Volume 126, Issue 12 e2021JB022437
Research Article
Open Access

Pressure Destabilizes Oxygen Vacancies in Bridgmanite

Hongzhan Fei

Corresponding Author

Hongzhan Fei

Bayerisches Geoinstitut, University of Bayreuth, Bayreuth, Germany

Correspondence to:

H. Fei,

[email protected]

Search for more papers by this author
Zhaodong Liu

Zhaodong Liu

Bayerisches Geoinstitut, University of Bayreuth, Bayreuth, Germany

State Key Laboratory of Superhard Materials, Jilin University, Changchun, China

Search for more papers by this author
Rong Huang

Rong Huang

Bayerisches Geoinstitut, University of Bayreuth, Bayreuth, Germany

Department of Earth Sciences, University College London, London, UK

Search for more papers by this author
Seiji Kamada

Seiji Kamada

Frontier Research Institute for Interdisciplinary Sciences, Tohoku University, Sendai, Japan

Department of Earth Science, Tohoku University, Sendai, Japan

Search for more papers by this author
Naohisa Hirao

Naohisa Hirao

Japan Synchrotron Radiation Research Institute (JASRI), Sayo, Japan

Search for more papers by this author
Saori Kawaguchi

Saori Kawaguchi

Japan Synchrotron Radiation Research Institute (JASRI), Sayo, Japan

Search for more papers by this author
Catherine McCammon

Catherine McCammon

Bayerisches Geoinstitut, University of Bayreuth, Bayreuth, Germany

Search for more papers by this author
Tomoo Katsura

Tomoo Katsura

Bayerisches Geoinstitut, University of Bayreuth, Bayreuth, Germany

Center for High Pressure Science and Technology Advanced Research, Beijing, China

Search for more papers by this author
First published: 19 November 2021
Citations: 3

Abstract

Bridgmanite may contain a large proportion of ferric iron in its crystal structure in the forms of FeFeO3 and MgFeO2.5 components. We investigated the pressure dependence of FeFeO3 and MgFeO2.5 contents in bridgmanite coexisting with MgFe2O4-phase and with or without ferropericlase in the MgO-SiO2-Fe2O3 ternary system at 2,300 K, 33 and 40 GPa. Together with the experiments at 27 GPa reported in Fei et al. (2020, https://doi.org/10.1029/2019GL086296), our results show that the FeFeO3 and MgFeO2.5 contents in bridgmanite decrease from 7.6 to 5.3 mol % and from 2 to 3 mol % to nearly zero, respectively, with increasing pressure from 27 to 40 GPa. Accordingly, the total Fe3+ decreases from 0.18 to 0.11 pfu. The formation of oxygen vacancies (MgFeO2.5 component) in bridgmanite is therefore dramatically suppressed by pressure. Oxygen vacancies can be produced by ferric iron in Fe3+-rich bridgmanite under the topmost lower mantle conditions, but the concentration should decrease rapidly with increasing pressure. The variation of oxygen-vacancy content with depth may potentially affect the physical properties of bridgmanite and thus affect mantle dynamics.

Key Points

  • MgFeO2.5, FeFeO3, and total Fe3+ contents in bridgmanite decrease with increasing pressure

  • Fe3+-linked oxygen vacancies in bridgmanite are destabilized by increasing pressure

  • MgFeO2.5 can be formed in Fe3+-rich bridgmanite under the topmost lower mantle conditions

Plain Language Summary

Bridgmanite is the most abundant mineral in the Earth's lower mantle. Although its basic chemical formula is MgSiO3, large amounts of Fe3+ can be added in the following two ways: (1) Two Fe3+ replace Mg2+ and Si4+ and form the FeFeO3 component. This is called charge-coupled substitution because the overall charge does not change. (2) One Fe3+ replaces one Si4+ and forms the MgFeO2.5 component. Here the loss of positive charge is compensated by a loss of oxygen and is therefore called oxygen-vacancy substitution. In this study, we measured the effect of pressure on the abundance of these two components of bridgmanite. We found that the MgFeO2.5 content decreases greatly with increasing pressure. Some oxygen sites may therefore be vacant in bridgmanite at the top of lower mantle, but the concentration of oxygen vacancies should decrease rapidly in deeper regions. The decrease of oxygen-vacancy concentration in bridgmanite will change the nature of the lower mantle, for example, rocks will become harder, and electrical conductivity will decrease with increasing depth.

1 Introduction

It is known that the disproportionation reaction of iron from Fe2+ to Fe3+ and Fe0 can occur in Earth's deep lower mantle (Armstrong et al., 2019; Frost et al., 2004). In particular, the separation of metallic iron and Fe3+-bearing silicate magma during core formation would raise the Fe3+/ΣFe ratio in silicate magma, resulting in high proportion of Fe3+ in deep mantle minerals that precipitated from the magma ocean (e.g., Andrault et al., 2018; Armstrong et al., 2019; Boujibar et al., 2016; Frost et al., 2008). On the other hand, the redox-induced density contrast may produce locally Fe3+-rich regions (Gu et al., 2016), whereas slabs may transport oxidized components into the deep mantle by subduction (Zhao et al., 2021). Therefore, it is expected that minerals in the deep mantle may contain large amount of Fe3+ (Fe3+/ΣFe up to 60% or more), at least locally (e.g., Armstrong et al., 2019; Bindi et al., 2020; Boujibar et al., 2016; Frost et al., 2004; Grocholski et al., 2009; Gu et al., 2016; Jackson et al., 2005; Kupenko et al., 2015; Kurnosov et al., 2017; Lauterbach et al., 2000; Li et al., 2006; McCammon, 1997; Piet et al., 2016; Prescher et al., 2014; Shim et al., 2017; Sinmyo et al., 2011), although they are under relatively reducing conditions with oxygen fugacity close to the iron-wüstite buffer (Frost & McCammon, 2008). Since Fe3+ may affect the chemical and physical properties of minerals by changing their defect chemistry (e.g., Creasy et al., 2020; Fei et al., 1994; Glazyrin et al., 2014; Holzapfel et al., 2005; Liu et al., 2018; Sinmyo et al., 2019; Wang et al., 2021), knowledge of mineral phase relations in Fe3+-rich systems is critical for investigating the structure, dynamics, and evolution of Earth's mantle.

