Detrital zircon provenance of north Gondwana Palaeozoic sandstones from Saudi Arabia

Abstract We present the first comprehensive detrital zircon U–Pb age dataset from Palaeozoic sandstones of Saudi Arabia, which provides new insights into the erosion history of the East African Orogen and sediment recycling in northern Gondwana. Five main age populations are present in varying amounts in the zircon age spectra, with age peaks at 625 Ma, 775 Ma, 980 Ma, 1840 Ma and 2480 Ma. Mainly igneous rocks of the Arabian–Nubian Shield are suggested to be the most prominent sources for the Ediacaran to middle Tonian zircon grains. Palaeoproterozoic and Archaean grains may be xenocrystic zircons or they have been recycled from older terrigenous sediment. A primary derivation from Palaeoproterozoic and Archaean basement is also possible, as rocks of such age occur in the vicinity. Approximately 4 % of the detrital zircons show Palaeozoic (340–541 Ma) ages. These grains are likely derived from Palaeozoic post-orogenic and anorogenic igneous rocks of NE Africa and Arabia. A few single grains gave up to Eoarchaean (3.6–4.0 Ga) ages, which are the oldest zircons yet described from Arabia and its vicinity. Their origin, however, is yet unknown. Detrital zircons with U–Pb ages of 1.0 Ga are present in varying amounts in all of the samples and are a feature of terrigenous sediment belonging to the Gondwana super-fan system with an East African – Arabian zircon province.


Introduction
The Gondwana supercontinent formed by the closure of the Mozambique Ocean and subsequent amalgamation of East and West Gondwana during the Pan-African Orogeny at the end of the Neoproterozoic Era (e.g. Stern, 1994;Johnson et al. 2011). This led to the formation of a~6000 km long accretionary orogen within Gondwana that extends from Arabia to East Africa and into Antarctica and is known as the East African Orogen (e.g. Stern, 1994;Johnson et al. 2011;Fritz et al. 2013), also referred to as the Transgondwanan supermountain (Squire et al. 2006). In NE Africa, the Pan-African Orogeny involved the collision of Archaean cratons and the Saharan Metacraton with the Arabian-Nubian Shield (Fig. 1a). The latter evolved between~565 and 870 Ma and primarily comprises Neoproterozoic juvenile oceanic island arcs (e.g. Stern, 1994;Johnson et al. 2011;Johnson, 2014).
After the assembly of Gondwana during the early Palaeozoic, a thick pile of quartz-rich sandstones was deposited across northern Gondwana (Fig. 1b), originally eroded from the Pan-African orogens (e.g. Garfunkel, 2002;Burke et al. 2003;Avigad et al. 2005;Meinhold et al. 2013). The high mineral maturity of the sediment can be attributed to either long transport distances and/or multi-stage sedimentary recycling (e.g. Garfunkel, 2002;Morag et al. 2011) or strong chemical weathering at the time of deposition (e.g. Avigad et al. 2005). The similarity of detrital zircon U-Pb age spectra in Palaeozoic sandstones throughout Gondwana let Squire et al. (2006) postulate the existence of large sediment fans that brought detritus from the East African Orogen towards the continental margins. Meinhold et al. (2013) extended the super-fan model to the northern Gondwana margin evaluating detrital zircon age spectra from Cambrian-Ordovician sandstones of North Africa, the Sinai Peninsula and the northwesternmost part of the Arabian Peninsula. Stephan et al. (2019a) used a more extensive dataset, including also peri-Gondwana units, and evaluated the detrital zircon age spectra, using multivariate data analysis. Detrital zircon data from Lower Palaeozoic sedimentary rocks revealed three contrasting provenance end-members of the former Gondwanan shelf, namely the Avalonian, West African and East African -Arabian zircon provinces (Stephan et al. 2019a, b). The East African -Arabian zircon province corresponds to the Gondwana super-fan system of Meinhold et al. (2013). Saudi Arabia would lie more distal along the sediment path to the NE. However, the Gondwana super-fan model has not been tested yet for Saudi Arabia. Besides that, a thorough provenance analysis of detrital zircon ages from the Palaeozoic succession of Saudi Arabia does not yet exist.
In the present study, we analysed 24 sandstone samples from the Palaeozoic succession of Saudi Arabia (Fig. 2) to determine their detrital zircon U-Pb age spectra using laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS). The data are essential for refining the model of the Gondwana super-fan system of Meinhold et al. (2013) at a more distal location and thus for refining the East African -Arabian zircon province of Stephan et al. (2019a, b). In addition, they provide complementary provenance information for the Palaeozoic sandstones on the Arabian Peninsula. A more thorough understanding of the origin of these sedimentary units is important for reconstructions of palaeosource areas and sediment transport and gives novel insights into the erosion history of the East African Orogen. Besides that, the recognition of single zircon grains with up to Eoarchaean (3.6-4.0 Ga) ages in Palaeozoic sandstones of Saudi Arabia contributes further to the discussion whether hitherto unknown Early Archaean basement can be found in the Middle East (see Paquette & Le Pennec, 2012, 2013Nutman et al. 2014) or whether such Early Archaean grains are rather recycled from old, distal sources.

