Ultralow Friction of Carbonate Faults Caused by Thermal Decomposition

See allHide authors and affiliations

Science  11 May 2007:
Vol. 316, Issue 5826, pp. 878-881
DOI: 10.1126/science.1139763

This article has a correction. Please see:


High-velocity weakening of faults may drive fault motion during large earthquakes. Experiments on simulated faults in Carrara marble at slip rates up to 1.3 meters per second demonstrate that thermal decomposition of calcite due to frictional heating induces pronounced fault weakening with steady-state friction coefficients as low as 0.06. Decomposition produces particles of tens of nanometers in size, and the ultralow friction appears to be associated with the flash heating on an ultrafine decomposition product. Thus, thermal decomposition may be an important process for the dynamic weakening of faults.

The strength of seismogenic faults, which is frictional resistance to fault slip during earthquakes, has been a major subject of debate in fault mechanics for 30 years (1, 2). Although the stress–heat flow paradox for the San Andreas fault (no heat-flow anomaly, contrary to the prediction from in situ stress measurement and laboratory data of rock friction) favors extremely low fault strength (3, 4), reasons for the weakness have been unclear. Recent work has shown that the dynamic weakening of faults during seismic slip can be caused by mechanisms such as frictional melting (59), thermal pressurization (1013), and silica-gel formation (14, 15). Fault gouge was also shown to exhibit pronounced slip weakening at high slip rates (16), presumably because of flash heating (13). Some analyses have predicted that slip-weakening distance, over which the initial peak friction drops to steady-state dynamic friction (8, 12), and fracture energy (13, 16) are of the same order as those parameters that are determined seismologically, narrowing the gap between laboratory studies of fault mechanics and seismology. Modeling the generation of large earthquakes is now becoming possible on the basis of the measured mechanical and transport properties of fault zones. Moreover, the dynamic weakening of faults may explain the lack of heat-flow anomaly after earthquake events along the San Andreas fault.

Thermal decomposition of rock-forming minerals at high ambient temperature and pressure can dramatically lower the strength of rocks because of the buildup of pore fluid pressure and the associated reduction of effective normal stress, provided that the sample is effectively undrained (17). Even at shallow crustal levels with low ambient temperature, thermal decomposition may occur at an elevated temperature because of coseismic frictional heating along fault zones. We demonstrated that a carbonate fault can lose frictional strength almost completely because of the thermal decomposition of calcite caused by frictional heating during high-velocity friction experiments on Carrara marble at seismic slip rates.

Forty-two friction experiments were conducted on precut bare surfaces of a pair of solid cylindrical specimens of Carrara marble (∼99% calcite) at room temperature and room humidity. The experiments were carried out at normal stresses of 1.1 to 13.4 MPa and at equivalent slip rates of 0.03 to 1.30 ms–1, with a rotary-shear, high-velocity friction apparatus at Kyoto University (18). The diameter and length of the specimen were 21.8 to 24.8 mm and about 20 mm, respectively (19). Because there is a slip-rate gradient across the fault due to cylindrical specimen geometry, we use the term “equivalent slip rate” (“slip rate” or “velocity” hereafter) (7, 18, 19).

At different slip rates (Fig. 1A), the friction coefficient of the specimens decreased nearly exponentially from peak friction, μp, to nearly steady-state friction, μss (20), with increasing displacement. The slip-weakening distance ranged from a few to several meters. The μss decreased markedly from ∼0.6 to a range of 0.04 to 0.11 with an increase in slip rate (Fig. 1B). A linear friction law holds for both peak and steady-state friction, yielding μp of 0.60 ± 0.01 and μss of 0.06 ± 0.01 (Fig. 1C). This μp value is more or less typical for marble (0.4 to 0.8) from conventional slow slip-rate (<1 mm s–1) experiments (21). But μss at a high slip rate (1.1 to 1.2 m s–1) was extremely low. In contrast, μss values remained high (0.46 to 0.63) at slow slip rates (0.03 to 0.08 m s–1).

Fig. 1.

Frictional properties of simulated faults in Carrara marble at subseismic to seismic slip rates. (A) Friction coefficient versus fault displacement for five runs conducted at different slip rates and at a normal stress of 7.3 MPa (except for HVR522 at 4.9 MPa). The dashed black rectangle shows an example of the range of data used for the estimation of steady-state friction. (B) μss plotted against the slip rate for 10 runs conducted at a normal stress of 7.3 MPa. Vertical bars show the SD of μss (shown only when the SD is greater than the box size). (C) Shear stresses plotted against normal stresses at peak and steady-state friction at slip rates of 1.14 to 1.18 m s–1. Open squares and circles indicate initial peak friction and steady-state friction, respectively. The slopes of the lines give frictional coefficients at peak and steady-state friction. τ, shear stress; σn, normal stress.

