On the suitability of the Thellier method of palaeointensity determinations on pseudo‐single‐domain and multidomain grains

Summary We report an experimental and theoretical study of non‐linear Arai–Nagata diagrams for samples containing pseudo‐single‐domain (PSD) and multidomain (MD) magnetite. Our aim is to reveal the physical reasons for the deviation of these plots from ideal straight lines. Contrary to expectations, the concavity of the Arai–Nagata diagrams is not related to the two most noticeable violations of the Thellier laws documented for non‐single‐domain particles: the tail of partial thermoremanence and the dependence of the magnitude of pTRM on the thermal history of the sample. Indeed, the contributions from these two factors mutually cancel each other. Phenomenologically, the non‐linear Arai–Nagata plots occur because samples during low‐temperature stages of the Thellier procedure lose too much remanence and recover too little of it. The excessive loss of the previously imparted total TRM is due at least partly to some partial demagnetization of high‐temperature TRM components and to progressive stabilization of domain structure during the repetitive heatings to moderate temperatures that form the basis of the Thellier procedure. For natural MD samples a linear fit to the low‐temperature data points on the Arai–Nagata plots leads to a palaeointensity overestimated by as much as 60 per cent. The samples containing hydrothermally grown or crushed and sieved MD magnetite provide low‐temperature apparent palaeointensities two to three times larger than the correct value. For small PSD samples the overestimate is less than 10–20 per cent and, in general, PSD samples can be used for the palaeointensity determinations.