Electrical conductivity of mantle minerals beneath East Asia revealed by geomagnetic observatory data

Abstract

The electrical conductivity of the earth’s mantle can provide important information about geodynamic phenomena. East Asia is associated with complex tectonics and geodynamic processes. Hence, it is necessary to better understand the deep structure beneath East Asia. In this study, geomagnetic data obtained from East Asian observatories are employed to image the conductivity structure of the mantle at depths ranging from 410 to 900 km. First, the data are processed using the modified bounded influence remote reference processing (BIRRP) method and the ratio method is used to correct for the ocean effect. Thereafter, the stable C-response curves at the 27 observatories are estimated, and 1D electrical conductivity models for these observatories are established using the L-BFGS method. The conductivity-depth profiles reveal a heterogeneous distribution of the electrical conductivity beneath East Asia. The mantle transition zone (MTZ) beneath East China and Japan is found to be more conductive, whereas the MTZ beneath central and southern regions of China is more resistive. In East China, the dehydration of the stagnant Pacific slab may lead to an increase in the conductivity of the mantle minerals. There is also the possibility of upwelling of the thermal material from the lower mantle beneath the Japanese Island arc. In Northwest China, there exists a large high-conductive body beneath the Tarim area, which could indicate an upwelling of the Tarim mantle plume. Our results provide insights into the deep structure of the earth at the mantle scale.

1 Introduction

A comprehensive understanding of the origin, evolution, and dynamics of the earth requires a thorough investigation of its structure at the mantle scale (Munch et al. 2018). Seismic, gravimetric, and electromagnetic studies significantly contribute toward inferring the deep structure of the earth (Kelbert et al. 2009).

East Asia is located in the southeast of the Eurasian continent, and its geophysical structure and tectonic activities have been influenced by the powerful interplay between the Philippine Sea plate, Pacific plate, and Indian plate. The Pacific and Philippine Sea plates are undergoing subduction beneath Japan and East China (see Fig. 1). In Southwestern China, the Indian–Asian collision has resulted in the elevation of the Tibetan plateau (Huang and Zhao 2006). The deep structure of East Asia is crucial for understanding the complex tectonics and geodynamic processes (Yuan et al. 2020).

Fig. 1

Topographic map and tectonic background distribution of East Asia. The colors represent the surface topography. The red circles indicate geomagnetic observatories employed in this study. The red triangles indicate the volcanoes in China. The red dashed line is the North–South gravity lineament (NSGL). The white lines indicate the plate boundaries

Multiple seismic images have illustrated the velocity structure of the mantle transition zone (MTZ) beneath East Asia (Li and van der Hilst 2010; Zhao et al. 2017). These images have significantly promoted the development of earth science. However, relying solely on seismic results leaves many unanswered questions.

Geomagnetic depth sounding (GDS) has emerged as a highly effective technique for obtaining detailed information about the deep structure of the earth (Zhang et al. 2020). This robust geophysical method can detect the electrical conductivity at depths ranging from 400 to 1200 km (Olsen 1998).

GDS has been widely employed to investigate the conductivity structure beneath different regions around the world (Olsen 1998; Neal et al. 2000; Shimizu et al. 2010; Koch and Kuvshinov 2015; Munch et al. 2018). However, there is relativity little research on the electrical conductivity structure deep beneath East Asia. Zhao et al. (2001) obtained the electrical conductivity structure beneath Northeast China using the Network-MT method. The MTZ beneath East China was found to exhibit a high conductivity. Zhang et al. (2020) estimated the C-response curves for 31 observatories and inverted these responses to obtain the electrical conductivity structure beneath East China. The observations suggested that the MTZ had a higher conductivity in the east than in the west. Yuan et al. (2020) drew largely the same conclusions from 45 geomagnetic observatories. However, due to the lack of data on the MTZ conductivity structure beneath Japanese Islands, research on the overall morphological structure of the subducting West Pacific Plate remains insufficient, which limits the study of the geodynamics deep beneath East Asia.

In this study, the geomagnetic data of 27 observatories were chosen in East Asia, and the C-responses from these observatories were estimated. A 1D inversion was performed to obtain the electrical conductivity structure of the mantle. Finally, the cause of the high-conductivity anomalies could be inferred by combining the geomagnetic and seismic results.

