Upper Mantle beneath the Myanmar and Surrounding Tomography: New Insight into Plate Subduction and Volcanism

Abstract

Myanmar and its surrounding areas have complex topography and strong tectonic movement, which has always been a challenge to most geoscientists. We used teleseismic tomography to study the subsurface velocity structure in this area. We present a new P-wave tomographic model beneath Myanmar and the surrounding areas by inverting 129,788 arrival-time data recorded by 372 stations. We found an inclined high-velocity subducting plate beneath central Myanmar, where the dip angle becomes smaller near 25°~26°N, and the seismic depth is limited below 200 km. The Indian oceanic lithosphere is being detached from the Indian continental lithosphere, which limits the depth of the earthquake. The active Tengchong volcano is underlain by a prominent low-velocity (low-V) anomaly in the shallow mantle, which may be caused by the subduction and dehydration of the Burma microplate (or Indian plate). The formation of the Singu volcano is related to the mantle flow of the Qinghai–Tibet plateau and the tearing of the Indian plate. The Yangtze craton (beneath the Sichuan Basin) shows a high-velocity anomaly, and both the shallow and deep parts have been destroyed, which may be related to the upwelling of deep heat flow.

1. Introduction

Myanmar is located in the southern part of the east Himalayan tectonic junction and adjacent to the northern part of the Himalayan continental collision orogen. The Andaman Sea oceanic subduction belt is in the south, and the Yangtze and Sunda continents are dominated by continents in the east, which are at the key position of the transition from continental collision to oceanic subduction (Figure 1). Curray et al. were the first to show that the elongated Burma plate connected the Himalayas of the ongoing Indo–Asian continental collision with the Indo–Australian plate, which was subducted southward under the Sunda plate [1]. The Burmese terrane constitutes most of the western part of Myanmar and is almost completely covered by the China–Burma Basin. The China–Burma Basin that includes two sub-basin side troughs located in the east and west of the Western Pacific Basin is a north–south strike belt of the Late Cretaceous–Cenozoic sedimentary basin [2]. There are also two distinct north–south trending tectonic zones: the Indo–Burma Mountains and the Central Burma Basin. The inner Indo–Burmese wedge consists of a sequence of Neogene clastic rocks deformed into a series of asymmetric folds that rose with the Indo–Burma Ranges [3]. The India–Myanmar mountain range is located at the corner between the Sumatra–Andaman and India subduction zones. It has a complex structure and is seismically active (approximately 150 km in depth). The distribution of moderate-depth seismic activity outlines a Benioff zone extending eastward to a depth of approximately 150 km [4,5]. The Indian plate subducts northward, while the region between the trough and the Sagaing fault descends eastward beneath the Eurasian plate [6].

The early subduction of the Indian Ocean lithosphere may have pulled down the continental lithosphere of Myanmar. There may be a transition from the continent to the ocean in this subduction zone because the subducting plate below is still converging at a rate of 18 mm/year [7]. The collision will continue only when driven by far-field forces [8]. Therefore, two modes for the rate for subduction have been proposed in this plate margin. Several researchers suggested that active subduction continues in the Burma region [9]. However, other researchers suggested that subduction along the Indo–Burmese arc has slowed down or may have stopped [6,10]. The Indian plate is subducting steeply under the Indo–Burma Mountains and slowly under the eastern Himalayas [11]. In the transitional zone, the difference in buoyancy between the continental and oceanic lithosphere may lead to plate tearing and fracture [12]. Below 100 km depth, Yao et al. found that the dip angle of the slab in the south is ~15° larger than that of the slab in the north, suggesting a possible slab tearing [13]. There are three volcanoes in Myanmar: Popa, Monywa and Singu. Recent volcanic activity at Popa during the Holocene is evidence of continued subduction of the Indian plate [14]. The shallow part of the Tengchong area (<400 km) is an obvious low-velocity anomaly [15,16]. Geochemical studies show that He3/He4 is relatively high [17], and three high geothermal gradient zones in the Tengchong area indicate the existence of a magma chamber [18]. The formation of the Tengchong volcano is not only related to the subduction and dehydration of the Indian slab but also associated with the distribution of the rift zones [19]. However, there is no unified result on how the Tengchong volcano was formed. The Yangtze craton has a widespread Archean basement overlain by a shallow crust partially reworked in the Proterozoic time [20]. The low heat flow at the surface (~40–60 mW/m²) suggests the Yangtze craton is a very stable block [21,22]. The Yangtze craton may have destructed due to the lateral extrusion of the upper mantle material under SE Tibet [23] and mantle upwelling at the base of the craton [24].

