There is a general consensus that the lower mantle is dominated by bridgmanite on the basis of high-pressure experimental data of density and sound velocities of the candidate minerals1,2,3. The models of the lower mantle are largely dependent on the chemical composition of bridgmanite throughout the entire lower mantle down to the top of D″ layer4. How the chemistry of bridgmanite especially the Fe3+/ ∑Fe ratio and total iron content change with bulk composition under the lower mantle conditions remains controversial. Under pressure-temperature conditions of the topmost lower mantle, up to 16% Fe3+ was obtained in bridgmanite (Mg, Fe)SiO3 synthesized from Fe2+-dominant material5,6, whereas higher concentrations of Fe3+ were observed in aluminous bridgmanite showing a nearly linear dependence of Fe3+/ ∑Fe with Al3+ content6,7. Furthermore, Fe metal was observed in coexistence with Fe3+ as a result of the disproportionation of ferrous iron in bridgmanite8.

Under pressure-temperature conditions of the deep lower mantle (>80 GPa), existing data of the Fe3+/∑Fe ratio in Al3+-bearing bridgmanite remain scattered ranging from 20 to 60% (Fig. S1) in part due to the differences in their starting materials9,10,11,12. The effects of pressure and Al3+ content on the Fe3+/∑Fe ratio of bridgmanite have not been fully clarified because of its complicating factors such as chemical composition, spin state of iron10, difficult-to-achieve equilibrium of site distribution of iron13, and possible iron oxidation induced by amorphization7,14. The ab initio calculations, however, suggested that the disproportionation reaction from ferrous iron to Fe3+ plus iron metal is energetically favorable, in both Al-free and Al-rich compositions, at all lower mantle pressures15. Properties sensitive to the Fe3+/∑Fe ratio of bridgmanite include element partitioning between lower mantle minerals10,16, spin state of iron17,18,19, and density and sound velocity profiles3,20 under the lower mantle conditions. To clarify the effect of Al3+ content on the Fe3+/∑Fe in bridgmanite, we should further take into account the H2O effect as the presence of a hydrous phase could drastically reduce the Al3+ content in bridgmanite with Al3+ preferentially partitioning into the coexisting hydrous phase relative to bridgmanite21,22,23,24,25. To the best of our knowledge, the H2O effect on the chemistry of bridgmanite has never been reported. The effects of Al3+ and H2O on the chemistry of bridgmanite under high-pressure-temperature conditions must be addressed in order to obtain an accurate model of the lower mantle and understand the origin of chemical heterogeneity in the deep lower mantle.

To understand the key factors controlling the iron oxidation state of bridgmanite, we designed experiments to separate the effects of Al3+ and H2O. First, we will determine the Fe3+/∑Fe ratio of Al3+-free bridgmanite as a function of pressure under hydrous conditions. Orthopyroxene (opx) (Mg0.85Fe0.15)SiO3 (Fs15) was sandwiched between hydrous silica gel as the starting material. Second, experiments on dry bridgmanite will be carried out for comparison. The previous study reported iron depletion in dry bridgmanite as a result of the disproportionation reaction26. Third, in order to evaluate the Al3+ effect, we conducted one experiment on Al3+-bearing bridgmanite in a hydrated basaltic composition to compare with the results of Al3+-free bridgmanite under similar pressure-temperature conditions. Our experiments were performed in laser-heated diamond anvil cells over the pressure and temperature range of 91-125 GPa and 1800-2400 K, close to mantle geotherm conditions27.

Results and discussions

Experimental conditions

Experimental results and conditions are listed in Table 1. The phase assemblages and Fe3+/∑Fe ratio of bridgmanite were obtained combining in situ X-ray diffraction (XRD) at high pressure with ex situ chemical analysis in a transmission electron microscope (TEM) on the samples recovered to ambient conditions (see “Methods). A thin section suitable for TEM analysis was precisely lifted out from the heated center in each sample. The heated area can be clearly recognized under microscope in contrast to the surrounding transparent unreacted sample (Fig. 1). The homogeneity of color is an indication of no obvious variation in iron content across the heated spot. To further examine the effect of temperature gradient on the chemical composition, we managed to prepare a thin section across the temperature gradient of the heated spot along the radial direction and confirmed the homogeneity of the bulk composition (Fig. S2 and Fig. S3). Element mapping and phase chemistry was obtained by energy dispersive spectroscopy (EDS). The bulk composition of the 5 × 5 μm2 area in the heated center was obtained with XFe = 0.15, where XFe is the iron content in atoms per two-cation formula unit, identical to the starting composition Fs15 (Fig. S3 and Table S1). Further, we carried out electron energy-loss spectroscopy (EELS) measurements at the Fe L2,3-edges to distinguish between Fe2+ and Fe3+. Previous studies have demonstrated that EELS is an ideal tool for quantitative determination of Fe3+/∑Fe ratio at the nanometer scale7,14. Specifically, to avoid possible iron oxidation induced by amorphization7, we carefully examined crystallinity of bridgmanite before and after each EELS measurement using selected area electron diffraction (SAED) in a TEM. In this study, only those data from crystalline bridgmanite are used to examine the pressure effect on the Fe3+/∑Fe ratio of bridgmanite.

