Late Permian High-Ti Basalt in Western Guangxi, SW China and Its Link With the Emeishan Large Igneous Province: Geochronological and Geochemical Perspectives

High-Ti (Ti/Y) flood basalts are widely distributed in the Late Permian Emeishan large igneous province (LIP), SW China, and their spatial distribution and genetic mechanism are important to reveal the role of plume-lithosphere interactions in the LIP origin. Western Guangxi is located on the eastern edge of Emeishan LIP. To explore the genesis of the high-Ti basalt in western Guangxi and any genetic link with the Emeishan LIP, we performed whole-rock geochemical and Sr-Nd isotope and zircon U-Pb-Hf isotope analyses on the Longlin basalts from western Guangxi. The results indicate that the Longlin basalt from Tongdeng area has relatively high SiO2 but low MgO and TFe2O3 contents. The rocks have zircon εHf(t) = −0.42 to 6.41, whole-rock (87Sr/86Sr)i = 0.707167–0.707345, and εNd(t) = −2.5 to −2.14. In contrast, the Longlin basalt from Zhoudong area has relatively low SiO2 but high MgO and TFe2O3 contents. The rocks have whole-rock (87Sr/86Sr)i = 0.706181–0.706191 and εNd(t) = −0.57 to 0.69. Four Longlin basalt samples display LREE enrichments and HREE depletions, and with indistinct δEu and δCe anomalies. LA-ICP-MS zircon U-Pb dating on three Longlin basalt samples (from different localities) yielded consistent weighted average age of 257.9 ± 2.6 Ma (MSWD = 0.55), 259.5 ± 0.75 Ma (MSWD = 3.0), and 256.7 ± 2.0 Ma (MSWD = 0.68), indicating a Late Permian emplacement. Considering the similar age and geochemical features between the Longlin basalt and Emeishan flood basalts, we interpret that the former is spatially, and temporally associated with the Emeishan LIP. Geochemical features show that the high-Ti basalts in western Guangxi resemble Deccan-type continental flood basalts (CFBs), which were derived by decompression melting of the mantle plume. Combined with previous geochemical studies, we suggest that the difference in Ti content and Ti/Y ratio in CFBs are related to the depth and melting degree of mantle source, in which high-Ti features may have been linked to low degree of partial melting in the deep mantle.


INTRODUCTION
Large igneous provinces (LIPs) are formed by voluminous mafic (and minor felsic) rocks emplaced in a short period of time, and are generally considered to be mantle plume-related but not directly related to normal ocean spreading and subduction (Coffin and Eldholm, 1994;Courtillot and Renne, 2003;Acocella, 2021), although recent works suggested that interaction between mantle plumes and subducting/subducted slabs may be plausible (Xu et al., 2019;Xu et al., 2021). LIPs are of great significance for studying of global environmental catastrophes and mass extinctions, mineral resources (esp. magmatic massive Cu-Ni-PGE sulfide and Fe-V-Ti oxide deposits), supercontinent reconstruction, and regional uplift (Wang and Zhou, 2005). Hence, petrogenesis of LIPs, such as the role of plume-lithosphere interaction in the origin of the continental flood basalts (CFB), and the origin of the high-Ti (Ti/ Y), and low Ti mafic magmatism, has attracted much research attention (Fodor, 1987;Peate et al., 1999;Lai et al., 2012).
The Emeishan LIP is located in the western margin of the Yangtze craton and the eastern margin of the Tibetan plateau, SW China, and likely related to Late Permian mantle plume processes (Chung and Jahn, 1995;Li et al., 2008;Zhang, 2009). Complex tectonic events in the Sanjiang fold belt had strongly deformed the original distribution of the Emeishan LIP rock units (Xu et al., 2013). Besides, whether the basaltic rocks in the Song Hien Tectonic Zone (NW Vietnam) are parts of the Emeishan LIP remains disputed (Halpin et al., 2016, and ref. therein). Therefore, it is of great significance to reveal the original spatial extent of CFB magmatism in the Emeishan LIP. Emeishan LIP consists mainly of flood basalts and associated mafic-ultramafic intrusions and (minor) felsic intrusions. The volcanics comprise mainly picrite, basalt and basaltic andesite, and can be divided into a high-Ti and low-Ti suite. This is similar to that in many other LIPs, such as the Siberian and Deccan Traps (Fodor, 1987;Peate et al., 1999;Fan et al., 2008). However, the validity of such Tibased distinction is still disputed Xu et al., 2004;Zhang et al., 2001;Shellnutt, 2014). Xu et al. (2001) considered that these high-Ti basalts were mantle plume-related, and formed by low-degree partial melting in the garnet stability field. However, Li et al. (2008) argued against substantial lithospheric mantle assimilation. Besides, some authors considered that the Emeishan basalts have a continuous TiO 2 range and should not be divided into two suites (Hou et al., 2011). For instance, Zhang (2009) considered that the high-Ti basalts may have formed by fractionation and accumulation of Fe-Ti oxides, whilst Hao et al. (2004) suggested that the high-Ti and low-Ti basalts are different fractionation products of the same parental magma.