Bridgmanite is stabilized in the pressure range 23–125 GPa (e.g., Ishii et al., 2018; Murakami et al., 2004) and is the dominant mineral in Earth (e.g., Irifune & Ringwood, 1987a1987b). It can incorporate large amounts of trivalent elements such as Al3+ and Fe3+ in its crystal structure (e.g., Andrault et al., 1998; McCammon, 1997; Navrotsky, 1999; Navrotsky et al., 2003; Shim et al., 2017) by the formation of XXO3 and MgXO2.5 components (X is Fe3+ or Al3+) via charge-coupled and oxygen-vacancy mechanisms, respectively (e.g., Huang, Boffa-Ballaran, McCammon, Miyajima, Dolejš & Frost, 2021; Huang, Boffa-Ballaran, McCammon, Miyajima, & Frost, 2021; Lauterbach et al., 2000; Liu, Akaogi, & Katsura, 2019; Liu, Ishii, & Katsura, 2017; Liu, Boffa-Ballaran, et al., 2019; Liu et al., 2020; Navrotsky, 1999; Navrotsky et al., 2003; Nishio-Hamane et al., 20052008; O'Neill & Jeanloiz, 1994). The different substitution mechanisms thus produce different types of defect species. The defect-controlled physical properties of bridgmanite such as atomic diffusivity, elasticity, plasticity, and electrical conductivity will depend on the substitution mechanisms (e.g., Andrault et al., 20072001; Boffa-Ballaran et al., 2012; Brodholt, 2000; Creasy et al., 2020; Daniel et al., 2004; Frost & Langenhorst, 2002; Saikia et al., 2009; Xu et al., 1998; Yagi et al., 2004; Yoshino et al., 2016; Zhang & Weidner, 1999).

The Al3+ substitution mechanism in Fe3+-free bridgmanite has been systematically studied (e.g., Andrault et al., 1998; Brodholt, 2000; Grüninger et al., 2019; Kojitani, Katsura, & Akaogi, 2007; Liu, Akaogi, & Katsura, 2019; Liu, Ishii, & Katsura, 2017; Liu, Boffa-Ballaran, et al., 2019; Navrotsky et al., 2003; Panero et al., 2006; Stebbins et al., 2003; Walter et al., 20042006; Yamamoto et al., 2003). It has been found that the MgAlO2.5 component is formed in MgO-excess systems, but not in SiO2-excess systems (Liu, Akaogi, & Katsura, 2019; Liu, Boffa-Ballaran, et al., 2019). The concentration of AlAlO3 increases with pressure and temperature, whereas the MgAlO2.5 content increases with temperature but decreases with pressure (Brodholt, 2000; Liu, Ishii, & Katsura, 2017; Liu, Nishi, et al., 2017; Liu, Akaogi, & Katsura, 2019).

In contrast to Al3+ substitution, the Fe3+ substitution mechanism in bridgmanite has been less studied. The majority of previous studies focused on Fe3+-Al3+ coupling (e.g., Frost & Langenhorst, 2002; Liu, Dubrovinsky, et al., 2019; Liu et al., 2020; Mohn & Trønnes, 2016; Nishio-Hamane et al., 2005; Richmond & Brodholt, 1998; Saikia et al., 2009; Vanpeteghem et al., 2006; Walter et al., 2004), and only a few studies examined Fe3+ substitution in Fe3+-rich systems. Earlier studies regarding Fe3+ substitution in Fe3+-rich bridgmanite show a dominance of the charge-coupled substitution mechanism (Andrault & Bolfan-Casanova, 2001; Catalli et al., 2010). However, recent studies with well-constrained chemical compositions (Fei et al., 2020; Huang, Boffa-Ballaran, McCammon, Miyajima, Dolejš & Frost, 2021; Hummer & Fei, 2012; Sinmyo et al., 2014) show that bridgmanite can contain 2–3 mol % of the MgFeO2.5 component in addition to the FeFeO3 component in the presence of ferropericlase at 25–27 GPa, that is, relatively low-pressure conditions of the bridgmanite stability field. It was also demonstrated that the FeFeO3 content increases with increasing temperature, whereas the MgFeO2.5 content is independent of temperature (Fei et al., 2020). However, the pressure dependence of Fe3+ substitution in bridgmanite is still unclear.

In this study, we investigated the substitution mechanism of Fe3+ in Al3+-free bridgmanite using a recently developed ultrahigh-pressure (>25 GPa) multianvil technique with tungsten carbide anvils (Ishii et al., 20162019) at 33 and 40 GPa at 2300 K. Although bridgmanite in the lower mantle contains Al3+, which could affect the Fe3+ substitution as mentioned above, we investigated the Al3+-free system to provide basic understanding of the roles of trivalent cations in bridgmanite chemistry. To maximize the MgFeO2.5 content in bridgmanite, experiments were performed in the MgO-SiO2-Fe2O3 system where bridgmanite coexists with MgFe2O4-phase, and with/without ferropericlase. Together with our recent work at 27 GPa (Fei et al., 2020), we show that the concentrations of both FeFeO3 and MgFeO2.5 in bridgmanite, and thus the total Fe3+ content, decrease with increasing pressure. Even coexisting with ferropericlase, the formation of oxygen vacancies is completely suppressed at about 40 GPa. Our results provide basic knowledge about the phase relations and Fe3+ substitution mechanisms in bridgmanite under Fe3+-rich conditions.

2 Experimental Procedure

2.1 High-Pressure Multianvil Experiments

The starting materials used in this study were identical to those in Fei et al. (2020), that is, mixtures with compositions of 5MgO +3SiO2 + 1Fe2O3 (MgO-rich sample) and 4MgO + 3SiO2 + 2Fe2O3 (Fe2O3-rich sample) prepared from SiO2, MgO, and Fe2O3 oxides (Figure 1). The purity of each oxide was >99.9%. Platinum chambers with inner diameter of 0.3 mm, outer diameter of 0.4 mm, and length of 0.3 mm were used as sample capsules, which were placed in a Al2O3 sleeve in the LaCrO3 furnace. A Cr2O3-doped MgO octahedron with edge length of 5.7 mm was used as the pressure medium (Figure 2).

Details are in the caption following the image

Chemical compositions of MgO-rich and Fe2O3-rich starting materials and run products in the ternary phase diagram with endmembers of Fe2O3, MgO, and SiO2. The 27 GPa run was already reported in Fei et al. (2020).

Details are in the caption following the image

Design of the 5.7/1.5 multianvil cell assembly for 33 and 40 GPa runs. M: MgO-rich starting material. F: Fe2O3-rich starting material.

High pressures were generated by tungsten carbide anvils with truncation edge lengths of 1.5 mm using the 15 MN multianvil press, IRIS-15, at the University of Bayreuth (Ishii et al., 2016). The experimental pressures were 33 and 40 GPa (Table 1). The temperature and annealing duration were 2,300 K and 24 hr, respectively. After annealing, the heating power supplier was switched off, by which the temperature decreased to less than 800 K within 1 s and to less than 400 K within 2–4 s. Afterward, the pressure was decreased to ambient conditions over durations exceeding 15 hr.

