Archeointensity estimates of a tenth-century kiln: first application of the Tsunakawa–Shaw paleointensity method to archeological relics

Abstract

Paleomagnetic information reconstructed from archeological materials can be utilized to estimate the archeological age of excavated relics, in addition to revealing the geomagnetic secular variation and core dynamics. The direction and intensity of the Earth’s magnetic field (archeodirection and archeointensity) can be ascertained using different methods, many of which have been proposed over the past decade. Among the new experimental techniques for archeointensity estimates is the Tsunakawa–Shaw method. This study demonstrates the validity of the Tsunakawa–Shaw method to reconstruct archeointensity from samples of baked clay from archeological relics. The validity of the approach was tested by comparison with the IZZI-Thellier method. The intensity values obtained coincided at the standard deviation (1σ) level. A total of 8 specimens for the Tsunakawa–Shaw method and 16 specimens for the IZZI-Thellier method, from 8 baked clay blocks, collected from the surface of the kiln were used in these experiments. Among them, 8 specimens (for the Tsunakawa–Shaw method) and 3 specimens (for the IZZI-Thellier method) passed a set of strict selection criteria used in the final evaluation of validity. Additionally, we performed rock magnetic experiments, mineral analysis, and paleodirection measurement to evaluate the suitability of the baked clay samples for paleointensity experiments and hence confirmed that the sample properties were ideal for performing paleointensity experiments. It is notable that the newly estimated archaomagnetic intensity values are lower than those in previous studies that used other paleointensity methods for the tenth century in Japan.

Introduction

The Earth’s magnetic field (geomagnetic field) is known to vary significantly with time. Variations with a comparatively short period between several decades and several millennia are referred to as geomagnetic secular variations, and their identification is needed to understand the dynamic behavior in the core using the dynamo calculation and to estimate the archeological age of excavated relics. Age estimation is done by comparing the value of the geomagnetic field reconstructed from baked materials to the secular variation curve, and this process is called the archeomagnetic dating method (e.g., Nakajima and Natsuhara 1981; Lanos 2004; Jordanova et al. 2004).

Numerous archeomagnetic studies have been conducted in Japan using oriented baked clay samples collected from several types of burnt relics, such as kilns (e.g., Hirooka 1977). These studies used samples from excavations performed by local government archeological research organizations. The archeomagnetic data can be extracted from the GEOMAGIA 50 database (Korhonen et al. 2008; Brown et al. 2015), which holds a total of 245 Japanese datasets (paleodirection and paleointensity), including 102 paleointensity datasets from archeological material (Nagata et al. 1963; Sasajima 1965; Sasajima and Maenaka 1966; Kitazawa 1970; Domen 1977; Sakai and Hirooka 1986).

All of the data collected before 1986 pertained to conventional experimental techniques, including the original Thellier–Thellier method (Thellier and Thellier 1959), the Domen–Thellier method (1977), and the Sakai–Thellier method (Sakai and Hirooka 1986). Other modern techniques have been proposed to improve reliability, such as the IZZI-Thellier method (Tauxe and Staudigel 2004) based on stepwise thermal demagnetization (ThD) and the Tsunakawa–Shaw method (Yamamoto et al. 2003; Mochizuki et al. 2004; Oishi et al. 2005) based on stepwise alternating field demagnetization (AFD). Studies have been conducted using the IZZI-Thellier method for both volcanic rocks (e.g., Cromwell et al. 2015; Yu 2012) and archeological material (e.g., Hong et al. 2013; Cai et al. 2014, 2015, 2017). For the Tsunakawa–Shaw method, many studies have focused on volcanic rocks (e.g., Yamamoto and Tsunakawa 2005; Mochizuki et al. 2006; Yamamoto et al. 2010; Mochizuki et al. 2011, 2013; Yamazaki and Yamamoto 2014; Ahn et al. 2016), but only one has applied the method to samples taken from clay from the floor of a reconstructed kiln from the present age that was burnt in a known magnetic field, demonstrating that such samples are suitable for archeointensity research (Yamamoto et al. 2015).

