Paleomagnetic studies on single crystals separated from the middle Cretaceous Iritono granite

Sample description

We studied zircon, quartz and plagioclase separated from the middle Cretaceous Iritono granite in the Abukuma massif, northeast Japan (Fig. 1). Cooling history of the Iritono granite is constrained by Wakabayashi et al. (2006) and Tsunakawa et al. (2009) using a thermal diffusion model of the granite body, and by two radiometric age determinations with different closure temperatures. The U–Pb zircon age is 115.7 ± 1.9 Ma (Tsunakawa et al. 2009) and the ⁴⁰Ar–³⁹Ar biotite age is 101.9 ± 0.2 Ma (Wakabayashi et al. 2006). The age of the Iritono granite corresponds to the middle part of the CNS, in which the polarity reversals had been stopped for a period as long as 20 Myr. The estimated cooling time for a lock-in of a paleomagnetic record is 4 × 10⁴ to 1.4 × 10⁷ years. Wakabayashi et al. (2006) and Tsunakawa et al. (2009) previously conducted rock-magnetic and paleomagnetic studies and paleointensity experiments on the whole-rock samples of the Iritono granite. The magnetic minerals in the Iritono granite were magnetite and pyrrhotite, and their fraction varied with sampling locations. Samples from site ITG09 showed the least contribution of pyrrhotite, and the primary magnetization was clearly distinguished from the secondary magnetization carried by low blocking temperature or low coercivity components in terms of natural remanent magnetization (NRM) direction. Tsunakawa et al. (2009) studied the paleointensity by both the Coe’s version of Thellier method (Thellier and Thellier 1959; Coe 1967) and the Tsunakawa–Shaw method using the whole-rock sample of site ITG09. Although the obtained paleointensities exhibit a bimodal distribution according to the different methods used, they were indistinguishable at the 2σ level and thus were combined into one site-mean. The resultant site-mean paleointensity was 58.4 ± 7.3 μT before applying the cooling rate correction, and 39.0 ± 4.9 μT after correction. This corresponds to a virtual dipole moment (VDM) of 9.1 ± 1.1 × 10²² Am². The present study used the mineral samples separated from the core samples of site ITG09.

Sample preparation

A granite sample core 2.54 cm in diameter was crushed with a non-magnetic mortar and pestle, and sorted by 850 μm and 350 μm mesh screens. Heavy fractions of the sample smaller than 350 μm were concentrated by an aqueous panning technique. Zircons with no visible cracks or opaque particles on the surface were hand-picked under a binocular stereoscopic microscope. Quartz and plagioclase were hand-picked from samples larger than 350 μm and smaller than 850 μm. These selected crystals were leached by hydrochloric acid (HCl) to remove any tiny magnetic particles on the sample surface. HCl concentration and leaching duration was 12 N and 4 days for zircon and quartz, and 6 N and 8 h for plagioclase, respectively. Samples were then sandwiched individually between layers of magnetically clean Scotch Magic Transparent Tape in the method of Sato et al. (2015) or were mounted individually on a glass holder (see below) for rock-magnetic measurements and paleointensity experiments.

Remanence measurements using SQUID magnetometer

A superconducting quantum interference device (SQUID) magnetometer (2G Enterprises Model 755-4.2 cm) was used for remanence measurements. We followed the method of single-crystal measurements by Sato et al. (2015). A sample holder made of acrylonitrile butadiene styrene (ABS) was used for measurements. Single-crystal samples sandwiched by tape or mounted on the glass holder were fixed on the edge of the ABS holder by double-stick tape. The magnetic moments of the ABS holder and double-stick tape were measured before and after sample measurement and subtracted from the sample moment. The detection limit of the method was 2 × 10⁻¹² Am², so we employed 4 × 10⁻¹² Am² as a threshold to distinguish significant remanence intensity from noise.

For stepwise thermal demagnetization (ThD) and paleointensity experiments, we made new thermally resistant holders for single-crystal measurement of the SQUID magnetometer based on the sample holder designed for SQUID microscope measurements (Fu et al. 2017). Images of the sample holder are shown in Fig. 2. Non-alkali high-temperature glass plates (Eagle XG, Corning, 1.1 mm thick) were cut into squares of 7 mm on a side. A 1-mm-diameter pit was drilled in the center of the glass plate, followed by intense cleaning in concentrated HCl. A single-crystal sample was put into the pit and fixed by stuffing SiO₂ powder with grain size of ~ 0.8 μm. This technique enabled us to conduct heating experiments on single crystals in a fixed sample coordinate. The blank magnetic moment of the glass holder after subtracting the moment of ABS holder and double-stick tape was well below the practical detection limit of the SQUID magnetometer.

