Paleomagnetic overprints in ocean sediment cores and their relationship to shear deformation caused by piston coring

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 107, NO. B4, 2067, /2001JB000518, 2002 Paleomagnetic overprints in ocean sediment cores and their relationship to shear deformation caused by piston coring Gary D. Acton, 1 Makoto Okada, 2 Bradford M. Clement, 3 Steve P. Lund, 4 and Trevor Williams 5 Received 12 March 2001; revised 9 October 2001; accepted 14 October 2001; published 13 April [1] We use paleomagnetic data from cores collected at Ocean Drilling Program Site 1062 on the Bahama Outer Ridge, as well as observations from prior cruises, to quantify the nature and to assess the causes of drilling overprints. The largest overprint arises from an isothermal remanent magnetization that is imparted to the core during coring or core retrieval. A second type of overprint is apparent only in soft sediments collected with a piston corer and has coercivities and unblocking temperatures that completely overlap those of the primary remanent magnetization, which means it cannot be removed by standard demagnetization methods. This demagnetizationresistant overprint can produce significant biases in paleomagnetic observations, particularly those from split core sections. We hypothesize that shearing, which occurs as a result of friction on the inside wall of the piston corer as it rips through sediments and cuts a core, deflects the ancient magnetization. We develop and test a model that predicts how shear deformation can rotate remanence-carrying grains and deflect the paleomagnetic remanence of a split core section. Using the shear model predictions and directions measured on split core sections and U channel samples from the undeformed core center, we estimate the average deformation sustained and then correct the split core data for biases caused by shear deformation. The reduction in the deviations between corrected split core data and U channel data is statistically significant, indicating that the shear model is capable of accounting for a significant portion of the systematic biases observed in paleomagnetic observations obtained from piston cores. INDEX TERMS: 1533 Geomagnetism and Paleomagnetism: Remagnetization; 1521 Geomagnetism and Paleomagnetism: Paleointensity; 3005 Marine Geology and Geophysics: Geomagnetism (1550); 1594 Geomagnetism and Paleomagnetism: Instruments and techniques; 4294 Oceanography: General: Instruments and techniques; KEYWORDS: Paleomagnetism, magnetic overprints, Ocean Drilling Program, JOIDES Resolution, piston coring, Site Introduction [2] A wealth of paleomagnetic data has been, and is continuing to be, obtained from drill cores, such as those collected on Ocean Drilling Program (ODP) and Deep Sea Drilling Project (DSDP) cruises. For example, on ODP cruises, paleomagnetic measurements before and after alternating field (AF) demagnetization are typically made continuously along split core sections cut from every core collected [e.g., Stokking et al., 1993; Keigwin et al., 1998] (Figure 1). Hence for every hole cored, a continuous downhole paleomagnetic record is acquired, making it possible to conduct a variety of paleomagnetic studies, such as high-resolution magnetostratigraphy, paleogeography, paleosecular variation, and geomagnetic field configuration studies [e.g., Schneider and Kent, 1990; Lund et al., 1998; Acton et al., 2000a]. 1 Ocean Drilling Program, Texas A&M University, College Station, Texas, USA. 2 Department of Environmental Sciences, Ibaraki University, Mito, Japan. 3 Department of Geology, Florida International University, Miami, Florida, USA. 4 Department of Earth Sciences, University of Southern California, Los Angeles, California, USA. 5 Borehole Research Group, Lamont-Doherty Earth Observatory, Palisades, New York, USA. Copyright 2002 by the American Geophysical Union /02/2001JB000518$09.00 [3] In many cases, the shipboard paleomagnetic record is the only one ever obtained for several reasons. First, repeating the shipboard measurements on subsamples taken from all of the cored interval would be extremely time consuming, and so subsequent shore-based studies instead focus on key intervals, which are usually those that appear to have interesting features as determined by shipboard measurements. Second, many of the intervals initially measured on the ship are subsequently sampled and consumed in a variety of other types of studies and thus are not available for further paleomagnetic investigations. Third, the shipboard measurements may be the only record of the paleomagnetic signal in some intervals because alteration of magnetic minerals can be rapid in some sediments when they are exposed to oxygen after being buried in reducing conditions [e.g., Richter et al., 1999; Yamazaki et al., 2000]. Thus obtaining reliable remanence measurements rapidly and continuously for every core maximizes the paleomagnetic information that can be gleaned from the relative rare core collections and helps direct more detailed shore-based analyses to intervals of greatest interest for paleomagnetic investigations. [4] Unfortunately, resolving the ancient magnetization can be difficult and in some cases impossible for both shipboard and shorebased studies because it is masked to some degree by drilling overprints, e.g., DSDP Leg 90 [Barton and Bloemendal, 1986], ODP Leg 104 [Bleil, 1989], ODP Leg 108 [Tauxe et al., 1989]; ODP Leg 116 [Hall and Sager, 1990], ODP Leg 115 [Backman et al., 1988], ODP Leg 134 [Collot et al., 1992]; ODP Leg 154 [Curry et al., 1995]; ODP Leg 157 [Fuller et al., 1998]; ODP Leg 160 [Roberts et al., 1996], ODP Leg 172 [Keigwin et al., 1998], and ODP Leg 186 [Sacks et al., 2000]. Drilling overprints are associated EPM 3-1

2 EPM 3-2 ACTON ET AL.: PALEOMAGNETIC OVERPRINTS IN OCEAN PISTON CORES Figure 1. Schematic diagram showing the Advanced Piston Corer (APC), drill ship, core, and paleomagnetic data. During an ODP cruise, the standard procedure is to cut the cores, which are typically 9.5 m or less in length, into 1.5- m-long sections with the deepest section possibly being shorter than 1.5 m. Each section is then split lengthwise into working and archive halves, called split core sections (see Figure 2 for orientation scheme). The magnetizations of the archive halves are measured continuously along each section in a long-core cryogenic magnetometer equipped with an in-line alternating field (AF) demagnetization unit (see the explanatory notes of Keigwin et al. [1998]). with continental as well as oceanic drill cores [e.g., Audunsson and Levi, 1989], though we will focus on ODP and DSDP coring. Herein we use the term drilling overprint in reference to any magnetic overprint acquired by the core during shipboard operations conducted up to the time the core is measured in the shipboard magnetometer. This includes piston or rotary coring and splitting or sawing the core for curation. We will focus particularly on drilling overprints in soft sediment cores, specifically on those collected with the Advanced Piston Corer (APC), which is the standard piston corer used by ODP (Figure 1). [5] In section 2 we present evidence for the existence of two primary drilling overprints: one that has a large magnitude but low coercivities and low-to-medium unblocking temperatures and can usually be removed by progressive AF or thermal