The Earth’s magnetic field is a vector which can be recorded by volcanic rocks, archeological material and sediments. Compared with the former two candidates, sediments have the advantages of continuity and broad distribution. The geomagnetic field is considered to originate from outer-core geostrophic flow^[1], and thus has globally consistent evolution.

Due to the global consistency of geomagnetic field variation, it provides another possibility, i.e. dating sediments. In the circum-Pacific of low to middle latitudes, benthic and planktonic calcareous foraminifera are usually used for the oxygen isotope stratigraphy and/or radiometric dating in studies of paleoceanography and paleoenvironment. However, in the Arctic and subarctic areas, the scarcity of foraminifera inhibits the application of these techniques. Organic carbon dating of bulk samples and paleontological percentage correlation are the common means to obtain chronology in Bering Sea sediments^{[2, 3, 4]}. While magnetic stratigraphy in the Bering Sea is rarely reported.

Recently, Barletta et al. (2008)^{[5, 6]} reported relative paleointensity and direction in the Arctic Chukchi Sea and Beaufort Sea that are comparable with those of lake and volcanic flows in North America. These geomagnetic secular variations are exploited for construction of the age model of Arctic sediment. Okada et al. (2005)^[7] presented the first relative paleointensity result for dating five cores from the Bowers Ridge and Northeast Slope in the Bering Sea which are used to build age models by correlating relative paleointensity with global intensity stack GLOPIS75^[8] and SINT800^[9].

In this work, we attempt to obtain PSV of the geomagnetic field, i.e. relative paleointensity and direction in using the core from a borehole on the northeast slope of the Bering Sea. Because submarine canyons and turbidity are very common in the Bering Sea slope area, we first examine the anisotropy of magnetic susceptibility (AMS) to detect the possible natural and artificial disturbance, and then analyze rock magnetic properties for the degree of magnetic uniformity. On the basis of these analyses, the relative paleointensity is obtained by two kinds of methods, normalization and pseudo-Thellier methods. A discussion about intensity and direction of geomagnetic field on the regional and global scale is followed to discuss the evolution of PSV since the last deglaciation.

The core sample B5-4 was retrieved in the First Arctic Expedition of China in 1999 at the foot of Northeastern slope in Bering Sea (58°05'16''N, 176°31'19''W), which is within a branch of Zhemchug submarine canyon with water depth of 3370 m and total length of 4.70 m. The topography of the whole slope is characterized by flat sections shallower than 1000 m and deeper than 3000 m, and steep section between 1000~3000 m where several huge canyons are present (Fig.1). The surface water above this core site is covered by Bering Sea Slope Current and therefore an area of high productivity called the green belt^[10].

(Fig.1)

Fig.1 Geographic location of borehole B5-4 and other related boreholes B2-9, B4-2 see Ref.[4], GAT-3A is pre-cruise borehole of Bering Sea expedition[7], 02023JPC in Ref.[11]. Contour of 200 m in the figure refers to the transition from shelf to slope. The inset shows the location of study area within Bering Sea and the open/close state of Bering strait when sea level is at 40 m by shaded area.

The lithology in core B5-4 is rather uniform, dark gray to gray green clayey silt rich in opal (smaller than 30%). Gray or light gray volcanic debris are common in the sediments that are usually silt and coarse sand. At 0.31~0.33 m and 4.48~4.70 m, a fine sand layer is visible and described as the tephra layer.