Seismic tomography can image deeper velocity structures, focal areas and fault structures that image seismic velocity structures beneath central Myanmar. For a long time, many researchers have conducted a lot of excellent work on Myanmar; they have studied the subduction system of Myanmar and the origin of the Tengchong volcano, but they have not reached a unified conclusion. Compared with previous work, in this study, the number of stations, seismic events and travel-time residuals in the selected data set is the largest, the ray coverage is more intensive, and the velocity structure model of 760 km underground is obtained. It provides a new way to solve the focus problems in this area.

2. Data and Method

2.1. Data

The sources of the seismic data used in this paper are the Incorporated Research Institutions for Seismology (IRIS) and the Global EHB database [25]. In order to meet the needs of continuous inversion calculation, we adopted the following standards: (1) The magnitude must be greater than 5.0, each event for ISC data must have a minimum of 15 arrivals and IRIS data should have a minimum of 5 arrivals; (2) All events must have epicenters within 30° to 90° of 25°N, 95°E so as to minimize the influence of the complex structure of the lower mantle and the core-mantle boundary (Figure 2); (3) After determining the source position in the 1D model, the absolute value of the travel-time residual of the sample data shall not exceed 4 s. Finally, 11,797 effective seismic events recorded by 372 stations were collected; in the subsequent inversion, 126,574 P-wave relative travel-time residuals were used.

To add the raw seismic waveform data obtained from IRIS to the dataset in this article, it was necessary to extract the relative travel-time residuals data. To improve the efficiency and accuracy of extracting the relative travel-time residuals, we chose the multi-channel cross-correlation method [26] and the adaptive stacking method [27] to stack the waveform data and compare the differences of the results obtained by the two methods. Data with large differences between the values or waveforms was eliminated. Compared with a single method of extracting data, the dual-method approach greatly improves the accuracy of the data and ensures the reliability of the final inversion results.

Figure 3 shows the average values of the relative travel-time residuals of seismic events received by various seismic stations in the region, ranging from −2.1 s to 2.1 s. Most of the seismic waves in the southern Shan Plateau and the southern Qinghai–Tibet Plateau arrive delayed, indicating that there is a low-velocity anomaly in the upper mantle. The early arrival of seismic waves in the Sichuan Basin indicates that there is a high-velocity anomaly in the upper mantle in this area. These can be used to test the reliability of the inversion results.

2.2. Method

Teleseismic tomography uses relative arrival times from distant earthquake sources recorded across an array of seismic stations to image the seismic structure of the upper mantle. It has been frequently used in various parts of the world [28,29].

In this study, we chose the following method (Figure 4). First, we determine the study area and assume that all travel-time residuals are caused by local anomalies. Second, the model space to be inverted is divided into grid nodes, and the initial one-dimensional model is replaced by grid points. Perturbation values of velocity or slowness are taken as unknown parameters to be solved. When seismic rays pass through a series of grid points, each grid point will affect the propagation time of the rays. Third, we use the ray-tracing technique in laterally inhomogeneous medium to calculate the difference between the theoretical travel time and the observed travel time from the source to the ground station. The travel-time residual is the sum of all travel-time perturbations on the ray, therefore, each seismic ray corresponds to a linear equation. There are many rays passing through each grid node; thus, the joint observation equations can be established. Finally, the velocity of each block or grid node is obtained by solving the equations, and the three-dimensional velocity structure distribution of the study area is obtained.

