The Hawaiian-Emperor volcanic chain is an iconic feature that appears in most introductory Earth science textbooks as the case example of age-progressive volcanism recording plate motion over a sublithospheric hotspot. The westward increase in island age, evident from volcano morphology, was recognized by the first native peoples and remarked upon by the first visiting naturalists (Darwin, 1842; Dana, 1849). With the successful application of the K-Ar method to some of the earliest age determinations on basaltic rocks (McDougall, 1964), a monotonic westward aging of the Hawaiian volcanoes was confirmed.

Wilson (1963) first noted the utility of the orientation of the Hawaiian-Emperor lineament and the rate of migration of volcanic activity along it as a reference frame for Pacific plate motion over a hotspot embedded in the mantle. Following the proposition that the Hawaiian hotspot is stationary, which is seemingly supported by the congruent geometries of several Pacific hotspot-related lineaments, Morgan (1971) proposed that Pacific plate motion could be quantified by a set of stage rotations fitting the hotspot volcanic trails, with angular velocities determined from radiometric ages of rocks sampled from the volcanoes. The great longevity of some of the volcanic trails (>100 m.y.) was ascribed to whole-mantle, upwardly convecting, narrow plumes that maintain a constant delivery of warmer-than-ambient mantle to the hotspots (Morgan, 1971).

Shortly thereafter, however, the fixed hotspot hypothesis was challenged. Given that plate tectonics is the expression of a convecting mantle, it is reasonable that any persistent feature embedded in the mantle, like a plume-fed hotspot, should also be moving with respect to other parts of the mantle and the spin axis. Analyses of global plate circuits suggested large relative motions between Hawaii and hotspots in the Atlantic and Indian Oceans (Molnar and Atwater, 1973). Subsequent studies have supported the results of these early plate circuit tests (Molnar and Stock, 1987; Cande et al., 1995; Norton, 1995; DiVenere and Kent, 1999; Raymond et al., 2000; Steinberger et al., 2004). Uncertainties in the plate circuits (e.g., potential motion between East and West Antarctica and tectonic complexities associated with the final transfer to the Pacific plate) have limited a general acceptance of these conclusions.

The important question, from the point of view of a useful frame of reference for plate motions, is the magnitude and constancy of this hotspot motion. An alternative approach to examine hotspot fixity is to determine the age and paleolatitude of volcanoes that form a given hotspot track. For the Hawaiian hotspot, the paleolatitudes of extinct volcanic edifices of the Emperor chain should match the present-day latitude of Hawaii (~19N) if the hotspot has remained fixed with respect to Earth's spin axis.

Previous paleolatitude estimates for Pacific Basin sites have been based on remote sensing data (modeling of marine magnetic anomalies of seamounts observed during magnetic surveys and of the skewness of marine magnetic anomalies). These analyses provided early important clues to Pacific plate motion. Whereas remote sensing data are still used in a few studies (e.g., Harada and Hamano, 2000), it is generally accepted that such data lack the resolution to address modern questions on hotspot motion (DiVenere and Kent, 1999). Instead, the most reliable indicators of paleolatitude are basaltic rocks, but enough time must be spanned by any section such that geomagnetic secular variation is sampled.

Recovery of such samples requires ocean drilling technology, specifically that available in the Ocean Drilling Program (ODP) and its predecessor program (the Deep Sea Drilling Project [DSDP]).

Paleomagnetic data provide estimates of motion relative to Earth's spin axis. A variable that must be considered in the interpretation of paleomagnetic data is the potential shift of Earth with respect to the spin axis, a process known as polar wander (Goldreich and Toomre, 1969). Addressing this variable is in fact directly related to the issue of hotspot motion: modern estimates of polar motion have been based on analyses of paleomagnetic data viewed in a fixed hotspot reference frame. The motion derived from this calculation is often called "true polar wander" (TPW) (e.g., Besse and Courtillot, 2002). Prior attempts to define TPW have largely been based on paleomagnetic data from the Atlantic-bordering continents. The availability of new paleomagnetic data sets from oceanic drilling of seamounts in the Pacific Ocean basin during the early 1990s provided an opportunity to test such models. The new paleomagnetic data conflicted with existing TPW models and instead indicated relative motion between Pacific hotspots and those in the Indo-Atlantic realm (Tarduno and Gee, 1995). Subsequent analyses of the continental record indicated only minor post-Cretaceous polar wander (Tarduno and Smirnov, 2001, 2002).

