Attribution: You must attribute the work in the manner specified by the author or licensor (but no in any way that suggests that they endorse you or your use of the work).Noncommercial ‒ you may not use this work for commercial purpose.No Derivative works ‒ You may not alter, transform, or build upon this work.Sharing ‒ Individual scientists are hereby granted permission, without fees or further requests to GSA, to use a single figure, a single table, and/or a brief paragraph of text in other subsequent works and to make unlimited photocopies of items in this journal for noncommercial use in classrooms to further education and science.


The fact that the natural remanent magnetization (NRM) intensity of mid-oceanic-ridge basalt (MORB) samples shows systematic variations as a function of age has long been recognized: maximum as well as average intensities are generally high for very young samples, falling off rather rapidly to less than half the recent values in samples between 10 and 30 Ma, whereupon they slowly rise in the early Tertiary and Cretaceous to values that approach those of the very young samples. NRM intensities measured in this study follow the same trends as those observed in previous publications. In this study, we take a statistical approach and examine whether this pattern can be explained by variations in one or more of all previously proposed mechanisms: chemical composition of the magnetic minerals, abundance of these magnetization carriers, vectorial superposition of parallel or antiparallel components of magnetization, magnetic grain or domain size patterns, low-temperature oxidation to titanomaghemite, or geomagnetic field behavior. We find that the samples do not show any compositional, petrological, rock-magnetic, or paleomagnetic patterns that can explain the trends. Geomagnetic field intensity is the only effect that cannot be directly tested on the same samples, but it shows a similar pattern as our measured NRM intensities. We therefore conclude that the geomagnetic field strength was, on-average, significantly greater during the Cretaceous than during the Oligocene and Miocene.


Linear magnetic anomalies provide a record of Earth's magnetic reversals and have led to the recognition of seafloor spreading (Vine and Matthews, 1963). In addition to the polarity of the magnetizations, the variations in magnetic amplitude of the anomalies have attracted widespread attention. These amplitude variations have, in turn, been shown to correspond to variations in the intensity of the natural remanent magnetization (NRM) (Bleil and Petersen, 1983; Johnson and Pariso, 1993, and references therein). A sharp decrease of NRM intensity back in time is observed to culminate in a low at ∼10–30 Ma, as can be seen in Figure 1. Before this time, the upper envelope of the NRM intensity for older basalts rises gradually with age to values that approach those of recent basalts near the mid-ocean ridges (Bleil and Petersen, 1983; Wittpenn et al., 1989; Johnson and Pariso, 1993). In essence, the NRM intensity values of the Cretaceous samples are more than double those determined for Tertiary samples with ages between 10 and 50 Ma (Fig. 1C). Moreover, Sayanagi and Tamaki (1992) have documented that this pattern is seen in the North Atlantic as well as the North Pacific Ocean, which suggests that this is a global phenomenon.

The lower NRM intensities in Oligocene and Miocene ocean-floor basalts are readily attributed to a degradation of the remanence due to low-temperature oxidation from titanomagnetite to nonstoichiometric titanomaghemite, because the latter has a lower intrinsic magnetic intensity (Irving et al., 1970; Marshall and Cox, 1972; Özdemir and O'Reilly, 1982; Prévot et al., 1983; Moskowitz, 1987; Beske-Diehl, 1990; Matzka et al., 2003; Wang et al., 2005). However, the increase to relatively high values (see Fig. 1) for mid-oceanic-ridge basalts (MORB) of Cretaceous age is rather puzzling, as low-temperature oxidation can hardly be expected to be a reversible process. If we can assume that low-temperature oxidation progresses steadily (even if not linearly) with age, and also can take for granted that NRM intensity keeps on diminishing with increasing oxidation (Xu et al., 1997), then it could be argued that the higher intensity of the Cretaceous rocks means that the paleofield intensity would have had to have been very high indeed, if all other factors are seen as inconsequential. Alternatively, alteration may not be such an important factor in influencing magnetic properties, including NRM intensity (Gee and Kent, 1999).