We used the software package FMTT [30,31] to perform an iterative nonlinear tomographic inversion of the subsurface P-wave velocity changes. The forward model was based on the Eikonal solver of the grid, namely, the fast marching method (FMM) algorithm, which is used to calculate the numerical value of the grid. It tracks the evolving interface along the narrow-band nodes and updates each grid node by solving the Eikonal equation using the finite difference approximation of the gradient vector satisfied by the upwind entropy. The main advantage is that it can generate reliable travel-time predictions in regions containing highly nonuniform velocity structures [32]. Our forward modeling was performed in parallel, which improves efficiency and facilitates travel-time prediction for large data sets. The inversion uses a subspace inversion scheme that adjusts the model velocity-node values to satisfy the data observations and travel-time residuals for damping and smoothing regularization constraints and eliminates non-uniqueness of the solution [33]. The subsurface structure is complex, and the velocity interface fluctuates greatly. Structure within the local 3D-model volume beneath the seismic array is represented using smoothly varying cubic B-spline volume elements, the values of which are controlled by a mesh of velocity nodes in spherical coordinates.

3. Checkerboard Resolution Tests

We used the same spatial distribution of stations and events as the final retrieval dataset and chose the checkerboard model to test the resolution of our dataset. We allocated 3% of the alternating positive and negative relative velocity values to the initial model. In order to ensure the accuracy of the inversion results, we chose different grid spacing to test and finally determined that the best grid spacing of horizontal inversion is 0. 7° and 0. 7°. In the vertical direction, we tested the grid resolutions of 70 km, 80 km and 85 km. Finally, the best inversion model was 0.7° × 0.7° × 85 km. The inversion grids were set vertically at depths of 100, 200, 300, 400, 500, 600 and 700 km in the vertical direction (Figure 5). A checkerboard consists of four inversion grids. Checkerboard inversion results show that our data set has a good recovery effect in 80% of the study area. The recovery effects of 300 km in the ocean area, Qinghai–Tibet Plateau and northern and western Shan Plateau are poor (Figure 5a–c). The main reason is that these regions have few stations and sparse rays. However, the shallow part of these areas is not the target area of our study. The Yangtze craton, the Burma region and the Tengchong area have been restored well. The more intensive the ray density and the greater the depth, the better the imaging effect, which provides support for us to study the deep state of the subducting plate and the heat source of the Tengchong volcano. Overall, our dataset is satisfactory in terms of checkerboard resolution. This provides good support for the study of hot issues in Myanmar and its surrounding areas.

4. Results

To apply the tomographic inversion scheme to the data set, we chose ak135 as a one-dimensional reference model and used the local model body to represent the velocity structure under the array. We divided the subsurface velocity structure into 21,645 velocity nodes with an approximate distance of 70 × 70 × 85 km between latitude, longitude and depth. The model spanned 2000 km and had a depth of 765 km. The data set, consisting of stations, seismic events and 129,788 relative arrival-time residuals, was subjected to five iterations of the tomographic inversion scheme to produce a solution model that satisfied the data to an acceptable level. The solution model reduced the data variance by 47.6% from 0.3596 s² to 0.1884 s², which corresponded to an RMS reduction from 387 ms to 201 ms (Figure 6).