Geological setting
In Saudi Arabia, the Palaeozoic succession crops out in a narrow band along the southern, eastern and northern margin of the Arabian Shield (Fig. 2a), which forms the eastern part of the Arabian-Nubian Shield (Fig. 1). The latter is a complex amalgamation of numerous intra-oceanic island arc terranes accreted between~630 and 870 Ma, with igneous activity extending until 565 Ma (Kröner, 1985;Stern, 1994;Johnson et al. 2011;Johnson, 2014). During the late Neoproterozoic, terrestrial sedimentary basins formed in which molasse was deposited under subaerial to shallow-water conditions, and variably interlayered with volcaniclastic rocks (Johnson et al. 2011. The Neoproterozoic rocks of the Arabian Shield are overlain with an erosional contact by Palaeozoic sandstones (Fig. 3a). Deposition took place from Cambrian to Permian times (Powers et al. 1966;Powers, 1968;Laboun, 2010) (Fig. 3a). Sedimentary environments varied, mainly ranging from fluvial braided stream setting to shallow marine, prodeltaic and open marine conditions (Dabbagh & Rogers, 1983;Stump & van der Eem, 1995;Al-Ajmi et al. 2015). The sediments were deposited on a stable continental shelf at a passive margin from the Cambrian until at least the Devonian (Stump & van der Eem, 1995;Sharland et al. 2001). The changes in accommodation space were primarily controlled by eustacy (Sharland et al. 2001). Low subsidence rates and the high stability of the Arabian Plate during most of the Palaeozoic resulted in a 'layer cake' stratigraphy (Bishop, 1995), although several erosional hiatuses and stratigraphic unconformities exist in the Palaeozoic succession (Stump & van der Eem, 1995). The erosional hiatuses separating sedimentary units are used for lithostratigraphic correlations (Stump & van der Eem, 1995;Laboun, 2010) (Fig. 3a). The Palaeozoic succession shows evidence for two Gondwana glaciations during the Late Ordovician (Hirnantian) (Clark-Lowes, 2005;Keller et al. 2011) andCarboniferous-Permian (McClure, 1980;Alsharhan & Mohammed, 1993;Keller et al. 2011). We refer to Keller et al. (2011), Al-Ajmi et al. (2015 and Bassis et al. (2016a) for details about sedimentary facies and depositional environment.
Siliciclastic sedimentary rocks, mainly quartz arenites, dominate the Lower Palaeozoic succession of Saudi Arabia (e.g. Fig. 3b;  Pollastro et al. 1998). For simplification, the Tethyan ophiolite complexes in Oman have been omitted. IS -Israel, JO -Jordan, KU -Kuwait, UAE -United Arab Emirates. (b) Geologic map of the southern Saudi Arabian study area ('Wajid area') (modified after Keller et al. 2011). (c) Geologic map of the southern Saudi Arabian study area ('Tabuk area') (modified after Pollastro et al. 1998). For simplification, the sample prefix AB-SA has been avoided in the sample localities shown in (b) and (c).

444
G Meinhold et al. Bassis et al. 2016a). The dominance of the stable minerals, such as zircon, tourmaline and rutile (high ZTR index, as defined by Hubert, 1962), indicating sediment maturity, among the transparent heavy mineral assemblage characterizes the Lower Palaeozoic sandstones (Babalola, 1999;Hussain et al. 2004;Knox et al. 2007;Bassis et al. 2016b). Upper Palaeozoic sandstones are more feldspar-rich ( Fig. 3b) and contain unstable transparent heavy minerals like garnet and/or staurolite (e.g. Knox et al. 2007;Bassis et al. 2016b). Whole-rock geochemistry and heavy mineral data reveal distinct changes in provenance between the Palaeozoic sandstones in central/northern (Tabuk area) and southern (Wajid area) Saudi Arabia (Bassis et al. 2016a, b) (Fig. 2a). Cambrian-Ordovician sandstones seem to be first-cycle sediments, probably sourced from the 'Pan-African' basement. The high mineralogical maturity of the sediment and the lack of apparent recycling trends seem to be a mixture of compositional variation of source rocks and reworking during deposition, coupled with intensive source area weathering and low sedimentation rates (see discussion in Bassis et al. 2016a). The overlying Upper Ordovician glaciogenic deposits likely consist of recycled Cambrian-Ordovician material. The Devonian-Permian sandstones show a significant influx of fresh basement material. The exact source for the Palaeozoic detritus is still under debate. The adjacent Arabian Shield most likely was a major contributor throughout the Palaeozoic, but far distant sources have to be taken into account as well (e.g. Al-Harbi & Khan 2008Bassis et al. 2016b). Possible source areas to the south include Archaean to Palaeoproterozoic terranes in Yemen (e.g. Babalola, 1999;Hussain et al. 2004;Bassis et al. 2016b), metamorphic terranes in Eritrea and Sudan, as well as the Mozambique Belt in East Africa (Bassis et al. 2016b).  Sharland et al. (2001) and (4) Haq & Schutter (2008). The dashed vertical line in the 'Sea-level changes' column represents an approximation of the present-day sea level. The 'Climate' column shows time interval with icehouse (cold) and greenhouse (warm) conditions. AP -Arabian plate sequence stratigraphy. PR -Proterozoic. (b) Thin-section photomicrographs of some representative samples from the Palaeozoic succession of Saudi Arabia under regular view (left-hand side) and with crossed polarizers (right-hand side). The Lower Palaeozoic sample is a highly mature and well-sorted quartz arenite from the Dibsiyah Formation (AB-SA79), typical for most of the Palaeozoic succession. The Upper Palaeozoic sample is an arkose from the Juwayl Formation (AB-SA98) with intense calcite cementation. Mineral abbreviations: qtzquartz; k-fspkali-feldspar; cccalcite. We refer to Bassis et al. (2016a, b) for details of the composition and maturity of the Palaeozoic succession of Saudi Arabia and the samples analysed in this study.

