G 3. AN ELECTRONIC JOURNAL OF THE EARTH SCIENCES Published by AGU and the Geochemical Society

Transcription

1 Geosystems G 3 AN ELECTRONIC JOURNAL OF THE EARTH SCIENCES Published by AGU and the Geochemical Society Article Volume 9, Number 2 9 February 2008 Q02005, doi: ISSN: Magnetostatic interactions in deep-sea sediments inferred from first-order reversal curve diagrams: Implications for relative paleointensity normalization Toshitsugu Yamazaki Geological Survey of Japan, AIST, Tsukuba, Ibaraki , Japan (toshi-yamazaki@aist.go.jp) [1] Understanding magnetostatic interactions in sediments is important for paleomagnetism and rock magnetism. Magnetostatic interactions are known to affect significantly the efficiency of anhysteretic remanent magnetization (ARM) acquisition, and hence the ratio of ARM susceptibility (c ARM ) to saturation isothermal remanent magnetization (SIRM), a magnetic grain-size proxy widely used for environmental applications, can be controlled by magnetostatic interactions. Relative paleointensity estimations may also be influenced by magnetostatic interactions because ARM is often used as a normalizer to correct for the difference in magnetizability of sediments. In this study, characterization of magnetic grains in deep-sea sediments from the North Pacific Ocean was conducted using first-order reversal curve (FORC) diagrams and IRM acquisition curves. The FORC diagrams indicate that the magnetic grains consist mainly of a noninteracting single-domain (SD) component and an interacting SD component, and additionally of a multidomain (MD) component. The relative abundances of these components were semiquantitatively estimated by curve fitting of cross sections along a line parallel to the axis of local interaction fields (H u ) and through a peak in the coercivity (H c ). Three components were fitted assuming a Gaussian distribution of interaction fields (H u ). The IRM acquisition curves can be described by two dominant components assuming a log-gaussian distribution. The mean coercivities of the two components are 40 and 100 mt, respectively. It is estimated that the former component is carried by biogenic magnetite, and roughly corresponds to the noninteracting SD component derived from the FORC diagrams, and that the latter component is carried by eolian maghemite, corresponding to the interacting SD component. The c ARM /SIRM ratio decreases with increasing concentration of the interacting SD component. This implies that the c ARM /SIRM ratio is significantly affected by magnetostatic interactions and that it does not necessarily reflect magnetic grain size. The effect of magnetostatic interactions on natural remanent magnetization acquisition in sediments is not well understood, but if it is similarly sensitive to the SIRM, normalization by ARM in relative paleointensity estimations would overcompensate for the concentration of magnetic grains and cause the significant coherence between the normalized intensity and the normalizer that has been reported in the literature. Components: 6630 words, 8 figures, 1 table. Keywords: FORC diagram; magnetostatic interaction; IRM; North Pacific; marine sediments; rock magnetism. Index Terms: 1540 Geomagnetism and Paleomagnetism: Rock and mineral magnetism; 1521 Geomagnetism and Paleomagnetism: Paleointensity; 1512 Geomagnetism and Paleomagnetism: Environmental magnetism. Received 20 August 2007; Revised 14 October 2007; Accepted 15 November 2007; Published 9 February Copyright 2008 by the American Geophysical Union 1 of 12

2 Yamazaki, T. (2008), Magnetostatic interactions in deep-sea sediments inferred from first-order reversal curve diagrams: Implications for relative paleointensity normalization, Geochem. Geophys. Geosyst., 9, Q02005, doi: / 2007GC Introduction [2] First-order reversal curve (FORC) diagrams [e.g., Pike et al., 1999; Roberts et al., 2000] have great potential for characterizing magnetic grain assemblages in paleomagnetism and rock magnetism. They provide information on the statistical distribution of coercivities (H c ) and local interaction fields (H u ) for the magnetic grain assemblage in a sample. Their performance has been examined by theoretical studies [e.g., Muxworthy et al., 2004; Newell, 2005; Egli, 2006] and experiments with well-characterized synthetic samples [e.g., Muxworthy and Dunlop, 2002; Carvallo et al., 2004; Muxworthy et al., 2005; Roberts et al., 2006]. FORC diagrams have also been obtained from natural samples. For example, selection of basalts suitable for Thellier-type paleointensity determinations was successfully conducted on the basis of domain states and/or interactions estimated from FORC diagrams [Wehland et al., 2005; Carvallo et al., 2006b]. Other studies have applied FORC diagrams to sediments [Roberts et al., 2000, 2006; Weaver et al., 2002; Rowan and Roberts, 2006], but systematic characterization of magnetostatic interactions in deep-sea sediments has not yet been reported, although it is particularly important in relation to anhysteretic remanent magnetization (ARM) acquisition. [3] ARM, which is imparted by superimposing a smoothly decaying alternating field (AF) with a small DC bias field, is used for various purposes in paleomagnetic and rock magnetic research. Banerjee et al. [1981] and King et al. [1982] proposed that the ratio of ARM to magnetic susceptibility (c) or to saturation isothermal remanent magnetization (SIRM) can be used as a proxy for magnetic grain size, and these parameters have been used widely in particular