A simple test for the presence of multidomain behavior during paleointensity experiments

1 Detecting and excluding non-ideal behavior during paleointensity experiments is critical to asserting the reliability of data. Our knowledge of detecting non-ideal behavior, in particular the influence of multidomain (MD) grains, has expanded considerably over the past decade and experimental procedures now commonly incorporate checks to detect the effects of MD behavior. However, many older studies were carried out before these checks were devised and provide no quantifiable means of testing for the presence of MD grains. An estimated one third of all entries in the most recent paleointensity database do not include some form of check for MD behavior. The reliability of these results is therefore questionable and can only hinder efforts to understand the evolution of the geomagnetic field and the geodynamo. I propose a simple phenomenological check that can be applied to previous studies, provided that the raw data are available, that will allow the exclusion of MD behavior and provide a means of identifying reliable data. The check is a quantification of the curvature, k, of data points on an Arai plot, a feature commonly associated with MD behavior. Analysis of paleointensity data from samples with known grain size indicates that this new parameter is significantly correlated with grain size and with the accuracy of the paleointensity estimates made from both limbs of the curved data. Analysis of 181 samples from five historical data sets indicates that k is significantly correlated with experimentally obtained MD and alteration check parameters, and the accuracy of the paleointensity estimate. A threshold selection value of k ≤ 0.164 can be defined using the samples with known grain sizes. Applying this cut-off value, combined with a threshold on the quality of the circle fit and a commonly used alteration check, to the historical data yields an accurate result with low scatter. When compared with previously published selection criteria that incorporate experimental checks for non-ideal behavior, the result of applying the criteria proposed here is an improvement. The application of these three criteria rejects over 65% of all inaccurate results and has the highest concentration of accurate results when compared with the other criteria sets tested. Other selection criteria can be subsequently used to improve on this result. While modern studies should always include experimental checks to identify MD behavior, this new criterion will provide a useful tool for future studies and, importantly, a method to assess the reliability of previously published data.