The largest plagiogranite on Earth formed by re-melting of juvenile proto-continental crust

The growth of continental crust through melt extraction from the mantle is a critical component of the chemical evolution of the Earth and the development of plate tectonics. However, the mechanisms involved remain debated. Here, we conduct petrological and geochemical analyses on a large (up to 5000 km2) granitoid body in the Arabian-Nubian shield near El-Shadli, Egypt. We identify these rocks as the largest known plagiogranitic complex on Earth, which shares characteristics such as low potassium, high sodium and flat rare earth element chondrite-normalized patterns with spatially associated gabbroic rocks. The hafnium isotopic compositions of zircon indicate a juvenile source for the magma. However, low zircon δ18O values suggest interaction with hydrothermal fluids. We propose that the El-Shadli plagiogranites were produced by extensive partial melting of juvenile, previously accreted oceanic crust and that this previously overlooked mechanism for the formation of plagiogranite is also responsible for the transformation of juvenile crust into a chemically stratified continental crust. The largest known plagiogranite on Earth, identified in the Arabian-Nubian shield, Egypt, formed through extensive partial melting of the previously accreted juvenile proto-continental crust, according to petrological and geochemical analyses.

Volumetrically, plagiogranites rarely exceed 1 vol. % of the oceanic crust, and range from dikes/veins a few millimeters to centimeters wide in MORs or ophiolites 4,5,8 , to large granite bodies several hundred meters to kilometers in dimension, such as the Wadi Suhaylay plagiogranite in the Oman ophiolite that has surface dimensions of 10 km × 8 km 6,7,14 . Until now, no plagiogranite larger than the Wadi Suhaylay plagiogranite has yet been reported. Here, we report the largest plagiogranitic complex (~5000 km 2 ) discovered to date, located in the Eastern Desert of Egypt in the Arabian-Nubian Shield (ANS). We investigate the petrogenesis of this large plagiogranitic complex based on whole-rock geochemical and Sr-Nd isotopic data, as well as zircon U-Pb-Hf-O isotope and trace-element data. We then discuss its petrogenesis in the context of Neoproterozoic global-scale tectonics, which was dominated by Rodinia break-up.
The ANS is one of the largest exposures of Neoproterozoic juvenile continental crust on Earth (Fig. 1a) 18 , consisting of numerous terranes with widespread ophiolites 19 . Most previous studies proposed that crustal growth in the ANS occurred predominantly through arc accretion 18,20 . Reymer and Schubert 21 and Stein and Goldstein 22 argued that the growth rate of the ANS significantly exceeded the rate of addition of mantle materials into the continental crust along modern subduction zones. They suggested that the large volume of magmas formed during the early stage of ANS crustal growth (within the Mozambique Ocean) may have been related to mantle plume magmatism (i.e., oceanic plateau). Gamal El Dien et al. 23 reported the occurrence of a voluminous ca. 700 Ma El-Shadli bimodal volcanic succession in the Eastern Desert region of the ANS, and speculated that a mantle plume may have provided sufficient heat to melt previously accreted oceanic terranes. In this work we adapt a similar plume model for the formation of the large plagiogranitic complex some 20-30 Myr prior to rift volcanism. We argue that contrary to conventional belief, the large plagiogranitic complex here formed through extensive re-melting of previously accreted oceanic crust driven by the emplacement of a mantle plume. Our work highlights the importance of mantle plumes in both the formation of exceptionally large plagiogranite intrusions, and in promoting crustal growth and consolidation in the ANS (and possibly continental crust in general), through post-accretionary re-melting of accreted oceanic crust.

Results
Geology of the El-Shadli plutonic complex. The study area in the southern Eastern Desert region of Egypt consists of six major basement rock assemblages (Figs. 1b and 2). These assemblages include the Wadi Ghadir ophiolitic complex (Wadi, W, the Arabic word for ephemeral watercourses where many exposures occur), the El-Shadli granite-diorite-gabbro plutonic complex, the El-Shadli bimodal volcanic rocks, the W. Hafafit core complex, as well as post-collisional alkalic granites and volcanic rocks (collision here refers to the late Neoproterozoic continental collision related to the final assembly of Gondwana 19 ) (Fig. 2).
The ca. 750 Ma W. Ghadir ophiolite assemblages (serpentinites, gabbroic rocks, sheeted dykes, and pillow basalts 24 ) are strongly deformed and variably metamorphosed 25 . They have a dominantly NW-trending foliation that is truncated by the El-Shadli granitoids 26 (Fig. 2), consistent with the intrusive relationship documented in the northern part of the study area and indicating that the granitoids formed post regional terrane accretion. These granitoids are part of the El-Shadli granitediorite-gabbro assemblage (the El-Shadli granitoids/plutonic complex) that crops out over much of the study area (Fig. 2). These granitoids have an almost continuous, arch-shaped map pattern in the central-south eastern part of the study region but are also exposed sporadically beneath the younger volcanic cover sequence (the El-Shadli bimodal volcanics), suggesting the possibility of a large and continuous plutonic complex underlying much of the region (Fig. 2). The estimated dimension of such a plutonic complex is ca.~100 km × 50 km.
The El-Shadli volcanic rocks are the largest bimodal suite exposed (80 km × 35 km), not only in the Eastern Desert of Egypt but also in the ANS 23,27 . They occur in a WNW-ESE trending belt flanked by granite-diorite-gabbro rocks (Fig. 2). The mafic endmember of the bimodal volcanic rocks consists of massive to pillow basaltic lava flows and dike swarms that intrude the granitoid rocks (Supplementary Fig. 1a). The felsic endmember consists of massive rhyolitic lava flows. The El-Shadli bimodal volcanic rocks had an imprecise Rb-Sr isochron age of~710 ± 24 Ma 27 . A recent study 23 provided a precise U-Pb zircon age of~700 Ma, and concluded that the El-Shadli bimodal volcanic rocks were generated in a post-accretionary riftrelated environment 27 , possibly induced by a mantle plume 23 .