Figure 7 shows the variation of P-wave velocity in Myanmar and the surrounding areas. The Indian subducted plate beneath Burma is a north–south, crescent-shaped and high-velocity anomaly, which is consistent with the findings of many authors (Figure 7a–c) [8,11,13,23]. With the increase in the subduction depth, the subduction dips to the east and seems to have a maximum depth of up to 600 km. The Himalayan tectonic zone located on the west side of the subduction shows a high-velocity anomaly in the east–west direction extending for about 200 km (Figure 8f). The two high-velocity anomalies are separated below 200 km and connected above 200 km. Anomalies of moderately high velocities are found at depths of 300~500 km beneath the Lhasa block (90°E, 31°N). The resolution test shows that the resolution of the structure at this depth is relatively high and has nothing to do with the shallow mantle structure. There is a low-velocity anomaly under the Tengchong volcano, which extends about 400 km (Figure 8f). There is an obvious high-velocity anomaly beneath the Yangtze craton, but the images of the Yangtze craton at different depths show variations (Figure 8h), and the shallow range is small and surrounded by low-velocity anomalies (Figure 7a–c). High-velocity anomalies appear on both sides of the Red River fault in the southeast corner of the area. At deeper depths, they may be the same high-velocity anomaly (Figure 7a–c). Under the Red River fault zone, there (102°E, 18°N) is an obvious NW–SE trending structure (Figure 7a–d) that disappears at 500–600 km. This NW–SE trend would be consistent with the orientation of the convergent margin of south Asia prior to collision of India [34]. An obvious low-velocity anomaly extending 300~400 km appears under the Thai volcanic cluster. As the depth increases, the low-velocity anomaly moves northward and, finally, an obvious low-velocity anomaly (103°E, 23°N) appears.

5. Discussion

The larger the data set, the better the effect of constraining the subsurface velocity structure. Based on our imaging results, the subduction of the Indian plate under Burma, the genesis of the Tengchong volcano and the related problems of the Yangtze craton are discussed.

5.1. Subduction Beneath Burma

Figure 8 shows the subducting plate beneath Myanmar. In the north–south direction, the Indian subduction under Myanmar occurs at 21°N and disappears at 27°N with the subduction depth ranging from 400 to 600 km, and the deepest may be more than 600 km. The subduction depth is consistent with the results of Yao et al. [13] and Ivan Koulakov [35], which are deeper than Replumaz et al. [36]. At depths greater than 100 km, its dip southward increases from ~60° for the northern segment to ~70° for the southern segment (Figure 8e,f). These variations in the plate dip are consistent with the abrupt changes in the Bouguer gravity anomaly (−150 to 175 mgal) toward the eastern part of the Northern Indoburman Range [37]. The position where the dip angle becomes smaller appears near 25°~26°N, which is consistent with the result of Khan [38]. Yao et al. showed that the angle changed at 22°~23°, which is 500 km north of the intersection of the Arakan Trench and the Bay of Bengal [13]. Comparing the position of the trench with the ophiolite belt on the eastern margin of the Indo–Burma Ranges, this trench recedes 200 km further westward than the southern trench [4]. Yao et al. assume that the steepening of the dip is related to the retreat of the trench [13]. Figure 8e shows that the northern part of the Indian subducting plate entered the MTZ under the Tengchong volcano, and there was expansion and stagnation. The subduction zone seismicity is consistent with the distribution of high-velocity anomalies in central Myanmar, which only extend to 200 km (Figure 7b–f). The earthquake termination may be due to plate dehydration embrittlement or runaway thermal shear stress [39]. The hydrated minerals in the subduction zone become anhydrous at shallow depths [40]. Plate tearing is most likely to occur at a depth of approximately 200 km, which is the place where the deepest seismic activity in the Wadati–Benioff zone along the Myanmar arc terminates [41]. The subduction part of the Indian lithosphere below the Indo–Burma Mountains clearly observed in the sectional image seems to have detached from other parts at approximately 25°N [11]. The Myanmar microplate is located at the transition of the subducting Indian plate from the ocean to the continental lithosphere. The continental lithosphere of the Indian plate subducted below the Himalayan–Tibet orogenic belt [42,43,44], and the oceanic lithosphere of the Indian plate subducted eastward below the Myanmar plate [33,45].

The continental crust may subduct to a depth of 150–200 km [46]. The nature of the deep and shallow subducting plate is different, the buoyancy is different, and it is easy to tear [12]. The eastward subduction and northward movement of the Indian plate promoted tearing [47]. Our tomography results show that, below 100 km depth, the dip angle of the slab in the south is ~15° smaller than that of the slab in the north.

