Intracrystalline nucleation during the post-garnet transformation under large overpressure conditions in deep subducting slabs



[1] The mechanism of the post-garnet transformation in natural single-crystalline pyropic garnet has been examined under large overpressure conditions of ∼31–43 GPa using the multianvil apparatus with sintered diamond anvils. Intracrystalline nucleation was found to be dominant above 35–38 GPa, while only grain-boundary nucleation is responsible at lower pressures. In addition, natural pyropic garnet transformed to a single phase of perovskite without decomposing to multiple phases above 38 GPa. Both intracrystalline nucleation and polymorphic growth of the post-garnet transformation under large overpressure conditions would strongly enhance the transformation kinetics. Consequently, the post-garnet transformation may pwroceed rapidly in the subducting slab at depths near ∼950 km, which would cause substantial rheological weakening of the subducting slab. The metastable post-garnet transformation may also cause the seismic discontinuities at a depth of 900–1080 km depth, which is indeed observed beneath some subduction zones.

1. Introduction

[2] High-pressure transformation of mantle minerals is considered to be kinetically inhibited in a slab subducted into the deep mantle due to the low temperatures. For example, both experimental kinetic studies and seismological observations suggest that metastable olivine likely survives without transforming to its high-pressure polymorph wadsleyite beyond the equilibrium phase boundary in the coldest central part of the subducting slab [e.g.,Kaneshima et al., 2007; Mosenfelder et al., 2001; Rubie and Ross, 1994]. This transformation basically proceeds by interface-controlled growth of wadsleyite after nucleation at the grain-boundary [e.g.,Kubo et al., 2004; Rubie and Ross, 1994]. On the other hand, intracrystalline nucleation has also been found to occur at relatively large overpressure conditions far from the equilibrium phase boundary [Kerschhofer et al., 1996, 2000]. This change of nucleation mechanism greatly enhances overall transformation rates and would affect rheological weakening of subducting slabs [Mosenfelder et al., 2001].

[3] Garnet is one of the major constituent minerals in the subducting slab. Previous experimental studies suggested that garnet decomposes to multiple phases at the uppermost lower mantle (∼720 km depth) at equilibrium [e.g., Kubo and Akaogi, 2000; Litasov et al., 2004; Oguri et al., 2000; Sano et al., 2006]. The post-garnet transformation is considered to proceed by nucleation on grain boundaries followed by sluggish growth of decomposed phases [Kubo et al., 2008, 2002]. According to this transformation mechanism, a significantly large amount of metastable garnet is predicted to exist in the subducting plate far beyond the equilibrium boundary. However, the large overpressure conditions of the post-garnet transformation in a deep subducting slab may cause intracrystalline nucleation. Also, the growth mechanism of the post-garnet transformation may change since the decomposed assemblage after garnet is different depending on the pressure and temperature [Funamori et al., 2000; Miyajima et al., 1999]. Thus, both the nucleation and growth mechanisms in the post-garnet transformation should be examined when we consider the metastability of garnet in the slab subducting deep into the lower mantle.

[4] Here we report experimental results on the nucleation and growth mechanisms of the post-garnet transformation under large overpressure conditions >31 GPa obtained using a Kawai-type high-pressure apparatus combined with sintered diamond anvils. Based on the present experimental results, occurrence of the post-garnet transformation at possible depths in deep subducting slabs is discussed.

2. Experimental Procedures

[5] Single crystals of natural pyropic garnet cut into 250 μm cubes were used as the starting material. An electron microprobe analysis shows that the typical composition of the sample is SiO2 = 41.72, Al2O3 = 21.52, Cr2O3 = 2.06, MgO = 20.99, CaO = 4.45, and FeO = 8.62 in wt%. The material was surrounded by a fine powder of MgO and enclosed in a graphite capsule. All the experiments were carried out using a Kawai-type high-pressure apparatus (MADONNA-II, at Ehime University) with a similar setup toStagno et al. [2011]. Sintered diamond anvils with a truncated edge length of 1.5 mm were used as the second stage anvils in order to generate pressures greater than ∼30 GPa. The sample assembly is composed of sintered (Mg,Co)O pressure medium, a cylindrical LaCrO3heater, and a Mo electrode. Temperature was monitored with a W3%Re-W25%Re thermocouple. The effect of pressure on the emf of the thermocouple was ignored. The pressure medium and the heater were dried at 1000°C for 3 hours before assembling the high-pressure cell. The cell assembly was then stored at 120 °C in a vacuum oven before the experiments.