First, we measured NRM intensity of 349, 455, and 268 grains for zircon, quartz, and plagioclase, respectively. On the basis of the NRM intensities, we then selected samples for further rock-magnetic and paleomagnetic measurements.

Rock-magnetic measurements

For the selected samples that showed significant NRM intensity (> 4 × 10⁻¹² Am² per grain), we conducted low-temperature remanence measurements using a magnetic property measurement system (Quantum Design model MPMS-XL5). Isothermal remanent magnetization (IRM) was first imparted at 2.5 T and 10 K after zero-field cooling from 300 K. The remanence was then measured during warming in zero-field (ZFC remanence). Subsequently, samples were cooled to 10 K in a 2.5 T field and then remanence was further measured during warming in zero-field (FC remanence).

Hysteresis loop measurements were taken for plagioclase grains and a quartz grain which contained magnetite using an alternating gradient magnetometer (LakeShore model MicroMag 2900). Samples sandwiched by tape were mounted on a transducer probe with a silica sample stage (Lake Shore model P1 probe). The blank saturation magnetization of the probe was 6 × 10⁻¹⁰ Am². Maximum field during hysteresis loop measurement was 0.5 T, and the field increment was 4 mT. Diamagnetic/paramagnetic corrections were applied to the obtained hysteresis loop by subtracting the average slopes at applied field of |B| > 300 mT. Results are exhibited in the Day plot (Day et al. 1977).

Stepwise ThD of NRM was performed on selected zircon and quartz samples using a TDS-1 thermal demagnetizer (Natsuhara Giken). For plagioclase samples, stepwise ThD of laboratory-imparted thermoremanent magnetization (TRM) was performed after paleointensity experiments. TRM was given by cooling from 610 °C in a 50 μT field in air.

To investigate the NRM to IRM (NRM/IRM) distribution of plagioclase, a room-temperature IRM was imparted to 75 plagioclase samples at 2 T by an MMPM10 pulse magnetizer (Magnetic Measurements), and the IRM intensity was measured using the SQUID magnetometer.

Paleointensity experiments

Remanence anisotropy measurements

To assess anisotropy effect, we measured the ARM anisotropy of 19 plagioclase samples including 13 samples which were subjected to the paleointensity experiments. ARM was imparted along three orthogonal axes (ARM_x, ARM_y, and ARM_z) to obtain the remanence anisotropy tensor. Measurements were taken after LTD treatments. TRM anisotropy tensor was also measured after paleointensity experiments for some samples and the consistency with the ARM anisotropy tensor was checked. The ARM and TRM measurements were also taken after AFD with a peak AC field of 50 mT. ARM anisotropy of whole-rock samples was checked based on measurement results obtained using a spinner magnetometer (Natsuhara Giken model SMD88).

Rock-magnetic properties of zircon

Sixteen out of 349 zircon samples had NRM intensities larger than the threshold (Fig. 3a). Low-temperature magnetometry and stepwise ThD measurements of NRM were taken on selected samples that had significant NRM intensity. Representative results are summarized in Fig. 4. Stepwise ThD treatment for NRM was performed on four samples. Two samples showed a characteristic magnetization component and pyrrhotite-like blocking temperature (Fig. 4a), but the other two did not show any stable remanence component (Fig. 4b). Low-temperature experiments were performed on additional five samples. One sample showed a phase transition of pyrrhotite at ~ 30 K (Fig. 4c), while the other four samples did not show any obvious transition (Fig. 4d). We concluded that the dominant magnetic inclusion in zircon is pyrrhotite and/or magnetically very soft materials. Since the whole-rock study determined that a low blocking temperature (< 400 °C) component is probably carried by pyrrhotite and hence was most likely remagnetized by a reheating event (Wakabayashi et al. 2006), we did not use zircon for paleointensity experiments.