for environmental applications of rock magnetism. The use of ARM in such grain size proxies is based on the fact that ARM acquisition is enhanced for grains in the single-domain (SD) size range, in particular in the finer part of the SD range [Maher, 1988]. In relative paleointensity estimations using marine sediment cores, paleomagnetists have tried to correct for differences in magnetizability of sediments using artificial remanent magnetizations [Tauxe, 1993], and ARM is often used for this purpose. The reason for using ARM is that its coercivity spectrum is often close to that of the natural remanent magnetization (NRM) [e.g., Levi and Banerjee, 1976; Kent and Opdyke, 1977; King et al., 1983]. [4] A fundamental problem is that the ARM acquisition efficiency is sensitive to magnetostatic interactions among magnetic grains. Sugiura [1979] experimentally showed that the ratio of ARM to SIRM decreases with an increase in the concentration of magnetite in artificial sediments. Yamazaki and Ioka [1997a] reported a similar relationship between ARM/c and c in natural sediments from the North Pacific Ocean. They interpreted that variations in ARM/c represent differences in the strength of magnetostatic interactions rather than magnetic grain-size variations because the variations were dependent on the R value of Cisowski [1981], which is indicative of the strength of magnetostatic interactions, and because grain-size variations are expected to be relatively small in pelagic environments such as the area that they studied. [5] In this paper, I characterize magnetic mineral assemblages in deep-sea sediments from the North Pacific Ocean using FORC diagrams and IRM acquisition curves. I semiquantitatively evaluate the relation between the magnetostatic interactions and ARM acquisition efficiency in these sediments. The results have significant implications for the use of magnetic grain-size proxies and the choice of a normalizer in relative paleointensity estimations. 2. Materials [6] Sediments from the middle latitudes of the North Pacific Ocean were used in this study (Table 1 and Figure 1). Two gravity cores, NGC65 and NGC69, each of about 7 m length, were analyzed here. These cores span the last 650 and 900 ka, respectively. Paleomagnetism and magnetic properties of these cores were reported by Yamazaki [1999] and Yamazaki and Kanamatsu [2007]. In addition, surface sediments were studied at 10 sites, which were taken using either a box 2of12

3 Table 1. Locations of Samples Sample Latitude Longitude Depth, m Cores NGC N E 4952 NGC N E 4858 Surface sediments NB N E 2505 NB N E 5741 NB N E 5814 NMC N E 5410 NB N E 4633 NK N E 2987 NK N E 3293 NB N E 4952 NB N E 5065 NB N E 4854 corer, a multiple corer, or a grab sampler. These sediments are part of the data set of Yamazaki and Ioka [1997a]; sites that are considered to belong to a similar depositional environment were selected for this study. Sediments from the Mid-Pacific Mountains and the Hess Rise (NB68, NK5, and NK6), where water depths are shallower than the carbonate compensation depth (CCD), are composed of calcareous clay. Sediments from below the CCD mainly consist of pelagic clay, and those northward of about 35 N contain abundant siliceous microfossil tests. [7] The NRM of these sediments is considered to be carried dominantly by magnetite/maghemite, although hematite also contributes to the magnetization [Yamazaki and Ioka, 1997b; Yamazaki, 1999]. This inference is based mainly on high S-ratios and subtle inflections indicative of the Verwey transition in low temperature SIRM measurements. The studied region belongs to the North Pacific Subtropical Gyre. Sediments in this region are considered to be mainly of eolian origin [Rea and Hovan, 1995; Weber et al., 1996], although biogenic magnetite also occurs [Yamazaki and Ioka, 1997b; Yamazaki and Kawahata, 1998]. This region is far from the eolian source region in the Asian continent, thus regional grain-size variations are small [Rea and Hovan, 1995]. Temporal variations of eolian grain size are also small during the late Pleistocene in this region, although some fluctuations have occurred with paleoclimate changes [Janecek and Rea, 1985]. [8] Magnetite dissolution during early diagenesis [e.g., Tarduno, 1994] is unlikely to have occurred in the studied sediments. The two gravity cores showed brownish color throughout the cores, which indicates an oxic condition. The oxic-redox boundary is inferred to be deeper than the bottom of the cores. Magnetite grains in the studied sediments are estimated to be partially maghemitized from the observation that thermal demagnetization curves of IRM from 6 to 300 K showed suppressed Verway transition being appeared only as a slight inflection [Yamazaki, 1999]. Reduction of maghemite to magnetite, which occurs in a certain depth in pelagic sediments [Smirnov and Tarduno, 2000], did not take place in these sediments, which is estimated from the observation that the shape of the low-temperature thermal demagnetization curves is almost identical with each other throughout the cores [Yamazaki, 1999]. 3. Measurements [9] Magnetic hysteresis and FORC measurements were conducted using an alternating gradient magnetometer (AGM, Princeton MicroMag 2900). For cores NGC65 and NGC69, 36 and 21 specimens, respectively, were selected from the entire depth ranges of the cores for this study. For the surface sediments, one specimen was measured from each site. FORC diagrams of two different resolutions were constructed. For high-resolution FORC diagrams, a field spacing between measurements was set to 0.5 mt. A total of 191 FORCs were measured, with H c between 0 and 60 mt, and H u between 15 and 15 mt. The narrow field spacing was adopted to precisely depict the shape of the Figure 1. Simplified bathymetry of the North Pacific Ocean and location of sediment samples used in this study. NGC65 and NGC69 (red) are gravity cores, and others (yellow) are surface sediments taken using either a box corer, a multiple corer, or a grab sampler. 3of12

