Surface-wave tomography of the Emeishan large igneous province (China): Magma storage system, hidden hotspot track, and its impact on the Capitanian mass extinction

Large igneous provinces (LIPs) are commonly associated with mass extinctions. However, the precise relations between LIPs and their impacts on biodiversity is enigmatic, given that they can be asynchronous. It has been proposed that the environmental impacts are primarily related to sill emplacement. Therefore, the structure of LIPs’ magma storage system is critical because it dictates the occurrence and timing of mass extinction. We use surface-wave tomography to image the lithosphere under the Permian Emeishan large igneous province (ELIP) in southwestern China. We find a northeast-trending zone of high shear-wave velocity (Vs) and negative radial anisotropy (Vsv > Vsh; v and h are vertically and horizontally polarized S waves, respectively) in the crust and lithosphere. We rule out the possibilities of rifting or orogenesis to explain these seismic characteristics and interpret the seismic anomaly as a mafic-ultramafic, dike-dominated magma storage system of the ELIP. We further propose that the anomaly represents a hidden hotspot track that was emplaced before the ELIP eruption. A zone of higher velocity but less-negative radial anisotropy, on the hotspot track but to the northeast of the eruption center in the Panxi region, reflects an elevated proportion of sills emplaced at the incipient stage of the ELIP. Liberation of poisonous gases by the early sill intrusions explains why the mid-Capitanian global biota crisis preceded the peak ELIP eruption by 2–3 m.y.


INTRODUCTION
Large igneous provinces (LIPs) are characterized by rapid emplacement of primarily mafic magma in the lithosphere and volcanic eruptions forming plateau basalts in an area >10 5 km 2 (Bryan and Ernst, 2008). Many LIPs have been associated with abrupt environmental catastrophes and mass-extinction events (Sobolev et al., 2011;Ernst, 2014). While LIP-related extinctions generally occurred during or shortly after the major volcanic phase of the corresponding LIPs (Bond and Wignall, 2014), some preceded the major phase, leaving the exact relationship between LIPs and mass extinctions enigmatic.
It has been proposed that the structure of magma storage systems of LIPs, particularly sill intrusions, is a key to understanding this relation (Svensen et al., 2009). Modern hotspots, such as Yellowstone (northwestern United States), display the magma storage system of active LIPs (Jiang et al., 2018). However, they do not provide a direct constraint on the timing of mass extinction. Ancient LIPs, by contrast, potentially preserve a complete record and thus offer an opportunity for studying their structures and assessing the environmental impacts.
We studied the seismic structure of the Emeishan large igneous province (ELIP) in southwestern China using a new surface-wave tomography model. The ELIP is, so far, the only candidate for having caused the global biota crisis in the mid-Capitanian (ca. 262 Ma) (Wignall et al., 2009;Bond et al., 2020), which preceded the major volcanic activity of the ELIP (260-257 Ma) (Shellnutt et al., 2012). We attempt to characterize the ELIP's magma storage system and explain the unusual asynchrony.

GEOLOGICAL BACKGROUND
The ELIP is located in the western Yangtze craton of the South China block (Figs. 1A and 1B). It comprises mainly tholeiitic continental flood basalts with a lesser amount of picrites, lamproites, gabbros, pyroxenites, and other lithologies (e.g., Chung and Jahn, 1995;Xu et al., 2004). These rocks were mainly derived from the sublithospheric mantle (Chung and Jahn, 1995;Xu et al., 2004;Zhou et al., 2005). The southwestern corner of the ELIP is offset left-laterally by the Cenozoic Ailao Shan-Red River shear zone and is poorly exposed. The larger, better-preserved northeastern part, covering at least ∼3 × 10 5 km 2 , is divided into three concentric zones (inner, intermediate, and outer) that represent a decrease in flood basalt thickness outward (He et al., 2003;Xu et al., 2004). Before the ELIP eruption, the Maokou Formation that was deposited on the Yangtze carbonate platform underwent denudation and karstification (He et al., 2003;Xiao et al., 2016). The denudation began in the northeastern Sichuan Basin and propagated southwestward in the midto late Guadalupian (Hu et al., 2012). A map of the remnant thickness of the Maokou Formation shows a northeast-trending zone of anomalously thin strata extending from the Sichuan Basin to the ELIP center (Fig. 1C).
Paleomagnetic studies show an equatorial paleolatitude for the ELIP at the peak eruption time and an overall clockwise rotation of ∼27° of the South China block since 260 Ma (Huang et al., 2018). Magmatic underplating occurred at the Moho depth in the inner zone of the ELIP, as evidenced by high density, high seismic velocity, high ratios of compressional to shear-wave velocity (Vp/Vs), and thickened crust (Chen et al., 2015). A pronounced northeast-trending positive residual gravity anomaly extends from the ELIP center to the Sichuan Basin and has been attributed to ELIP intrusions (Deng et al., 2014).