In this study, we applied the Tsunakawa–Shaw method to actual archeological remains for the first time. We also performed a variety of rock magnetic experiments, mineral analysis, and paleodirection measurements to evaluate the suitability of baked clay samples for paleointensity experiments.

Samples

Sueki (Sue ware) is a type of earthenware that was widely used until the latter part of the Heian period to the early part of the Kamakura period (around the twelfth century) in Japan and that originated on the Korean peninsula and came to Japanese islands in the middle part of the Kofun period (the fifth century). The surface of Sueki exhibits a grayish blue and glassy luster because it was fired at temperatures greater than 1000 °C in an Anagama-type kiln (e.g., Mizoguchi 2013).

Samples used in this study were collected from the remains of the Sayama Higashiyama-Oku kiln, which is located in the Sayama area of Bizen City, Okayama Prefecture, Japan (34°41′N, 134°11′E, Fig. 1a, b). It was excavated by the Okayama University of Science. The excavated artifacts of the kiln were mainly Sueki fragments and kiln wall blocks. The types of Sueki fragments are Tsuki-type goblets, bowls, small plates, pots, Tsubo-type jars, Kame-type jars, and Fuhji-ken-type inkstones. The Tsuki-type goblets and bowls consisted of 228 circular-foot-style fragments, 87 flat-foot-style fragments, and 84 flat-base-style fragments. The mean diameters of the basilar parts of each style are as follows: 7.0–8.5 cm for the circular-foot-style, 7.0–7.4 cm for the flat-foot-style, and 7.0–7.4 cm for the flat-base-style fragments [see Kameda et al. (2014) for further archeological information]. Comparing this archeological information with that reported in previous studies (e.g., Nagayama 1936; Itoh 1987; Nishikawa 1966; Kameda 1996), it is estimated that the kiln was operated during around the first half of the Heian period (~ AD 900–1000). The geology of the area is characterized by rhyolite (National Institute of Advanced Industrial Science and Technology 2017).

Fig. 1

a Map around Japan. b Enlarged map around Sayama Higashiyama-Oku kiln, which is located at the east end of the Oku old kiln complex. c Entire photograph of remains of Sayama Higashiyama-Oku kiln (kiln body). d Enlarged photograph around baking chamber of the kiln, where 8 block samples (fixed using plaster) were collected

Baked clay blocks (~ 10 cm square) were collected from the kiln floor of the central part of the burning chamber (Fig. 1c, d) in the following way:

• 1. Rectangular grooves were dug into the kiln floor using a tile hammer.

• 2. The surface of the blocks surrounded by the grooves was reinforced with plaster.

• 3. The orientations of the blocks were measured using a Brunton compass as a strike and a dip, and this direction information was marked on the upper surface of the blocks.

• 4. The blocks were peeled from the floor. They were named as BH 1 to BH 8 (see Fig. 1d for the placement of samples in the kiln).

To avoid the less-fired and/or unfired parts, we used only the surficial parts of the blocks (within 1.5 cm from the kiln floor) for experiments. For AFD measurements, each 1.5-cm cubic specimen was cut from the block samples and then placed into a 7-cm³ plastic case with paper clay. For measurements of archeointensity and ThD, a 1.5-cm³ specimen cut from the block samples was placed into a silica glass case with glass wool and high-temperature cement. For the magnetic and mineralogical experiments, crushed specimens were prepared from the block samples. In the archeodirection measurements, the declination collection was set to be − 7.3°, based on the IGRF-12 model (International Association of Geomagnetism and Aeronomy, Working Group V-MOD 2010).