We assume that slab tearing is expected to occur at oceanic slab–slab junctions [48,49]. This slab tearing is most likely to occur at a depth of ~200 km under the Sagaing fault, where the deepest seismicity of the Wadati–Benioff zone along the Burma arc terminates. The deep-sea crust sinks into the MTZ with the influence of negative buoyancy (Figure 8d,e). During continental subduction, the buoyancy and strength of the lithosphere are conducive to the separation of the marine lithosphere and the adjacent continental lithosphere [50]. The angular changes observed on the profile may prove that the continental and oceanic plates had separated, which also led to the termination of the deepest focal depth of the earthquake.

5.2. The Tengchong Volcano

The Tengchong active volcano is located in the eastern edge of the convergence zone between India and Eurasia. The last eruption was in 1609, and the area is thought to be still active. In recent years, several researchers suggested that the origin of the Tengchong volcano was affected by plate subduction [8,15,16,51]. The Holocene calc-alkaline eruption of the Popa volcano is volcanic evidence of the continuous subduction of the Indian plate [14]. The mid-deep earthquake (>150 km) may indicate that the oceanic crust of the Indian subduction plate is dehydrating. This inferred dehydration will result in a large amount of magma normally observed in subduction zones [13]. Below the Indo–Burmese Mountains, the descending part of the subducting plate is separated from the curved part and sinks at a steep angle [11]. The results of Lei et al. show that the low-velocity anomaly below the Tengchong volcano originated at a depth of ~400 km; it is speculated that the Tengchong volcano is a rift-related volcano formed by the subduction and dehydration of the Indian plate and the action of mantle wedge flow [19]. However, some authors assume that it originated from the melting of the mantle beneath Myanmar. Two different models have been proposed for the formation of Quaternary volcanic rocks in Tengchong and Myanmar: 90°E ridge subduction [52] and the roll-back followed by slab “break-off” [53]. Our results suggest that the deep subduction of the Indian plate at this location may have contributed to the formation of the Tengchong volcano. According to the latest geochemical research, the Cenozoic volcanic rocks related to the rift in the Tengchong area were formed by partial melting of enriched mantle sources, possibly due to the assimilation of oceanic crust subduction plates and sediments in the Neo–Tethys Basin. These volcanic rocks have distinct isotopic characteristics [54]. Incompatible trace element and Sr–Nd–Pb isotope compositions reflect the suprasubduction-zone fluid enrichment of the asthenospheric mantle wedge beneath the Burma–Tengchong terrane as a consequence of eastward under thrusting of the Indian continental lithosphere following the India–Asia collision at 55 Ma [55].

According to a study [56], the MTZ below central Myanmar (25.5°N) exhibits low conductivity, indicating that the MTZ may contain water. Meanwhile, the receiver function results show that the upper boundary of the MTZ is convex upward and the lower boundary is concave downward [38]. This position is shown as a low-velocity anomaly in the slice, which excludes the influence of the subduction zone. This is because the mineral water content increases the stability interval [57]. This also proves that the MTZ under the central part of Myanmar is wet. Our profile results indicate that the subducting slab has entered the MTZ and undergone extensive dehydration (Figure 8e). When the subducting slab interacts with the wet MTZ, this triggers partial melting and forms low-velocity seismic waves and inland volcanoes in front of the trench [58]. Our results show that the magma of the Tengchong volcano comes from the MTZ, and the dehydration of deep subduction slabs accelerates the upward migration of high-temperature materials. (Figure 9 and Figure 10).

