【作者】 吴耀；

【导师】 金振民； Yingwei Fei；

【作者基本信息】 中国地质大学 ， 构造地质学， 2009， 博士

【摘要】 自上世纪80年代以来，在世界范围内主要的陆-陆碰撞造山带榴辉岩中均发现了柯石英（假象）、金刚石（假象），以及高压矿物出溶体，表明这些陆壳岩石曾至少经历了压力＞3-4 GPa（相当于120-150 km上地幔深度）的高温高压变质作用，并折返出露于地表。这一发现极大地改变了人们对传统地壳变质作用温压范围的认识，使得榴辉岩和与其密切相关的超高压变质作用成为目前国际地学界研究的热点。与此同时，在陆-陆碰撞带超高压岩石中发现越来越多的证据，如含钾单斜辉石出熔体，α-PbO₂结构金红石，尤其是超硅石榴石表明，这些现今出露于地表的岩石可能来源于＞200km的地幔深部。近年来地震层析（Seismic tomography）资料已经明确揭示了直达下地幔底部的大洋深俯冲带的存在。表明镁铁质的大洋板片甚至可以俯冲至核幔边界。目前，人们对这些经历超高压变质作用的岩石能够折返回地壳深度并出露于地表的地球动力学机制尚缺乏了解。浮力作用折返可能是超高压榴辉岩的重返地壳深度的重要手段。然而，已有的高温高压实验研究结果并不支持镁铁质岩石（榴辉岩和橄榄岩）在自身浮力作用下发生折返。俯冲的洋壳玄武岩在进入上地幔直到核幔边界，其密度一直大于地幔岩石。同样在橄榄岩体系中，随着压力增加，俯冲板片密度亦大于同深度的地幔岩石，超高压镁铁质岩石不可能通过自身浮力折返。因此，超高压榴辉岩和橄榄岩折返过程中需要巨量低密度陆壳物质作为载体，并提供上浮力。然而，由于超高压镁铁质岩石的围岩—各种片麻岩和大理岩等抗后期变质作用叠加和风化能力不强，导致其超高压变质作用并不明显，在超高压变质作用产生之初，曾经有过对其围岩“原地”（in situ）和“异地”（foreign）成因的激烈争论。但随着超高压榴辉岩围岩片麻岩锆石中柯石英和金刚石包体的发现，超高压变质岩及其围岩“原地”成因已经为越来越多的人所接受。特别是最近在阿尔金变质岩中发现斯石英假象，表明长英质岩石曾经俯冲至上地幔350 km而后又折返回地表。然而，与相对稳定的大洋深俯冲不同，低密度的大陆地壳物质是否能够发生深俯冲并折返回地表的动力学过程是当前超高压岩石动力学研究的难点所在。要解决这一问题，需要了解表壳岩石在不同俯冲深度的高温高压相变行为，从而了解深俯冲陆壳与对应深度上地幔橄榄岩密度之间的关系。国际上目前对天然表壳岩石的高温高压实验研究还非常少，缺乏系统和深入的详细研究，这在很大程度上制约了我们对陆壳深俯冲和折返动力学过程的认识和理解。本次研究中我们使用活塞-圆筒高温高压装置（中国地质大学）和多面砧高温高压实验装置（美国卡内基研究所）进行压力范围为3．5-24 GPa，温度为750-1800℃，相当于上地幔100 km至地幔转换带底部（700 km）深度的高温高压实验研究。实验起始材料为大别山双河地区黑云母斜长片麻岩和二长花岗质片麻岩（粒度15-20μm）。获得了如下实验结果：压力为3．5-6 GPa实验，相当于深俯冲陆壳岩石进入上地幔120-200 km深度，黑云斜长片麻岩相变为主要由柯石英、硬玉、石榴石和多硅白云母组成的柯石英硬玉岩，其中初始样品中黑云母和钾长石基本被消耗生成叶片状的多硅白云母，仅有小部分残留的钾长石存在。而在高温实验中，黑云母／多硅白云母发生分解生成具有myrmekitic结构的石榴石+K-mica+熔体的集合体。表明，深俯冲陆壳物质在进入软流圈后，在较低的温度条件下多硅白云母将（Si=3．5-3．8 pfu）成为主要的含钾矿物。此外，3．5 GPa时原始样品中的绿帘石仍然稳定存在。当压力增加到6 GPa时，绿帘石在低温条件下（＜900℃）转变为硬柱石，而当温度升高时绿帘石发生分解生成钙铝榴石。这一压力范围产物最重要的特点是生成了两种赋存形式的柯石英。一种具有完整晶貌的柯石英，另一种则是以蠕虫或针状与硬玉形成交生结构（Symplectitic stucture）；9-14 GPa实验：相当于上地幔300 km至地幔转换带顶部。随着压力的增加，9 GPa时产物中出现斯石英，同时伴随多硅白云母的脱水反应生成K-锰钡矿。在这一压力段陆壳片麻岩相变为以斯石英+硬玉+K-锰钡矿+石榴石为主要组成矿物的斯石英硬玉岩。多硅白云母逐渐分解形成一种富钾的含水矿物K-mica，而硬柱石在9-14 GPa压力条件下已不能稳定存在；14-18 GPa实验：相当于整个地幔转换带压力范围。实验产物中最显著的特点是含水矿物如多硅白云母、K-mica均已消失。起始黑云斜长片麻岩转化为由斯石英+硬玉+K-锰钡矿+高硅石榴石组成的硬玉斯石英岩，并且不存在含水矿物；24 GPa实验：对应深度相当于下地幔顶部（660 km）。片麻岩俯冲至下地幔顶部的矿物组合为斯石英+硬玉+K-锰钡矿+石榴石+钙钛矿。实验新生成高压矿物的化学成分受到温度和压力的控制。单斜辉石／硬玉：cpx出现在每个实验产物中。在较低压力条件下（3．5-14 GPa），cpx中NaAlSi₂O₆的含量约为55-65％。当压力到14 GPa时，辉石中硬玉（NaAlSi₂O₆）接近90-100％。而压力为14-24 GPa范围时，单斜辉石成分未发生显著变化；石榴石：随着压力增加，石榴石中Si含量逐渐增高，3．5 GPa时石榴石中Si=3 pfu（每分子含量），当压力增加到6 GPa时，石榴石中Si=3．1，从6 GPa到9 GPa，Si含量从3．1增加到3．22，9 GPa至14 GPa，石榴石中Si含量保持为3．22不变，而当压力增加到18-24 GPa时石榴石分子中Si含量增加到3．3。Al在石榴石中含量则与Si具有相反的趋势，随着压力增加，石榴石中A1含量降低。此外，石榴石Ca和Mg的含量则强烈受控于与之共生的K-锰钡矿和Ca-钙钛矿，9 GPa时，K-锰钡矿出现，大量的Ca溶解于锰钡矿中形成Ca₂Al₂Si₂O₈，导致石榴石中钙铝榴石含量降低，相应的镁铝榴石含量增加。当压力超过18 GPa时，实验产物中出现另一个新的含Ca高压矿物Ca-钙钛矿，石榴石中钙铝榴石含量进一步减少；林根石在K-锰钡矿中的溶解度：常温常压条件下钾长石和云母是陆壳岩石中最重要的两种含钾矿物。然而钾长石和云母在高压下均将最终转变为K-锰钡矿。因此，当陆壳岩石至8-9 GPa时（相当于上地幔250-300 km深度），K-锰钡矿将成为钾元素在俯冲带中最主要的载体。在相同温度条件下，随着压力增加，K-锰钡矿中KAlSi₃O₈含量逐渐减低，Lingunite含量增加，到24 GPa，1400℃时K-锰钡矿中含有近20％的Lingunite。而CaAl₂Si₂O₈含量在低温下似乎与压力无关。压力相同时，K-锰钡矿中Lingunite和CaAlSi₂O₈随温度增加而迅速增加（图3．16）。