Cubic plastic boxes are pushed into the working half section at an interval of 4 cm, the arrow on the bottom of the box points downward. No paleomagnetic samples are available at 1.68~1.93 m since sediments are disturbed, and a total of 106 samples are obtained. Prior to any treatment, anisotropy of magnetic susceptibility (AMS) is measured on Kappabridge3s susceptibility meter for all samples. Alternate demagnetization and remanence measurement are undertaken on 2G superconducting magnetometer with step of 5 mT before 50 mT and 10 mT after till 80 mT or 100 mT depending on sample types. Anhysteresis remanece is measured in the 80 mT/0.05 mT alternate and direct field and then the same demagnetization steps as natural remanence demagnetization are used. The measurement procedure of S ratio is to add a 0.1 T and 0.3 T direct field first and then 1 T strong field along the same direction. The saturation remanence of 1 T for the whole sample exceeds the upper limit of superconducting magnetometer 2×10^-4 Am², so powder samples are used for measuring saturation remanence and corrected to saturation remanence for the whole sample by multiplying the weight ratio of whole and powder sample. The assumption of this method is that 1 T is strong enough for saturating both oriented and powder samples so the remanence is only related with weight. The formula of S ratio is S_0.1T = IRM_0.1T/SIRM_1T, S_0.3T = IRM_0.3T/SIRM_1T. A small amount of powder samples are dried naturally and grinded to powder and magnetic susceptibility changes with heating and cooling cycles are monitored on the Kappabridge3s susceptibility meter from room temperature to 700°C.

Normalization and pseudo-Thellier methods are used for the relative paleointensity. Three normalization parameters include magnetic susceptibility (κ), anhysteresis remanence (ARM) and saturation remanence (SIRM). The pseudo-Thellier method is to calculate the linear slope of demagnetization of NRM and ARM in the same coercivity interval to represent the relative paleointensity.

AMS of oriented samples show that most inclinations of the short axis (90/105) are steeper than 60° and those of long (98/105) and medium axis (104/105) shallower than 30° and no dominant directions for the latter two (Fig.2a). Corrected anisotropy degree (P') is linearly related with foliation (F) (R² = 0.92, n = 105), but not with lineation (L) (R² = 0.01, n = 105) (Fig.2b). The distribution plot of lineation and foliation indicates magnetic ellipsoid is mainly oblate (Fig.2c), i.e. the typical original sedimentary fabric. Both the orientation of three principal axes and the relationship between anisotropy parameters imply that magnetic grains are naturally aligned within water column and compacted by gravity. Except stronger water current revealed by 14% short axes departing from vertical position over 30°, relatively quiet water and no obvious disturbance of erosion or turbidity are hinted by most other samples. Under the coordinate of core splitting (facing upper surface of the core, downward push of box is -X, right is +Y , left is -Y , and downward along the core is Z+), declination of long and short axes do not orient along Y and X axis (Figs. 2d and 2e), therefore AMS observed here is not artificially induced by splitting process^[12].

(Fig.2)

Fig.2 AMS uncorrected to absolute direction in B5-4 (a) Projection of long (square), medium (triangle) and short (circle) axis to lower hemisphere; (b) Relationship between corrected anisotropy degree and lineation (solid) and foliation (open); (c) Flinn plot of lineation and foliation, oblate predominate; (d) Declination of short axis (D - Kmin) and (e) Declination of long axis (D - Kmax) in the coordinates of core splitting. Two inclinations of short axis close to horizontal is sample 28 and 88, correspond to depth of 1.14~ 1.16 m and 3.90~3.92 m where no visible disturbance is observed.

The κ - T curve for the bulk sample reveals Curie point close to 600°C during heating/cooling processes, obviously magnetite (Fig.3a). No other Curie points other than magnetite’s are observed, indicative of possible absence of iron sulphide. The heating and cooling curves almost overlap during 600~700°C implying minor contribution from hematite. κ after heating to 700°C and cooling back to room temperature is 3.5 times of original κ, and the increase on the cooling curve occurs at 580°C, which shows transformation from iron-bearing clay to magnetite. S_0.1T and S_0.3T are larger than 0.7 and 0.9, respectively (Figs. 4(d, e)), the evidence of low coercivity with magnetite dominance. At the same time, middle destructive field of remanence (MDF) is usually lower than 40 mT (Fig.4f), also evidence of low coercivity of minerals carrying remanence.

