areas outlined in Figure 12C.

In the less extensively deformed parts of the fold, quartz and plagioclase occupy roughly the same percentage of the area (compare least deformed Area 3 to intensively deformed Areas 1 and 2 in Table 4). Both minerals are partially intermixed but prefer to cluster with their own phase. In the flanks of the fold, a more well-defined layering of quartz, biotite and anorthite occurs, where both quartz and biotite are sandwiched between layers of anorthite. In the core of the fold, hardly any quartz minerals are observed. Biotite is associated with anorthite (Figures 12C and 13). This is illustrated

for association data for the three areas in Figure 12C (see Table 4). Area 1, derived from the core of the fault, and Area 2, derived from the flank, show that biotite is strongly associated with plagioclase (45.3 area% and 48.7 area%, respectively), while in the least deformed Area 3 this association is only 34.0 area%. Plagioclase in the flanks of fold (Area 2) is strongly associated with biotite (45.3 area%), and the association with quartz is the least of the three areas (34.5 area%). Plagioclase in the core of the fold (Area 1) is forming the sandwiching layer between biotite (37.8 area% association) and quartz (40.9 area%). In the least deformed part of the sample, plagioclase is strongly associated with quartz (51.6 area%) and much less with biotite (21.3 area%), which is partially caused by a much lower area% of biotite in this section.


**Table 4.** Area% and association data for biotite, quartz and plagioclase in the three areas indicated in Figure 12C. TS = thin section.

Folding in the area occurred at ca. 600 ◦C [37,41]. Under these conditions all three minerals (quartz, anorthite and biotite) deform plastically under most strain rates [44,45]. Quartz, which under these conditions is more plastic than plagioclase, moves to the flanks of the fold, while the more rigid plagioclase is pressed towards the cores of the fold (compare Areas 1 and 2 in Figure 12C and Table 4). Biotite rotates by a combination of recrystallization and mechanical fracturing and healing to accommodate to the main stress in the region and can be used to read the record of the main foliation outside the isoclinal fold (Figure 13), i.e., 0◦. However, the movement of quartz and plagioclase also force the biotite to follow the local foliation, leading to a stacking of horizontally oriented minerals into a local foliation of ca. 30◦ to the horizontal fold axial plane (compare to Figure 13). Continued stress on the rock will remove the evidence for a fold core from the record of the rock, leaving the sample layered by mineral phase and with clusters of flat-lying biotite. We are thus able to study the generation of a schistose layering from an initially little deformed rock in a single sample by studying individual areas of the fold core.