2 Data

2.1 Data selection

To obtain the deep electrical conductivity structure of East Asia, a large amount of geomagnetic data were collected from East Asia. The data were obtained from the National Geomagnetic Network of China and the World Data Center for Geomagnetism. After careful selection, 27 observatories were selected for the C-response estimation, with a minimum record length of at least 10 years.

3 C-response estimation

In a 1D layered model, $P_{1}^{0}$ can be used to represent the approximated magnetospheric source. The C-response can be defined in terms of the GDS data (Banks 1969):

$$C(\omega ) = - \alpha_{0} \tan \theta H_{r} (\omega )/2H_{\theta } (\omega )$$

(1)

where ω is the angular frequency; ${\alpha }_{0}$ is the mean radius of the Earth; θ denotes the geomagnetic colatitude; ${\text{H}}_{\text{r}}$ and H_θ are the vertical and northward components of the geomagnetic fields, respectively. Under 1D conditions, tanθ in Eq. (1) is often considered a compensation term for the source space structure.

The data error of the C-response can be evaluated as follows:

$$\delta C(\omega ) = \left| C \right|\sqrt {(1 - coh^{2} /coh^{2} )(1/\beta )^{1/(L - 1)} }$$

(2)

Based on the study by Schmucker (1999), the quality of the estimated C-responses can be evaluated using the squared coherence (coh²) expressed in Eq. (2). To obtain coh² over sections for each frequency, block averaging was applied, which was derived from the bounded influence remote reference processing (BIRRP) software package (Chave et al. 2004). Notably, 1−β represents the level of confidence, which is typically set to 0.9; L denotes the number of sections for a specific frequency.

When processing the geomagnetic data, the outliers in the original time series should be removed, and the secular variations that were obtained from the International Association of Geomagnetism and Aeronomy were also removed. Since the forward calculation of the GDS is performed in the geomagnetic system, the observed data should be rotated from the geographic coordinate system to the geomagnetic coordinate system to obtain the corresponding magnetic field components. Finally, based on the BIRRP data processing software (Welch 1967; see Percival and Walden 1993), the self-reference method was used instead of the remote reference method (Semenov and Kuvshinov 2012) to calculate the C-responses with a period of 1.3–113 days.

GG, geodetic geographic coordinates system; GM, geomagnetic coordinates system.

Figure 2 shows a comparison of the C-responses estimated using the BIRRP software and those obtained from the study by Semenov and Kuvshinov (2012). Evidently, the BIRRP-estimated curves are smoother and more stable, which can be characterized by the lower error bars. This method can effectively suppress the outliers in the horizontal and vertical components. Therefore, this method performs well when processing geomagnetic data with significant disturbances. After careful data processing and considering the data distribution, 27 observatories could produce stable C-response curves and were employed in this study. Figure 1 shows the spatial distribution of these observatories, while Table 1 presents more detailed information.

Fig. 2

Comparison of the C-responses estimated using the BIRRP method and those obtained from the study by Semenov and Kuvshinov (2012). The blue circles and lines indicate the results obtained by Semenov and Kuvshinov (2012), while the red circle and lines indicate the BIRRP-estimated C-responses

Table 1 Geomagnetic observatories used for inversion, along with their acronyms

4 Inversion method

4.1 Ocean effect correction

Oceans are high-conductivity bodies, and when the distance between the measuring point and the coastline is less than the skin depth of the low-frequency electromagnetic signal, the presence of an ocean will affect the distribution of the electromagnetic field. Therefore, for C-responses with short periods, the east–west current in highly conductive seawater will undergo severe distortion on encountering a high-resistance continent. These perturbations in the C-responses need to be particularly considered for coastal observatories.

For 1D inversion, the ratio method can be used to correct for the ocean effect (Everett et al. 2003; Kuvshinov and Olsen 2006; Püthe et al. 2015a). The specific correction equation is as follows:

$$C^{{ {\text{obs}},{\text{corr}}}} = C^{{{\text{obs}}}} k = C^{{{\text{obs}}}} (C^{1D} /C^{{1D + 3D{\text{surface}}}} )$$

(3)

where the corrected C-response is denoted by ${\text{C}}^{\text{ obs}\text{,}{\text{corr}}}$, while the observed response is denoted by ${\text{C}}^{\text{ obs}}$, and k is the correction coefficient. ${\text{C}}^{\text{ 1}{\text{D}}}$ is the C-response derived from the 1D conductivity model (Püthe et al. 2015b), while the predicted response ${\text{C}}^{\text{ 1}{\text{D}}+ \text{3} \text{D }{\text{surface}}}$ incorporates the variability in the ocean conductivity considering the heterogeneity described by Everett et al. (2003) and Kelbert et al. (2014).