Microstructural observations and electron probe microanalysis of deformed specimens were conducted on thin sections normal to the fault and parallel to the slip direction. We started experiments with precut surfaces of Carrara marble, but gouge zones formed very quickly on both sides of the slip surface, which nearly coincides with the precut surface (Fig. 2A). Samples were collected from the slip surface for observations with a field-emission scanning electron microscope (FE-SEM) and a transmission electron microscope (TEM) and for x-ray diffraction (XRD) analyses. Those analyses revealed that the gouge zone forms at a very early stage of displacement (<2 m), when the outer and inner zones of gouge consist of calcite and lime (CaO) and/or hydrated lime [Ca(OH)2], respectively. In the host rock adjacent to the fault gouge, fracturing of calcite grains occurred, and the size of calcite fragments decreased toward the slip surface to become calcite gouge. Calcite thermally decomposes into lime and CO2 gas at about 720 to 900°C (22, 23), and the lime can be transformed to hydrated lime by absorbing moisture when it is exposed to the atmosphere. Thus, thermal decomposition of calcite occurred from a very early stage of slip.

Fig. 2.

Textures and decomposition products in fault zones. (A) A cross–polarized light photomicrograph of a thermally induced decomposition zone developed in Carrara marble, deformed at a slip rate of 1.30 m s–1 and at a normal stress of 4.6 MPa (HVR398, total slip = 125 m). DFs are between the wall rock and the decomposition zone. Slip was highly localized along the surface denoted by “slip surface.” (B and C) SEM and TEM photomicrographs, respectively, illustrating microstructures of the lime (CaO) aggregates on the slip interface. (D) XRD spectra of an undeformed specimen, a preheated specimen, and specimens deformed at different conditions. Diffraction intensity is shown in 1000 × counts per second (cps) on all vertical axes.

Many fractures were present over a wide decomposition zone between the decomposition fronts (DFs) (Fig. 2A); those fractures must have increased permeability. The decomposed zone consisted of grainlike aggregates ranging from about 100 to a few hundred nanometers in diameter (Fig. 2B), but each aggregate was composed of ultrafine grains that were several to a few tens of nanometers in size (Fig. 2C and fig. S3). Calcite decomposition was confirmed in specimens at fast slip rates (>0.4 m s–1), but no evidence of decomposition was present in a specimen deformed at a slow slip rate of 0.08 m s–1 (run number HVR522; compare XRD curves in Fig. 2D). We did not recognize glass or amorphous material in any of the specimens (fig. S3).

To determine the timing of decomposition with respect to the slip-weakening behavior, we measured the emission of CO2 released from a deforming sample by using two solid electrolyte-type CO2 sensors (19). Sensor 1 (without a filter) was set very close to the fault (about 30 mm away) to detect the onset of decomposition. It took about 0.9 to 1.0 s for this sensor to begin to detect CO2. Sensor 2 (with a filter) had an initial response time of about 2 s and a 90% response time of ∼90 s. This sensor was used to determine the total amount of CO2 emission. At a high normal stress (12.2 MPa) and the largest slip rate (1.17 m s–1), μp of about 0.6 dropped to μss of about 0.04 (Fig. 3A). Sensor 2 showed a continuous increase toward a CO2 concentration of about 29,000 parts per million. This concentration roughly agrees with the expected emission of CO2 from the volume of decomposed calcite that was estimated on photomicrographs. The output from sensor 1 showed a reduction of CO2 concentration after the end of a run because of dissipation of concentrated CO2 near the fault (Fig. 3A). The first detectable output from sensor 1 was recognized at 1.0 s (Fig. 3B). The initial response time was 0.9 to 1.0 s for this sensor, so there must have been an emission of CO2 almost immediately after the onset of slip (24). Another run at a lower slip rate (0.17 m s–1) and at a normal stress of 9.8 MPa also indicated that the slip weakening was concurrent with the thermal decomposition of calcite (fig. S4).

Fig. 3.

Monitoring of CO2 gas emission and temperature measurement. (A) Friction coefficient and outputs from two CO2 sensors plotted against time. Sensor 1 (without a filter) is a quick-response sensor for detecting the onset of CO2 emission, and sensor 2 (with a filter) is a slow-response sensor for monitoring the amount of emitted CO2. V, slip rate. (B) Enlargement of (A) for the first 5 s of slip, with an output only from sensor 1 shown with mechanical data. The timing of peak friction and the first detectable change in output from sensor 1 are indicated by arrows. Vertical gray bars in both (A) and (B) indicate the response time of sensor 1 (about 1 s). (C) Friction coefficient (black), slip rate (blue), and temperature measured with a radiation thermometer (red) plotted against time in HVR511, which was conducted at about the same sliding condition (normal stress and slip rate of 12.1 MPa and 1.18 m s–1, respectively) as that in HVR601. (D) Enlargement of the decreasing slip-rate phase of the experiment after turning off the magnetic clutch in (C).