(Fig.3)

Fig.3 Rock magnetic property during heating and at room temperature in core B5-4 (a) κ - T of sample B5-4-96 (depth 4.26~4.28 m, in air). Thick/thin line represent heating/cooling; (b) Relationship between κARM and κ. Two lines indicate grain size of 0.1 μm and 1.0 μm according to King et al. (1983)’s model. Note the samples of 0~0.44 m(solid circle) have finest grain size, and gradually become coarse downward. The samples from the bottom of the core (triangle, 4.48~4.70 m) have a little coarser magnetic grain size and increased concentration. The sample with highest κ in Fig.3b is #102(4.50~4.52 m) which is close to the tephra (4.48~4.50 m); (c) Relationship of ARM and SIRM. Except the upper part (0~0.44 m), the whole core has constant ARM/SIRM, implying uniform magnetic size. #102 (triangle with highest ARM) has relatively low SIRM which should be the abnormity. The ratio of two remanences avoid paramagnetic and superparamagnetic influence.

(Fig.4)

Fig.4 Rock magnetic properties in borehole B5-4 (a) κ; (b) ARM; (c) SIRM; (d) S0.1T; (e) S0.3T; (f) MDF.

Anhysteresis susceptibility (κ_ARM) is calculated by normalizing ARM with the direct field. When magnetic carrier is magnetite, the relationship between κ_ARM and κ is useful to judge the magnetic grain size and concentration: the larger κ_ARM/κ is, the finer the magnetic grain size, and also the farther from origin, the higher the concentration of magnetic minerals. Except 0~0.44 m, other samples have an almost linear distribution of κ_ARM and κ between 0.1 μm and 1.0 μm, single domain range suggested by King et al.^[26] (Fig.3b). Because the κ of this core is low and so the value might be influenced by paramagnetic and superparamagnetic minerals, we further apply two remanence parameters to evaluate magnetic grain size^[14]. In the figure of two remanences, a similar uniform magnetic grain size is indicated by linear relationship of κ_ARM and κ except the upper 0~0.44m and tephra (Fig.3c).

In Fig.4, three concentration-dependent parameters, κ, ARM and SIRM, vary mainly at the upper part 0~0.44 m and/or the bottom 4.46~4.68 m with increases of 2~3, 4~5 and 2~3 times, respectively, and are rather constant at other intervals (Fig.4(a-c)). Two remanences-ARM and SIRM have peaks above 0.44 m and below 4.5 m, while κ only peaks at the bottom. Compared with SIRM, amplitude of ARM peak is large possibly due to the fine grain size because ARM increases drastically from 0.06 to 0.02 μm (single domain range) than from 100 to 1 μm (pseudo single domain and multi-domain)^[14]. Absence of κ peak at the upper part may be caused by increased water content. At the bottom of the core below 4.5 m, synchronous increase of remanence and κ suggests higher input of magnetite, not only in the tephra (4.48~4.50 m).

Vector projection of alternate field demagnetization (AFD) shows that the direction of remanence is stable and clearly towards the origin (Fig.5). Decay curves of normalized remanence reveal smaller MDF than 40 mT. MDF of four samples near the surface of the core is 44~46 mT (Fig.4f). The fine grain size revealed by κ_ARM and κ and the increasing MDF in the same samples imply the cause and effect relationship. Most remanences decline to 10% of original values after 80 mT AFD and only small amount to 20% (Fig.5). 100 mT AFD fails to reduce the remanence further and the direction remains the same (Fig.5a B5-4-5). The low MDF suggests the validity of the AFD method to retrieve characteristic remanence.

(Fig.5)

Fig.5 Orthogonal vector projection and remanence decay curve in core B5-4 (a) Typical samples; (b-e) Samples from the first to the fourth sections of shallow inclination Samples without labeled unit have unit of 10-8 Am2.

Relative declination of remanence circles around 0° except 0~0.28 m (Fig.6a). Most inclinations vary around the expected axial dipole latitude (72.7°), and no negative inclination is found. However, inclinations smaller than 60° are observed at 0.58~0.72 m, 2.54~2.62 m, 3.22~3.60 m and 4.42~4.68 m (Fig.6b). The four sections of shallow inclinations have similar clear and stable directions on the vector projection plot (Figs. 5(b-e)). The stability and single directions from principal component analyses suggest the origin of the geomagnetic field. The orthogonal vector projection also shows viscous remanence is removed before 20 mT so NRM_20mT is selected as the characteristic remanence to be normalized in the relative paleointensity.