当压力24 GPa，1800℃时K-锰钡矿中分别含有42％的Lingunite和1 1％的CaAl₂Si₂O₈。本实验在较低的温度条件下成功生成了KAlSi₃O₈含量最低的K-锰钡矿。可能表明，在多矿物体系中K-锰钡矿和Lingunite具有更宽的固溶区域；斯石英中Al和Fe的含量：压力小于9 GPa实验中，产物中的二氧化硅矿物为石英/柯石英，其电子探针成分分析结果显示SiO₂含量＞99．5％，而Al12O₃，Fe₂O₃，MgO和CaO含量则＜0．1％；当压力增加到＞9 GPa时斯石英开始稳定存在，斯石英中SiO₂随着压力增加而逐步减少，而Al₂O₃，Fe₂O₃，MgO和CaO含量则随压力增加而增高，尤其是Al₂O₃在斯石英中的溶解度更是升高一个数量级以上。在9-14 GPa压力范围内，斯石英中Al₂O₃约为0．15-0．45％，Fe₂O₃含量为0．1-0．15％，两者之间不存在某种特定的比列关系。Al₂O₃在斯石英中的溶解度还显示强烈的温度相关性，在18 GPa，1200℃时，斯石英中Al₂O₃含量为1．15％，而相同压力下1600℃时Al₂O₃含量增加到2．15％，24 GPa，1800℃条件下斯石英中Al₂O₃含量达到最高为4．23％。我们通过质量平衡研究了深俯冲陆壳片麻岩在上地幔不同深度时的矿物组成。表壳岩石主要由长石+石英+云母和其他矿物如角闪石，石榴石，榍石和绿帘石等组成。随着表壳岩石进入俯冲带，这些矿物逐渐被高压矿物所取代。深度小于120 km时，俯冲表壳岩石中斜长石分解产生硬玉，形成硬玉石英岩；当表壳岩石俯冲深度进入柯石英稳定域内（＞120 km深度），石英相变为柯石英，随着压力增加生成多硅白云母和硬柱石，此时俯冲岩石转变为柯石英硬玉岩，其中石榴石含量随压力增加而增加，而柯石英含量几乎不变，多硅白云母为主要的含钾矿物，多硅白云母和硬柱石为主要的含水矿物。在深俯冲陆壳岩石中硬柱石一直稳定至约9 GPa，相当于上地幔270-300 km，此后随压力和温度增加发生熔融转变为石榴石+熔体；而多硅白云母则具有更高的压力稳定性，并可稳定至上地幔转换带顶部（14GPa），在9-14 GPa（300-410 km）范围内，多硅白云母逐渐发生分解生成另一种结构未知的富钾云母K-mica+石榴石+K-锰钡矿。这一压力区间，柯石英将相变为斯石英。当深俯冲陆壳物质进入上地幔转换带后，陆壳岩石体系中所有的含水矿物将不再稳定。此时的俯冲岩石将转变为相对“干”的含K-锰钡矿和石榴石的斯石英硬玉岩。当然一些名义上的无水矿物如硬玉、斯石英和K-锰钡矿将扮演水载体的作用，将水输运至更大的上地幔深度。从520 km地震波不连续面（18 GPa）至下地幔顶部（24 GPa），深俯冲陆壳岩石中的主要矿物组合为斯石英+K-锰钡矿+硬玉+Ca-钙钛矿+石榴石，仍然为斯石英硬玉岩。我们计算了陆壳岩石沿三条不同地温曲线发生深俯冲的密度剖面。考虑到陆壳俯冲的平均地温梯度小于洋壳俯冲带，我们以冷的洋壳俯冲带温度曲线代表陆壳俯冲带的平均地温。陆壳岩石沿这三条地温曲线发生深俯冲过程中最大的区别在于随着温度的增加，含水矿物的种类和稳定性发生变化。低密度表壳岩石（地表平均密度约2．7 g／cm³）俯冲至3．5-6GPa，相当于上地幔120-200 km深度时密度将随温度变化增大至3．0-3．2 g／cm³，而同深度地幔橄榄岩的密度为3．4 g/cm³；如果表壳岩石进一步俯冲，俯冲物质密度将迅速增加，8 GPa时（250km）沿冷地温曲线俯冲的陆壳岩石密度已经与地幔橄榄岩密度相当（3．4 g／cm³），而沿正常地幔温度曲线俯冲的陆壳岩石要到9 GPa时（～300 km）才具有与地幔橄榄岩相当的密度；当压力大于9GPa时，俯冲陆壳岩石中出现斯石英和K-锰钡矿，导致俯冲岩石密度急剧增加至3．7-3．8 g／cm³，远大于此深度处的地幔橄榄岩（3．4-3．5 g／cm³）；从9 GPa（300km）至660 km地幔转换带底部（24 GPa），俯冲陆壳岩石密度随压力呈线性增加至～3．9 GPa，在这一深度区间上地幔橄榄岩中将出现两种新的高压矿物瓦兹利石（～14 GPa）和林伍德石（～18 GPa），在660 km地震波不连续面深度时地幔橄榄岩的密度与俯冲陆壳岩石的密度相近。但随着压力继续增加，进入下地幔时，林伍德石分解形成镁-铁钙钛矿和方镁铁石（（Mg，Fe）SiO₃+（Mg，Fe）O），下地幔顶部岩石密度将＞4．3 g／cm³，将远远大于深俯冲陆壳岩石。综上分析，我们在国际上第一次对天然陆壳副片麻岩进行了涵盖整个上地幔温度和压力范围的高温高压相变动力学实验研究，获得了如下四方面具有重要科学意义的结论：（1）大陆深俯冲带中表壳长英质岩石在压力为3．5-9 GPa（相当于上地幔120-300 km深度）范围内转变为柯石英硬玉岩，并含有大量含水矿物（主要为多硅白云母和绿帘石／硬柱石）；当压力为9-24 GPa时，深俯冲陆壳岩石将最终转变为超硅石榴石斯石英硬玉岩（密度为3．7-3．9 g／cm³）并导致俯冲的陆壳物质密度超过此深度处的地幔橄榄岩（3．4-3．7 g／cm³）；（2）通过实验表明巨量表壳物质深俯冲作用是可以发生的。陆-陆的碰撞俯冲首先应该在俯冲洋壳驱动下发生，如果俯冲洋壳能够将陆壳低密度岩石拖拽至250-300 km（临界深度），此后俯冲陆壳岩石密度将比相同深度地幔橄榄岩大约0．2-0．3 g／cm³，因相变导致的重力失稳将使陆壳岩石进一步俯冲至上地幔底部（660 km），并最终被迫停留在660 km深度参与地幔的化学演化；（3）在低密度的大陆地壳岩石深俯冲过程矿物相变和分解导致岩石密度增加是造成俯冲带重力不稳定的重要因素。作者认为250-300 km是大陆深俯冲物质能够通过上浮力发生折返的最大深度，这一深度也与绝大多数出露于地表的UHP岩石的最大深度是一致的。本次研究表明密度差异是陆壳物质能够发生深俯冲作用并折返回地表的一个至关重要的驱动力；（4）多硅白云母和绿帘石／硬柱石是深俯冲陆壳岩石中最主要的两种含水矿物，在大陆深俯冲过程中能够携带0．5wt％左右的水（相当于地表陆壳岩石含水量的一半）进入上地幔深部。当深俯冲陆壳岩石进入地幔转换带时，所有的含水矿物将不再稳定存在，但名义上无水矿物单斜辉石（硬玉）、K-锰钡矿和含Al斯石英依然能够容纳0．5 wt％的水，表明深俯冲陆壳岩石在进入地幔转换带后不太可能发生大规模的脱水事件。本文的研究结果表明，在前期洋壳俯冲作用的带动下巨量表壳物质深俯冲作用是可以发生的。“浮力折返”可能是俯冲深度小于250-300 km陆壳岩石的重要折返机制。考虑到陆壳俯冲带尚存在一定量的榴辉岩和其他镁铁质岩石，实际上这一深度可能略小于上述值。一旦大陆地壳在俯冲洋壳的拖拽下进入这一“临界深度”，那么此时即使下覆俯冲洋壳因拆沉作用而与陆壳分离，俯冲陆壳依然能够在自身重力作用下继续俯冲到660km上-下地幔边界。在深俯冲过程中陆壳岩石中矿物相变和分解导致岩石密度增加是造成俯冲带重力不稳定的重要因素。更多还原

