For the digital image processing and interpretation, Landsat-8 OLI and Sentinel-2A multispectral images were used because their spectral ranges are more adequate for lithological discrimination and mineral detection. This is due to important absorption mineral features in the near-infrared (NIR) and short-wave infrared (SWIR) that must be covered by the specificc band’s spectral ranges (Cardoso-Fernandes et al., 2019). For detailed information on the spectral and spatial characteristics of Landsat-8 OLI and Sentinel-2A, see the supplementary data file (Supplementary Section S1 and Supplementary Table S1).

Two cloud-free and georeferenced scenes of level-1T Landsat-8 and level-1C Sentinel-2A were obtained from the USGS platform and processed in this study. The preprocessing techniques of pixel resampling, subsetting, radiometric, and atmospheric corrections were considered and performed. The atmospheric correction is used to minimize the influences of atmospheric factors in multispectral data (Liang and Wang, 2020; Van der Meer et al., 2014). The Landsat-8 digital numbers (DNs) were converted to top of the atmosphere (TOA) reflectance. The Landsat-8 is subjected to the dark object subtraction (DOS) atmospheric correction to convert TOA reflectance to surface reflectance (Chavez, 1996). The Sentinel-2A image is already provided in TOA reflectance. The internal average relative reflection (IARR) method was applied to the Sentinel-2A data. The IARR technique for mineral mapping does not need prior knowledge of the geological features (Ben-Dor et al., 1994). The IARR normalizes the images to a scene with an average spectrum. The image processing is performed using ENVI v. 5.3 and the Sentinel Application Platform (SNAP v. 7) software for the image processing where the output of geologic and geospatial maps are carried out by the ArcGIS v. 10.5 packages.

Major, trace, and rare earth elements (REEs) were carried out by (ICP-MS) in ACME Labs, Vancouver, Canada, following fusion with lithium tetraborate/metaborate with digestion by nitric acid. Calibration of the ICP-MS was carried out using standards. The analytical precision, as calculated from replicate analyses, was 0.5% for major oxides and varied from 2 to 20% for trace elements. The detection limits for trace elements and major oxides were 0.01–0.5 ppm and 0.001 wt%-0.04 wt%, respectively. Some samples were tested for major elements using the method of Shapiro and Brannock (1962) using lithium tetraborate as a fux at the laboratories of the Nuclear Materials Authority, Cairo, Egypt. The corundum was analyzed with a JEOL JXA-733 electron probe microanalysis (University of New Brunswick) equipped with ds spec and dQant32 automation (Geller Micro Analytical Labs). An accelerating voltage of 15 kV and a probe current of 30 nA. The peak counting times were 30 s (Al and Si), 90 s (Ti, V, Cr, and Cs), and 120 s (Na, Mg, K, Ca, Mn, Fe, and Ni. The environmental scanning electron microscope (ESEM) investigation was done at the Nuclear Material Authority (NMA). The ESEM is equipped with link analytical AN-1000/855 energy dispersive X-ray spectrometer (EDX) calibrated using natural standards to identify elements and detect, semi-quantitatively, the corundum compositions.

Generally, all the mineral deposits show characteristic spectral signatures on multispectral spectral remote sensing systems based on their mineralogy and texture. As mentioned above the detection of pegmatites and associated mineralization in the examined area is based on the identification of the country rocks (ultramafics and gneisses) and the hydrothermally altered zones. The hydrothermal alteration zones are characterized by substantial contents of hydrous ferromagnesian silicate and iron oxides. Therefore, the band ratio technique was applied to Landsat-8 and Sentinel-2A data to recognize types of hydrothermal alteration. For the Landsat-8, the band ratios 4/2, 6/7, 4/3, and 6/5 (Figures 5C,E, Supplementary Figures 3A,C, respectively) are used to map iron oxides, OH-bearing minerals, ferric iron, and ferrous minerals, respectively. Where the band ratios of sentinel-2A of 4/2, 11/12, 11/8A, and 12/11 (Figures 5D,F, Supplementary Figures 3B,D, respectively) are used to map iron oxides, hydroxyl minerals, ferric oxides, and ferrous silicate, respectively. A density slice method was applied to enhance the band ratio results. The color ramp of this method offers the lower values (e.g., cyan) in cold colors and the higher values in hot colors (e.g., red). The resultant images emphasize subtle gray scale differences that may be imperceptible (Sabins, 1996).