Bleil and Petersen (1983) proposed that the higher Cretaceous MORB intensity could be explained by cation ordering at very high oxidation levels, and this would be testable by examining the change in degree of oxidation (described by the z-parameter, which varies from 0 to 1). Similarly, we can test for other rock-magnetic effects, such as those related to grain or domain size, volumetric abundance of remanence carriers, or systematic chemical composition variations. In addition, we have demagnetized our MORB samples to investigate whether the presence (or absence) of a multivectorial nature of the NRM might have resulted in lower (higher) total NRM intensities. One other possible factor is less easily examined and involves the strength of the ancient geomagnetic field; however, if this produced the higher NRM intensity (and if no other factors seem likely explanations), we can expect to find a correlation between paleointensity values and NRM intensities. Thus, in this study, we examine all of the above possible causes in a statistical manner by investigating whether trends are found as a function of age that mimic the NRM intensity patterns.

Analytical Techniques

Scanning electron microscope (SEM) observations have been performed on polished thin sections with a Hitachi S3200N SEM fitted with secondary electron and back-scattered electron (BSE) imaging systems. Ion-milled rock samples were prepared from thin sections using a Gatan Dual Ion Mill and were subsequently examined with analytical electron microscopy (AEM) and transmission electron microscopy (TEM) using a Philips CM-12 TEM fitted with a Kevex Quantum energy dispersive system (EDS). The TEM was operated at 120 kV.

The degree of oxidation (z) of individual titanomaghemite grains was determined with a TEM-AEM technique (Zhou et al., 1999), using higher-order Laue zone (HOLZ) lines in convergent beam mode to determine lattice parameters. If compositions (x values, based on the fraction of ulvöspinel in the titanomagnetite) of the same grains are determined based on thin film theory using AEM, then the oxidation state (z) can be calculated (Zhou et al., 1999). Electron microprobe analyses were used to verify x-values, employing a Cameca SX100 with five wavelength dispersive X-ray spectrometers. The electron microprobe was operated at 15 kV with a beam current of 20 nA.

Rock magnetic properties were measured at the Institute for Rock Magnetism (IRM) at the University of Minnesota. Hysteresis loops of bulk rock samples were measured on a vibrating sample magnetometer (MicroMag VSM), in field values up to 1.5 T. High field thermal curves were measured for selected samples using the VSM, and low-field measurements of susceptibility versus temperature were obtained by using a Kappa Bridge. These two kinds of high-temperature measurements were both performed with the samples contained in a vessel filled with noble gas.

Sample Selection

We have obtained drill core samples of MORB from the Ocean Drilling Program (ODP) and Deep Sea Drilling Project (DSDP) sample distribution sites. A new collection for this particular study consists of 24 basalt samples from 5 drilling sites with ages older than 100 Ma (01Table 1) and 55 samples from 16 sites with ages less than 100 Ma (Fig. 2). The Jurassic–Early Cretaceous MORB samples have been characterized in terms of rock magnetic parameters, as well as composition (including Ti content [x]; see 02Tables 2 and 033) and oxidation degree (z) (02Table 2), because very few results had been available for the latter parameter for these ages until this study. For MORB of Late Cretaceous and Cenozoic age, oxidation state was already sufficiently known from previous results obtained by our group (Zhou et al., 2001), and these have been included in the analysis. The new rock magnetic determinations for MORB < 100 Ma are listed in 04Table 4.

The ages of the basalts have been taken from the Proceedings of the Ocean Drilling Program, with earlier results as summarized by Johnson and Pariso (1993). The one note about age assignments that we must highlight concerns Site 802, which is located in the western Pacific plate (to the east of the Philippine plate), in the center of the East Mariana Basin, where the water depth is 5674 m. The sediment sequence is 500 m thick and ∼50 m of tholeiitic basalt pillow lava were drilled below Cretaceous sediments. This is consistent with the 115 Ma age obtained from Ar-Ar dating (Pringle, 1992). The magnetic anomaly pattern, however, indicates a somewhat different age (Lancelot et al., 1990). We have used the isotopic age of the rocks at this drilling site for our analysis, noting that the alternative would not influence our conclusions.

Parameters for a total of 275 samples have been compiled including the new observations of this study (94 samples), results of Zhou et al. (2001; 88 samples), and results from Matzka et al. (2003; 93 samples). Of these, 15 samples from our collection and 5 samples from Matzka's have been excluded from the analyses because they have very high Curie temperatures and very low x-values, which indicate that they underwent high-temperature oxidation-exsolution; SEM observations of ilmenite and low-Ti magnetite in some samples confirm these observations (see also Matzka et al., 2003).

