3.1 Data Requesting

The broadband teleseismic data used in the study were recorded by 43 stations including permanent stations from the II network (active since 1990) and temporary stations in the XL array (for the period from 2012 to 2016) distributed in the area of 47°N to 52°N and 95°E to 105°E (Figure 1). All the seismograms were requested from the Seismological Facility for the Advancement of Geoscience (SAGE) Data Management Center (DMC). To balance the quantity and quality of the requested data, the epicentral distance range used for data request is from Δ_min = 30° to Δ_max = 100°, and the cutoff magnitude (M_c) is defined by the epicentral distance (Δ) and focal depth (D) subjecting to the equation M_c = 5.2 + (Δ − Δ_min)/(Δ_max – Δ_min)−D/D_max, where D_max = 700 km (Liu & Gao, 2010).

3.2 Data Processing

Data processing procedures used in this study are described in the SI section and S. S. Gao & Liu (2014b), and are briefly summarized here. The three-component seismograms are filtered utilizing a 4-pole, 2-pass Bessel filter with corner frequencies of 0.02 and 0.2 Hz. To attenuate the prominent PP phase, an exponential taper function with a half-width of 30 s is applied, centered on the theoretical PP arrival time. The frequency-domain water-level deconvolution algorithm (Ammon, 1991; Clayton & Wiggins, 1976) is implemented on the vertical component for selected seismograms that exhibit first-arrival signal-to-noise ratio (SNR) exceeding 4.0 to compute radial RFs. Following visual inspection, a total of 13,307 RFs displaying distinct initial P pulses from 5,026 events (Figure S1a in Supporting Information S1) are utilized in the analysis. The RFs are spatially grouped into circular bins with a radius of 0.5° based on the ray-piercing point coordinates (Figure 1) calculated at the midpoint of the MTZ (depth of 535 km). These binned RFs undergo moveout correction and stacking to generate depth profiles for each 0.5° interval. While the vast majority of the d410 and d660 arrivals are robust (Figure 2 and Figure S2 in Supporting Information S1), in a few instances, the depth windows are manually adjusted considering the depth and morphology of arrivals in adjacent bins. Specifically, if there are multiple arrivals with similar amplitudes, the one that shows the maximum similarity in the depth and waveform is selected. For a given bin, the mean depth for a MTZ discontinuity and the SD value are calculated using 50 bootstrap resampling iterations.

4.1 Apparent Depths

The obtained MTZ discontinuity depths (Figure S3 and Table S1 in Supporting Information S1) using the aforementioned procedures are apparent rather than true depths due to the fact that the 1-D IASP91 Earth model is utilized in performing the moveout correction. Measurements for the d410 or d660 are obtained from 225 bins in total, including 217 bins for the d410, 214 bins for the d660 and 206 bins for both the d410 and d660. The measurements for each of the 225 bins can be found in Table S1 of Supporting Information S1. The spatially continuous resulting depths for the two MTZ discontinuities and the MTZ thickness are generated by applying a continuous curvature surface griding algorithm with a tension factor of 0.5 (Smith & Wessel, 1990).

Two sub-areas (A and B, Figure 3, Figure S3 and Table S2 in Supporting Information S1) display distinctive observations and thus are the focuses of the following discussions. Beneath Area A which is situated from the west to south of the Siberian craton, the d410 is apparently depressed by about 13 km and d660 is depressed about 30 km, leading to a MTZ thickness of 267 ± 13 km. Inside Area A, in the area marked as A1 (Figure 3), there is an area with greatly depressed d660. This area is located in the close vicinity of Cenozoic volcanic fields in the East Sayan Mountains (Figure 1). Beneath this area, the greatest MTZ thickening in the entire study area with an average thickness of 297 ± 5 km is detected, and the maximum thickness is about 302 km. The MTZ thickening is mostly caused by the apparent depression of the d660 which corresponds to statistically significant positive amplitudes above the baseline (Figure 2a).

For Area B which includes the Tuva-Mongolian Massif, an apparent d410 depth of 433 ± 9 km and an apparent d660 depth of 662 ± 14 km are revealed, leading to an ∼21 km MTZ thinning that is mostly caused by the depression of the d410 (Figure 2b). Relative to the adjacent bins, there is a slight shallowing of the d660 for the same bins with depressed d410 (Figure 2b).