【Abstract】 Previous works have demonstrated that deep subduction of continental crust is required toexplain the formation of coesite,micro-diamond and other index minerals in the ultra-highpressure metamorphic（UHPM）rocks from continental affinities.The possibility of continentalcrust being subducted into greater depth;perhaps even to the mantle transition zone has also beenproposed with the findings of oriented exsolutions of ilmenite and chromite/magnetite andtitanite and omphacite containing exsolved coesite rods and plates.More recently,subduction ofcontinental crust into depth of at least 350 km and return to the surface has been suggested withthe finding of former stishovite pseudomorphy in UHPM rocks from the Altyn Tagh,westemChina and relevant experiments.The idea of deep subduction of continental crust,along withUHP eclogite has also been supported by geochemical isotope signatures,and more recentlycoesite-bearing zircons from paragneisses within UHPM terrains.Subduction and recycling ofthese components into the mantle play a significant role in the evolution of mantle heterogeneityin terms of trace element abundances,volatile contents and radiogenic isotope systems.The mineralogy of subducted continental sediments at high pressure has been inferred fromhigh P-T experiments performed on compositionally analogies of average continental crust.Inorder to avoid complication related to reaction kinetics and to easily establish chemicalequilibrium,most of the previous experimental studies carried out on gels,glasses and sinteredoxide mix starting materials with bulk compositions close to that of pelagic and terrigenoussediments and hypothetical continental crust.Only few experiments were performed usingnatural rock powders as starting material.Experimental results suggest that at shallow depth thesubducted continental material,mainly consisting of coesite,clinopyroxene,orthoclase,garnet,and minor hydrous phases,is buoyant compared to the surrounding mantle in terms ofzero-pressure density.The intrinsic buoyancy of continental crust would oppose the entrainmentof subducted slab and would provide buoyancy-based exhumation force for UHMP rocks.However,it is not well established experimentally that to what depth this buoyancy contrastwould be reversed,thus the fate of the subducted continental crust in deep mantle is questionable.In addition to the continental crust,oceanic trench where a few hundred meters to severalkilometer thick sediments eroded from continental crust and deposited on the seafloor is anotherenvironment in which continental crust material may enter deep part of the upper mantle.It hasbeen shown that the subducted oceanic crust becomes much denser than the surrounding mantlethroughout the upper mantle,the sedimentary layer could be delivered,along with the basaltic crust,into the deep mantle and produce