4 Figure 2. (a) Relationship between c ARM /SIRM ratio (anhysteretic remanent magnetization susceptibility to saturation isothermal remanent magnetization) and SIRM. FORC diagrams for representative samples on the righthand side are for specimens from depths of (b) 0.04 m in surface sediments from grab-sample NK5, (c) 4.18 m in core NGC65, and (d) 1.31 m in core NGC69. A total of 191 FORCs were measured, and a smoothing factor (SF) of 3 was used [Roberts et al., 2000]. peak near the H c axis representing a noninteracting SD grain assemblage, although this may amplify noise. For normal-resolution FORC diagrams, a field spacing between measurements was set to 1.3 mt. A total of 150 FORCs were measured, with H c between 0 and 80 mt, and H u between 50 and 50 mt. Both the high- and normalresolution measurements were carried out for surface sediment specimens. For cores NGC65 and NGC69, high-resolution FORC diagrams were constructed for all specimens, but the normalresolution measurements were made only for selected specimens. The maximum applied field was 1.0T for both the high- and normal-resolution modes. The averaging time spent at each data point was 200 ms except for specimens with extremely weak magnetizations (NB68, NK5, and NK6), for which a 400 ms averaging time was used. A smoothing factor (SF) [Roberts et al., 2000] of 3 was adopted. The Forcobello software of Michael Winklhofer and Chris Pike was used for data processing [Winklhofer and Zimanyi, 2006]. [10] Different instruments were used for ARM acquisition for the surface sediments [Yamazaki and Ioka, 1997a] and for the two gravity cores [Yamazaki, 1999]. An AF of 100 mt was applied for the former study with a Natsuhara-Giken DEM- 86 AF demagnetizer and a homemade DC-field coil, while an 80 mt field was used for the latter with the in-line ARM apparatus linked to a passthrough magnetometer system (2G Enterprises model 760). The decay rates and frequencies of the AFs were also different. ARM acquisition strongly depends on the experimental process by which the ARM is imparted [Sagnotti et al., 2003], so the difference between the instruments was calibrated by imparting ARMs to the samples from core NGC65 using the two instruments. Out of a total of 260 discrete samples (10 cm 3 volume), every other specimen was subjected to ARM acquisition using one instrument, and the rest were magnetized using the other instrument. From the two data sets, the difference in the ARM acquisition efficiency for the two instruments was determined to be approximately 30%, and a correction was applied to ARM data from the surface sediments. The IRM for calculating c ARM /SIRM ratios was imparted using a pulse magnetizer. The IRM imparted at 2.5T is regarded as the SIRM. [11] IRM acquisition curves were obtained using the AGM. One hundred measurements were made at equidistant field steps on a log-scale ranging 4of12

5 Figure 3. (a) Normal-resolution FORC diagram for a specimen from a depth of 0.04 m in surface sediments from box-core NB70. The field spacing is 1.3 mt, and 150 FORCs were measured (SF = 3). (b) Cross section of Figure 3a through the peak of the FORC distribution parallel to the H u axis (dashed yellow line). (c) High-resolution FORC diagram for the same specimen as in Figure 3a. The field spacing is 0.5 mt, and 191 FORCs were measured (SF = 3). (d) Cross section of Figure 3c through the peak of the FORC distribution parallel to the H u axis. (e) Close-up of Figure 3c near the H c axis. (f) Cross section of Figure 3e through the peak of the FORC distribution parallel to the H c axis. from 3 mt to 1.4T. Specimens used for IRM acquisition were not identical to those used for the FORC measurements, but they were taken from the same discrete samples. The IRM acquisition curves were decomposed into magnetic coercivity components using the method of Kruiver et al. [2001] assuming that the IRM acquisition curves are a linear addition of components represented by cumulative log-gaussian functions. 4. Results [12] In a plot of c ARM /SIRM versus SIRM (Figure 2a), the data set from cores NGC65 and NGC69 has an inverse correlation, as presented by 5of12

6 Figure 4. Relationship between the c ARM /SIRM ratio and the proportion of an interacting (I) single-domain (SD) component to a noninteracting (N-I) SD component. Open squares represent specimens from surface sediments, red squares are from core NGC65, and blue squares are from core NGC69. Yamazaki [1999]. Data from the surface sediments extend the trend from the distribution of data points for the two cores toward lower SIRM values. This observation supports the correction for the instrumental difference in ARM acquisition efficiency applied to the surface sediments. [13] Typical examples of FORC diagrams are shown on the right-hand side of Figure 2. The closed contours indicate the dominance of SD grains [Pike et al., 1999; Roberts et al., 2000]. The H c values for the peak of the distributions range between 20 and 30 mt, which are typical