Methods

Rock magnetic experiments

Measurements of hysteresis parameters (saturation magnetization, Ms; saturation remanent magnetization, Mrs; coercivity, Hc; coercivity of remanence, Hcr) and first-order reversal curves (FORCs) were taken using a vibrating sample magnetometer (VSM; MicroMag 3900, Princeton Meas.) at the Kochi Core Center (KCC), Kochi, Japan. Hysteresis loops were measured in the coercivity range of − 1.4 to 1.4 T. FORC diagrams were generated using the UNIFORC code (Winklhofer and Zimanyi 2006; Egli et al. 2010). The thermomagnetic curve was measured in air using a magnetic balance (NMB-89, Natsuhara) of KCC. Crushed specimens were gradually heated from approximately 30 °C to 700 °C and then gently cooled to approximately 50 °C at a rate of 10 °C/min. Throughout the temperature cycle, a field of 300 mT was applied to the sample. Acquisition curves of isothermal remanent magnetization (IRM) were obtained using a magnetic property measurement system (XL5, Quantum Design) at the Machine Center, Okayama University of Science, Japan. IRM was initially applied at 5 T in the reverse direction and subsequently applied in steps with 1–5000 mT sections divided into 60 intervals at logarithmic intervals in the positive direction. The curves were analyzed to estimate coercivity components using the IRM unmix v2.2 program (Heslop et al. 2002).

Mineralogical experiments

X-ray diffraction (XRD) measurement was taken using a Rigaku X-ray Diffractometer RINT 2100 V at Kyushu University with Cu kα radiation monochromatized by a curved graphite crystal. Data analyses such as determination of XRD peak positions and identification of minerals were performed using Mac Diff 4.2.5 (Petschick 2000). The measurements were taken to estimate the firing temperature in the kiln and the absence of specific mineral assemblages, which determines the degree of thermal transformation that occurs in the samples during the firing procedure and, as a consequence, can provide an estimate of the firing temperature during the operation of the kiln (e.g., Schomburg 1991; Berna et al. 2007).

Archeodirection measurements

Archeodirection was measured in (1) stepwise ThD experiments and (2) stepwise AFD experiments.

In the ThD experiments, we used a thermal demagnetizer (TDS-1, Natsuhara-Giken), an automated spinner magnetometer with AF demagnetizer (DSPIN, Natsuhara-Giken) at the Information Processing Center, Okayama University of Science (IPC), and a thermal demagnetizer (TDS-1, Natsuhara-Giken) and spinner magnetometer (ASPIN, Natsuhara-Giken) at Kyushu University. The demagnetization was performed between 100 and 700 °C, and the heating hold time was 15 min. The specimens were cooled in the thermal demagnetizer using a fan.

The stepwise AFD experiment was also conducted using an automated spinner magnetometer with an AF demagnetizer (DSPIN, Natsuhara-Giken) at the KCC, and the demagnetization was performed in the AF range between 2 and 180 mT.

The results were projected onto a Zijderveld diagram (As and Zijderveld 1958) and Schmidt net to determine the direction of the primary remanence by principal component analysis (Kirschvink 1980).

Archeointensity measurements

We used the Tsunakawa–Shaw and IZZI-Thellier methods in the archeointensity experiments.

We followed the procedures described in Yamamoto et al. (2003), Mochizuki et al. (2004), and Oishi et al. (2005) for the Tsunakawa–Shaw method and used an automated spinner magnetometer with an AF demagnetizer (DSPIN, Natsuhara-Giken) to measure and demagnetize remanence and to impart anhysteretic remanent magnetization (ARM). Laboratory TRM was imparted in a vacuum using a thermal demagnetizer (TDS-1, Natsuhara-Giken) with an applied field of 50 μT. We held the maximum temperature at 610 °C for 15 (30) min for the first (second) heating, and then, the specimens were cooled slowly to room temperature over 2–3 h. Low-temperature demagnetization (e.g., Ozima et al. 1964; Heider et al. 1992) was performed in a shielded case made of permalloy: a specimen was put in a dewar filled with liquid nitrogen for 10 min and subsequently pulled out to warm up to room temperature using a fan for 20 min.

Specimen-level experimental results were analyzed using criteria similar to those adopted in Yamamoto et al. (2003), which are summarized in the upper column of Table 1. They were further analyzed to determine site-level results based on the criteria shown in the lower column of Table 1.