remarkable geochemical heterogeneous.Another poorly constrained issue is the maximum stability of some hydrous minerals andthe role of continuous dehydration reactions in the subducted continental crust.The emergingquestion of water recycling at continental subduction zones involve not whether water is recycledbut how it is recycled.What phases are involved in which reactions under what conditions?Phase transformations in the subducted continental slabs are mostly described on the basis ofexperiments in the simplified model systems.In these systems,solid solutions are inhibited bythe absence of proper chemical components.This would further favor discontinuous dehydrationreaction which implies a focused release of fluid at specific uninvariant P-T conditions.Incontrast,solid solutions are more common in natural minerals,thus continuous reaction plays anessential role in volatile releasing in complex natural systems,even though some hydrousminerals do not show significant compositional variability.The breakdown reaction of somehydrous minerals may involve a number of solid solutions and is therefore continuousdehydration.Consequently,in most natural systems dehydration reactions in the subducted slabsare continuous,and fluid release tends to be smeared out over a wide pressure and temperaturerange rather than at a specific P-T conditions.Thus,it requires experimental investigations onnatural rocks to better understand these complex issues.In order to determine the mineralogy and density of subducted upper continental crust,wecarried out experiments on natural biotie bearing para-gneiss and granitic gneiss powders fromShuanghe UHP terrain,Dabie Mt.,Eastern China.Those rocks are compositionally similar to thatof the classic rocks from UHPM terrains over a wide pressure and temperature ranges of 3.5GPa-24 GPa and 750℃-1800℃,corresponding to the P-T path through the upper mantle.Theexperiments provide data on the subsolidus phase relations and buoyancy relationship of thesubducted upper continental crust and its surrounding mantle as a function of pressures andtemperatures.The fate of the subducted upper continental crust and water releasing andtransportation during continental subduction will be discussed in light of the new experimentaldata.We also explore the micro-structural and textural characteristics of the phase transformationas a result of breakdown reactions and replacement by high-pressure phases,especially ofhydrous minerals.The observations provide a close link between the high-pressure experimentsand the occurrences of natural UHPM minerals.The experimental charges show well-crystallized textures of the coexisted phasescharacterized with straight or flat grain boundaries and homogeneous chemical compositions,implying that close chemical equilibrium was achieved.The starting material was firsttransformed into an assemblage consisting of