(Fig.6)

Fig.6 Paleomagnetic result in core B5-4 and the correlation with other records (a) Relative declination; (b) Inclination and error. These are results from principal component analysis; (c) Relative paleointensity by pseudo-Thellier method; (d) NRM20mT/SIRM; (e) NRM20mT/κ; (f) (NRM/ARM)20mT; (g) Relative paleointensity NRM/SIRM from ODP983[17]. The link line between f and g is tie points, the upper number is depth in m, lower is age in ka. Arrow is AMS14C dating; (h) Absolute intensity of Holocene[18]; (I) RPI global stack SINT 200[19].

In sum, the change range of magnetic concentration is within 5 times, and main carrier of magnetic signal is magnetite, the magnetic grain size is stable single domain of 0.1~1 μm except a small part smaller than 0.1 μm. AMS presents normal sedimentary fabric with nearly horizontal long and medium axes and nearly vertical short axis. S ratio and MDF vary to some degree at the interval of 3.5~4.5 m, but magnetic properties are rather uniform. The remanence directions are also consistent with the axial dipolar model. Therefore these sediments are suitable to retrieve paleomagnetic secular variation, i.e. direction and relative paleointensity of the geomagnetic field. In the following part, we apply two kinds of methods for the relative paleointensity estimation-normalization and pseudo-Thellier methods and principal component analysis of Kirschvink^[15] for direction analysis.

The conventional normalization procedure for relative paleointensity is to normalize characteristic remanence with concentration parameters to eliminate effects of various concentration, grain size and mineralogy (Figs. 6(d-f)). The pseudo-Thellier method is based on that of Tauxe (1995)^[16], in which the slope and error of NRM and ARM demagnetization at the same coercivity interval are calculated by linear fit (Fig.6c). The results from the two methods basically agree with each other, both show a general trend of increasing upward (Figs. 6(c-f)).

An absolute age is required for establishment of relative paleointensity. Only one AMS¹⁴C dating is available from cold species N. pachyderma(sin) at 4.54~4.56 m. The age is 12.250±0.05 ka (Woods Hole) and 13.317~13.435 cal. ka(1σ) after correction using calib 601, so the middle age of 13.376 ka is taken as the final calendar age. Carbon reservoir in Bering Sea is 700 a (△R = 300)^[11]. Isotope dating from U-Th shows no excess ²³⁰Th in the sediments (²³⁰Th_ex=²³⁰Th_(dpm/g)-²³⁴U_(dpm/g)), a sign of quick accumulation after deglaciation (Chen Zhihua’s unpublished data).

Based on the AMS ¹⁴C dating result, the relative paleointensity from B5-4 is compared with North Atlantic core ODP 983^[17], absolute intensity of Holocene synthetic database^[18] and global stack from marine sediment SINT 200^[19] (Figs. 6(g-I)). The consistency between our result and ODP 983 is impressively high and an increasing trend in the latter two for the last 10 ka and 16 ka is also consistent with our result. The comparison with ODP 983 produces three tie points(Figs. 6(f, g)), and other ages are linearly interpolated and extrapolated. Nearly linear sedimentation rates indicate this comparison is reasonable and the average rate is 35 cm/ka for the whole core(Fig.7).

(Fig.7)

Fig.7 Age model of core B5-4 from RPI

Absolute intensity of global average for the last 12 ka has the highest values in 1~3 ka B.P., (10.5 ~ 11.5) × 10²² Am², and medium values larger than 8×10²² Am² in 8~10 ka B.P.^[18] (Fig.8g), with resolution 500~1000 a. The record curve from sediment stack SINT 200 monotonically increases for the last 14 ka except a local low at 6 ka (Fig.8g gray). Absolute intensity database GEOMAGIA50 includes 8000 data within 50 ka^{[20, 21]} (web data renewed till 2011.11.23). These data are from archeological material such as baked clay, ceramic, tile and volcanic rocks. Based on these data, we calculate the spherical harmonic model at the site of the core using the Korte & Constable (2005)’s model CALS3k, CALS7k and CALS10k^[22] whose variations mimic those of Yang et al.(2000)^[18] (Fig.8f). The model shows more variations than absolute intensity and SINT 200, e.g. the local peak at 4~6 ka B.P. which is the same as SINT 200 (Fig.8g).