The prior work on plate circuits, the failure of TPW to explain Pacific paleomagnetic data, and the inferences on hotspot motion stemming from these analyses set the stage for additional work. Specifically, the desire to obtain a new and more extensive Pacific data to test hotspot motion motivated ODP Proposal 523 (J.A. Tarduno, R.D. Cottrell, and B. Steinberger, Motion of the Hawaiian Hotspot during Formation of the Emperor Seamounts: A Paleomagnetic Test [earth.rochester.edu/pmag/odp-proposal523.html]). The study proposed eventually became ODP Leg 197.

Below, we review the paleomagnetic results from Leg 197, as well as the geophysical survey profiles that were essential for site selection; these data, together with the geochronological results (Duncan and Keller, 2004), form the basis for the hotspot motion test. We follow this description with a brief summary of other work related to hotspots that has followed Leg 197. In addition, data bearing on the long-term history of the geodynamo resulting from analyses of Leg 197 cores are also reviewed. These data, together with the results of the hotspot motion test, provide new insight into the nature of large-scale mantle convection. A companion program of Leg 197 drilling was to describe the volcanic development and geochemical variability within the Emperor Seamounts, compared with the Hawaiian Islands, in order to assess the effect of changing lithospheric thickness and the contributions of mantle plume and oceanic lithosphere to melting throughout the history of the Hawaiian hotspot. This program is also briefly summarized.

The reader is referred to additional reports of biostratigraphic studies (Bordine et al., Siesser, and Tremolada and Siesser, all this volume) and logging (Gaillot et al., this volume), which we do not review.

Paleomagnetic Results from Leg 197

Prior paleomagnetic analyses of 81-m.y.-old basalt recovered from the Emperor Seamounts (Detroit Seamount, ODP Site 884) yielded a paleolatitude of ~36N (Tarduno and Cottrell, 1997), clearly discordant with the latitude of Hawaii. Only one other data set that averaged secular variation was available prior to Leg 197. Data from ~61-m.y.-old basalt (Sharp and Clague, 2002) from Suiko Seamount define a paleolatitude of ~27N (Kono, 1980). Together, these data sets suggest that the Emperor Seamounts record southward motion of the hotspot plume in the mantle (Tarduno and Cottrell, 1997).

Leg 197 sought to test the hypothesis of southward motion of the Hawaiian hotspot by drilling additional basement sites in the Emperor chain (Fig. F1). Detailed stepwise alternating-field (AF) demagnetization data were collected aboard the drillship JOIDES Resolution (Tarduno, Duncan, Scholl, et al., 2002). All analyses were conducted on standard paleomagnetic minicores, and directions were fit using principal component analysis. Although these shipboard data are of high resolution, they alone are insufficient to define paleolatitudes. Magnetic minerals with intermediate to high coercivities, carrying magnetizations resistant to AF demagnetization, are commonly formed during subaerial or seafloor weathering. The presence of these minerals, which include titanomaghemite, hematite, and a range of oxyhydroxides, had been inferred on the basis of rock magnetic measurements, and they have been directly observed in some of the recovered lava flow samples using reflected-light microscopy. The magnetizations of these mineral phases are easily resolvable with thermal demagnetization (Doubrovine and Tarduno, 2004a, 2004b). Accordingly, detailed (25C steps; 50–625C range) thermal demagnetization data, measured on a high-resolution 2G Enterprises SQUID (super-conducting interference device) magnetometer, were collected as part of Leg 197 postcruise work.

At least several millennia of geomagnetic field behavior must be sampled such that the axial dipole term becomes dominant, allowing a statistically significant estimate of the paleolatitude. The angular dispersion of inclination averages (McFadden and Reid, 1982) from independent lava flows (inclination groups) were compared with global lava data of the same age (McFadden et al., 1991) to examine whether secular variation had been adequately sampled (Cox, 1970; Brock, 1971). Analyses of the Leg 197 data utilized observations of the physical aspects of the lava flows, as well as petrologic and geochemical data, to group cooling units into lava units. This approach has resulted in a more conservative assignment of inclinations groups than used in prior studies.

Marine sediments can also provide useful paleolatitude information, but the results must be interpreted with care; in general, they provide minimum values of paleolatitude because of potential compaction-induced inclination flattening (e.g., Anson and Kodama, 1987; Arason and Levi, 1990; Tarduno, 1990). An advantage magnetizations from sediments have with respect to those from basalt, however, is that they can record significant time intervals. Similarly, chemical remanent magnetization (CRM), carried by minerals formed during weathering, can preserve stable magnetizations that provide insight into the time-averaged field.