demagnetization. It is characterized by a large vertical component, which results in initial natural remanent magnetization (NRM) directions with steep inclinations, along with a smaller radial-horizontal component, which biases the declination toward 0 or 180 (Figures 2, 3, and 4). This overprint arises from an isothermal remanent magnetization (IRM) that is imparted to the core during coring or core retrieval. A second overprint generally has a small magnitude but coercivities and unblocking temperatures that appear to overlap completely those of the primary remanent magnetization. Hence it cannot be isolated and removed by standard demagnetization methods, and so we refer to it as a demagnetization-resistant overprint. It does, however, produce systematic biases in the remanence direction and intensity that allow it to be identified (Figures 4 and 5 and discussion in sections 2 and 5). [6] A goal of this paper is to explain the origin of the demagnetization-resistant overprint that is observed in split core paleomagnetic data from APC cores. We hypothesize that shearing, which occurs as a result of friction on the inside wall of the piston corer as the corer rips through the sediments and cuts a core, produces deflections of the ancient magnetization (Figures 1 and 6). In sections 4 and 5 we develop and test a new model for how shear rotates remanence-carrying grains and produces a deflection of the remanent magnetization vector in piston cores. 2. Nature of the Drilling Overprints 2.1. Isothermal Remanent Magnetization (IRM) Overprints [7] Numerous ODP and DSDP paleomagnetic studies have found that the most prominent part of the overprinting is associated with an IRM that is imparted to the core during drilling, while the core is being pulled through the drill pipe to the rig floor, or, though less likely, during subsequent cutting or splitting of the core as it is curated. The overprints are not surprising given that the core is collected with a metal cutting shoe or drill bit in the vicinity of a highly magnetic bottom hole assembly (BHA) and must pass through a long metal drill string. Moreover, the core resides in a metal core barrel, separated from the core by only a thin plastic core liner, and passes near rig floor equipment that produces large magnetic fields. Magnetic field measurements around the BHA, core barrel, and other drilling equipment have indicated that fields may locally exceed 1 T, though generally the fields generated near these are less than 20 mt [Keating, 1984; Sager and Hutton, 1986; Stokking et al., 1993; Fuller et al., 1998]. [8] The IRM component, while very large, can generally be removed by low AF demagnetization (<25 mt), though for some magnetic mineralogies, particularly those with low coercivities, the IRM may completely overprint any remanent magnetic signal. Similarly, thermal demagnetization of C is typically sufficient to remove the IRM component. In some cases, however, thermal demagnetization up to 500 C is necessary to completely remove the IRM [e.g., Acton et al., 2000a], which is similar to IRM

3 ACTON ET AL.: PALEOMAGNETIC OVERPRINTS IN OCEAN PISTON CORES EPM 3-3 AF demagnetization, with the remainder being related to deformation. Following Fuller et al. [1998, p. 48], we use the term radialhorizontal rather than just radial because as they note, even when the radial-horizontal component is present, the inclination can actually be steep and remain so throughout demagnetization. The radial-horizontal overprint mainly biases the horizontal component, which given the ODP orientation system used in measuring the remanence of archive half sections (Figure 2), results in declinations biased toward 0 or 180, depending on whether the overprint is directed toward or away from the center of the core, respectively. Inward directed radial-horizontal overprints appear to be the most common, but both forms exist. As Audunsson and Levi [1989] suggested, such a drilling-related IRM may be acquired near the drill bit or cutting shoe where the ambient magnetic field lines might be expected to have a significant inward component on the interior and an outward component on the exterior as they radiate from the end of the drill string. Indeed, Fuller et al. [1998] observed radial, as well as vertical, magnetic fields up to 5 mt in size from measurements made inside and around a core barrel fitted with an APC cutting shoe. Figure 2. Illustration of standard ODP core orientation system with radial and vertical overprints shown. Given this orientation system and given that the archive half sections are measured in the shipboard magnetometer, an inward directed radial overprint, such as shown here, results in measured declinations biased toward 0. Outward directed radial overprints would result in declinations biased toward 180. components in continental rocks from outcrops that have been struck by lightning [e.g., Acton et al., 2000b]. [9] The IRM drilling overprint is characterized by its nearvertical orientation, which can be either downward or upward directed (Figure 3). In a rare, but important example, D. Schneider and D. Vandamme (as cited by Backman et al. [1988, pp ]) noted intervals with steep negative inclinations and intervals with moderate-to-steep positive inclinations associated with overprinting within a single core. The pattern of overprinting was shown to be similar in other cores when the same core barrel was used, implicating the core barrel as a source of at least part of the IRM. From our experience on Legs 165, 172, 178, and 186, the IRM drilling overprint is nearly constant within a core barrel and between core barrels, with the NRM inclinations prior to AF demagnetization being steep and positive (60 90 ). Possibly, the core barrels are completely remagnetized, preferentially with a vertical orientation, over time by some common mechanism, such as traveling up and down the drill string. Subsequently, cores collected within them have nearvertical overprint owing to the core barrel magnetization and to the trip through the drill string. [10] In addition to the near-vertical IRM, a smaller radialhorizontal component also appears to be present in oceanic drill cores (Figures 3, 4, and 5a), and has been documented in continental drill cores as well [Audunsson and Levi, 1989]. As we will discuss further in section 2.2, only part of the radialhorizontal component is an IRM overprint that can be removed by Figure 3. Orthogonal vector demagnetization diagrams for two samples from Hole 1062E illustrating the existence of a steep downward directed drilling overprint and a smaller radial overprint (see Figure 2 for orientation system). The steep overprint, which produces a strong deflection of the inclination (open squares), and the radial component, which produces a small deflection of the declination, are removed following mt AF demagnetization. In this example the radial component appears (top) to be outward directed (declination deflected toward 180 ) for the reverse polarity sample and (bottom) to be inward directed (declination deflected toward 0 ) for the normal polarity samples. We suggest that the removed overprints are isothermal remanent magnetization (IRM) components imparted during drilling. Other drilling overprints, which are resistant to AF demagnetization, may still be lurking in the linear demagnetization paths that decay toward the origin of the plot.