Western Guangxi in SW China is located in the southwestern margin of the Yangtze craton ( Figure 1A). It tectonically lies in the intersection between the Tethyan and Pacific tectonic domains. Mafic magmatism in the region is widespread and closely related to gold mineralization (Huang et al., 2015). Among the many mafic magmatic units, the Late Permian basalts are coeval with the Emeishan CFB, and generally display high-Ti (or high-Ti/Y) characteristics (Fan et al., 2008;Lai et al., 2012;Zhang and Xiao, 2014;Huang et al., 2015). However, petrogenetic setting of this magmatism is variably suggested to be 1) intra-oceanic setting related to the eastern Paleo-Tethys (Wu et al., 1993;Wu et al., 1997); 2) mantle plume related as part of the Emeishan (Liao et al., 2013;Zhang and Xiao, 2014;Huang et al., 2015). Besides, there are various views on the genesis of these rocks, including partial melting of the subcontinental lithospheric mantle (Lai et al., 2012), and mixing of plume-derived magma and continental lithosphere materials (Fan et al., 2008). In order to decipher the genesis and petrogenetic setting of these Late Permian basalts in western Guangxi, we conducted detailed field and petrographic observations on the high-Ti basalts in the Longlin area (western Guangxi), followed by zircon U-Pb-Hf isotope, as well as whole-rock elemental and Sr-Nd isotope analyses. Our findings constrain the formation timing and petrogenetic setting of the Longlin basalts, and explore the spatial distribution of Emeishan LIP and the genesis of high Ti-basaltic magmatism in western Guangxi.

GEOLOGICAL BACKGROUND AND PETROGRAPHIC FEATURES
The study area is located in the northern part of the Nanpanjiang-Youjiang rift basin in western Guangxi (Figure 1). Local tectonism is strong, and has undergone the Caledonian (Early Paleozoic), Variscan (Late Paleozoic), Indosinian (Triassic), Yanshanian (Jurassic-Cretaceous), and Himalayan (Cenozoic) tectonic cycles (Xu et al., 2013). The Caledonian orogeny may have closed the early Paleozoic South China ocean basin, forming the folded basement of the region. Afterward, the Indosinian orogeny terminated the pelagic sedimentation in South China, forming the complex tectonic framework of the region.
The study area is mainly covered by pelagic sedimentary rocks of Cambrian and Devonian to Triassic ages ( Figure 2). The Cambrian rocks (total outcrop size: 60 km 2 ) comprise mainly carbonates, which are mainly distributed at Shechang and minor at Maxiong, Shuidong, Ganhao, and Poyan of the Xinzhou-Pingban counties (Li et al., 2019). Upper Paleozoic rocks (total outcrop size: 814 km 2 ) comprise mainly carbonates and minor clastics, of which the carbonate rocks are mainly distributed in the platforms and the limbs of the Shechang anticline, as well as in the northeastern part of the study area (Li et al., 2019). The clastic rocks are mainly distributed in the periphery of the platforms and in the southern limb of the Shechang anticline. Thick Triassic turbidite sequences are widely distributed in the deep-water basins between the carbonate platforms (total outcrop size: 1,005 km 2 ) (Li et al., 2019).