Rock-magnetic properties of quartz

Similar to the zircon samples, very few samples of quartz (7 out of 455) had NRM intensities larger than the threshold (Fig. 3b). We took stepwise ThD measurements of NRM on two quartz grains. In both samples, magnetization decreased generally toward the origin in the orthogonal plot. One sample shows a high blocking temperature suggesting magnetite inclusion (Fig. 5a) and the other sample exhibits a lower, pyrrhotite-like blocking temperature (Fig. 5b). We further took low-temperature magnetometry measurements and hysteresis loop measurements on the former sample. The Verwey transition of magnetite was recognized near 120 K (Fig. 5c) that indicate titanium-poor magnetite (Özdemir et al. 1993; Moskowitz et al. 1998). High coercivity (B_c> 10 mT) exhibited in the slope-corrected hysteresis loop suggests the existence of fine-grained magnetite. We concluded that quartz is a potentially ideal sample for paleomagnetic study. However, we decided not to use quartz for paleointensity experiments because it was difficult to find enough magnetite bearing samples to study.

Rock-magnetic properties of plagioclase

A histogram of NRM intensity for the plagioclase crystals is shown in Fig. 3c. In contrast to the zircon and quartz samples, very high population of the plagioclase samples (224 out of 268; 84%) exhibited the significant NRM intensities. Figure 6a shows a diagram between NRM intensities and IRM intensities for the 75 plagioclase grains. NRM/IRM distributes in a narrow range, namely around 0.1 (Fig. 6b), indicating the similar magnetic carrier and NRM origin among the plagioclase grains. The NRM/IRM ratio around 0.1 is higher than the TRM(50 μT)/IRM ratio for synthetic samples which resemble rocks (Yu 2010), and consistent with, but somewhat lower than the previous reports on plagioclase crystals (Usui et al. 2015) or rocks containing exsolved magnetite (Selkin et al. 2007).

We performed magnetic hysteresis and low-temperature magnetometry measurements on four selected grains with different NRM/IRM ratios. All of the four samples exhibited similar features in both hysteresis and low-temperature magnetization. Results on the hysteresis measurements fall in the PSD region of a Day plot (Fig. 6d) and are concentrated in a narrower region of the diagram compared to the whole rock. This indicates that the magnetite in plagioclase crystals has narrower range of grain size than that in the whole rock. The Verwey transition of magnetite was clearly observed at approximately 120 K (Fig. 6e), indicating a very low titanium content of magnetite (Özdemir et al. 1993; Moskowitz et al. 1998). Larger remanence in the FC curve relative to the ZFC curve (Fig. 6e) also suggests a dominance of fine-grained magnetite (Moskowitz et al. 1993; Carter-Stiglitz et al. 2001, 2002; Kosterov 2003).

After paleointensity measurements, we took hysteresis measurements on four samples and low-temperature measurements on one sample. The results were similar to those shown in Fig. 6c, e, which implies that the double heating during paleointensity measurements did not severely affect the magnetic characteristics of plagioclase grains. A distribution of the blocking temperature was investigated on four samples after paleointensity experiments. Results show a very narrow blocking temperature distribution around 530–580 °C (Fig. 6f).

Figure 7 is a microscopic image of a polished plagioclase sample (sample no. 68 in Table 2). The tiny opaque minerals with rounded to needle-like shapes are uniformly distributed in the host plagioclase and showed no association with cracks. These opaque minerals are probably magnetite, and the texture implies that the magnetites were not generated by secondary alternation but rather have a primary origin such as incorporation during plagioclase crystallization or exsolution at subsolidus conditions. Furthermore, the needle-like shape of magnetite and their preferred alignment relative to the feldspar suggests an origin via exsolution because magnetite tends to form equant octahedral crystals when crystallizing from a magma. The needle-like grains could be categorized to SD state due to particle length (few micron in most but up to > 10 um) and width length ratio of < 0.1 (Dunlop and Özdemir 1997). The round-shaped grains are possibly in the PSD state.

To summarize, rock-magnetic measurements of plagioclase samples indicate that the plagioclase crystals contain nearly-pure needle-like-shaped SD and PSD magnetite with width less than a few micron and various aspect ratio and are suitable for paleointensity measurements.

Paleointensity experiments of plagioclase

We conducted Tsunakawa–Shaw paleointensity experiments on 17 plagioclase grains. In consideration of the sensitivity of the instrument, samples with NRM intensities larger than 5 × 10⁻¹¹ Am² were chosen for the experiments. Taking weak remanences of single-crystal samples into account, we employed a slightly different selection criteria from the study by Yamamoto and Tsunakawa (2005) which worked on strong remanences of volcanic whole rocks. The criteria we adopted are:

1. 1.

