Magnetic field intensity is one of Earth's fundamental properties and its temporal behavior has implications in fields ranging from geodynamics to archeology. Thermal remanent magnetization (TRM) has the strongest theoretical basis of all the forms of natural remanent magnetization, and natural and archeological materials have been used to estimate paleointensities for decades. Although founded on sound theory for ideal samples (those that produce linear Arai plots), paleointensity estimation is challenging with non-ideal samples, which are more abundant in nature and widely used in experiments.
We examined the behavior of natural samples using both original and laboratory- acquired TRMs during paleointensity experiments and characterized them based on proxies for domain state including curvature, k, and bulk domain stability parameters. We then investigated their capacity to retain a record of the magnetic field. Samples taken from previous experiments were separated into straight and curved groups representing single-domain-like from multi-domain-like remanences, respectively, based on a critical threshold value, k = 0.164.
Specimens from the two sets were given a fresh TRM in a 70 μT laboratory field and subjected to an infield-zerofield, zerofield-infield (IZZI)-type paleointensity experiment. Straight specimens recovered the laboratory field with high precision while curved specimens produced more scattered results. However, both sets closely recovered the average laboratory field, which suggests that experiments containing a sufficient number of specimens can avoid large biases in the field estimate.
We also found that the dependence of cooling rate on the laboratory TRM was significant in most samples. However, it did not depend on their inferred domain states and should be estimated for all samples whose cooling rates differ from the laboratory field. Our results confirm that while ideally behaved specimens can produce accurate and precise paleofield estimates, non-ideal, or curved, specimens produce more scattered, although unbiased, estimates.