Hunt (1977) pointed out that the corundum Al₂O₃ mineral type provides strong absorption in 2.10 µm and shows an intense absorption feature in 2.40 μm of the electromagnetic spectrum. Moreover, the standard spectral curve of corundum in the USGS spectral library shows that it provides high reflectance in 1.5 µm and high absorption in 2.1 µm (Figure 6A). Therefore, the band ratios 6/7 of Landsat-8 and 11/12 of Sentinel-2A are very helpful in detecting the corundum-bearing pegmatites and Al₂O₃ rich rocks in bright pixels in the study area. The constrained energy minimization (CEM) technique developed by Harsanyi (1993)has been used to solve the typical adaptive beam-forming problem in signal processing. CEM has become one of the most widely used techniques today for minerals mapping (Gabr et al., 2010). This technique was adopted in the present study as a spectral mapping method for detecting corundum-bearing pegmatites. It was applied to Landsat-8 bands to detect corundum minerals with reference to the USGS spectral library. The potential distribution of corundum-bearing pegmatites is dropped over gray scale band ratio of 6/7 of Landsat-8 (Figure 6B) as well as in El Qarn area and south El Hima areas (Figures 6C,D, respectively).

The aforementioned processed color composite and gray scale images of Landsat-8 and Sentinel-2A integration with field investigations, petrographical, and geochemical data interpretation have been used to update the geological map of El Ramly and Greiling (1988) of the study area (Figure 1B). In addition, the new geologic map recognized the pegmatite bodies, veins, and lenses, which contain corundum mineralization in the study area.

Representative electron microprobe analysis (EMPA) of the examined corundum revealed that the Al₂O₃ is the dominant constituent (Supplementary Table S6). The Al₂O₃ reaches up to 99.72 wt.% in pegmatite-gneiss and up to 100 wt.% in pegmatite-ultramafics (Um Nimr area). Other constituents such as SiO₂, MnO, CaO, MgO, FeO, Cr₂O₃, TiO_2, Ga₂O₃, Na₂O, and K₂O have an average of less than 0.5 wt.%. The main differences in the examined corundum of all localities are in chromophores (traces) such as Ti, Cr, Fe, and Mg. The studied corundum of pegmatite hosted in-gneiss contain the lowest amounts (Maximum concentrations) of TiO₂ (0.02 wt.%), FeO (0.36 wt.%) and Cr₂O₃ (0.29 wt.%) relative to those of pegmatite hosted in ultramafics (0.12, 0.42, and 1.88, respectively).

The tested false-color composites are apt to discriminate between the different lithological units, lithological contacts, and structures in the study area. The composites that better identify hydrothermally altered rocks are RGB-567 for L8 in faint green tones (Supplementary Figure 2A). The RGB-2812 for S2A demarcates the iron oxides and clay minerals in whitish pink color (Supplementary Figure 2B). The two composites of RGB-327 for L8 (Figure 4A) and RGB-3212 for S2A (Figure 4B) manifest rocks enriched with ferromagnesian minerals by strong dark pixels, and the degree of pixel opacity decreases slightly for rocks with lower ferromagnesian content. The highly folded ultramafics and metagabbros can be easily detected in the pinkish-gray color of the white gneissic granodiorite. Al₂O₃ rich rocks are enhanced by the light to dark buff color as shown in Figure 4A.