Methods
Samples analysed in this study were previously studied using thinsection petrography and whole-rock geochemistry (Bassis et al. 2016a) and conventional heavy mineral analysis including electron microprobe studies to determine the chemical composition of garnet and rutile grains (Bassis et al. 2016b). Samples were taken from surface outcrops in the Tabuk and Wajid study areas, respectively (Fig. 2). Lithology, stratigraphy and geographic coordinates of the studied samples are given in Table 1. After rock crushing, milling and dry sieving to obtain the 63−125 μm fractions, the heavy mineral fractions were separated using sodium polytungstate with a density of 2.85 g mL −1 (see Bassis et al. 2016b for details). Zircon selection from the heavy fractions was achieved by hand-picking under a binocular microscope. Zircon grains were fixed in epoxy resin mounts and polished to expose the interior of the grains. Prior to the analyses, transmitted light photomicrographs were taken with a polarizing microscope equipped with a camera system to study the zircon shape and roundness. Cathodoluminescence (CL) imaging was applied using a JEOL JXA 8900 RL electron probe microanalyzer (EPMA) equipped with a CL detector at the University of Göttingen (Department of Geochemistry, Geoscience Center) to reveal the internal structures (e.g. growth zones) and to guide spot placement in the zircon grains.
The U-Pb age determination was performed on a sector-field ICP-MS (Element2, ThermoFisher) coupled to a 193 nm Photon Machines Analyte G2 Excimer laser ablation system at the University of Münster (Institute of Mineralogy) following the procedure described in Lewin et al. (2020a). Isotope data were acquired on masses 202, 204, 206, 207 and 238. The mass 202 was used to quantify interference of 204 Hg on 204 Pb. A common Pb correction was only applied to an analysis if the fraction of common 206 Pb to total 206 Pb exceeded 1 %. A summary of the laser ablation ICP-MS operation parameters is given in Table S1 in the Supplementary Material available online at https://doi.org/10. 1017/S0016756820000576. Data reduction follows the procedure described by Kooijman et al. (2012). To account for the lower precision of 207 Pb/ 206 Pb values for young zircons, the data were filtered using two criteria as in previous detrital zircon provenance studies (Löwen et al. 2017;Lewin et al. 2020a): accepted were all zircon ages (1) with 90-110 % concordance [100 × ( 206 Pb/ 238 U/ 207 Pb/ 235 U)] for grains younger than 1200 Ma, and (2) with 90-110 % concordance [100 × ( 206 Pb/ 238 U / 207 Pb/ 206 Pb)] for grains older than 1200 Ma. This age was chosen due to the natural gap of zircon ages in the analysed samples. The R-package Provenance (Vermeesch et al. 2016) was used for visualization of the zircon age spectra as kernel density estimates (KDE) and for multi-sample comparison using multidimensional scaling (MDS). MDS is a useful technique to interpret, for example, large detrital zircon U-Pb age datasets by statistically comparing the data and visualizing all in a simple two-dimensional map in which 'similar' samples plot close together and 'dissimilar' samples plot far apart (e.g. Vermeesch, 2013;Vermeesch & Garzanti, 2015). The international chronostratigraphic chart of Cohen et al. (2018) was used as a stratigraphic reference for data interpretation.

Results
From the 24 sandstone samples of the Palaeozoic succession of Saudi Arabia, in total, 1915 zircon ages have been obtained, of which 1734 ages (91 % of all zircons) are 90-110 % concordant (Table 1) using the respective data filters described in Section 3. Complete datasets used here are reported in Table S2 in the Supplementary Material available online at https://doi.org/10. 1017/S0016756820000576.
The majority of the detrital zircons are clear or translucent. Subrounded to rounded grains are dominant. Most of the zircons show undisturbed oscillatory (magmatic) zoning in the CL images (Fig. 4). Many grains also show regrowth features suggesting that these grains have subsequently recrystallized during late-or postmagmatic cooling (Corfu et al. 2003). Zircons with xenocrystic cores and clearly zoned rims are minor. In the following, the zircon data are presented independently for both study areas starting for each with the oldest stratigraphic formation. We also mention the youngest zircon age for each sample/formation. However, we caution against overinterpretation of these ages in regard to estimation of the maximum age of deposition. A small amount of Pb loss can shift the U-Pb zircon age along the concordia curve toward younger ages, but the age will remain concordant using the filtering criteria as outlined in Section 3. Nonetheless, Dickinson & Gehrels (2009) could exemplarily show for the Mesozoic sedimentary succession of the Colorado Plateau that the youngest single grain zircon ages are compatible with depositional age for >90 % of the samples, and lie within 5 Ma of depositional age for~60 % of the samples. Also, the weighted arithmetic mean of the youngest population of detrital zircon ages is often used for the estimation of the maximum age of deposition (Dickinson & Gehrels, 2009). However, we did not apply that here as (1) not each sample provided the required minimum number (n ≥ 3) of grains for the youngest zircon population that overlap in age at 2 standard deviations and (2) this approach does also have its shortcomings (see Spencer et al. 2016 for details).