Table 1 Experimental conditions and results
Fig. 1: Representative microscopic images of the samples recovered to ambient conditions.
figure 1

The heated area of the samples recovered from (A) 125 GPa and 2400 K (Run#344-112) versus (B) 91 GPa and 1950 K (Run#275-81).

Pressure effect on the iron oxidation state of Al3+-free bridgmanite under hydrous conditions

A thin layer of Fs15 opx was sandwiched between symmetric layers of hydrous silica gel containing ~2 wt.% H2O. In the first run, the Fs15 sample was cold compressed to 93 GPa and then heated at 2250 K for 15 mins, corresponding to 105 GPa after accounting for the thermal pressure (Run#390-93, Table 1). The heating duration was 15 mins after the target temperature was reached in all the runs unless otherwise specified. The two-dimensional XRD scan on the sample after temperature quench showed the coexistence of bridgmanite and NiAs-type silica phases (Fig. S4). The sample was then recovered to ambient conditions and prepared for TEM analysis. The EDS mapping of the TEM section showed a homogeneous bridgmanite phase except for a few very small grains of iron metal (Fig. 2A). The chemical composition of bridgmanite was obtained based on multiple EDS analyses with XFe = 0.14 very close to the starting material Fs15 (Table S1), consistent with the observation of a nearly pure bridgmanite phase (Fig. 2A). The SAED data confirmed its crystallinity (Fig. 2B) and the EELS data (Fig. 3) revealed an Fe3+/∑Fe ratio of 0.30(1) by a linear combination of Fe L3 reference spectra of the ferrous and ferric iron standards (Fig. S5).

Fig. 2: TEM images and electron diffraction of the recovered crystalline bridgmanite.
figure 2

A, B Run#390-93, showing a nearly pure bridgmanite phase with grain boundaries recovered from 105 GPa and 2250 K; (C) and (D) Run#188-96, recovered from 108 GPa and 2250 K; (E) and (F) Run#332-82, coexistence of bridgmanite and Fe metal recovered from 92 GPa and 1850 K. The SAED data confirmed crystallinity of bridgmanite in all the runs. The black dots in (A) are marks for damages induced by electron irradiation during the EELS measurement.

Fig. 3: EELS measurements showing ferrous-iron-dominant bridgmanite above 105 GPa and 2250 K under hydrous conditions.
figure 3

The EELS data were recorded for the crystalline bridgmanite samples recovered from 105-119 GPa and 2250-2400 K, in comparison with the EELS data of the starting material Fs15, siderite FeCO3 (Fe2+/ΣFe = 100%) and hematite Fe2O3 (Fe3+/ΣFe = 100%) measured under the same instrumental conditions. The maxima of Fe2+ (708.1 eV) and Fe3+ (709.8 eV) at the L3-edge are indicated by the black dotted lines.

To evaluate the pressure effect on the Fe3+/∑Fe ratio under hydrous conditions, we synthesized another two separate samples at 108 GPa and 2250 K (Run#188-96) and 119 GPa and 2400 K (Run#390-106), respectively (Table 1). The EELS data revealed the Fe3+/∑Fe ratios of 0.14(1) and 0.06(3) for these two runs (Fig. 3), respectively, while the bridgmanite phase in both runs remained crystalline after the recovery (Fig. 2C–F). In the pressure range where bridgmanite is dominant in ferrous iron, we did not observe a temperature dependence of the iron content and Fe3+/∑Fe ratio. In conclusion, the EELS data confirmed the stability of ferrous-iron-dominant bridgmanite above 105 GPa and 2250 K under hydrous conditions (Fig. 4).