of the coercivity of SD magnetite [Dunlop and Özdemir, 1997]. The FORC distributions have a narrow ridge along the H c axis, with a small vertical spread, indicating that magnetostatic interaction is weak (Figures 2b 2d). However, there is a small but notable difference in the vertical spread at the foot of the ridge among the three examples; the spread for the specimen at the upper left-hand corner of the c ARM /SIRM versus SIRM plot is smaller than that for the specimen at the lower right-hand corner of the plot. This suggests that the former specimen has smaller magnetostatic interactions for the part of the magnetic mineral assemblage responsible for this part of the FORC distribution. [14] A cross section through the peak of the FORC distribution parallel to the H u axis has a sharp peak with broad shoulders (Figures 3b and 3d). The sharp peak represents an assemblage of noninteracting SD grains [Pike et al., 1999; Roberts et al., 2000]. Magnetostatic interactions produce FORC distributions with a significant vertical spread. The broad shoulders of the profile therefore indicate that an assemblage of interacting SD grains also exists. In addition, contours that diverge from the H c axis on the normal-resolution FORC diagrams (Figure 3a) indicate the occurrence of multidomain (MD) grains [Pike et al., 2001], although their concentration must be relatively small. I semiquantitatively estimate the relative contributions of the noninteracting SD component (hereafter called the N-I component), the interacting SD component (referred to as the I component), and the MD component by curve fitting of the cross sections with three components assuming a Gaussian distribution of H u for each component. The standard deviation of each component was fixed to be 0.7 mt (N-I component), 6 mt (I component), and 23 mt (MD component), respectively, and the relative abundance of the three components was varied by trial and error so as to achieve an optimal fit, which is similar to the IRM decomposition of Kruiver et al. [2001] in the fitting method. For the fitting, cross sections were obtained from the highresolution FORC diagrams (Figure 3d). The MD component is represented nearly as a straight line on these cross sections (the 23 mt standard deviation of the MD component was estimated from cross sections of the normal-resolution FORC diagrams (Figure 3b)). FORC diagrams are inherently asymmetrical [Pike et al., 1999; Muxworthy et al., 2004; Newell, 2005], and it is only practical to fit components to one side of the profiles. The lower half (negative H u ) of the FORC diagrams was therefore used for the component fitting. [15] ARM acquisition efficiency decreases with increased strength of magnetostatic interaction, therefore it is expected that the c ARM /SIRM ratio of the N-I component would be larger than that of the I component, and that an increase in the proportion of the I component causes a decrease in c ARM /SIRM ratio. This is exactly what is observed in Figure 4, in which it is clear that the c ARM /SIRM ratio inversely correlates with the ratio of the I component to the N-I component. [16] Component analyses of IRM acquisition curves reveal that the IRM is mainly carried by two components, the low-coercivity component (hereafter the L component) and the middlecoercivity component (the M component) (Figure 5). 6of12

7 Variations in the shape of IRM acquisition curves can be described by changes in relative abundance of the L and M components with little change of the mean coercivity and dispersion. The mean coercivity of the L component ranges between 37 and 42 mt. The dispersion of the coercivity is relatively small; the dispersion parameter (DP) of Kruiver et al. [2001] is around The southern four sites (NB68, NG69, NB70, and NMC2) are exceptions, where the mean coercivity is a little lower, about 35 mt. The mean and dispersion of the M component are between 98 and 105 mt and between 0.26 and 0.30, respectively. On the basis of the mean coercivities, the L and M components are interpreted to represent magnetite and maghemite, respectively. In addition, relatively small amounts of a high-coercivity component (mean coercivity of 0.6T, DP of 0.4) and a verylow-coercivity component (mean coercivity of 15 mt, DP of 0.33) are required for optimal fitting. The occurrence of the very-low-coercivity component is partly due to the measurement technique of IRM acquisition curves; grains near the SD-superparamagnetic (SP) boundary would contribute to IRM on short timescales associated with rapid AGM measurements, whereas they do not contribute to IRM when using a pulse magnetizer and off-line measurement on a magnetometer because of thermal relaxation. Possible contribution of a SP component is indicated by the fact that the FORC distributions for the SD component intersect the Hc = 0 axis (Figure 3f). 