4.2 Velocity Corrected Depths

The apparent depths are converted into velocity corrected depths by using two global (Lu et al., 2019; Obayashi et al., 2013) and three regional velocity models (Chen, Niu, Liu, Tromp, et al., 2015; Tao et al., 2018; Wu et al., 2021) according to the velocity correction procedure presented in S. S. Gao & Liu (2014a). The results can be found in Figure S4 of Supporting Information S1. Due to the significant disparity observed among these velocity models, we computed the average corrected d410 and d660 depths (Figure 3 and Table S3 in Supporting Information S1) for each bin from the five velocity models. The average corrected depths for Area A are 420 ± 6 km and 685 ± 14 km for d410 and d660, respectively, with a mean MTZ thickness of 265 ± 13 km. For Area A1, these values are 423 ± 2, 717 ± 4, and 294 ± 5 km; and for Area B, they are 428 ± 9, 657 ± 12, and 229 ± 11 km. The corrected results (Table S4 in Supporting Information S1) are within a few kilometers from the apparent measurements (Table S2 in Supporting Information S1) as a result of minor lateral variations in upper-mantle and MTZ velocities. As demonstrated in numerous receiver function studies of the MTZ, the corrected depths of the d410 and d660 are frequently positively correlated (e.g., S. S. Gao & Liu, 2014b). This is primarily due to insufficient correction of apparent depths, as the velocity anomaly amplitudes in most tomography models are lower than their true values due to the application of a spatial damping factor. Therefore, in the following discussion, we place greater emphasis on the corrected MTZ thickness rather than the individual corrected discontinuity depths.

5.1 Constraining the Spatial Distribution of Slab Remnants in the MTZ

The spatial variations in MTZ thickness and the depths of seismic discontinuities provide key insights into the geometry and depth penetration of the slab segments beneath the region. For Area A, slab remnants from previous plate convergence such as the closure of the Paleo-Mongol-Okhotsk ocean and subduction of the Pacific plate beneath Eurasia (Chen, Niu, Liu, & Tromp, 2015) are possibly the main causes leading to the thickened MTZ. Under the assumption that the Clapeyron slope for the d410 is 2.9 MPa/K and that for the d660 is −1.3 MPa/K for the post-spinel transition (Fei et al., 2004), an MTZ thickening of 17 km corresponds to a positive temperature anomaly of 144 K. If a scale factor (dVp/dT) of −4.8 × 10⁻⁴ km s⁻¹ K⁻¹ (Deal et al., 1999) is used, a Vp velocity anomaly of about +0.75% is expected. Among the various tomography models (Figure S5 in Supporting Information S1), the one by Wu et al. (2021) fits the estimated value the best. In addition, the spatial distribution of the high velocities in this model (Figure S5e in Supporting Information S1) fits the area with thickened MTZ the best among all the models shown in Figure 3. It is important to note that, due to the wide range of previously reported Clapeyron slopes (Tauzin & Ricard, 2014) and the temperature-pressure dependence of the scale factor, the estimated temperature and velocity anomalies here and in the next section may carry significant uncertainties. However, these uncertainties do not impact the primary objective of this part of the study, which is to delineate the spatial distribution of thermal anomalies rather than their accurate amplitudes.

5.2 Possible Thermal Upwelling Beneath the Tuva-Mongolian Massif

A key feature of the observed MTZ discontinuity depths is the significant depression of the d410 by 23 km, while the d660 remains at a nearly normal (662 ± 14 km) depth beneath the Tuva-Mongolian Massif (Area B). After velocity corrections, the d410 is depressed by 18 km and the d660 is uplifted by 3 km. The MTZ beneath Area B is approximately 21 km thinner than the global average (Figure 3). Such a characteristic has been observed elsewhere such as SE Alaska (Dahm et al., 2017) and the NW Tibetan Plateau (Miao et al., 2024), and can be interpreted as thermal upwelling originating from inside the MTZ, which depresses the d410 but has no effect on the d660. In Area B, a viable mechanism for the deepening of the d410 is thermal upwelling associated with the western edge of the subducted Pacific slab. This interpretation is supported by a low-velocity zone (LVZ) that extends from the surface to the upper MTZ (Chen, Niu, Liu, Tromp, et al., 2015; Tao et al., 2018; Wu et al., 2021). Assuming a Clapeyron slope of +2.9 MPa/K for the d410 (Bina & Helffrich, 1994), the temperature anomaly near the d410 is calculated to be +220 K, and using the scaling factor of Deal et al. (1999), the corresponding velocity anomaly near the d410 is about −1.2%, which is approximately consistent with the velocity models of Tao et al. (2018) and Wu et al. (2021) (Figure S6 in Supporting Information S1).

5.3 Slab Dehydration as a Possible Cause for the Origin of Intraplate Cenozoic Volcanism

The spatial distribution of Cenozoic intraplate volcanism in the study region, encompassing the Jom-Bolok, Azas Plateau, Hamar-Daban Range, Dgida Basin, Khanuy Gol, and Orkhon Gol, predominantly coincides with the periphery of Area A. Such a spatial correspondence suggests that slab dehydration could be the ultimate cause of the Cenozoic volcanism. Mineral physics investigations indicate that subducting slabs older than 50 million years possess the capacity to deliver water into the MTZ through hydrous minerals and dense hydrous magnesium silicates, particularly hydrous ringwoodite, and phases E, D, and H (Pearson et al., 2014; Thompson, 1992). As temperature increases within these slabs, the water retention ability of wadsleyite and ringwoodite diminishes (Ohtani et al., 2004), thereby triggering continuous and prolonged dehydration processes. This process involves the release of volatiles from hydrous minerals within the slab as it undergoes phase transitions in the MTZ, potentially triggering flux melting in the overlying mantle wedge. Therefore, slab dehydration is a possible formation mechanism of the widespread recent magmatism in the peripheral regions of Area A (Figure 4).

5.4 A Possible Lithospheric Drip Traversing the MTZ

The greatest MTZ thickening is observed beneath the East Sayan Mountains, situated just south of the Azas Plateau and Jom-Bolok Cenozoic volcanic fields, with an average MTZ thickness of 297 ± 5 km within a circular region approximately 220 km in diameter (Area A1 in Figure 3). This thickening is primarily caused by a depression of the d660. Higher-than-normal seismic velocities are also observed in the lower MTZ beneath this area (Figure 2a). A thickened MTZ can result from either low temperatures associated with a subducted slab (Gu et al., 1998) or elevated water content linked to older slab segments (Ghosh et al., 2013). Because high water content corresponds to lower-than-normal seismic velocities, which is not observed, we propose that the most plausible explanation for the nearly circular MTZ thickening in Area A1 is a lithospheric drip detaching into the lower MTZ, as detailed below.

Several independent observations support this interpretation. First, the presence of garnet-bearing minerals on the Azas Plateau points to an anhydrous mantle origin (Litasov et al., 2002), consistent with the expected geochemistry of a lithospheric drip rather than slab dehydration. Second, the Nb-Ta enrichment in local alkaline basalts suggests a high time-integrated 238U/204Pb (HIMU) component, indicative of recycled lithosphere (Aulbach et al., 2008; Litasov et al., 2002) and supports the involvement of an upwelling related to lithospheric dripping (Figure 4). Third, the anomalously thick lithosphere in Area A1, coupled with thin lithosphere in surrounding regions (Figure S7 in Supporting Information S1), can be explained by lithospheric material being drawn downward from adjacent areas. In contrast, the observed geometry and seismic characteristics do not favor a stalled subducted slab interpretation. Subducted slabs typically produce elongated zones of MTZ thickening, whereas the East Sayan anomaly is circular. Moreover, regions with elevated water content often correlate with negative seismic velocity anomalies, which are not observed here. Note that the strongest evidence supporting the lithospheric drip model is the thickened MTZ observed in this study within a circular-shaped area (Figure 3c), supplemented by geochemical observations. Although the circular lithospheric thickening shown in the Litho-1.0 global model (Figure S7 in Supporting Information S1) aligns well with the predicted geometry by the drip model, we acknowledge that the relatively low resolution of Litho-1.0 limits its ability to serve as definitive support on its own.

If we assume that (a) the drip was triggered by a passing mantle plume (Hales et al., 2005), (b) the commencement of volcanism during the Mid-Miocene epoch at about 16 Ma (Rasskazov, 1994) was associated with lithospheric thinning in areas surrounding the lithospheric drip, and (c) the Eurasian plate has been maintaining the current westward movement and 21 km/myr rate (Gripp & Gordon, 2002) since 16 Ma, we can estimate that the current location of the mantle plume is located about 350 km east of Area A1. Coincidentally, various seismic tomography studies (Wu et al., 2021; Zhao et al., 2006) have revealed a column of low seismic velocities extending from the MTZ to the uppermost mantle beneath the western Lake Baikal (Figure 1). The low velocities present an elliptical or circular shape and are interpreted as indicative of an active mantle upwelling. Furthermore, the region proposed as the current location of this plume is characterized by an anomalously thin lithosphere (Figure S7 in Supporting Information S1), providing additional evidence for the existence of the plume. Based on these observations, we suggest that the mantle plume now located beneath western Lake Baikal is the same entity responsible for the initiation of the lithospheric drip when Area A1 was directly above the plume several million years ago. Despite the seemingly strong supporting evidence from the circular-shaped MTZ thickness anomaly, seismic tomography, geochemical observations, lithospheric thickness estimates, and the apparent alignment of the proposed lithospheric drip with the “Baikal plume” and recent plate motion, a comprehensive geodynamic modeling effort is required to test the viability of the proposed drip model and to elucidate the interconnections among the various physical properties.