TOTAL ENERGIES OF DIFFERENT ANTIGORITE STRUCTURE MODELS: A DFT STUDY

The serpentine mineral antigorite is known to be stable up to 6 GPa (200 km) at 620°C [1]. Under more extreme conditions, it breaks down to a dehydrated assemblage (Atg = Ol + Tlc (En) + H2O). Earthquakes at intermediate mantle depth may be triggered by dehydration embrittlement caused by antigorite breakdown. In spite of the importance of antigorite at subduction zones, little is known of its properties at high pressures. Knowledge of the elasticity of antigorite is important for the interpretation of seismological observations at subduction zones and for the detection of antigorite in this and other geological settings. This knowledge may be achieved, in principle, either experimentally or computationally. However, both routes are troublesome. Experimentally, one is faced with the high density of defects usually affecting single-crystal sized specimens, and with the difficulties in dealing with such a large unit cell and a DAC-shadowed reduced dataset. Computationally, the antigorite system is quite large and thus very costly. Moreover, a good guess of the crystal structure is mandatory to enhance the chances to successfully achieve any reliable result. We report here the results of an ab initio quantum mechanical investigation of the antigorite structures proposed in the recent literature. The structures investigated are the m = 17 and the m = 16 polysomes, XRD refined by Capitani & Mellini [2, 3] and a second m = 17 polysome proposed by Dódony & co-workers [4]. Basically, the first two structure models differ from the last by the presence of tetrahedral 8-reversals and the absence of octahedral offsets. The total free energy for the three models after structural relaxation are -6.983189, -6.986766, -6.977770 eV/atom, respectively. From these results it turns out that the models with tetrahedral 8-reversals and without octahedral offsets are energetically favoured with respect to the model excluding tetrahedral 8-reversals and including octahedral offsets. For c