Magmatic rocks are uncommon in the study area, and consist mainly of mafic volcanic and (minor) plutonic rocks (total outcrop size: 30.75 km 2 ) in the central and southwestern parts of western Guangxi, e.g., at Nawei, Tongdeng, Boli, and Xiaoren. The basalts occur mainly as interlayers among the Middle Permian Sidazhai Formation and Upper Permian Linghao Formation ( Figure 3A). Minor tuffaceous rocks were also found, and there is no discernible alteration in the wallrocks. Grain size decreases toward the intrusive margin, and vesicle and almond/pillow structure are uncommon. As shown in Figure 4, there are three layers of basalt in the Linghao Formation, with the thickness of 65, 15, and 10 m from bottom to top. The bottom of the first basalt layer consists of the Linghao Formation siltstone, and the top of each layer consists of volcaniclastic rocks. Eight leastaltered samples (from Tongdeng, Xiaoren, and Zhoudong) were collected from the Linghao Formation basalt. The rocks are grey-greyish green and porphyritic ( Figure 3B). Compositionally, the rocks contain mainly pyroxene (40-45%) and plagioclase (47-52%), together with minor biotite, ilmenite, pyrite, and pyrrhotite ( Figure 3C). The basalts have phenocrysts of pyroxene and plagioclase (grain size: 0.5-1.2 mm). The pyroxene is subhedralanhedral elongated and partly ilmenite-altered, whilst the plagioclase is elongated and partly epidote-altered. Finegrained pyroxene, sphene, and serpentine occur as interstitial minerals ( Figure 3D).
Analytical Chemistry and Testing Services (ALS) Chemex Co. Ltd. (ALS Geochemistry method ME-XRF26). Fused glass disks with lithium borate were used and the analytical precisions were better than ± 0.01%, estimated from repeated analyses of the standards OREAS 146 and SY-4. Trace element concentrations of the samples were determined by a Perkin-Elmer ELAN 6000 inductively coupled plasma mass spectrometry (ICP-MS) after acid digestion in high-pressure Teflon bombs at the GIG-CAS (ALS Geochemistry method ME-MS81). The analytical errors were <5% for rare earth element (REE) and high field strength element (HFSE), and 5-10% for the other elements, based on repetitive analyses of standards NCSDC47009 and SARM-5. Detailed analytical procedures followed those outlined by Zhang et al. (2019).
Whole-rock Sr-Nd isotope analysis was conducted at the Isotope Geochemistry Laboratory, Wuhan Center of Geological Survey (China Geological Survey), using a Triton and MAT 261 thermal ionization mass spectrometer. The chemical analysis was monitored by the GBW04411, GBW04419 and NBS607 procedures, and the equipment conditions by the NBS987, NBS981 and JNDI procedures. Detailed analytical procedures were as described by Wang et al. (2015).
The CL images were performed using the TIMA analysis in the Chengpu Geological Testing Co. Ltd., Langfang, China. LA-ICP-MS zircon U-Pb dating was conducted at the Key Laboratory of Institute Continental Collision and Plateau Uplift, Tibetan Plateau Research (Chinese Academy of Sciences) analysis. The analysis used a UP193 F × 193 nm ArF ablation system coupled with an Agilent 7500a ICP-MS. Analytical conditions include 193 nm wavelength, < 4 ns pulse width, 35 μm spot size. 91,500 zircon was used as the external standard, and was analyzed between every six samples. The sample isotopic and elemental contents were calculated with the Glitter 4.5 software, whilst the Pb isotope calibration and U-Pb age calculation was performed with the ComPbcorr#3_18 (Andersen, 2002), and Isoplot 4.15 (Ludwig, 2003), respectively.
Zircon Lu-Hf isotope analysis was conducted at the State Key Laboratory of Geological Processes and Mineral an energy density of ∼7 J/cm 2 , and helium as the carrier gas. The standards (91,500 and GJ-1 zircon) were analyzed once for every eight zircon samples. Detail procedures for data calibration and processing are as described in Ye et al. (2013).  Frontiers in Earth Science | www.frontiersin.org September 2021 | Volume 9 | Article 729955 5 ANALYSIS RESULTS

Whole-Rock Major and Trace Element Geochemistry
In this study, we analyzed four fresh Longlin basalt samples (two samples from Tongdeng and Zhoudong each) for their wholerock geochemical compositions ( Table 1) Xu et al. (2001). In the Zr/TiO 2 -Nb/Y and SiO 2 -Nb/Y discrimination diagrams ( Figure 5), all samples fall inside (near) the alkali basalt field.
