When the Compass Stopped Reversing Its Poles
Subir K. Banerjee* 

 Earth's magnetic field reverses a few times every million years at random intervals, as a result of positive feedbacks to magnetohydrodynamic instabilities within the liquid iron core (1). Occasionally, however, the dynamo mechanism in Earth's liquid iron core stops its dance of random dipole field reversals and for 30 to 50 million years maintains either "normal" polarity (like today) or the reversed state (in which a compass would point south). At least two such superchron periods are known in the recent geological past. Between 118 and 83 million years ago (Ma) (the Cretaceous superchron), the field maintained constant normal polarity, and it remained reversed from 312 to 262 Ma. Why did the random dipolar reversal suddenly stop, and what made the stochastic dance start again?

 To answer this question, we require accurate data of geomagnetic field behavior during a superchron, so as to devise and test models of geomagnetic dynamo behavior that could cause it. On page 1779 of this issue, Tarduno et al. (2) report the latest in a series of innovative attempts to accurately determine the magnitude of the magnetic field at Earth's surface and the virtual dipole moment (VDM) at Earth's center that is responsible for it before, during, and after the Cretaceous superchron (3-5). Unlike low values found by previous authors, however, Tarduno et al. find that during the superchron, the time-averaged VDM was 12 x 1022 A/m2--twice as high as the average for the past 160 Ma (4) and 50% higher than today.

[Figure 1] 
A site of more than scientific interest. This Buddha sculpture is at the famous Buddhist school at Nalanda, near the Rajmahal volcanic field.

Obtaining a truly accurate datum that is not contaminated by local effects unrelated to the global dipole field is a real tour de force. Tarduno et al. used a multistep polishing and etching method to extract 149 single crystals of plagioclase feldspar from eight independent lava flows that erupted at different times between 113 and 115 Ma. Other researchers have previously performed paleomagnetic and rock magnetic studies at the same site, the Rajmahal volcanic field near the Bihar/West Bengal state boundary in eastern India. But unlike previous investigators, Tarduno et al. concentrated on measuring the weak (~10-11 A/m2) but geologically stable magnetic moment from millimeter-sized single crystals of plagioclase, rather than whole rocks. The reason was simple: Although whole rocks ~2.5 cm in size may have more easily measurable magnetic moments, laboratory reheating inevitably causes some degree of chemical alteration, compromising the reliability of the measurement. Tarduno et al. show that whole rock samples contain clay minerals formed since crystallization by weathering at Earth's surface. These rapidly break down during laboratory reheating and produce new, very effective carriers of magnetic remanence (the magnetization that remains after the removal of an external field). This excess magnetite conspires to produce a remanent magnetization of the same magnitude as observed today in nature (the natural remanent magnetization, NRM) but in a weaker field than the original geomagnetic paleointensity. In contrast, the tiny (100- to 350-nanometer diameter) magnetite inclusions sealed inside the plagioclase crystals are safe from chemical change and thus yield the true paleointensity during the Cretaceous superchron. This leads to the deduced high value for the VDM.

 Of the 149 single crystals of plagioclase that were studied, 56 satisfied rigorous reproducibility tests. The latter come from eight lava flows that erupted at different times within 0.1 to 1 Ma. The average value from all the flows and all the 56 crystals with reproducible data can thus be taken to represent the true dipolar field during Cretaceous superchron. The averaging process removes potential sources of error in VDM magnitude due to the highly variable (in space and time) nondipole components of the total geomagnetic field. These have shorter than 10,000-year periodicities and hence cancel out when data are averaged over 0.1 to 1 Ma.

 The fact that it represents a reliable time average distinguishes Tarduno et al.'s VDM value from those deduced with an earlier, equally innovative method that targeted submarine basaltic glasses found in ocean sediments (6). The magnetite carriers of NRM are again protected from chemical change in the laboratory by their natural glass armor. However, unlike the layered stratigraphy of the Rajmahal volcanic lavas, individual submarine glass samples cannot be shown to be from sources that erupted over a sufficiently long time interval to cancel out nondipole contributions. This may explain why studies based on submarine basaltic glasses (4, 5) have yielded equivocal answers as to whether the VDM was unusually high or low during the Cretaceous superchron.

 One may argue that compared with the 35-Ma length of the superchron, we have so few reliable VDM values that the jury is still out. But there is some conceptual and model support for the high value of Tarduno et al. The late Allan Cox suggested (7) that during times of high dipole field strength, nondipoles cannot randomly drive the total field to reversal and furthermore (8) that even though core reversals may be stochastic, their frequency may be set and reset in intervals of ~100 Ma by processes in the core-mantle boundary. And in a recent three-dimensional numerical modeling study by Glatzmaier et al. (9), the field did not reverse when the imposed core-mantle boundary heat flux pattern was in phase with the convected flux from the liquid core. This led to very high dipole field values.

 The tales of Panchatantra, a Buddhist book of fables, were written within a few hundred kilometers of the volcanic field of Rajmahal. It tells the well-known story of five blind men in complete disagreement about the shape and size of an elephant that they could all touch (but at different spots on the elephant). We need many more "seeing eyes" to explain why superchrons occur, but we are one step closer to knowing what characterized them.

1. R. T. Merrill et al., The Magnetic Field of the Earth (Academic Press, San Diego, CA, 1996) [publisher's information].
2. J. A. Tarduno, R. D. Cottrell, A. V. Smirnov, Science 291, 1779 (2001).
3. M. Perrin, V. Shcherbakov, J. Geomagn. Geoelectr. 49, 601 (1997).
4. M. T. Juarez et al., Nature 394, 878 (1998) [GEOREF].
5. P. Selkin, L. Tauxe, Philos. Trans. R. Soc. London Ser. A 358, 1065 (2000).
6. T. Pick, L. J. Tauxe, J. Geophys. Res. 98, 17949 (1993).
7. A. J. Cox, J. Geophys. Res. 73, 3247 (1968).
8. ------, Rev. Geophys. Space Phys. 13, 35 (1975).
9. G. A. Glatzmaier et al., Nature 401, 885 (1999).

The author is at the Institute for Rock Magnetism, School of Earth Sciences, University of Minnesota, Minneapolis, MN 55455, USA. E-mail:

Volume 291, Number 5509, Issue of 2 Mar 2001, pp. 1714-1715.
Copyright © 2001 by The American Association for the Advancement of Science.