There are 54 samples with ages less than 5 Ma (including 10 from our study, as listed in 04Table 4). Their NRM intensities have been included in the plots of this study in order to allow us to present a complete overview of intensity variations; however, because we are interested mainly in comparing the properties of early-mid Cenozoic and Mesozoic MORB samples, we have excluded these Plio-Pleistocene MORB samples from our subsequent analyses, which, consequently, are based on a total of 201 samples. 01Tables 1–4 list locations, ages, compositional and rock-magnetic parameters for the new samples reported on in this study.

Testing the Possible Causes of the Intensity Variations

The NRM intensities of the samples used in our analysis are shown in Figure 1C, together with the NRM-intensity patterns published previously. Clearly, the Cretaceous samples reveal higher intensities than those with ages between 10 and 40 Ma. In this and all other diagrams, we portray the observations of the Mesozoic samples as filled (blue) symbols, and those of the Cenozoic as open (green) symbols.

As a first approach to correlate properties with NRM intensities, we plot compositional and rock-magnetic parameters versus JNRM (Fig. 3), but this does not reveal any visual identification of trends. Regression lines (not shown) have minor slopes only and, given the very low correlation coefficients (R2 less than 0.17), they are without any correlation trends. In the following subsections, we further address systematically the possible causes for the Mesozoic-Cenozoic intensity contrasts, and describe the tests used for our analyses, which generally are based on plots of parameters versus age instead of JNRM.

Composition and Oxidation State

Figures 3A and 4A show the variations in composition (x) and Figures 3B and 4B illustrate z as a function of JNRM and age. The z-values have been obtained by convergent-beam electron diffraction yielding HOLZ-line patterns, as described by Zhou et al. (1999, 2000, and 2001). The x-values were obtained by AEM (02Table 2), or electron microprobe (03Table 3), and represent the fraction of ulvö spinel in the titanomagnetite solid-solution series, such that x = 3Ti/(Ti+Fe). Figure 5 shows that the Curie temperatures of all these samples follow the well-established correlation trend with z (Nishitani and Kono, 1983), validating our use of the HOLZ-line technique.

Figure 4B shows that there is no systematic or significant variation in z with age, whereas Figure 4A shows a very minor trend, such that x is slightly higher for the interval 30–95 Ma than before or after. We note, though, that x at 95 Ma is as high as x at 40 Ma, whereas the intensity of the NRM is rather high for the latter and low for the former (Fig. 1C). Figures 3A and 3B show that there is no obvious correlation with NRM intensity. We conclude that it is unlikely that the higher NRM intensity values of Cretaceous MORB represent a lesser (or, more generally, different) oxidation state. The chemical composition (x) of the Tertiary and Cretaceous MORB samples shows perhaps a minor decrease with increasing age, but this trend is hardly significant and is inadequate to explain a doubling of NRM intensity.

Variable Abundance of Magnetic Minerals

To test whether increased NRM-intensity is related to increased abundance of potentially magnetic minerals, we examine measures that reflect the capacity of MORB samples to carry a remanence (Mrs) or the magnitude of the saturation magnetization (Ms). Johnson and Pariso (1993) concluded, based on the data available to them, that elevated magnetization intensities of older MORB samples were due primarily to a greater abundance of magnetic oxides in Cretaceous and older oceanic crust.

Figures 3C, 3D, and 6 show that there is no significant overall variation in Ms or Mrs with either JNRM or age. Matzka et al. (2003) noted that the interval of 10–40 Ma was characterized by lower Ms for titanomaghemite samples, but when we add the data from Zhou et al. (2001), as was done in Figure 6A, this is no longer noticeable. Higher than average values of Ms are seen at ca. 65–70 Ma, but this interval does not at all correspond to the higher NRM intensities of Figure 1 nor to the interval for which Johnson and Pariso (1993) determined their correlation.

Figure 7A and 7B presents JNRM divided by Ms and Mrs. If abundance of remanence carriers were to be the explanation for the Cretaceous-Cenozoic intensity contrasts, we would expect that normalization of NRM intensity by either saturation remanence or saturation-induced magnetization would eliminate the contrast. Figure 7 negates this idea.

The anticipated increases in either Ms or Mrs, which should be observed if the higher NRM-intensity in the Cretaceous were to be caused by an increase in volumetric percentage of magnetic minerals, simply do not seem to be there. Qualitatively this is confirmed by inspection of many SEM images (not shown), which do not show greater abundance of Ti-Fe-oxides in Cretaceous MORB.

