To understand the chemical reactions of clay minerals in a fault zone during an earthquake, we analyzed the clay minerals in the Chelungpu fault, which slipped during the 1999 Chi-Chi earthquake. X-ray diffraction spectroscopy showed that kaolinite and smectite contents were lower in the black gouge zone than in the surrounding gray gouge, breccia, or fracture-damaged zones. We applied a chemical kinetics approach to examine whether dehydroxylation of kaolinite and dehydration of interlayer water, dehydroxylation, and illitization of smectite occurred during coseismic frictional heating, and found that the first two reactions could complete under the temperature-time profile of the Chi-Chi earthquake, reconstructed by a previous study. Because dehydration of smectite interlayer water and dehydroxylation of kaolinite would have completed 3.6 and 8.2 s after the beginning of slip, the resulting release of water might have affected the frictional mechanism during the earthquake.
 Frictional heating frequently occurs during an earthquake, and it constitutes the largest part of the total seismic energy budget of the earthquake [e.g., Scholz, 2002]. Estimation of the temperatures reached during an earthquake can provide important information about possible faulting mechanisms, such as frictional melting or thermal pressurization, and dynamic shear stress during the earthquake. Not only production of pseudotachylyte [McKenzie and Brune, 1972] but also chemical reactions of minerals such as thermal decomposition of carbonate minerals [Han et al., 2007] have been reported to be induced by frictional heating during an earthquake, and investigations of these reactions are important for understanding the faulting mechanism.
Hirono et al. [2006a, 2006b, 2007a] reported high magnetic susceptibility values in the fault gouge of the Taiwan Chelungpu fault (Figure 1), which slipped during the 1999 Taiwan Chi-Chi earthquake (Mw 7.6) [Ma et al., 1999], and Mishima et al.  proposed that these high magnetic susceptibility values may have resulted from the formation of magnetic minerals from paramagnetic minerals at high temperature (> 400°C). Hirono et al. [2006a] reported that the inorganic carbon content is low within the same fault gouge of the Chelungpu fault, and suggested that thermal decomposition of carbonate minerals occurred during the earthquake. Hirono et al. [2007b] estimated the dynamic shear stress during the Chi-Chi earthquake from the chemical kinetics of the decomposition reaction of calcite, and suggested that a similar kinetic approach could be applied not only to thermal decomposition of carbonates but also to the dewatering of clay minerals. The dewatering of clay minerals during an earthquake might significantly affect the faulting mechanism, because clay minerals can release water rapidly. Therefore, in this study, we first analyzed relative amounts of clay minerals within the Chelungpu fault, and then used a chemical kinetics approach to investigate whether dewatering could have occurred during the Chi-Chi earthquake. We also discuss the amount of water that might have been expelled from clay minerals during the earthquake.
2. Clay Mineral Analyses of the Taiwan Chelungpu Fault Zone
 The Taiwan Chelungpu-fault Drilling Project (TCDP) was undertaken to investigate the faulting mechanism during the Chi-Chi earthquake, and core samples were recovered from two holes: Hole A (total depth, 2003.00 m) and Hole B (total depth, 1352.60 m). In Hole B (Figure 1), the depth interval from 948.42 to 1040 m stratigraphically corresponds to the Pliocene to Pleistocene Cholan Formation, which is composed dominantly of sandstone and a sandstone-siltstone alternation with weak to heavy bioturbation. The Pliocene Chinshui Shale occurs from 1040 to 1280 m depth and consists predominantly of siltstone with weak bioturbation. From 1280 to 1352.60 m depth is the late Miocene to early Pliocene Kueichulin Formation, which is composed dominantly of massive sandstone with minor siltstone. Hirono et al. [2006b, 2007a] reported three dominant fault zones, FZB1136 (i.e., fault zone around 1136 m depth in Hole B), FZB1194, and FZB1243, within the Chinshui Shale (Figure 1) and interpreted them as segments of the Chelungpu fault. Because FZB1136 has been identified as the zone that most likely slipped during the 1999 Chi-Chi earthquake [Kano et al., 2006; Ma et al., 2006; Wu et al., 2007], we collected a total of 14 samples from that fault zone (Figure 1).
