Thellier-Type Paleointensity Data from Multidomain Specimens

Friday, 19 December 2014
Greig A Paterson1, Andrew John Biggin2, Emma Hodgson2 and Mimi J Hill3, (1)Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing, China, (2)University of Liverpool, Liverpool, United Kingdom, (3)University of Liverpool, Liverpool, L69, United Kingdom
The manifestation of multidomain (MD) grains in paleointensity data is one of the primary reasons for experimental failure and ambiguity in the reliability of accepted results. To better characterize MD effects we take the novel approach of incorporating realistic levels of experimental noise into a phenomenological MD model and systematically exploring the parameter space of paleointensity experiments (the choice of protocol, laboratory and ancient field strengths, and angular dependences). Our model predictions qualitatively and quantitatively compare well with real data from MD specimens and predict recently observed MD behavior. We also explore the quantification of the fraction of natural remanent magnetization used to make a paleointensity estimate and find that FRAC, a recently proposed alternative quantification, is consistently better at isolating accurate results than the traditionally used f. On the basis of ensuring at least an accurate average paleointensity estimate, we recommend minimum FRAC values of 0.65, 0.45, 0.65, and 0.55 for the Thellier, Coe, Aitken, and IZZI protocols, respectively. Building upon this, we use the models and stochastic optimization to develop new sets of selection criteria (MCRIT) designed to maximize the likelihood of accepting accurate estimates from a suite of specimens influenced by MD behavior. Using a quasi-independent synthetic data set generated from the MD model and a fully independent real data set compiled from control paleointensity experiments, we demonstrate that the MCRIT criteria outperform their original counterparts, in terms of their ability to isolate accurate results with low scatter. The use of independent constraints in the data selection process are vital if we wish to remove the arbitrariness that hinders the identification of reliable paleointensity data.