No evidence was seen by SEM for secondary magnetite produced by any form of alteration of the silicate minerals. The observed clay minerals are of the trioctahedral smectite type, with the species name saponite, which does not transform to illite. Rather, it transforms to chlorite. The Fe in saponite is limited in amount, and easily substitutes into the neoformed chlorite. We also note that Figure 6 does not provide any evidence for the formation of secondary magnetite and that the Mrs/Ms ratio, to be discussed later, does not show this either.

Vectorial Composition of the Remanence

Thermal and alternating field demagnetization of the NRM of MORB samples can reveal a multivectorial nature of the remanence; even though the samples are unoriented, an antiparallel overprint superposed on a higher unblocking or coercivity component would be readily recognizable. It is anticipated that multivectorial remanence will be revealed in thermal demagnetization, because the blocking temperatures of the original titanomagnetite and its alteration product titanomaghemite are very different. The example of Figure 8B is very representative; only a couple of samples had overprints that represented a fraction of the total NRM that was higher than in this 112 Ma sample, whereas many samples showed small or insignificant overprints. The example of Figure 8A suggests that a small overprint may be directionally masked by complete overlap of components in the 20–250 °C interval, but there is no proof of this suggestion because thermal demagnetization up to more than 500 °C cannot be repeated on a given sample because of alteration at higher temperatures. Details of the demagnetization analysis, including many directional plots, are presented elsewhere (Wang, 2005) and are available upon request.

If multivectorial NRM is present and if the directions of the magnetic components are such that their (lower or higher) resultant vector could be expected to correlate with (lower or higher) NRM intensities, then two factors will be of importance. One consists of the angle 𝛉 between the different vectors, which can range from 0° (entirely parallel) to 180° (perfectly antiparallel). The other is a correction to the original NRM intensity by summing arithmetically the magnitudes of the individual magnetic vectors. The angle 𝛉 should correlate with NRM intensities if differing multivectorial behavior is an explanation for the Cenozoic-Mesozoic contrast. In other words, if two components co-exist while being parallel to each other, their resultant vector sum must be relatively high, whereas two antipodal components would have a much lower total NRM intensity. To test whether multivectorial magnetizations could be held responsible for the NRM intensity patterns of Figure 1, we constructed Figure 9, whereas Figure 7C illustrates the corrected NRM intensity versus age.

The plots of 𝛉 versus age and 𝛉 versus NRM intensity (Fig. 9) do not show any significant signs of correlation. High, low, and intermediate values of 𝛉 are found at the oldest and youngest ages, although one could argue that there is a dearth of intermediate values for the Cretaceous and the Neogene. Similarly, high, low, and intermediate values of weight-normalized NRM intensity are found for low and high 𝛉 values. We conclude that Figure 9 does not present any evidence for a link between multivectorial NRMs and NRM intensity.

Figure 7C presents JNRM corrected for overprint components; in this process, the corrected intensity is calculated as the summed lengths of all components, after aligning them with each other. If overprints were to be the correct explanation for enhanced or diminished NRM intensity, we would expect the plot of the corrected NRM-intensity values to no longer show any variation with age. This, however, is not what is observed. The values for the mid–late Cenozoic are still much lower than those for the Cretaceous. We conclude that variable overprints are not an appropriate explanation for the Cenozoic-Mesozoic JNRM patterns.

Magnetic Grain Sizes or Predominant Domain Sizes

Smaller grains and a predominance of single-domain grains will generally carry a higher-stability component of magnetization, which shows up in demagnetization diagrams like those of Figure 8 as components removed with higher unblocking temperatures or higher coercivity. The assumption we are making in testing whether the higher Cretaceous NRM-intensities are attributable to grain size effects is that the smaller (single-domain type) grains, carrying a more stable remanence, will lead to a higher NRM intensity when they are more abundant. When the magnetic remanence carriers are more multi-domain in nature, then the remanence presumably is likely to be lower in intensity (see also Gee and Kent, 1999).

To test this, we plot a proxy for grain/domain size (Mrs/Ms) as a function of JNRM in Figure 3E and as a function of age in Figure 10. There is only one significant trend and that is that older MORB samples have lower Mrs/Ms ratios, indicative of a larger (more pseudo-to multi-) domain size, and opposite to what would be needed to explain the higher NRM intensities.