The Sagaing fault shows an approximate 20 mm/yr right-lateral offset along the eastern edge of the Burma plate [4]. The latest geochemical research shows that the magma of Singu is a mixture of two different melts from the Indian and Tibetan asthenospheres [12]. The strike-slip fault may trigger the present magmatic activity and promote the upward migration of mantle melt [52]. Most interpretations of the volcanism in these regions are related to return flow from the top of the lower mantle due to the long history of subduction in the area [61,62,63,64]. The tomograhic image from Li et al. and Raoof et al. indicates that the upwelling of the Indian asthenosphere is probably triggered by the slab detachment where the Indian oceanic lithosphere is being detached from the Indian continental lithosphere [8,11]. Our results show that there may be a low-velocity corridor from the mantle transition zone to the underside of the Singu volcano (Figure 8e, red arrow). Although our results do not have enough resolution, several researchers had similar results [8,11,25]. This may be magma formed in the transition zone between the subducting plate and the mantle, pushed by strike-slip faults through the torn Indian subducting plate to the Singu volcano (Figure 9 and Figure 10). Future high-resolution regional seismic studies as well as detailed geodynamic modeling may shed further light on this interesting phenomenon.

5.3. Destruction of the Yangtze Craton

The Yangtze craton is a very stable Precambrian craton, which is adjacent to the Qinling–Dabie orogenic belt and the North China craton in the north, and adjacent to the Longmenshan fault and the Red River fault and the Qinghai–Tibet Plateau in the west and southwest. We directly interpret the high-velocity anomaly in the northeastern corner of the study area as a stable Yangtze craton confined to the Sichuan Basin. Our images show that most of the Yangtze Craton has a persistent high-velocity anomaly at a depth of 300 km. The shape of the high-velocity anomaly is similar to a wedge, which widens horizontally with the increase in depth (Figure 8h). Some research results indicate that in the jelly sandwich model, the strength of the mantle lithosphere decreases with increasing depth [65,66,67]. There is a significant mantle flow in southeastern Tibet [23,68]. Some research results suggest that the Indian plate continuously extrudes the upper mantle material forming a significant mantle flow in southeastern Tibet and causing damage to the shallow part of the Yangtze craton [23]. However, according to the latest geochemical results, the material flowing out of the Qinghai–Tibet Plateau can directly reach the Andaman Sea [12]. After Myanmar has undergone extensive subduction, the subducting plate has entered the MTZ and formed a stagnant plate [69]. Therefore, subduction promotes large-scale upwelling of the lower mantle, thereby reducing the velocity and significantly extending the upper mantle [18]. As happened under the North China craton, the stable craton was destroyed [44,70]. Our results show that the Yangtze craton is surrounded by low-velocity structures originating from the MTZ, so we prefer that the shallow and bottom parts of the Yangtze craton are affected by heat upwelling from the mantle transition zone or deeper (Figure 8h). Therefore, they need to be further confirmed by geophysics and numerical simulation.

6. Conclusions

A large number of high-quality teleseismic travel-time residuals in Myanmar and its surrounding areas were used to determine the high-resolution, three-dimensional imaging model of the region. The results provide new insights into the outstanding geological problems in this area.

The Myanmar microplate lies in the transition zone of the subducted Indian plate from oceanic to continental lithosphere. Beneath the transition zone, the Indian subducted plate is torn, and part of the oceanic plate has entered the mantle transition zone. At depths greater than 100 km, under the action of different negative buoyancy, its dip southward increases from ~60° for the northern segment to ~70° for the southern segment.

There is an obvious low-velocity anomaly under the Tengchong volcano, which extends to 400 km underground. The subduction of the Indian plate into the wet mantle transition zone resulted in the extrusion of a large amount of water, resulting in the melting of the upper mantle. Movement of slip faults and dehydration of stagnant plates both promote magma upwelling. This provided the conditions for the formation of the Tengchong volcano.

The low-velocity anomalies are likely upwelling mantle through gaps in the deep slab and provided the conditions for the formation of the Singu volcano. However, further geodynamic study is required to better understand them.

The Yangtze craton shows high-velocity anomalies. The shallow and bottom parts are damaged, which may be related to the upwelling of deep magma.