The 680-630 Ma W. Hafafit Core complex is exposed in the NW part of the study area and is composed mainly of granitoids, gneisses, and amphibolites 28 (Fig. 2). The post-collisional alkaline granites (~640-550 Ma) are located in the eastern part of the study area and intrude both the El-Shadli plutonic complex and bimodal volcanics 29,30 (Fig. 2). Post-collisional alkaline volcanic rocks and their pyroclastic equivalents occur around W. Ranga (W. Ranga volcanics) in the eastern part of the study area (Fig. 2). This sequence comprises two slightly metamorphosed volcanic sequences: (a) intermediate-mafic lava flows (andesites, basaltic andesites, and basalts) in the north near Gabal (G) El-Sarobi and (b) felsic rocks (rhyolitic and dacitic lava flows) in the south. Previously undated, these volcanic rocks were interpreted to reflect either a post-collisional setting 30 or pre-collisional intra-oceanic arc assemblage 29 . In addition, previous work 29,30 inferred that the W. Ranga volcanic rocks were intruded by the El-Shadli granite-diorite-gabbro assemblage, although contact relationships are unclear in the field. However, these interpretations are not supported by a new zircon U/Pb age of 585 ± 10 Ma 31 , which indicates that the volcanic rocks are significantly younger than the El-Shadli granite-diorite-gabbro assemblage (see below).
Sample description and petrography. The El-Shadli granitoids occur as massive plutons. Within individual batholiths, 1 × 1 to 5 × 2 m gabbroic diorite enclaves represent over 50% of the outcrop area and provide evidence of extensive magmatic mixing zones (Fig. 3a, a-I and Supplementary Fig. 1b). The enclaves show no obvious chilled margins and have irregular, scalloped contacts with the granite (Fig. 3a-I). A similar magmatic mixing zone has been described in the Oman ophiolite plagiogranitic suite 6 . Smaller (centimeter-sized) ultramafic/mafic enclaves occur within the studied gabbroic diorite rocks and granites ( Fig. 3b-d).
Gabbroic samples S11-1 and S12 have a composition of SiO 2 = 45.   In general, the studied samples show a distinctive flat REE pattern with slight LREE depletion, similar to that of the overlying El-Shadli bimodal volcanics 23 (Fig. 5a). The El-Shadli granites (including diorite samples) show similar MORB-like REE patterns with 5-40 times enrichment relative to chondrite 33 Fig. 7 and Supplementary Data 1). The calculated ɛNd(t) values range from +5.55 to +8.18 ( Supplementary Fig. 7), similar to those of the El-Shadli bimodal volcanics 23 , with the corresponding T DM model ages 35,36 ranging from 0.78 to 1.2 Ga.
Zircon U-Pb-Hf-O isotopes and trace elements Diorite sample R10-1. Zircon grains from sample R10-1 have equant to sub-rounded morphologies and range from 100->200 μm in length ( Fig. 6a and Supplementary Fig. 8a). In cathodoluminescence (CL) images, some zircon grains display relatively subhedral to euhedral rims with clear concentric oscillatory zoning ( Fig. 6a and Supplementary Fig. 8a), but the majority have anhedral corroded rims with oscillatory, convolute, or convoluted/mottled zoning patterns ( Fig. 6a and Supplementary Fig. 8a). We carefully excluded from our analysis those grains with either mottled or convoluted zones. For sample R10-1, 14 U-Pb Sensitive High-Resolution Ion Microprobe (SHRIMP) analyses conducted on 14 grains yielded variable U (142-680 ppm) and Th (69-1310 ppm) contents and high Th/U ratios (0. 48-4.42) (Supplementary Data 2). The high Th/U ratios are typical of zircons crystallized from low SiO 2 melts (i.e., gabbro-diorite rocks) 37 . The U-Pb isotopic analyses yield a weighted mean 206 Pb/ 238 U age of 733 ± 7 Ma (±95% conf., mean square of weighted deviates [MSWD] = 0.76, N = 13; Fig. 6a), which is interpreted as the crystallization age of the gabbro-diorite complex. Eighteen laser ablation split stream (LASS) analyses on the same sample were conducted on 18 grains, including analyses over the same 14 spots utilized for SHRIMP analyses. The LASS analyses yield a similar 206 Pb/ 238 U age of 732 ± 5 Ma (±95% conf., MSWD = 1.4, N = 18; Supplementary Fig. 9a). Chondrite (CI)normalized 33 REE patterns of the zircons exhibit a negative slope  23 and Jabal Hamatah quadrangle 98 ). The rock units include the El-Shadli bimodal volcanic rocks, the El-Shadli plagiogranites-diorite-gabbro assemblage (this study), the W. Ghadir ophiolitic assemblage, the W. Hafafit core complex, post-collision alkaline granites, and post-collisional volcanics. Age data are from Kroner et al. 25 , Stern et al. 27 , Gamal El Dien et al. 23 , Kroner et al. 28 Supplementary Fig. 9a). Zircon grains have variable Hf contents (7843-10,390 ppm), Nb contents (0.59-2.46 ppm), and U/Yb ratios (0.06-0.23) (Fig. 7a, b, Supplementary Fig. 10a, and Supplementary Data 3). The Lu-Hf isotopic analyses of these 18 spots yield 176 Lu/ 177 Hf ratios of 0.004320 ± 0.000340 to 0.008358 ± 0.000099 and 176   with two-stage Hf model ages (T DM crustal) of 0.77-1.02 Ga. Although spot R10-1-15 has εHf (t) = 15.21 ± 1.16 and T DM = 0.66 Ga, the young T DM age of this spot compared to the crystallization age of the sample suggests recent Pb loss. As a result, the data from this spot were not included in the εHf (t) weighted mean calculations. The T DM crustal calculation assumes a 176  Trondhjemite sample R11-1. Zircon grains from sample R11-1 are commonly >200 μm in length and are equant to prismatic with euhedral rims. These grains show homogenous textures, some with well-developed oscillatory zoning in CL images, but others with faint and broad zoning ( Fig. 6b and Supplementary Fig. 8). Sixteen U-Pb SHRIMP analyses on 16 grains yielded variable U (135-1241 ppm) and Th (53-703 ppm) contents, and Th/U ratios of 0.37-0.66 (Supplementary Data 2), typical of zircons that crystallized in a high-SiO 2 magma 37 . The 16 analyses form a tight cluster that yields a weighted mean 206 Pb/ 238 U age of 733 ± 3 Ma (±95% conf., MSWD = 0.11, N = 16; Fig. 6b), interpreted to be the crystallization age of the Trondhjemite. Thirty-eight LASS analyses of 38 grains included 16 grains that were analyzed by SHRIMP, usually over the same spots. Taken together, these LASS analyses yield a 206 Pb/ 238 U age of 730 ± 3 Ma (±95% conf., MSWD = 0.87; Supplementary Fig. 9b and Supplementary Data 3) that is within analytical uncertainty of the SHRIMP age. CI-normalized 33 43 showing that the studied plagiogranites were likely derived from a low-K mafic magma source. d Tectonic setting plot 40 . Data sources: previously reported plagiogranites are as referenced in the text; arc granitoids (gray shade in (a)) are extracted from open access GeoRoc repository 101 ; Archean trondhjemite-tonalite-granodiorite (TTG: purple shades) 47 ; other ANS I-and A-type granitoids 44,45 ; the El-Shadli bimodal volcanics 23 .
