Plate reconstructions reveal that two secular centers of convergence formed beneath eastern Eurasia and North America no later than 200 Ma. The cause of these convergence centers, which featured flat subduction, slab stagnation, and/or continental margin subduction, remains uncertain. Here, we propose that upper-mantle thermal inhomogeneity, particularly an anomalously cool Northern Hemispheric upper mantle, was a fundamental driver of this long-lived convergence. By considering the pattern of observed thermal inhomogeneity, our numerical models show that flow-induced asymmetrical subduction will tend to develop toward cold mantle domains, even when the subducting plate is buoyant. The models can reproduce the diverse subduction styles observed in the Northern Hemisphere by including proposed pre-subduction plate distributions and/or properties.Two major convergence centers gradually developed in the Northern Hemisphere since no later than 200 Ma, evidenced by the secular convergence of the Siberian, Mongolian, Tibetan, Indian, Arabian, and (paleo-) Tethyan and (paleo-) Pacific plates toward eastern Eurasia, as well as the convergence of the terranes and the Farallon and Pacific plates toward North America (Fig. 1; Hall, 2002; Müller et al., 2016). Surprisingly, these asymmetric subductions largely verged toward the two centers during the past 200 m.y. (Müller et al., 2016; Wan et al., 2019). Furthermore, a series of representative but “less usual” subduction events that involved buoyant slabs occurred around both centers, implying that the buoyancy contrast between adjacent plates was not the sole factor determining the subduction style. The observed less usual subduction styles include: (1) flat subduction and slab stagnation of paleo-Pacific, Farallon, and/or Pacific plates (Liu et al., 2010; Peng et al., 2021b; Wu et al., 2019); (2) ridge subduction events during (paleo-) Tethyan and (paleo-) Pacific subduction (Kapp and DeCelles, 2019; Müller et al., 2016; Peng et al., 2021a; Zhang et al., 2019); and (3) the subduction of young oceanic basins below older ones, e.g., the (proto-) South China Sea (SCS) basins (Arcay et al., 2020; Hall, 2002). Intuitively, some enduring “attractor” structures seem to have induced repeated subduction of surrounding plates beneath eastern Eurasia and western North America (Fig. 1).Various mechanisms have been proposed to explain the long-lived convergence processes in the Northern Hemisphere. Based on the observations that two large low-shear-velocity provinces (LLSVPs) exist below Africa and the Pacific plates, some studies have highlighted the role of ridge/plume push in fragmenting Pangea and further in dispersing tectonic plates away from the LLSVPs (Li and Zhong, 2009). Other groups have emphasized the role of slab pull. Specifically, subducted slabs could induce strong mantle downwelling that sucks surrounding tectonic plates downward (Becker and Faccenna, 2011; Peng