For the predicted ${\text{C}}^{\text{ 1}{\text{D}}+ \text{3} \text{D }{\text{surface}}}$, the staggered-grid finite difference method is used for the forward estimation (Uyeshima and Schultz 2000). A thin-shell layer of 12.6 km is used with a resolution of 1° × 1° (Munch et al. 2018) to simulate the surface conductance of the earth.

Figure 3 shows the ocean effect on the C-responses at Changli (CHL) and Chongming (CHM) stations along the coastal region. A gap can be observed between the primary C-responses and the corrected ones at small periods particularly when the period is less than 10 days, while the ocean effect disappears at larger periods.

Fig. 3

Comparison of the C-responses with and without ocean effect at Changli (CHL) and Chongming (CHM) stations. The C-responses of the model incorporated with the variability in the ocean conductivity are indicated by red lines, while the C-responses of the 1D conductivity model with no surface conductivity layer are indicated by the blue lines

5 One-dimensional inversion method

The L-BFGS inversion method can efficiently solve nonlinear optimization problems (Nocedal 1980; Liu et al. 1989). It has been widely used in electromagnetic inversions owing to its lower memory requirement and higher convergence stability (Avdeev and Avdeeva 2009; Liu et al. 2013; Weng et al. 2015).

The basic iteration format of L-BFGS is as follows:

$$m_{k + 1} = m_{k} + \alpha_{k} p_{k}$$

(4)

$$p_{k} = - {\rm H}_{k}^{ - 1} \nabla \Phi_{k}$$

(5)

$$\nabla \Phi_{k} = \left. {\left( {\partial \Phi /\partial m_{1} , \partial \Phi /\partial m_{2} , \cdots , \partial \Phi /\partial m_{N} } \right)^{T} } \right|$$

(6)

In the above equations, the number of iterations is represented by k. $\alpha_{k}$ denotes the search step, and T is the transposition of the matrix. The search direction is denoted by $p_{k}$. ${\text{H}}_{\text{k}}^{-1}$ is the inverse of the Hessian matrix. By utilizing the second derivative of the objective function to adjust the search direction, as expressed in Eq. (5), the convergence can be accelerated.

6 Results

The accuracy of the inversion results significantly depends on the initial model. Several initial models were tested to find a suitable model for the inversions at the different observatories. Figure 4 shows the corresponding inversion results. During the tests, two types of earth models were designed: a complex model, as shown in Fig. 4a, and a simplified model comprising only five layers, as shown in Fig. 4b.

Fig. 4

Initial model tests with two different forward models. The different colored lines indicate the results of different initial models used in the 1D inversion; 38, 41 and 54 layers represent the initial models, which divide the earth into various layers, and “Kuvshinov and Olsen 2006” represents the model developed by Kuvshinov and Olsen (2006), which divides the earth into more than 130 layers. a 41-layer model tests; the model is the global mean conductivity model developed by Püthe et al. (2015a, b); b 5-layer model tests

The forward models represent the models that were designed to simulate the earth’s structure. The corresponding C-responses were calculated, and 3% Gaussian noise was added to constitute the observed responses. Figure 4a and b shows the inversion results of the two designed forward models under five types of initial models, respectively. For all these initial models, the earth was divided into different numbers of layers, and every model was a half-space model of 1 Ω⋅m. When the change in objective function less than 10^–5 or the regularization factor is less than 10^–4, the inversion terminates.

Figure 4 shows that most of the initial models could effectively restore the forward model in the depth range of 600–1200 km, whereas the 54-layer model was much closer to the designed model regardless of the complexity of the earth’s structure. Therefore, in the inversion of the real data, the 54-layer model was chosen as the initial model.