The Tongdeng and Zhoudong samples have total REE (ΣREE) 250.70-259.89 ppm and 210.26-211.33 ppm and La N / Yb N 13.24-13.44 and 9.78-10.12, respectively. This indicates that the former has higher ΣREE and LREE/HREE fractionation than the latter. Samples from both Tongdeng and Zhoudong have indistinct Eu and Ce anomalies ( Figure 6A; Tongdeng: δEu 0.93-0.94, δCe 0.95-1.01; Zhoudong: δEu 1.09-1.15, δCe 1.01-1.02). In the primitive mantle-normalized multi-element diagram ( Figure 6B), the Zhoudong samples have distinct positive anomalies for some large ion lithophile elements (LILEs, e.g., Rb, K) and Sr, but negative anomalies in Nb and Y. In comparison, the Tongdeng samples have generally higher LILE contents, and have more distinct negative Sr and Zr-Hf contents than the Zhoudong samples. HREE contents of the Tongdeng and Zhoudong samples are similar.
The Longlin basalt samples have low loss on ignition (LOI 2.68-4.09 wt%), indicating minimal alteration/ weathering, which is consistent with the petrographic observation.

Zhoudong Samples
Zircons (n 25) from sample ZD-1 show oscillatory zoning in CL images ( Figure 7E). Most of them have length 50-80 μm, width 40-60 μm, and length: width ratio 1.1-1.4, and have high Th/U 0.32-1.87 (avg. 0.72). All the zircons cluster around the concordia and yielded a concordia age of 256.8 ± 1.1 Ma (MSWD 9.5) ( Figure 7F), which is similar to the weighted average age of 256.7 ± 2.0 Ma (MSWD 0.68) ( Table 2). For Hf isotopes (Table 3) Table 4). The Nd model age for the Tongdeng and Zhoudong samples are 1,227-1,217 Ma and 1,002-1,017 Ma, respectively. In the ( 87 Sr/ 86 Sr) i -εΝd(t) discrimination diagram ( Figure 8A), all the Tongdeng and Zhoudong samples fall close to the MORB field and overlap with the Emeishan high-Ti basalts. In the εΝd(t)-εHf(t) discrimination diagram ( Figure 8B), the Tongdeng samples all fall far below the mantle evolution line, whilst the Zhoudong samples fall on or slightly below it.

Permian Mafic Magmatism in Western Guangxi
In this study, our new zircon LA-ICP-MS U-Pb dating results on three Longlin basalt samples yielded weighed mean 206 Pb/ 238 U age of 257.9 ± 2.6 Ma (Tongdeng), 259.6 ± 1.4 Ma (Xiaoren) and 256.7 ± 2.0 Ma (Zhoudong). Zircons in Longlin basalt are generally characterized by (weak) oscillatory zoning on CL images and high Th/U ratios (0.32-1.87), resembling typical magmatic zircons (Hoskin and Schaltegger, 2003 In recent years, many Late Permian magmatic rocks have been discovered in Western Guangxi.

Crustal Assimilation
As shown in the ( 87 Sr/ 86 Sr) i -εΝd(t) diagram ( Figure 8A), the samples plot above the evolutionary trend defined by mid-ocean ridge basalt (MORB) and lower or middle continental crust (LCC/MCC). This, and the relatively wide zircon εHf(t) range and linear correlations between MgO, Nb/La, and SiO 2 (Figure 9), suggest that parental magma may have experienced crustal assimilation during their ascent. Crustal   Sun and McDonough, 1989), as found in both the Tongdeng (Th/Ta 2.55-2.63) and Zhoudong (Th/Ta 1.95-1.97) samples. The Tongdeng samples may have had stronger crustal assimilation influence as their Th/Ta ratios are clearly higher than their Zhoudong counterparts. This conclusion is also supported by the higher SiO 2 and Nb/La, the lower MgO, zircon εHf(t) and whole-rock εNd(t) values, together with the more xenocrystic zircons in the Tongdeng samples.
