Low-temperature evolution of OH bands in synthetic forsterite, implication for the nature of H defects at high pressure

We performed in situ infrared spectroscopic measurements of OH bands in a forsterite single crystal between -194 and 200 A degrees C. The crystal was synthesized at 2 GPa from a cooling experiment performed between 1,400 and 1,275 A degrees C at a rate of 1 A degrees C per hour under high silica-activity conditions. Twenty-four individual bands were identified at low temperature. Three different groups can be distinguished: (1) Most of the OH bands between 3,300 and 3,650 cm(-1) display a small frequency lowering (< 4 cm(-1)) and a moderate broadening (< 10 cm(-1)) as temperature is increased from -194 to 200 A degrees C. The behaviour of these bands is compatible with weakly H-bonded OH groups associated with hydrogen substitution into silicon tetrahedra; (2) In the same frequency range, two bands at 3,617 and 3,566 cm(-1) display a significantly anharmonic behaviour with stronger frequency lowering (42 and 27 cm(-1) respectively) and broadening (similar to 30 cm(-1)) with increasing temperature. It is tentatively proposed that the defects responsible for these OH bands correspond to H atoms in interstitial position; (3) In the frequency region between 3,300 and 3,000 cm(-1), three broad bands are identified at 3,151, 3,178 and 3,217 cm(-1), at -194 A degrees C. They exhibit significant frequency increase (similar to 20 cm(-1)) and broadening (similar to 70 cm(-1)) with increasing temperature, indicating moderate H bonding. These bands are compatible with (2H)(Mg) defects. A survey of published spectra of forsterite samples synthesized above 5 GPa shows that about 75 % of the incorporated hydrogen belongs to type (1) OH bands associated with Si substitution and 25 % to the broad band at 3,566 cm(-1) (type (2); 3,550 cm(-1) at room temperature). The contribution of OH bands of type (3), associated to (2H)(Mg) defects, is negligible. Therefore, solubility of hydrogen in forsterite (and natural olivine compositions) cannot be described by a single solubility law, but by the combination of at least two laws, with different activation volumes and water fugacity exponents.