Juwayl Formation (Carboniferous-Permian): From two samples (100 and 98), in total, 114 zircon ages have been obtained, Fig. 4. Cathodoluminescence images of representative zircon grains from samples analysed in this study with the location of the LA-ICP-MS analysis spot and the corresponding 206 Pb/ 238 U age reported with ±2σ uncertainty in Ma for grains younger than 1.2 Ga and 207 Pb/ 206 Pb age reported with ±2σ uncertainty in Ma for grains older than 1.2 Ga. The number after the sample name corresponds to the analysis spot as indicated in Table S2 in the Supplementary Material available online at https://doi.org/10.1017/S0016756820000576.

Discussion
The first comprehensive detrital zircon age analysis of Palaeozoic sandstones from Saudi Arabia presented here revealed three prominent age populations at~625 Ma,~775 Ma and~980 Ma, accompanied by two minor age populations at~1840 Ma and 2480 Ma (Fig. 5). These age populations are similar to those identified by Garzanti et al. (2013) in two Palaeozoic sandstone samples from the Tabuk area. However, we do not further discuss these two samples due to their uncertain stratigraphic assignment. For example, Garzanti et al. (2013) described an Ordovician sandstone sample, which they assigned to the 'Tabuk Formation'. However, that term has been discarded long ago (see Laboun, 2010).

5.a. Possible source areas
As most of the zircons show oscillatory zoning and some show regrowth features in the CL images (Fig. 4), igneous rocks are the primary source of these grains. The Ediacaran-Cryogenian (541 to <720 Ma) and the Tonian (720 to <900 Ma) age groups represent igneous events associated with the Pan-African Orogen. Igneous crustal domains in the Arabian-Nubian Shield and the Eastern Granulites-Cabo Delgado Nappe Complex in the Mozambique Belt are likely sources, in agreement with palaeocurrent data, as igneous rocks with ages between 541 and 900 Ma are well known from these areas (e.g. Kröner, 1985;Stern, 1994;

450
G Meinhold et al. Johnson et al. 2011;Fritz et al. 2013;Johnson, 2014). Although it is difficult to narrow down the exact source, the~625 Ma zircon age peak in the studied samples (Fig. 5) is likely related to the youngest crustal growth phases in the Arabian-Nubian Shield, which are found in the eastern Arabian Shield and northern Nubian Shield. Also,~628 Ma old plutonic rocks occur in the western Nubian Shield (e.g. granodiorite at Gebel el-Asr in southern Egypt: Zhang et al. 2019a) and 630-500 Ma igneous rocks are known from the southern Nubian Shield (e.g. granitoids in the Southern Ethiopian Shield: Teklay et al. 1998;Yibas et al. 2002;Stern et al. 2012). The~775 Ma zircon age peak (Fig. 5) is likely related to the oldest crustal growth phases, which are found in the western Arabian Shield and the southern Nubian Shield (Fritz et al. 2013). The few early Palaeozoic ages (439-541 Ma) reported in almost all of the samples (Fig. 7) could be derived from post-orogenic igneous rocks. Cambrian post-orogenic plutonic rocks are known from the Arabian-Nubian Shield and SE Africa and Madagascar (Fritz et al. 2013). Cambrian and Ordovician alkaline and peralkaline plutonic rocks occur in Egypt and Sudan (Höhndorf et al. 1994;Woolley, 2001;Veevers, 2007). In addition, early Palaeozoic post-orogenic plutonic rocks, volcanic rocks and related products, such as volcanic ash beds (K-bentonites), as described from the Middle Ordovician of Libya (Ramos et al. 2003), and ascribed to volcanic activity along the northern margin of Gondwana during that time, could be an additional or alternative source (Meinhold et al. 2011;Lewin et al. 2020a). Late Devonian -Early Carboniferous zircons may come from within-plate, anorogenic igneous rocks from localized regions of NE Africa and Arabia, as known, for example, from Levant and its vicinity (e.g. Stern et al. 2014;Golan et al. 2018). The youngest zircon ages in the studied samples should not be overinterpreted, as they are commonly single ages. Similar to the detrital zircons from Ethiopia (Lewin et al. 2020a), some of the early Palaeozoic zircons from sandstones of Saudi Arabia were corrected for common Pb, and thus they may be over-corrected, leading to younger ages.