5. Discussion Figure 5. Examples of IRM component analyses. Squares represent data points that define the gradient of IRM acquisition curves, which can be described by the sum (black curve) of two dominant components: a lowcoercivity (L) component (red curve) and a middlecoercivity (M) component (blue curve), and two minor components, a high-coercivity component (green) and a very-low-coercivity component (gray). Variations in the shape of IRM acquisition curves can be explained by varying the relative abundance of the four components without significantly changing the mean coercivities and dispersions. The FORC diagram for each example is shown in Figures 2b 2d Origin of Magnetic Minerals [17] FORC diagrams reveal that there are three types of magnetic grains in the studied central North Pacific sediments, including noninteracting and interacting SD components and a MD component. Here I infer the origin of these components. The ratio of the I component to the N-I component correlates with the ratio of the M component to the L component (Figure 6). The correlation is significant at the 95% confidence level, although the scatter is relatively large. This suggests that the I and N-I components roughly correspond to the M and L components, respectively. The scatter in Figure 6 is considered to be caused by several kinds of uncertainties. The correlation of components resulted from FORC analysis and those from IRM decomposition may not be one to one because the two methods will yield different coercivities due to inherent difference in measurements: mag- 7of12

8 Figure 6. Relationship between the proportion of an interacting SD component (I component) to a noninteracting SD component (N-I component) derived from FORC diagrams and the proportion of a middlecoercivity component (M component) to a low-coercivity component (L component) derived from IRM component analyses. netization under applied fields versus remanent magnetization. In addition, IRM component analysis can be influenced by magnetostatic interactions [Kruiver et al., 2001], and FORC diagrams may not be linearly additive in the presence of strong magnetostatic interactions [Carvallo et al., 2006a]. [18] In a pelagic environment far from land, eolian dust and biogenic magnetite are the only likely sources of magnetic minerals. It is reasonable to consider that maghemite is the main constituent of the magnetic minerals of eolian dust, which is transported from arid regions in Asia. The M and I components are hence inferred to be of eolian origin. The reason why the eolian component has significant magnetostatic interaction is not clear, but we infer that the eolian component contains titanomagnetites (and/or titanomaghemites) of volcanic and/or plutonic origins with ilmenite lamellae, which results in significant magnetostatic interaction. The relative contribution of the MD component increases with the I component (Figure 7), which suggests that the MD component is also of eolian origin. The N-I and L components are, on the other hand, considered to be carried by biogenic magnetite. In the central North Pacific, biogenic magnetite is ubiquitous, and the predominant morphology of the biogenic magnetite is octahedral [Yamazaki and Kawahata, 1998]. Biogenic magnetites can have high c ARM /SIRM ratios when the chains of magnetosomes remain almost intact and magnetostatic interactions are not significant [Kirschvink et al., 1992; Moskowitz et al., 1993]. The relatively small DP of the L component is consistent with the narrow grain-size range of biogenic magnetite. From Figure 4, a c ARM /SIRM ratio of or higher is expected for the N-I component when the proportion of the I component goes to zero. The c ARM /SIRM value of the N-I component, and the mean and dispersion of the coercivity of the L component are comparable to those of the BS (bacteria soft) component of Egli [2004a, 2004b], although it should be noted that the method of coercivity analysis and ARM instrumentation are different in this study. The BS component is considered to represent magnetosomes with octahedral morphology [Egli, 2004b], which is consistent with the predominance of this morphology in the studied area. [19] The inverse correlation between c ARM /SIRM and SIRM (Figures 2a and 8c) can be conceptually explained by significantly varying the eolian input over time and space and by maintaining a relatively constant concentration of biogenic magnetite. An increased flux of the I component with low c ARM / SIRM values would increase the SIRM but not the c ARM (Figures 8a and 8b). The three data points in the lower-left-hand corner of Figures 8a and 8b correspond to the sites above the CCD (NB68, NK5, and NK6); calcium carbonate dilution has Figure 7. Relationship between the interacting SD component (I component) and the MD component. Note that the apparent larger values of the vertical axis than the horizontal axis resulted from the MD component being integrated over a larger range of H u values and do not mean significance of the MD component. 8of12

9 Figure 8. (a) Relationship between SIRM and the proportion of interacting (I) to noninteracting (N-I) components. (b) Relationship between ARM susceptibility and the proportion of I to N-I components. (c) Relationship between c ARM /SIRM ratio and SIRM. (d) Relationship between c ARM /SIRM ratio and coercivity (H c ). caused the low values of SIRM and c ARM in these cases. Other sites lie near or below the CCD, and SIRM and c ARM variations (Figure 8) are not controlled by carbonate contents in general Magnetic Grain-Size Proxy [20] This study has revealed that c ARM /SIRM values in sediments from the central North Pacific Ocean are dominantly controlled by the relative abundance of noninteracting SD biogenic magnetite and interacting SD eolian maghemite. ARM acquisition efficiency is sensitive to magnetostatic interaction; therefore a larger proportion of the noninteracting component will cause higher c ARM /SIRM ratios. It is thus considered that magnetostatic interaction is the main factor that controls the c ARM /SIRM ratio in these sediments, and that the c ARM /SIRM ratio does not necessarily reflect magnetic grain-size changes. When utilizing c ARM /SIRM (or c ARM /c) as a magnetic grain-size proxy, it is necessary to take into consideration the effect of magnetostatic interaction on ARM acquisition. In the studied sediments, there is a possibility that the average grain size is larger when the c ARM /SIRM ratio is lower because the sediments include a MD component, which is associated with the eolian I component. [21] We should aware of magnetic grain-size changes during reduction diagenesis even in a pelagic environment [Tarduno, 1994; Tarduno and Wilkison, 1996], but this is not the case in the studied sediments because these sediments are above the oxic-redox boundary. The tendency for the surface sediments to larger c ARM /SIRM than that of the gravity cores (Figures 2a and 8c) can be explained as follows. The flux of the eolian I 9of12