Trondhjemite sample S01-1a. Zircon grains from sample S01-1a are euhedral with lengths <200 μm. Most grains exhibit welldeveloped concentric oscillatory zoning, although some have homogenous cores ( Fig. 6c and Supplementary Fig. 8). Twenty U-Pb SHRIMP analyses on 19 grains yielded relatively low U (28-153 ppm) and Th (13-257 ppm) contents, and Th/U ratios of 0.38-0.65 (Supplementary Data 2) that are indicative of a magmatic origin. Spot #S01-1-2 has a high common 206 Pb (6.9%), and this analysis is not included in the age calculation. The remaining 19 analyses yield a weighted mean 206 Pb/ 238 U age of 729 ± 7 Ma (±95% conf., MSWD = 0.37, N = 19; Fig. 6c), which is interpreted as the crystallization age of this Trondhjemite sample. Thirty LASS analyses were conducted on 30 grains, with 20 analyses from the same spots as the SHRIMP analyses, usually over the SHRIMP spot. These analyses yield a 206 Pb/ 238 U age of  Fig. 9c and Supplementary Data 3). The zircon grains have Hf contents ranging from 8182 to 9859 ppm, Nb from 0.73 to 3.06 ppm, and U/Yb ratios from 0.04 to 0.12 ( Fig. 7a, b, Supplementary Fig. 10a, and Supplementary Data 3). The Lu-Hf isotopes of these 30 spots yield 176 Lu/ 177 Hf ratios ranging from 0.001548 ± 0.00002 to 0.00668 ± 0.00020 and 176  Trondhjemite sample S14. Zircon grains from sample S14 are equant to sub-rounded, with euhedral rims. They range in length from 50 to 150 μm. Zircon grains commonly show oscillatory zoning in CL images, although some display homogenous cores ( Fig. 6d and Supplementary Fig. 8). Nineteen U-Pb SHRIMP analyses conducted on 19 grains yield U and Th contents of 48-937 and 20-1451 ppm, respectively, with Th/U ratios ranging from 0.31 to 1.55 (Supplementary Data 2). These values are consistent with a magmatic origin. Seventeen of the 19 U-Pb isotopic analyses yield a weighted mean 206 Pb/ 238 U age of 722 ± 7 Ma (±95% conf., MSWD = 0.41, N = 17; Fig. 6d), which is interpreted as the crystallization age of the sample. Nineteen LASS analyses located over the same 19 Fig. 9d and Supplementary Data 3). Zircon grains have Hf concentrations ranging from 8515 to 11,322 ppm, Nb from 0.80 to 9.37 ppm, and U/Yb ratios from 0.07 to 0.47 (Fig. 7a, b, Supplementary Fig. 10a, and Supplementary Data 3). Lu-Hf isotopic analyses of the same 19 spots give 176 Lu/ 177 Hf ratios of 0.000936 ± 0.000009 to 0.011387 ± 0.000083 and 176 Hf/ 177 Hf (t) ratios of 0.282577 ± 0.000041 to 0.282724 ± 0.000045 (Supplementary Data 3). Calculated εHf (t) values range from +9.39 ± 0.62 to +13.19 ± 1.08 and give a weighted mean of +11.38 ± 0.56 (MSWD = 3.3) (Fig. 7c, d, Supplementary Fig. 10b, 101 , Archean TTG (purple solid line) 47 , and the El-Shadli bimodal volcanics 23 . b Global plagiogranite database compared to that of MORB and arc granitoids. c, d Other ANS granitoids data 44,45 compared to that of global MORB and arc granitoids: I-type granitoids (c) and A-type granitoids (d).
Geochemically, the El-Shadli granitoids have very low K 2 O contents (<1 wt. %), typical of known plagiogranite rock suites worldwide (Fig. 4b), suggesting a low-K mafic source 43 (Fig. 4c). On discrimination diagrams of Rb vs. Y + Nb and Nb vs. Y 40 , the El-Shadli granites plot in the ORG (ocean ridge granite) field, which is typical of plagiogranites from ocean ridge settings ( Fig. 4d and Supplementary Fig. 12b). In addition to their high contents of Y (38-98 ppm), and low Nb (0.78-2.83 ppm), they are low in Rb (a fluid-mobile element) (2.06-13.45 ppm; average = 6.5 ppm) (Fig. 4d and Supplementary Fig. 12b) indicating derivation from a highly depleted source and no interaction with (or contamination by) continental crustal materials. Such geochemical features are typical of the MOR-and/or oceanic plume-related granites 40 (Fig. 4d,  Supplementary Fig. 12b, and Supplementary Data 1).