The pre-900 Ma zircon ages in the studied samples may have originally been xenocrystic zircons in Pan-African igneous rocks of the East African Orogen. This, however, will not be applicable to the majority of the pre-900 Ma ages recorded in the studied samples. Stern et al. (2010) have shown that only 5 % of individually dated zircons from igneous rocks of the Arabian-Nubian Shield are older than 880 Ma, with prominent age populations at 0.9-1.15 Ga (Tonian-Stenian), 1.7-2.1 Ga (late Palaeoproterozoic), 2.4-2.8 Ga (Palaeoproterozoic-Neoarchaean) and 3.2-3.5 Ga (Palaeoarchaean). Furthermore, most of the zircons analysed here have oscillatory (magmatic) zoning in the CL images, with xenocrystic cores a very minor component. Therefore, other sources than 'xenocrystic' zircons need to be considered for the pre-900 Ma detrital zircons.
The nearest basement rocks with Tonian-Stenian (900 to <1200 Ma) ages are documented from Sudan and the Sinai Peninsula. Kröner et al. (1987) described high-grade metasedimentary gneisses at Sabaloka in Sudan containing detrital zircons ranging in age from 2.6 to 0.9 Ga. Küster et al. (2008) described~900-920 Ma-old metamorphic zircons from basement rocks of the Bayuda terrane in Sudan, which indicate an orogenic event preceding the East African Orogeny. Both the Saboloka and Bayuda basement units belong to the eastern Saharan Metacraton (Fig. 1b). Be 'eri-Shlevin et al. (2009) reported 0.9-1.13 Ga-old zircons from the Sa'al schist in Sinai. This, together with the euhedral shape of some of these grains and the lack of other pre-Neoproterozoic zircon ages in the Sa'al schist is taken as evidence for the presence of 0.9-1.1 Ga-old crust at the NE margin of western Gondwana prior to the formation of the Arabian-Nubian Shield (Be'eri-Shlevin et al. 2009). Such crust is preserved as calc-alkaline volcanic and intrusive rocks (1.02-1.03 Ga) within the Sa'al metamorphic complex on Sinai (Be'eri-Shlevin et al. 2012). Although the Sinai basement is in proximity to the study area, at least to the Tabuk area, sediment supply from the~1 Ga basement seems unlikely, as palaeocurrent data point toward a southerly provenance (Dabbagh & Rogers, 1983;Stump & van der Eem, 1995;Babalola, 1999;Hussain et al. 2000). However, at least for the Wajid area, a prevailing northward transport direction cannot be inferred for the entire Palaeozoic succession, as opposing (NW to SE) transport directions occur in the glaciogenic units of the Upper Ordovician Sanamah Formation and the Carboniferous-Permian Juwayl Formation (Hinderer et al. 2009;Keller et al. 2011;Al-Ajmi et al. 2015). A change of the prevailing northward transport direction seems to have started during the Devonian because, in the Wajid area, dip measurements of cross-bedding indicate palaeocurrent directions to the north and NNE in the lower Khusayyayn Formation and already more easterly directions in the upper Khusayyayn Formation (Stump & van der Eem, 1995). The Arabian-Nubian Shield including its northernmost part on the Sinai Peninsula could thus be a possible source area. However, minor amounts of~1 Ga zircons are present in Neoproterozoic diamictites of northern Ethiopia (Avigad et al. 2007) and the Eastern Desert of Egypt . Thus, some~1 Ga zircons may have been recycled from Neoproterozoic siliciclastic sediments, but this probably does not account for the majority of the~1 Ga zircon grains in the studied Palaeozoic sandstones. The source of the~1 Ga zircons will be further discussed in Section 5.c in regard to the Gondwana super-fan system and the East African -Arabian zircon province, respectively.
Palaeoproterozoic and Archaean detrital zircons recorded in Palaeozoic sandstones of Saudi Arabia may have been recycled from Neoproterozoic sediments of the Arabian-Nubian Shield, as these sediments contain Palaeoproterozoic and Neo-and Mesoarchaean detrital zircon grains (Agar et al. 1992;Wilde & Youssef, 2002;Ali et al. 2010). Alternatively, they may represent xenocrystic cores originally derived from igneous rocks of the Arabian-Nubian Shield, but this would only account for a small number of grains (see Stern et al. 2010). A primary derivation from Palaeoproterozoic and Archaean basement is also likely, as metaigneous rocks of such age occur in the vicinity of the study area, e.g.~1.66 Ga crust (and probably also older crust based on Sm-Nd isotope data) in the Khida terrane of the eastern Arabian Shield (Whitehouse et al. 2001), and~2.55-2.95 Ga crust in the Arabian Shield of Yemen (Whitehouse et al. 1998). The eastern Saharan Metacraton also contains Palaeoproterozoic and Archaean crustal components (e.g. Küster et al. 2008;Bea et al. 2011;Kongnyuy et al. 2012;Zhang et al. 2019b) and thus can be seen as a possible source area. This may be supported by the fact that the eastern and southeastern parts of the Saharan Metacraton were exposed at the surface and thus were prone to erosion throughout most of the Palaeozoic, according to palaeogeographic reconstructions (e.g. Torsvik & Cocks, 2013).