(Fig.8)

Fig.8 Relative paleointensity on age in core B5-4 and synthetic comparison with other RPI and absolute intensity of geomagnetic field (a) RPI in ODP 983[17]; (b) RPI (NRM/ARM)20mT in core B5-4, half-filled diamond is RPI tie points, filled diamond is AMS14C dating; (c) NRM/κ; (d) NRM/SIRM; (e) RPI from pseudo-Thellier: slope of NRM and ARM demagnetization; (f) Spheric harmonic model at Being Sea (58.09°N, 183.48°E) for the last 3 ka, 7 ka, 10 ka[20, 21]; (g) Absolute paleointensity (dark line and right axis) and RPI and RPI global stack SINT 200[19] (gray line and left axis); (h) RPI global stack GLOPIS75[8]. http://geomagia.ucsd.edu/geomagia/query.php

Records of longer time span and higher resolution come from GLOPIS75^[8] and North Atlantic ODP 983^[17]. The former includes 24 cores from the Atlantic Ocean, Mediterranean Sea, and Indian Ocean, where sedimentation rates are from 7 to 34 cm/ka (ODP 983 included). The main characteristic of GLOPIS75 is the appearance of three peaks at 1~3 ka, 8~10 ka and 12~14 ka which decline downward sequentially, noticing their drastic consistency with Holocene absolute intensity (Figs. 8(g, h)). The sedimentation rate in ODP 983 is highest for the last 14 ka, 130 a before 12 ka BP and 30~40 a after 12 ka B.P.. So ODP 983 shows higher frequency of changes than GLOPIS75 and bears more similarity with our result. GLOPIS75 is the stack of records around the globe, while ODP 983 reflects a character of the subarctic area in the North Hemisphere. This can partially explain respective similarity between GLOPIS75 and Holocene intensity (Figs. 8(g, h)), ODP 983 and the borehole of this work and also the small difference between records from single core and global average.

According to the dating and comparison of paleointensity, the single sample in B5-4 sediments represents ~120 a, so resolution is close to ODP983. The big peak at 1~3 ka on absolute record sprays into four (Figs. 8(b-d)) or five small peaks (Fig.8e), and the shape and peak-trough distribution correspond one to that of ODP 983. Low intensities at 3~8 ka have low amplitude and high frequency (Figs. 8(b-e)), which makes comparison with ODP 983 difficult and no more tie points produced. A declining trend is visible from 14 ka to 8 ka but no obvious tie points. Although there are three peaks at 12~10 ka B.P. in B5-4, the uncontinuity in ODP 983 inhibits further contrast. In spite of the scarcity of more tie points, the consistency between our result and the global stack, spherical harmonic model, especially with ODP983 is conspicuous, which proves that the relative paleointensity of this work is truly the refection of geomagnetic behavior rather than lithological characters, and the persistent magnetic flux at the core-mantle boundary proposed by Kort & Holem (2010)^[23].

Since the amplitude of paleointensity for the last 14 ka is small, especially at 4~10 ka B.P., so the high resolution records do not facilitate more tie points (Figs. 8(a, b)). We therefore turn to direction of the geomagnetic field, hoping to have more tie points. The direction reports of the geomagnetic field are mainly from lake deposits, including the Lake Baikal^[24], lakes in North American including East Canada and east US^{[25, 26]}, North Europe^{[27, 28]}, and England^[29] (Fig.9).

(Fig.9)

Fig.9 Relative declination and inclination in core B5-4 on age and comparison with other records from North Hemisphere Left/right panel is relative declination/inclination with (a) spherical harmonic model at 58.09°N183.48°E according to Korte et al. (2009) (CALS3k)[30] and Korte & Constable (2005) (CALS7k)[22]; (b) B5-4; (c) Lake Baikal[24]; (d) St Lawrence Estuary in east Canada[25]; (e) Lakes in east America[26]; (f) Lake Nautaj¨arvi in Finland[27]; (g) Varve lake in Sweden[28]; (h )England lake[29].