Other geochemical features typical of plagiogranites are the flat REE patterns 2,3,5,7,8,10 (Fig. 5a, b), low Sr (43-200 ppm), low Rb/ Sr (0.02-0.21), and low Sr/Y ratios (0.44-5.29) 2,3,5 . Their low abundance of LILE such as Cs, Rb, Ba, Th, and U also distinguishes them from arc-related granitoids and demonstrates a lack of interaction between the parental magma and subducting crustal and sedimentary materials 40 (Supplementary Fig. 6). These geochemical features therefore suggest that the El-Shadli granitoids have a different origin and formed in a different geodynamic setting from other coeval ANS granitoids 44 Zircon trace-element contents are a powerful tool for tracking the origin of granites and for distinguishing between the different granitic types 38 . Chondrite-normalized REE patterns of the El-Shadli zircon grains plot in the ocean-crust zircon field, as defined by zircons from oceanic crust plagiogranites and gabbroic rocks 38 ( Supplementary Fig. 9). Also, the low U/Yb and Gd/Yb of the El-Shadli zircon grains are similar to MOR plagiogranite zircons 38 (Fig. 7a, b and Supplementary Fig. 10a). Their δ 18 O values of 4.52 ± 0.14 to 4.83 ± 0.09‰ extend~1‰ below values typical of mantle zircon (5.3 ± 0.3‰) 46 and are similar to δ 18 O values of plagiogranites ( Fig. 7d and Supplementary Fig. 10c), which range from 3.9 to 5.6‰ (average = 4.9 ± 0.6‰) 7 .
( Supplementary Fig. 7) indicate the juvenile composition of a source rock that was itself extracted from a depleted MORB-like mantle source.
The Sr-Nd isotopic data of the El-Shadli plagiogranites and associated gabbros, as with the overlying El-Shadli bimodal volcanics 23 , are comparable to the nearby~750 Ma W. Gerf and W. Ghadir N-MORB mafic ophiolites 24,48 (Supplementary Fig. 7). Collectively, the geochemical and isotopic characteristics of the El-Shadli plagiogranites and associated gabbros suggest derivation from a MORB-like depleted source, similar to that of the El-Shadli bimodal volcanics 23 . These characteristics are all typical of plagiogranites that are generally interpreted to reflect juvenile and depleted sources 2,3,5,7,8,10,40 . The enrichment in some LILE (Cs, Rb, Ba, Th, and U) in the El-Shadli plagiogranites and associated gabbros relative to N-MORB are likely an inherited feature and could be related to hydrothermal alteration of their source in the oceanic crust 11,49 . The same enrichments in these LILE have been reported in the nearby~750 Ma W. Gerf and W. Ghadir N-MORB mafic ophiolites 24,48 .
The juvenile El-Shadli plagiogranite lithogeochemistry, together with the absence of pre-Neoproterozoic zircon inherited cores or xenocrysts in any of the samples ( Fig. 6 and Supplementary Fig. 9), implies that there is no evidence for interaction with old crustal materials 7 . The low U/Yb ratios (mostly <0.1) of the El-Shadli plagiogranite zircons are similar to modern ocean-crust zircons from a depleted source with MORB-like composition and contrasts with continental and arc zircons 38 (Fig. 7a, b and Supplementary Fig. 10a). The low Nb/Yb and Gd/Yb ratios of the El-Shadli plagiogranite zircons are also well-defined features of zircon extracted from a depleted MORB-like source 23,38 (Fig. 7b and Supplementary Fig. 10a). The high positive εHf (t) values (weighted mean from +10.82 ± 0.17 to +11.56 ± 0.51) of the El-Shadli plagiogranite zircons ( Fig. 7c and Supplementary Fig. 10b) also indicate a juvenile source with no measurable contribution from old continental crust 50,51 . These data indicate the El-Shadli plagiogranites were derived from a highly depleted source and indicate a different mode of origin from that of the coeval ANS granitoids 44 (Fig. 7c). It is worth noting that the felsic endmember (sample S02-5) of the overlying~700 Ma El-Shadli bimodal volcanics shares the same zircon isotopic and trace-element characteristics as the plagiogranite, suggesting derivation from a similarly depleted MORB-like source 23 (Fig. 7).
The uniform δ 18 O values (weighted mean from 4.52 ± 0.14 to 4.83 ± 0.09‰) of the El-Shadli plagiogranite zircons are lower than that of typical mantle values (5.3 ± 0.3‰) 46 but are consistent either with a source that has previously undergone interaction with high-temperature hydrothermal fluids or with sub-solidus alteration of plagiogranite by post-magmatic fluidrock interaction 7,9 ( Fig. 7d and Supplementary Fig. 10c). The former possibility is favored for several reasons. First, petrographic observations and the low LOI indicate the plagiogranites are very fresh. Second, zircon resists post-magmatic modification, and zircon within the plagiogranite preserves primary magmatic signatures such as crystal shapes with oscillatory zoning (Fig. 6 and Supplementary Fig. 8), high Th/U (mostly >0.4) 52 , and U contents mostly <1000 ppm 37,52 (Supplementary Fig. 11 and Supplementary Data 1). In addition, there is no correlation between Th/U and U vs. δ 18 O ( Supplementary Fig. 11), and our U-Pb data yield concordant to nearly concordant ages ( Fig. 6 and  17 based on experimental work to evaluate the composition of plagiogranitic melts resulting from fractional crystallization of MORB 55 and partial melting of hydrated oceanic crust (i.e., gabbros, amphibolites, and basalts) 15,17,54 . b TiO 2 (wt%) vs. SiO 2 (wt%) diagram 11 showing the lower limit for plagiogranitic melts resulting from fractional crystallization of MORB. c TiO 2 WR (wt%) vs. δ 18 Ozr (‰) diagram 7 showing the field of plagiogranites produced by fractional crystallization of MORB. Error bars represent 2 sigma. d La (ppm) vs. SiO 2 (wt%) diagram was modified after Brophy 16 , suggesting a partial melting origin for the studied rocks. Global plagiogranite database is as referenced in the text and data for the El-Shadli bimodal volcanics are that of Gamal El Dien et al. 23 . Supplementary Fig. 9). All these features indicate that the δ 18 O compositions may be attributed to hydrothermal fluid contamination of their source rock 7,9 , rather than to post-magmatic alteration processes.