et al., 2021b). However, given the nearly “equatorial” locations of the LLSVPs, the ridge/plume push models cannot directly explain why convergence centers only developed in the Northern Hemisphere. Regarding the slab-pull scenarios, a chicken-egg problem also exists, which is why the Northern Hemisphere should first foster the initial subduction (Fig. 1).Recently, Earth's upper mantle was found to have an inhomogeneous distribution of temperature (Adam et al., 2021; Debayle et al., 2020). Following the approach of Debayle et al. (2020), we recalculated the upper-mantle thermal state with a recent attenuation tomography model (Fig. 2; Karaoğlu and Romanowicz, 2018). As shown in Figure 2, (1) cooler and warmer mantle domains can have thermal contrasts of up to 300 °C, with the cooler domains largely lying below Eurasia and North America; (2) the upper-mantle flow tends to converge toward or diverge from cooler mantle domains in the Northern Hemisphere (Fig. 2), demonstrating a potential role for these thermal anomalies in shaping upper-mantle convection. Previous studies have proposed that thermal inhomogeneity–induced mantle convection can explain episodic plate reorganizations (King et al., 2002; Peng et al., 2021b). Here, based on numerical exploration, we further suggest that this mechanism could have played a fundamental role in determining the subduction polarity during long-term overall plate convergence.We explored 80 two-dimensional (2-D) models to investigate the effects of upper-mantle thermal inhomogeneity. In the model domain, an oceanic plate is separated from the major continental plate (with/without a craton) by a thin continental block (or a marginal basin) (Fig. 3A). The oceanic plate contains two portions: the subducted slab and the incoming ocean. We assume that the tip of the incoming ocean has a thermal age of 10–20 m.y., which is 10–50 m.y. younger than the continental margin (see sensitivity test summary in Fig. S5 in the Supplemental Material1), so that spontaneous subduction is inhibited (Stern and Gerya, 2018). The density contrast between adjacent plates and the inhomogeneous upper-mantle thermal state dynamically drives the model evolution. Once the two oceanic portions tear at the “weak seed” (Fig. 3A), the incoming ocean contacts the continental margin along a roughly vertical boundary. Thus, no oblique weak zone is preset to promote a certain subduction polarity. This setup mimics the scenario where the former subduction in the Northern Hemisphere has ceased, with the subsequent ocean's future subduction polarity left free to develop.To introduce thermal inhomogeneity into the upper mantle, we initially set a cold region below the continent 300 °C cooler than the background mantle. The cold region is separated from the oceanic upper mantle by the subducted slab (Figs. 2 and 3A). We further assumed that this cold region has experienced limited melt extraction; thus, it is mainly composed of pargasite (amphibole)-bearing peridotite (Green and Falloon, 2005; Niu and Green, 2018). The compositional density of a mixture of pargasite (6%, 3120 kg/m3; Mineralogy Database, http://www.webmineral.com/) and fertile peridotite (94%, 3390 kg/m3) is ~3374 kg/m3 (Green and Falloon, 2005). Therefore, for a thermal expansion coefficient of ~3 × 10−5 K−1, the net density of the cold mantle domain is ~0.42% larger than the background mantle. Without consideration of a compositional effect, a 140 °C temperature reduction would be able to achieve the same net density difference.Similar to previous studies (Peng et al., 2021b), strong mantle convection forms after the initial slab tearing, eliminating the barrier between the cold and hot mantle domains (Figs. 3B, 3E, 3H, and 3K). The downwelling in the cold mantle domain, not the slab, enhances the vigor of convection (Fig. 3). In the experiments, mantle convection induced by thermal inhomogeneity enhances the contraction between adjacent tectonic plates. Within 19 m.y., asymmetric subduction occurred in ~88% of the tested models toward the cold mantle domain. In ~43% of the models, the initial young incoming ocean subducted below the older one (see sensitivity test summary in Fig. S5). In cases with an initially homogeneous upper-mantle thermal state, the timing of subduction initiation could reach up to ten times longer, and subduction of the young plate below the older plate never occurred (Movies S5 vs. S6 in the Supplemental Material).After incorporating upper-mantle thermal inhomogeneity (Fig. 2), the numerical models predict that the subduction polarity will tend to develop toward the cooler upper-mantle domain (Fig. 3). Therefore, upper-mantle thermal inhomogeneity may have been a fundamental factor in determining the observed nearly unchanged Northern Hemisphere subduction polarity. Compared to models with ridge/plume push, since most of the current cooler upper-mantle domains exist at present in the Northern Hemisphere (Fig. 2), our models (also see King et al., 2002) may better explain why the post–200 Ma convergence centers mainly developed there. On the other hand, because atypical subduction largely occurred prior to peaks in regional subduction flux (Fig. 4A), the mantle convection induced by this thermal inhomogeneity could have triggered these events. In addition, simulated convergence rates after subduction initiation roughly reproduce the reconstructed values during Izu-Bonin-Mariana subduction (Fig. 4B; Müller et al., 2016), further demonstrating the potential robustness of these results.Among the tested models, most of the subducted oceanic plates develop steep subduction, which finally form a curved slab wall in the lower mantle (cf. run 48 in Fig. 3 vs. Sigloch and Mihalynuk, 2013). By further considering the proposed pre-subduction lithospheric distribution and properties (Hall, 2002; King et al., 2015; Larvet et al., 2023; Yang et al., 2019), these generic models also reproduce the diverse subduction styles observed in the Northern Hemisphere (cf. Fig. 1 vs. 3): In models where the continental interior remains intact, stagnation of the subducted oceanic plate in the mantle transition zone (MTZ) happens in varying ranges (run 72 in Fig. 3; Peng et al., 2021a). After considering the weakening effects of the slab-derived fluids/melts in the back-arc region (Yang et al., 2019) and a smaller compositional density of the oceanic plate (King et al., 2015), slab stagnation can reach >1000 km away from the trench (Fig. 2F). The average time of the slab's stagnation in the MTZ is ~24.4 m.y., and its sinking rate in the lower mantle is ~1.02 cm/yr, consistent with previous estimates for East Asia (Liu et al., 2017; Wu et al., 2016).In the models where the major continent breaks, and one continental fragment drifts oceanward, flat subduction of the incoming oceanic plate can quickly develop (run 37 in Fig. 3; Yang et al., 2019). In the later stages of flat subduction, the deeply subducted portion breaks up with the flat one, and the latter fragment then detaches from the overriding plate (Fig. 3J; Liu et al., 2010). At the end of the model run, the juxtaposition of nearly vertical slab structures is crudely like that seen beneath modern North America (cf. Figs. 1C vs. 3J).In models where a thinned continental margin (or marginal sea) subducts, the following continent can be further torn apart, causing the opening of a new marginal sea (Fig. 3M), e.g., the SCS (Hall, 2002; Larvet et al., 2023). The subduction direction is away from the cold mantle domain (run 61 in Fig. 3) because the large thermal contrast in the upper mantle is close to the continent-ocean boundary; thus, the induced lithospheric contraction focuses at that location. Due to the buoyancy of the margin slab, slab stagnation in the MTZ also develops (Fig. 3M). In these models, the time delay of 14 m.y. between the former subduction initiation and the opening of a new marginal basin is roughly consistent with that between the pre-SCS subduction initiation and its subsequent opening (Arcay et al., 2020; Hall, 2002).In models where the continental interior remains intact, stagnation of the subducted oceanic plate in the mantle transition zone (MTZ) happens in varying ranges (run 72 in Fig. 3; Peng et al., 2021a). After considering the weakening effects of the slab-derived fluids/melts in the back-arc region (Yang et al., 2019) and a smaller compositional density of the oceanic plate (King et al., 2015), slab stagnation can reach >1000 km away from the trench (Fig. 2F). The average time of the slab's stagnation in the MTZ is ~24.4 m.y., and its sinking rate in the lower mantle is ~1.02 cm/yr, consistent with previous estimates for East Asia (Liu et al., 2017; Wu et al., 2016).In the models where the major continent breaks, and one continental fragment drifts oceanward, flat subduction of the incoming oceanic plate can quickly develop (run 37 in Fig. 3; Yang et al., 2019). In the later stages of flat subduction, the deeply subducted portion breaks up with the flat one, and the latter fragment then detaches from the overriding plate (Fig. 3J; Liu et al., 2010). At the end of the model run, the juxtaposition of nearly vertical slab structures is crudely like that seen beneath modern North America (cf. Figs. 1C vs. 3J).In models where a thinned continental margin (or marginal sea) subducts, the following continent can be further torn apart, causing the opening of a new marginal sea (Fig. 3M), e.g., the SCS (Hall, 2002; Larvet et al., 2023). The subduction direction is away from the cold mantle domain (run 61 in Fig. 3) because the large thermal contrast in the upper mantle is close to the continent-ocean boundary; thus, the induced lithospheric contraction focuses at that location. Due to the buoyancy of the margin slab, slab stagnation in the MTZ also develops (Fig. 3M). In these models, the time delay of 14 m.y. between the former subduction initiation and the opening of a new marginal basin is roughly consistent with that between the pre-SCS subduction initiation and its subsequent opening (Arcay et al., 2020; Hall, 2002).There are several potential causes for the thermal inhomogeneity in the upper mantle (Figs. 2 and 4C). Previous studies have noted that the reconstructed locations of large igneous provinces younger than 200 Ma appear to be distributed mainly above the edges of the LLSVPs (Torsvik et al., 2010), implying that the LLSVPs may somehow have stable positions and that the heat from them may be the most critical source for hot upper-mantle anomalies (Fig. 4C). On the other hand, if craton-bearing regions possess a higher thermal conductivity than their surrounding mantle, as previously hypothesized (Petitjean et al., 2006), then their underlying upper mantle could be more effectively cooled. This possible cooling effect of cratons appears to be consistent with the possible drip-like structures that have been proposed beneath the North American cratons (Fig. 1C; Cawood et al., 2023) and the depressed elevation of most cratons in the Northern Hemisphere (Wang et al., 2022). In this perspective, relatively cold upper-mantle domains could be repeatedly created beneath cratonic regions.There is an additional potential feedback between this mode of upper-mantle flow and flow in the lower mantle. The LLSVPs may be the upwelling axial “polar regions” of an exceptionally stable degree-2 pattern of convection in the lower mantle (Morgan and Vannucchi, 2023). In this hypothesis, long-lived lower-mantle subduction preferentially occurs in the corresponding “equatorial” downwelling ring of this degree-2 flow structure (Morgan and Vannucchi, 2023). When cooler upper-mantle regions overlie this ring of lower-mantle downwelling—as they currently do in the Northern