If the rock formation was influenced by crustal assimilation, the rocks would have (Th/Ta) PM ≈ 1 and (La/Nb) PM > 1 (by lower crustal material); and (Th/Ta) PM > 2 and (La/Nb) PM > 2 (by upper crustal material) (Peng et al., 1994). The Longlin basalts have (Th/Ta) PM 0.94-1.27 and (La/Nb) PM 1.17-1.20, showing mainly lower crustal influence. In the ( 87 Sr/ 86 Sr) i -εΝd(t) discrimination diagram ( Figure 8A), the Longlin basalts plot above the evolutionary trend defined by MORB and LCC/MCC, indicating also lower crustal involvement. Assuming that the Longlin basalts were depleted mantle-derived, zircon Hf isotope calculation indicates that crustal input is of 8% for the Zhoudong samples and 40% for the Tongdeng samples. In the magma mixing modelling, the εHf(t) value of the depleted mantle-derived, ancient crustal-derived, and parental magmas of the Longlin basalts are 16 [depleted mantle εHf(t) value at 259 Ma], −13.22 [minimum measured Longlin basalt εHf(t) value] and −1/−10 [average measured Zhoudong/Tongdeng basalt εHf(t) value], respectively. The assumed Hf contents of the depleted mantle and crustal material are 0.309 μg/g (Taylor et al., 1981) and 4.5 μg/g (Yangtze middle-lower crustal average; Chi and Yan, 2007), respectively. Considering that the mantlesourced parental magma of the Longlin basalts has lower/equal εHf(t) value than/to that of the depleted mantle, and that the assimilated crustal material has lower εHf(t) value than the lowest εHf(t) value (−13.22) of the samples, the Zhoudong and Tongdeng samples should have their actual degree of crustal assimilation below 8 and 40%, respectively.
The Lu-Hf and Sm-Nd isotopic systems have similar geochemical characteristics, and hence the εHf(t) and εNd(t) values show strongly positive correlation . Nd-Hf decoupling of magmatic rocks are commonly related to residual garnet or zircon enrichment in the magma source region Vervoort et al., 2000;Zeng et al., 2021a), and/or crustal assimilation (Patchett et al., 1984;Carpentier et al., 2009). From Figure 8B, it is shown that the Tongdeng samples  . have clear Nd-Hf decoupling, whereas for the Zhoudong samples such decoupling is indistinct. Because not all samples of the Longlin basalts show Nd-Hf decoupling, the decoupling is probably unrelated to the mantle source region. As aforementioned, the Zhoudong samples have higher degree of crustal assimilation than the Tongdeng samples, implying that their Hf-Nd isotope decoupling was likely assimilation related. Since Hf 4+ and Zr 4+ have similar ionic radii, Hf is commonly enriched in zircon through isomorphism, leading to very low zircon Lu/Hf value (distinct lower than most rock-forming minerals) (Vervoort et al., 2011). After prolonged accumulation, εHf(t) of zircon would be clearly lower than that of other minerals, and thus the sediments or rocks with high zircon contents would have lower εHf(t) than those that lack zircon (e.g., mudstone, shale, or pelagic sediments) (Patchett et al., 1984;Carpentier et al., 2009). Meanwhile, the Sm-Nd  (Zindler and Hart, 2003); LCC, MCC, and UC (Jahn et al., 1999); global sediments from ; global lower crust from Vervoort et al. (2000); IMM and PMM from Kempton et al. (2012). Frontiers in Earth Science | www.frontiersin.org September 2021 | Volume 9 | Article 729955 isotope system is unaffected by zircon. Hence, the εHf(t) value of zircon-rich sediments or rocks would deviate negatively from its εNd(t) value with reference to the mantle array, whereas the εHf(t) value of zircon-poor sediments or rocks would deviate positively from its εNd(t) value with reference to the mantle array. The εHf(t) values of the Tongdeng sample are obviously lower than its εNd(t) value, indicating that the crustal materials assimilated by it should be zircon-rich sediments or rocks. The lesser crustal assimilation for the Zhoudong samples may have caused the indistinct Hf-Nd decoupling. The content of REE in the crust is higher than that in the mantle (Taylor and McLennan, 1985;Sun and McDonough, 1989). Therefore, compared with the Zhoudong samples, the Tongdeng samples experienced stronger crustal contamination, which may be one of the reasons for their higher REE content. In addition, the contents of Zr and Hf in Tongdeng samples are obviously higher than Zhoudong samples, which is consistent with the fact that the crustal materials assimilated by them are rich in zircon.