Below, we briefly summarize the postcruise thermal demagnetization data collected from each of the Leg 197 sites that were reported in a summary article (Tarduno et al., 2003). We also review more recent rock magnetic analyses.

Koko Seamount (Site 1206)

Site 1206 (Fig. F1) was positioned on the southeastern side of the lower summit terrace on Koko Seamount using crossing underway seismic profiles. The central area of Koko Seamount is capped by a relatively thick (~200 m) sequence of sedimentary deposits overlying a plateau of igneous basement. To avoid penetrating the sedimentary cover, which probably includes reefal debris, Site 1206 was positioned over an acoustically imaged sequence of flat-lying lava flows and volcaniclastic beds underlying the thinly sedimented southeast flank of the seamount's summit platform (Kerr et al., this volume). Thin intercalations of limestone, volcaniclastic sandstone, and a deeply weathered flow top were also recovered, providing geological evidence of time between lava flow units.

Plateaus in 40Ar-39Ar incremental heating spectra from six whole-rock samples yield a mean age of 49.15 0.21 Ma (2 uncertainty quoted; Duncan and Keller, 2004).

Seventeen inclination groups, spanning three polarity zones (Chrons 21n-21r-22n) (Cande and Kent, 1995) were identified in the thermal demagnetization data having a mean inclination (It = 38.3 +6.9/–9.3; hereafter all uncertainty regions are 95% confidence interval unless otherwise noted), nearly identical to that isolated by AF treatment (Iaf = 38.5 +8.4/–10.9, based on 14 inclination groups sampled) (Fig. F2). The angular dispersion of the thermal data (Sf = 15.3 +4.3/–2.7) is within error of that predicted by global 45- to 80-Ma lava flows (McFadden et al., 1991). Comparison of the inclination units based on thermal demagnetization with a synthetic Fisher (Fisher, 1953) distribution (Fig. F2) suggests that the basalt sequence represents well the time-averaged geomagnetic field. Furthermore, a highly stable magnetization carried by hematite (likely a CRM with unblocking temperatures >580C) in samples of the deeply weathered basalt yields a mean inclination that is indistinguishable from that of the lava flows.

Nintoku Seamount (Site 1205)

At Site 1205 on Nintoku Seamount (Fig. F1), a sequence of 25 subaerially erupted a'a and pahoehoe lavas and interbedded sediment and soil horizons were recovered in 283 m of basement penetration. Two different types of igneous basement produce distinctly different seismic signatures in the vicinity of Site 1205 (Kerr et al., this volume). The upper unit is imaged as laterally coherent, low frequency, and west-dipping reflections that coincide with a bowl-shaped sequence of dominantly alkalic basalt lava flow units interbedded with soil layers. The lower acoustic unit at the base of the drilled section is more acoustically massive and rises gently upward to the southeast. The strongly reflective sequence of lava flows and soil interbeds of the upper acoustic unit thins in this direction. The more massive acoustic character of the lower unit is linked to the less common occurrence of soil horizons below ~203 meters below seafloor (mbsf). Plateaus in 40Ar-39Ar incremental heating spectra from six whole-rock basalt samples spanning the section give a mean age of 55.59 0.25 Ma (Duncan and Keller, 2004). The thermal demagnetization data (22 inclination groups) yielded a mean (It = –44.3 +10.3/–6.3) that is similar to that of the AF data (Iaf = –45.7 +10.5/–6.3). The estimated angular dispersion of the thermal data (Sf = 19.9 +4.8/–3.2) is higher than predicted by models based on average data from 45- to 80-Ma lavas (McFadden et al., 1991) and may point to higher-frequency changes in secular variation with time. Nevertheless, the geologic evidence for time, together with a comparison of the data versus a Fisherian distribution (Fig. F2), suggests the mean value represents the time-averaged field.