[6] The samples were compressed to pressures of 31–43 GPa at room temperature, then heated to 1000–1200°C for 1 hour. The heating and cooling rates were controlled to be about ∼100 K/min in all the experiments. After the experiments, recovered samples were mounted in epoxy resin and polished with diamond powder. The microstructures and chemical compositions of the recovered samples were examined using a field emission-scanning electron microscope (FE-SEM, JSM-7000F) and an analytical transmission electron microscope (ATEM) with an energy-dispersive X-ray spectrometer (EDS). Thin foils of typically 100 nm thick for ATEM analyses were prepared using a focused ion beam (FIB) system (JEOL JEM-9310FIB).

3. Results

[7] We carried out seven experiments at 31–43 GPa, which is far from the equilibrium boundary of the post-garnet transformation (∼24 GPa) [Kubo and Akaogi, 2000; Oguri et al., 2000]. All the samples were recovered before the reaction was completed. Back scattered electron (BSE) images of the recovered samples showed that the transformation occurred at the grain boundaries in all cases (Figures 1a and 1b). Intracrystalline nucleation was also observed in some run products from relatively high pressure conditions (Figures 1c–1f). Post-garnet assemblages of recovered samples consist of very fine grains of less than a hundred nanometers. The phase assemblages of some recovered samples were analyzed by EDS using ATEM (Figure S1 in the auxiliary material) based on the phase reactions reported in previous experimental studies [Funamori et al., 2000; Miyajima et al., 1999; Sano et al., 2006]. The texture of the decomposition assemblages was complicated and shows no structural coherency between the host and newly formed phases in both the intracrystalline and grain-boundary transformations. The phase assemblage in this study may metastable because of the low heating temperature and short annealing time. The experimental conditions, phase assemblages identified, and thickness of the reaction rim are summarized inTable 1.

Figure 1.

Microscopic images of the recovered samples. (a) A BSE image and (b) a TEM bright-field image of the sample recovered from the run at 31 GPa and 1200°C. The post-garnet transformation occurred at the grain boundary of garnet. (c) A BSE image of the sample recovered from the run at 35 GPa and 1200°C. (d) A BSE image and (e, f) TEM bright-field images of the samples recovered from 43 GPa and 1100°C. The electron diffraction pattern of parent garnet (bottom right in Figure 1e) is also shown. Intracrystalline nucleation along the dislocation cores (arrows in Figures 1e and 1f) in addition to the grain-boundary nucleation occurred due to the large overpressure in the latter runs. Abbreviations: gt, garnet; pgt; post-garnet assemblage; pc, periclase.

Table 1. Experimental Conditionsa
Run NumberPressure (GPa)Temperature (°C)Growth Distance (μm)Intracrystalline NucleationDecomposed Phases
  • a

    Decomposed phases were identified by EDS spectra taken from analytical TEM according to the post-garnet phase assemblages reported in previous studies [Funamori et al., 2000; Miyajima et al., 1999; Sano et al., 2006] The uncertainties were estimated from the standard deviations of ∼10 measurements.

  • b

    A small amount of intracrystalline nucleation was observed only in TEM observations.

  • c

    The phase was found only in intracrystalline nucleation.

  • d

    Almost single phases with very small amount of aluminous phase. Abbreviations: Mg-Pv, Mg-rich perovskite; Ca, Ca-rich phase (Ca-perovskite and/or CM-perovskite); St, stishovite; Al, aluminous phase (calcium-ferrite type phase and/or NAL phase); Co, corundum; Fp, ferropericlase; CMA-Pv, Ca-Ma-Al-perovskite.

PGS053112003.9 ± 0.4×Mg-Pv, Ca, St, Al, Co,
PGS0735120014.3 ± 1.9Mg-Pv, Ca, St, Al, Fpc
PGS033810001.4 ± 0.7× 
PGS043811007.0 ± 1.8bMg-Pv (CMA-Pv)d
PGS0238120026.8 ± 4.6 
PGS084111003.3 ± 0.7× 
PGS064311005.1 ± 0.6Mg-Pv (CMA-Pv)d

[8] At 31 GPa and 1200°C, the phase transformation occurred only at grain boundaries and no intracrystalline nucleation was observed (Figures 1a and 1b). The phase assemblage of the post-garnet transformation consists of Mg-perovskite, stishovite, corundum, an Al-rich phase (calcium-ferrite type phase and/or NAL phase), and a Ca-rich phase (Ca-perovskite or Ca-Mg(CM)-perovskite) [Asahara et al., 2005]. This phase assemblage is almost consistent with previous experimental studies on the post-garnet transformation [Miyajima et al., 1999; Oguri et al., 2000]. The breakdown to multiple phases and nucleation at the grain boundaries should cause slow kinetics of the post-garnet transformation [Kubo et al., 2008].

