Healing of gabbro-built faults under hydrothermal conditions

Fault frictional healing Δμ controls the storage of the elastic strain energy in the fault wall rocks and the re-occurrence of earthquakes in pre-existing faults. In the last 40 years, fault healing has been investigated with laboratory slide-hold-slide (SHS) experiments aimed at reproducing the seismic cycle. Experiments performed with different rocks types (e.g., granite, limestone, basalt) revealed that (1) Δμ increases with hold time t_h and, (2) the frictional healing rate β =Δμ/log t_h >0. This increase in fault frictional strength with t_h is interpreted as due to the increase (1) in the real area of contact or (2) of chemical bond strength. However, most of these experiments were conducted under room conditions, whereas natural earthquakes generally nucleate at ambient temperatures T  >150℃ and in the presence of pressurized fluids. Under these ambient conditions, fluid-assisted and thermally-activated processes (pressure-solution transfer, stress corrosion, etc.) may impact on the magnitude of Δμ and on β.

In this study, SHS experiments were performed on gabbro-built gouges (grain size <88 mm) in a rotary shear machine equipped with a pressurized vessel to explore frictional healing under hydrothermal conditions. All experiments were conducted at a constant effective normal stress (σ_eff =50MPa), and temperature (T) ranging from 25 to 400 ℃  under dry or pore fluid (deionized H₂O) pressure (P_f=30 MPa) conditions. In the SHS sequence, the imposed slip velocity was V=10 μm/s, and hold time t_h varied from 3 to 10000 s. For each experiment, two SHS sequences separated by a slip displacement interval of 40 mm were conducted.

Under dry conditions at all tested temperatures and under hydrothermal conditions but at T  <100℃, Δμ increases with t_h, consistent with previous experiments. Moreover, the Δμ and β values in the 2^nd SHS sequence are slightly higher than those in the 1^st sequence, possibly due to the smaller grain size at the larger displacement that promotes fault healing. By contrast, in the experiments performed under hydrothermal conditions but T >200℃, Δμ decreases and β switches to negative values (<0) when the hold is longer than a threshold hold time. In detail, at T=300℃: β= 0.0161±0.0017 for holds <300s and -0.0074±0.0043 for holds >300s, and at T=400℃: β= 0.0057±0.0020 for holds <100s and -0.0227±0.0042 for holds >100s.

The underlying mechanism responsible for the decrease in Δμ and the transition from β > 0 to β < 0 with the hold time, which could result in the transition from seismic to aseismic fault behavior in nature, is still poorly understood. However, high-resolution microstructural analyses conducted by scanning electron microscopy on experimental fault products rule out the formation of weak minerals (e.g., clays) in the gouge layer.  Consequently, the weakening of the fault is probably related to the decrease in bond strength at the asperity contacts.

The experimental data presented here suggest that fault healing of natural faults is controlled by the feedback of multiple physico-chemical processes associated with the slip history and type of fluid-rock interaction under hydrothermal conditions.