Multivariable Dependence of Fe-Mg Partitioning in the Lower Mantle Ho-kwang Mao,* Guoyin Shen,† Russell J. Hemley High-pressure diamond-cell experiments indicate that the iron-magnesium partitioning between (Fe,Mg)SiO 3 -perovskite and magnesiowu ¨ stite in Earth’s lower mantle depends on the pressure, temperature, bulk iron/magnesium ratio, and ferric iron content. The perovskite stability field expands with increasing pressure and temperature. The ferric iron component preferentially dissolves in perovskite and raises the apparent total iron content but had little effect on the partitioning of the ferrous iron. The ferrous iron depletes in perovskite at the top of the lower mantle and gradually increases at greater depth. These changes in iron-magnesium composition should affect geochemical and geo- physical properties of the deep interior. At depths greater than 670 km in the lower mantle, olivine and orthopyroxene transform to (Fe,Mg)SiO 3 -perovskite (pv) and (Fe,Mg)O-magnesiowu ¨stite (mw) (1), which are likely to form the most abundant mineral assemblage in the Earth (2, 3). The Fe-Mg partition coefficient of pv and mw, K = (X Fe pv /X Mg pv )/(X Fe mw /X Mg mw ), where X is the molar fraction of Fe (Fe 2+ and Fe 3+ ) or Mg in pv or mw, plays a crucial role in deter- mining the composition (4–6) and physical properties (7–10) of the lower mantle, the interaction between the solid oxide mantle and the molten metallic core (11–13), the partitioning of siderophile elements (14– 16), and the stability of pv and mw (17). However, reported K values have varied enormously, from 0.08 to 1 (Table 1), ren- dering lower mantle geochemistry essential- ly unconstrained. The upper limit of Fe in pv, X Fe pv , has also been a subject of contro- versy. Yagi et al. (18) have reported X Fe pv as high as 0.21 for pv synthesized in diamond cells, whereas X Fe pv  0.15 has never been observed in pv synthesized with multianvil apparatus (17). Guyot et al.(19) have reported that K increased sharply with pressure from 25 to 40 GPa but remained constant above 40 GPa. The constancy of K at higher pressures has generally been accepted and used for models of the deep lower mantle. Multian- vil experiments have provided information on the dependence of K on other variables, including the Fe (20) and Al 2 O 3 content (21, 22), the presence of a B 2 O 3 catalyst (23), the equilibration time (23, 24), and the temperature (24). With the addition of 4% Al 2 O 3 (21), the Fe content in pv in- creases drastically to become equal to that in mw. The extremely large K value in the presence of Al 2 O 3 , however, implies (22) an extremely high Fe 3+ /Fe fraction (50 to 90%) (25), a composition unlike that ex- pected for the lower mantle. The central issue pertinent to the composition of the lower mantle remains the role of ferrous iron, and the behavior of Fe 2+ and Fe 3+ must be evaluated separately. We used a laser-heated diamond cell and synchrotron x-ray diffraction (26) to deter- mine the K value as a function of pressure (P), temperature (T), and composition (X) at conditions deep into the lower mantle. Starting samples were synthetic olivines (Fo 100 , Fo 82 , Fo 65 , Fo 60 , and Fo 35 ) or or- thopyroxenes (En 100 and En 60 ) with all Fe in ferrous form (27). To investigate the effects of Fe 3+ , we added 4  1 mole per- cent hematite. The diamond cell is a closed system, in which the bulk Fe 3+ /Fe 2+ ratio in the starting material is preserved in the absence of other transition elements (unless metallic Fe is produced). We used the dou- ble-sided, multimode laser-heating tech- nique (26) to attain uniform and constant temperatures (50 K) in heated areas 30 to 50 m wide (Fig. 1A). Samples were sand- wiched between two NaCl thermal insula- tion layers (26). Temperatures were in- creased and fixed for at least 5 min at 1500, 1800, and 2000 (50) K (Fig. 1B), which are sufficiently below the melting point to minimize preferential diffusion (28). Com- positional uniformity within the 50-m heated area was confirmed by x-ray diffrac- tion and electron microprobe. Pressures were calibrated with ruby chips situated near the hot spot; all reported pressures refer to measurements after heating. A total of 38 experiments were per- formed at myriad P-T-X conditions within the pv stability field. In each experiment, the sample had transformed to pv + mw whereas the surrounding unheated area re- mained as the starting phase. A 5-m syn- chrotron x-ray beam (29) was used to probe the sample after it was quenched (30). We used the unit-cell parameters of pv and mw to calculate X Fe pv (17) and X Fe mw (31). We tested the effects of variable heating time and found that the value of K did not change significantly (10%) in heating times lasting 5 to 30 min, indicating that 5 min was sufficient. Equilibrium and revers- ibility were tested by approaching the same P-T point along different thermodynamic paths. Geophysical Laboratory and Center for High Pressure Research, Carnegie Institution of Washington, 5251 Broad Branch Road, NW, Washington, DC 20015 –1305 USA. * To whom correspondence should be addressed. †Present address: Consortium for Advanced Radiation Sources, University of Chicago, Chicago, IL 60637 USA. Table 1. High P-T partition coefficients (K ) obtained by multianvil and diamond-cell experiments. The list was selected to highlight extreme values and multivariable dependence; ts, this study. Dashes indicate that temperatures were not measured or reported for that experiment. K Samples T (K) P (GPa) Reference Multianvil 0.09 Fo 80 1873 26 (20) 0.15 Fo 90 1873 26 (20) 0.16 Fo 70 + B 2 O 3 1900 23 (23) 0.35 Fo 91 + B 2 O 3 1900 23 (23) 0.26 Fo 91 1573 26 (24) 0.23 Fo 91 1873 26 (24) 0.17 Fo 89 1600 25 (21) 1.0 Fo 89 + Al 2 O 3 1600 25 (21) Diamond cell 0.08 Fo 93 -Fo 59 – 25 –35 (35) 0.16 Fo 83 2500  750 25 (19) 0.29 Fo 83 2500  750 40 (19) 0.13 Fo 85 -Fo 73 1673 26 (32) 0.12 Fo 90 + Cr, Ni, Mn – 30 (16) 0.50 Fo 90 + Cr, Ni, Mn – 25 (16) 0.04 Fo 60 1500  50 32 ts 0.29 Fo 82 2000  50 50 ts 0.09 Fo 65 1500  50 40 ts 0.12 Fo 65 + Fe 2 O 3 1500  50 37 ts 0.11 Fo 82 1800  50 31 ts 0.24 Fo 82 + Fe 2 O 3 1800  50 29 ts REPORTS SCIENCE  VOL. 278  19 DECEMBER 1997  www.sciencemag.org 2098