Comment on “Experimental reassessment of the Shaw paleointensity method using laboratory‐induced thermal remanent magnetization” by Y. Pan, J. Shaw, R. Zhu, and M. J. Hill

[1] Pan et al. [2002] experimentally investigated the validity of the Shaw paleointensity method [Shaw, 1974; Kono, 1978; Rolph and Shaw, 1985] using laboratory-induced thermal remanent magnetization (TRM) to simulate natural remanent magnetization (NRM). They prepared fully demagnetized Cretaceous basalt samples by exposing them to a 150-mT alternating field (AF), and then heating them in an oven to 600°C for 20 min in a laboratory field of 50 μT to produce the simulated NRM. The samples were then divided into three sets (A, B, and C) and subjected to Shaw-type experiments. The TRM acquisition parameters differed among the sets: the 10 samples of set A were heated to 600°C for 20 min in a laboratory field of 50 μT, set B comprised 20 samples that were heated to 600°C for 30 min in a field of 30 μT, and 20 set C samples were heated to 700°C for 40 min in a field of 50 μT. After applying Rolph's correction [Rolph and Shaw, 1985], 29 out of the 30 samples in sets A and B yielded paleointensities close to the expected value (50 ± 5 μT), whereas only 9 out of the 20 samples in set C yielded values close to the expected. Since 8 out of the remaining 11 samples in set C showed incorrect intensities, Pan et al. [2002] concluded that monitoring rock magnetic properties such as the “P value” (defined below) was very important when applying the Shaw method with Rolph's correction. Although in my view it would be better to further examine the validity of Rolph's correction by double heating [Tsunakawa and Shaw, 1994], study by Pan et al. [2002] is important to reassess the reliability of Shaw-type experiments because of several recent criticisms [e.g., Goguitchaichvili et al., 1999; Vlag et al., 2000]. [2] However, because I do not consider the data processing used to construct the paleointensity plots of Pan et al. [2002] to be correct, different conclusions can be drawn from their study. For example, sample C20 yielded a value of 62.6 μT from the NRM-TRM* plot [Pan et al., 2002, Figure 2], which is 25% higher than the expected intensity. This plot showed excellent linearity in spite of the absence of linearity in the original NRM-TRM diagram. According to Pan et al. [2002], this linearity was produced by Rolph's correction. However, I do not consider the correction to be valid because the original NRM-TRM and ARM1-ARM2 plots (where ARM is anhysteretic remanent magnetization) had inverse convexity; in other words, they were dissimilar to each other. In Rolph's correction, linearity in an NRM-TRM* plot is generally produced by the resemblance between the NRM-TRM and ARM1-ARM2 plots (for example, sample B04 in set B [Pan et al., 2002, Figure 2]). [4] Contrary to these requirements, Pan et al. [2002] use nonsubtracted ARMs. Therefore the paleointensity results obtained by Rolph's correction are wrongly calculated. This is particularly important for set C, because a significant fraction of the TRMs in this set survived after the maximum AF cleaning step [Pan et al., 2002, Figure 2]. On the other hand, the paleointensities of the samples from sets A and B were not significantly affected because of the negligible amount of remanence after the maximum AF cleaning step. [5] Therefore the conclusion derived from set C by Pan et al. [2002] should be reconsidered after proper calculations. An example is shown for sample C20 (Figure 1): (1) the apparent magnitudes of NRM, TRM, ARM1 and ARM2 for various AF steps (NRM′[Hc], TRM′[Hc], ARM′1[Hc], and ARM′2[Hc]) were digitized from the original figure [Pan et al., 2002, Figure 2], and then the true remanences (NRM[Hc], TRM[Hc], ARM1[Hc], and ARM2[Hc]) were calculated by scalar subtraction of the apparent remanence at the maximum AF step (NRM′[120mT], TRM′[120mT], ARM′1[120mT], and ARM′2[120mT]) from that at the corresponding AF step (NRM′[Hc], TRM′[Hc], ARM′1[Hc], and ARM′2[Hc]); (2) TRM*[Hc] was then calculated by TRM[Hc] × (ARM1[Hc]/ARM2[Hc]). We did not perform vector subtraction because the remanences were presented not in vector but in scalar form in the original figure. The resultant diagram (Figure 1) does not have good linearity. It is quite different from the original figure [Figure 2; Pan et al., 2002], and I consider this difference to be mainly due to the erroneous correction. [7] A final comment is that the definition of the “P value” [Pan et al., 2002, section 4.2] seems invalid when considered in terms of the principles of ARM. Pan et al. [2002] defined the “P value” as the percentage difference in residual magnetization between ARM2 and ARM1 divided by ARM1. They gave those ARMs in a maximum peak AF field of 150 mT with an associated direct field of 100 μT. Since these remanences were subjected to stepwise AF demagnetization up to almost the same peak field, the ARMs are completely demagnetized in principle, so no “residual” ARMs should be observed. Therefore the observed remanences are not residual ARMs but remaining NRM or TRM. I think that the “P value” should be defined as the percentage difference in residual magnetization between TRM and NRM divided by NRM, that is (residual TRM - residual NRM)/(initial NRM) × 100. [8] I thank Nobutatsu Mochizuki, Tokyo Institute of Technology, for several valuable discussions. I also appreciate the constructive comments of Toshitsugu Yamazaki, Geological Survey of Japan, AIST. I acknowledge Carlo Laj for his comments regarding the English.