Hemisphere—then long-lived upper-mantle downwelling could easily occur (Fig. 4C). In contrast, if a region of upper-mantle downwelling were to migrate above an upwelling lower-mantle LLSVP, then upper-mantle subduction would become inhibited, and the region of upper-mantle downwelling would either migrate laterally or shut off (Fig. 4C).We acknowledge the helpful suggestions from John Armitage on adding the periodic boundary condition. This study was supported by the National Natural Science Foundation of China (grant nos. 92355302, 41890812, 42104105, 42074114, and 42150710533), the Tuguangchi Award for Excellent Young Scholar (TGC202101), and the “14th Five-Year Plan” independent plan of the Guangzhou Institute of Geochemistry, Chinese Academy of Sciences.

中文翻译：

---

由上地幔热不均匀性驱动的北半球长寿辐合系统

板块重建表明，在欧亚大陆东部和北美下方不迟于 200 Ma 形成了两个长期的汇聚中心。这些汇聚中心以平坦俯冲、板片停滞和/或大陆边缘俯冲为特征的原因仍不确定。在这里，我们提出，上地幔的热不均匀性，特别是异常凉爽的北半球上地幔，是这种长期汇聚的根本驱动力。通过考虑观察到的热不均匀性模式，我们的数值模型表明，即使俯冲板块有浮力，流动引起的不对称俯冲也将倾向于向冷地幔域发展。这些模型可以通过包括拟议的俯冲前板块分布和/或属性来重现在北半球观察到的各种俯冲样式。自不晚于 200 Ma 以来，两个主要的辐合中心在北半球逐渐形成，这可以从板块的长期辐合证明。西伯利亚、蒙古、西藏、印度、阿拉伯、（古）特提斯和（古）太平洋板块朝向欧亚大陆东部，以及地体和法拉隆板块和太平洋板块朝向北美的汇聚（图 1；霍尔） ，2002 年；穆勒等人，2016 年）。令人惊讶的是，在过去的 200 年里，这些不对称俯冲基本上向两个中心靠拢（Müller 等，2016；Wan 等，2019）。此外，在两个中心周围发生了一系列涉及浮力板片的代表性但“不太常见”的俯冲事件，这意味着相邻板块之间的浮力对比并不是决定俯冲类型的唯一因素。观察到的不太常见的俯冲类型包括：（1）古太平洋板块、法拉隆板块和/或太平洋板块的平坦俯冲和板片停滞（Liu等，2010；Peng等，2021b；Wu等，2019） ; （2）（古）特提斯和（古）太平洋俯冲期间的山脊俯冲事件（Kapp和DeCelles，2019；Müller等，2016；Peng等，2021a；Zhang等，2019）； (3) 年轻的海洋盆地俯冲到较古老的海洋盆地之下，例如（原始）南海（SCS）盆地（Arcay 等人，2020 年；Hall，2002 年）。直观上，一些持久的“吸引子”结构似乎引起了欧亚大陆东部和北美西部下方周围板块的反复俯冲（图1）。人们提出了各种机制来解释北半球长期存在的汇聚过程。基于对非洲和太平洋板块下方存在两个大型低剪切速度省（LLSVP）的观察，一些研究强调了山脊/羽流推动在盘古大陆破碎以及进一步使构造板块远离 LLSVP 方面的作用（Li和钟，2009）。其他团体也强调了板坯拉力的作用。具体来说，俯冲板块可能会引起强烈的地幔下降流，将周围的构造板块吸向下方（Becker 和 Faccenna，2011；Peng 等，2021b）。然而，鉴于 LLSVP 的位置接近“赤道”，山脊/羽流推动模型无法直接解释为什么辐合中心仅在北半球发展。对于板块拉动情景，同样存在先有鸡还是先有蛋的问题，这就是为什么北半球应该首先促进初始俯冲（图1）。最近，人们发现地球上地幔的温度分布不均匀（Adam et al.等人，2021；德拜尔等人，2020）。遵循 Debayle 等人的方法。 （2020），我们用最近的衰减断层扫描模型重新计算了上地幔热状态（图2；Karaoğlu和Romanowicz，2018）。如图2所示，（1）较冷和较温暖的地幔域的热对比可达300°C，较冷的域主要位于欧亚大陆和北美以下； （2）上地幔流倾向于向北半球较冷的地幔区域汇聚或发散（图2），这证明了这些热异常在形成上地幔对流方面的潜在作用。先前的研究提出，热不均匀性引起的地幔对流可以解释幕式板块重组(King et al., 2002; Peng et al., 2021b)。在这里，基于数值探索，我们进一步表明，这种机制可能在长期整体板块收敛过程中确定俯冲极性方面发挥了根本性作用。我们探索了 80 个二维 (2-D) 模型来研究上层板块的影响。 -地幔热不均匀性。在模型域中，海洋板块通过薄大陆块（或边缘盆地）与主要大陆板块（有/无克拉通）分开（图3A）。海洋板块包含两部分：俯冲板块和流入海洋。我们假设传入海洋尖端的热年龄为10-20 my，比大陆边缘小10-50 my（参见补充材料1中图S5中的敏感性测试摘要），因此自发俯冲为抑制（Stern 和 Gerya，2018）。相邻板块之间的密度对比和不均匀的上地幔热状态动态地驱动着模型的演化。一旦两个海洋部分在“弱种子”处撕裂（图3A），流入的海洋就会沿着大致垂直的边界接触大陆边缘。因此，没有预设倾斜弱带来促进一定的俯冲极性。这种设置模仿了北半球之前的俯冲已经停止的情况，随后海洋未来的俯冲极性可以自由发展。为了将热不均匀性引入上地幔，我们最初在大陆下方设置了一个温度低 300°C 的寒冷区域比背景地幔。冷区通过俯冲板块与大洋上地幔分开（图2和3A）。我们进一步假设这个寒冷地区经历了有限的熔体提取；因此，主要由含闪石（角闪石）的橄榄岩组成（Green and Falloon，2005；Niu and Green，2018）。细闪石（6%，3120 kg/m3；矿物学数据库，http://www.webmineral.com/）和肥沃橄榄岩（94%，3390 kg/m3）混合物的成分密度约为 3374 kg/m3（格林和法隆，2005）。因此，对于约 3 × 10−5 K−1 的热膨胀系数，冷地幔域的净密度比背景地幔大约 0.42%。在不考虑成分效应的情况下，温度降低140℃也能达到相同的净密度差。