In summary, the whole-rock petrological, geochemical, and isotopic data and zircon U-Pb-Hf-O-trace-element data suggest that the El-Shadli plagiogranites were derived from a MORB-like, highly depleted mafic source that had previously undergone hydrothermal alteration in an oceanic-and/or ridge/rift-related setting.
Experiments show that the composition of plagiogranitic melt resulting from fractional crystallization of MORB 55 differs markedly from that of plagiogranite melt generated by partial melting of gabbros and amphibolites from the lower oceanic crust and oceanic crustal basalts 15,17,54,56 , particularly with regard to TiO 2 , SiO 2, and K 2 O contents (Fig. 8a). The El-Shadli plagiogranites have a compositional range that overlaps with melts produced by partial melting of hydrated gabbroic and amphibolitic rocks of the lower oceanic crust (Fig. 8a). The TiO 2 content of plagiogranitic melts derived by fractionation from tholeiitic magmas is controlled by source composition and redox conditions 11,55 . Experiments on tholeiitic primitive MORB melts 55 show that Fe-Ti oxides are generally stable under oxidizing conditions, and so SiO 2 -enriched felsic melts have significantly lower Ti contents. On the other hand, under reducing conditions, the felsic melts produced by fractionation are characterized by high TiO 2 and FeO contents. As the differentiation of MORB melts typically occurs under more reducing conditions, the resultant felsic melts should have high Ti contents 11,55 . In contrast, gabbros in the lower oceanic crust are highly depleted in Ti, and partial melting of this source generally occurs under hydrous (i.e., oxidizing) conditions 15 . The resulting felsic melts should, therefore, have very low Ti contents 15 . Thus, Koepke et al. 11,15 proposed that the TiO 2 content of plagiogranites is a powerful indicator for discriminating between the more felsic plagiogranitic melts generated from anatexis of oceanic lower crustal gabbroic rocks (TiO 2 < 1 wt. %) and those generated by fractionation of MORB melts (TiO 2 > 1 wt. %). The El-Shadli trondhjemitic and tonalitic rocks have TiO 2 concentrations below the lower limit for experimental melts produced during MORB fractionation, again supporting an origin involving partial melting of mafic (gabbroic) crust (Fig. 8b, c).
Based on theoretical modeling, Brophy 16 proposed that plagiogranitic melts formed by hydrous partial melting and exhibit either a flat or slightly decreasing REE with increasing SiO 2 content. Their REE patterns can be slightly higher, and/or overlap with, co-magmatic mafic rocks. On the other hand, plagiogranitic melts produced by fractional crystallization show a positive correlation between REE and SiO 2 contents 16 . The El-Shadli plagiogranites and associated gabbroic rocks have similar and overlapping flat REE patterns (Fig. 5a) and lack a positive correlation between La and SiO 2 (Fig. 8d), features typical of melts formed by hydrous partial melting of lower oceanic crust 16 . This process contrasts with that of the Iceland lavas, which display a positive correlation between La with SiO 2 from mafic to felsic endmembers and are thought to have been formed by fractional crystallization 16 .
Koepke et al. 11,15 proposed that hydrous partial melting of preexisting gabbros produces a plagiogranitic melt with amphibole as residual phase (Olivine + clinopyroxene + plagioclase (A) + H 2 O = amphibole + orthopyroxene + plagioclase (B) + plagiogranitic melt). As REE abundances (particular LREE to MREE) and HFS elements such as Nb are very sensitive to the presence of amphiboles in the residue, the resultant melt should be depleted in LREE and Nb 6,11,15,54 . This is indeed the case for the El-Shadli plagiogranites ( Fig. 8d and Supplementary Fig. 6). The lack of HREE depletion (Fig. 5a) indicates the absence of residual garnet, implying partial melting occurred in a low-pressure environment (i.e., in plagioclase stability field).
Zircon δ 18 O values can also distinguish between plagiogranitic rocks generated by fractional crystallization (typical mantle-like δ 18 O values of 5.2 ± 0.5‰) 7 and those generated by re-melting of hydrothermally altered gabbroic lower oceanic crust (δ 18 O = 4.9 ± 0.6‰) 7 . The El-Shadli zircons have low δ 18 O values ranging between 4.52 and 4.83‰ (Fig. 7d and Supplementary Fig. 10c). There are two possible ways to cause such low δ 18 O values: one is by melting hydrothermally altered source rocks, and the other is by assimilating hydrothermally altered wall rock during magma intrusion 7,9 . However, a plagiogranitic magma contaminated by wall-rock assimilation would be expected to contain zircon xenocrysts and Hf-O isotopic data comparable to that upper crustal rocks 9 , but neither of these features have been observed in the studied rocks (Fig. 7d). Also, normal oceanic crust is characterized by δ 18 O values as low as 2‰ for the gabbroic lower crust due to interaction of the mafic magma with hydrothermal fluids 7,14 . The systematic low δ 18 O values are typical of global ophiolitic plagiogranites, indicating that these values are primary signatures and reflect partial melting of the lower oceanic crust without overprinting by late, lowtemperature fluids 7 . The above observations support the formation of the El-Shadli plagiogranites by re-melting of altered gabbroic lower oceanic crust. On the plot of TiO 2 (WR) vs. δ 18 Ozr (Fig. 8c), the El-Shadli plagiogranites also plot outside the field characterizing plagiogranites with a fractional crystallization origin but overlap with plagiogranites with a known partial melting origin from the Oman and Troodos ophiolites 7,57 .
It has been demonstrated that high-temperature hydrothermal fluids can penetrate to upper mantle depths, and are thus capable of interacting with the oceanic gabbroic lower crust 49 . The circulation of such hydrothermal fluids has been recognized by their seawater signatures such as high radiogenic 87 Sr/ 86 Sr values and high Cl and B contents in basaltic glass from MORB and secondary amphibole in the gabbroic section of ophiolites 11,15,17,49,54 . The penetration of such hydrothermal fluids is facilitated by high-temperature shear zones/ detachment fault systems in the oceanic crust 15,49 .
