Résumé
A refined palaeointensity experiment, accompanied by rock-magnetic studies, has been carried out on six lava flows from 1910 and 1928 from Mt. Etna. The purpose of the study was to try to understand why these very young basaltic flows are generally unable to provide a correct estimate of the magnitude of the ambient magnetic field during flow cooling. Susceptibility versus temperature curves and ore microscopic studies show that 3 types of magnetic minerals (phases "h", "m" and "l") are present in these flows, some samples containing a single largely dominant magnetic phase while others contain a mixture of several phases. Phase "h" is a thermally stable, near magnetite phase resulting from titanomagnetite oxyexsolution. Phase "l" is a thermally stable titanomagnetite with a Curie temperature of approximately 200ºC. Phase "m" is a titanomagnetite phase of Curie temperature between 450-490ºC which is unstable at temperatures above 400ºC. In addition to the usual reliability checks of the Thellier method (NRM-TRM linearity, pTRM checks), our paleointensity experiments included additional heatings allowing determination of the MD or PSD-SD character of each pTRM and determination of CRM or transdomain remanences possibly acquired during heating. From the 28 samples studied 20 provide a linear NRM-TRM plot over about 1/4 or more of total NRM. However, only six of them, all containing near-magnetite as a single phase, display positive pTRM checks. Nevertheless, these six samples yield a mean paleointensity of about 52 µT, which exceeds the real field paleomagnitude (42 µT) by some 25%. The reasons for this almost-total failure of paleointensity experiments are diverse. For samples with a dominant "l" phase, pTRMs present a behaviour typical of large MD grains, with as much as 1/3 of remanence with unblocking temperatures exceeding the blocking range. No CRM is acquired. Yet a remanence does develop during heating in a field (followed by cooling in zero field). We suggest that this remanence is a transdomain remanence resulting from domain rearrangements. These two observations are in conflict with some of the basic requirements of the Thellier method. Phase "m" seems chemically stable up to 400°C but pTRM changes start at lower temperatures. This thermal instability and the development of a significant transdomain remanence seem to be the causes of the failure of paleointensity experiments. The reason for the rather large (and quite unexpected) error in the average paleointensity provided by the samples containing only a near magnetite (phase "h") may lie in the fact that the low to medium temperature pTRMs, which represent a significant fraction of the total TRM, seem to be carried by small MD particles. In conclusion, several modifications of the Thellier method are proposed.