Thus, a fault in Carrara marble clearly shows pronounced slip weakening at high slip rates while undergoing calcite decomposition along the slip surface. This means that a fault can become very weak as a result of frictional heating caused by its own motion. Such dynamic weakening should destabilize fault slip and foster the generation of large earthquakes (13). We next address the question of what causes such a pronounced weakening of a fault. One may consider that a buildup of pore pressure in a fault zone owing to the release of CO2 gas is the most likely cause for fault weakening. To test this possibility, we conducted a critical experiment using preheated and decomposed specimens of Carrara marble (left in an oven at 900° to 904°C for 1.5 hours). We quantitatively confirmed complete decomposition by measuring weight loss after heating and by XRD analyses (Fig. 2D). The decomposed specimens could no longer emit CO2 gas (19). The behavior is markedly similar between faults in Carrara marble and in decomposed specimens (Fig. 4). We have not measured the permeability of the decomposed zone in marble yet. But fractures in the decomposed zone in Fig. 2A suggest that the permeability of the decomposed zone is large enough for CO2 to escape and to prevent the buildup of high pore fluid pressures. Enhanced permeability during the dehydration of serpentinite at elevated ambient temperature (not due to frictional heating) was also confirmed recently (25).

Fig. 4.

Comparison of frictional behavior of a fault in Carrara marble (ordinary) and a fault in preheated and completely decomposed specimens of Carrara marble.

These results indicate that the weakening is attributed not to CO2 pressure but to the low frictional strength of newly formed ultrafine lime grains. Among other possibilities, frictional melting can be immediately removed because calcite decomposition occurs before melting. Also, CaO melting would be unlikely because its melting temperature (∼2572°C) is much higher than the temperatures recorded at the slip interface. Indeed, we did not detect any glass or amorphous materials in fault zones (fig. S3). Wrinkle-like pulses or normal separation of a fault along a bimaterial interface (26) is also unlikely because there is no material contrast across a fault in our experiments.

In view of the existing data, we considered that flash heating (13) at interfaces of ultrafine particles is critical for pronounced weakening of carbonate fault. The weakening by flash heating or transient local heating at asperity contacts has been proposed to occur via local melting at asperity contacts and/or by strength degradation of asperity contacts at submelting temperatures (13). The latter case is more likely for decomposed calcite because lime has a very high melting temperature. However, exact deformation mechanisms along sliding asperity or grain contacts still remain to be explored.

To demonstrate the importance of temperature rise during high-velocity sliding, we measured the temperature along the fault of the specimens (Fig. 3, C and D), using a radiation thermometer (19, 27) during a run conducted at about the same sliding condition as that in HVR601. The thermometer measured an average temperature higher than 550°C over an area of 0.4 mm in diameter with a fast response time (<0.1 s). The measured temperature reached a maximum of 950°C at 30 s (high enough for calcite decomposition) after attaining about 650°C during the first 9 s (Fig. 3C). The local temperature at sliding asperities should be higher than the measured surface temperature even in the early stage (<9 s), in view of the CO2 emission data (Fig. 3, A and B). The friction and thermal evolution in the final stages of the experiment are very interesting. After the specimen was disconnected from the motor, fault slip decelerated and stopped in about 3 s (Fig. 3, C and D). Friction increased at an accelerating rate as the temperature fell. The inverse relation between friction and temperature strongly suggests that the immediate strength recovery could be related to a rapid drop in temperature.

The simulated faults of Carrara marble exhibit lower friction than does the Nojima fault gouge, although their overall behaviors are similar (16). A possible reason for this difference is a very effective production of ultrafine grains that are tens of nanometers in diameter (Fig. 2C) by a decomposition reaction (no time for grain growth during seismic-fault motion). The process may be similar to the cases of intermediate- to deep-focus earthquakes, for which the formation of ultrafine reaction products may play a decisive role in earthquake generation (28). Other gouge materials have to undergo grain comminution to form ultrafine grains, which requires extra work in fault zones, resulting in higher friction. For slip on faults in Carrara marble, understanding friction between nanometer-scale particles seems to be a key for delineating the exact mechanisms of the dynamic weakening of faults.

Our results have important implications for earthquake geology and fault mechanics. Marked decomposition weakening may be a widespread phenomenon, because fault gouges commonly contain sheet silicate minerals that decompose even at lower temperatures than that for calcite decomposition, although thermal decomposition of sheet silicates may be followed by frictional melting (29). Also, thermally induced decomposition may leave geological evidence (other than pseudotachylytes) of seismic-fault slip, contrary to geologists' opinion that faults do not preserve a record of seismic slip, except for the small percentage of faults containing pseudotachylyte (30). Indeed, we have shown that coseismic decomposition of siderite produces a stable mineral, magnetite (31). Thus, the clear demonstration of thermal decomposition during seismic slip opens up a new series of investigations in integrated fault and earthquake studies.

Supporting Online Material

Materials and Methods

Figs. S1 to S4


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article