   A primary component found in an orthogonal plot of NRM demagnetization

    
2. 2.

   f > 0.3 in a NRM–TRM1* plot

    
3. 3.

   R > 0.90 in a NRM–TRM1* plot

    
4. 4.

   Slope of a TRM1–TRM2* plot within 1 ± 0.1

    
5. 5.

   R > 0.95 in a TRM1–TRM2* plot

    

where f is the NRM fraction of the primary component and R is the correlation coefficients. Primary components were identified on an orthogonal plot of NRM, and it associated with MAD values < 16°. By the LTD treatment, 5–15% of the NRM was demagnetized. The demagnetized components by LTD are attributed to the remanence carried by PSD magnetite (Heider et al. 1992). By the above criteria, 9 out of 17 results were selected: two results were rejected by the criteria of 1 and/or 3; six results were rejected due to the criterion of 4. Typical examples of successful and rejected results are, respectively, presented in Figs. 8, 9 and 10. Results of all 17 samples are summarized in Table 1. Paleointensities obtained from the nine results range between 43.1 and 77.9 μT, yielding an average of 57.4 μT and a standard deviation of 11.8 μT. Figure 11 shows an individual plagioclase paleointensity as well as their average together with the average paleointensity reported from the whole rocks. The plagioclase average is in good agreement with the average paleointensity reported from the whole rocks, though the dispersion of plagioclase paleointensity is slightly larger than that of the whole rocks.

Sample ID	NRM0 (pAm2)	HL (mT)	f	MAD (°)	ARM0 (pAm2)	NRM–TRM1*		TRM1–TRM2*
						Slope	R	Slope	R	F (μT)
Successful samples
9004	134.6	0	1.000	9.2	248.7	0.945 ± 0.064	0.9647	1.054 ± 0.050	0.9827	47.2 ± 3.2
9009	93.0	40	0.700	9.8	147.2	1.142 ± 0.100	0.9706	1.094 ± 0.021	0.9972	57.1 ± 5.0
9020	484.0	45	0.486	9.2	616.5	1.253 ± 0.083	0.9658	1.048 ± 0.026	0.9944	62.6 ± 4.2
9036	344.0	45	0.499	10.2	794.1	0.862 ± 0.120	0.9075	1.082 ± 0.021	0.9966	43.1 ± 6.0
9099	193.1	0	1.000	11.0	319.0	1.558 ± 0.126	0.9456	1.051 ± 0.056	0.9744	77.9 ± 6.3
9054	264.8	35	0.570	15.9	355.1	1.457 ± 0.187	0.9075	1.078 ± 0.061	0.9710	72.8 ± 9.4
9070	126.7	15	0.942	12.9	336.0	0.974 ± 0.096	0.9259	0.968 ± 0.032	0.9900	48.7 ± 4.8
34,019	104.0	20	0.868	11.0	265.5	1.129 ± 0.106	0.9394	0.960 ± 0.031	0.9908	56.4 ± 5.3
44,092	64.7	10	0.992	10.6	117.1	1.007 ± 0.096	0.9301	1.008 ± 0.041	0.9851	50.3 ± 4.8
									Mean	57.4 ± 11.8
Failed samples
9013	90.8	35	0.677	7.9	106.9	1.424 ± 0.095	0.9785	0.887 ± 0.051	0.9728	71.2 ± 4.7
9016	79.5	35	0.600	11.1	156.0	0.714 ± 0.058	0.9714	1.297 ± 0.065	0.9805	35.7 ± 2.9
9035	269.6	No primary magnetization			552.5	–	–	0.961 ± 0.035	0.9872
9047	181.6	Low linearity in NRM/TRM1* plot			392.7		–	0.971 ± 0.029	0.9909
9061	174.4	20	0.860	8.9	335.7	0.935 ± 0.086	0.9420	1.232 ± 0.070	0.9705	46.7 ± 4.3
9113	495.7	25	0.855	8.9	624.4	1.611 ± 0.111	0.9662	1.173 ± 0.065	0.9722	80.6 ± 5.6
44,106	88.5	20	0.923	5.9	120.5	1.216 ± 0.104	0.9465	0.849 ± 0.078	0.9276	60.8 ± 5.2
44,108	76.2	0	1.000	5.3	132.4	1.310 ± 0.062	0.9847	0.627 ± 0.119	0.7706	65.5 ± 3.1