 We used X-ray diffraction spectroscopy (Spectris PANalytical, X'Pert PRO MPD) to analyze the clay minerals in the samples. First, the samples were powdered in an agate mortar, and then the <2-μm fraction was separated from coarser material. Oriented slides of the clay-sized (<2-μm) fraction were saturated with ethylene glycol vapor before analysis. All scans were run at 45 kV/40 mA with CuKa radiation at the following machine settings: divergence slit, 1°; receiving slit, 0.1 mm; and step width, 0.01°. We calculated the relative abundances of smectite, illite, kaolinite, and chlorite by using Biscaye's  method (Figure 1).
 In the fracture-damaged and breccia zones, the averaged abundances of smectite, illite, kaolinite, and chlorite were 6.0%, 64.1%, 10.1%, and 19.8%, respectively; in the gray gouge zone, they were 6.1%, 71.0%, 7.3%, and 15.6%, respectively; and in the black gouge zone, they were 3.0%, 87.0%, 0.5%, and 9.5%, respectively (Figure 1). The black gouge zone within FZB1136 was characterized by the disappearance of kaolinite, lower relative abundances of smectite and chlorite, and a greater relative abundance of illite.
3. Chemical Kinetics of Clay Mineral Reactions
Hirono et al. [2006a, 2006b, 2007a] and Mishima et al.  reported relatively higher magnetic susceptibility values and a relatively low carbonate mineral content in the same black gouge zone within FZB1136, and they suggested that frictional heating had occurred in that zone during the Chi-Chi earthquake. Taking our results and these previous findings into consideration, we hypothesize that the changes observed in the clay mineral content of the black gouge zone were caused by dehydroxylation of kaolinite and dehydration of interlayer water, dehydroxylation, and illitization of smectite at high temperature. To assess whether these reactions could have occurred during the earthquake, we calculated the chemical kinetics of each reaction and compared the calculated results with the analyzed abundance of each clay mineral.
 The reaction mechanism and the kinetic parameters of the dehydroxylation of kaolinite and dehydration of interlayer water and dehydroxylation of smectite have been extensively studied [e.g., Girgis et al., 1987]. The kinetics of a chemical reaction are commonly expressed as follows:
where α is the reacted fraction of a substance (0 ≤ α ≤ 1, α = 1 if the entire substance is reacted), t is the reaction time, k is the reaction rate, and f(α) is the kinetic function determined by the reaction mechanism. If the reaction is carried out in an isothermal manner, the integrated form of equation (1) is
where g(α) is also a kinetic function determined by the reaction mechanism. We adopted a first-order reaction for the dehydroxylation of kaolinite [Saikia et al., 2002] and the dehydration of interlayer water and dehydroxylation of smectite [Bray and Redfern, 1999; Guler and Sarier, 1990], and f(α) and g(α) are expressed as
 The relationship between the reaction rate and temperature, as expressed by the Arrhenius equation, is
where A is a constant (pre-exponential term), Ea is the activation energy necessary for a reaction to occur, R is the gas constant (8.31447 JK−1mol−1), and T is temperature (K). From equations (2), (4), and (5), the relationship between reaction time and temperature in the isothermal condition is therefore
Saikia et al.  reported that for dehydroxylation of kaolinite, the values of the parameters of the Arrhenius equation are A = 9.57 × 108 s−1 and Ea = 196.8 kJ mol−1. Bray and Redfern  and Guler and Sarier  reported that Ea was 32.2 and 50.7 kJ mol−1 for dehydration of interlayer water and dehydroxylation of smectite, respectively, and we determined the respective values of A (553 and 8.80 s−1) graphically from their figures showing the relationship between the reaction rate and temperature because their exact values were not reported. The remaining unknown kinetics parameters are thus α, time, and temperature. If we assume that α = 0.99 (most of the substance has reacted), then Figure 2 shows graphically how the time and temperature of the reactions are related.
