© 2008 Nature Publishing Group LETTERS Evidence for a very-long-term trend in geomagnetic secular variation ANDREW J. BIGGIN*, GEERT H. M. A. STRIK AND COR G. LANGEREIS Palaeomagnetic Laboratory Fort Hoofddijk, Budapestlaan 17, Universiteit Utrecht, Utrecht 3584 CD, Netherlands *e-mail: biggin@geo.uu.nl Published online: 4 May 2008; doi:10.1038/ngeo181 The Earth’s inner core is believed to inhibit rapid fluctuations in the geomagnetic field from developing into full polarity reversals 1,2 . Consequently, during the Precambrian, the smaller size of the inner core might suggest that polarity reversals could occur more readily. It is therefore surprising that there are indications that reversals were rare during this period 3,4 . Here we use new and existing palaeomagnetic data from three continents to examine the stability of the Earth’s magnetic field from 2.82 to 2.45 billion years ago. We show that, on average, geomagnetic secular variation (the field variations produced by normal geodynamo action) during the late Archaean and early Proterozoic was different from that of the past 200 million years; specifically, the apparent variability of the geomagnetic pole as viewed at low and mid-latitudes was reduced relative to the past 200 million years. According to both dynamo simulations 4 and more recent palaeomagnetic field observations 5 , the observed pattern of secular variation suggests a lower frequency of polarity reversals 2.5 billion years ago. This may imply that the geodynamo is becoming progressively less stable over long timescales, consistent with some numerical simulations 6,7 , possibly as a result of changing outer-core geometry that has accompanied inner-core growth. Our confidence in claims 3,4 that the geomagnetic field reversed its polarity less frequently in the Precambrian period (>542 Myr) than in the Phanerozoic is limited by the lack of long continuous Precambrian magnetostratigraphic sections. Fortunately, analyses of geomagnetic palaeosecular variation (PSV)—the record of ancient secular variation—provide an independent means of assessing the stability of the geomagnetic field and have the added advantage that they do not require rocks to form continuous time series as do magnetostratigraphic studies. A recent study analysed PSV (ref. 8) in the late Archaean to earliest Proterozoic and concluded that it was similar to that of the past 5Myr. The present study focuses on approximately the same time period (about 2.45–2.82 Gyr), as it is the oldest for which there is a sufficient (albeit limited) number of suitable data available. However, our dataset is 3.5 times larger than that used previously and is more strictly filtered. A detailed discussion of the differences between the two analyses is given in the Supplementary Information. The new and published palaeomagnetic data (87 and 109 site mean directions respectively; see Table 1) we used for this study were produced from fast-cooled igneous rock units from three continents that met strict criteria of suitability (see the Supplementary Information). We are confident that the magnetic directions used in this study represent the geomagnetic field close to the time the rocks recording them were formed because they are associated with positive fold, conglomerate and/or baked contact tests. Reversal tests were also positive, which suggests that secular variation was adequately sampled. The PSV analysis was kept consistent as far as possible with a previous analysis of PSV in the period 0–5 Myr (ref. 9) (see the Supplementary Information). Figure 1 plots the angular dispersion of the 2.45–2.82 Gyr virtual geomagnetic poles (VGPs) against palaeolatitude, calculated from the inclination of the mean direction assuming that the field was a geocentric axial dipole. Despite the relatively small numbers of site mean data (relative to studies focusing on more recent times), the plotted dispersions are evidently accurate enough to show a clear relationship with palaeolatitude. Furthermore, Fig. 1 shows that the basic shape of the VGP dispersion curves is extremely robust with respect to the specific details of how we carried out our analysis. Figure 1d shows our preferred dataset, which we consider to be a reliable first-order description of PSV in the late Archaean–early Proterozoic. This VGP dispersion curve is markedly different from that produced by the 0–5 Myr dataset (also shown in Fig. 1), suggesting that the nature of PSV was different in these two time periods. Specifically, the angular dispersion of VGPs from the 2.45–2.82 Gyr sites at low to mid-palaeolatitudes is lower whereas the rate of increase in VGP dispersion with palaeolatitude is slightly higher. The data used to produce the 0–5Myr curve shown in Fig. 1 are currently being superseded by the Time Averaged Field Initiative (TAFI) project. These and other recent studies of PSV act to strengthen the conclusions of the present study, as is discussed in the Supplementary Information. The shape of a VGP dispersion curve is dictated by several factors and is consequently difficult to interpret directly in terms of physical processes 10 . Nonetheless, although its physical significance is disputed, Model G (ref. 11) (see the Methods section) has proved effective at fitting the observed variations in the shapes of curves for time windows from the past 195 Myr (ref. 5) (Fig. 2a). Furthermore, the evolution of its shape parameters has been shown 5 to correspond well to the average geomagnetic polarity reversal frequency in the particular time window for which the PSV analysis was carried out (Fig. 2), which suggests a link between mean PSV and reversal frequency. Therefore, Model G may be safely used here to describe the shapes of VGP dispersion curves, regardless of whether interpretations 11 of the physical significance of these shapes are correct. The late Archaean–early Proterozoic model has both shape parameters within errors of the model fit to the Cretaceous Normal Superchron time window (point 5 in Fig. 2). Consequently, the empirical relationship shown in Fig. 2b predicts that the nature geoscience VOL 1 JUNE 2008 www.nature.com/naturegeoscience 395