ALBERT

Description: The determination of the natural remanent magnetization (NRM) of archaeological features can be used for magnetic modelling, joining of shards, archaeomagnetic dating or the investigation of the firing–cooling–collapsing order of ancient buildings. The measurement of NRM is normally conducted on cylindrical or cubic samples in the laboratory. Nevertheless, archaeological finds should preferably not be destroyed, and laboratory instruments are high in costs. Therefore, we propose a lightweight and portable measurement set-up including already available field magnetometers (preferably caesium magnetometers) in which the archaeological sample of arbitrary shape, in our case a piece of daub, is mounted inside a gimbal to be rotated in all directions. The magnetic field of the sample is measured at a large number of rotational positions with the magnetometer kept at fixed position. In these measurements, the unknown direction of the NRM vector of the sample is rotated, whereas the average magnetic susceptibility of the sample and the ambient magnetic field are constant and known. Hence, the vector of NRM can be determined through least-squares inversion. For the inversion computation, the sample volume is discretized either as voxel model or approximated as an equivalent sphere. Under certain conditions depending on sample–sensor distance, dipole moment and radius of the sample, the approximation by a sphere is valid without effect on the accuracy of results. Empirically determined functions quantifying these conditions for different sensor sensitivities and noise levels are provided. Validation with laboratory measurements on palaeomagnetic subsamples from the destroyed daub samples indicate that the NRM can be determined by our proposed method with a maximum error in inclination of 2°, in declination of 20° and in magnetization of ±0.6 A/m. This is accurate enough, for example, to determine from daub pieces of burnt house remains whether the building was burnt and cooled before or after it collapsed.