The ultramafic rocks can be well identified in the band ratio of RGB-6/7, 4/3, 5/4 for (L8) (Figure 4C) with the dark to bright bloody color. These rocks are separated with sharp contacts from the dark blue pixels of biotite gneisses and pinkish-gray color of gneissose granodiorite. The pegmatite veins and bodies exhibit orange color. The band ratio RGB-11/7, 12/5, 11/12 of (S2A) (Figure 4D) shows the ultramafics in lemon green, gneisses in bright green, the highly injected white gneissic granodiorite with pegmatites in blue colors. The pegmatites are demarcated by the light magenta color. The Chica-Olma ratio RGB-6/7, 6/5, 4/2 of (L8) (Supplementary Figure 2C) highlights the ultramafics rocks in reddish yellow, gneisses in dark green, gneissic granodiorite-tonalite in light violet to dark blue color. The pale yellow color of pegmatite bodies can be detected well from the dark background colors of gneissic granitic rocks. In addition, the ultramafic rocks in the six studied areas of Um Maiat, Abu Merikhat, Um Nimr, North El Hima, South El Hima, and Um Shegera exhibit bright green colors in the band ratio RGB-12/8A, 11/8A, 12/11 of (S2A) (Supplementary Figure 2D). These ultramafics are injected by pale color pegmatite veins and lenses.

Loughlin (1991) pointed out that the magnitude and sign (positive or negative) of the eigenvectors depend on which spectral properties of rocks and/or minerals are responsible for the statistical variance mapped into each principal component (PC). The eigenvector loadings of PC1, PC2, PC3, and PC4 show high contributions with bands 5, 6, 7 of Landsat-8 (Supplementary Table S2). Therefore, PCA123 (Figure 4E) and PCA432 (Supplementary Figure 2E) were performed to highlight areas enriched with iron oxides and clay minerals, which are associated with the pegmatite veins. The dark red and blue pixels have a high surface abundance of iron oxide/hydroxide minerals, which are mostly associated with ultramafic rocks (Figure 4E and Supplementary Figure 2E). Al–OH mineral groups which occurred in the gneissic granitic rocks were enhanced by cyan and light green colors in Figure 4E and Supplementary Figure 2E, respectively. The pegmatite bodies are brought out by the dark red pixels in the two images. In accordance with the eigenvector loadings of the Sentinel-2A, the PC1, PC3, PC4, and PC6 have strong loadings of bands 8A, 11, and 12 (Supplementary Table S3). Therefore, the PCA145 (Figure 4F) and PCA643 (Supplementary Figure 2F) were produced to map the gneisses in pink pixels in Figure 4F and in pinkish-blue in supplementary Figure (2f). Al₂O₃ rich rock enhanced with bright green pixels in Figure 4F and yellowish pale green in Supplementary Figure 2F. Moreover, the MNF-531 of Landsat-8 (Figure 5A) and the decorrelation stretch for the RGB-12118 (Figure 5B) of Sentinel-2A shows good and useful discrimination of ultramafics (in green and light blue pixels), gneisses (in red pixels), biotite gneisses (greenish cyan pixels), and white gneissose granite (in cyan pixels). A good appearance of the pegmatite veins and lenses in a bright yellowish-green cut the ultramafics and gneissic background in the two images (Figures 5A,B).

To enhance the mineralization zones in the study area, specific band ratios were applied to Landsat-8 and Sentinel-2A data. The band ratios 4/2 to Landsat-8 (Figure 5C) and Sentinel-2A (Figure 5D) are used to delineate the iron oxides commonly in bright pixels in the gray scale version. The present study offers the color density slice version of the two gray scale ratios with red areas corresponding to the highest surface reflectance values and easy to detect. The band ratios 6/7 to Landsat-8 (Figure 5E) and 11/12 to Sentinel-2A (Figure 5F) enhanced in red color the areas rich with hydrothermal minerals, which are associated with the ultramafics and pegmatite bodies. To map the ferric oxides band ratios of 4/3 to Landsat-8 (Supplementary Figure 3A) and 11/8A to Sentinel-2A (Supplementary Figure 3B). Where the band ratios 6/5 to landsat-8 (Supplementary Figure 3C) and 12/11 to Sentinel-2A (Supplementary Figure 3D) are successful to map the ferrous minerals. The resultant images of the previous band ratios provide coinciding results from the Landsat-8 and Sentinel-2A band ratios and confirm each other. The distribution of the iron oxides and hydroxyl minerals helps in detecting the pegmatite veins and associated mineralization.

