**4. Results**

The following results are presented as comparison and verification of SEM-MLA and Raman mapping. Emphasis is given on imaging, discrimination of magnetite and hematite and modal mineralogy in area % with respect to the time demands for measurements and processing.

#### *4.1. Classification and Imaging*

The mineral distribution maps of both classifications (SEM-MLA and Raman) correspond well with the thin section of the reflected light photograph (Figure 2). Both methods recognized the present minerals. The classification with MLA was manually adjusted with a BSE grey level range from 150 to 190 for hematite and 191 to 250 for magnetite. Mineral boundaries are pre-defined based on the BSE grey values with the SEM-MLA. This function is not ye<sup>t</sup> available for Raman imaging resulting in a rasterized resolution. General features, textures of the sample material and the abundance and location of important minerals are visible with both methods.

#### *4.2. Discrimination of Magnetite and Hematite*

In the reflected light image of the polished sample of apatite iron ore, magnetite, hematite, and gangue minerals (mostly apatite, carbonates, and silicates) can be seen (Figure 3). In SEM-based backscattered electron images, iron oxides are hardly distinguishable from each other due to similar BSE grey values. The processed false-colored mineral maps from MLA and Raman show all present phases. However, the precise discrimination of magnetite and hematite varies with both techniques. A smaller area (see the red square in Figure 3) was selected for high-resolution Raman mapping. Six minerals were spectrally identified within this smaller area, corresponding to magnetite, Ti-free hematite, Ti-bearing hematite, quartz, ankerite, bornite, and chalcopyrite. The di fferent Raman spectra for hematite were assigned to Ti-bearing hematite by EDS data.

## *4.3. Modal Mineralogy*

The modal mineralogy can be extracted in the Dataview software for SEM-MLA measurements. Raman does not commonly o ffer this data analysis feature as an in-built tool. However, quantification of phase abundances is possible based on the number of pixels identified for each phase in relation to the total number of pixels in the scanned area.

The modal mineralogy of the major and minor components obtained for the area displayed in Figure 2 shows similar results between the slow and the fast SEM-MLA and Raman measurements (Table 2). However, magnetite occurs in higher modal abundance in the measurement with lower step size (35.54 area %) compared to the fast measurement (33.62 area %) and Raman (34.9 area %). Contrary, the modal mineralogy of hematite and Ti-bearing hematite is elevated in the fast GXMAP but lower in Raman (Table 2). Quartz, apatite, and ankerite show slightly lower values compared to SEM-MLA. The modal mineralogy of calcite is in the range of ± 0.2 area % for all measurements.


**Table 2.** Modal mineralogy (area %) of major and minor components obtained by SEM + MLA and Raman measurements on sample HA-81155-2. Fe-oxides are highlighted for better comparison.

## *4.4. Time Constraints*

For this study, SEM-based MLA measurements were performed with two di fferent resolutions and step sizes on the same sample in addition to Raman mapping.

## 4.4.1. Measurement Time

The measurement time for fast GXMAP mapping with a resolution of 500 × 500 pixels and a step size of 36 μm was 6 h. Mapping with a step size of 12 μm and the same resolution took 12 h. The identical area of the same sample measured with Raman was the most time-intensive with 42 h using a step size of 30 μm. The measurement time of the detailed Raman map (Figure 3) took 1 h using a step size of 1 μm.

#### 4.4.2. Data Processing Time

The processing time is dependent on many variables, including the operator in charge, technical equipment, e.g., computational power, software requirements, and functions. In SEM-MLA data processing of faster measurements using larger step sizes resulted in more unknowns and misidentification of phases. In this case, additional processing time was required due to manual correction of the sample set.

Data processing in Raman imaging depends strongly on the signal-to-noise ratio of the acquired spectra. Since processing is done mainly o ffline after data acquisition, it is crucial to ensure the best acquisition parameters (e.g., laser power, exposure time) before measurement. To discriminate similar minerals (e.g., Ti-bearing and Ti-poor hematite) longer acquisition times are recommended.