Detroit Seamount (Site 1204)

Two holes were drilled at Site 1204 on Detroit Seamount (Holes 1204A and 1204B), penetrating 60 and 138.5 m of basalt, respectively. Detroit Seamount, one of the northernmost volcanic edifices of the Emperor Seamount chain, is a broad volcanic plateau constructed of little deformed but locally faulted basement of lava and volcaniclastic beds. Flows and volcaniclastic units are distinguishable on the seismic records. Basement is overlain by a capping of drift sediment, the Meiji Drift, as thick as 800–900 m. Detailed descriptions of the seismic data used to position Leg 197 drilling sites in nonstructural settings are provided in Kerr et al. (2005). Diamictite and volcanic ash–rich sediment directly overlying basement contain nannofossils of Campanian age (CC22–CC23, ~73–76 Ma) (Berggren et al., 1995); nannofossils of this zone were also recovered in a sediment interbed in the last core from Hole 1204B, suggesting the entire basement section drilled is of this age.

Reflected-light microscopy showed a dominance of titanomaghemite in basalt samples. Approximately 30% of the samples examined failed to show a simple AF demagnetization behavior; instead, the presence of at least one unresolved magnetic component was suggested. Thermal demagnetization revealed a clear two-component structure in many samples. A reversed polarity component was defined at low unblocking temperatures (~100 to 275–350C), whereas a normal polarity component was defined at higher temperatures. Because of this unusual behavior, a detailed investigation of the Site 1204 basalts was conducted (Doubrovine and Tarduno, 2004a).

Extensive rock magnetic tests coupled with X-ray diffraction (XRD) and scanning electron microscopy (SEM) characterization established that the lower unblocking temperature component in the Site 1204 basalts (as well as those from ODP Site 883) represents a partial self-reversal of the primary magnetization. Self-reversal of natural remanent magnetization, the remarkable ability of some rocks to acquire a magnetization antiparallel to that of Earth's geomagnetic field, has been established in only a few, arguably rare, igneous rocks bearing hemoilmenite. However, hemoilmenite is absent from the Site 1204 rocks. Instead, the self-reversed component is carried by titanomaghemite, which formed by in situ low-temperature oxidation; the self-reversal process is consistent with ionic reordering as described by Verhoogen (1956, 1962). Subsequent measurements of basalt from other sites seem to suggest that the "self-reversing" compositional field is even more restrictive than that envisioned in the classic work of Verhoogen (1956, 1962) and O'Reilly and Banerjee (1966). When present, self-reversal in titanomaghemite connotes extreme oxidation at low temperature (needed to maintain the cation-deficient structure of titanomaghemite) (Doubrovine and Tarduno, 2005). This could be achieved through sustained fluid flow and Fe removal; therefore, self-reversals may provide one way of identifying such processes in oceanic crust.

Because the self-reversed component is not typically isolated by AF demagnetization, we exclude from consideration paleomagnetic results based on AF results from Detroit Seamount, including a study in which the two-component magnetic structure was missed (e.g., Sager, 2002). Instead, we restrict our analysis to the high unblocking temperature magnetization. Although carried by titanomaghemite, it is nevertheless this magnetization that best reflects the primary field direction (see additional discussion in Doubrovine and Tarduno, 2004b).

Only a single inclination unit was identified in Hole 1204A, with a mean inclination It = 64.5 +10.2/–19.0. On the basis of the lithofacies succession in Hole 1204B, five inclination groups were identified. Thermal demagnetization results indicate a mean inclination of (It =) 60.1 +5.2/–5.5. Angular dispersion estimates are low (Sf = 3.1 +2.0/–0.9) and at face value suggest that the sequence does not adequately sample secular variation. However, thermal demagnetization data from samples of a breccia interbedded with the lavas show a consistent direction, similar to that of the basalt flows. This observation, together with the presence of titanomaghemite, suggests that CRMs acquired over time might be important at Site 1204.

Detroit Seamount (Site 1203)

Eighteen compound lava flow units and 14 volcaniclastic sedimentary interbeds were drilled in 453 m of basement penetration at Site 1203 on Detroit Seamount. Plateaus in 40Ar-39Ar incremental heating spectra from three whole-rock basalt samples and two feldspar separates yield a mean age of 75.82 0.62 Ma (Duncan and Keller, 2004). Thermal demagnetization of the two uppermost pillow basalt flows of the sequence revealed a two-component structure similar to that observed at Site 1204. In all other basalt samples, a simple univectorial decay was observed after the removal of minor overprints. The thermal demagnetization results (16 inclination groups) yield a mean (It = 48.6 +7.0/–10.6) similar to that calculated from AF data (Iaf = 50.0 +7.3/–10.6, for 14 inclination units). The estimated angular dispersion of the thermal demagnetization data (Sf =18.4 +6.9/–3.7) is slightly higher than that expected from global lava flow data (McFadden et al., 1991).