Using the estimated C-response mentioned above, the Ocean effect was corrected, and then a 1D L-BFGS inversion was performed, and the initial model for the inversion comprised a half-space with a conductivity of 1 S⋅m⁻¹. The earth was divided into 54 layers, with the first layer having a thickness of 10 km and the subsequent layers increasing in thickness by a factor of 1.05. The regularization parameter initially took a value of 2 and decreased by a coefficient of 0.5. Figure 5 shows the typical corresponding fitting curves. Evidently, the inversion curves exhibit an excellent fit with the observed C-responses at small periods, while the fitting gaps become significant toward the large periods.

Fig. 5

Fitting curves of the inversion result at typical stations. The blue circles indicate the observed C-responses after ocean effect correction, and the red lines represent the C-responses of the inversed model

To better display the fitting effect of the response curves at these observatories, a map of the root mean square (RMS) of the 1D inversion was drawn. As depicted in Fig. 6, most observatories have a low RMS of less than 0.4, indicating a better fitting between the response curves of the inverted model and the observed data. Some of the observatories in the south and west of China have higher RMS, but the values are still less than 0.5, implying that the inversion results are convincing.

Fig. 6

RMS distribution of the 1D inversion at different observatories. The red color represents a poor fitting, whereas the blue color represents a good fitting

7 Discussion

7.1 Heterogeneity in the conductivity

Figure 7 presents the typical conductivity-depth profiles from six geomagnetic observatories. As shown by the inverted models, it is apparent that the electrical conductivity varies significantly beneath the different regions in East Asia. For the CNH (Changchun) observatory, the conductivity significantly increased at the bottom of the MTZ, which has also been identified through the magnetotelluric network (Ichiki et al. 2001). The MZL (Manzhouli) and WMQ (Wulumuqi) observatories also showed enhanced conductivities in the MTZ and the upper lower mantle. The four observatories in Northwest China all showed a high conductivity in the depth range of 410–900 km, forming a large high-conductivity area. In comparison, the JYG (Jiayuguan) observatory maintained a low conductivity for depths less than 670 km, though the conductivity significantly increased at the lower mantle. For the CDP (Chengdu) observatory, the conductivity was significantly lower than the global mean for depths shallower than 900 km, consistent with the conclusions drawn by Yuan et al. (2020). The MIZ (Mizusawa) observatory, which is located in the Japanese Islands and near the ocean, exhibited a considerably higher conductivity than the global average for depths greater than 670 km. These variations in the electrical conductivity in the depth range of 410–900 km indicate that the mantle structure of East Asia has heterogeneity.

Fig. 7

Typical conductivity-depth curves of the observatories corresponding to different regions in East Asia. The thick black line indicates the results of the global mean model which is modified from Kelbert et al. (2009)

8 Map view of the conductivity

To more effectively illustrate the conductivity variations in the MTZ and in the upper lower mantle, electrical conductivity maps were constructed based on the 1D inversion results obtained from 27 observatories (Fig. 8). These maps display the conductivity at depth intervals of 410–520, 520–670, and 670–900 km to highlight the heterogeneity in the conductivity more vividly, where every slice has its own color bar. The global-scale conductivity model suggested the existence of a high-conductivity anomaly in the MTZ of Eastern China (Karato 2011). Moreover, 3D inversions have confirmed this phenomenon in East China (Kelbert et al. 2009; Semenov et al. 2012).

Fig. 8

Map view of the electrical structure of East Asia at different layers and corresponding seismic images: a Conductivity distribution at different depths; b Maps of the logarithm of the conductivity; c Maps of the seismic images at different depths (Huang and Zhao 2006). The pink dashed line indicates the NSGL, and the pink lines indicate the plate boundaries

Evidently, the conductivities in the MTZ of East China and Japan are considerably higher than that in the MTZ of Western China (as shown in Fig. 8a). Notably, in the depth range of 520–670 km, the high and low conductivity characteristic is consistent with the undulation of China’s surface topography, and many scholars believe that the dividing line (NSGL) can be considered the front edge of the subducting Pacific plate (Huang et al. 2006; Yuan et al. 2020; Zhang et al. 2020). Furthermore, the four observatories in Northwestern China distinctly indicate the presence of a high-conductivity anomaly in the depth range of 410–900 km, which supports the findings reported by Guo et al. (2021).