Table 1. List of Run Conditions and Chemical Compositions of the Run Products
Starting material Run. No. Assembly P (GPa) T (h) Phase N MgO wt% SiO2 wt% Fe2O3 wt% Total wt% Mg pfu Si pfu Fe pfu Bridgmanite (mol %) MgFe2O4-phase (mol %) Ferropericlase (mol %)
FeFeO3 MgFeO2.5 MgSiO3 Mg2SiO4 MgFe2O4 Fe8/3O4 MgO FeO
MgO-rich I574 (Fei et al., 2020) 7/3 27 9 Bridgmanite 17 35.94 (99) 52.29 (80) 13.47 (42) 101.70 (123) 0.923 (15) 0.902 (13) 0.175 (7) 7.6 (15) 2.2 (27) 90.2 (13)
MgFe2O4 11 19.82 (34) 2.85 (9) 76.25 (95) 98.92 (119) 0.957 (9) 0.093 (3) 1.858 (10) 9.5 (3) 78.3 (15) 12.2 (13)
(Mg·Fe)O 7 89.75 (117) 0.25 (17) 11.17 (36) 101.16 (85) 0.932 (5) 0.002 (1) 0.065 (3) 93.5 (5) 6.5 (3)
I873 5.7/1.5 33 24 Bridgmanite 16 36.38 (81) 53.68 (99) 9.34 (98) 99.40 (88) 0.944 (11) 0.934 (11) 0.122 (14) 5.6 (13) 1.0 (20) 93.4 (11)
MgFe2O4 16 13.78 (25) 2.20 (22) 81.17 (66) 97.14 (75) 0.716 (11) 0.077 (7) 2.130 (14) 7.6 (7) 55.4 (21) 37.1 (15)
(Mg·Fe)O 11 79.57 (235) 0.20 (19) 20.60 (53) 100.37 (261) 0.882 (5) 0.002 (1) 0.115 (2) 88.4 (3) 11.6 (3)
I909 5.7/1.5 40 24 Bridgmanite 24 37.67 (66) 56.14 (103) 8.31 (59) 102.12 (134) 0.947 (13) 0.947 (11) 0.106 (8) 5.3 (13) 0.0 (23) 94.7 (11)
MgFe2O4 14 11.70 (60) 2.52 (43) 85.94 (102) 102.40 (41) 0.584 (26) 0.084 (14) 2.165 (34) 8.4 (14) 41.5 (18) 50.1 (18)
(Mg·Fe)O 17 78.49 (56) 0.21 (56) 23.69 (32) 102.40 (41) 0.812 (6) 0.001 (4) 0.124 (2) 86.8 (1) 13.2 (1)
Fe2O3-rich I574 (Fei et al., 2020) 7/3 27 9 Bridgmanite 15 29.72 (94) 43.57 (194) 26.67 (211) 99.96 (170) 0.821 (18) 0.807 (25) 0.372 (36) 17.9 (18) 1.4 (26) 80.7 (25)
MgFe2O4 13 16.98 (37) 2.11 (7) 79.99 (50) 99.09 (67) 0.846 (14) 0.071 (2) 2.013 (12) 7.1 (2) 70.4 (17) 22.5 (16)
I646 (Fei et al., 2020) 7/3 27 20 Bridgmanite 9 31.45 (94) 46.78 (126) 23.40 (246) 101.63 (67) 0.843 (19) 0.841 (18) 0.316 (35) 15.7 (19) 0.2 (13) 84.1 (18)
MgFe2O4 9 18.23 (50) 4.55 (23) 74.25 (137) 97.04 (140) 0.8W85 (19) 0.148 (8) 1.819 (26) 15.1 (8) 60.2 (22) 24.7 (21)
I873 5.7/1.5 33 24 Bridgmanite 20 35.50 (83) 52.10 (97) 12.58 (142) 100.18 (95) 0.924 (13) 0.910 (11) 0.166 (20) 7.6 (13) 1.4 (14) 91.0 (11)
MgFe2O4 13 12.37 (21) 1.75 (7) 83.57 (92) 97.69 (83) 0.652 (11) 0.062 (3) 2.224 (13) 6.0 (3) 51.4 (12) 42.6 (12)
I909 5.7/1.5 40 24 Bridgmanite 14 37.35 (100) 55.81 (80) 9.07 (84) 102.23 (99) 0.941 (16) 0.944 (12) 0.115 (12) 5.9 (16) −0.2 (26) 94.4 (12)
MgFe2O4 13 9.90 (67) 1.5 (21) 88.93 (102) 100.07 (45) 0.502 (33) 0.042 (7) 2.276 (27) 4.2 (7) 41.7 (31) 54.1 (31)
  • Note. All experiments were performed at 2,300 K. Runs I574 and I646 at 27 GPa are already reported in Fei et al. (2020). P: pressure. t: annealing time. N: number of analyzed points by electron microprobe. The error bars are one standard deviation of N points from electron microprobe analysis.

2.2 Sample Analysis

  • (1)

    Scanning Electron Microscopy (SEM). Cross sections of the recovered assemblies were prepared and analyzed by SEM. Backscattered electron images (BSE) were taken on the cross sections (Figure 3). The presented phases on the cross sections were examined by an energy dispersive detector

  • (2)

    X-ray Diffraction. Microfocus X-ray diffraction analyses were performed using a microfocus X-ray diffractometer (Brucker AXS D8 Discover) with a microfocus source of Co-Kα radiation. The beam diameter was about 100 µm focused on the cross sections of the recovered samples. The acceleration voltage and beam current were 40 kV and 500 µA, respectively. The exposure time was 5–6 hr for each analysis. Examples of the diffraction patterns are shown in Figure 4

  • (3)

    Mössbauer Spectroscopy. Synchrotron Mössbauer source (SMS) spectroscopy analyses were performed under ambient conditions on all the samples at beamline BL10XU, SPring-8, Japan. The detailed setup and analytical conditions of SMS spectroscopy are given in Hirao et al. (2020). The spectra were fitted using MossA with Lorentzian doublets (Prescher et al., 2012) (Figure 5)

  • (4)

    Electron Microprobe Analysis. The chemical compositions of bridgmanite and coexisting phases were obtained by electron probe microanalyzer (EPMA) at the University of Bayreuth. The acceleration voltage was 15 kV, the beam current was 5 nA, and the counting time was 20 s for each point analysis. An enstatite standard was used for Mg and Si, whereas metallic iron was used for Fe. Tests were also made using a Fe2O3-standard for analysis of Fe in the samples, which did not show any meaningful difference compared to results using a metallic-Fe standard. Grains near the Pt capsule wall were avoided in the analyses

2.3 Calculation of Chemical Formula

Assuming that MgSiO3 + MgFeO2.5 + FeFeO3 = 100% in the recovered bridgmanite samples, the molar concentrations of the three components were obtained from the equation (Grüninger et al., 2019; Liu, Ishii, & Katsura, 2017),
urn:x-wiley:21699313:media:jgrb55339:jgrb55339-math-0001(1)
where the atomic ratio of Mg, Fe, and Si (a: b: c) in bridgmanite was taken from EPMA. The presence of an MgFeO2.5 component can be indicated by higher Mg atomic content than Si (a > c), whereas there should be no MgFeO2.5 component if a = c. The Mg content could be lower than Si (a < c) if Fe3+ is partially reduced, which is not the case in this study since Fe3+/ΣFe ≈ 100% based on Mössbauer analysis as described later.