Table 1 List of the selection criteria for the Tsunakawa–Shaw method

For application of the IZZI-Thellier method, temperature steps were set to be every 50 °C between room temperature and 500 °C and every 35 °C between 500 and 600 °C. The pTRM checks (Coe 1967) were performed at 300, 400, 500, and 565 °C. All of the heating was performed in the air using a thermal demagnetizer (TDS-1, Natsuhara-Giken) at the KCC and IPC. For each heating run, the specimen was maintained at the destination temperature for 15 min and then cooled to room temperature using a fan. For the in-field step, laboratory TRM was applied by a DC field of 50 μT, and its direction was set to be orthogonal to the initial NRM direction of the specimen. The measurement of remanence was taken using a spinner magnetometer (ASPIN, Natsuhara-Giken) and an automated spinner magnetometer with an AF demagnetizer (DSPIN, Natsuhara-Giken).

Specimen-level experimental results were analyzed on Arai plots (Arai 1963; Nagata et al. 1963) using the Thellier GUI program included in the “PmagPy” software package by Shaar and Tauxe (2013), based on criteria summarized in the upper column of Table 2, which were similar to those adopted in Cromwell et al. (2015). Site-level results were determined based on the criteria shown in the lower column of Table 2, which are identical to those adopted in the Tsunakawa–Shaw experiment.

Table 2 List of the selection criteria for the IZZI-Thellier method

Results

Rock magnetic experiments

Representative hysteresis loops are shown in Fig. 2a, b. A narrow, saturated loop is a characteristic typical to titanomagnetite, which mainly consists of pseudo-single-domain (PSD) particles (e.g., Tauxe et al. 1996). The hysteresis parameters are distributed across the ranges of 0.27 < Mr/Ms < 0.38, and 1.6 < Hcr/Hc < 1.9. These values are plotted in the area close to the PSD region in a day plot (Day et al. 1977) (Fig. 2c). It appeared that induced magnetizations almost reached saturations at ± 1.0 T. Therefore, we did not carry out the correction of the hysteresis parameters for unsaturated loops pointed out in Doubrovine and Tarduno (2006). For most of the specimens, the FORC diagrams are characterized by a dominance of single-domain (SD)-like components (Fig. 2d, e), while multidomain (MD)-like components with wider vertical spreads are superimposed in some specimens (Fig. 2e).

Fig. 2

a, b Typical hysteresis loops before (blue) and after (red) paramagnetic correction, respectively. c Day plot. d, e Typical FORC diagrams. f–h Typical thermomagnetic curves heated in air. Red line represents the heating procedure. Blue line represents the cooling procedure. i Typical IRM acquisition curve. j Result of IRM decomposition of (i). Black dots represent first-order differentiation of the originally measured IRM acquisition curve. Red solid lines represent the decomposed coercivity components. Green solid line represents an accumulation of all components indicated by red solid lines

Thermomagnetic curves are classified roughly into three types (Fig. 2f–h). In the first type (Fig. 2f), the curve is almost completely reversible without showing any thermal alteration. The second type (Fig. 2g) exhibits a heating curve that overwhelms the cooling curve. This indicates that some titanomagnetites might change into hematites upon heating. The third type (Fig. 2h) shows that the heating curve is always lower than the cooling curve. This infers that some new magnetic mineral might be produced upon heating. Overall, alterations that occurred during heating–cooling cycles were negligible and/or small, suggesting that the present samples are resistant to thermal alteration during laboratory heating. Because the Curie temperatures of the samples are around 500–550 °C, the major magnetic carriers are considered to be Ti-poor titanomagnetites (Nagata and Akimoto 1961).

A typical result of the IRM acquisition curves is shown in Fig. 2i. By decomposing the curves, three components with coercivities at (1) 20 mT, (2) 60 mT, and (3) 130 mT were confirmed in all samples (Fig. 2j). They are interpreted to be Ti-poor titanomagnetites with three different grain sizes, considering the magnetic characteristics.

Mineralogical experiments

On the XRD patterns of the sample BH1, an XRD peak set (e.g., 5.39 Å for 110 peak, 2.89 Å for 001, 2.69 Å for 220) corresponding to mullite and the strongest peak (4.04 Å for 101) of cristobalite was recognized. These XRD peak sets were also found on the XRD patterns of the samples BH7 and BH8, indicating the presence of mullite and cristobalite (Fig. 3).