Paleolatitude History

The inclination groups, averaged by site, form a time-progressive sequence of decreasing paleolatitude with time (Fig. F3). This is inconsistent with the fixed-hotspot hypothesis and instead indicates the plume was moving with respect to the spin axis. Whereas many scientists believe that long-standing nondipole terms are insignificant in the geomagnetic field during the last 200 m.y. (Courtillot and Besse, 2004), it should be noted that if the one model calling for minor octupole fields is correct (Van der Voo and Torsvik, 2001), our paleolatitude data would underestimate the true rate of hotspot migration. As discussed earlier, polar wander, the rotation of the entire solid Earth with respect to the spin axis (Goldreich and Toomre, 1969), is an alternative physical mechanism that might explain paleolatitude trends. However, polar wander can be tested by examining globally distributed data, and these tests have failed to confirm it (Tarduno and Gee, 1995; Cottrell and Tarduno, 2000b; Tarduno and Smirnov, 2001, 2002; Torsvik et al., 2002; Besse and Courtillot, 2002).

Paleolatitude estimates for Detroit Seamount are available from lava flows and sediments, based on thermal demagnetization. Lava emplacement was probably less frequent at Site 884 relative to the other sites because it is on the flank of Detroit Seamount. Hole 1204B lavas clearly reflect low angular dispersion, as do the basalts at Site 883 (Doubrovine and Tarduno, 2004b). However, these rocks might carry a CRM acquired over a time interval longer than that of a cooling lava flow, explaining the agreement of their mean inclination with that of the Site 884 basalt section and the Site 1203 sediments. In fact, the mean inclination from the Site 1203 sediments should be a minimum because of potential inclination shallowing. These comparisons further suggest that the mean inclination derived from the basalts at the same site is shallower because the available lavas underrepresent higher inclination values.

Tarduno et al. (2003) presented two averaging scenarios for the Detroit Seamount paleomagnetic data. In one, the paleomagnetic results from the Site 884 basalts, Hole 1204B basalts, and Site 1203 sediments best represent the field estimate (paleolatitude Model A). In another, all the individual basalt inclinations were combined to form a grand mean. Both averaging models indicate high average rates of motion (Model A = 57.7 19.2 mm/yr; Model B = 43.1 22.6 mm/yr) consistent with the hypothesis that the Hawaiian hotspot moved rapidly southward from 81 to 47 Ma. The higher rate of southward motion is in slightly better agreement with the estimates of hotspot motion based on independent relative plate motions (e.g., Raymond et al., 2000), and Doubrovine and Tarduno (2004b) further discuss reasons why the higher-latitude mean value from Detroit Seamount may be accurate. Nevertheless, it is not possible to distinguish between these models with the presently available data. The slow latitudinal plate motion at this time is consistent with plate tectonic driving forces (Cottrell and Tarduno, 2003). The paleomagnetic data are also consistent with geodynamic models of the interaction of a plume with large-scale mantle flow (Steinberger and O'Connell, 1998; Tarduno et al., 2003; Steinberger et al., 2004) (Fig. F4).

Developments after Leg 197

Following the development of ODP Proposal 523 and publication of the Leg 197 Initial Results, several important studies concerning hotspot motion appeared in the literature. Antretter et al. (2002) studied the paleomagnetism of basalt cores recovered from Kerguelen Plateau (ODP Leg 183). The authors concluded that the data supported motion of the Kerguelen hotspot, consistent with geodynamic models of mantle flow. Paleomagnetic analyses of lavas drilled on Ontong Java Plateau during ODP Leg 192 defined a difference between observed paleolatitudes and those predicted by fixed hotspots (Riisager et al., 2003a). These data are consistent with hotspot motion as defined in studies of other Cretaceous Pacific sites (Tarduno and Gee, 1995) and the Emperor Seamounts (Tarduno and Cottrell, 1997). Moore et al. (2004) and Pares and Moore (2005) present detailed analyses of equatorial crossing data based on Pacific DSDP and ODP sedimentary cores. They conclude that the Hawaiian hotspot moved southward since 53 Ma, consistent with Leg 197 results.