10 component is considered to be lower during interglacial periods, which include the Holocene when the surface sediments were deposited. Furthermore, it is expected that sediments shallower than the CCD would have higher flux of the biogenic N-I component because of possible higher organic carbon contents. Decreasing abundance of magnetotactic bacteria with water depths was reported in the Atlantic [Petermann and Bleil, 1993]. If dissolution of a finer fraction of magnetite takes place, a decrease of magnetic concentration and an increase of average grain size will occur, which contradict increasing SIRM with a decrease in c ARM /SIRM in the studied sediments (Figures 2 and 8c). No correlation between c ARM /SIRM and the coercivity (H c ) determined from hysteresis loops (Figure 8d) also indicates that magnetite dissolution did not occur Normalizer in Relative Paleointensity Estimations [22] In relative paleointensity estimations from cores NGC65 and NGC69, Yamazaki [1999] and Yamazaki and Kanamatsu [2007] preferred SIRM as a normalizer to ARM for the following two reasons. First, NRM intensities normalized by SIRM for the two cores agreed well with each other, while those normalized by ARM did not. This implies that differences in NRM intensity due to differences in magnetic mineral concentration could not be compensated by ARM. Second, a cross-correlation analysis showed that significant coherence occurs between the normalized intensity and the normalizer for ARM over wide frequency ranges, which is not the case for SIRM. Yamazaki [1999] and Yamazaki and Kanamatsu [2007] estimated that the failure of ARM in this case is due to a strong control of magnetostatic interactions on ARM acquisition. The present study confirms the significance of magnetostatic interaction in the studied sediments. These cores contain a mixture of noninteracting and interacting SD assemblages, and because ARM acquisition efficiency of the two components is significantly different, it is unlikely that the magnetizability of the sediments can be represented by ARM. The effect of magnetostatic interactions on detrital remanent magnetization acquisition is not yet well understood, but if it is less sensitive to interactions than the ARM and similarly sensitive to the SIRM, normalization by ARM could overcompensate for the concentration of magnetic grains and cause significant coherency between the normalized intensity and the normalizer. [23] In relative paleointensity studies, ARM has been widely used as a normalizer, partly because the coercivity spectrum of ARM is considered to be closer to that of the NRM than SIRM [e.g., Levi and Banerjee, 1976; Kent and Opdyke, 1977; King et al., 1983]. However, some authors use SIRM as a normalizer because NRM/SIRM has a smaller dependence on the strength of AF demagnetization fields than NRM/ARM [Channell et al., 1998, 2004; Channell and Kleiven, 2000], which indicates that the coercivity of the SIRM in their sediments resembles that of the NRM. Furthermore, SIRM has increasingly been used as a normalizer because coherence or correlation between NRM/SIRM and SIRM is often smaller than between NRM/ARM and ARM [Tauxe and Shackleton, 1994; Lehman et al., 1996;Channell et al., 1998; Williams et al., 1998; Yamazaki, 1999; St-Onge et al., 2003; Yamazaki and Kanamatsu, 2007]. This suggests that a possible over-compensation by ARM, due to magnetostatic interactions, could occur widely in sediments. Marine sediments often contain noninteracting biogenic magnetite, and they are usually mixed with detrital and/or eolian components that can be affected by significant magnetostatic interactions. When selecting a normalizer, the effect of magnetostatic interactions should therefore be carefully evaluated. 6. Conclusions [24] Characterization of magnetic mineral assemblage in deep-sea sediments from the North Pacific Ocean using FORC diagrams and IRM acquisition curves have led to the following conclusions. [25] 1. FORC diagrams showed that the sediments consist of assemblages of noninteracting SD grains (the N-I component), interacting SD grains (the I component), and MD grains. The proportion of the I component to the N-I component in abundance inversely correlates with the c ARM /SIRM ratio. This indicates that magnetostatic interaction is the main factor that controls the c ARM /SIRM ratio in these sediments, and that the c ARM /SIRM ratio does not necessarily reflect magnetic grainsize changes. [26] 2. Component analyses of IRM acquisition curves revealed that the IRM is mainly carried by two components, the low-coercivity component with the mean coercivity of 40 mt (the L component) and the middle-coercivity component with the mean coercivity of 100 mt (the M component). The proportion of the M compo- 10 of 12