Field observation of irregularly shaped enclaves of gabbroic blocks within the plagiogranite indicate the presence of magmatic mingling (Fig. 3a, a-I and Supplementary Fig. 1b). Such contact relationships are typical of mingling between two contemporaneous magmas 6 . Their interpreted coeval origin is further supported by the overlapping ages for the two rock types (Fig. 6 and Supplementary Fig. 9). The El-Shadli gabbroic rocks exhibit an evolved signature, characterized by low Mg# [MgO/(MgO + FeO T )*100 < 60], Cr (<500 ppm), and Ni (<100 ppm) (Supplementary Data 1). Such low values are significantly lower than that of primary mantle melts (Ni > 500 ppm, Cr > 1000 ppm, and Mg# > 72) 58 and suggest these gabbroic rocks are not primary mantle melts. These characteristics, together with the negative correlation of SiO 2 with MgO and CaO ( Supplementary Fig. 4), indicate that the source rock of the mafic magmas underwent olivine and clinopyroxene fractionation. Thus, the magma compositions of the gabbroic rocks originated by extensive partial melting of a depleted, juvenile mafic source.
Collectively, whole-rock petrology, geochemical and isotopic data, and zircon Hf-O isotopes support partial melting of the hydrated gabbroic lower oceanic crust as the origin for the El-Shadli plagiogranites. Some "arc-like" trace-element patterns, such as Nb and Ti negative anomalies, and LILE enrichment, can be interpreted as the result of (1) hydrothermal fluid contamination of the source rocks and/or (2) a two-stage process (i.e., generation followed by re-melting of highly depleted source) for the production of the plagiogranitic melt.
Geodynamic context and formation mechanism for the El-Shadli plutonic complex. The geochemical and isotopic data distinguish the El-Shadli plagiogranites from other arc-related granitoids (Figs 4, 5, and 7 and Supplementary Figs 6, 10, and 12) and suggest their formation in a ridge/within-plate tectonic environment ( Fig. 4d and Supplementary Fig. 12b). The El-Shadli plutonic complex (plagiogranites and associated gabbros) intruded pre-existing poly-deformed and variably metamorphosed ophiolitic complexes, as evidenced by cross-cutting relationships with strongly foliated ophiolitic country rocks (Fig. 2). The lack of significant deformation or metamorphism in the El-Shadli plutonic complex is consistent with its emplacement in a withinplate tectonic setting. In addition, the El-Shadli plutonic complex is overlain by rift-related, within-plate, and slightly younger (~700 Ma) bimodal volcanic rocks 23,27 (Fig. 2). These relationships further support a post-accretionary, within-plate/rifting tectonic setting for the El-Shadli plutonic complex. Geochemical and isotopic data for the El-Shadli plagiogranites suggest an origin involving the anatexis of hydrated gabbroic lower oceanic crust. Their highly depleted and juvenile composition indicates that the magma was likely derived from partial melting of a juvenile mafic proto-continental crust (i.e., accreted oceanic crust) along a continental margin (Fig. 9b, c). This juvenile mafic proto-continental crust was accreted along the flanks of an ocean (i.e., the Mozambique Ocean) that bordered the Rodinia supercontinent (Fig. 9b, c). Such accreted mafic juvenile crust was likely thinner and denser than normal continental crust and was thus mostly submerged (Fig. 9b, c), thereby explaining the submarine setting of the slightly younger (~700 Ma) bimodal volcanics and the formation of pillow basalts 23,27 . The El-Shadli plagiogranites were formed by melting of an oceanic crust that was accreted 20-10 Myr earlier (i.e., by melting of older ophiolitic complexes), unlike other plagiogranites found in ophiolites and MOR settings worldwide, in which plagiogranites formed contemporaneously with MOR magmatism 2,4,57 .
It has been suggested that the volume of plagiogranites reflects the formation process (fractional crystallization vs. partial melting) 15 . Fractional crystallization of basaltic magma normally produces small volumes of plagiogranites in MOR environments 4,7,15 . On the other hand, larger volumes of plagiogranitic melt can be generated by partial melting processes and are more likely to occur in ophiolites 7,15 . Such partial melting processes are made possible by high H 2 O activity and possibly with an excess heat source in MOR environments. The H 2 O could be provided through penetration of hydrothermal fluids from the ocean via high-temperature detachment fault systems/shear zones along the MOR 15,49 (Fig. 9a). The excess heat source is generally believed to have been provided by underlying melt lenses beneath the MOR 6,15,54 , but Amri et al. 5 suggested an upwelling mantle diapir as a heat source for the formation of the largest plagiogranites (up to 8 km in outcrop dimension) within the Oman ophiolites. As the El-Shadli plagiogranites formed after terrane accretion instead of in a MOR environment, and the 730-700 Ma magmatism is coeval with global continental rifting and the break-up of Rodinia (~825-680 Ma) 59,60 , the heat source for their formation could have been related to emplacement of a mantle plume during Rodinia break-up 59,60 (Fig. 9c), as first speculated by Gamal El Dien et al. 23 for the formation of thẽ 700 Ma rift volcanics found in the same region. Given that the El-Shadli plagiogranites formed in a post-accretionary within-plate setting, the mantle upwelling could also have been induced by slab break-off 61 . However, the composition of the El-Shadli plutonic complex is inconsistent with granitoid magmas produced by slab break-off 62 (Fig. 4d). Thus, a plume-induced model 23 is preferred based on evidence discussed below.
First, the exceptional large volume of the El-Shadli plagiogranite (a lateral dimension of ca.~100 km × 50 km) can best be explained by mantle plume-induced large-scale re-melting the accreted mafic lower crust that generated both the super-large El-Shadli plagiogranites and the overlying bimodal volcanics (80 km × 35 km with a thickness of >10 km) 23,27 . Second, there is evidence for mantle plume activity of that age in the region, such as the layered mafic-ultramafic intrusions at Korab Kansi (741 ± 24 Ma) 63 and at G. Dahanib (710 ± 7 Ma) 64 (Fig. 1b). Third, the hiatus between the 730-720 Ma plagiogranite and the overlying 700 Ma bimodal rift volcanics (Fig. 9d) can be interpreted as plume-induced regional doming 23,65 .