METHOD AND RESULTS
We developed a radially anisotropic shearwave velocity model of the lithosphere under the ELIP region from Rayleigh wave and Love wave phase velocity at periods of 8-167 s (see the Supplemental Material 1 ). Rayleigh and Love waves are primarily sensitive to vertically (Vsv) and horizontally (Vsh) polarized shearwave velocities, respectively. Radial anisotropy, defined as 100% × (Vsh − Vsv) / Vs, where Vs is the Voigt average of Vsh and Vsv, exploits the variations between the two waves by assum-ing hexagonal anisotropy with a vertical symmetry axis (Babuska and Cara, 1991). Radial anisotropy may be induced by lattice preferred orientation of minerals such as mica, olivine, and pyroxene (Silver and Chan, 1991;Shapiro et al., 2004) or by shape preferred orientation of structures such as fractures, foliation, and magma (Silver and Chan, 1988;Emmermann and Lauterjung, 1997).
Our shear-wave velocity model (Figs. 2 and 3; Figs. S12-S14 in the Supplemental Material) shows high velocity at 15-220 km depths beneath the Yangtze craton east of the Xianshuihe-Xiaojiang fault and a low-velocity zone in the crust of the southeastern Tibetan Plateau. These results agree with those of previous studies: high-velocity anomaly dominates the Yangtze craton due to the cold and thick cratonic lithosphere, and the low-velocity zone is related to the thickened, partially molten Tibetan crust (Huang et al., 2010).
Radial anisotropy results show lateral and vertical variations in the crust and mantle lithosphere (Figs. 2 and 3; Fig. S15). Positive radial anisotropy (Vsh > Vsv) is found in the mid-and lower crust (depths of 15 km to the Moho) under the southeastern Tibetan Plateau, which has been interpreted as a result of the subhorizontal alignment of mica due to ductile deformation induced by the India-Asia collision (Shapiro et al., 2004). Another zone of positive radial anisotropy is located around the southern Xianshuihe-Xiaojiang fault. It is associated with low shear-wave velocities. We explain this as a result of subhorizontal ductile deformation in the middle and lower crust due to the lateral expansion of the Tibetan Plateau. In the lithospheric mantle below 80 km, positive radial anisotropy exists ubiquitously. Similar positive radial anisotropy is present in other cratons (Nettles and Dziewoński, 2008). This largely results from the lattice preferred orientation of olivine in horizontal planes.
The most intriguing phenomenon is a northeast-trending coherent zone of negative radial anisotropy (Vsv > Vsh) paired with high shearwave velocity under the inner and intermediate zones of the ELIP (Figs. 2 and 3). It appears as a broad zone in the shallow crust (0-15 km) and can be traced down to ∼80 km depth. Such a zone of reduced radial anisotropy is still detectable down to 120 km depth ( Fig. 2; Fig. S15). The amplitude of the negative radial anisotropy is mostly ∼−2%, but is as strong as −4% in the lower crust under the Sichuan Basin. A similar body of negative radial anisotropy was imaged, but not interpreted, at 20-35 km depths by Huang et al. (2010). Another study (Xie et al., 2013) found a narrow zone of negative anisotropy in the middle and upper crust under eastern Tibet and ascribed it to faults and cracks.

Geological Interpretation
We focus on the northeast-trending zone of high velocity and negative radial anisotropy  Table S1 for data sources (see footnote 1).
in this study. Negative radial anisotropy in the lithosphere is commonly explained by subvertical alignment of mineral crystals such as mica and olivine, macroscopic structures such as faults and shear zones, or partial melts (Shapiro et al., 2004;Xie et al., 2013).
To interpret this anomaly, we first consider three regional tectonic events: Cenozoic Tibetan orogenesis (Shapiro et al., 2004), Permian rifting (Cong, 1988), and Neoproterozoic rifting (Li et al., 1999). Faulting and/or fracturing caused by any of these events would have resulted in low, rather than high, seismic velocities in the craton. Neither the Xianshuihe-Xiaojiang fault nor the Ailao Shan-Red River shear zone is parallel to the observed northeast-trending anomaly, which does not favor the Cenozoic orogenesis interpretation. While the Panxi rift was active in Permian, a hypothetical northeastern extension of the failed rift triggered by the ELIP is unlikely because it would predict northeastward expansion of denudation starting from the end-Guadalupian; this is in conflict with the observation that denudation initiated in the mid-Guadalupian and propagated southwestward (Hu et al., 2012). A Neoproterozoic rift cannot explain the Permian volcanism or denudation, either. Because of the lack of geologic support, we favor an alternative explanation for the seismic anomaly.
We propose an interpretation that connects the seismic anomaly with the shape preferred orientation of magma systems (Fig. 4A). The magma storage system of a LIP consists of a network of dikes, sills, laccoliths, and magma chambers by which magma is transported through and stored within the lithosphere (Ernst et al., 2019). The geometry of such a system can be simplified into two end members in the scope of seismic radial anisotropy: vertical dikes and horizontal sills. The thickness of a single dike or sill is usually below the detection limit of seismic waves; in groups, however, they can be seismically detectable (Backus, 1962).
For a transversely isotropic medium of mafic to felsic composition with a vertical or horizontal symmetry (Eshelby, 1957), the presence of dikes and sills of mafic-ultramafic com-position can effectively alter the bulk elastic properties of the medium (Fig. 4A). The bulk Vs and radial anisotropy of a medium with a dike-dominated system both decrease at the time of intrusion; after the magma solidifies and cools, Vs increases and the radial anisotropy remains depleted. For a sill-dominated system, Vs decreases first then increases after cooling, while the radial anisotropy remains elevated.
The observed body of negative radial anisotropy and high Vs (such as anomalies "a" and "c" in Fig. 3) can thus be interpreted as the solidified magma storage system of the ELIP that is dominated by sub-seismic-scale, densely spaced mafic-ultramafic dikes. Extensive dike swarms of the ELIP are observed on the surface (Li et al., 2015; Fig. S19). This is consistent with our interpretation for the subsurface. Anomaly "b" (Fig. 3) is characterized by thicker crust, higher Vs, and less-negative radial anisotropy than anomalies "a" and "c", indicating a greater amount of intrusion with a higher fraction of sills there.