Fig. 4: Ferric iron concentration in bridgmanite under hydrous lower mantle conditions.
figure 4

The blue solid circles represent the Fe3+/ΣFe ratios measured in recovered crystalline bridgmanite samples over the pressure range of 90-130 GPa. The light blue curve is a guide to the eye for the Fe3+/ΣFe ratio of bridgmanite. The dash line indicates the corresponding depth (~2300 km) where the Fe3+/ΣFe ratio falls below 0.2.

In an experiment conducted at 91 GPa and 1950 K, the spotty diffraction pattern indicated the formation of a well-crystallized bridgmanite phase at high pressure (Fig. S2), but the TEM analysis on the recovered sample revealed an amorphous bridgmanite phase (Run#275-81, Fig. S6), implying that the sample lost its crystallinity during the recovery. The element mapping showed the coexistence of Fe-bearing bridgmanite, metallic particles and a silica phase (other than the pressure medium), consistent with in situ XRD observation at high pressure (Fig. S6). The EELS measurements revealed all Fe3+ in the amorphous bridgmanite phase.

In order to recover a crystalline bridgmanite phase at this pressure, we synthesized another two samples at slightly lower temperatures. First, a reverse experiment was conducted (Run#332-9582). An Fe2+-dominant bridgmanite phase was synthesized at 107 GPa and 2200 K based on the experimental conditions mentioned above. The sample was then decompressed to 82 GPa and heated again at 1900 K and 92 GPa for 20 mins. The recovered sample was an amorphous Fe3+-bridgmanite phase similar to the results of Run#275-81. We further conducted another experiment at 92 GPa and 1850 K (Run#332-82). Eventually, a crystalline bridgmanite phase was preserved after the recovery in coexistence with some iron metal (Fig. 2E, F). The EELS measurements on the crystalline bridgmanite phase revealed a mixture of Fe2+ and Fe3+ and small variations of the Fe3+/∑Fe ratio was observed across the sample possibly due to the relatively low temperature for the synthesis (Fig. S7). We obtained an average Fe3+/∑Fe ratio of 0.56(6) at 92 GPa and 1850 K, in comparison to dominant ferrous iron in bridgmanite above 105 GPa and 2250 K. The amorphization of bridgmanite in those runs at slightly higher temperatures might indicate the instability of the Fe3+-dominant bridgmanite phase under ambient conditions. These results combined have showed a dramatic pressure effect on the Fe3+/∑Fe ratio of bridgmanite under hydrous conditions with dominant ferrous iron in bridgmanite at depth greater than 2300 km (Fig. 4).

Fe-bearing bridgmanite under dry versus hydrous conditions

To understand the role of H2O in controlling the iron oxidation state, we further compare the chemistry of bridgmanite between dry and hydrous conditions under similar pressure-temperature conditions. We conducted three separate sets of experiments in Fs15 sandwiched between dry silica layers at 112-115 GPa after accounting for the thermal pressure (Table 1). We conducted the first experiment at 114 GPa and 2200 K (Run#344-102d). The TEM images of the recovered sample showed the coexistence of Fe-depleted bridgmanite, a mixture of Fe-rich grain and a silica-rich amorphous phase (Fig. S3). The chemical analysis revealed Fe-depletion in bridgmanite with XFe = 0.11, while the bulk composition of 5×5 μm2 area in the heated center is obtained with XFe = 0.15, identical to the starting composition Fs15 (Table S1). To examine the temperature effect on Fe-depletion of dry bridgmanite, we conducted a separate experiment at a higher temperature of 2350 K and 115 GPa (Run#344-102d-2). We observed a greater Fe-depletion in bridgmanite with XFe = 0.02 at 2350 K compared to the run at 2200 K (Table S1). In agreement with the previous study26, we confirmed Fe-depletion in bridgmanite as a result of the disproportionation reaction. The temperature effect on the disproportionation reaction and crystallization of the H-phase (Fig. S8) will require further investigation (see Supplementary Note 1 for details). The observation of Fe-depletion in bridgmanite under dry conditions, in contrast to the ferrous-iron-dominant bridgmanite without Fe loss under hydrous conditions, demonstrates that H2O stabilizes ferrous iron in bridgmanite under the deep lower mantle conditions at >2300 km depth.

To reveal how H2O stabilizes ferrous iron, we conducted experiments in the Fe2O3-SiO2-H2O system under similar high-pressure-temperature conditions, and obtained FeO in coexistence with hydrous NiAs-type SiO2 in the run products (Fig. 5), indicating that the stability of ferrous iron is a combined result of H2O effect and high pressure. At slightly lower pressures, a hexagonal hydrous phase Fe12.76O18Hx was obtained in the Fe2O3-H2O and FeO-H2O systems, respectively28. The results further demonstrated that the iron valence state under H2O-saturated deep lower mantle conditions is independent on the iron valence state in the starting materials.