4 EPM 3-4 ACTON ET AL.: PALEOMAGNETIC OVERPRINTS IN OCEAN PISTON CORES Figure 4. Orthogonal demagnetization diagrams for five samples taken from a 2-cm-thick whole round section. Samples 2 5 are from the perimeter of the core, and sample 1 is from the center. AF demagnetization steps are every 2.5 mt. Open symbols are the projection of the remanence direction onto the vertical plane, and solid symbols are the projection onto the horizontal plane. Owing to the drilling overprint, the four samples from the perimeter of the core have directions that are initially steep and point toward the center of the core. They also have intensities roughly an order of a magnitude higher than the sample from the center. Following demagnetization of mt, the directions become shallower, though only a small portion of the radial overprint is removed from the perimeter samples. The primary remanence direction is interpreted to be pointing southeast and upward at a shallow angle and is resolved in demagnetization steps higher than 12.5 mt only for the center sample. Taken from Figure 6 of Curry et al. [1995]. [11] The IRM overprints are not homogeneous across the core, and the vertical and radial-horizontal components are not always equally easy to remove with low AF demagnetization. Closer examination illustrates that the vertical and radial-horizontal drilling overprints are more extensive near the periphery than in the center of the core (see D. Schneider and D. Vandamme, as cited by Backman et al. [1988, p. 478, Figure 12], note that the inclination axis on their Figure 12 appears to have been inadvertently inverted; P. Roperch, L. Stokking, and X. Zhao as cited by Collot et al. [1992, pp ]; and C. Richter, D. Schneider, and J.-P. Valet as cited by of Curry et al. [1995, pp ]). Furthermore, the vertical component can generally be removed by AF demagnetization more readily than radial-horizontal component for samples near the periphery. The radial-horizontal component, while small near the center of the core and presumably completely canceled at the very center of the core, appears to be only slightly affected by AF demagnetization near the periphery of the core (Figure 4). This suggests that the radial-horizontal component is more than just a IRM overprint Demagnetization-Resistant Overprints [12] Overprints that have coercivities or unblocking temperature spectra similar or identical to that of the primary magnetization elude identification and removal in standard paleomagnetic analyses, which rely on progressive AF or thermal demagnetization, orthogonal vector demagnetization diagrams, and principal component analyses. For a drilling overprint with this property,

5 ACTON ET AL.: PALEOMAGNETIC OVERPRINTS IN OCEAN PISTON CORES EPM 3-5 Figure 5. Comparison of split core and U channel paleomagnetic data for several cores following AF demagnetization at 20 mt. (a) Declination data from Core 1062E-8H illustrating a systematic bias that depends on the azimuthal orientation of the core, particularly the measured declination of the remanence vector relative to the orientation of the split core surface. When the measured declination is <180, the split core declinations are less than the U channel declination, whereas when the measured declination is >180, the split core declinations are greater than the U channel declinations. The effect is that of an inward pointing radial-horizontal overprint that biases the declination of split core data toward values of 0 or 360. Inclination data (b) for Core 1062E-8H illustrating that split core directions can be systematically steeper than those from U channels and (c) for Core 1062E-5H illustrating an inclination bias in the opposite sense. orthogonal vector demagnetization plots may reveal linear demagnetization paths with progressive AF or thermal demagnetization following the removal of the IRM overprint with its low coercivity and low-to-medium unblocking temperature (Figure 3). The characteristic remanent magnetization (ChRM) direction defined by the linear paths is thus not the primary magnetization alone but is instead composed of the primary magnetization plus a demagnetization-resistant drilling overprint. [13] Observing and characterizing this overprint has been complicated because it is typically small relative to the IRM overprint

6 EPM 3-6 ACTON ET AL.: PALEOMAGNETIC OVERPRINTS IN OCEAN PISTON CORES and hidden from standard demagnetization analyses. It is the nature of the bias, however, that has allowed it to be observed. The most obvious evidence is the radial-horizontal component, which biases declinations toward 0 or 180, as discussed in section 2.1. A particularly good example of the demagnetization-resistent radialhorizontal component is shown in Figure 4 (taken from Figure 6 of Curry et al. [1995]), where four samples were collected around the periphery of a core. Even after AF demagnetization in excess of 20 mt, the declinations for the four samples differ significantly, and all point toward the center of the core. Though less obvious, the declination bias can also be observed in comparisons of split core and U channel data collected from the same core (Figure 5a). In addition, this overprint is sometimes apparent from one core to the next because it can also produce a nearly constant bias in the inclination (e.g., Figures 5b and 5c) and intensity within a core [Keigwin et al., 1998, pp ]. The sense of the biases in inclination, declination, and intensity is dependent on the direction of the primary magnetization relative to the azimuthal orientation of the core, which is an important clue as to the origin of the overprint (Figure 5). Another clue is that these types of demagnetization-resistant overprints are observed in soft sediment piston cores but not in lithified sediments or hard rock formations [Fuller et al., 1998]. [14] Some of the characteristics of the demagnetization-resistant overprint are common to the drilling-related IRM overprint; that is, both