**1. Introduction**

#### *1.1. Automated Quantitative Mineralogy (AQM)*

Little work has been done in the use of scanning electron microscopy (SEM)-based automated quantitative mineralogy (AQM) in the evaluation of mineralogy and microstructures of fault gouges. Physicochemical processes taking place in these rocks are recorded in the mineralogy and the microstructures of the fault gouge itself. However, the subsequent mineralogical assemblages are incredibly fine-grained, and therefore it is challenging for microanalytical applications to provide a high-resolution quantitative mineralogy of a large area in order to understand these physicochemical processes.

"Automated mineralogy", or more accurately described as automated quantitative mineralogy (AQM), has been available since the early 1980s with technologies such as QEMSCAN [1] and the more recent TIMA-X [2]. The key application area of this technique was to provide an automated and routine analysis capability to quantify the mineralogy and textures of ore mineralogy (modal mineralogy, liberation, association, etc.) to aid mineral processing plant development, optimization, and troubleshooting [1,3].

These automated mineralogy techniques control the electron beam to step across the sample surface, at a user-defined spatial resolution (i.e., 5-μm steps) acquiring unquantified energy-dispersive

spectroscopy (EDX) spectrum. This unprocessed (no matrix corrections or spectrum quantification) spectrum is matched to a list of known referenced EDX spectra to provide a mineral name [2,4]. Without a direct measured chemical quantification from that analysis point, spectral fingerprinting on validated known mineral standards using external techniques such as electron microprobe is used [2,4,5].

The EDX spectrum, once acquired, is matched to the referenced spectrum from the mineral to classify the individual analysis. Whilst this technique is widely applied within AQM, it only provides a "crude" assessment of the mineralogy where elemental concentrations exceed 10 wt.% [5]. As such, this historic methodology falls short of providing a direct chemical analysis or correct analytical procedure to provide the necessary quality of analysis or standardized results.

Since the inception of AQM technologies, no significant methodological development of this capability occurred with little technological advancement despite the huge strides made in EDX technology and analytical capability. Multiple examples in the literature point toward the quantitative measurement capability of modern day EDX on major elements, which are able to achieve relative errors of less than 2% on a standardized analysis of polished samples [6,7].

However, the development of Mineralogic provides a step change in the analytical capabilities of AQM with the integration of these new EDX capabilities. This AQM software acquires an EDX spectrum, and performs matrix corrections and peak deconvolution for each analysis point. Minerals are subsequently classified based on stoichiometry values [8,9], providing new analytical capabilities for AQM and opening up new application areas. Concentrations of elements that cannot be analyzed correctly (e.g., H), can be assigned. The chemistry in each pixel is saved by the Mineralogic software, and the pixel's mineralogy can therefore be reinterpreted after the analysis. Thus, each EDX analysis provides one pixel on the false-colored mineral map [9].

In dispersive spectrometry, fundamental electron beam interaction with the sample still poses challenges. All electron beam microanalysis users will be aware of the challenges of carrying out an analysis at low spatial resolutions. As a result of the energy of the primary electron beam (acceleration voltage; kV) hitting the sample, X-rays are generated from an interaction volume, which may consequently be measured by the SEM's energy-dispersive X-ray spectrometer (EDX). The size of this interaction volume is a function of the primary energy of the beam and the mean atomic number of the sample, and it is usually presented as the diameter of the beam inside the sample material projected in the X-Z plane, which is also called the diameter of the footprint of the electron beam [10]. The interaction volume is larger than the diameter of the beam on the landing spot. In geological samples, typically containing silicates, interaction volumes can be in a region of 2–5 μm, with a 20-kV primary electron beam. Therefore, providing a chemical analysis and mineral classification based on an EDX spectrum can be challenging for fine-grained material, due to the "mixing" of compositions of a number of grains. For those grains, the interaction volume can result in the generation of X-rays from multiple minerals at a single point, providing complex mixed mineralogical compositions. Whilst these challenges remain, there is little alternative substitute for the quantification of such mineral assemblages, as gray level alone is not enough to separate and quantify the mineralogy and textures.

In this study, we provide a proof of concept in showing the capability of a new generation of SEM-based AQM to map the mineralogy and investigate the grain size of the fault gouge of varying composition at a 200-nm spatial EDX resolution.