 On the other hand, the smectite illitization reaction is expressed as follows [Huang et al., 1993]:
where S is the molar fraction of smectite and [K+] is the concentration of dissolved potassium (M). If this reaction is carried out in an isothermal manner, the integration of equation (7) gives
where S0 is the initial molar fraction of smectite. From equations (5) and (8), the relationship between reaction time and temperature in the isothermal condition is therefore
Here we assumed that S0 = 1 (the initial substance before the reaction is only smectite), Ea = 28 kJmol−1, and A = 8.08 × 10−4 s−1M−1 [Huang et al., 1993]. Because the concentration of potassium in the interstitial fluid within the Chelungpu fault has unfortunately not been analyzed, we assumed it to be the average value in seawater, 9.7 × 10−3 M [Millero, 1996] or, as an upper limit, the saturated concentration with respect to potassium chloride in pure water at 48°C, 5.7 M [Sunier and Baumbach, 1976]. Therefore, the relationship between the time and temperature of the reactions, assuming S = 0.01 (99% of the smectite converted to illite), is as shown in Figure 2.
4. Clay Mineral Reactions During an Earthquake
 At a relatively low temperature (e.g., <200°C), only dehydration of interlayer water of smectite can occur during a short period (e.g., <1000 s), whereas the other reactions require a longer time (e.g., >105 s) (Figure 2). At higher temperatures (e.g., >800°C), dehydroxylation of smectite and kaolinite can also occur during a short period, but illitization of smectite requires a much longer time (>105 s) at temperatures from 100 to 1100°C.
 Because the temperature in a fault zone during an earthquake is not constant but changes with time because of frictional heating and heat conduction, quantitative estimation of the reacted fraction is necessary to construct a realistic temperature-time profile. From equations (1) and (3), the fraction of reacted matter, α, for dehydration of interlayer water and dehydroxylation of smectite and dehydroxylation of kaolinite is expressed as
The reaction rate, k(t), is not constant, but changes with time. This rate also depends on the temperature, as described by the Arrhenius equation (equation (5)), so that α is expressed as
where T(t) is the temperature (K) in the fault zone. On the other hand, from equation (7), the fraction of reacted matter, 1-S, for illitization of smectite is expressed as
The reaction rate, k(t), changes with time and is related to temperature as described by the Arrhenius equation (equation (5)). Therefore, 1-S is expressed as
Hirono et al. [2007b] reconstructed the temperature-time profile in the black gouge zone within FZB1136 during the Chi-Chi earthquake from the chemical kinetics of the thermal decomposition of carbonate minerals (Figure 3a). We adopted this profile for use with equations (11) and (13), and then solved for the values of α and 1-S by the quadrature method. For the other parameters we used the values as indicated in section 3.
 The resulting changes of α and 1-S with time are shown in Figure 3b. The dehydration of interlayer water of smectite reaction can complete in 3.6 s (α = 0.99), dehydroxylation of kaolinite can complete in 8.2 s (α = 0.99), and dehydroxylation of smectite can complete in 5.4 × 107 s (α = 0.99). However, illitization of smectite does not complete; after 1.58 × 108 s (the period between the Chi-Chi earthquake and sample recovery from Hole B), the calculated reacted fractions are 0.038 for [K+] = 9.7 × 10−3 M and 0.96 for [K+] = 5.7 M.
5. Discussion and Conclusions
 The results of these kinetic analyses of clay mineral reactions are consistent with the observed disappearance of kaolinite in the black gouge zone within FZB1136. The disappearance of kaolinite might thus have been caused by dehydroxylation because of frictional heating during the earthquake. After dehydroxylation, the decomposed kaolinite may remain as secondary minerals such as amorphous aluminosilicates [e.g., Temuujin et al., 1998]. The lower abundance of smectite might also have been caused by dehydroxylation. The higher abundance of illite might have been caused by illitization of smectite. The illitization reaction is strongly related to the concentration of potassium (see equation 13). If the concentration is higher than the value in seawater, illitization will be enhanced. However, the concentration of potassium in the interstitial fluid within the Chelungpu fault has unfortunately not been analyzed, so we cannot strictly be certain of the cause of the higher abundance of illite. We were unable to find information on the kinetics of chlorite dehydroxylation or its Arrhenius parameters in the literature; thus, we cannot discuss the cause of the lower abundance of chlorite.