have a radial-horizontal component that can bias the declination. Similarly, both could steepen the inclination. Thus separating the two can be difficult, and in some instances the IRM may be strong enough to completely remagnetize the core, particularly when the remanent-carrying minerals have low-to-medium coercivities. In such cases, the drilling-related IRM would cause a demagnetization-resistant overprint in the sense that magnetic cleaning would not reveal the primary magnetization direction. Generally, however, the drilling-related IRM overprint is magnetically soft and can be removed by low field AF demagnetization. As mentioned in section 2.1, it is also nearly always observed with a steep downward component. Removal of the drilling-related IRM is thus usually clear in orthogonal demagnetization plots such as those shown in Figure 3. We emphasis that the demagnetizationresistant overprint that we seek to explain is revealed in the biases it causes and, unlike the drilling-related IRM, cannot be removed by AF or thermal demagnetization. [15] The preferred explanation for the demagnetization-resistant overprints has been that the cores are partially, or in some cases fully, remagnetized by mechanical disturbance that occurs in the presence of strong magnetic fields, such as those around the drilling bits, cutting shoes, core barrels, or core-splitting equipment [Burmester, 1977; Gravenor et al., 1984; Stokking et al., 1993; Jackson and Van der Voo, 1985; Roberts et al., 1996; Fuller et al., 1998; D. Schneider and D. Vandamme, as cited by Backman et al., 1988]. The disturbances proposed vary from vibration to fluidization and extreme core deformation such as suck in, which is where sediment has been sucked up into the core liner, often by tens of centimeters. An alternate explanation is that the periphery of the core has sustained a strong IRM that completely remagnetized sediments near the core liner, whereas those sediments toward the center of the core are exposed to smaller fields and are only partially remagnetized. Both explanations are likely valid in specific situations. For example, some sediments that have relatively soft magnetizations may be com- Figure 6. (opposite) (top and bottom) Core deformation in soft sediment cores resulting from piston coring. The downward bent layers illustrate that the amount of shear rapidly decays away from the core liner. (middle) Deformation of a horizontal feature in a split core as predicted by the function Z(r) for three different degree of deformation (b) values (see text).

7 ACTON ET AL.: PALEOMAGNETIC OVERPRINTS IN OCEAN PISTON CORES EPM 3-7 Figure 7. Schematic cross section of a split core illustrating the relationship between the trend of sheared laminations and the amount a particle or piece of core is rotated. pletely remagnetized along the periphery of the core by the drilling-related IRM. Likewise, strong deformation of the core may result in complete remagnetization of the sediment in the direction of the ambient field, in a manner analogous to remanence acquisition in clay bricks [Games, 1977]. We believe that complete remagnetization during extreme deformation dominates in suck-in cores [e.g., Roberts et al., 1996] and may even be common for strongly deformed sediment that occurs within a millimeter or a few millimeters of the periphery but that it cannot be the only overprinting mechanism. [16] In fact, both explanations fail to explain cases where the measured inclination is significantly shallower than expected and, more importantly, significantly shallower than the drilling-related IRM (e.g., Figure 5c), which presumably represents the orientation of the ambient field as the sediments are shaken, stirred, and remagnetized. Furthermore, the IRM mechanism is incompatible with paleomagnetic measurements, which indicate that at least part of the primary remanence is still present near the periphery of the core, and with rock magnetic analyses, which indicate that the demagnetization-resistant overprint has characteristics different from an IRM remagnetization [Fuller et al., 1998]. Some other overprinting mechanism must therefore be invoked. 3. Paleomagnetic Data [17] To test our hypothesis and to examine the average effect of shearing on ODP sediment cores, we choose paleomagnetic data from an interval that extends over tens of meters and hence over several different cores. Shear deformation within a short interval within a core or over an entire interval spanned by a core can and does vary. Over several cores, the variations tend to average out. To thoroughly test the model, the data should also span more than one core because our model predicts deflections of the original remanence vector that are ultimately related to the azimuthal orientation of a piston core, which should vary randomly from core to core. Finally, we want to look at data that also span both normal and reversed polarity because our model predicts opposite orientations for the radial overprint when the paleomagnetic vector is pointing downward versus that when it points upward. [18] Here we use standard paleomagnetic data from split core measurements made on archive half core sections and U channel samples taken from the opposing working half core sections. The U channel samples are strips of sediment collected from the undeformed center of the core sections, each 2 cm by 2 cm in cross section and up to 1.5 m long [Tauxe et al., 1983; Nagy and Valet, 1993]. We use the standard ODP orientation system for all samples measured (Figure 2). [19] The U channels were collected from core sections for Cores 6H 10H from Hole 1062E drilled in a mud wave on the Bahama Outer Rise. These cores contain fine-grained sediments that alternate between carbonate-rich and clay-rich lithologies associated with deposition during interglacial and glacial periods, respectively. The sediments come from an interval cored from 46 to 98 m below seafloor (mbsf ) and vary in age from 0.35 