Similarly, the concentrations of MgFe2O4, Mg2SiO4, and Fe8/3O4 components in the MgFe2O4-phase were calculated from the EPMA results by assuming Fe3+/ΣFe ≈ 100%, whereas MgO and FeO components in ferropericlase were calculated by assuming all Fe as ferrous (Table 1).

3 Experimental Results

3.1 Phase Assemblages in the Recovered Samples

The recovered MgO-rich and Fe2O3-rich samples contain bridgmanite, a phase close to MgFe2O4 composition (hereafter MgFe2O4-phase), and either with (MgO-rich samples) or without (Fe2O3-rich samples) ferropericlase (Table 1 and Figure 1), as demonstrated by the backscattering images (Figure 3) and X-ray diffraction (Figure 4). No observable inhomogeneity of phase compositions was found throughout the capsules, indicating that chemical equilibrium was reached.

Details are in the caption following the image

SEM images of samples recovered from 33 (a, b) and 40 GPa (c, d). Brg: bridgmanite. Fpc: ferropericlase. MgFe2O4: MgFe2O4-phase.

Details are in the caption following the image

X-ray diffraction spectra of the recovered samples (MgO-rich conditions). Pt and Al2O3 peaks are from the sample capsule and the sleeve outside of the capsule in the cell assembly (Figure 2), respectively, owing to the limited spatial resolution of the diffractometer. The MgFe2O4-phase show consistent peaks at different pressures, indicting the same structure. The spectra are background subtracted. The identified peaks of bridgmanite are labeled in the figure (hkl), and their d-spacings and fitted lattice parameters are given in the online Supporting Information (Table S1).

The MgFe2O4-phase was previously assigned to be a CaMn2O4-type structure (Andrault & Bolfan-Casanova, 2001; Fei et al., 2020) or CaTi2O4-type structure (Greenberg et al., 2017), and recently suggested to be a new structure (modified Na-Fe-Ti oxide-type) of post-spinel (Ishii et al., 2020). Our study primarily focused on the bridgmanite phase without considering the structural complexities of the MgFe2O4-phase.

3.2 Fe3+/ΣFe in the Run Products

Mössbauer spectra unambiguously indicate the predominance of Fe3+ in bridgmanite (Figures 5 and 6). Although fitting models are not unique due to the high degree of line overlap, the Fe2+ in bridgmanite as Fe2+SiO3 component, whose hyperfine parameters are expected to have center shift (CS) and quadrupole splitting (QS) of 0.9–1.2 and 1.5–2.5 mm/s, respectively (e.g., Huang, Boffa-Ballaran, McCammon, Miyajima, Dolejš & Frost, 2021; Hummer & Fei, 2012; Lauterbach et al., 2000; McCammon, 1998; Sinmyo et al., 2019), was not detected in any plausible fit. Instead, doublets of Fe3+ or Fe2.5+ in nonmagnetic MgFe2O4-phase were identified in the run products from all high-pressure conditions, whereas Fe2+ in nonmagnetic MgFe2O4-phase and Fe3+ in magnetic MgFe2O4-phase were fitted in the 40 GPa samples (Figures 5 and 6). Because of the small proportion of ferropericlase, doublets of ferropericlase are not observed within the experimental data scatter.

Details are in the caption following the image

Synchrotron Mössbauer source spectra of the recovered samples. Left: MgO-rich samples. Right: Fe2O3-rich samples. A baseline was subtracted from the raw spectra determined from a calibration using the single-line absorber K2Mg57Fe(CN)6. The dark green doublets correspond to Fe3+ in bridgmanite, the blue doublets are Fe3+ or Fe2.5+ in nonmagnetic MgFe2O4-phase, and the light green doublets are Fe3+ in magnetic MgFe2O4-phase.

Details are in the caption following the image

Center shift and quadrupole splitting derived from fits of the Mössbauer spectra. There is no detectable Fe2+ in bridgmanite, which should have center shift and quadrupole splitting of about 0.9–1.2 and 1.5–2.5 mm/s, respectively, based on the literature data as shown by open squares (Huang, Boffa-Ballaran, McCammon, Miyajima, Dolejš & Frost, 2021; Lauterbach et al., 2000; McCammon, 1998; Sinmyo et al., 2019).

3.3 Composition of Bridgmanite, MgFe2O4-Phase, and Ferropericlase

By comparison of bridgmanite compositions at 33 and 40 GPa with that at 27 GPa from Fei et al. (2020), it is found that the Fe3+ content in bridgmanite under MgO-rich conditions decreases dramatically from ∼0.17 pfu at 27 GPa (Fei et al., 2020) to ∼0.11 pfu at 40 GPa (Figure 7a). As expected, the Fe2O3-rich samples have higher Fe3+ content than the MgO-rich samples, and Fe3+ content decreases from 0.37 to 0.12 pfu at 27–40 GPa (Figure 7a). The Mg/Si ratios in bridgmanite are slightly higher than unity in the MgO-rich samples, whereas they are essentially unity in the Fe2O3-rich samples (Table 1).

Details are in the caption following the image

Chemical composition of bridgmanite, ferropericlase, and MgFe2O4-phase in run products. (a) Fe3+ content in bridgmanite calculated to O = 3. (b) Mg, Si, and Fe3+ contents in the MgFe2O4-phase calculated to O = 4. (c) Mg and Fe content in ferropericlase from MgO-rich samples calculated to O = 1 assuming that all Fe is ferrous in ferropericlase. Black symbols represent the Fe3+ content in bridgmanite that coexists with ferropericlase reported in previous studies (Frost & Langenhorst, 2002; Frost et al., 2004; Huang, Boffa-Ballaran, McCammon, Miyajima, Dolejš & Frost, 2021; Hummer & Fei, 2012; Lauterbach et al., 2000), all of which are lower than this study because their experimental temperatures are lower, and/or MgFe2O4-phase did not appear (namely Fe3+ is not saturated). The data points with blue symbols at 27 GPa are from Fei et al. (2020). The error bars represent one standard deviation of the analyzed points by electron microprobe as shown in Table 1.