coesite+jadeite+garnet+phengite+epidote/lawsonite in the pressure range of 3.5-9 GPa at temperatures up to 900℃.Withincreasing pressure,the assemblage transforms to stishovite+jadeite+K-hollandite+garnet+phengite or K-mica depending on temperature in the pressure range of 9-14 GPa.Epidote/lawsonite was smeared out above 9 GPa,while phengite was eliminated at a pressure of14 GPa.The assemblage consisting of stishovite+jadeite+K-hollandite+garnet remains to bestable at pressures up to 18 GPa,where Ca-perovskite was identified.The mineral assemblages in the pressure range of 18-24 GPa showed minor variation.However,significant variations of thechemical compositions and modal proportion of minerals were observed.Jadeite and coesiteformed fine-grained aggregates as breakdown products of plagioclase in experiments at pressurerange 3.5-6 GPa.Both jadeite and coesite from these aggregates were identified by Ramanspectroscopy with the diagnostic peaks at 699 cm^-1for jadeite and 523 cm^-1for coesite.Therewere also porphyroblastic or coarse grained coesite,characterized by the most intensive Ramanpeak at 523 cm^-1,formed from the direct transformation of the original quartz in the startingmaterial.Symplectitic intergrowth of phengite with Si content of 3.5-3.8 pfu rimmed along theinterfaces of K-feldspar and biotite and of plagioclase and biotite,was observed in thelow-temperature experiments.At higher pressures and temperatures,fine garnet grains together with K-rich mica andK-hollandite formed myrmekitic microstructure indicate dehydration reactions of biotite.Coesitewas replaced by stishovite and K-hollandite was present in runs at pressures＞9 GPa.Similar tocoesite occurrence,stishovite appeared in two textures（fine and coarse grains）and wasconfirmed by its characteristic Raman shift at 759 cm^-1and 589 cm^-1.K-hollandite wasconfirmed by diagnostic Raman shift at 765 cm^-1.The fine grain stishovite is the product of thephengite/K-mica dehydration;whereas coarse grains were formed from the direct transformationof coarse grain quartz.Both phengite and K-mica were eliminated at 14 GPa through thereactions 5 and 6 above.Epidote/lawsonite was observed only in experiments below 9 GPa and attemperature＜1000℃.It was eliminated and replaced by garnet at higher temperature throughmelting.It is worth noticing that no major hydrous minerals were found in experimentsexceeding 14 GPa.The transformed assemblage consists of stishovite,jadeite,garnet andK-hollandite.There is no significant change of phase assemblage within this pressure range.The chemical compositions of minerals showed systematic changes with both pressure andtemperature.For example,clinopyroxene gradually dissolved into garnet which becameincreasingly majoritic and Na rich with increasing pressure between 14-24 GPa.Likewise,atconstant temperature,the jadeite content in pyroxene tends to increase with increasing