to 1.1 Ma. They were chosen because (1) they have a mean intensity that can be accurately measured (varying between 10 2 and 10 4 A/m), (2) they have stable magnetizations with negligible change in the remanence direction after AF demagnetization above 10 mt, indicating the IRM overprint is removed by 10 mt or less, (3) the remanence directions and intensity agree very well with paleomagnetic results from coeval sediments from six other holes cored at the same site, (4) hysteresis and low-temperature rock magnetic analysis indicates the remanence is carried primarily by pseudo-single-domain magnetite, the most common remanence carrier observed in paleomagnetic studies, and (5) both polarities are present. [20] The split core results are from shipboard measurements made during Leg 172 in a pass-through cryogenic magnetometer (Model 760R from 2G Enterprises) following 20 mt of AF demagnetization [Keigwin et al., 1998]. U channel samples were measured at the University of California, Davis, in a similar magnetometer (Model 755R from 2G Enterprises), and again we use data following 20 mt AF demagnetization. The major difference between the magnetometers is in the size of their sensor coils and the resulting sensitivity. The magnetometer at Davis has smaller diameter coils that are nearer the sample as it passes through the sensor region, which results in this magnetometer measuring a narrower interval of sediment (<7 cm) than the shipboard magnetometer (<10 cm). Some differences between

8 EPM 3-8 ACTON ET AL.: PALEOMAGNETIC OVERPRINTS IN OCEAN PISTON CORES Figure 8. Inclination, declination, and intensity anomalies expected as a function of true and observed declination for different true inclination values. To find the true values for the inclination and declination, the inclination anomaly and the declination anomaly should be added to the measured inclination and declination, respectively. To find the true intensity, the observed intensity should be multiplied by the intensity anomaly value. The measured declination is that defined for the ODP archive half coordinate system (see Figure 3). The degree of deformation is held constant at 0.05.

9 ACTON ET AL.: PALEOMAGNETIC OVERPRINTS IN OCEAN PISTON CORES EPM 3-9 the measurements made on the same interval can therefore be related to magnetometer differences, though these are usually limited to where the intensity or directions of the remanence vector changes rapidly over an interval comparable in size to the sensor region [e.g., Nagy and Valet, 1993; Roberts et al., 1996]. The magnetometer at Davis also resides in a magnetically shielded room, whereas the shipboard magnetometer resides on a moving metal ship, resulting in higher noise and lower sensitivity for the shipboard magnetometer. The lower sensitivity may lead to measurement differences when sediments have magnetizations <10 4 A/m. [21] Split core measurements were made every 5 cm on each core section. After removing any measurement made within 5 cm of the end of a core section and any measurement made within intervals with strong coring disturbance or gas voids, all of which can potentially give bogus results, we were left with 797 remanence measurements. U channel samples were measured every centimeter, allowing us to select the same 797 intervals for comparison. 4. Model for Shear Deformation [22] The coring process for the ODP Advanced Piston Corer (APC) is virtually instantaneous as the corer is shot into the sediment with a force of up to 28,000 pounds (125,000 N). The APC has been successful in retrieving nearly pristine, continuous sedimentary records from around the world. Even so, most APC cores have some form of deformation, most of which have insignificant impact on subsequent geoscience investigations. The most common deformation is that related to the shoe of the piston corer cutting through the sediment and ensuing drag of the sediment on the inside walls of the corer as the sediment slides up into the plastic core liner. The result is that sediment near the core liner is typically bent or sheared downward, and within <1 mm of the core liner, a zone of vertically smeared sediment often exists. The downward bending can be seen to rapidly decrease toward the center of the core in cores with horizontal layers or laminae (Figure 6). Similar deformation has been observed and described in the past for a variety of soft sediment corers [Hvorslev, 1949]. [23] During Leg 172, M. Okada (as cited by Keigwin et al. [1998, pp ]) noted that if the radial thickness of the region with deformed sediment is >1 cm, which is commonly the case, then the deformed sediment volume exceeds the undeformed sediment volume for a standard split core section. Therefore the deformed sediment may contribute significantly to the mean remanent magnetization measured on split core sections. He then proposed a simple shear model that successfully predicted many of the observed changes in mean inclination that were observed from one core to the next in several of the holes cored on Leg 172 [e.g., see Keigwin et al., 1998, p. 104, Figure 25]. More importantly, it illustrated that the size of this overprint was strongly dependent on the orientation of the remanence vector relative to the azimuthal orientation of the split core section being measured. For example, a remanent vector pointing downward and away from the split core face would tend to get steeper, whereas a remanent vector pointing downward and toward the split core face would get shallower. Hence variable inclination and intensity anomalies could occur from one core to the next even when the sediment had a constant paleomagnetic direction and even when the core had a constant amount of shear deformation. Figure 9. (opposite) Inclination, declination, and intensity anomalies expected as a function of true declination for different degrees of deformation. The true inclination is held constant at 45. Anomaly values are as described in the Figure 8 caption.