The composition of the MgFe2O4-phase deviates from the MgFe2O4 endmember (Figure 1). With increasing pressure from 27 to 40 GPa, the Fe3+ content increases from 1.9 to 2.3 pfu, and the Mg content decreases from 0.9 to 0.5 pfu. The Si content is low but detectable (∼0.1 pfu) (Figure 7b). Additionally, up to 13.2 mol % Fe was found in ferropericlase (Figure 7c).

4 Discussion

4.1 Fe3+ and Fe2+ Partitioning Between Bridgmanite and Ferropericlase

The Fe contents [Fe/(Fe + Mg)] in ferropericlase of the MgO-rich samples are 6.5–13.2% (Figure 7c). If all Fe were ferrous in ferropericlase, the partition coefficient of Fe2+ between bridgmanite and ferropericlase would be nearly zero based on the absence of Fe2+ in the current bridgmanite samples. The partition coefficient is thus much smaller than that suggested by previous studies (e.g., Nakajima et al., 2012; Prescher et al., 2014; Xu et al., 2017), who reported values of 0.2–0.4 at 25–40 GPa. We emphasize that this discrepancy cannot be caused by undetectable Fe2+ in our bridgmanite samples. If the Fe2+/Mg partition coefficient given by previous studies were followed by our samples, Fe2+/ΣFe should be 15–50% in bridgmanite. Such a significant fraction of Fe2+ would definitely be detectable by both in-house (Fei et al., 2020) and synchrotron (this study) Mössbauer spectroscopy because the hyperfine parameters of Fe2+ doublets in bridgmanite are well known and would not overlap with other components in our spectra (e.g., Huang, Boffa-Ballaran, McCammon, Miyajima, Dolejš & Frost, 2021; Sinmyo et al., 2019; Yoshino et al., 2016, Figure 6). Therefore, the appearance of 6.5–13.2 mol % FeO in ferropericlase does not suggest the presence of Fe2+ in bridgmanite. Since the experiments in previous studies (e.g., Nakajima et al., 2012; Prescher et al., 2014) were mostly performed under relatively reducing conditions with high Fe2+/ΣFe ratios, one explanation for this discrepancy is that Fe2+ is almost entirely incorporated into ferropericlase when the bulk Fe2+ content in the system is extremely low, that is, the partition coefficient may have a substantial compositional dependence.

4.2 Fe3+ Content in Bridgmanite

Based on the phase rule, the Fe3+ content in bridgmanite under MgO-rich conditions should be uniquely constrained because three phases coexist in the system. Although some Fe2O3 might be reduced to FeO in high-pressure experiments as indicated by the presence of Fe in ferropericlase in MgO-rich samples (Figure 7a), the number of components in bridgmanite is still three because the Fe3+/ΣFe ratio is close to 100% in bridgmanite as demonstrated by Mössbauer spectroscopy.

The Fe2O3-rich samples in this study and some other studies (e.g., Andrault & Bolfan-Casanova, 2001; Liu et al., 2018; Wang et al., 2021) show much higher Fe3+ contents in bridgmanite (up to 1.0 pfu) than the current MgO-rich bridgmanite samples. However, these high Fe3+-content bridgmanite samples did not coexist with ferropericlase (bridgmanite + MgFe2O4-phase in Fe2O3-rich samples in this study and only bridgmanite in Andrault & Bolfan-Casanova, 2001, Liu et al., 2018, and Wang et al., 2021). When bridgmanite coexists with ferropericlase, the Fe3+ content in bridgmanite will be limited because of the formation of the MgFe2O4-phase from FeFeO3 and MgO (Andrault & Bolfan-Casanova, 2001). When bridgmanite does not coexist with MgO, the Fe3+ content in bridgmanite depends on the starting material. If the bulk Fe3+ content in the starting material is high, the Fe3+ content in bridgmanite can accordingly be high based on the phase relations in Figure 1, for example, Fe3+ can reach 1.0 pfu as shown in Liu et al. (2018) and Wang et al. (2021). This is understandable because the molar volume of hematite (30.5 cm3/mol) is slightly larger than the FeFeO3 component in bridgmanite (29.55 cm3/mol, Huang, Boffa-Ballaran, McCammon, Miyajima, & Frost, 2021). Therefore, Fe3+ may tend to be incorporated in bridgmanite by a MgSiO3-Fe2O3 solid solution instead of forming hematite, consequently, the Fe3+ solubility in bridgmanite is high. The upper limit of Fe3+ content should be obtained in the system with coexistence of bridgmanite and hematite, which was not investigated in this study. Additionally, the formation of the FeAlO3 component will also increase the Fe3+ content in Al-bearing bridgmanite, which causes the high Fe3+ content (about 0.7 pfu) in Liu, Dubrovinsky, et al. (2019) and Liu et al. (2020).

The MgFe2O4-phase may have structural complexities (e.g., Andrault & Bolfan-Casanova, 2001; Greenberg et al., 2017; Ishii et al., 2020), which may affect Fe3+ partitioning between bridgmanite and the MgFe2O4-phase and thus affect the Fe3+ content in bridgmanite. Some studies reported phase transitions among polymorphs of MgFe2O4 at high temperatures below 25 GPa (e.g., Ishii et al., 2020; Uenver-Thiele et al., 2017), and at ambient temperature in the pressure range 25–40 GPa (Greenberg et al., 2017). However, no phase transition of MgFe2O4 has been reported at the conditions of our experiments, that is, 2,300 K and 27–40 GPa. The absence of a phase transition of the MgFe2O4-phase in this study has also been indicated by X-ray diffraction of the recovered samples (Figure 4). Therefore, the systematic decrease of Fe3+ content in bridgmanite is not expected to be caused by complex polymorphism of MgFe2O4.

4.3 Pressure Dependence of Fe3+ Substitution in Bridgmanite

In the MgO-rich samples, the FeFeO3 content in bridgmanite decreases from 7.7 to 5.3 mol %, whereas the MgFeO2.5 content decreases from 2.2 to ∼0% at 27–40 GPa (Figure 8a). The extrapolation of data agrees well with the maximum MgFeO2.5 content of 3.5% reported by Hummer and Fei (2012) at 25 GPa and 1,970–2,070 K (Figure 8a). In contrast, Fe2O3-rich samples have FeFeO3 contents ranging from 17.9% to 5.9%, which is higher than the MgO-rich samples, and have MgFeO2.5 contents that are essentially zero within error over the entire investigated pressure range (Figure 8b).