NRM0 NRM intensity before LTD, H_L the lowest AFD step used in paleointensity estimate, f NRM fraction of the primary magnetization used in paleointensity estimate, ARM0 intensity of ARM0 before LTD, R correlation coefficient, F estimated paleointensity

Two protocols were employed for handling the anisotropy bias on paleointensity ("Paleointensity experiments" section). In both protocols, it was technically difficult to impart ARM0 accurately parallel to ChRM. Therefore, the anisotropy bias on each sample would not be corrected completely. The angular differences between ChRM and ARM0 were 24° at most. The possible canceling of anisotropy bias by averaging a number of samples is discussed in "Anisotropy effect on paleointensity" section. The protocol in which ARM1, ARM2, TRM1 and TRM2 were given along the Y axis seems to be more reproducible for the present sample configuration, though the number of studied samples was not enough to determine which protocol was more suitable.

We found that ARM was larger than TRM in all plagioclase samples, in contrast to the whole-rock sample. This has been reported as a peculiar feature of exsolved magnetite by Usui et al. (2015).

Remanence anisotropy of plagioclase

ARM anisotropy tensors were estimated from the measured ARM_x, ARM_y, and ARM_z for each plagioclase grain. Eigenvalues and anisotropy parameters, corrected anisotropy degree P_j, and shape factor T_j (Jelinek 1981) were calculated (Table 2). Positive and negative values of T_j indicate that the shape of the anisotropy ellipsoids is oblate and prolate, respectively. Median of P_j is 3.35 and T_j varies from + 0.72 to − 0.80. Typical results of analysis on anisotropy directions and anisotropy parameters are shown in Fig. 12. The directions of the anisotropy axes are identical for ARM and TRM, and do not change by AFD. Hence, we used the ARM anisotropy tensor as a proxy for TRM anisotropy tensor in the discussion in "Anisotropy effect on paleointensity" section. P_j increased after AFD, which could be reasonable considering that grains with high aspect ratio correspond to high coercivity components. The whole-rock ARM was isotropic (P_j = 1.2, T_j = − 0.28).

Sample name	Mean ARM (10−10 Am2)	Eigen values of remanence tensor			P j	T j
		w 1	w 2	w 3
9020	3.22	1.70	0.81	0.49	3.51	− 0.20
9036	4.49	1.56	1.01	0.43	3.72	0.32
9099	1.77	1.50	0.92	0.58	2.60	− 0.03
9054	2.14	1.42	0.93	0.65	2.19	− 0.08
9070	2.14	1.36	0.90	0.74	1.87	-0.38
34,019	1.64	1.60	1.07	0.33	5.18	0.49
44,092	0.95	1.43	1.09	0.48	3.10	0.50
9035	3.06	1.40	0.86	0.74	1.96	− 0.51
9047	2.54	1.23	1.04	0.73	1.70	0.37
9061	1.58	1.64	0.88	0.48	3.41	0.00
9113	3.08	1.57	0.83	0.60	2.68	− 0.31
44,106	0.85	1.58	0.85	0.57	2.78	− 0.23
44,108	0.71	1.48	0.80	0.71	2.19	− 0.67
11	1.16	1.65	0.86	0.49	3.35	− 0.07
12	2.22	1.95	0.68	0.37	5.41	− 0.26
16	1.81	1.56	0.99	0.44	3.58	0.28
36	2.14	1.64	0.92	0.44	3.76	0.12
55	1.91	1.48	1.20	0.32	5.17	0.72
68	1.35	1.79	0.64	0.57	3.52	− 0.80
Median		1.56	0.90	0.49

Mean ARM average of the eigenvalues of ARM anisotropy tensor, w₁, w₂ and w₃ maximum, medium and minimum eigenvalue of ARM anisotropy tensor normalized by the mean ARM