[9] At 35 GPa and 1200°C, intracrystalline nucleation was also observed in addition to grain-boundary nucleation (Figure 1c), indicating that the intracrystalline nucleation becomes responsible for the post-garnet transformation at such a pressure. The growth rate of the post-garnet assemblage following intracrystalline nucleation was slow compared to that associated with grain-boundary nucleation due to the different growth mechanism. Although the phase assemblage observed was almost the same as that of the run at 31 GPa (Figure S1), ferropericlase was found only as a product of intracrystalline nucleation. The content of Al2O3in Mg-perovskite clearly increased compared to that observed at 31 GPa, which is consistent with previous studies that demonstrated the alumina content in Mg-perovskite increases with increasing pressure [e.g.,Kubo and Akaogi, 2000; Hirose et al., 2001].

[10] At 38–43 GPa and 1100°C, both intracrystalline and grain-boundary nucleation were observed (Figures 1c–1f). TEM observations showed the presence of aligned dislocations from which the nucleation of post-garnet phase(s) was initiated (Figures 1e and 1f). The electron diffraction pattern of the parent garnet (bottom right in Figure 1e) indicates that dislocation cores appear to be aligned parallel to {111} planes. Such linear alignments of newly nucleated grains are also clearly visible in the BSE images (Figures 1c and 1d) taken from more extensive areas. Garnet transformed to an almost single phase of Mg-perovskite rather than decomposed to multiple phases due to the increase in Al and Ca contents in perovskite. This phase is possibly the metastable CMA-perovskite reported in previous studies at relatively low temperatures although the thermodynamic stability of this phase is still controversial [Funamori et al., 2000; Sano et al., 2006]. Note that the chemical compositions of the host garnet and transformed perovskite are almost identical (Figure S2).

4. Discussion

[11] The present results clearly demonstrate that intracrystalline nucleation becomes more responsible for the post-garnet transformation mechanism than grain boundary nucleation, when large overpressures are applied. Generally, transformation kinetics of minerals depend on the grain size of parent phases because the transformation occurs at the grain boundaries in many cases [e.g.,Kubo et al., 2008]. Therefore, intracrystalline nucleation strongly enhances the overall transformation rate by significantly reducing the effective grain size [Kerschhofer et al., 2000; Mosenfelder et al., 2001]. Consequently, the rate of the post-garnet transformation increases discontinuously at a depth where intracrystalline nucleation becomes predominant.

[12] Figure 2 shows the width of the reaction rim of the grain boundary against pressure (Figures 2a and 2b) and inverse temperature (Figure 2c). The growth rate of the reaction rims increases with increasing pressure in the range of 31–38 GPa under isothermal conditions (Figure 2a), while it is approximately constant above 38 GPa (Figure 2b). The increase of reaction rate with pressure (Figure 2a) is most likely due to the change of the growth mechanism in the post-garnet transformation. At the relatively low pressures of 31–35 GPa, the garnet decomposed to multiple phases. This decomposition should cause slow growth kinetics in the post-garnet transformation [Kubo et al., 2008]. On the other hand, garnet transforms to a single phase of perovskite at higher pressures above 38 GPa. Generally, such a polymorphic transformation proceeds by an interface-controlled growth mechanism, where the growth rate is constant and much faster than that of the decomposition growth [e.g.,Hogrefe et al., 1994]. Thus, the mechanism change from decomposition to polymorphic growth enhances the transformation kinetics in the present experiments (Figure 2a). The rate contrast of the two growth mechanisms would become much larger on geological time scales of slab subduction because of the significant difference in the time dependency.

Figure 2.

(a) Width of the reaction rim against pressure of 31–38 GPa at 1200°C, (b) against pressure of 38–43 GPa at 1100°C, and (c) against inverse temperature of 1000–1200°C at 38 GPa. The uncertainties were estimated from the standard deviations of about 10 measurements for each run.

