| Abstract |
Rock magnetism and magnetic potential field data are influenced by a number of factors, including mineral composition, hydrothermal alteration, faults, and past and present temperatures. Laboratory measurements on rock samples from Reykjanes show that mid-oceanic ridge basalts (MORB) possess large remanent magnetization and have two or three different Curie temperatures. This makes the inversion and interpretation of such data challenging. The magnetization of minerals and rocks can be seen a battle between quantum-mechanical order and thermal disorder. The transition between these domains takes place around the Curie/Neel temperature. Below the Curie temperature the rock is ferromagnetic or ferrimagnetic, above the Curie temperature, only a weak paramagnetism or diamagnetism remains. For this reason, we may expect that magnetic data and the magnetization may be a useful source of information about the presence of working geothermal systems. The carriers of the magnetization of the mid-oceanic basalts are primarily titanomagnetite and secondarily titanomaghemite, the latter created by oxidation of the former. Titanomagnetite is known to be a solid solution of ulvöspinel and magnetite, formed when the basaltic lavas are cooled and solidified. Upon further cooling, the solid solution becomes unstable. Exsolution takes place when the titanomagentite enters the miscibility gap of the phase diagram, and leads to decomposition of initially homogeneous titanomagnetite into ulvöspinel-rich and magnetite-rich domains. We present a Bayesian inversion method for estimation of subsurface temperature from magnetic potential field data. The inversion is performed in two steps: First, we compute the magnetization, including both the induced and the remanent parts, by inversion of magnetic data. Due to the poor depth-resolution of potential field data, we use a Marquardt-Levenberg type map inversion method, computing laterally varying magnetization averaged over the relevant depth interval. The inputs to the inversion are (scalar) total magnetic anomaly and magnetic background field, including inclination and declination. In young lavas, we assume that the paleomagnetic field at the time of lava deposition is approximately the same as the present-day field. Second, we use a phenomenological model relating magnetization to temperature to perform Bayesian inversion for subsurface temperature. The phenomenological model is based on spinodal exsolution of titanomagnetite and the Ising model from quantum statistics. The Ising model accounts for spin-spin interactions and interactions between spin and the external magnetic field. Therefore, it represents both the ferromagnetic and paramagnetic domains and reduces to the Currie-Weiss law in the high-temperature limit. The solid exsolution model describes the reorganization of initially homogeneous titanomagnetite into magnetite-rich and ulvöspinel-rich domains, with two different Curie temperatures. The spinodal exsolution model is calibrated to laboratory measurements from the literature. The exsolution model is aimed at fast-cooled lavas with small or moderate oxidation of titanomagnetite. The proposed Bayesian inversion scheme is demonstrated on high-resolution magnetic data acquired at Reykjanes, SW Iceland. Preliminary results indicate that lateral variations in subsurface temperature and hydrothermal alteration can be detected by the proposed inversion scheme. Inversion of magnetic data, like all geophysical data, gives ambiguous results. Therefore, the aim and purpose of the proposed magnetic inversion are to include it in a multigeophysical inversion scheme, together with electromagnetic, seismic and gravity data. |