## *1.2. Geological Background*

The sample that was used for this study was collected in southern West Greenland from the northern coast of Ikkattup nunaa Island in the Ikkattoq anlanngua fjord (Figure 1). This locality is situated in the southern part of the Ravn Storø Supracrustal belt, which is intruded by a well-known Meso-Archaean Fiskenæsset anorthosite-leucogabbro complex. The fault rock sample represents an amphibolite intercalated with orthogneiss. The amphibolite is part of the ca. 3.0 Ga Ravn Storø Supracrustal complex, whereas the intruding orthogneiss was dated to 2.88 Ga. Later tectonic activity in the area includes two folding and thrusting events during the Neo-Archaean, during which the Kvanefjeld block to the south collided with the Bjørnesund block (which includes the sample locality and the Fiskenæsset complex to the north of the outcrop). The thrust zones were reactivated in an extensional setting during the latest Neoarchaean or earliest Palaeoproterozoic, and the fault gouge studied here most likely was formed during this extensional event ([11]; and references therein).

**Figure 1.** Geological map showing the Bjørnesund region in southern West Greenland. Sample locality, as well as the major tectonic boundary between the Bjørnesund Block and the Kvanefjeld block are indicated [12]. Map adapted from the Geological Survey of Denmark and Greenland (GEUS) [13]; reproduced with permission from GEUS.

#### **2. Materials and Methods**

The fault gouge was studied in a polished thin section and shows a mixture of grains that were originally part of the orthogneiss and of the amphibolite. Intensive mixing of the different minerals in the fault gouge prevented healing of the fault gouge to larger grain sizes [14]. Therefore, many comminuted grains of sizes smaller than 200 nm are still present in the sample. However, mineral reactions can be observed: epidote, illite, and chlorite (chamosite) formed to replace the original mineral assembly.

To perform the microanalysis, the thin section across the fault rock was coated with carbon and studied with a Zeiss SIGMA 300VP Scanning Electron Microscope (SEM) equipped two Bruker XFlash 6|30 energy-dispersive spectroscopy detectors (EDX), with 129-eV energy resolution and with the Zeiss Mineralogic automated quantitative mineralogy software platform. The region was selected and imaged to provide a high-resolution back-scattered electrons contrast (BSE) mosaic of the region of interest. In addition, on this region of interest, a 200-nm quantitative mineralogical analysis was carried out using Mineralogic, creating a mineral map (Figure 2).

**Figure 2.** Large area overview image of the scanned sample area. (**a**) Displays the high-resolution BSE image (**b**) Displays the false-colored mineral map and key, displaying the mineralogy across this location.

To analyze the fine-grained fault gouge, we reduced the acceleration voltage of the primary electron beam to 10 kV to reduce the interaction volume [15–17], and therefore to minimize the amount of mixed pixel data, and used the 60-μm aperture providing a 1.8-nA beam current. Despite this reduction of primary electron energy to 10 kV, the interaction volume still has a diameter of up to approximately 1 μm for minerals with the lowest average atomic number (Z) and down to 400 nm for a high Z phase on top of a low Z phase (Figure 3). 200-nm EDX pixel size would greatly provide oversampling and a large volume of mixture pixel signatures.

Despite this, the data below outlines how the analysis and advanced mineral classifications employed in the Mineralogic mineral library were able to deconvolute any mixed signals and provide a reliable quantitative mineralogy and textural assessment. We were able to do this due to the access to fully quantitative EDX classifications for each pixel within the Mineralogic software and the ability to quantify the weight percent contribution of the elements present on a per-pixel basis (see also [9]). This enabled the creation of highly discriminatory mineral classifications whereby "contamination" of the chemical analysis provided by the above described mixed signals were able to be identified and factored into the mineral classifications of the mineral library to enable the correct classification (Figure 4). This capability and deconvolution procedure allowed the creation of true mineral classifications without the need for creating false mixed signal classifications. The mineral classifications could be used on particles down to 200 nm, even though the interaction volume of the beam was larger than 200 nm. Therefore, the obtained resolution of the analysis is 200 nm, which is the applied pixel size for the mineral, despite the larger size of the interaction volume.

**Figure 3.** Monte Carlo simulations for the trajectories of 20,000 electrons in quartz (SiO2) and pyrite (FeS2) in carbon-coated samples. Arrows show the width of the interaction volume at 10 kV for 50% and 90% of the electrons. (**a**) Quartz, (**b**) pyrite, (**c**) 200-nm pyrite on top of quartz, (**d**) 200-nm quartz on top of pyrite. Simulations were made with Casino [18,19].

**Figure 4.** Detailed view of a false-colored mineral map and the associated BSE image. Each pixel represents 200 × 200 nm. Grains of ca. 200 nm can be recognized and classified from the energy-dispersive spectroscopy (EDX) data for that pixel. Yellow boxes display example EDX normalized weight percentages and the interpretation of the operator for the mineralogy.