Fractional Crystallization
It is generally considered that mantle-sourced primitive melts may have had relatively high Ni (> 400 ppm), Cr (> 1,000 ppm) and Mg# (73-81) contents (Mahoney, 1997;Wilson, 1989), which are not found in the Longlin basalts . This shows that fractionation of mafic minerals (e.g., olivine and pyroxene) has likely occurred (Frey and Prinz, 1978;Wang et al., 2016;Zeng et al., 2021b). Due to the insufficient element geochemical data, it is hard to determine the fractionated minerals from the Harker-type diagrams. Using the lowest Cr content of mantle-sourced melt (1,000 ppm; Wilson, 1989), calculation suggests that the petrogenesis of the Tongdeng samples requires ∼50% olivine or ∼20% pyroxene fractionation, whilst that of the Zhoudong samples requires ∼30% olivine or ∼10% pyroxene fractionation (partition coefficient from GERM website: http://www.earthref. org/). Because crustal assimilation can also reduce the magma Cr content, the degree of crystallization differentiation of mafic minerals in Zhoudong and Tongdeng samples must be lower than the above calculated values, especially the Tongdeng samples with higher degree of crustal contamination. The lack of Eu anomalies in the Longlin basalts indicates insignificant plagioclase fractionation.

Nature of Magma Source
All Longlin basalt samples in this study have lower εHf(t) and εNd(t) values than the depleted mantle, suggesting that the basaltic magma has undergone crustal assimilation during the magma ascent, and/or the magma was sourced from the mantle metasomatized by subducting crustal material, possibly from the altered Paleotethyan slab (Zeng et al., 2018;Xu et al., 2021).
The magma source region can be determined by the most primitive samples of the Longlin basalts (i.e., Zhoudong samples, with the least crustal assimilation and fractionation). Since Yb is more compatible in garnet than in pyroxene and spinel, whilst Sm is incompatible in all these minerals, Sm/Yb can also be used to constrain the mantle source mineralogy (Aldanmaz et al., 2000). In the Sm vs. Sm/Yb diagram ( Figure 10A), the Zhoudong samples plot between the partial melting curves of garnet lherzolite and spinel-garnet facies (1:1) lherzolite. From the Sm/Yb vs. La/Sm diagram ( Figure 10B) (Özdemir and Güleç, 2014), the compositions of Longlin basalt are consistent with magmas formed by low-degree melting (1.0-2.0%) and >80% garnet lherzolite contribution. It is commonly accepted that the depth of the garnet-spinel stability zone is 75-85 km in the upper mantle (McKenzie and O'Nions, 1991;Klemme and O'Neill, 2000). Therefore, magma for Longlin basalt was likely formed in the spinel-garnet transition zone at 75-85 km depths.

Geodynamic Significance
In many tectonic discrimination diagrams (Figure 11), the Longlin basalts fall inside the within-plate basalt field, similar to most mafic rocks from western Guangxi. According to Sun et al. (2007), within-plate basalts have w (Nb)/w (Zr) > 0.04 and w (Th)/w (Nb) > 0.11, of which continental riftrelated basalts and continental extension-related basalts have w (Th)/w (Nb) 0.11-0.27 and >0.27 (generally 0.27-0.67), respectively. The Longlin basalts have their Nb/Zr (0.13-0.16) similar to that of within-plate basalts and their w (Th)/w (Nb) (0.12-0.15) similar to that of continental riftrelated basalts of the within-plate basalts. This shows that the Late Permian western Guangxi was located in a continental rifting setting.
Considering the role of mantle plume and lithosphere in the CFB formation, Turner et al. (1996) divided the CFB into two genetic types, i.e., Deccan-type and Parana-type. Deccantype CFB has geochemical characteristics similar to OIB, which is derived from decompression melting of asthenospheric peridotite in the mantle plume. The Parana-type CFB is rich in incompatible elements, and is formed by partial melting of the lithospheric mantle (driven by mantle plume-derived heating). As shown in Figure 6, the Late Permian basalts in western Guangxi are enriched in LILEs and LREEs, with no discernible Nb, Ta and Ti anomalies. The distribution curves are obviously different from NMORB and EMORB, but similar to OIB. In addition, the Late Permian basalt samples in western Guangxi fall mainly in the OIB field ( Figure 12). Therefore, we suggest that the Late Permian basalt in western Guangxi belongs to the Deccan-type, and its formation is closely related to the decompression melting of a mantle plume. In the Th/Nb-TiO 2 /Yb diagram ( Figure 12C), most of the western Guangxi basalt samples fall near the type-IIIB line (plume-SZLM interactions), which also demonstrates a mantle plume source with lithospheric input. This is consistent with our conclusion that the Longlin mafic magma has experienced crustal assimilation.