Figure 6B shows the potential areas to the presence of the corundum-bearing pegmatites based on the CEM technique appear in red color pixels drapped over the band ratio 6/7 of Landsat-8. The validation of this map is carried out using 32 ground control points or sample locations of corundum-bearing pegmatites. The accuracy assessment attained 90.6% where 29 points are failed over the detected potential pegmatites pixels. Figures 6C,D shows the matching of the sample locations with the corresponding potential pegmatites pixels in the Wadi El Qarn area (No. 2, corundum-bearing pegmatites hosted in gneisses, Figure 6C) and South Wadi El Hima area (No. 8, corundum-bearing pegmatites hosted in ultramafics, Figure 6D). One hundred and forty-one (141) pegmatite veins and lenses have been detected in the study area (Figure 1B). Their lengths are ranging between 0.06 and 1.16 km and with an average length of 0.21 km. Figure 11A shows the density map of these pegmatite veins drapped over the shade relief map of the study area. These pegmatite veins are trending mostly NW-SE which is affected by the foliation trend of the gneisses, and the foliated ultramafics. The distribution of Al₂O₃ content percent in the corundum-bearing pegmatites (Supplementary Table S6) drapped over the potential map of the pegmatite (Figure 11B) to reflect the coincide results of the geochemistry and remote sensing. Figure 11B shows the high Al₂O₃ content in the corundum-bearing pegmatites in the Um Shagara, North, and South El Hima areas but offers a moderate content in the rest of the studies areas.

These rocks are remnants of oceanic crust that are related to Neoproterozoic ophiolitic rocks (Figure 12A). Serpentinization processes reflect low-grade metamorphism (Greenschist facies) affecting the rocks. Bulk-rock composition of the examined ultramafics can be used to manifest post magmatic processes affecting the protolith, including metasomatism and partial melting (El Mezayen et al., 2018; Lasheen et al., 2021a).

Despite different tectonic signatures of the ophiolitic rocks in the Eastern Desert of Egypt, recent studies can infer the applicable environment based on primary mantle minerals certainly Cr-spinels and olivine relics. Most studies suggest that the Egyptian ophiolitic ultramafics are formed in sub-arc and emplaced in fore-arc (e.g., Azer and Stern, 2007; El Mezayen et al., 2018; Lasheen et al., 2021a). Bulk rocks analysis (hydrous) show that they have SiO₂ content that varies from 39.14 to 42.67 wt.%, MgO from 41.05 to 44.56 wt.%, Fe₂O₃ from 8.36 to 10.56 wt.%, CaO from 0.25 to 0.93 wt.%, and TiO₂ from 0.05–0.24 wt.%, suggesting high partial melting degrees (Uysal et al., 2012). Further constraints, the studied ultramafic rocks have high contents of MgO (40.05 wt.%), Ni (1745 ppm), Cr (2,134 ppm), and Co (92 ppm), reflecting their depleted mantle sources (Lasheen et al., 2021a). Moreover, their lower contents of Nb and Zr suggest their depleted source (Geng et al., 2011). SiO_2, as well as Al₂O₃ of the ultramafics, are relatively immobile compared to CaO (e.g., Niu, 2004) and their ratios can be used to infer the tectonic setting (Hattori and Guillot, 2007). The studied ultramafics possess an Al₂O₃/SiO₂ ratio (av. 0.05) lower than that of PM (McDonough and Sun, 1995) and MgO/SiO₂ ratios (av. 1.06) higher than that of PM (∼0.85), which compares with fore-arc peridotites. The examined samples straddle in and are akin to (due to higher content of Al₂O₃) fore-arc field (Ishii et al., 1992; Pearce et al., 2000; Niu, 2004; Figure 12A). Further supporting, Ahmed and Gharib (2018) concluded that the Cr-spinel composition (Cr# > 0.7) of Hafafit ultramafics reflects high partial melting degrees and formed within the fore-arc environment. On the other hand, mantle wedge, subducted, and abyssal ultramafics can be separated from each other’s using some trace and REEs. In terms of, U, Ba, and Cs (ppm) against Yb (ppm), the examined rocks are located far from the terrestrial array in the subducted field (Figures 12B–D).