Although Palaeo-and Eoarchaean zircon ages make up only 0.6 % of the obtained concordant age spectrum (Fig. 5), they are of great interest, as pre-3.2 Ga crust is not known from anywhere in the surrounding area (e.g. Hargrove et al. 2006;Stern et al. 2010;Fritz et al. 2013;Johnson, 2014). The exception is uppermost Palaeoarchaean basement rock in SW Egypt, i.e. 3.22 ± 0.02 Ga gneiss at Gebel Kamil (Bea et al. 2011) and 3.245 ± 0.05 Ga gneiss at Gebel Uweinat (Zhang et al. 2019b). The Devonian samples yielded the highest amount of Palaeo-and Eoarchaean zircon ages compared to all other studied samples and, surprisingly, they yielded the oldest zircon ages yet recorded from Arabia and the surrounding area: 3771 ± 37 Ma (#: 82) and 3850 ± 25 Ma (#: 38) in the Jubah Formation sample 152 (Figs 4, 5). Similar old ages (3729-3769 Ma, n = 4) were reported from inherited zircon crystals in Neogene ignimbrites of Central Anatolia in Turkey, and this was suggested to be an indication for the presence of Early Archaean basement in the Anatolian subsurface (Paquette & Le Pennec, 2012, 2013. However, an alternative explanation such as derivation from a Gondwanan source and later reworking by sedimentary and/or igneous events has been put forward (Köksal et al. 2013). Nutman et al. (2014) reported an inherited zircon core of 3579 Ma from a Pan-African granite of the Sanandaj-Sirjan Zone and discussed it as the oldest crustal material yet found in Iran. The source of these old grains is not yet known. Archaean zircons may represent xenocrystic zircons assimilated from terrigenous sediment, may reflect assimilation of cryptic Archaean continental crust that underlies the juvenile crust of the Arabian-Nubian Shield (Hargrove et al. 2006), or may indicate a more distal source area such as the Saharan Metacraton. It can be speculated that yet unknown terrigenous sediment and/or igneous rocks containing inherited up to Eoarchaean components may occur in the Saharan Metacraton. These rocks were later eroded, and the detrital material was transported by continental-scale river systems and/or glaciers toward the Arabian Peninsula.

5.b. Sample comparison
Multidimensional scaling (MDS) has successfully been applied to detrital zircon age datasets from Palaeozoic sandstones of northern Gondwana (Stephan et al. 2019a;Lewin et al. 2020a) and is applied here to Palaeozoic sandstones from Saudi Arabia (Section 5.b.1), followed by a comparison with samples from Ethiopia (Section 5.b.2). However, it is worth keeping in mind that quantitative comparisons of U-Pb age distributions can be biased because, for example, there is a grain-size dependence on age-data acquisition and interpretation (e.g. Yang et al. 2012;Ibañez-Mejia et al. 2018).

5.b.1. Comparison of samples from Saudi Arabia
First, among the samples from Saudi Arabia, the Cambrian and Ordovician samples (Fig. 8a) are separately discussed from the Devonian to Permian samples (Fig. 8b) before discussing all samples in a single MDS plot (Fig. 8c). Among the Cambrian and Ordovician samples, there is no clear separation according to stratigraphic age or study area, i.e. Wajid versus Tabuk. However, among the Wajid samples, the Cambrian-Ordovician Dibsiyah Formation samples (74 and 79) differ from those of the Upper Ordovician Sanamah Formation (62 and 73), and the two samples from the Sanamah Formation are very similar in their zircon spectra (Fig. 8a). The Cambrian-Ordovician Saq Formation sample 169 from the Tabuk area shows some resemblance to the Cambrian-Ordovician Dibsiyah Formation sample 74 of the Wajid area whereas the Cambrian-Ordovician Saq Formation sample 170 from the Tabuk area shows the least similarity to the Cambrian-Ordovician Dibsiyah Formation of the Wajid area (Fig. 8a). The Ordovician Qasim Formation samples from the Tabuk area form two sample groups. Group 1 comprises the older samples 144/1 and 145 whereas group 2 comprises the younger samples 124 and 126 (Figs 7, 8a). Overall, these patterns among the Cambrian and Ordovician samples suggest times of a homogeneous sediment source and times of sediment supply from different sources to the Wajid and Tabuk areas.
Among the Devonian to Permian samples, there is a clear separation according to stratigraphic age, Devonian versus Carboniferous-Permian, but not according to the study area, i.e. Wajid versus Tabuk (Fig. 8b). Thus, during the Devonian the sediment provenance was relatively similar for both the Wajid and Tabuk areas. The sediment source changed in the Carboniferous-Permian but again it was similar throughout both study areas.
Comparing the Devonian to Permian samples with the Cambrian and Ordovician samples in an MDS plot reveals some interesting similarities (Fig. 8c). According to that, zircon populations of the Upper Palaeozoic sandstones resemble those of the Lower Palaeozoic sandstones in both study areas. However, the Upper Palaeozoic sandstone composition cannot simply be explained by sediment recycling of Lower Palaeozoic sandstones because the latter are mainly quartz arenites (Fig. 3b) with a dominance of zircon, tourmaline and rutile among the transparent heavy minerals (Babalola, 1999;Hussain et al. 2004;Knox et al. 2007;Bassis et al. 2016b) whereas the Upper Palaeozoic are more feldspar-rich (Fig. 3b) and also contain unstable transparent heavy minerals like garnet and/or staurolite (e.g. Knox et al. 2007;Bassis et al. 2016b). Thus, it is more likely that fresh basement rocks with zircon ages resembling that from basement rocks, which supplied sediment to the Lower Palaeozoic sandstones, also supplied detritus to the Upper Palaeozoic sandstones. Minor sediment recycling from Lower Palaeozoic sandstones may have occurred, but it certainly cannot account for the majority of the detrital material in the Upper Palaeozoic sandstones.