与之前的研究类似(Peng et al., 2021b)，板片初始撕裂后形成强地幔对流，消除了障碍冷地幔域和热地幔域之间（图3B、3E、3H和3K）。冷地幔域而不是板片的下降流增强了对流的强度（图3）。在实验中，热不均匀性引起的地幔对流增强了相邻构造板块之间的收缩。在 19 my 内，约 88% 的测试模型发生了向冷地幔域的不对称俯冲。在约 43% 的模型中，最初的年轻传入海洋俯冲到较旧的海洋下方（参见图 S5 中的敏感性测试摘要）。在最初具有均匀上地幔热状态的情况下，俯冲起始时间可能会延长十倍，并且年轻板块从未俯冲到较老板块下方（补充材料中的电影 S5 与 S6）。考虑到上地幔的热不均匀性（图2），数值模型预测俯冲极性将倾向于向较冷的上地幔区域发展（图3）。因此，上地幔的热不均匀性可能是决定观测到的几乎不变的北半球俯冲极性的基本因素。与脊/羽流推力模型相比，由于当前大多数较冷的上地幔区域目前存在于北半球（图 2），我们的模型（另见 King 等，2002）可以更好地解释为什么–200 Ma汇聚中心主要发育在那里。另一方面，由于非典型俯冲主要发生在区域俯冲通量峰值之前（图4A），因此这种热不均匀性引起的地幔对流可能引发了这些事件。此外，俯冲起始后的模拟收敛速率大致再现了伊豆-博南-马里亚纳俯冲期间的重建值（图4B；Müller等人，2016），进一步证明了这些结果的潜在稳健性。在测试的模型中，大多数俯冲的海洋板块发展出陡峭的俯冲作用，最终在下地幔中形成弯曲的板壁（参见图 3 中的运行 48 vs. Sigloch 和 Mihalynuk，2013）。通过进一步考虑所提出的俯冲前岩石圈分布和性质(Hall, 2002; King et al., 2015; Larvet et al., 2023; Yang et al., 2019),这些通用模型还再现了在北半球观察到的各种俯冲类型（参见图 1 与图 3）：在大陆内部保持完整的模型中，俯冲海洋板块在地幔过渡带 (MTZ) 中的停滞发生在不同的范围（图 3 中的运行 72；Peng 等人，2021a）。考虑到弧后区域板片衍生流体/熔体的弱化作用(Yang等，2019)和海洋板块较小的成分密度(King等，2015)，板片停滞可以达到>距海沟1000公里（图2F）。板片在MTZ的平均滞留时间约为24.4 my，其在下地幔的下沉速率约为1.02 cm/yr，与之前对东亚的估计一致(Liu et al., 2017; Wu et al., 2016）。在主要大陆破裂且一块大陆碎片向海洋漂移的模型中，传入的海洋板块的平坦俯冲会迅速发展（图3中的运行37；Yang等人，2019）。在平坦俯冲的后期，深俯冲部分与平坦俯冲部分分裂，然后后者从上覆板块分离（图 3J；Liu et al., 2010）。在模型运行结束时，几乎垂直的板状结构的并置粗略地类似于现代北美下方的情况（参见图 1C 与 3J）。在大陆边缘（或边缘海）俯冲变薄的模型中，接下来的大陆可能会进一步撕裂，导致新的边缘海的出现（图3M），例如南海（Hall，2002；Larvet等，2023）。俯冲方向远离冷地幔域（图3中的61），因为上地幔热反差大，靠近大陆-海洋边界；因此，引起的岩石圈收缩集中在该位置。由于边缘板片的浮力，MTZ 中的板片停滞也发生了（图 3M）。在这些模型中，前俯冲起始和新边缘盆地打开之间的时间延迟了14 my，与南海前俯冲起始和随后的打开之间的时间延迟大致一致（Arcay等，2020；Hall，2002） ）。在大陆内部保持完整的模型中，俯冲海洋板块在地幔过渡带（MTZ）中的停滞发生在不同的范围内（图3中的72；Peng等人，2021a）。考虑到弧后区域板片衍生流体/熔体的弱化作用(Yang等，2019)和海洋板块较小的成分密度(King等，2015)，板片停滞可以达到>距海沟1000公里（图2F）。板片在MTZ的平均滞留时间约为24.4 my，其在下地幔的下沉速率约为1.02 cm/yr，与之前对东亚的估计一致(Liu et al., 2017; Wu et al., 2016）。在主要大陆破裂且一块大陆碎片向海洋漂移的模型中，传入的海洋板块的平坦俯冲可以迅速发展（图 3 中的运行 37；Yang 等人，2019）。在平坦俯冲的后期，深俯冲部分与平坦俯冲部分分裂，然后后者从上覆板块分离（图 3J；Liu et al., 2010）。在模型运行结束时，几乎垂直的板状结构的并置粗略地类似于现代北美下方的情况（参见图 1C 与 3J）。在大陆边缘（或边缘海）俯冲变薄的模型中，接下来的大陆可能会进一步撕裂，导致新的边缘海的出现（图3M），例如南海（Hall，2002；Larvet等，2023）。俯冲方向远离冷地幔域（图3中的61），因为上地幔热反差大，靠近大陆-海洋边界；因此，引起的岩石圈收缩集中在该位置。由于边缘板片的浮力，MTZ 中的板片停滞也会发生（图 3M）。在这些模型中，前俯冲起始和新边缘盆地打开之间的时间延迟了14 my，与南海前俯冲起始和随后的打开之间的时间延迟大致一致（Arcay等，2020；Hall，2002） ）。上地幔热不均匀性有几个潜在的原因（图2和4C）。先前的研究指出，小于 200 Ma 的大型火成岩省的重建位置似乎主要分布在 LLSVP 的边缘上方（Torsvik 等，2010），这意味着 LLSVP 可能在某种程度上具有稳定的位置，并且来自它们可能是上地幔热异常的最关键来源（图4C）。另一方面，如果克拉通区域具有比其周围地幔更高的热导率，正如之前的假设（Petitjean 等，2006），那么它们下面的上地幔可能会被更有效地冷却。克拉通的这种可能的冷却效应似乎与北美克拉通下方可能存在的滴水状结构一致（图 1C；Cawood 等人，2023）以及北半球大多数克拉通的凹陷海拔（图 1C；Cawood 等，2023）。王等人，2022）。从这个角度来看，相对较冷的上地幔区域可能会在克拉通区域下方重复产生。这种上地幔流动模式和下地幔流动之间存在额外的潜在反馈。 LLSVP 可能是下地幔异常稳定的 2 级对流模式的上涌轴向“极区”（Morgan 和 Vannucchi，2023）。在这个假设中，长期的下地幔俯冲优先发生在这种 2 级流动结构的相应“赤道”下降环中（Morgan 和 Vannucchi，2023）。当较冷的上地幔区域覆盖在这个下地幔下降流环之上时——就像目前在北半球所做的那样——那么很容易发生长期存在的上地幔下降流（图4C）。相反，如果上地幔下降流区域迁移到上升流的下地幔LLSVP上方，则上地幔俯冲将受到抑制，上地幔下降流区域将横向迁移或关闭（图4C） ）。我们感谢 John Armitage 关于添加周期性边界条件的有用建议。该研究得到国家自然科学基金（批准号：92355302、41890812、42104105、42074114、42150710533）、土光池优秀青年学者奖（TGC202101）、“十四五”自主计划的资助中国科学院广州地球化学研究所.