Anisotropy effect on paleointensity

Our paleointensity value could be either larger or smaller than the true value depending on angles between the anisotropy axes and directions of ancient or laboratory fields (Paterson 2013). In our paleointensity experiments, anisotropy bias is mainly caused by the directional difference between the external field that gave the NRM, and the laboratory field that gave the ARM0. Since the whole-rock sample was isotropic, the plagioclase grains should be randomly oriented in the host rock assuming that the ARM of whole rock is mainly carried by plagioclase-hosted magnetite. Therefore, the direction of the external field which gave NRM should be random against the anisotropy axes of each plagioclase grain. We calculated the anisotropy bias between two remanence vectors given by randomly oriented external fields with same intensity using the typical anisotropy tensor with eigenvalues (w₁, w₂, w₃) = (1.56, 0.90, 0.49), assuming that the laboratory field was also randomly oriented against the anisotropy axes. Figure 13 shows the anisotropy bias (ratio of the intensity of two remanence vectors) as a function of the angular difference between two remanence vectors (NRM and ARM0). In the present study, we obtained paleointensity results from nine plagioclase samples; the angular difference between NRM and ARM0 for each sample was below 25°. In this condition, the anisotropy bias averaged for nine samples was within 1 ± 0.1, and the standard deviation was below 25%. Therefore, anisotropy bias is likely canceled by averaging paleointensity results from nine samples in this study. Also, the variation of the experimental results (~ 20% of the mean value) was consistent with our calculation. We concluded that accurate paleointensity information can be derived from the mean paleointensity values of an assembly of single plagioclase crystals, while the large dispersion of paleointensity values is constitutional for the randomly oriented anisotropic grain assemblage.

Based on anisotropy measurements of plagioclase crystals separated from an Archean granitoid, Usui et al. (2015) demonstrated that (1) geometric mean instead of arithmetic mean should be used, and (2) tens of crystals would be needed to achieve reliable paleointensity estimates. In the present study, the geometric mean and the arithmetic mean are similar in the range of standard deviation, so fewer crystals are required. This difference can be attributed to the variation of the anisotropy effect; the anisotropy tensor they used for estimates was more anisotropic (corresponding to P_j = 6.21 and T_j = 0.34) than that used in the present study (corresponding to P_j = 3.18 and T_j = 0.04). Since the shape and fabric of exsolved magnetite vary among samples, the anisotropy effect and how to get rid of it need to be studied carefully for each rock.

In addition to remanence anisotropy, nonlinear TRM acquisition is a major issue of exsolved magnetite paleomagnetism. In the case of the studied sample, the NRM/IRM ratio was lower than the TRM/IRM ratio of previously studied plagioclase crystals (Usui et al. 2015) and rocks containing exsolved magnetite (Selkin et al. 2007). This implies that the nonlinear TRM acquisition may be insignificant for the obtained paleointensity range (~ 60 μT). Also, there is a possibility that NRM of exsolved magnetite is a thermochemical remanent magnetization (TCRM) rather than a TRM since the formation temperature of exsolved magnetite in plagioclase is not clear (Feinberg et al. 2005). In that case, obtained paleointensity could give the lower limit of the field strength at the age, as TCRM acquisition is less efficient than TRM acquisition (Stacey and Banerjee 1974; Usui and Nakamura 2009).

Comparison of magnetic carriers of plagioclase and whole-rock samples

Wakabayashi et al. (2006) and Tsunakawa et al. (2009) predicted that most of the stable remanence of the Iritono granite was carried by magnetite inclusions in plagioclase. However, detailed rock-magnetic experiments on plagioclase grains compared to the previously reported whole-rock studies revealed that the distribution of blocking temperatures and grain sizes are different between plagioclase crystals and the whole rock. The pTRM distributions do not show any concentration in a particular temperature interval below 550 °C for plagioclase samples, while about 10% of the whole-rock TRM is carried by a low blocking temperature (300–500 °C) component. Because the Iritono granite contains magnetite and pyrrhotite ("Rock-magnetic properties of zircon, Rock-magnetic properties of quartz, Rock-magnetic properties of plagioclase" sections; Wakabayashi et al. 2006; Tsunakawa et al. 2009), the low blocking temperature component above 350 °C found in the whole-rock TRM can be attributed to coarse-grained PSD and MD magnetites. Hysteresis loop measurements also indicate that magnetite in plagioclase has a narrow range of grain size compared to the whole rock (Fig. 6d). In addition, the results of the whole-rock experiments exhibit a bimodal distribution according to different paleointensity methods which suggest the influence of non-ideal magnetic minerals, and alternation of such minerals which could not be detected or suppressed completely. On the other hand, plagioclase samples mostly contain nearly-pure, fine-grained magnetite as the magnetic carriers.