Torsvik et al. (2002) review the debate over rapid Cretaceous TPW (Prevot et al., 2000; Camps et al., 2002) vs. hotspot motion (Tarduno and Smirnov, 2001, 2002) and find little evidence for TPW. Courtillot et al. (2003) use Leg 197 results in a comparison of Pacific and Indo-Atlantic paleomagnetic data relative to hotspots in the respective hemispheres. They conclude that there were significant differences between the two data sets, which by definition cannot be explained by TPW (i.e., rotation of the entire Earth with respect to the spin axis). This result supersedes earlier conclusions reached by Besse and Courtillot (2002) and confirms an earlier finding of relative motion between groups of Pacific and Atlantic hotspots (Tarduno and Gee, 1995). An alternative plate circuit has been recently suggested that invokes a path through the Lord Howe Rise (Steinberger et al., 2004). This plate path also predicts hotspot motion consistent with that defined by the Leg 197 paleomagnetic data.

Koppers and Staudigel (2005) examine the ages of purported hotspot tracks of the central Pacific. They conclude that bends in these tracks, typically taken to be coeval with the Hawaiian-Emperor bend, are in fact of differing age. They offer several explanations, including motion between hotspots within the Pacific Basin and strong lithospheric control on the expression of some hotspot magmatism. For some apparent hotspot tracks this can be taken as support for nonplume models that relate volcanism to extension. The results of Koppers and Staudigel (2005) highlight that attempts at backtracking based on geometry alone, including "hotspotting" (Wessel and Kroenke, 1997, 1998; Wessel and Lyons, 1997; Kroenke et al., 2004), cannot be used to test hotspot motion; instead they must be combined with accurate and complete age information.

The nature of mantle plumes has recently been the subject of numerous debates at meetings and in the literature (e.g., Foulger and Natland, 2003; Sleep, 2003). Some workers see different categories of plumes with different sources within the mantle (e.g., Courtillot et al., 2003). Some recent tomographic investigations have also been interpreted as representing a variety of mantle plumes (Montelli et al., 2004), but the interpretations have sometimes differed from other classifications. Others hold to an interpretation that sees no role for lower mantle plumes (Anderson, 2000, 2004). Whereas the resolution of the tomographic images is itself the subject of lively debate at meetings, it seems likely that the approach will eventually yield conclusive results. Regardless of the possibility of different hotspot types, at the time of the writing of this report no viable alternative exists to the mantle plume model that explains all the data available from the Hawaiian-Emperor chain.


Another goal of Leg 197 was to better define characteristics of the Late Cretaceous–Paleogene geomagnetic field. A particular focus was on past field strength, or paleointensity. Several postcruise efforts have been devoted toward that end, using various natural materials recovered during Leg 197.

Whole-Rock Basalt Studies

A classic approach to paleointensity study is to analyze whole-rock basalt samples. A sample's natural remanent magnetization (NRM) is demagnetized over a given temperature range in a zero-field environment. Next, the sample is reheated to the same temperature in the presence of a known laboratory field. As a result, a thermoremanent magnetization (TRM) is imparted to the sample. Given the NRM lost, TRM gained, and applied field, the paleointensity can be calculated. This approach is attributed to Thellier and Thellier (1959), as modified by Coe (1967), and it is arguably the most rigorous experimental method for the collection of paleointensity data.

The rock magnetic requirements for Thellier paleointensity analysis, however, are severe. The magnetic carriers must be extremely fine grained (having single-domain behavior), they must carry an original TRM, and they must not alter during the Thellier experiments. Carvallo et al. (2004a, 2004b) applied rock magnetic tests on many Leg 197 lavas, seeking samples that might meet these criteria. Ultimately, most of the samples that showed indications of suitable fine magnetic grain sizes (and associated domain states) also showed that the magnetic minerals had undergone some low-temperature oxidation (i.e., much less severe than that displayed by some Site 1204 basalts but still detectable). Under normal seafloor conditions, original titanomagnetite can be partially converted to titanomaghemite (therefore the process is also often called "maghemitization"). Evidence of maghemitization was observed in reflected-light microscopy studies of the Leg 197 lavas (analyses by Clive Neal, reported in Tarduno, Duncan, Scholl, et al., 2002). Maghemitization transforms a TRM into a chemical remanent (or crystallization remanent) magnetization. The accuracy of CRM in recording past field strength is unclear, but it is generally accepted that the process is less efficient than TRM (Dunlop and Ozdemir, 1997; Smirnov and Tarduno, 2005). Submarine basaltic rocks affected by low-temperature magnetization are thought to yield anomalously low paleointensity data (e.g., Gromme et al., 1979).