5.b.2. Comparison of samples from Saudi Arabia with those from Ethiopia
In Ethiopia, Palaeozoic successions are present around the Mekelle Basin in the Tigray province of northern Ethiopia and, to a minor extent, in the Blue Nile region in the west of the country, and comprise Upper Ordovician and Carboniferous-Permian glaciogenic sandstones (e.g. Lewin et al. 2018, and references therein). Comparing detrital zircon data from Upper Ordovician sandstones of Ethiopia (Lewin et al. 2020a) with those from Saudi Arabia (this study) in an MDS plot reveals that Upper Ordovician samples from the Wajid area resemble those from northern Ethiopia whereas Upper Ordovician samples from the Tabuk area resemble those from northernmost Ethiopia (Fig. 9). The slight dissimilarity of the age spectra among the glaciogenic sandstones leads to the assumption that a change in provenance occurred during the Late Ordovician (Hirnantian) glaciation. Overall, however, the high similarity of many of the Cambrian-Ordovician age spectra to those of the Upper Ordovician glaciogenic sandstones (Fig. 8a) suggests that no change in provenance occurred with the onset of glaciation. Rather, the glaciers reworked sediment from the underlying Cambrian to Middle Ordovician successions to account for the high maturity of the Upper Ordovician siliciclastics (e.g. Lewin et al. 2018Lewin et al. , 2020a. Thus, large-scale sediment recycling occurred across NE Africa and Arabia during the Hirnantian Gondwana glaciation. The Carboniferous-Permian samples from Ethiopia plot distinct from those of Saudi Arabia, but with one sample from each of the Wajid and Tabuk areas being close to those from Ethiopia (Fig. 10). Carboniferous-Permian glaciogenic sandstones from 452 G Meinhold et al.
Ethiopia and Saudi Arabia show a similar sediment composition, i.e. being feldspar-bearing and containing unstable heavy minerals (Bassis et al. 2016a, b;Lewin et al. 2018Lewin et al. , 2020b. The detritus of the Carboniferous-Permian glaciogenic sandstones of Ethiopia was probably derived from the south, i.e. from basement rocks of the Nubian Shield as inferred from detrital zircon data (Lewin et al. 2020a), and a northward transport direction was inferred by Bussert (2010) based on the orientation and geometry of palaeolandforms, such as roches moutonnées (Lewin et al. 2020a). In northern Ethiopia, the glaciogenic sandstones were deposited in a large NNE-trending trough in which glacial erosion may have followed and reinforced this preglacial topography (Lewin et al. 2020a). Before the opening of the Red Sea, the Arabian Peninsula was attached to NE Africa (Fig. 1b), and thus the Wajid area was in proximity to Ethiopia. Taking all data together, it is suggested that the Carboniferous-Permian glaciogenic sandstones of Saudi Arabia received the majority of their detritus due to erosion of fresh basement rocks likely from the Nubian Shield such as in Ethiopia. On their way to the northeast, only minor amounts of detritus were admixed from the Arabian Shield. (Colour online) Non-metric multidimensional scaling maps for the detrital zircon age spectra of Palaeozoic sandstone analysed in this study. Only ages >500 Ma are used here due to the low reliability of younger ages following Lewin et al. (2020a). The colours in the circles correspond to the colours used for the stratigraphy shown in Figure 3a.

5.c. Gondwana super-fan system and East African -Arabian zircon province
Based on the similarity of detrital zircon U-Pb age spectra in Palaeozoic sandstones throughout Gondwana, Squire et al.  (Meinhold et al. 2013). The~1 Ga zircons are a key feature of the Gondwana super-fan system and were likely transported from regions in the centre of the East African Orogen, i.e. from the Mozambique Belt, or from regions of the SE Saharan Metacraton towards the continental margin of Gondwana during the early Palaeozoic. Intensive source area weathering, low sedimentation rates, and reworking during deposition lead to the high mineralogical maturity of the Lower Palaeozoic sediment (see discussion in Bassis et al. 2016a) contain-ing~1 Ga zircons. According to the Gondwana super-fan model, Saudi Arabia would lie more distal along the sediment path to the northeast. Based on a more extensive detrital zircon dataset including also peri-Gondwana units and using multivariate data analysis, Stephan et al. (2019a) showed the existence of three contrasting provenance end-members in Lower Palaeozoic sedimentary rocks of the former Gondwanan shelf, namely the Avalonian, West African and East African -Arabian zircon provinces (see also Stephan et al. 2019b). The East African -Arabian zircon province corresponds to the Gondwana super-fan system of Meinhold et al. (2013). To test whether the East African -Arabian zircon province is also a feature of the Lower Palaeozoic sandstones of Saudi Arabia, we compared in an MDS plot the newly obtained detrital    Lewin et al. (2020a). Note that the published data are those compiled by Meinhold et al. (2013) augmented by those from Altumi et al. (2013) and Lewin et al. (2020a). The orange field outlines the Gondwana super-fan sediments of NE Africa and Arabia, as defined by Meinhold et al. (2013), which are characterized by detrital zircons with 'Pan-African' U-Pb ages accompanied by zircon grains with Tonian-Stenian (~1 Ga) and minor older ages. Sediments with such ages correspond to the East African -Arabian zircon province of Stephan et al. (2019a).
