Recent studies comparing density (&rgr;) from the seismological model PREM to that predicted for various perovskite and magnesiow¿stite phase assemblages under lower mantle conditions suggest that the lower mantle is a few percent denser than reasonable candidates for upper mantle composition such as pyrolite. This has been interpreted as evidence for an iron-enriched lower mantle. In addition to density, bulk sound velocity (√&PHgr;) provides an important constraint upon proposed lower mantle compositions. Being independent of the shear modulus. √&PHgr; has several desirable characteristics. Experimentally, it can be determined in the laboratory, by density measurements under static compression; seismologically, it is not affected by attenuative dispersion and is expected to be much less laterally heterogeneous than P- or S-wave velocities. We have calculated &rgr; and √&PHgr; along lower mantle adiabats, as functions of iron-magnesium content (&KHgr;Mg) and silica content (XPv), and compared to them to values obtained from PREM. For a reasonable temperature estimate at the top of the lower mantle (Tlm) of 2000 K, we find that the &rgr; data are most compatible with a lower mantle enriched in Fe, in agreement with previous studies.

In addition, the √&PHgr; data require that silica enrichment accompany this Fe enrichment. Furthermore, increasing Tlm increases both the required Fe and Si content. With a high assumed Tlm (>2700 K), the &rgr; data can be satisfied by low XMg values, but even a pure perovskite lower mantle is too slow compared to √&PHgr;. Thus, assuming no free SiO2 in the lower mantle, this provides an upper bound on Tlm of approximately 2700 K. Finally, conclusions regarding the composition and temperature of the lower mantle are strongly dependent upon uncertainties in estimated thermoelastic parameters and upon choice of a seismological model. Most importantly, neither Fe nor Si enrichment is required if the zero-pressure volume coefficient of thermal expansion of perovskite is about 2.5¿10-5 K-1, rather than 4.0¿10-5 K-1, at high temperatures. Alternatively, but less probably, Fe and Si enrichment are not required if the average &rgr; of the lower mantle is 1.0% less than given by PREM. The Si enrichment is not required if the average √&PHgr; of the lower mantle is 0.5% slower than PREM, closer to the value given by the body-wave model derived from the Jeffreys-Bullen tables, or if the lower mantle is anomalously cold, with Tlm≈1700 K. ¿ American Geophysical Union 1990