Fig. 3

X-ray powder diffraction patterns for samples BH1, BH7, and BH8

Consequently, the mineralogical constituents of these samples affirm that the firing was at temperatures that were sufficient to result in full-TRM acquisitions of the samples. The firing temperature can be estimated to be higher than 1000 °C, which is higher than the Curie temperatures of magnetite and hematite. Doi and Sakamoto (2002) and Shiraishi and Kyoguro (2002) reported that the presence of cristobalite and mullite suggests firing at temperatures higher than 1000 °C for the clay artifacts of Bizen City (these materials have a mineral composition similar to the baked clay samples used in this study).

Archeodirection measurements

• 1. The ThD experiment was performed for a total of 6 cubic samples. All the detailed experimental results are shown in Table 3, and typical Zijderveld diagrams are shown in Fig. 4a–c. It is common to all samples that the remanence consists mostly of one stable characteristic, remanent magnetization and that the secondary magnetization is very small (the secondary magnetization is almost all demagnetized at 100 °C). The maximum blocking temperature of all specimens distributes around 550 °C. The equal area projection of the paleodirection, which was obtained after principal component analysis (Kirschvink 1980) for these Zijderveld diagrams, is shown in Fig. 5a, and the mean paleodirection is shown in Table 5.

  Table 3 Measurement results of the ThD experiment

  Fig. 4

  

  a–c Typical Zijderveld plot for ThD experiments. d–f Typical Zijderveld plot for AFD experiments. Solid circles represent the horizontal projection. Open circles represent the vertical projection. The inserted graph is the stepwise demagnetization curve

  Fig. 5

  

  a Direction of the remanent magnetization plotted on the Schmidt net for ThD experiment. b Direction of the remanent magnetization for AFD experiments. c Orange point with orange dashed circle indicates the mean direction with α95 of ThD and AFD experiment. Purple dots represent the reference curve between AD 0 and 2000 at every 50 years proposed by Hirooka (1977)

• 2. The AFD experiment was conducted for a total of 33 cubic specimens. All the detailed experimental results are shown in Table 4, and typical Zijderveld diagrams are shown in Fig. 4e–g. Like the ThD experimental data, for all samples, the remanence consists primarily of one stable characteristic, remanent magnetization, and the secondary magnetization is very small (the secondary magnetization is almost all demagnetized at 8 mT). The maximum angular deviation (MAD; Kirschvink 1980) values are mostly less than 1°. The median destructive field (MDF) is higher than 30 mT, which indicates that the remanence is very stable and hard (there are specimens with an MDF higher than 70 mT). Equal area projection of the paleodirection is shown in Fig. 5b, and the mean paleodirection is shown in Table 5.

  Table 4 Measurement results of the AFD experiment

  Table 5 Mean direction obtained by the ThD and the AFD experiment

The mean directions in the AFD and ThD experiments coincide within the range of α₉₅ (95% confidence limit). In Fig. 5c, we plot the mean direction of ThD and AFD experiments. In comparison with the reference curve by Hirooka (1977), the mean direction of the remains in this study is plotted close to the paths in (1) AD 500–550 and (2) AD 900–950, within the range of α₉₅. In particular, the mean value overlaps the path in (2) AD 900–950. This age coincides with the archeologically estimated age (AD 900–1000). Therefore, assuming that the reference curve by Hirooka (1977) is accurate, this suggests that the remanent magnetization was acquired at the time of the last burning in AD 900–950.

Archeointensity measurements

The Tsunakawa–Shaw experiment was performed on a total of 8 specimens. All of the specimens yielded results passing the selection criteria, which are listed in Table 6. Typical results are shown in Fig. 6. All the archeointensity values are determined from more than 75% of the total NRMs (f _N  ≥ 0.75). The slopes of the ARM₀–ARM₁ diagrams (slope_A0) are almost in unity (0.989–1.07) except for the specimen BH8L, suggesting little influence of laboratory alterations and anisotropies of the remanent magnetizations. The site-mean archeointensity is calculated to be 48.1 μT with a coefficient of variation $H_{\delta } \%$ of $6.4\%$ ($H_{\text{past}} = 48.1 \pm 3.1 \,\upmu{\text{T}}$) (Table 6).