Nevertheless, the magnetic carrier of plagioclase crystals and the whole rock was different in terms of distribution of grain size and blocking temperatures, and the estimated paleointensity was consistent among them. Therefore, we conclude that a reliable paleointensity was obtained successfully. Because of the more ‘ideal’ magnetic carrier, paleointensity experiments on single plagioclase with exsolved magnetite inclusions can potentially give more reliable and informative paleointensity results than the conventional whole-rock experiments.

Effect of cooling rate on paleointensity

The extremely slow cooling of granitic rocks compared to laboratory timescales may require a correction to the paleointensity estimate due to the time dependence of the acquisition of TRM, which varies by size and aspect ratio of the magnetic grains (Halgedahl et al. 1980; Selkin et al. 2000; Yu 2011). Results on magnetic hysteresis measurements of plagioclase grains are plotted in the PSD region on the Day plot (Fig. 6d), which could be interpreted as a mixture of a range of grain size and aspect ratio. The stable remanence that is involved in paleointensity measurements is carried by SD to PSD magnetite. Based on SD theory (Halgedahl et al. 1980; Selkin et al. 2000) and the estimated cooling time of the Iritono granite body, Tsunakawa et al. (2009) argued that the ratio of TRM in nature to TRM in laboratory would be 1.5 for the SD components. On the other hand, PSD grains have insignificant cooling rate dependence on TRM acquisition (Yu 2011). The cooling rate corrected paleointensity value of 38.2 ± 7.9 μT assuming SD magnetite gives the lower limit of the paleointensity, since the PSD magnetite should give higher paleointensity value. Therefore, corresponding cooling rate corrected VDM value of 8.9 ± 1.8 × 10²² Am² using the paleointensity value of plagioclase crystals and inclination of the H component of the whole rock in Wakabayashi et al. (2006) can impose a constraint on the lower limit of paleointensity at the age of 115 Ma.

Significance of Shaw-type paleointensity methods on single crystals

This is the first report applying the Tsunakawa–Shaw paleointensity method to single grain samples. Considering that several results were rejected because of severe alternation during laboratory heating, a Shaw-type method, in which number of heating in laboratory are minimized, seems to be more appropriate than a Thellier-type method. Furthermore, the ThD curves of plagioclase crystals (Fig. 6f) show a very narrow distribution of blocking temperature below the Curie temperature of magnetite (530–580 °C), while the AFD curve (Top-right diagram in Fig. 8) show broad distribution of coercivity (50–150 mT). This emphasizes an advantage to estimate a paleointensity not in a blocking temperature space (by a Thellier-type method) but in a coercivity space (by a Shaw-type method), especially for a magnetically weak sample such as single crystals. Thus the Tsunakawa–Shaw method could be more suitable than the Thellier–Thellier methods in the case of plagioclase sample containing exsolved magnetite, while these methods should be compared in the future paleointensity study using appropriate samples.

Paleointensity during middle CNS

Considering the possible TCRM origin of NRM and the contribution of PSD grains on the cooling rate correction, the VDM value of 8.9 ± 1.8 × 10²² Am² gives the lower limit of the time-averaged field strength during the middle age of CNS. Average field strength of the periods of frequent reversals have been estimated as the VDM value of the past 5 million years from the Society Islands volcanic rocks (3.6× 10²² Am²; Yamamoto and Tsunakawa 2005), and the virtual axial dipole moment (VADM) value of 0–160 Ma excluding CNS period from submarine basalt glass samples (4.8× 10²² Am²; Tauxe 2006). The present result suggests that the time-averaged field strength during middle CNS was several times as large as that of non-superchron periods, supporting the prediction by dynamo models and simulations (e.g. Larson and Olson 1991; Glatzmaier et al. 1999; Kutzner and Christensen 2002; Christensen and Aubert 2006; Olson and Christensen 2006; Courtillot and Olson 2007; Takahashi et al. 2008; Olson et al. 2010). By applying the present paleointensity method to various granitic rocks from different ages, we may be able to improve our understanding of the long-term behavior of the geomagnetic field concerning the mantle convection process without the complications of unideal magnetic minerals that often compromise such work.