Tectonic and Environmental Implications
We place the South China block in a plate reconstruction framework (Fig. 4B). Paleomagnetic studies show that the South China block was moving northward continuously from 300 to 260 Ma and has experienced an overall ∼27° clockwise rotation since then (Huang et al., 2018). Assuming a stationary Emeishan hotspot, the predicted hotspot track should have aligned north-south in the middle Permian; i.e., northeast-southwest at present. Another prediction would be the southwestward propagation of denudation via dynamic topography. Indeed, the seismic anomaly trends northeast ( Fig. 2; Fig. S15) and overlaps the positive residual gravity anomaly (Deng et al., 2014) and the thinned zone of the Maokou Formation (Fig. 1C). Therefore, we infer that this 700-km-long anomaly represents the Emeishan hotspot track that was emplaced before the ELIP eruption and is concealed in the Yangtze craton.
Lithospheric thickness variations affect the route of magma transport and storage and the location of major eruptions. A schematic section A-A″ illustrates how the magma system developed through time (Figs. 4C and 4D). While the northern half of the hotspot track has no radiogenic age constraint yet, from the denudation record (Hu et al., 2012) and plate reconstruction (Huang et al., 2018), we infer that a mantle upflow reached the base of the thick craton under the Sichuan Basin in the mid-Guadalupian. It caused gentle uplift of the surface, leading to denudation and karstification of the Maokou carbonate rocks.
As the South China block drifted northward, melts started to form and infiltrate the craton, mainly as dikes. They did not find an "easy way" out until a thinner lithosphere was above the mantle upflow. In the upper crust, the Xiaojiang normal fault and the Panxi rift were (re)activated before and during the onset of the ELIP (Cong, 1988;He et al., 2003). On the surface, the denudation zone expanded southward with respect to the north-drifting South China block.
Magma underplating (Chen et al., 2015) must have started at anomaly "b" (Fig. 4C) by 263 Ma (Zhou et al., 2005), causing abundant intrusions with a higher proportion of sills in the Panxi region. Emplacement of laterally extensive sills tends to take place in mechanical layers (Kavanagh et al., 2006) such as crust-mantle interface, near the brittle-ductile transition, and within sedimentary basins. Sills efficiently store magma over large areas. When emplaced in a shallow basin, they can liberate large volumes of greenhouse gases and mercury that may trigger catastrophic environmental change (Svensen et al., 2009). Indeed, above "b", the 263 Ma Panzhihua layered mafic intrusions intruded and metamorphosed the Dengying dolomitic limestones in the shallow crust (Zhou et al., 2005), which could have caused degassing that led to the mid-Capitanian biota crisis.
As the South China block kept moving northward, the major eruption occurred at 260-257 Ma near the cratonic margin, associ-ated with domal uplift and dike swarm intrusions (He et al., 2003;Shellnutt et al., 2012;Li et al., 2015). The release of volatiles and mercury peaked, causing the end-Guadalupian mass extinction (Huang et al., 2019). After solidification, the 700-km-long magma storage system of the ELIP left a frozen-in seismic anisotropy and became a stiff body in the western Yangtze craton, which shaped the geometry of the Cenozoic Himalayan-Tibetan orogen (Xu et al., 2021).

CONCLUSIONS
Our study used surface-wave tomography to characterize the structure of the magma storage system of the ELIP, discovered the concealed Emeishan hotspot track, and established a connection among hotspot volcanism, plate motion, and a biodiversity crisis. Similar studies can be conducted for other LIPs, which may enable future models to evaluate the environmental impacts in a more quantitative way.  Babuska, V., and Cara, M., 1991