The ξ potential has been inferred from streaming potential measurements with crushed rock samples as a function of pH and electrolyte concentration for various salts. The value obtained for crushed Fontainebleau sandstone at pH = 5.7 and a KCl solution with a resistivity of 400 Ω m is −40±5 mV, where the error is dominated by sample to sample variations. The sensitivity of the ξ potential to the electrolyte resistivity for KCl is given experimentally by ρf0.23±0.014 where ρf is the electrolyte resistivity. The point of zero charge (pzc) is observed for pH = 2.5±0.1, and the ξ potential is positive for pH < pzc and negative for pH > pzc. For pH > 5 the variations of the ξ potential with pH can be approximated by ξ(pH)/ξ(5.7) = 1 + (0.068±0.004)(pH‐5.7) for ρf = 100 Ω m. The ξ potential has been observed to be sensitive to the valence of the ions and is approximately reduced by the charge of the cation, unless specific adsorption takes place like in the case of Al3+. The experimental results are well accounted for by a three‐layer numerical model of the electrical double layer, and the parameters of this model can be evaluated from the experimental data. The sensitivity of the ξ potential to the rock minerals has also been studied. The ξ potential obtained for granitic rocks is comparable to that obtained for Fontainebleau sandstone but can be reduced by a factor of 2–4 for sandstones containing significant fractions of carbonates or clay. To take into account the effect of the chemical composition of the electrolyte, a chemical efficiency is defined as the ratio of the ξ potential to the ξ potential measured for KCl. This chemical efficiency is measured to be ∼80% for typical groundwater but can be as low as 40% for a water with a high dissolved carbonate content. The set of empirical laws derived from our measurements can be used to assess the magnitude of the streaming potentials expected in natural geophysical systems.

Streaming potential and resistivity measurements have been performed on Fontainebleau sandstone and Villejust quartzite samples in a triaxial device during compaction, uniaxial compression, and rupture. Measurements on individual samples do not show any clear intrinsic dependence of the streaming potential coefficient with permeability. An apparent dependence of the streaming potential coefficient with permeability is, however, observed during deformation. The effect of surface conductivity is taken into account and is small compared with the observed changes in the streaming potential coefficient. The observed dependence is therefore interpreted in terms of a difference in the evolution of the electrical and hydraulic connectivity patterns during deformation. This effect causes the streaming potential coefficient, and consequently the inferred ξ potential, to be reduced by a geometrical factor R G representing the electrical efficiency of the hydraulic network. Estimates of the R G factor varying between 0.2 and 0.8 for electrolyte resistivity larger than 100 Ωm are obtained by comparing the values of the ξ potential inferred from intact rock samples with the values obtained from crushed rock samples, where the geometrical effects are assumed to be negligible. The reduction of the streaming potential coefficient observed during compaction or uniaxial compression suggests that the tortuosity of the hydraulic network increases faster than the tortuosity of the electrical network. Before rupture, an increase in the streaming potential coefficient associated with the onset of dilatancy was observed for three samples of Fontainebleau sandstone and one sample of Villejust quartzite. The changes in streaming potential coefficient prior to failure range from 30% to 50%. During one experiment, an increase in the concentration of sulfate ions was also observed before failure. These experiments suggest that observable streaming potential and geochemical variations could occur before earthquakes.

Electric potential variations have been recorded from November 1995 to February 1996 and continuously since October 1996 at 14 measurement points on a one km wide ridge separating two lakes in the French Alps. The levels of the lakes vary by several tens of meters on a yearly cycle, inducing stress variations and fluid percolation. At one point, unambiguous variations as large as 120 mV are observed over a year, linearly correlated with the levels of the lakes with a magnitude of 2 mV per meter of water level change. This particular measurement point lies at the edge of a SP anomaly, which supports the presence of a localized zone of ground water flow forced by the lake level, suggesting an electrokinetic mechanism. The observed correlation implies a ζ‐potential of the order of ‐8 mV for a 60 Ωm electrolyte, in agreement with laboratory measurements.