Possible formation mechanisms for the Emeishan high-Ti and low-Ti basalts are still under debate, and have been attributed to the different magma source regions, degrees of partial melting, fractional crystallization, and crustal assimilation processes, or a combination of these factors (e.g., Xu et al., 2001;Hao et al., 2004;Li et al., 2008;Zhang, 2009). In this study, formation of the Tongdeng and Zhoudong samples have clearly different degrees of crustal assimilation, yet they all have high-Ti features, which shows no obvious Frontiers in Earth Science | www.frontiersin.org September 2021 | Volume 9 | Article 729955 relation between the high-Ti features and crustal assimilation. Since Fe-Ti oxides commonly fractionated at the late magma evolution stage (when MgO < 4 wt%) (Xu et al., 2013). The Zhoudong basalt samples have MgO > 4.42 wt%, which suggests that the high-Ti features are unlikely to be related to ilmenite fractionation. As afore-discussed, Late Permian basalts from western Guangxi (e.g., Longlin) were mainly sourced from low-degree partial melting of the mantle in the spinel-garnet transition zone, features that may have shaped their high-Ti characters. Since lithospheric thinning is commonly present above the mantle plume axis, partial melting could have extended to shallower depths (e.g., spinel stability field) and become more extensive (i.e., higher-degree partial melting). Conversely, distal from the mantle plume head, the lithospheric mantle is thicker and mantle decompression melting occurred at greater depths (in the spinel-garnet transition zone and garnet stability field) with lower intensity (i.e., low-degree partial melting). Therefore, the location of western Guanxi basalts on the Emeishan LIP margin may explain the high-Ti character of the basalts in this region.
High-Ti and low-Ti basalts also coexist in other LIPs around the world, such as the Paraná, Deccan, and Siberia (Peate et al., 1999;Xia et al., 2012). Fodor (1987) and Arndt et al. (1993) considered that high-Ti and low-Ti basalts are caused by different degrees of melting of the same mantle source, with the degree of partial melting degree for high-Ti basalts being lower than their low-Ti counterparts. McKenzie and Bickle (1988) suggested that the lithospheric mantle thickness is the principal factor that controls the depth and degree of partial melting in the mantle source region. Xia et al. (2012) considered that high-Ti basalts are formed by lower degree of partial melting in the deeper part of the mantle plume (tail), whereas low-Ti basalts are formed by higher degree of partial melting in the shallower part of the mantle plume (head). The above viewpoints are consistent with our conclusion that the difference of mantle source depth and degree of partial melting have caused the significant Ti content variation of magmatic rocks in the Emeishan LIP.

1) Longlin basalts from different places in western Guangxi
yielded similar Late Permian zircon U-Pb ages, i.e., 257.9 ± 2.6 Ma (Tongdeng), 259.5 ± 0.75 Ma (Xiaoren), and 256.7 ± 2.0 Ma (Zhoudong), coeval with the Emeishan flood basalt emplacement. 2) Element and Sr-Nd-Hf isotope geochemistry show that the Longlin basalts were formed from low-degree partial melting in the spinel-garnet transition zone at depths of 75-85 km. The parental magma may have undergone mafic mineral (e.g., olivine and pyroxenes) fractionation and varying degrees of lower crustal assimilation. 3) Late Permian basalts in western Guangxi have similar emplacement age and geochemical features to the Emeishan high-Ti flood basalts. This suggests that the Emeishan LIP activity extended to western Guangxi. 4) Late Permian basalts in western Guangxi have OIB-type geochemical characteristics, and their formation is closely linked to the decompression melting of a mantle plume. The high-Ti features may have caused by the low-degree partial melting in the deep mantle.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article, further inquiries can be directed to the corresponding author.

AUTHOR CONTRIBUTIONS
CGZ and RYZ designed the research and drafted the manuscript. CML and JJ collected and processed the data. XJS prepared