 Because dehydration of interlayer water of smectite and dehydroxylation of kaolinite could almost complete within the frictional heating stage, that is, during the fault slip (Figure 3b), their abrupt release of water might have affected the frictional behavior dynamically. We here estimate the possible amounts of expelled water. Smectite generally contains two molecules of H2O within the interlayer space, and the H2O is released by dehydration. Smectite, with the chemical formula Na0.33(Al1.67Mg0.33)Si4O10(OH)2˙2H2O, has a molecular mass of 403.0, so the mass ratio of expelled H2O (molecular mass of 2H2O = 36.0) from the interlayer to 1 mole of smectite is 0.0893. The relative amount of smectite in the black gouge zone averaged 3.0 wt %, but the absolute clay mineral content of the bulk sample was not measured. We thus assumed various values for the bulk clay mineral content, and estimated the amount of expelled H2O as a result of coseismic frictional heating (Table 1). On the other hand, 4 moles of OH (thus, 2 moles of H2O) is produced by dehydroxylation of 1 mole of kaolinite. For kaolinite with the chemical formula Al2Si2O5(OH)4, the mass ratio of expelled H2O to 1 mole of kaolinite is 0.139 (molecular mass of kaolinite, 258.2). If we assume that the initial relative amount of kaolinite in the black gouge zone was equal to the averaged values in the fracture-damaged and breccia zones, that is, 10.1 wt %, then the difference between the initial content and measured content (0.5 wt %) in the black gouge zone (10.1 − 0.5 = 9.6 wt %) corresponds to the amount of dehydroxylated kaolinite. Assuming various values for the absolute clay mineral content of the bulk sample, we estimated the amount of H2O expelled from kaolinite (Table 1). Water may thus have been produced in the black gouge zone by coseismic frictional heating.
Table 1. Estimation of the Amount of H2O Expelled by Dehydration of Interlayer Water of Smectite and by Dehydroxylation of Kaolinitea
Clay Mineral Content of the Bulk Sample
20 wt %
40 wt %
60 wt %
80 wt %
Assuming a sample density of 2.2 g/cm3 [from Hirono et al., 2006b], the amount of H2O per 1 cm3 was calculated.
Smectite content (wt %)
Interlayer H2O (wt %)
5.36 × 10−2
1.07 × 10−1
1.61 × 10−1
2.14 × 10−1
Interlayer H2O (g)
1.18 × 10−3
2.36 × 10−3
3.54 × 10−3
4.72 × 10−3
Amount of kaolinite that disappeared (wt %)
H2O from kaolinite (wt %)
2.67 × 10−1
5.34 × 10−1
8.01 × 10−1
H2O from kaolinite (g)
5.87 × 10−3
1.17 × 10−2
1.76 × 10−2
2.35 × 10−2
 Whether these clay mineral reactions occur may depend upon whether conditions are dry or wet and whether the system is open (drained) or closed (undrained) with respect to water. Under dry conditions in an open system with respect to H2O, dehydration of interlayer water and hydroxylation of clay minerals can certainly occur, but under wet conditions in a closed system, these reactions may be prohibited. In general, under dry conditions, a system is open with respect to H2O, whereas under wet conditions, the system is likely to be closed. In a fault zone, these conditions depend strongly on mass transfer properties, such as permeability, of the fault gouge and the surrounding wall rock. Unfortunately, we have no in situ data as to whether the black gouge zone was dry or wet or whether the system was open or closed at the time of slip. Thus, in the strict sense, our kinetic modeling result includes uncertainty. Further investigation of in situ conditions in the fault zone and quantitative estimation of the conditions at the time of slip, whether dry or wet in an open or closed system, are necessary in the future.
 Finally, we emphasize that a kinetics approach based on chemical reactions may be a valid and useful method for explaining the abundance of clay minerals in the fault gouge, and, moreover, we propose that the abrupt release of water from clay minerals can affect the frictional mechanism during an earthquake. For more accurate estimation, however, it is important to determine the reaction mechanisms, their expression in terms of f(α) and g(α), and the parameters of the Arrhenius equation (A and Ea) under in situ conditions in the fault zone. Experimental or theoretical demonstration of the dynamic friction weakening during an earthquake is necessary.
 We thank Yakov Khazan and anonymous reviewers for their constructive comments, and editors Fabio Florindo and Eric Calais for editing this paper.