Magnetotelluric imaging of a fossil oceanic plate in northwestern Xinjiang, China

Because an oceanic plate colliding with a continental plate will usually be subducted and recycled into the deep mantle, a fossil oceanic plate after the closure of an ancient ocean has rarely been imaged in the subcontinental lithospheric mantle. This has led to a long-standing debate about the fate of subducted ocean plates. The problem can be addressed by imaging the lithosphere in a continental accretion zone with past ocean subduction. We present a study using long-period magnetotelluric data that reveals a large shallow-mantle conductor in a Phanerozoic accretion area in northwestern Xinjiang, China. This conductor extends >300 km laterally at depths from 120 to 220 km and resembles a segment of a fossil oceanic plate. The reduced resistivity is ascribed to the volatile-bearing metasomatic minerals, based on its relatively fertile nature and low temperature. Our results demonstrate that an oceanic plate can be trapped in continental lithosphere, underscoring the significance of oceanic plate subduction to continental accretion, and shedding new light on our understanding of continental formation and evolution. INTRODUCTION Geophysical imaging studies have shown some relics of subduction zones in the subcontinental lithospheric mantle (SCLM; Boerner et al., 1999; Chen et al., 2009). However, no such study has unambiguously revealed a complete fossil oceanic plate in the SCLM, which may be due to the fact that ancient oceanic plates are seldomly preserved in the lithospheric mantle, and, even if they are, they are difficult to image. This hinders our understanding of the fate of subducted oceanic plates, and requires the estimation of mass of recycled oceanic plates in the upper mantle to heavily rely on geochemical data (Meibom and Anderson, 2003). Imaging resistivity of modern subduction zones helps us to study the subduction processes because resistivity is very sensitive to phases such as fluids and melts (e.g., Soyer and Unsworth, 2006; Wannamaker et al., 2009). Recent years have witnessed an increasing interest in imaging resistivity of the SCLM in the study of craton evolution, by tracing relatively conductive compositions associated with ancient suture zones, metamorphism, and/or magmatic metasomatism (e.g., Jones et al., 2001; Chen et al., 2009; Evans et al., 2011). The western Junggar region in northwestern Xinjiang (Fig. 1), China, is located in the southwestern part of the Central Asian orogenic belt (CAOB). Recent studies based on magnetotelluric (MT) and seismic data (Xu et al., 2016; Zhang et al., 2017; Wu et al., 2018; Liu et al., 2019) suggested that remnants of a fossil intraoceanic subduction zone are preserved in this region. However, these studies only imaged the crustal structures of the region, due to the limited spatial coverage and bandwidth of seismic and MT data. Here, we used MT data from a long-period array to image the possible trapped oceanic plate in the upper mantle of the western Junggar region, which may shed new light on the fate of the subducted oceanic plate and the evolution of continental lithosphere. METHODS We used six GMS07e systems (Metronix Geophysics, Germany) and deployed an array of 101 long-period MT stations, with site spacing ranging from 25 to 45 km, across the western Junggar region and Chinese northwestern Tianshan (Fig. 1). The MT data from stations in desert regions were collected for more than 20 d, and other stations recorded data for at least 1 wk. The data were recorded on two electrical channels and three magnetic channels. The impedances (Z) and tippers (T) were estimated in a band of 392 Hz to 10,000 s by the remote referencing technique of Gamble et al. (1979) and/or the robust single-station method of Egbert and Livelybrooks (1996). The impedance estimation scheme strongly depends on the coherence between one horizontal electrical field and its orthogonally horizontal magnetic field, and the pattern of electromagnetic noise. We used the ModEM program (Kelbert et al., 2014; Egbert and Kelbert, 2012) to simultaneously invert the measured Z and T data from 101 stations at 29 periods, from 2.1 s to 10,000 s for Z, and from 2.1 s to 8000 s for T. To balance the data fit across all periods, we assigned an error floor of 5% of |Zxy| for Zxy and Zxx, and 5% of |Zyx| for Zyx and Zyy and 0.03 for T. We used an initial model of three layers: from the surface to 410 km with a resistivity of 100 Ω·m, from 410 km to 660 km with a resistivity of 10 Ω·m, and below 660 km with a resistivity of 3 Ω·m. *E-mail: xyxian@zju.edu.cn Published online 6 February 2020 †These authors contributed