10 EPM 3-10 ACTON ET AL.: PALEOMAGNETIC OVERPRINTS IN OCEAN PISTON CORES Figure 10. Split core inclination, declination, and intensity are shown before (solid dots) and after (open dots) the shear correction was applied. U channel measurements (solid squares) are shown for comparison. The agreement between the split core and U channel data improves significantly after the split core data are corrected for shear deformation. [24] Even though the results were encouraging, this initial model was an over simplification and it failed to explain any of the observed radial-horizontal overprint. In particular, the model had a physically unrealistic aspect in that only the radial component of the remanence vector was reoriented by shearing, resulting in a net change in only the length of the vertical component. In reality, micron to submicron size grains, some of which are the remanent-carrying magnetic minerals, would be physically rotated. Thus the entire remanence vector is rotated by shearing, not just the radial component. [25] Here we propose a new model that accounts for rotation of particles caused by shearing with associated rotation of the remanence carried by magnetic particles. As a result, shearing not only produces changes in the inclination and intensity but also produces a change in the declination that results in a radial overprint. In the model the physical rotation of portions of the sediment that result from shearing are analogous to folding that occurs at dome-shaped structures on outcrop scale. When paleomagnetists sample such large-scale features, they would simply unfold the sedimentary layers carrying a remanence to get the predeformation paleomagnetic direction. The paleomagnetic fold correction is done by rotating the remanence vector about an axis by an amount that would restore the layers to horizontal. In the center of a dome structure, folding is absent or negligible, and no fold correction would be required. Away from the center, progressively larger rotations about horizontal axes tangential to the dome would be needed to restore the remanence vector to its prefold direction. In a manner analogous to this, shearing in piston cores bends the sedimentary layers and rotates the remanence vector near the core liner downward, with the amount of rotation increasing progressively away from the undeformed center of the core. [26] Similar to M. Okada (as cited by Keigwin et al. [1998]), we use the function Zr ðþ¼ b½lnð1 r=rþþr=rš r ¼ U to R ð1þ to determine the amount of shear or bending of a horizontal line as a function of distance r from the center of the core, where b determines the magnitude of bending and is so referred to as the degree of deformation (Figure 6). R is the radius of a core (3.3 cm for a standard ODP piston core) and U is the distance from the center of the core where shear deformation begins. The function Z(r) was derived to produce curves that mimic the visual deformation. Visual inspection of cores indicates that the inner 1 cm of a core is rarely deformed (U 1 cm) and that b values fall between 0.0 and 0.4, with lower values (<0.2) being most common. More highly deformed cores exist but would likely be classified as suck-in cores and avoided for paleomagnetic study. [27] We next divide the cross section of a split core into a large number of tiny pieces of sediment and determine how much each piece would be rotated for a certain degree of deformation. We used 828 pieces in our calculations and found that using more than 500 pieces resulted in negligible changes in the model predictions. We estimate the rotation sustained by each piece of sediment that lies some distance r from the center of the core by calculating the dip of Z(r), which is dz/dr. The pieces affected by shear are then rotated about horizontal rotation axes that are perpendicular to the radial vector to a particular piece of sediment; that is, the axes are horizontal and tangential to the core liner (Figure 7). The vector sum of all the pieces gives the model predicted remanence vector that would be observed from magnetometer measurements of split core sections affected by shear. The deflection results in a bias or anomaly, which is the difference between this predicted value and the original or true remanence vector (Figure 8). Thus the true values for the inclination (I T ), declination (D T ), and intensity (J T ) are defined as I T ¼ I M þ I A ; D T ¼ D M þ D A ; J T ¼ J M J A ; where I M, D M, and J M are the model predicted (or, as shown below, measured) inclination, declination, and intensity, respectively, for split core sections, and I A, D A, and J A are the anomalies or biases in inclination, declination, and intensity, respectively, ð2þ ð3þ ð4þ

11 ACTON ET AL.: PALEOMAGNETIC OVERPRINTS IN OCEAN PISTON CORES EPM 3-11 Figure 11. Deviations in inclinations and declinations between U channel samples and split core sections from Hole 1062E are shown (a and b) before and (c and d) after correction for shear deformation with a constant degree of deformation of 0.07 (solid dots). For comparison purposes we show the model predictions (dashed line) in Figures 11a and 11b for the case where the true inclination is 47, which is the expected geocentric axial dipole inclination for Hole 1062E. Actual model predictions vary with inclination as discussed in the text. caused by shear deformation. Here the true and model declinations are for azimuthally unoriented cores in the ODP orientation system Model Predictions [28] The size of the deflection of the remanence vector depends strongly on the declination of the remanence vector relative to the split core coordinate system, as well as on the inclination (Figure 8) and the degree of deformation (Figure 9). In the archive half coordinate system used by ODP the model predicts that any interval with positive true remanent inclinations will have a radial-horizontal overprint that gives declinations biased toward 0, whereas intervals with negative inclinations will be biased toward 180. The deflection of the declination is largest for steep remanence vectors, where the declination can flip by 180. Significant inclination deflections are also predicted even for relatively small deformations; for example, deflection by as much as ±10 can occur when the degree of deformation is only The intensity is least affected by shear in that the relative change in intensity versus declination is typically a few percent. For example for a deformation with b = 0.1, the observed intensity will be on average 9% smaller than the true intensity, but it only varies by ±2% about this average. For studies of relative paleointensity the biases in the split core intensities caused by shear are therefore negligible, except perhaps where deformation is strong and varies significantly from one core to the next Correcting Split Core Remanence Measurements [29] The predicted paleomagnetic directions that result from shear for a given degree of deformation can be used to correct the observed directions from split cores. We do this by calculating the deflection of the true direction for all possible combinations of true