11 nent to the L component correlates with the proportion of the I component to the N-I component, which suggests that the M and L components correspond to the I and N-I components, respectively. It is estimated that the M and I components represent eolian maghemite, and L and N-I components represent biogenic magnetite. [27] 3. ARM is often used as the normalizer for relative paleointensity estimations. Because the studied sediments contain a mixture of noninteracting and interacting SD assemblages, and because ARM acquisition efficiency of the two components is significantly different, it is unlikely that the magnetizability of the sediments can be represented by ARM. The effect of magnetostatic interactions on detrital remanent magnetization acquisition is not yet well understood, but if it is less sensitive to interactions than the ARM, normalization by ARM could overcompensate for the concentration of magnetic grains and cause significant coherency between the normalized intensity and the normalizer that has been reported in literature. Acknowledgments [28] I thank Etsuko Usuda and Seiko Inoue for help with measurements and Hirokuni Oda and Nobutatsu Mochizuki for reading the manuscript. The paper was fundamentally improved by the thoughtful review by Andrew Roberts. Constructive review comments of Claire Carvallo, John Tarduno, Mike Jackson, and Catherine Kissel are also greatly appreciated. This study was partly supported by a Grant-in-Aid for Scientific Research ((A)(2) ) from the Japan Society for the Promotion of Science. References Banerjee, S. K., J. King, and J. A. Marvin (1981), Rapid method for magnetic granulometry with applications to environmental studies, Geophys. Res. Lett., 8, Carvallo, C., Ö. Özdemir, and D. J. Dunlop (2004), First-order reversal curve (FORC) diagrams of elongated single-domain grains at high and low temperatures, J. Geophys. Res., 109, B04105, doi: /2003jb Carvallo, C., A. R. Muxworthy, and D. J. Dunlop (2006a), First-order reversal curve (FORC) diagrams of magnetic mixtures: Micromagnetic models and measurements, Phys. Earth Planet. Inter., 154, Carvallo, C., A. P. Roberts, R. Leonhardt, C. Laj, C. Kissel, M. Perrin, and P. Camps (2006b), Increasing the efficiency of paleointensity analyses by selection of samples using first-order reversal curve diagrams, J. Geophys. Res., 111, B12103, doi: /2005jb Channell, J. E. T., and H. F. Kleiven (2000), Geomagnetic palaeointensities and astrochronological ages for the Matuyama-Brunhes boundary and the boundaries of the Jaramillo Subchron: Palaeomagnetic and oxygen isotope records from ODP Site 983, Philos. Trans. R. Soc. London, Ser. A, 358, Channell, J. E. T., D. A. Hodell, J. McManus, and B. Lehman (1998), Orbital modulation of the Earth s magnetic field intensity, Nature, 394, Channell, J. E. T., J. H. Curtis, and B. P. Flower (2004), The Matuyama-Brunhes boundary interval ( ka) in North Atlantic drift sediments, Geophys. J. Int., 158, Cisowski, C. (1981), Interacting vs. non-interacting single domain behavior in natural and synthetic samples, Phys. Earth Planet. Inter., 26, Dunlop, D. J., and Ö. Özdemir (1997), Rock Magnetism: Fundamentals and Frontiers, 573 pp., Cambridge Univ. Press, New York. Egli, R. (2004a), Characterization of individual rock magnetic components by analysis of remanence curves, 1. Unmixing natural sediments, Stud. Geophys. Geod., 48, Egli, R. (2004b), Characterization of individual rock magnetic components by analysis of remanence curves, 3. Bacterial magnetite and natural processes in lakes, Phys. Chem. Earth, 29, Egli, R. (2006), Theoretical aspects of dipolar interactions and their appearance in first-order reversal curves of thermally activated single-domain particles, J. Geophys. Res., 111, B12S17, doi: /2006jb Janecek, T. R., and D. K. Rea (1985), Quaternary fluctuations in the Northern Hemisphere trade winds and westerlies, Quat. Res., 24, Kent, D. V., and N. D. Opdyke (1977), Palaeomagnetic field intensity variation recorded in a Brunhes epoch deep-sea sediment core, Nature, 266, King, J., S. K. Banerjee, J. Marvin, and Ö. Özdemir (1982), A comparison of different magnetic methods for determining the relative grain size of magnetite in natural materials: Some results from lake sediments, Earth Planet. Sci. Lett., 59, King, J. W., S. K. Banerjee, and J. Marvin (1983), A new rockmagnetic approach to selecting sediments for geomagnetic paleointensity studies: Application to paleointensity for the last 4000 years, J. Geophys. Res., 88, Kirschvink, J. L., A. Kobayashi-Kirschvink, and B. J. Woodford (1992), Magnetite biomineralization in the human brain, Proc. Natl. Sci. U.S.A., 89, Kruiver, P. P., M. J. Dekkers, and D. Heslop (2001), Quantification of magnetic coercivity components by the analysis of acquisition curves of isothermal remanent magnetization, Earth Planet. Sci. Lett., 189, Lehman, B., C. Laj, C. Kissel, A. Mazaud, M. Paterne, and L. Labeyrie (1996), Relative changes of the geomagnetic field intensity during the last 280 kyear from piston cores in the Açores area, Phys. Earth Planet. Inter., 93, Levi, S., and S. K. Banerjee (1976), On the possibility of obtaining relative paleointensities from lake sediments, Earth Planet. Sci. Lett., 29, Maher, B. A. (1988), Magnetic properties of some synthetic sub-micron magnetites, Geophys. J. Int., 94, Moskowitz, B. M., R. B. Frankel, and D. A. Bazylinski (1993), Rock magnetic criteria for the detection of biogenic magnetite, Earth Planet. Sci. Lett., 120, Muxworthy, A. R., and D. J. Dunlop (2002), First-order reversal curve (FORC) diagrams for pseudo-single-domain magnetites at high temperature, Earth Planet. Sci. Lett., 203, Muxworthy, A. R., D. Heslop, and W. Williams (2004), Influence of magnetostatic interactions on first-order-reversal- 11 of 12