[13] Figure 3shows the boundaries of nucleation and growth mechanisms of the post-garnet transformation in the present experiments. Both the peridotitic and MORB layers in subducting slabs yield large amounts of majoritic garnet (∼30–90 vol.%) under the pressure and temperature conditions of the mantle transition region [Ganguly et al., 2009; Irifune, 1993]. Considering the mechanisms of grain-boundary nucleation and the inward growth of decomposed phases at relatively low overpressure conditions (∼24–35 GPa), the post-garnet transformation cannot proceed at a substantial rate in geological time scales of 106–107 years at ∼1300°C when the slabs penetrate into the lower mantle [Kubo et al., 2008]. In this case, large amounts of metastable garnet survive in the colder subducting slab beyond the equilibrium post-garnet phase boundary. On the other hand, our study demonstrates that both the nucleation and growth mechanisms during the post-garnet transformation change at larger overpressure conditions above 35 GPa. Although further information on the nucleation and growth kinetics is required, these mechanisms can trigger a sharp and extensive post-garnet transformation at a pressure around 35 GPa due to the enhanced transformation kinetics if the dislocation densities of subducting garnet is sufficiently high. Also, the critical pressure for the post-garnet transformation by intracrystalline nucleation in subducting slabs may be lower than our experimental result due to the large difference in timescale between geological processes and laboratory experiments. Precise determination of the realistic transformation depth requires more quantitative information on the nucleation kinetics, which will be determined by carefully designed experiments in our future project. In addition, the critical pressure is also variable with the chemical composition of the original garnet. Thus, it is clearly different in the case in peridotitic and MORB layers [Irifune, 1993] and also influenced largely by the metastable persistence of non-majoritic garnet due to the slow kinetics of the pyroxene-garnet transformation [e.g.,Nishi et al., 2008].

Figure 3.

Boundaries of nucleation and growth mechanisms of the post-garnet transformation in the present experiment. The thin line and dashed line show the boundaries between grain-boundary and intracrystalline nucleations and between the decomposition and polymorphic transformation of garnet, respectively, which are determined in the present experiments. Filled symbols show the occurrence of intracrystalline nucleation, while open symbols indicate that only grain boundary nucleations are observed. Circles and triangles show the decomposition and the polymorphic transformation, respectively. Open squares show the grain-boundary nucleation and decomposition growth process examined by the previous kinetic study [Kubo et al., 2008]. The geothermal profile of the cold subducting slab is also shown (shaded arrow) [Kirby et al., 1996]. Asterisk indicates that intracrystalline nucleation was assumed to occur although we failed to observe it by BSE image because the nucleation was found in the run at lower pressure conditions using TEM.

[14] It has been recognized that the strength of the subducting slab decreased by the grain-size reduction of the constituent minerals through the phase transformations, especially when the transformation occurs under larger overpressure conditions [Karato et al., 2001; Rubie, 1984]. Because the volume fraction of garnet at the mantle transition zone (30–40 vol.%) is high enough to cause an interconnection [Wray, 1976], the bulk rheology of the subducting slab should be controlled by a weak post-garnet assemblage with a small grain size. Thus, strong rheological weakening of the subducting slab may be induced at a depth where the metastable post-garnet transformation occurs. Consequently, subduction of the slab into depths greater than 900 km (∼35 GPa) in the lower mantle may be prevented and stagnated over these depths. Slab stagnation far below the 660 km discontinuity observed in some seismic tomography models [Fukao et al., 2001; Niu et al., 2003] may reflect strong rheological weakening due to the metastable post-garnet transformation.

[15] Some other seismological observations indicate the presence of seismic discontinuities at depths of about 900–1080 km (35–42 GPa) [Kawakatsu and Niu, 1994; Niu and Kawakatsu, 1997]. Major phase transitions corresponding to the depths of such seismic discontinuities have been found neither in peridotitic mantle nor in subducted MORB [e.g., Ono et al., 2001]. The present study suggests that the metastable post-garnet transformation may also be responsible for these mid-mantle discontinuities in the lower mantle.


[16] We thank K. Fujino, T. Kubo, N. Nishiyama, and Y. Nishihara for their helpful comments. We are also grateful to W. Fulong for assistance with the high-pressure and high-temperature experiments. A constructive review by D.C. Rubie also improved the manuscript. M. Nishi was supported by Research Fellowships of the Japan Society for the Promotion of Science for Young Scientists.

[17] The Editor thanks David Rubie and Eiji Ohtani for their assistance in evaluating this paper.