12 EPM 3-12 ACTON ET AL.: PALEOMAGNETIC OVERPRINTS IN OCEAN PISTON CORES directions (i.e., 0 declination >360 and 90 inclination 90 ) rounded to the nearest degree. For each true direction we obtain a predicted direction (i.e., that which would be expected to be observed) and the shear anomaly or correction for a constant degree of deformation. The shear correction is composed of the inclination, declination, and intensity anomaly components as defined in (2) (4). The observed (measured) direction can then be matched to the nearest model predicted direction, and the shear correction can be applied by adding the inclination and declination anomaly corrections and multiplying by the intensity anomaly correction. In Figure 10 we illustrate the affect of correcting the split core data for a 0.08 degree of deformation, which brings the split core data into much better agreement with U channel results from the same interval A Method for Estimating the Average Amount of Shear [30] An estimate of the mean value for the degree of deformation (b) can be obtained by minimizing the difference between the true paleomagnetic direction and the shear-corrected direction for a single core interval or for multiple intervals down core. Of course, we do not know the true directions, though they can be estimated from U channel paleomagnetic data, where the U channel sample has been taken from the undeformed center portion of a core section. The shear-corrected directions are the split core data corrected for shear for a given b value. By computing the shear-corrected directions from multiple b values, an estimate of the amount of shear can be obtained for intervals where both U channel samples and split core sections have been measured. [31] We choose to minimize a chi-square statistic defined as c 2 ¼ Xn i¼1 ðd i Þ 2 s d ; ð5þ where d i is the angular distance between the ith U channel and shear-corrected split core paleomagnetic direction for n observations. The average standard error for angular distance is given by s d, which is taken to be twice the standard deviation of the angular distance between U channel and split core results when no shear correction has been applied. This assignment of the standard error is somewhat arbitrary but results in minimum chi-square values that are reasonable; that is, the reduced chi-square statistic, which is the chi-square divided by the number of degrees of freedom (number of observations minus the number of parameters being estimated), is one or less at its minimum. [32] To find the minimum chi-square, an iterative search is conducted for different b values. Each time the value of b is changed, a new set of shear-corrected directions are calculated, and chi-square is recalculated. The 95% confidence limits for b are estimated by constant chi-square boundaries [Press et al., 1992, pp ]. Since we are searching for the boundaries for only one parameter (b), the 95% confidence interval occurs where chisquare is 3.84 more than the minimum chi-square. 5. Results [33] We seek to estimate the degree of deformation that minimizes the misfit between the split core and U channel data and then to test whether the reduction in this misfit is significant enough to justify making a shear correction. Using the 797 paleomagnetic observations where both split core and U channel measurements have been made, we correct the split core observations for shear using a constant degree of deformation. We then compared these corrected values to the U channel data and compute the chi-square statistic. After repeating this process for degree of deformation values that vary from 0.0 to 0.40, we found a best fit value of 0.07 with a 95% confidence limit of [34] Relative to no shear correction, the best fit shear correction decreases chi-square by 8.5% (from 422 to 386) and gives a minimum reduced chi-square of with 796 degrees of freedom. An F ratio test for the statistical significance of an additional adjustable parameter [e.g., Bevington, 1969, pp ] gives a value of F = 74. The probability of F being this large by chance is less than (the 0.1% risk level for 1 versus 796 degrees of freedom), indicating that the inclusion of a shear correction is highly significant. [35] The shear model is further supported graphically by comparing the model predictions versus the observed differences between split core and U channel inclinations and declinations (Figure 11). The azimuthal dependence predicted by the model for these differences is apparent in the observations prior to making the shear correction (Figures 11a and 11b) but is absent after making the correction (Figures 11c and 11d), which is what would be expected if the shear correction has removed systematic variations and only random variations remain. 6. Discussion [36] The above results indicate that the shear model is capable of explaining a significant portion of the systematic biases observed in paleomagnetic data obtained from APC cores, specifically those biases that cause demagnetization-resistant overprints in sediment from near the periphery of a core. The shear model is also physically appealing since visual examination of APC cores illustrates that most have some form of shear deformation, noted by the downturned sediments near the core liner. [37] Admittedly, the actual rotation of rigid particles in a deforming medium is more complicated than the simple bending analogy that we use. For example, continuum mechanic models, which have been applied to crustal block rotations in shear zones, indicate that the size of rotation of a rigid particle in the continuum depends on many factors. In particular, the rotation of a rigid particle varies as a function of the shape and orientation of a particle relative to the velocity field of the medium [Lamb, 1987]. Also, the rotation of the particle is approximately twice as large for a rigid particle that is actively being driven by bounding faults (i.e., the pinned model in which blocks are pinned at pivot points to bounding faults) rather than when a particle floats passively in the medium [McKenzie and Jackson, 1983]. The bending model that we use predicts the correct net sense of rotation caused by shear deformation, with the predicted size of rotation being similar to that of the pinned model everywhere except within a few millimeters from the core periphery (Figure 7). Near the periphery of the core, our model never predicts a rotation >90, which is likely an underestimate if the particles behave as ball bearings as shown in Figure 7. Given the complicated manner in which rigid particles may rotate in a deforming medium, our estimate of the deformation factor is probably only good to a factor of 2. [38] Similarly, the reduction in the difference between U channel and split core data that we see after correcting the split core data for shear deformation would presumably be reduced even further if the complicated rotations of magnetic particles could be described more accurately. An even larger reduction would result if the degree of deformation was allowed to vary with depth rather than be held constant as we have done. For example, the short interval analyzed in Figure 10 indicates a slightly greater average degree of deformation than does the entire interval from Cores 6H 10H. [39] Other types of coring deformation likely contribute additional biases to the paleomagnetic vector, though no other core deformation is a prevalent or consistent from core to core as the bending or shearing deformation. The fact that the relatively simple model explains a significantly portion of the disparity between U channel and split core data supports our hypothesis that shear