12 curve (FORC) diagrams: A micromagnetic approach, Geophys. J. Int., 158, Muxworthy, A. R., J. G. King, and D. Heslop (2005), Assessing the ability of first-order reversal curve (FORC) diagrams to unravel complex magnetic signals, J. Geophys. Res., 110, B01105, doi: /2004jb Newell, A. J. (2005), A high-precision model of first-order reversal curve (FORC) functions for single-domain ferromagnets with uniaxial anisotropy, Geochem. Geophys. Geosyst., 6, Q05010, doi: /2004gc Petermann, H., and U. Bleil (1993), Detection of live magnetotactic bacteria in South Atlantic deep-sea sediments, Earth Planet. Sci. Lett., 117, Pike, C. R., A. P. Roberts, and K. L. Verosub (1999), Characterizing interactions in fine magnetic particle systems using first order reversal curves, J. Appl. Phys., 85, Pike, C. R., A. P. Roberts, M. J. Dekkers, and K. L. Verosub (2001), An investigation of multi-domain hysteresis mechanisms using FORC diagrams, Phys. Earth Planet. Inter., 126, Rea, D. K., and S. A. Hovan (1995), Grain size distribution and depositional processes of the mineral component of abyssal sediments: Lessons from the North Pacific, Paleoceanography, 10, Roberts, A. P., C. R. Pike, and K. L. Verosub (2000), Firstorder reversal curve diagrams: A new tool for characterizing the magnetic properties of natural samples, J. Geophys. Res., 105, 28,461 28,475. Roberts, A. P., Q. Liu, C. J. Rowan, L. Chang, C. Carvallo, J. Torrent, and C.-S. Horng (2006), Characterization of hematite (a-fe 2 O 3 ), goethite (a-feooh), greigite (Fe 3 S 4 ), and pyrrhotite (Fe 7 S 8 ) using first-order reversal curve diagrams, J. Geophys. Res., 111, B12S35, doi: /2006jb Rowan, C. J., and A. P. Roberts (2006), Magnetite dissolution, diachronous greigite formation, and secondary magnetizations from pyrite oxidation: Unraveling complex magnetizations in Neogene marine sediments from New Zealand, Earth Planet. Sci. Lett., 241, Sagnotti, L., P. Rochette, M. Jackson, F. Vadeboin, J. Dinarès- Turell, A. Winkler, and Mag-Net Science Team (2003), Inter-laboratory calibration of low-field magnetic and anhysteretic susceptibility measurements, Phys. Earth Planet. Inter., 138, Smirnov, A. V., and J. A. Tarduno (2000), Low-temperature magnetic properties of pelagic sediments (Ocean Drilling Program Site 805C): Tracers of maghemitization and magnetic mineral reduction, J. Geophys. Res., 105, 16,457 16,471. St-Onge,G.,J.S.Stoner,andC.Hillaire-Marcel(2003), Holocene paleomagnetic records from the St. Lawrence Estuary, eastern Canada: Centennial- to millennial-scale geomagnetic modulation of cosmogenic isotopes, Earth Planet. Sci. Lett, 209, Sugiura, N. (1979), ARM, TRM and magnetic interactions: Concentration dependence, Earth Planet. Sci. Lett., 42, Tarduno, J. A. (1994), Temporal trends of magnetic dissolution in the pelagic realm: Gauging paleoproductivity, Earth Planet. Sci. Lett., 123, Tarduno, J. A., and S. L. Wilkison (1996), Non-steady state magnetic mineral reduction, chemical lock-in, and delayed remanence acquisition in pelagic sediments, Earth Planet. Sci. Lett., 144, Tauxe, L. (1993), Sedimentary records of relative paleointensity of the geomagnetic field: Theory and practice, Rev. Geophys., 31(3), Tauxe, L., and N. J. Shackleton (1994), Relative paleointensity records from the Ontong-Java Plateau, Geophys. J. Int., 117, Weaver, R., A. P. Roberts, and A. J. Barker (2002), A late diagenetic (syn-folding) magnetization carried by pyrrhotite: Implications for paleomagnetic studies from magnetic iron sulfide-bearing sediments, Earth Planet. Sci. Lett., 200, Weber, E. T., II, R. M. Owen, G. R. Dickens, A. N. Halliday, C. E. Jones, and D. K. Rea (1996), Quantitative resolution of eolian continental crustal material and volcanic detritus in North Pacific surface sediment, Paleoceanography, 11, Wehland, F., R. Leonhardt, F. Vadeboin, and E. Appel (2005), Magnetic interaction analysis of basaltic samples and preselection for absolute palaeointensity measurements, Geophys. J. Int., 162, Williams, T., N. Thouveny, and K. M. Creer (1998), A normalised intensity record from Lac du Bouchet: Geomagnetic palaeointensity for the last 300 kyr?, Earth Planet. Sci. Lett., 156, Winklhofer, M., and G. T. Zimanyi (2006), Extracting the intrinsic switching field distribution in perpendicular media: A comparative analysis, J. Appl. Phys., 99, 08E710, doi: / Yamazaki, T. (1999), Relative paleointensity of the geomagnetic field during Brunhes Chron recorded in North Pacific deep-sea sediment cores: Orbital influence?, Earth Planet. Sci. Lett., 169, Yamazaki, T., and N. Ioka (1997a), Cautionary note on magnetic grain-size estimation using the ratio of ARM to magnetic susceptibility, Geophys. Res. Lett., 24, Yamazaki, T., and N. Ioka (1997b), Environmental rockmagnetism of pelagic clay: Implications for Asian eolian input to the North Pacific since the Pliocene, Paleoceanography, 12, Yamazaki, T., and T. Kanamatsu (2007), A relative paleointensity record of the geomagnetic field since 1.6 Ma from the North Pacific, Earth Planets Space, 59, Yamazaki, T., and H. Kawahata (1998), Organic carbon flux controls the morphology of magnetofossils in marine sediments, Geology, 26, of 12