Table 6 Acceptable results of the Tsunakawa–Shaw experiment

Fig. 6

a, b Typical results of the Tsunakawa–Shaw method. All the samples show similar graph patterned in common (NRM-TRM1 and ARM0-ARM1 ratios show extremely high linearity). The 3 diagrams on the left side show results from the first laboratory heating. The 3 diagrams on the right side show results from the second laboratory heating

The IZZI-Thellier experiments were conducted on a total of 18 specimens. Only the results from 3 specimens passed the selection criteria (success rate is 17%), which are listed in Table 7. The Arai plots of the successful results show excellent linearities (Fig. 7a–c). These results come from samples not showing noticeable dominance of the MD component in the FORC diagram (Fig. 2d). Conversely, the rejected 15 results show nonlinear trends in the Arai plots, such as concave-up curves (Fig. 7d), systematic failures of the pTRM checks with increasing temperatures (Fig. 7e), and zig-zagging behavior (Fig. 7f). These results come from samples both showing and not showing noticeable dominance of the MD component in the FORC diagram (Fig. 2d, e). It is likely that these nonlinearities are due to the alteration of the specimens and/or the effect of MD particles. The site-mean archeointensity is calculated to be 45.3 μT with a coefficient of variation $H_{\delta } \%$ of $9.7\%$ ($H_{\text{past}} = 45.3 \pm 4.4 \,\upmu{\text{T}}$) (Table 7).

Table 7 Acceptable results of the IZZI-Thellier experiment

Fig. 7

Typical examples of Arai plots of the IZZI-Thellier method of the accepted specimens (a–c) and the rejected specimens (d–f). Solid circles represent the general NRM/TRM ratios. Open triangles represent the points of pTRM check. Green lines indicate the selected straight segment. Red circles (blue circles) correspond to ZI steps (IZ steps)

Discussion

In this study, we applied the two different modern paleointensity methods to the baked clay samples taken from the kiln floor of a Japanese Sueki old kiln. The site-mean archeointensity values resulted in $48.1 \pm 3.1 \,\upmu{\text{T}} \left( {n = 8} \right)$ by the Tsunakawa–Shaw method and $45.3 \pm 4.4 \,\,\upmu{\text{T}} \left( {n = 3} \right)$ by the IZZI-Thellier method. The two means are indistinguishable at the 1σ level, with very small coefficients of variation (Fig. 8a), which indicates that they are reliable archeointensity estimates. It is considered that the Tsunakawa–Shaw method can be employed to recover archeointensities with a high reliability and high success rate from archeological relics.

Fig. 8

a Comparison of the paleointensity value between the different experimental methods (on the specimen level). Red circles are the values by the IZZI-Thellier method, and the red dashed line is their mean intensity. Green circles are the values by the Tsunakawa–Shaw method, and the green dashed line is their mean intensity. b Paleointensity values (site mean) of this study and previous studies. Green circle is the mean intensity by the Tsunakawa–Shaw method. Red circle is the mean intensity by the IZZI-Thellier method. Gray circles are the mean value reported by the previous studies in Japan extracted from the GROMAGIA 50 database. The errors are standard deviations (1σ)

The reliability of the archeointensity estimates is also ensured from the viewpoint of rock magnetism and mineralogy. The rock magnetic and mineralogical results indicate that the samples are suitable for paleointensity experiments: (1) PSD-like magnetic particle assemblage; (2) little magnetic interaction; (3) single ChRM components; (4) NRMs of TRM origins; (5) continuous blocking temperature distribution; (6) little laboratory alteration; (7) (titano-)magnetite as the main magnetic minerals; and (8) crystallization temperatures as high as 1000 °C for the constituting minerals. All these characters fulfill nearly, although not perfectly, the conditions of “the ideal behavior on the paleointensity experiments” first suggested by Coe (1967) as (1) the additivity law of pTRM, (2) independence law of pTRM, and (3) proportion law of TRM.