As plumes tend to preferentially nucleate along the margins of a superplume 66 , a plume heat source is in accord with the proposed paleogeographic position of the ANS over the edge of the Rodinia superplume during the break-up/continental rifting of the supercontinent Rodinia between 825 and 680 Ma 59,60 . Coeval plume-related magmatism along the margins of Rodinia including granitoids, mafic-ultramafic dikes, and rift-related bimodal volcanics found in South China, Australia, Southern Africa, and Laurentia 60,67,68 . The proposed plume-induced model for the El-Shadli plutonic complex (~730-720 Ma) is coeval with the global plume/rift-related magmatism such as the well-known Franklin large igneous provinces event (~727-720 Ma) in Laurentia 60,68 that extend until~712 Ma 69 , and the Gannakouriep event (~720 Ma) in South Africa and Namibia 67 .
In summary, we propose that the exceptionally large El-Shadli plagiogranite was formed by partial melting of previously accreted oceanic crust (juvenile proto-continental crust) above a mantle plume, instead of the previously well-known formation mechanisms in MOR environments. The preferred tectonic model (Fig. 9) also has implications for the crustal growth of the ANS. Most previous models focus on arc accretion 18,20 , but mantle plume and rift-related-magmatism have also been suggested to have played an important role in crustal evolution 21,22 . However, previous studies 21, 22 advocating plume involvement only inferred that it occurred within an oceanic environment (i.e., the formation of oceanic plateaus) before accretion and that the products of plume activity were subsequently accreted with island arcs to form a juvenile continental crust. The model presented herein, modified from Gamal El Dien et al. 23 , suggests that plumes may have also been important in the post-accretionary evolution of the ANS. Our model emphasizes the importance of mantle plume emplacement in the transformation of newly accreted ANS Neoproterozoic juvenile crust into a coherent normal continental crust by promoting intensive crustal re-melting, and possibly by the addition of new melts from the mantle into the crust. As the ANS represents one of the largest post-Archaean juvenile continental crustal growth events on Earth, our new mechanism for the formation of plagiogranite and the transformation of accreted juvenile terranes into consolidated and stratified normal continental crust could potentially serve as an analog to the early crust formation on early Earth.
Thin sections of the rock samples were cut and polished with progressively finer grades of diamond paste (9-1 µm thick) by service provider Yu'neng Petrology and Mineral Service Company, China. The relative abundance of minerals (both transparent and opaque) and their textural relationships were determined on a Nikon Eclipse optical microscope at Curtin University using transmitted and reflected light. Also, whole thin section images were collected for some samples using a Zeiss Axio Imager M2m Imaging System at Curtin University, at 5× magnification under transmitted and reflected light.
Phase mineral distribution maps for samples R10-1, R11-1, S01-1a, S14, S23-1, S23-2, S24, S25-1, and S25-2 were acquired using a TESCAN Integrated Mineral Analyzer (TIMA) housed at the John de Laeter Centre, Curtin University ( Supplementary Fig. 2). The analysis was performed on carbon-coated whole thin sections using Liberation analysis in "Dot Mapping" mode adopting 3 μm dot spacing for backscattered electron images and 27 μm for X-ray acquisition operating at 25 kV acceleration voltage, 15 mm working distance, and a magnification of 185 times. Data processing (include calculation of modal composition) was performed with the TESCAN TIMA version 1.6.71 software.
Whole-rock major and trace-element geochemistry. Rock samples were examined first by optical microscopy. Selected rock samples were crushed with a polyethylene-wrapped hammer into <0.5 cm small chips, ultrasonically cleaned in distilled water, and subsequently dried and handpicked to avoid altered pieces and visible contamination. The samples were then grounded with ethyl alcohol in an agate ring mill to grain size <50 μm, and the resulting powders were used for the analyses of major and trace elements, and Sr-Nd isotopes (Supplementary Data 1).
Whole-rock major element oxides were determined by XRF at Bureau Veritas Lab, Perth, Western Australia. Each whole-rock sample was pulverized in a vibrating disc pulverizer. The sample powder was ignited at 1000°C for 2 h to determine the LOI using a robotic TGA system. Subsequently, the sample was cast using a 66:34 flux with 4% lithium nitrate to form a glass pill (fuse bead) by melting at 1080°C. The major oxides (Al 2 O 3 , Cr 2 O 3 , Fe 2 O 3 , K 2 O, MgO, MnO, Na 2 O, P 2 O 5 , SiO 2 , and TiO 2 ) were determined using XRF on the oven-dried (105°C) fuse beads.
The trace and REE of the samples were analyzed using an Agilent 7500 quadrupole ICPMS at Macquarie (MQ) GeoAnalytical lab, Macquarie University, Australia. Acid digestion with hydrofluoric acid (HF) was used to digest geological materials for trace-element determination. Powdered samples of 100 mg each were weighed into 15 mL Savillex Teflon beakers. The samples were first refluxed in 1.5 mL conc. HF (Merck Suprapur grade) +1.5 mL of Teflon distilled HNO 3 at 130°C overnight, then ultrasonicated and dried. This process was repeated and then ten drops of conc. HClO 4 (Merck Suprapur) and 2 mL of conc. HF (Merck Suprapur) was added and evaporated to incipient dryness at 170°C before adding five drops of conc. HClO 4 . The sample beakers were tightly capped and placed on a hot plate for 24 h at 190°C, then cooled to room temperature. Samples were finally refluxed in 2 mL of 6 N distilled HCl at 150°C for 3 h and dried down at 190°C, and then reflux again in 2 mL of 6 N distilled HNO 3 at 150°C for 3 h. For traceelement analyses, the samples were diluted in 100 mL of 2% HNO 3 + 0.5% HF, and heated at 80°C until a clear solution resulted. Reference materials BCR-2 (basalt), BHVO-2 (basalt), BIR-1 (basalt), and GSP-2 (granodiorite) were analyzed at the beginning and end of each analytical session, and their measured values agree with recommended GeoRem (http://georem.mpch-mainz.gwdg.de) and USGS values.
The precision of the measurements by repeated analyses of reference samples is better than ±5% for trace elements and REEs.