Different origins of the examined gneiss can be deduced using several discrimination diagrams. To infer their igneous origin, the examined samples were plotted on a binary diagram of Tarney (1977). The analyzed samples occupy the igneous source (Figure 13A). Therefore, they are treated as igneous (plutonic/volcanic) rocks in their classification, magma type, and tectonic regime. According to Cox et al. (1979), the analyzed gneisses plotted in the granite field (Figure 13B) and in granite and alkali granite based on De la Roche et al. (1980) (Figure 13C) diagram. Due to the enrichment of total alkalis of the examined gneisses, their parental magma has the chemical composition of A-type granites (Figure 13D) according to Whalen et al. (1987). Therefore, the parent rocks of the examined gneisses can be inferred using some geochemical diagrams such as Nb against Y of Pearce et al. (1984). The protolith of the examined gneisses seems to be formed within the plate tectonic regime (Figure 13E).

Corundum is commonly found in pegmatite either associated with margarite (marundite) or plagioclase (plumasite) (Bucher et al., 2005; Dill, 2015). Corundum-bearing pegmatites in the Hafafit area have both plumasite and marundite (Ahmed and Gharib, 2018). Corundum can be formed by magmatic as well as metasomatic/metamorphic processes (Peucat et al., 2007; Giuliani et al., 2014; Karmakar et al., 2017). Chromophores (e.g., Cr, Mg, V, Ga, Fe, and Ti) can be used as reliable for corundum origin (Peucat et al., 2007; Giuliani et al., 2014; Ahmed and Gharib, 2018).

The magmatic types can be distinguished by high Ga and low Cr contents with Cr₂O₃/Ga₂O₃ ratio less than one, whereas the metamorphic suites are characterized by high Cr and low Ga contents with Cr₂O₃/Ga₂O₃ ratio of more than 3 (Sutherland et al., 1998). The composition of the examined corundum has Cr₂O₃/Ga₂O₃ values ranging from 2.5 to 134, reflecting their metamorphic origin. In addition, the Ga/Mg ratio is used to differentiate between magmatic and metamorphic types. Magmatic corundum is defined by Ga/Mg more than six and less than three for metamorphic and metasomatic types (Giuliani et al., 2014), due to Ga substitute for Al and the Ga/Al ratio is constant in the continental crust (Peucat et al., 2007).

It is noticeable that the investigated corundum has Ga/Mg ratio ≤3, reflecting metamorphic/metasomatic sources (Figure 14A). Further constraints, Fe/Mg ratio is used to differentiate between magmatic and metamorphic corundum. This ratio is high for magmatic (>100) and low for metamorphic and metasomatic types (<100), (Peucat et al., 2007). The studied corundum has a composition that contains Fe/Mg less than 100, suggesting its metamorphic origin. Fe-Mg*100-Ti*10 diagram can be used to differentiate between metamorphic and magmatic corundum (Peucat et al., 2007). The investigated corundum has low Fe and Ti and high Mg contents, therefore it straddles metamorphic/metasomatic field (Figure 14B).

Therefore, the examined corundum is originated through interaction processes between the intrusive pegmatite and the host rocks (ultramafics and gneiss). Desilication of these pegmatites led to an excess of alumina in the residual melt producing corundum mineralization. The studied corundum contains low Fe and high Mg contents as typical metamorphic corundum, which is indicated by Mg-Fe-Ti ternary diagram (Peucat et al., 2007). Figure 15 shows a simple schematic model that illustrates the metasomatic interaction between felsic magma (pegmatite) and ultramafic rocks and gneisses. The emplacement of pegmatite (silica and alumina rich) within ultramafic rocks (serpentinites/silica deficient rocks), the mineralized zone has been formed with new mineral associations in serpentinites. The newly formed minerals include actinolite, anthophyllite, and vermiculite. The residual melt becomes more Al-rich due to silica diffusion. Therefore, alumina-rich minerals have been formed such as corundum, hercynite, and almandine. On the other hand, the tourmaline-rich zone was formed at pegmatite and gneiss (containing B-rich fluids) contact. In addition, the small scale of corundum-bearing pegmatites within ultramafics rocks represent the main factor for preventing ruby (red corundum) due to less remobilization of Cr and Fe.