zircon data from Saudi Arabia (this study) with those from the literature for north and northeast Africa, Israel and Jordan (Fig. 11). The majority of the Lower Palaeozoic samples from Saudi Arabia reveal a high similarity to samples from Libya, Israel, Jordan and Ethiopia (Fig. 11). This suggests that the Gondwana super-fan system was extensive and supplied detritus with a signature of an East African -Arabian zircon province also to the Lower Palaeozoic succession of Saudi Arabia (Fig. 12). The Algerian samples are not part of the cluster. In addition, some Cambrian and Lower Ordovician samples from Libya, Israel and Jordan are not part of the cluster. This is because the Gondwana super-fan system with an East African -Arabian zircon province seems to have been active at different times across NE Africa and Arabia, with the main phase of activity starting at some point between the Late Cambrian and late Middle Ordovician (Meinhold et al. 2013).
As discussed in Section 5.a, the Sa'al metamorphic complex on the Sinai Peninsula comprises~1 Ga basement rocks (Be'eri-Shlevin et al. 2012) and it is in proximity to the study area, at least to the Tabuk area. However, sediment supply from the Sinai basement to the Ordovician succession of Saudi Arabia is rather unlikely taking into account the roundness of the detrital zircon grains (this study) and the palaeocurrent data that point toward a southerly provenance (Dabbagh & Rogers, 1983;Stump & van der Eem, 1995;Babalola, 1999;Hussain et al. 2000). For example, palaeocurrents of Upper Ordovician sandstones from the southeastern Tabuk area show sediment transport away from the Arabian Shield toward the northeast (Clark-Lowes, 2005).

Conclusions
The erosion history of the East African Orogen and sediment recycling in northern Gondwana during the Palaeozoic have been topics of much debate. Our new detrital zircon U-Pb ages from 24 sandstone samples from the Cambrian to Permian succession of Saudi Arabia provide new insights into these topics.
Five main zircon age populations have been identified in varying amounts, with age peaks at~625 Ma,~775 Ma,~980 Ma, 1840 Ma and~2480 Ma. Igneous rocks are the primary source of these grains, as most of the zircons show undisturbed oscillatory zoning and some also show regrowth features in the CL images. Ediacaran to middle Tonian zircon grains are likely derived from igneous rocks of the Arabian-Nubian Shield. Palaeoproterozoic and Archaean grains may be xenocrystic zircons. However, this likely only accounts for a small number of grains. Recycling from  Stephan et al. (2019b) showing the main zircon provinces recognized in Cambrian-Ordovician siliciclastic sediments of northern Gondwana and its periphery. The East African -Arabian zircon province defined by Stephan et al. (2019a) corresponds to the zircon province of the Gondwana super-fan system of Meinhold et al. (2013). According to the detrital zircon U-Pb ages from Palaeozoic sandstones of Saudi Arabia (this study), the East African -Arabian zircon province can be extended to the central and southern Arabian Peninsula. White arrows show main sediment transport direction. older terrigenous sediment is possible. A primary derivation from Palaeoproterozoic and Archaean basement is also likely, as metaigneous rocks of such age occur in the vicinity of the study area. A minor amount (4 %) of the detrital zircons show Palaeozoic (340-541 Ma) ages. Early Palaeozoic zircons are likely derived from Cambrian and Ordovician post-orogenic igneous rocks, as present in the Arabian-Nubian Shield and to the south. Late Palaeozoic zircons may come from within-plate, anorogenic igneous rocks, as known from northern Arabia.
It is worth emphasizing that the zircon grains with up to Eoarchaean (3.6-4.0 Ga) ages are the oldest zircons yet presented from Arabia and its vicinity. It can be speculated that these grains represent xenocrystic cores originally derived from igneous rocks of the Arabian-Nubian Shield. However, such old ages have not yet been described from basement rocks of NE Africa and Arabia, and thus their origin is enigmatic. Tonian-Stenian (900 to <1200 Ma) zircons are present in varying amounts in all of the samples and are a feature of terrigenous sediment belonging to the Gondwana super-fan system and the East African -Arabian zircon province, respectively, in early Palaeozoic time. Multidimensional scaling (MDS) reveals similarities between detrital zircon age spectra from Ordovician sandstones of Saudi Arabia and time-equivalent successions of Ethiopia, Libya, Israel and Jordan. Thus, the Gondwana super-fan system with an East African -Arabian zircon province had a larger size than previously thought, as it also reached the Arabian Peninsula during the early Palaeozoic. Glaciogenic sandstones from Saudi Arabia related to the Late Ordovician (Hirnantian) Gondwana glaciation resemble timeequivalent successions from Ethiopia, suggesting prominent sediment transport and recycling with similar provenance. Besides that, MDS shows dissimilarities between detrital zircon spectra from Devonian and Carboniferous-Permian sandstones of Saudi Arabia, suggesting a prominent provenance change with the onset of the late Palaeozoic Gondwana glaciation.