Details are in the caption following the image

Substitution mechanisms in bridgmanite and MgFe2O4-phase. (a) Fe3+ substitution in bridgmanite under MgO-rich conditions. The maximum MgFeO2.5 content reported by Hummer and Fei (2012) at 25 GPa, 1,970–2,070 K is also shown for comparison. (b) Fe3+ substitution in bridgmanite under Fe2O3-rich conditions. (c). Substitution mechanisms in the MgFe2O4-phase. The data points with blue symbols at 27 GPa are from Fei et al. (2020). The error bars represent one standard deviation of the analyzed points by electron microprobe as shown in Table 1.

Note that smaller Pt capsules were used in the runs at 33 and 40 GPa in this study compared to the 27 GPa runs in Fei et al. (2020). Because Pt capsules may absorb Fe from the samples and thus release O2, more O2 might be released in 33 and 40 GPa runs relative to sample volumes. However, the O2 formed by Fe dissolution in Pt should not cause MgFeO2.5 content to decrease with increasing pressure. Since Fe3+/ΣFe ≈100% in bridgmanite in all runs, the chemistry of bridgmanite in MgO-rich samples is uniquely constrained with maximized Fe3+ content and maximized Fe3+/ΣFe ratio. Excess O2 cannot further oxidize bridgmanite. Therefore, the chemistry of bridgmanite from MgO-rich samples in this study will not be affected by excess O2. Although the excess O2 may produce peroxide components (Hu et al., 2016; Zhu et al., 2019), it requires pressures >70 GPa, which is not the case in this study.

Because of the uncertainties in the Mg, Si, and Fe contents obtained from EPMA analysis (Table 1), the error bars are relatively large for the relatively small MgFeO2.5 contents (Figure 7a). Additionally, the reproducibility of MgFeO2.5 content obtained in different runs under the same pressure and temperature conditions is about ±0.3–1.25 mol % (Table 1 and Fei et al., 2020), which is not negligible. These problems make it challenging to obtain a definite conclusion regarding the pressure dependence of the MgFeO2.5 content. However, plots of Mg and Si contents in bridgmanite versus Fe3+ content show that data at 27 GPa and 1,700–2,300 K under MgO-rich conditions clearly deviate from the theoretical Mg and Si contents of pure FeFeO3 substitution, whereas they essentially follow the trend of the pure FeFeO3 substitution mechanism in Fe2O3-rich samples (Figure 9 and Fei et al., 2020). This behavior demonstrates the presence of MgFeO2.5 components at 27 GPa under MgO-rich conditions. In contrast to the 27-GPa data, the 33-GPa data are closer to the pure FeFeO3 substitution, and the 40-GPa data are exactly on the trend of pure FeFeO3 substitution even in MgO-rich samples (Figure 9), which suggests a decrease of MgFeO2.5 content with increasing pressure. Therefore, the reduction of MgFeO2.5 content with increasing pressure is convincing despite relatively large uncertainties of absolute MgFeO2.5 contents.

Details are in the caption following the image

Mg and Si contents (per formula unit) as a function of Fe3+ content in bridgmanite in MgO-rich samples. The data points at 27 GPa are from Fei et al. (2020) at 1,700–2,300 K, whereas those at 33 and 40 GPa are from this study with a temperature condition of 2,300 K. The theoretical Mg and Si contents with pure FeFeO3 substitution and pure MgFeO2.5 substitution mechanisms are shown by thin solid and dashed lines, respectively. The error bars represent one standard deviation of the analyzed points by electron microprobe as shown in Table 1.

The suppression of Fe3+-linked oxygen vacancies with increasing pressure can be understood by the volume increase associated with MgFeO2.5 formation. Partial molar volume of the MgFeO2.5 component in bridgmanite is estimated to be 27.65 cm3/mol (Huang, Boffa-Ballaran, McCammon, Miyajima, & Frost, 2021), whereas the molar volume of the MgFe2O4-phase is 41.4 cm3/mol (Ishii et al., 2020), and that of MgO is 11.25 cm3/mol at ambient conditions (Dorogokupets, 2010; Tange et al., 2012). Thus, the volume change in the reaction,
urn:x-wiley:21699313:media:jgrb55339:jgrb55339-math-0002(2)
is about +2.7 cm3/mol at ambient conditions. This volume change will be even larger (4.3–4.7 cm3/mol) by adjusting the pressure to 27–40 GPa using reported equation of states (Dorogokupets, 2010; Ishii et al., 2020; Tange et al., 2012), because the MgFe2O4-phase has a much smaller bulk modulus (164 GPa) than bridgmanite (257 GPa) (Ishii et al., 2020; Tange et al., 2012). The reaction should thus be significantly suppressed by increasing pressure. As a result, the concentration of MgFeO2.5 content decreases rapidly with increasing pressure.
The negative pressure dependence of MgFeO2.5 content is identical to that of the MgAlO2.5 content in the MgO-SiO2-Al2O3 system (Liu, Ishii, & Katsu, 2017), which can also be understood by the positive volume change of the reaction:
urn:x-wiley:21699313:media:jgrb55339:jgrb55339-math-0003(3)
in which MgAlO2.5 and MgAl2O4 have molar volume of 26.6 and 36.5 cm3/mol, respectively (Huang, Boffa-Ballaran, McCammon, Miyajima, & Frost, 2021; Kojitani, Hisatomi, & Akaogi, 2007; Liu, Akaogi, & Katsura, 2019; Sueda et al., 2009).
The FeFeO3 content in our samples decreases with increasing pressure as well. The reaction of Fe3+ between bridgmanite and MgFe2O4-phase can be written as:
urn:x-wiley:21699313:media:jgrb55339:jgrb55339-math-0004(4)
The FeFeO3 component has a molar volume of 29.55 cm3/mol (Huang, Boffa-Ballaran, McCammon, Miyajima, & Frost, 2021). Although the volume change of the above reaction is negative at ambient conditions (ΔV = −0.6 cm3/mol), it becomes positive (+0.5 to +0.8 cm3/mol) after adjusting to 27–40 GPa using the equation of state for each phase (Dorogokupets, 2010; Ishii et al., 2020; Tange et al., 2012). As a result, the FeFeO3 content in bridgmanite decreases with pressure.
In contrast to the FeFeO3 component in bridgmanite, the AlAlO3 content in Fe3+-free bridgmanite increases with increasing pressure (Liu et al., 2016, Liu, Nishi, et al., 2017). Since bridgmanite coexists with corundum (Al2O3) in Liu, Ishii, and Katsura (2017) and Liu, Nishi, et al. (2017), the exchange of Al between bridgmanite and corundum can be written as:
urn:x-wiley:21699313:media:jgrb55339:jgrb55339-math-0005(5)
Al2O3 (corundum) and AlAlO3 (bridgmanite) have nearly identical molar volumes (25.6 and 25.8 cm3/mol, respectively) (Dewaele & Torrent, 2013; Huang, Boffa-Ballaran, McCammon, Miyajima, & Frost, 2021; Liu, Akaogi, & Katsura, 2019).
On the other hand, in the case that bridgmanite coexists with MgAl2O4 and MgO, the AlAlO3 component in bridgmanite can be formed by,
urn:x-wiley:21699313:media:jgrb55339:jgrb55339-math-0006(6)
which has a small, but positive volume change (ΔV = +0.5 cm3/mol at ambient conditions) using the molar volume of each component reported previously (Huang, Boffa-Ballaran, McCammon, Miyajima, & Frost, 2021; Kojitani, Hisatomi, & Akaogi, 2007; Liu, Akaogi, & Katsura, 2019; Sueda et al., 2009). Thus, the small but positive volume changes of reactions (5) and (6) suggest that the AlAlO3 content in bridgmanite should slightly decrease or be nearly constant with increasing pressure, which contradicts the tendency reported by Liu et al. (2016) and Liu, Nishi, et al. (2017). A possible cause for this discrepancy could be a large uncertainty in the reported molar volume of AlAlO3 component in bridgmanite since it is extrapolated from the volume of bridgmanite with relatively low AlAlO3 content (up to 14 mol %) (Huang, Boffa-Ballaran, McCammon, Miyajima, & Frost, 2021; Liu, Akaogi, & Katsura, 2019; Liu, Boffa-Ballaran, et al., 2019), or that the AlAlO3 component has a much smaller bulk modulus than MgSiO3-bridgmanite since Al3+ has a smaller ionic radius than Mg2+ (0.50 Å versus 0.65 Å), leading to negative ΔV for reactions (5) and (6) at high pressures. In contrast, Fe3+ has a comparable ionic radius (0.64 Å) with Mg2+ and larger than Si4+ (0.42 Å). Therefore, the FeFeO3 component should be less compressible than AlAlO3 and MgSiO3 components.