13 ACTON ET AL.: PALEOMAGNETIC OVERPRINTS IN OCEAN PISTON CORES EPM 3-13 deformation is at least a significant source of the demagnetizationresistant drilling overprint. [40] Throughout the analysis we have assumed the U channel directions are representative of the true paleomagnetic direction because the sediment in the U channel is collected from the central part of the core that is undisturbed or nearly undisturbed and is more distant from magnetic sources during coring. Unfortunately, we do not know the true paleomagnetic direction and so cannot preclude the possibility that even the central part of the core has some degree of drilling overprint. Another concern, though one beyond the scope of this paper, are biases that may occur from any form of paleomagnetic sampling in soft sediment. In particular, what biases in the remanence vector occur in U channel and discrete samples (7 cm 3 oriented cubes), both of which are typically collected by pushing the sample holder into the sediment. Statistically significant deflections of the remanence directions have been observed for discrete samples [e.g., Gravenor et al., 1984], though we are unaware of any study that has documented the size or mechanism for a large number of discrete samples or for any number of U channel samples. It is therefore possible that some of the differences observed between split core and U channel results are related to biases caused by U channel sampling. Given that (1) deformation caused by U channel sampling is at least visually negligible relative to that of the piston corer, (2) radial-horizontal components have been frequently observed in split core results but not in U channel results, (3) biases in inclination from one core to the next are notable in split core results [e.g., Keigwin et al., 1998, pp ] but not in U channel results, and (4) the directional differences between split core and U channel results have azimuthal dependencies that are well explained by the shear deformation model (Figure 11), it seems highly probable that most or all of the difference between split core and U channel results can be attributed to deflections of the true paleomagnetic vector in split core results. The U channel results thus represent the best estimate of the true paleomagnetic vector that we can obtain. Support for U channels giving accurate estimates of the true vector comes from the mean U channel inclination from the Brunhes age sediments analyzed at Site 1062, which differs by <0.5 from the expected inclination for a geocentric axial dipole Avoiding Core Deformation and Drilling Overprints [41] Ideally, we would prefer to collect cores without the shear deformation and without exposing the core to large magnetic fields. To achieve this would require improvements to the coring tools used, with the realization that they have been optimized to some degree already. Past experiments with piston corers, such as those by Hvorslev [1949], provide valuable insights into where improvements could be made. We suggest that attempts be made to redesign the cutting shoe by reducing its wall thickness, flattening the taper on its cutting edge further, and increasing its diameter slightly relative to the plastic core liner. These steps reduce the amount of sediment displaced by the cutting shoe, reduce excess sediment from entering the core barrel, and reduce friction on the core as it slides into the core liner, respectively. A more drastic change would be to go to larger diameter cores, which would increase the amount of undeformed sediment relative to deformed. With the extra core material it would be preferable to collect and measure U channel samples on board the drill ship instead of split core sections. Minor changes to the current design might also be useful, such as simply polishing or lubricating the inside of the cutting shoe prior to collecting each core. This would reduce friction on the sediment. Unfortunately, lubricants can have deleterious side effects on geochemical studies, so their use is unlikely to be acceptable. A final suggestion is to use nonmagnetic materials where possible, particularly for the cutting shoe and core barrel, with the goal of reducing drilling-related IRM overprints. [42] Alternatively, if improvements to the coring tools prove to be impractical, immediate steps can be taken to improve the quality of the paleomagnetic results by improving coring, sampling, and measuring strategies. Considering that the primary goal of paleomagnetists on many cruises is to obtain continuous, highresolution records of the geomagnetic field (direction and intensity), several obvious, though not inexpensive, steps could be taken. First, we would recommend that three or more holes be drilled at a site, which is a strategy used on many cruises dedicated to paleoceanographic studies. Multiple holes aid in the recovery of a complete sedimentary section at a site, allow construction of composite paleomagnetic records in which biases caused by core deformation or other sources are more likely to be noted and avoided or averaged out, and provide greater opportunity for U channel samples to be collected and measured on the drill ship. Second, collecting U channels from the center of each working half core section and then measuring them, rather than the split core sections, immediately would avoid the shear deformation and ephemeral magnetization problems while providing the high-resolution data necessary for many paleomagnetic studies Shear Deformation and Core Recovery [43] Since ODP began using the APC system, core recovery for APC-cored intervals have commonly exceeded 100%, usually by 1 3%. Part of the artificially high recovery percentages have been attributed to decompression of sediments as they are brought from below the seafloor to the surface [Farrell and Janecek, 1991; Hagelberg et al., 1995; MacKillop et al., 1995; Moran, 1997]. Additional expansion results from curation practices, in which soupy core materials commonly occurring at the top of many cores are curated as part of the core. In reality, much of the soupy material results from sediment falling into the hole or from sediment being stirred at the bottom of the hole. This happens as the roller cone bit, which is part of the bottom hole assembly (BHA), advances from the top of the core previously recovered to the top of the core that is next to be recovered. Other factors such as ship motion and associated motion of the BHA likely contribute further [Ruddiman et al., 1987; Robinson, 1990]. We suggest that shear deformation is another factor that adds to the excess core recovery because the deformation may reflect the entrance of excess sediment (sediment displaced by the cutting shoe), which increases friction that causes the convex bending of sediment as the sediment slides into the corer. 7. Conclusions [44] Sources for overprints are many and complex in paleomagnetic studies of ocean drill cores. We have focused on one form of overprinting that appears to be prevalent in sediment cores collected with hydraulic piston corers. Our results indicate that sediment shear originating from friction on the inside wall of the piston corer produces a statistically significant portion of observed demagnetization-resistant drilling overprints in split core paleomagnetic data. Through an iterative search that minimizes differences in the paleomagnetic directions measured on split cores and U channel samples from Hole 1062E, we find a best fit estimate for the mean degree of deformation of 0.07 ± This degree of deformation corresponds to downward bending of sediment near the periphery of the core that is comparable to that commonly observed, though the value should only be consider representative of the sediments analyzed in this study. [45] Should the mean degree of deformation be found to vary little from one sedimentary environment to another, it would be possible to remove the mean bias caused by shear deformation through the correction procedure we have developed. Complete removal of shear-related overprints will likely be difficult because of variations in the amount of shear within and between cores.