Carvallo et al. (2004a, 2004b) identified only one lava unit that had apparently escaped the more dramatic effects of low-temperature oxidation. Thellier experiments of these samples yield field strengths ranging from 34.2 to 39.4 T. These results alone cannot constrain geomagnetic field behavior because they do not span a period sufficient to yield time-averaged field behavior. Nevertheless, as discussed below, the Carvallo et al. (2004a, 2004b) study provides key data for the evaluation of new paleointensity approaches and for understanding the veracity of existing data in the paleointensity database.

Submarine Basaltic Glass

It has been argued that submarine basaltic glass can be a high-resolution recorder of past field strength; the magnetization is thought to be carried by fine-grained magnetic particles within the glass. Because these particles are so Fe rich, questions have been raised about the process of magnetization. Some authors favor a low-temperature CRM process (e.g., Heller et al., 2002; Goguitchaichvili et al., 2004) as opposed to a thermoremanent magnetization; in this case, submarine basaltic glass would not meet the basic requirements for Thellier paleointensity analysis. On the other hand, support for the accuracy comes from studies that show that young submarine glass generally yields field values that agree with known modern field strengths (Pick and Tauxe, 1993).

Time-averaged results are not available from most of the submarine basaltic database (Selkin and Tauxe, 2000). Nevertheless, these data are generally of high technical quality.

Recently, relatively high paleointensity values have been reported from submarine basaltic glass samples of one locality (Troodos ophiolite) formed during the Cretaceous Normal Polarity Superchron (Tauxe and Staudigel, 2004). These have been interpreted as reflecting a higher field strength during the Cretaceous Normal Polarity Superchron, superseding prior interpretations based on submarine basaltic glass that the Superchron field was of normal or weak intensity (Pick and Tauxe, 1993). Notwithstanding these data, the aggregate of individual sample results available from submarine basaltic glass from the 10- to 160-Ma interval suggest a weak mean field strength for the last 160 m.y. that is close to half the present-day field value (Tarduno and Smirnov, 2004).

Submarine basaltic glass was recovered at a few Leg 197 sites, especially Site 1203 on Detroit Seamount. Smirnov and Tarduno (2003) sought to evaluate the accuracy of this glass as a paleointensity recorder through a rock magnetic, transmission electron microscope, and paleointensity study. They observe neocrystallization of magnetite in Thellier heatings of submarine basaltic glass from three other Cretaceous ODP sites. This alteration resulted in artificially low paleointensity values (Fig. F5). Glasses are thermodynamically unstable; with time and/or increased temperature, they become crystalline. The chemical bonds in glass break over a temperature transitional range (Bouska, 1993; Donth, 2001), and this range overlaps with temperatures applied during typical Thellier experiments. Therefore, the tendency of older submarine basaltic glass to record lower and more variable paleointensity values could reflect subtle experimental alteration, enhanced by natural weakening of bonds and hydration with age, lowering the glass transition temperature.

Smirnov and Tarduno (2003) suggest that monitor samples (i.e., splits of the submarine basaltic glass specimens) be used during the Thellier experiments. Specifically, Smirnov and Tarduno (2003) advocate the use of magnetic hysteresis measurements on the monitor samples after each heating. This experimental approach was adopted by Riisager et al. (2003b) in a study of submarine basaltic glass from Ontong Java Plateau (Leg 192). Experimentally induced alteration similar to that objected from Site 1203 was reported, further suggesting that this could be a general characteristic of older submarine basaltic glass.

Plagioclase Feldspar Studies and Long-Term Field Behavior

Single plagioclase crystals can contain minute magnetic inclusions (often 50–350 nm) that are shielded by the silicate matrix and less susceptible to natural and experimental alteration (e.g., Smirnov et al., 2003). Thellier paleointensity analyses of plagioclase crystals separated from a modern lava flow on Kilauea, Hawaii, provide a benchmark for this approach (Cottrell and Tarduno, 1999). In that study, paleointensity values were obtained that matched the field value known from magnetic observatory data.

Many of the lavas from Nintoku Seamount Site 1205 contained plagioclase feldspar potentially suitable for Thellier analyses. To test the potential of these crystals, Tarduno and Cottrell (2005) conducted several rock magnetic tests, including the measurement of first-order reversal curves (FORCs) (Pike et al., 1999; Roberts et al., 2000). In a FORC diagram the horizontal axis (Hc) is a gauge of microcoercivity, whereas the vertical axis (Hb) is a measure of the interaction field between magnetic grains. Neither the whole rocks nor the single crystals from Site 1205 show evidence for large magnetic interactions (Fig. F6). But the plagioclase FORC distributions are clearly more favorable for paleointensity studies. The whole-rock FORC diagrams and individual magnetic hysteresis curves show evidence for very low microcoercivities, consistent with the presence of multidomain grains. Such grains violate Thellier requirements (e.g., Dunlop and Ozdemir, 1997; Dunlop et al., 2005).