Figure 8b illustrates the variation in the conductivity with the depth in the form of a unified logarithmic conductivity color bar. With the increase in the depth, the conductivity of the study area showed an upward trend. The conductivity ranged from 0.01 to 1 S/m. In the MTZ, the conductivity of the north and east of the study region is higher than other regions.

9 Distribution of the water contents

The main minerals in the MTZ are wadsleyite and ringwoodite, which are the Earth’s main reservoirs of water, and the existence of water has a significant impact on the conductivity of the mantle (Karato 1990). Moreover, physical experiments on rocks under high-temperature and high-pressure conditions have shown that the electrical conductivity of wadsleyite and ringwoodite in the MTZ is closely related to the temperature and water content (Xu and Shankland 1998; Huang et al. 2005).

Huang et al. (2006) determined the effects of water and temperature on the conductivity of minerals wadsleyite and ringwoodite to infer the water content of the MTZ (as shown in Fig. 9). It also reveals that conductivity of these minerals depends strongly on water content but weakly on temperature. The presence of water can elevate their conductivity by up to three orders of magnitude (Koyama et al. 2006; Yoshino et al. 2008).

Fig. 9

The distribution of temperature and water content on electrical conductivity modified from Huang et al (2005): a Wadsleyite; b ringwoodite. The numbers on the curves represent electrical conductivity (in S/m); the orange lines represent the temperature used in the estimation

To further reveal the cause of the anomalies beneath different regions, the inverted conductivity values of different stations are put in Fig. 9 to obtain the water content at each station. Since the temperature has relative small influence on conductivity, we set the temperature of MTZ in the study area uniformly to 1850 K.

The estimated water content of MTZ at each geomagnetic station is shown in Fig. 10. At the depth of 410–520 km, the northeast region of China (Such as WMQ, WUS and HOQ stations) has relatively high water content of about 0.05–0.1 wt%. In the Bohai Sea and eastern coastal areas of China, the average water content is about 0.08 wt%. In addition, the water content of the Japanese island arc is also relatively high, especially for the stations of YSS and SSO where the water contents are close to 0.15 wt%.

Fig. 10

The distribution of water content in the MTZ: a 410–520 km; b 520–670 km

At the depth of 520–670 km, the water content is lower. In the northeast China and the eastern coastal areas of China, the water content is about 0.04 wt%; and in the Japanese island arc, the water content is about 0.02 wt%; in addition, CHL and YSS stations have high water content more than 0.2 wt%.

9.1 Properties of anomaly bodies beneath Eastern China

In the MTZ of Eastern China, a stripe of high-conductivity and high-velocity anomaly bodies can be observed, as shown in Fig. 8. As mentioned above, both temperature and water content can increase the conductivity of the minerals in MTZ. However, based on the temperature alone, a high velocity should correspond to a relatively low conductivity in the MTZ (Huang et al. 2005). Therefore, the water content of the minerals in the MTZ beneath Eastern China should be considered. In addition, Research on rock physics has revealed that the seismic wave velocities are only weakly sensitive to water content (Karato 2011). Therefore, the high-conductivity and high-velocity anomaly bodies may represent the material with high water content and density. Moreover, many global and regional tomographic images indicate that the Pacific slab has reached the MTZ of China and that it is horizontally distributed over a range of over 1000 km on a discontinuous interface at a depth of 660 km in Eastern China (Fukao et al. 2001; Zhao et al. 2009; Chen et al. 2017). It has been estimated that the water content in this region is about 0.08 wt%, according to Guo and Yoshino (2013), the stagnant Pacific slab has the capacity to convey an immense volume of water into the MTZ. Consequently, this high-conductivity body in the MTZ of Eastern China could be attributed to this water-enriched slab.

9.2 Properties of anomaly bodies beneath Japanese islands arc

It is worth noting that, due to the use of geomagnetic data of Japanese islands, we also obtained the conductivity structure beneath the Japanese islands arc. Beneath the Japanese Islands, the MTZ and upper lower mantle can still be characterized by a high conductivity. However, in view of the seismic images, this high-conductivity body corresponds to a low velocity. Evidently, the property of this body differs from that of the body found beneath Eastern China, and the main factor contributing to the high conductivity in the MTZ beneath the Japanese Islands may not solely be the water content.