equally to this work. Downloaded from https://pubs.geoscienceworld.org/gsa/geology/article-pdf/48/4/385/4972188/385.pdf by Macquarie University user on 12 June 2020 386 www.gsapubs.org | Volume 48 | Number 4 | GEOLOGY | Geological Society of America We discretized the model using a horizontal grid with nodes at 8 km spacing in the core, surrounded by seven layers of nodes with the size increasing exponentially outward by a factor of 1.2. We also divided the model from the surface to 1500 km depth into six layers: 0–1 km, 1–20 km, 20–100 km, 100–410 km, 410–660 km, and 660–1500 km. We then divided each layer into a number of logarithmically equidistant thin sublayers (7, 19, 15, 12, 3, and 3 from the top to the bottom). The fine topmost layer was used to accommodate the near-surface distortion in the observed data. The parameterization resulted in a total of 65 × 80 × 54 nodes, plus an additional seven air layers above the surface. The final model after 140 NLCG (nonlinear conjugate gradient) iteration yielded a normalized root mean square misfit (nRMS) of 2.0. The tests on the sensitivity and resolution of the MT data, as well as the reliability of the results, are provided in the GSA Data Repository1. RESULTS Three-Dimensional (3-D) Resistivity Model The resistivity model from MT data, using the scheme described above, is shown in Figure 2. All conductors with resistivity 60 Ω·m through a relatively lower-resistivity zone of 25–60 Ω·m at depths of 60–120 km (Fig. 2D). Constraints from the Thermochemical Nature of the SCLM The bulk magnesium number (Mg#) of the lithospheric mantle is defined as 100 × MgO / (MgO + FeO), and it is a diagnostic indicator of the depleted or fertile nature of the mantle. Globally, the Mg# of highly depleted Archean SCLM is 93.1, while that of fertile (primitive) upper mantle is ∼89.3 (Griffin et al., 2009). The Mg# across the entire region imaged in our study is <90.5 (Zhang et al., 2019), suggesting typical Phanerozoic SCLM (Griffin et al., 2009). The imaged SCLM conductor corresponds to a relatively fertile zone with Mg# varying from 89.1 to 89.7, and a cold geotherm with the temperature ranging from ∼850 °C to 1250 °C (Zhang et al., 2019, their figure 6). The conductor begins at a temperature transition associated with electrical conduction from H-dominated to F-dominated (H—hydrogen ion and hydroxyl, F—fluoride ion), as illustrated by Li et al. (2017, their figure 7), and ends at a temperature approaching the incipient melting point for peridotites (Sifré et al., 2014). DISCUSSION Origin of the Large Conductor in the Shallow Upper Mantle There are different origins for low resistivity in the SCLM in different geodynamic settings, such as ponded melt in a hotspot, concentration of conductive minerals or water, and relics of metasomatism beneath stable continents. We can rule out the hotpot origin based on the tectonic history of the region, and we discuss the other two possible origins here. Partial Melting and Water We do not think that the SCLM conductor can be attributed to partial melting because the temperature profile is lower than the water-saturated solidus (Fig. 3A; Katz et al., 2003; Green, 2015). To evaluate the contribution of water to the SCLM conductor, we used the bulk resistivity from the MT data and the main oxides from multiple observables (Zhang et al., 2019), as well as the calibrated relationship between resistivity and water content from laboratory data (Table DR1). The estimated water content is plotted in Figure 3A. If the low resistivity were totally due to water, the required minimum water content would be at the maximum of the blue-shaded area in the Figure 3A, varying from more than 4000 ppmw at 60 km depth to less than dozens of parts per million at 220 km depth, a range that is substantially larger than the water storage capacity of nominally anhydrous minerals (NAMs; Férot and Bolfan-Casanova, 2012) at depths <160 km. It is thus implausible for water alone to explain the observed resistivity. Furthermore, eclogitization in a water-rich environment would be greatly promoted for a basaltic oceanic crust trapped in the shallow mantle (e.g., Griffin et al., 2009), which would result in foundering of 1GSA Data Repository item 2020108, supplementary information on methods of MT imaging and the calculations of electrical conductivity, with supplementary figures and Table DR1, is available online at http://www.geosociety.org/datarepository/2020/, or on request from editing@geosociety.org. A B C Figure 1. Locations of the research area and deployed magnetotelluric (MT) sites in the western Junggar region in northwestern Xinjiang, China. (A) Simplified tectonic f