NRM/TRM data from 44 of 86 plagioclase crystals measured from 11 lava flow units from Site 1205 that meet experimental reliability criteria (Cottrell and Tarduno, 2000a; Tarduno and Cottrell, 2005) yield a mean field value of 42.8 13.6 T (1 uncertainty) (Fig. F7). Plagioclase crystals large enough for paleointensity analyses are not available from the one aphyric lava for which Carvallo et al. (2004a) obtained Thellier paleointensity data seemingly unaffected by low-temperature oxidation. But the range of paleointensity values obtained from the whole rocks (34.2–39.4 T) is compatible with that available from the plagioclase crystals from the Site 1205 sequence.

Perhaps more importantly, Carvallo et al. (2004a) also performed Thellier experiments on some Site 1205 lavas that had undergone low-temperature oxidation. Interestingly, these yielded nominal whole-rock paleointensity values lower than those obtained using plagioclase crystals from the same units (Tarduno and Cottrell, 2005). This observation is consistent with the hypothesis that low-temperature oxidation and the acquisition of CRM has led to a low field bias in the paleointensity database from whole rocks (Tarduno and Smirnov, 2004).

The plagioclase crystals come from lavas that span the basement sequence. We can be confident that time elapsed between the lavas because some lava tops are deeply weathered, whereas in other cases, soil horizons separate the lavas (Tarduno, Duncan, Scholl, et al., 2002). Paleomagnetic directions from whole rocks and their associated estimate of angular dispersion (Tarduno et al., 2003) further indicate that the flows are independent in time and average secular variation. Thus, the virtual dipole moments obtained from each lava flow together comprise a time-averaged paleomagnetic dipole moment. This moment, 8.9 x 1022 A-m2 (Tarduno and Cottrell, 2005), is significantly weaker (at the 95% confidence level) than those derived from plagioclase crystals from Cretaceous Normal Polarity Superchron lavas (Cottrell and Tarduno, 2000a; Tarduno et al., 2001, 2002). The standard deviation of the Site 1205 virtual dipole moments (32%) is similar to that of the modern field but nearly three times greater than those recorded by the Cretaceous Normal Polarity Superchron plagioclase crystals (Fig. F8).

Available plagioclase-based paleointensity data now sample the range of field behavior displayed by the geodynamo during the last 200 m.y. The pattern suggested by the ensemble of single-crystal paleointensities traces an inverse relationship between field strength and reversal frequency (Fig. F8), as was suggested in the pioneering work of Cox (1968) (see also discussion by Banerjee, 2001). Such a relationship has also been suggested in some subsequent models (e.g., Larson and Olson, 1991) and analyses of marine magnetic anomaly intensities (McElhinny and Larson, 2003).

There are several competing interpretations of the geomagnetic reversal chronology; it is useful to mention these before discussing further the importance of the Site 1205 paleointensity results. In one interpretation, nonstationary intervals bound the Cretaceous Normal Polarity Superchron (McFadden et al., 1991). In another interpretation, the reversal history is best represented by a series of stationary regimes (Lowrie and Kent, 2004). In still another interpretation, the Cretaceous Normal Polarity Superchron is thought to be part of a nonstationary interval spanning 130–25 Ma, followed by a stationary interval. The last interpretation has particular importance because it has been used to further suggest that superchrons reflect nonlinear dynamo processes (Hulot and Gallet, 2003).

In contrast, superchrons may reflect times when the nature of core/mantle boundary heat flux allows the geodynamo to operate at peak efficiency, as suggested in some numerical models (Glatzmaier et al., 1999; Roberts and Glatzmaier, 2000; Olson and Christensen, 2002; Christensen and Olson, 2003), whereas the succeeding period of reversals may signal a less efficient dynamo with a lower and more variable dipole intensity. This interpretation is consistent with the observations from the Site 1205 plagioclase crystals and the overall plagioclase paleointensity data set (Tarduno and Cottrell, 2005). The timescale for the transition between these states is consistent with an active lower mantle, controlling the nature of the geodynamo.