Fig. 5: Formation of the FeO phase in the Fe2O3-SiO2-H2O system under high pressure-temperature conditions of the deep lower mantle.
figure 5

A The starting material was Fe(OH)3 sandwiched between dry SiO2 layers (Run#335-95). The pyrite-structured FeOOHx (py-FeOOH) is in coexistence with rhombohedral(R) FeO when H2O is over-saturated in NiAs-type SiO2 (Nt). B A hydrous gel sample with a molar ratio of Fe2O3: SiO2 = 1:4 containing ~2 wt.% H2O was loaded in Ne medium (Run#332-99). The CaCl2-type silica (Ct) is in coexistence with FeO(R) when H2O is unsaturated in the coexisting Nt phase. The X-ray wavelength was 0.6199 Å.

Al3+ effect on the iron oxidation state of bridgmanite under hydrous conditions

The basaltic and pyrolitic compositions contain about 15 wt.% and 3-5 wt.% Al2O3, respectively29. To examine the Al3+ effect on the iron oxidation state under hydrous conditions, we obtained aluminous bridgmanite phase in coexistence with an Al3+-rich hydrous silica phase at 120 GPa and 2200 K using a hydrous gel starting material with all iron in Fe3+. The MgO-Al2O3-Fe2O3-SiO2 gel sample containing 4 wt% H2O has a simplified basaltic composition and has been used in the previous studies23,28. In this sample, gradual amorphization of aluminous bridgmanite was observed, which led to an increase of the Fe3+/∑Fe ratio from 0.46 to 0.75 between two consecutive EELS measurements on one selected grain (Fig. S9). The results indicate that the Fe3+ content of aluminous bridgmanite is coupled to its Al3+ concentration under hydrous conditions. We obtained an atomic ratio Al/(Fe+Al) of ~0.40 in the recovered bridgmanite phase (Table S1), which is equal to the measured Fe3+/∑Fe ratio within the uncertainties. About 60% of iron in aluminous bridgmanite is still in the Fe2+ state despite all iron in Fe3+ in the starting material. Furthermore, our observation showed that amorphization of aluminous bridgmanite could lead to an overestimation of its Fe3+/∑Fe ratio. Future EELS measurements should be performed on crystalline aluminous bridgmanite to establish a relationship between Fe3+ and Al3+ content under hydrous lower mantle conditions.

Bridgmanite is nearly dry in coexistence with a hydrous phase24. Water can be stored in the high-pressure phases of silica in a basaltic composition23,30,31,32. Under the deep lower mantle conditions, the Al3+-rich NiAs-type silica phase with an approximate formula Si0.7Al 0.3O1.85Hx23 could contain up to 4.6 wt.% H2O via the Si4+  =  Al3+ + H+ charge-coupled substitution30,33. On the other hand, the solid solution of δ-phase34 and phase H35, AlOOH–MgSiO2(OH)2, was found stable in coexistence with bridgmanite in a pyrolitic lower mantle system21,22,25. In a system where the water content is lower than the level as simulated in our experiments, we would expect a decrease of water content in the hydrous phase or a smaller proportion of the hydrous phase. As Al3+ preferentially partitions into the coexisting hydrous phase relative to bridgmanite21,22,23,24,25, the Fe3+ content in bridgmanite will be reduced accordingly due to the coupled substitution of Fe3+ and Al3+.

In summary, we investigated the combined effects of H2O and pressure on the chemistry of bridgmanite and obtained the following results: (1) ferric-iron-rich bridgmanite (Mg, Fe)SiO3 was observed under hydrous conditions at depth <2000 km; (2) the presence of H2O in a coexisting hydrous phase stabilizes ferrous iron in bridgmanite at depth >2300 km, in contrast to Fe-depletion in dry bridgmanite (Mg, Fe)SiO3 as a result of the disproportionation; and (3) our preliminary results at 120 GPa and 2200 K indicate that the Fe3+ content is coupled to its Al3+ concentration in bridgmanite under hydrous conditions. The experiments in a hydrated pyrolitic composition have not been conducted yet due to the technical challenges.