To compare our data with previously published data, we extracted the site-mean archeointensity data from the GEOMAGIA 50 database (Korhonen et al. 2008; Brown et al. 2015) by applying site-level selection criteria: a minimum of three successful results for a site ($n$ $\ge 3$) and successful results providing a site mean with a standard deviation less than 20% of the mean (coefficient of variation, $H_{\delta } \%$ ≤ 20%). It is noted that our data obviously fulfill these criteria. In the GEOMAGIA 50 database, there are two archeointensity results from archeological artifacts in Japan for AD 900–1000: (1) $58.9 \pm 5.9 \,\upmu{\text{T}}$ at site K-14 of Sasajima (1965) (baked earth, original Thellier method) and (2) $55.1 \pm 3.7 \upmu{\text{T}}$ of Sakai and Hirooka (1986) (ceramic, Sakai–Thellier method). Regarding data (1), its age was registered as AD $1000 \pm 20$ (originally reported in Sasajima 1965) but was corrected to AD $960 \pm 20$ in Sasajima and Maenaka (1966). We adopted the corrected age value for the data (1; Fig. 8b). Furthermore, the age as AD $1120 \pm 20$ of (3) $58.2 \pm 1.2\,\,\upmu{\text{T}}$ at site O-53 (baked earth, original Thellier method) in Sasajima (1965) was corrected to AD $1000 \pm 20$ in Sasajima and Maenaka (1966). Therefore, we also added the site-mean intensity value in this article (Fig. 8b). All of them were significantly higher than the results obtained by this study at the 1σ level (Fig. 8b).

One of the possible causes is thought to be undetected laboratory alterations in the previous study data because they were yielded by Thellier-type methods without pTRM checks. Another possible cause is that the linear segments in Arai plots might be selected in a biased way from low blocking temperature portions constituting a part of concave-up curves in previous studies. In recent studies, such Arai plots are interpreted to be affected by MD particles, and therefore, the intensity values calculated from such plots should be rejected because the reliability is suspicious (Tauxe 2010; Paterson 2011; Tanaka and Yamamoto 2016). Among the three previous data, Arai plots from sites K-14 and O-53 are, respectively, presented in Sasajima (1965) and Sasajima and Maenaka (1966). The Arai plot of site K-14 shows a straight line between room temperature and 500 °C, but some portion of NRM remains for a temperature range higher than 500 °C and is not associated with the pTRM check: the plot might be potentially convex downward. The two Arai plots of site O-53 are linear throughout all temperature ranges, but they were not associated with the pTRM checks.

Published site-mean archeointensity results in Japan selected from the GEOMAGIA 50 database for the last 2 kyr (Nagata et al. 1963; Sasajima 1965; Sasajima and Maenaka 1966; Sakai and Hirooka 1986) are all higher than the archeointensities obtained by the present study (Fig. 8b). They are mostly from Sakai and Hirooka (1986), and some of the Arai plots presented in their paper appear to be convex downward. It is speculated that they might systematically overestimate the past variation of the geomagnetic field intensity around Japan at the time.

Conclusion

In this study, we tested the suitability of the Tsunakawa–Shaw paleointensity method for analyzing baked clay samples collected from the floor of the Sayama Higashiyama-Oku kiln by comparing the results with those collated using the IZZI-Thellier method.

The intensity from the Tsunakawa–Shaw method was 48.1 ± 3.1 μT (N = 8) and that from the IZZI-Thellier method was 45.3 ± 4.4 μT (N = 3). The results from both methods coincide within the range of their standard deviations (1σ). Therefore, these experimental results suggest that the Tsunakawa–Shaw method is suitable for reconstructing paleointensity values from real archeological materials with high reliability.

From rock magnetic experiments, mineral analysis, and paleodirection measurements, it was revealed that the main magnetic mineral in the samples is titanomagnetite, the alteration during heating is very small, and the recorded NRM is the only single component of TRM acquired in the operating time at temperatures greater than 1000 °C. Thus, we confirmed that the samples used in this study are suitable for paleointensity experiments.

The two intensity values from the modern experimental methods are lower than those from previous studies in Japan.