Declination in B5-4 resembles the most with Lake Baikal, then with Saint Lawrence Estuary cores. Totally six tie points divide declination changes into seven periods (first to seventh downward) that can be compared peak to peak with the Lake Baikal (Figs. 9(b, c)). A large peak during 8.8~9.3 ka BP has no counterpart in other records and only a medium peak in the Finland record (Figs. 9(b, f)). Obviously, the degree of resemblance in declination variation is correlated with distance between cores. The degree of similarity between the Bering Sea, Lake Baikal and East Canada is high (Figs. 9(b-e)) and they differ with European record in the fifth period (Figs. 9(f-h)). The age results from the Lake Baikal and East Canada are uncorrected ¹⁴C and the reservoir corrected calendar age, respectively, so the latter ages have been adopted for our correlation. Problem arises with the sixth tie point because the East Canada record is not long enough. In the Lake Baikal, this tie point has the age of 10.69 ka, from which we subtract the age difference of the fifth tie points of uncorrected Lake Baikal and East Canada calendar ages (9.53~9.457 ka). Result of the sixth tie point is 10.617 ka calendar age.

Inclinations in core B5-4 fluctuate highly frequent between 50 to 80°. Similar to the declination, the variation pattern in inclination of our result mimics those of Lake Baikal (Figs. 9(b, c)). At a millennial scale, characteristic lows in inclination of B5-4 can also be found in North America and Europe records (Figs. 9(b-h)). Four tie points from inclination (Fig.9 dash line) do not contradict with those from declination (Fig.9 solid line). Tie points from directions differ slightly from intensity linear fit with age difference smaller than 250 a and AMS¹⁴C dating and last inclination tie point all see a transition of sedimentation rates towards higher values around 12 ka (Fig.10a). On the synthetic age model, relative paleointensity of B5-4 fits better with ODP 983 (Fig.10b), and inclination and declination vary more consistent with that of East Canada record (Figs. 10(c, d)). So the combination of intensity and direction of the geomagnetic field can provide more tie points and a more accurate age model.

(Fig.10)

Fig.10 The synthetic age model from both intensity and direction and the re-comparison with other records (a) Age model, solid, square and triangle is tie points from intensity, declination and inclination, the scale in the figure is 250 a; (b) Relative paleointensity on synthetic age model fit with ODP 983; (c) Declination; (d) Inclination with East Canada record.

Since the high sedimentation rates and limited core length, the time span of sediments at high latitudes are usually confined to Holocene^{[27, 31, 32, 33]}, and the inclination report beyond 10 ka is even rare. The report from Lake Baikal reaches 84 ka but the ages are uncorrected and lake sediments are accessible to old carbon from land and re-deposition of autochthonous algal^[24], resulting in large difference with our calendar age. For comparison we cite our results previously reported from the Eastern China Sea, Yellow Sea and Japan Sea to discuss the two sections of shallow inclination in the core of this work (Fig.11). Two sections of shallow inclination are within 9~14 ka(smaller than 60°), 9.023~9.94 ka and 12.87~14.14 ka (Fig.11b), respectively. This event is embodied in two cores (EY02-1, NHH01) in the East China Sea by complete reversal (Fig.11(a, c)) and several jumps between normal and reversal polarity during 8.5~17 cal ka, which is regarded as Gothenburg event^{[34, 35]}. The above age differences are caused by the sparse dating and interpolation. According to the direction instability and time range, the discontinuous shallow inclinations during 9~14 ka in core B5-4 probably are the same event (Fig.11b).