Geophysical and geochemical implications

The deep lower mantle structure is dominated by two large low-shear-velocity provinces (LLSVPs) beneath the Pacific and Africa at depth greater than 2300 km36,37,38,39. The calculations of seismic velocities suggested that the LLSVPs with lowered shear-wave speeds and higher-than-average density can be well explained by iron enrichment in a bridgmanite dominant composition40,41,42. The stability of ferrous-iron-dominant bridgmanite under hydrous conditions, in contrast to the disproportionation and iron-depletion in dry bridgmanite26, has important consequences for the deep lower mantle at depth >2300 km. Under deep lower mantle conditions above 60 GPa, the calculated shear-wave-velocity of iron silicate perovskite with composition 25 mol%Fe2O3-75mol%FeSiO3 as a function of pressure43 is nearly parallel to those of MgSiO3 and FeSiO344,45. Considering iron-enrichment in bridgmanite from XFe = 0.10 to 0.15, we obtain about 1.1% increase in density, 1.0% decrease in shear-wave velocity and 0.3% in bulk-sound velocity on the basis of the previous calculations45. The H2O-induced iron-enrichment and stability of ferrous iron in bridgmanite is in general consistent with the character of the LLSVPs46,47, providing an alternative to the basal magma ocean hypothesis48. To constrain geophysical and geochemical models of LLSVPs more quantitively, future research will be needed to examine Al and Fe partitioning between bridgmanite and coexisting phases in both hydrated basaltic and pyrolitic compositions, respectively. In particular, the occurrence of iron spin-pairing in ferropericlase49,50,51 coupled with the Fe3+/∑Fe ratio in bridgmanite could affect the Fe partitioning10.

Importantly, the seismic imaging observations revealed a spatial connection between broad plume-like conduits rooted at the base of the mantle and major hotspots52, and the correlation of hotspot locations within or at the borders of the LLSVPs further supports such a connection53,54. The presence of H2O, even in small concentrations, strongly influences the rheological properties under the mantle conditions55, although the rheological properties under the deep lower mantle conditions remain poorly understood. Whether the primordial noble gases and volatiles56,57,58 are stored in LLSVPs is the subject of continuing debate56,59. However, a dense, low-viscosity layer at the base of the lower mantle may influence plume chemistry and dynamics and be critical in establishing the long-lived conduits in the lower mantle60. In addition, when the upwelling plumes contain such H2O-bearing ferrous-iron-dominant material, disproportionation of ferrous iron would produce Fe metal plus ferric iron at a shallower depth, thus contributing to some variations in the redox conditions of the mantle.


Synchrotron X-ray diffraction

The experiments were conducted in laser-heated diamond anvil cells. Diamond anvils with flat culet diameters of 150 µm beveled at 10° up to 300 μm were used to generate pressure. Each sample was compressed to a target pressure and then heated using a double-sided heating system equipped with Ytterbium fiber lasers. The measured temperature uncertainties were within ±150 K28. Pressure was calibrated by the Raman shift of diamond anvil61 at room temperature after temperature quench. The thermal pressures can be estimated in this pressure range based on the equation Pth (GPa) = (T − 300)  0.006228. The phase assemblages were characterized by XRD measurements conducted at 15U1 beamline of Shanghai Synchrotron Radiation Facility (SSRF) with an X-ray wavelength of 0.6199 Å or at the P02.2 beamline of PETRA III with an X-ray wavelength of 0.2900 Å.

Transmission electron microscope (TEM) analysis

After a sample was recovered to ambient conditions, a cross-section was lifted from the center of heated area and thinned to 50–80 nm in thickness using a FEI Versa-3D focused ion beam (FIB). Elemental mapping and phase chemistry was obtained in a JEOL field emission TEM operating at 200 kV equipped with an EDS system. The EELS data were collected with an aperture of 5 mm, dispersion of 0.05 eV per channel, energy resolution of about 0.65 eV, dwell time from 0.01 to 0.1 s, and scan integration durations of 10–60 s. To enhance the signal-to-background ratio, the EELS data of Run#332-82 were collected with a dispersion of 0.15 eV per channel (Fig. S7). We determined the Fe3+/∑Fe ratio in our recovered samples by a linear combination of Fe L3 reference spectra of the ferrous (FeCO3) and ferric iron (Fe2O3) standards from 703 to 717 eV following the method described by van Aken and Liebscher14. The numbers in parenthesis are one standard deviation based on multiple analyses. We obtained Fe3+/∑Fe = 0.03(1) for our starting material Fs15 opx as a reference (Fig. S5).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.