4.4 Chemistry of MgFe2O4-Phase

The detailed substitution mechanism of the MgFe2O4-phase is unknown, but it is assumed to be composed of MgFe2O4, Mg2SiO4, and Fe3+8/3O3 (or Fe2+Fe3+2O4) components (Huang, 2020; Huang, Boffa-Ballaran, McCammon, Miyajima, Dolejš & Frost, 2021; Liu, Akaogi, & Katsura, 2019). The Mg2SiO4 substitution mechanism occurs by replacement of two Fe3+ sites by Mg2+ and Si4+. Fe2+Fe3+2O4 substitution might have occurred instead of Fe3+8/3O4 if significant Fe in the MgFe2O4-phase were ferrous.

Based on the atomic contents of Mg, Si, and Fe from EPMA analysis (Table 1), the concentration of each component can be obtained by assuming the MgFe2O4-phase to be composed of MgFe2O4, Mg2SiO4, and Fe8/3O4. As shown in Figure 8c, the Fe8/3O4 content increases significantly from 12–25% to ∼50% with increasing pressure from 27 to 40 GPa. Correspondingly, the MgFe2O4 content decreases from 60–80% to ∼40% (Figure 8c). The variation of Fe8/3O4 and MgFe2O4 contents in MgFe2O4-phase could be caused by the reaction of FeFeO3 (bridgmanite) = ¾ Fe8/3O4 (MgFe2O4-phase) or MgFe2O4 (MgFe2O4-phase) = ¾ Fe8/3O4 (MgFe2O4-phase) + MgO (periclase). The Mg2SiO4 content has a much smaller pressure dependence than the Fe8/3O4 content: it slightly decreases from 7–15 to 4–8 mol % with increasing pressure from 27 to 40 GPa (Table 1). These observations imply that the negative pressure dependence of Fe3+ content in bridgmanite is dominated by dissolution of Fe3+ from bridgmanite and formation of the Fe8/3O4 component in the MgFe2O4-phase, but the replacement of Fe3+ in bridgmanite by Mg2+ and Si4+ released from the MgFe2O4-phase also partially contributes to the decrease of Fe3+ in bridgmanite.

4.5 Implications for Chemistry of Fe3+-Rich Bridgmanite

4.5.1 Fe3+-Bearing and Al-Free System

Because bridgmanite coexists with ferropericlase and MgFe2O4-phase in the MgO-SiO2-Fe2O3 ternary system under MgO-rich conditions, the experimentally determined MgFeO2.5, FeFeO3, and total Fe3+ contents should represent their maximum contents in Al-free bridgmanite when ferropericlase is present. Fei et al. (2020) demonstrated that the concentration of FeFeO3 component increases with increasing temperature, whereas the MgFeO2.5 content has no clear temperature dependence when bridgmanite coexists with ferropericlase at 27 GPa. On the other hand, the MgFeO2.5, FeFeO3, and total Fe3+ contents are found to decrease with increasing pressure in the present study (Figures 7a and 8a). By combining the pressure and temperature effects following the geotherm in the lower mantle (Katsura et al., 2010), the MgFeO2.5 content in Al-free bridgmanite coexisting with ferropericlase decreases rapidly by more than two orders of magnitude from 4 to 5 mol % at the topmost lower mantle to nearly zero at 1,000–1,200 km depth. In contrast, the FeFeO3 content decreases from 8% at 700 km depth to a minimum of ∼4% at about 800 km depth, and is nearly constant or slightly increases to 5% at 1,200 km depth because of the negative and positive pressure and temperature dependences, respectively (Figure 10).

Details are in the caption following the image

FeFeO3 and MgFeO2.5 contents as a function of depth according to the pressure dependences determined in this study, temperature dependences given by Fei et al. (2020), and geotherm from Katsura et al. (2010). The MgAlO2.5 content in Fe3+-free bridgmanite is based on the pressure and temperature dependences given by Liu, Ishii, and Katsura (2017) and Liu, Akaogi, & Katsura (2019).