(Fig.11)

Fig.11 Comparison of polarity excursion in early Holocen (a) Inclination and error from East China Sea core EY02-1[34]; (b) Inclination and error of B5-4 on new age; (c) Inclination and error in Yellow Sea core NHH01[35]; (d) Inclination and error in Japan Sea core KCES[39]. (a) and (c) carbon reservoir unified to △R = -139 ± 59 a, arrow indicate AMS14C dating, upper is depth in m and lower is calendar age in year

The drift of the geomagnetic polar around 10 ka is generally considered a Gothenburg event, and was discovered on the Beijing profiles^{[36, 37]}, Okinawa Trough^[38], and East China Sea^{[34, 35]}. The reported ages vary from 14.230~13.690 ka B.P. (traditional radiometric age)^{[36, 37]}, 12.911~11.953 ka B.P. (calendar age)^[38] to 17.047~10.686 ka B.P. (calendar age)^[34] and 9.444~7.978 ka (calendar age)^[35]. The latter two ages resulted from interpolation and extrapolation of AMS¹⁴C dating results (Fig.11(a, c)), so not accurate. In addition, uncertainty of carbon reservoir is another reason for the age uncertainty. Despite of this, the records of unsTable polarity around 10~14 ka are accumulating in numbers that advocates the argument of geomagnetic field origin.

The inclination consistency between B5-4 and Yellow Sea core (resolution is ~45 a) is obvious after the Gothenburg event, and five peaks (I_P1-5) and four troughs (I_T1-4) are identified among which I_T3 presents as a complete reversal in the Yellow Sea at ~3 ka with a time span of 100 a (Figs. 11(b, c)). However a core of the Japan Sea with similar resolution has no identifiable shallow and negative inclinations (Fig.11d). Absence of real reversal in B5-4 might be linked with early diagenesis, e.g., the finer magnetic grain size and related S ratio change on the upper part probably are signs of early diagenesis (Fig.3-4). As to the Japan sea core, the signal of early diagenesis is much more stronger^[39]. Also smoothing effect could be another factor for the smearing out of short negative imprint^[40]. We note that although there is early diagenesis of different degrees and smoothed inclination in the Bering Sea and Japan Sea, the relative paleointensity still preserves comparable characteristics^[41]. This may suggest the influence of early diagnesis on direction is larger than on intensity. Sediments with high deposition rates and free from early diagenesis and smoothing could reveal millennial or even centennial characters of declination and inclination. Consequently, the combination of relative paleointensity and direction can serve as tools to date sediments.

(1) Anisotropy of magnetic susceptibility in core B5-4 on the continental slope of the Bering Sea has nearly horizontal long and medium axes, short axis nearly vertical, suggesting undisturbed and original sedimentary structure.

(2) S ratios, median demagnetization field indicate that magnetite of low coercivity are dominant magnetic carriers. Except 0~0.44 m in the suface, magnetic grain size is stable single domain within 0.1~1 μm, and concentration range by less than 5 times. The magnetic uniformity is verified to record relative paleointensity.

(3) Relative paleointensity from normalization and pseudo-Thellier method generally increased in the last 14 ka and peaked at 1~3 ka, 8~10 ka and 12~14 ka, respectively. The common features from this work, absolute intensity, global stack GLOPIS 75 and ODP 983, have millennial consistency and possibly centennial comparability in ODP 983.

(4) Declination and inclination in B5-4 are comparable well with those of the Lake Baikal and North America which provide extra tie points in addition to relative paleointensity.

(5) Inclination in core B5-4 has six peaks or troughs for the last 14 ka and the shallow sections between 9 and 14 ka possibly are Gothenburg events by analog with other records. And the shallow inclinations after 10 ka BP in core B5-4 could find counterparts in the Yellow sea core.

The core used in this study was recovered in the First Arctic Expedition of China in 1999 financed by the Ministry of Finance of China and organized by Chinese Arctic and Antarctic Administration. We thank the staff of “Snow Dragon” vessel and scientists of Geology Group for sampling the core. We thank J E T Channell in University of Florida for sharing data of core ODP 983 and Yang Xiaoqiang in Sun Yat-Sen University for providing the direction data. Han Yibing, Lu Yao, Wang Kunshan and Wu Yonghua participated the sampling in the lab. This work was supported by National Key Basic Research Program of China (2013CB429704) and Natural Science Foundation Committee Program (41376072, 40876036, 41076136, 41106166), Key International Cooperation Program (40710069004), Arctic and Antarctic Joint Investigation Project (CHINARE2013-03-02, CHINARE2013-04-01), and Program from Key laboratory of Marine Hydrocarbon Resources and Environmental Geology, Ministry of Land and Resources (MRE201211).