We add a brief note about the high Mrs/Ms values (≫ 0.5) for some Cenozoic MORB, which have been discussed before in terms of somewhat erroneous Ms values (Matzka et al., 2003; Fabian, 2005), because the applied field may have failed to saturate the samples. The very high Mrs/Ms values are indeed convincingly associated with rather low Ms values, as can be seen in 04Table 4. We do not think, however, that this effect has any consequences for our study. As can be seen in Figure 10, the high Mrs/Ms values are observed mostly for samples younger than 45 Ma. If there is to be a correction to Ms, it would increase its Cenozoic values and the diagram of Figure 6A would show fewer low Ms values for the Cenozoic. This, however, would not provide an explanation for the relatively higher NRM intensity values in Mesozoic MORB as being caused by higher magnetic mineral content. It is also possible that the association of high Mrs/Ms with low Ms may be caused by spatial variations within pillow fragments (Gee and Kent, 1999). A more in-depth discussion of the trends in the Mrs/Ms ratio versus age (Fig. 10), or JNRM (Fig. 3E), or Ms (04Table 4), is outside the scope of this study.

The Ancient Geomagnetic Field Intensity as the Cause of Higher NRM Intensity

The last of the six possible effects that we proffer as possible causes for the higher Cretaceous NRM is not testable by determining paleointensities from the very same MORB samples of various ages that we describe in this study. Thus, we must resort to analyses of ancient geomagnetic field strength (i.e., paleo-intensity values) determined in previous publications and on more suitable materials, such as submarine basaltic glass and single-crystals of plagioclase (Selkin and Tauxe, 2000; Tarduno et al., 2001; Tauxe and Staudigel, 2004). To aid in our discussion of whether the paleointensity patterns are amenable to interpret the higher remanence of Cretaceous MORB, we plot, in Figure 11, paleointensity values of the ancient geomagnetic field (from Tarduno et al., 2001 and Tauxe and Staudigel, 2004).

Figure 11 (bottom) shows that before 5 Ma the average field strength was generally lower than today, with the exception of some data points in the Cretaceous. Note that the data indicate higher as well as lower paleointensity values, but the upper envelopes of the data sets clearly demonstrate patterns that resemble those of Figure 1. Thus, the aggregate of available and reliable paleointensity measurements suggests that the higher NRM intensities in the Cretaceous may well be attributable to higher geomagnetic field intensities for that period.


Three decades of measurements of the NRMs of MORB samples have repeatedly demonstrated that the NRM intensities show a sharp decrease back in time, which culminates in a low at ca. 10–30 Ma, whereas before that time the intensities are seen to rise gradually to values that approach those of recent basalts near the mid-ocean ridges (Bleil and Petersen, 1983; Wittpenn et al., 1989; Johnson and Pariso, 1993). In turn, these variations in NRM intensities are often held responsible for variations in the amplitudes of marine magnetic anomalies (e.g., Johnson and Pariso, 1993). Examination of several compositional, petrological, geo-, rock- and paleomagnetic parameters yields only one acceptable correlation with the well-known NRM-intensity patterns for the last 160 Ma and that is for the ancient strength of the geomagnetic field. All other factors examined, such as chemical composition of the magnetic minerals, abundance of these magnetization carriers, vectorial superposition of parallel or antiparallel components of magnetization, magnetic grain or domain size patterns, or low-temperature oxidation to titanomaghemite, fail to show convincing correlations with NRM-intensity patterns.

If we are correct in concluding that the main cause for relatively high Cretaceous NRM intensities is a relatively high geomagnetic field strength at that time, that would provide support for the idea that the long Cretaceous Normal Superchron was a period not only of stability in terms of (few or no) reversals, but also one of—on average—higher paleointensity. The plots of Figure 11 do, of course, suggest this, but the fact remains that low mid-Cretaceous paleointensities have also been observed. We note that low NRM intensity values are present as well as relatively higher ones. It could be that the field was characterized by rather variable short-time variations in strength, but the more likely explanation is that variability in parameters correlates with location within a single cooling unit (Marshall and Cox, 1972; Gee and Kent, 1999).

Corresponding author: +1-734-764-8322, +1-734-763-4690 (fax)

We thank Jeff C. Alt and Eric J. Essene for insightful discussions and the Ocean Drilling Program (ODP) for supplying samples. Ulrich Bleil provided extremely valuable comments on an early draft of the manuscript, and the journal's reviewers, Dennis Kent and Sue Beske-Diehl, made many helpful suggestions for improving the manuscript. This research was supported by a grant from the Division of Earth Sciences of the National Science Foundation (EAR 0207257) and a visiting fellowship from the Institute for Rock Magnetism (IRM). The IRM is funded by grants from the Keck Foundation, the National Science Foundation, and the University of Minnesota.