High-temperature crystal chemistry of hydrous Mg- and Fe-cordierites

Abstract

Structural refinements have been completed for a Mg-rich cordierite using data recorded at 24,375o,775o and 24C (after heating to 775) and for an Fe-rich cordierite at 24 and 375'C. The mean T-O bond lengths in both cordierites remain unchanged but the mean octahedral bonds (M-O) lengthen upon heating. The unusually low thermal expansion of the Mg-cordierite is the result of its relatively rigid tetrahedral framework and the anisotropic expansion ofoctahedra isolated from each other. This anisotropic expansion leads to a slight rotation of the six-membered rings, a concomitant collapse of the structure parallel to c, and an expansion parallel to a and 6. In the Fe-cordierite, the octahedron is more flattened, resulting in c being smaller and a and b being larger than the cell dimensions of the Mgcordierite. Upon heating Fe-cordierite, there is no evidence for a rotation of the rings, and a, b, and c increase as the M-O bonds expand. X-ray Ap maps calculated for the Mg-cordierite showed approximate positions and relative amounts ofchannel constituents. The peak ascribed to the alkali and other atoms that centers the six-membered rings becomes elongated parallel to c upon heating through 375.c. However, the peak ascribed to the oxygen associated with HrO in the 24' and 375' maps is absent in the 775'C maps. It reappears in maps computed from the 24 (after heating) data. In both cordierites, small amounts of hematite were produced during heating (prematurely halting data collection on the Fe-cordierite), and apparently formed by combination of octahedral and channel iron with oxygen from the channel water molecules. A re-examination of the water-orientation the channels o[ the Mg-cordierite using neutron and X-ray Ap maps does not clearly show either type I [H-o-H in the (100) plane with the H-H vector parallel to cl or type II [H-o-H in the (100) plane with the H-H vector parallel to b] water, as previously suggested by spectroscopic studies. Instead, our ap maps indicate that the water molecule lies in a plane tilted -29o from (100) and that the H-H vector is tilted -l9o from c.