Whole-rock Sr-Nd isotopes. Sr and Nd isotopic ratios were obtained by thermal ionization mass spectrometry (TIMS) on a Thermo Finnigan Triton system at Macquarie (MQ) GeoAnalytical Lab, Macquarie University. Analytical procedures including cleaning and column preparation, sample digestion, chromatographic separation to collect Sr and Nd, and sample loading on Re filament and data collection from TIMS are detailed in Tilhac et al. 71 . BHVO-2 reference material was first measured in each analytical session to check instrument status and sensitivity for Sr and Nd and was followed by unknown samples. Ratios were normalized to 86  Zircon mount preparation. Zircon grains were separated from crushed rock samples (1-5 kg/each) using standard methods (Frantz magnetic separator and heavy liquids) at Yu'neng Petrology and Mineral Service Company, China. The best-quality grains, characterized by homogenous transparency, color, and fluorescence, were handpicked and mounted in 24 mm diameter epoxy resin discs and polished to expose the interior of the grains. Zircon grains from each sample were mounted along with zircon standards for age calibration (Plešovice) and O isotopes (Penglai and Qinghu) in Guangzhou Institute of Geochemistry, Chinese Academy of Sciences (GIGCAS). The zircon crystals were characterized by CL imaging using a Tescan MIRA3 scanning electron microscope in the Microscopy and Microanalysis Facility, John de Laeter Centre, Curtin University using 12 kV, spot size 6 μm, and working distance~15 mm ( Supplementary Fig. 8). In the CL images, high-U regions appear as dark regions, and low U regions appear as bright regions. Analytical spots, selected through CL imaging, were transparent, smooth, and without obvious inclusions. The polished mounts were cleaned in high-purity ethanol using an ultrasonic bath, and then vacuum-coated with high-purity gold before the SHRIMP and SIMS analyses.
SHRIMP U-Pb zircon analyses. Zircon U-Pb isotope measurements were done on the SHRIMP II instrument at the John de Laeter Centre, Curtin University (Supplementary Data 2). Curtin ion-microprobe SHRIMP II analytical procedures broadly followed those described by Compston et al. 72 and Williams 73 . A 25-30 μm diameter spot size was used for all the analyzed grains with a primary beam current of 2.5-3.0 nA, mass resolution of~5000, and analysis sensitivity of~18 cps ppm −1 nA −1 . Data for each spot were collected in sets of six scans through the mass range of 196 Zr 2 O + , 204 Pb + , background, 206 Pb + , 207 Pb + , 208 Pb + , 238 U + , 248 ThO + , and 254 UO + . The measured isotopic ratios were corrected for common Pb based on the measured 204 Pb and using the two stages common Pb evolution curve of Stacey and Kramers 74 to get the common Pb compositions. Multiple standard analyses were interspersed between analyses during each session (once every four unknowns). Pb/U ages were normalized to a recommended value of 337.1 Ma determined by conventional U-Pb analysis of Plešovice zircon standard 75 . The measurement of 206 Pb/ 238 U and 207 Pb/ 235 U ages requires normalization to results of the standard analyses, and the reported results were corrected for the uncertainties associated with the measurements of the Plešovice standard. The correction formula for Pb/U fractionation is 206 Pb + / 238 U + = a( 238 U 16 O + / 238 U + ) b76 using the parameter values of Black et al. 77 All isotopic measurements were reduced, processed, and interpreted using SQUID II, Isoplot, and Isoplot R programs [78][79][80] . The errors associated with individual analyses are 1σ uncertainties and include errors from U-Pb calibration based on the reproducibility of U-Pb standard measurements, counting statistics, and the common Pb correction. The weighted mean ages are quoted at the 95% confidence interval (2 SD). The ages quoted through the text are 206 Pb/ 238 U ages that are more precise than 207 Pb/ 206 Pb ages for Neoproterozoic zircons due to the lower content of 207 Pb in these zircons 81 .
Laser ablation split stream inductively coupled plasma mass spectrometry (LASS-ICPMS) zircon U-Pb, Lu-Hf isotopes, and trace-elements analyses. Zircon U-Pb, Lu-Hf isotopes, and trace-element abundances were measured using LASS inductively coupled plasma mass spectrometry (LASS-ICPMS) in the Geo-History Facility at the John de Laeter Centre, Curtin University (Supplementary Data 3). The analyses were conducted on the same grain as the SHRIMP analysis, commonly over the same spots. Additional LASS-ICPMS spots were targeted based on CL images. The measured U-Pb and Lu-Hf isotopes and trace-element data were collected simultaneously using an excimer laser (Resonetics S-155-LR 193 εHf ðTÞ ¼ ½ 176  where, f C , f Z , and f DM are the f Lu/Hf values of the continental crust, zircon sample, and the depleted mantle, respectively. Z = analyzed zircon, CHUR = chondritic uniform reservoir; DM = depleted mantle, T DM C = Two-stage Hf model ages. SIMS zircon O-isotope analyses. Zircon O isotopic compositions were measured using a Cameca IMS 1280-HR at the high-precision SIMS Laboratory of the GIGCAS, following the analytical procedures of Yang et al. 93 (Supplementary  Data 3). The Cs + primary ion beam was used to sputter oxygen ions from unknown zircon samples that accelerated at 10 kV and~2 nA in intensity, with the mass resolution being~2500. The target area was~20 μm in diameter that includes a 10 μm spot diameter and a 10 μm raster diameter. 16 O and 18 O isotopes were measured in multi-collector mode using two off-axis Faraday cups. Qinghu and Penglai reference zircons 94,95 have been used to evaluate the precision (reproducibility) and accuracy. Penglai zircons were used as external reference material to calibrate the instrumental mass fractionation (IMF). Qinghu zircons were treated as an unknown sample, which yielded a precision of 5.45 ± 0.03‰ (2 SD; MSWD = 0.58) that is consistent with the recommended value of 5.4 ± 0.2‰ (2 SD) 94 . During each session, one Penglai zircon analysis was conducted with every five unknown spots (including a Qinghu zircon standard that was treated as an unknown). The measured 18