pressure,while it decreases with increasing temperature at constant pressure.At pressures greater than 14GPa,the composition of clinopyroxene from the experiments approaches that of pure jadeite.Garnet contains a significant amount of majorite component（Si=3.08-3.30 pfu）at pressuresexceed 11 GPa as compared with stoichiometric garnet（Si=3 pfu）.The Ca content in garnetdecreases rapidly with increasing pressure as a result of the occurrence of CaSiO3 perovskite atpressures＞18GPa.The composition of coesite from experiments at 3.5 and 6 GPa were close tothat of pure SiO₂.Less than 0.1 wt% Al₂O₃ and Fe₂O₃ were soluble in coesite.The solubility ofAl₂O₃ and Fe₂O₃ in stishovite increases with pressure and it slightly increases with increasingtemperature at pressures exceeding 9 GPa where coesite transforms to stishovite.At 24 GPa and1800℃,stishovite contains 4.23 wt% Al₂O₃.The solubilities of Lingunite（NaAlSi₃O₈）andCaAl₂Si₂O₈ in K-hollandite showed strong positive correlations with increasing temperature.Themaximum solubility of Lingunite and CaAl₂Si₂O₈ in K-hollandite occurred at the highestpressure and temperature of this study（24 GPa,1800℃）.At this P-T condition,42 mol% NaAlSi₃O₈ and 11 mol% CaAl₂Si₂O₈ were dissolved into K-hollandite.We used third-order high-temperature Birch-Murnaghan（HTBM）equation of state,combined with mineral proportions present in subducted continental crust at elevated pressuresand temperatures obtained in this study,to calculate the density profiles of the subductedcontinental crust rocks along geotherms of cold subduction,hot subduction and normal mantle,respectively.The comparison of the calculated density profiles of the subducted continental crustwith that of surrounding mantle provides some insight into the fate of the subducted continentalcrust.The subducted continental crust is relatively buoyant compared to the surroundingmantle at depths showller than c.250 km.The intrinsic buoyancy of the subducted continentalcrust would oppose entrainment to greater depth,and the buoyancy force would be an importantexhumation mechanism.Our experimental results demonstrate that the density of continentalcrust would be equal to or denser than that of mantle rock when those rocks have beentransported to depth＞250 km（8-9 GPa）.Thus,the‘point of no return’derived from ourexperimental results would be located at 250 km-300 km.Accordingly,once these rocksbeing subducted together with eclogite and/or dragged by oceanic subduetion exceeding thecritical depth,the high-pressure phases in the subducted continental crust would creategravitational instability which would favor continental subduction into the lower part of thetransition zone.In that case,entrainment by the sinking slab could become the predominantinfluence until the base of the mantle transition zone,which may play an important role in thegeochemical evolution of the Earth.更多还原