Meanwhile, evident low-wave-velocity anomaly zones could be seen at or near the front edge of most stagnant slabs (Hayes et al. 2018; Goes et al. 2017), and these zones showed high-conductivity features (Li et al. 2020; Kelbert et al. 2009; Semenov and Kuvshinov 2012).

Without considering the temperature changes, the estimated water content in this area is about 0.15 wt%, which exceeds the estimated critical water content in the upper mantle (Bell and Rossman 1992), suggesting that partial melting may occur. Considering the above factors, it can be inferred that the high-conductivity and low-velocity body in the MTZ beneath the Japanese Island could be attributed to the combined effect of the temperature and water content. The water content can lower the solidus of the minerals, and the high temperature can likely cause partial melting.

Nakajima and Hasegawa (2007) revealed the presence of a distinct low-velocity anomaly in the upper mantle beneath Southwestern Japan through seismic tomography. It has been speculated that this anomalous region was caused by temperature variations exceeding 200 K and that it is associated with a melt fraction of less than 1% (Drewes 2009). This mushroom-shaped low-velocity body may be the remnant of an upwelling mantle plume (He 2019). Owing to the proximity of the hot mantle upwelling to the subducting plate, its elevated temperature may induce partial melting at the base of the West Pacific subducting plate (Yang and Faccenda, 2020).

Therefore, it is speculated that there is a possible partial melting layer underneath the subducting slab. This partial melting layer may facilitate the deflection and flattening of the plate at the base of the MTZ, potentially accounting for slab stagnation (Mao and Zhong 2018).

9.3 Properties of anomaly bodies beneath northwest China

The inversion results also indicate notable high-conductivity anomalies in the MTZ and the upper lower mantle beneath the northwestern region of China, particularly in the upper lower mantle at Urumqi station. Seismic imaging results have shown a large-scale lateral anomaly characterized by a low velocity beneath the Tarim Basin (Lei et al. 2007; Wei et al. 2020), and these two types of anomalies almost overlap.

The high conductivity in the MTZ is typically related to increased temperature or water content (Huang et al. 2005). The water content in this area is estimated about 0.15 wt%, and such a high water content is typically related to the dehydration of the stagnant slab. Since there is no evident stagnant slab in the MTZ beneath Tarim Basin, and for the lower mantle, the water content is extremely low (generally not exceeding 20 ppm (Hirschmann et al. 2009)), the high conductivity beneath Tarim Basin could be due to the high temperature.

The conductivity at different observatories near Tarim Basin could be fitted by a rise in temperature, and this rise in temperature may have caused changes in the mineral morphology, which correspond to the low velocity inferred from the seismic images.

In summary, it can be speculated that there is a hot material upwelling beneath the Tarim basin (Lei et al. 2007), originating from the lower mantle and penetrating the discontinuous interface of 670 km into the MTZ. This also provides electrical evidence for the possible existence of mantle plumes beneath the Tarim basin.

9.4 Properties of anomaly bodies beneath north China

For North China (around MZL observation), it can be seen from Fig. 8 that at greater depths, within the depth slice of 670–900 km, the high-conductivity body enlarges, which corresponds to a low velocity. Since the volcanic rock is widely distributed in Great Xing’an Mountains (Xu et al. 2013), this high-conductivity and low-velocity anomaly may related to the deep material sources of the volcanic activities. Therefore, this high-conductivity and low-wave-velocity anomaly zone may be the hot melten phase material.

10 Conclusions

The following are the main results and inferences obtained by combining electrical conductivity modeling and seismic imaging of the earth’s mantle structure:

The subducting Pacific slab can convey a large amount of water into the MTZ beneath East China, and the dehydration of the stagnant slab may result in a high conductivity of the MTZ.

A low-resistance and low-velocity body, different from the cold subduction slab, was discovered beneath the Japanese Islands. It could be partially melted material and may have played a certain role in the stagnation of the Pacific slab.

In the WMQ and surrounding regions, a large low-resistance and low-velocity body was discovered in the MTZ. It could be a hot upwelling material and may be associated with the Tarim mantle plume.

In future, all these C-responses processed in this study will be used for inverting a finer 3-D electrical structure of the MTZ and the uppermost lower mantle beneath China, which could provide a more reliable model for the deep structure of earth.