4.5.2 Fe3+ and Al-Bearing System

The pure Fe3+-bearing (MgO-SiO2-Fe2O3) system is ideal model for understanding the chemistry of bridgmanite. A more realistic approach for bridgmanite in the lower mantle considers both Fe3+ and Al3+ (e.g., Frost & McCammon, 2008; Irifune, 1994). The substitution mechanism will be controlled by the ratio of Al3+ and Fe3+ because of the formation of the FeAlO3 component (e.g., Huang, Boffa-Ballaran, McCammon, Miyajima, Dolejš & Frost, 2021; Liu et al., 2020; Mohn & Trønnes, 2016; Richmond & Brodholt, 1998; Walter et al., 2006; Zhang & Oganov, 2006). When the Al3+/Fe3+ atomic ratio is larger than unity, all Fe3+ will be consumed by the FeAlO3 component, whereas excess Al3+ will form the AlAlO3 and MgAlO2.5 components (Huang, Boffa-Ballaran, McCammon, Miyajima, Dolejš & Frost, 2021; Huang, Boffa-Ballaran, McCammon, Miyajima, & Frost, 2021; Mohn & Trønnes, 2016). In contrast, when the Al3+/Fe3+ ratio is smaller than unity, excess Fe3+ will form the FeFeO3 and MgFeO2.5 components in addition to the FeAlO3 component (Huang, Boffa-Ballaran, McCammon, Miyajima, Dolejš & Frost, 2021; Mohn & Trønnes, 2016).

Although the Fe3+ content in bridgmanite under deep lower mantle conditions is considered to be relatively low in comparison with Al3+ (Frost & McCammon, 2008; Irifune & Ringwood, 1987a; Kurnosov et al., 2017; Lauterbach et al., 2000; Liu et al., 2020; Nakajima et al., 2012; Prescher et al., 2014; Shim et al., 2017; Sinmyo et al., 2019) and therefore Fe3+ should mainly form FeAlO3, Fe3+/Al3+ could be larger than unity in some regions. For example, bridgmanite in the topmost lower mantle has relatively high Fe3+ solubility (Fei et al., 2020; Liu et al., 2018; Wang et al., 2021) but low Al3+ solubility (Liu et al., 2016; Panero et al., 2006), subducted slabs may have relatively high oxygen fugacity conditions and thus should be Fe3+-enriched (Zhao et al., 2021), and harzburgitic rocks are depleted in Al3+ (e.g., Irifune & Ringwood, 1987b). All of these regions may have relatively high Fe3+ content in bridgmanite, and thus MgFeO2.5 and FeFeO3 components could be formed. Their concentrations should decrease rapidly with increasing pressure because of the negative pressure dependence of their solubilities as determined in this study.

4.6 Implications for Lower Mantle Dynamics

The presence of MgFeO2.5 and FeFeO3 components in bridgmanite may affect its physical and chemical properties as predicted from the effects of MgAlO2.5 and AlAlO3 components (e.g., Andrault et al., 20012007; Boffa-Ballaran et al., 2012; Brodholt, 2000; Daniel et al., 2004; Frost & Langenhorst, 2002; Saikia et al., 2009; Xu et al., 1998; Yagi et al., 2004; Zhang & Weidner, 1999). Because the MgFeO2.5 component contains oxygen vacancies, the atomic diffusivity, which is proportional to the defect concentration, is expected to be enhanced. On the other hand, although the FeFeO3 component does not produce vacancies, it should strongly distort the crystal structure of bridgmanite by substitution of Fe3+ on the Si site compared to other components such as MgSiO3, AlAlO3, FeSiO3, and FeAlO3 due to the much larger ionic radius of Fe3+ compared to Si4+ and Al3+. As a result, the FeFeO3 component is expected to enhance element diffusivities as well. Therefore, the decrease of both FeFeO3 and MgFeO2.5 content with pressure may cause decreasing atomic diffusivities in bridgmanite, which may affect diffusion-controlled physical and chemical processes and thus affect the mantle dynamics.

One example is mantle rheology. The creep of minerals is controlled by diffusion of the slowest species (e.g., Herring, 1950; Nabarro, 1967). Although the viscosity of bridgmanite is controlled by Mg and Si diffusion rather than O because Mg and Si diffuse slower than O (e.g., Dobson et al., 2008; Holzapfel et al., 2005; Xu et al., 2011; Yamazaki et al., 2000), both Mg and Si are fully surrounded by O in polyhedrons. The hopping of Mg and Si ions from/into the polyhedron should become easier when an oxygen ion is missing. Hence, oxygen vacancies may enhance the diffusion of Mg and Si and thus reduce the viscosity. Therefore, it is predicted that the decrease in both FeFeO3 and MgFeO2.5 contents in bridgmanite from 700 to ∼1,000–1,200 km depth could suppress Mg and Si diffusivities in bridgmanite, which may contribute to the large viscosity increase in the midmantle inferred from geoid analysis (Rudolph et al., 2015).

Another example is electrical conductivity in the lower mantle. The electrical conductivity of bridgmanite is dominated by the ionic conduction mechanism at relatively high temperatures (e.g., Dobson, 2003; Xu & McCammon, 2002; Yoshino et al., 2016), which is controlled by atomic diffusion of the fastest species, that is, O in bridgmanite (Dobson et al., 2008). Therefore, based on the Nernst-Einstein relation, the ionic conductivity of bridgmanite should be enhanced by the presence of the MgFeO2.5 component. The decreasing of MgFeO2.5 content with depth may contribute to the decrease in observed conductivity at >800 km depth based on magnetotelluric sounding (e.g., Civet et al., 2015; Civet & Tarits, 2013).

The above examples are based on qualitative interpretation. To constrain the role of MgFeO2.5 and FeFeO3 components on mantle dynamics in more detail, further investigations about their effects on the physical and chemical properties of bridgmanite are required. Additionally, as mentioned above, bridgmanite in the lower mantle contains Al3+, which could affect the substitution mechanism of Fe3+ (e.g., Huang, Boffa-Ballaran, McCammon, Miyajima, Dolejš & Frost, 2021; Huang, Boffa-Ballaran, McCammon, Miyajima, & Frost, 2021; Liu et al., 2020). More experimental studies on the pressure and temperature dependences of Fe3+ substitution in both Fe3+ and Al3+ bearing bridgmanite are therefore necessary following the pattern of the detailed investigated in Huang, Boffa-Ballaran, McCammon, Miyajima, Dolejš and Frost (2021); Huang, Boffa-Ballaran, McCammon, Miyajima, and Frost (2021); at a single condition (25 GPa, 1970 K) corresponding to the topmost lower mantle.

Acknowledgments

We acknowledge the European Research Council (ERC) research Grant (No. 787527) and the Deutsche Forschungsgemeinschaft (DFG) (KA3434/9-1, KA3434/11-1, KA3434/12-1) to T. Katsura, and the annual budget of Bayerisches Geoinstitut to H. Fei for financial support to this study. The Synchrotron Mossbauer source spectroscopy analysis was performed with the approval of the Japan Synchrotron Radiation Research Institute (2019B1136). Open access funding enabled and organized by Projekt DEAL.

    Data Availability Statement

    The EPMA, XRD, and Mössbauer data for this paper are given in Zenodo (https://doi.org/10.5281/zenodo.5661686).