Comment on ‘ ‘ Experimental reassessment of the Shaw paleointensity method using laboratory - induced thermal remanent

[1] Pan et al. [2002] experimentally investigated the validity of the Shaw paleointensity method [Shaw, 1974; Kono, 1978; Rolph and Shaw, 1985] using laboratoryinduced 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 mT 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 Awere heated to 600 C for 20 min in a laboratory field of 50 mT, set B comprised 20 samples that were heated to 600 C for 30 min in a field of 30 mT, and 20 set C samples were heated to 700 C for 40 min in a field of 50 mT. 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 mT), 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 Shawtype 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 mT 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 NRMTRM* 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]). [3] This erroneous correction is probably the result of calculation without vector subtraction. In Rolph’s correction method, individual TRM data for a certain coercivity are corrected by multiplying by the ratio of its equivalent anhysteretic remanent magnetizations (ARMs) given in the same field before and after the laboratory heating that produces the TRM. If the original and corrected TRMs for a certain coercivity Hc are defined as TRM[Hc] and TRM*[Hc], and the ARMs before and after the heating as ARM1[Hc] and ARM2[Hc], then TRM*[Hc] is calculated as follows: