TY - JOUR
AU - Spitzer, Klaus
AB - Summary The spectral complex conductivity of a water-bearing sand during interaction with carbon dioxide (CO2) is influenced by multiple, simultaneous processes. These processes include partial saturation due to the replacement of conductive pore water with CO2 and chemical interaction of the reactive CO2 with the bulk fluid and the grain-water interface. We present a laboratory study on the spectral induced polarization of water-bearing sands during exposure to and flow-through by CO2. Conductivity spectra were measured successfully at pressures up to 30 MPa and 80 °C during active flow and at steady-state conditions concentrating on the frequency range between 0.0014 and 100 Hz. The frequency range between 0.1 and 100 Hz turned out to be most indicative for potential monitoring applications. The presented data show that the impact of CO2 on the electrolytic conductivity may be covered by a model for pore-water conductivity, which depends on salinity, pressure and temperature and has been derived from earlier investigations of the pore-water phase. The new data covering the three-phase system CO2-brine-sand further show that chemical interaction causes a reduction of surface conductivity by almost 20 per cent, which could be related to the low pH-value in the acidic environment due to CO2 dissolution and the dissociation of carbonic acid. The quantification of the total CO2 effect may be used as a correction during monitoring of a sequestration in terms of saturation. We show that this leads to a correct reconstruction of fluid saturation from electrical measurements. In addition, an indicator for changes of the inner surface area, which is related to mineral dissolution or precipitation processes, can be computed from the imaginary part of conductivity. The low frequency range between 0.0014 and 0.1 Hz shows additional characteristics, which deviate from the behaviour at higher frequencies. A Debye decomposition approach is applied to isolate the feature dominating the data at low frequencies. We conclude from our study that electrical conductivity is not only a highly sensitive indicator for CO2 saturation in pore space. When it is measured in its full spectral and complex form it contains additional information on the chemical state of the system, which holds the potential of getting access to both saturation and interface properties with one monitoring method. Electrical properties, Gas and hydrate systems, High-pressure behaviour 1 INTRODUCTION Electrical conductivity is known to be highly sensitive to changes in the pore filling of a rock and, consequently, electromagnetic methods are a promising approach for a non-invasive monitoring of geotechnologies such as carbon dioxide (CO2) sequestration but also deep geothermal energy. The key of every electromagnetic monitoring is the knowledge of how the measured physical quantities relate to the desired reservoir parameters (e.g. porosity, gas saturation, pore water composition, hydraulic permeability, mineral dissolution/precipitation processes). It is the task of petrophysical investigations to reveal these relationships and to provide models for the direct and robust derivation of reservoir characteristics from geophysical measurements. The first and most fundamental petrophysical relationship between reservoir properties and electrical measurements is Archie's law, which describes the electrolytic conduction with a direct proportionality between electrical conductivity and the conductivity of the pore-filling fluid (Archie 1942). Especially at low salinities or significant clay content, a second conduction mechanism has been observed, which is not covered by Archie's law. This conduction is associated with the grain-water interface and additionally contributes to rock conductivity (Waxman & Smits 1968; Rink & Schopper 1974). For example, Vinegar & Waxman (1984) showed that conductivity in shaly sands has a polarization component as well and the electrical rock conductivity is, therefore, a complex and frequency-dependent quantity. Numerous studies have investigated the impact of different parameters on electrolytic and surface conductivity. Theoretical and experimental contributions are concerned with the origins and dependencies of polarization in porous media (e.g. Dissado & Hill 1984; Olhoeft 1985; Börner & Schön 1991; Revil & Glover 1997; Weller et al.2010). The factor which most strongly influences rock conductivity is water saturation (Archie 1942; Vinegar & Waxman 1984). Salinity is an important factor as well since the ion content of naturally occurring pore fluids varies strongly (Revil & Skold 2011; Weller & Slater 2012; Weller et al.2015). Surface conductivity is also known to be influenced by the pH of the pore water since the proton supply changes the properties of the interface layer (Skold et al.2011). Other studies extend the focus to the full frequency dependence of complex conductivity. Vinegar & Waxman (1984) show data for the frequency range between 1 Hz and 1 kHz. Olhoeft (1985) presents spectra for different rock types and frequencies ranging from 1e-3 Hz to 1 MHz. More recent studies show data for wide frequency ranges and address the derivation of pore space characteristics from spectral induced polarization (SIP; e.g. Leroy et al.2008; Nordsiek & Weller 2008). Measurements at different pressure levels were presented by Zisser et al. (2010) for frequencies above 1 kHz. The presence of a reactive gas such as CO2 in pore space gives rise to several processes, which change multiple system parameters at the same time. Basically, CO2 acts on the electrical rock conductivity in the following three ways (Börner et al.2013): CO2 partially replaces the pore water and consequently reduces the water saturation. CO2 chemically interacts with the pore water by dissolution and the dissociation of carbonic acid. CO2 interacts with the grain surfaces by mineral dissolution and/or precipitation. Although the partial saturation dominates the overall conductivity, the chemical interactions are important to understand and to quantify when electrical measurements are used as monitoring methods. The impact of CO2 dissolution and the dissociation of carbonic acid on pore water conductivity has been shown to be a superposition of an increased ion supply on the one hand and a decreased ion mobility on the other hand (Börner et al.2015a). This effect has to be expected to act on both electrolytic and surface conductivity. Since CO2 is an acidifying agent an effect caused by the low-pH environment is also probable. Very few studies considering the effect of CO2 on the complex electrical properties exist (for a time-domain IP field study see e.g. Doetsch et al.2015). Most existing investigations are restricted to the real part of conductivity so far (e.g. Kim et al.2011). Limited information is available about the effect of chemical interaction on surface conductivity. Measurements on sandstone cores before and after treatment with supercritical CO2 for the frequency range between 400 Hz and 1 MHz by Nover et al. (2013) indicate that complex conductivity is sensitive to changes caused by CO2. To our knowledge no measurements exist for the complex conductivity under in-situ conditions and the influence of CO2. However, such data are important to evaluate the possibilities for electromagnetic monitoring techniques. It remains to motivate why we investigate the sand-brine-CO2 system with measurements of the spectral complex conductivity. Measuring complex conductivity provides information on both electrolytic and surface conductivity and therefore holds the potential to reveal more details about the CO2 - rock interaction then the basic direct-current measurement. For example, besides the water saturation, which should be recovered by an electrical monitoring, we can additionally get access to the chemical interaction between CO2 and the mineral matrix due to the sensitivity of surface conduction to the properties of the grain-water-interface. We consequently present a laboratory study of the spectral complex electrical conductivity of sand samples containing pore waters of varying salinity representing reservoir conditions down to 3000 m depth. Using a clean quartz sand with its high permeability and porosity allows for a better separation of the chemical interaction between the mobile phases on the one hand and the interaction between the rock matrix and the pore fillings on the other hand. Due to the chemical stability of quartz, the latter may be neglected for clean sands. Static conditions and drainage by an active CO2 flow are investigated at different pressures, temperatures and pore water salinities. Thereby, we are concerned to address how stable complex conductivity may be measured under the challenging conditions, how the chemical interaction manifests itself in changes of the complex conductivity, and how measurements of complex conductivity under reservoir conditions may be utilized to monitor both fluid saturation and alterations of the grain-water interface. 2 THEORY 2.1 Frequency-dependent complex conductivity Measurements of the frequency-dependent complex conductivity, also known as SIP, use a harmonic alternating current of distinct frequencies. The formulations for SIP are derived from Maxwell's equations in the frequency domain, where the relation between total current density Jtot and electric field E is given by (Maxwell 1881; Nabighian 1989): \begin{equation} \mathbf {J}_{{\rm tot}}(\omega ) =\hat{\sigma }^{\ast }\mathbf {E}+i\omega \hat{\epsilon }^{\ast }\mathbf {E}, \end{equation} (1) where |$\hat{\epsilon }^{\ast }$| and |$\hat{\sigma }^{\ast }$| denote dielectric permittivity and electrical conductivity, respectively. Both are complex functions (denoted with asterisk), which depend on the angular frequency ω (e.g. Olhoeft 1979). Generally, the real (i.e. in-phase) part of eq. (1) is the conduction component of the total current density, whereas the imaginary (i.e. quadrature) part is associated with its displacement component. If we consider the complex nature of permittivity and conductivity (primed parameters refer to the real part and double primed parameters refer to the imaginary part of the complex quantity), eq. (1) extends to \begin{equation} \mathbf {J}_{\rm tot}(\omega ) =((\hat{\sigma }^{\prime }+\omega \hat{\epsilon }^{\prime \prime })+i(\hat{\sigma }^{\prime \prime }+\omega \hat{\epsilon }^{\prime }))\mathbf {E}. \end{equation} (2) When general earth media are considered, dielectric permittivity can be included in the definition of effective conductivity σ*, therefore (Nabighian 1989) \begin{equation} \mathbf {J}_{{\rm tot}}(\omega ) =\sigma ^{\ast }\mathbf {E} = (\sigma ^{\prime } + i\sigma ^{\prime \prime })\mathbf {E}. \end{equation} (3) In the exact same manner conductivity could be absorbed into permittivity. Both formulations are equivalent to one another. Effective conductivity and effective permittivity are related via \begin{equation} \sigma ^{\ast } =i\omega \epsilon ^{\ast }. \end{equation} (4) Calculations and measurements presented in this work are based on the formulations for effective conductivity σ*, which may be considered as the transfer function between the applied current density and the measured electric field. σ* contains information about both conduction and polarization processes. Due to the low conductivity and multiple conduction mechanisms within real rocks it is not possible to separate the contributions although the contribution from |$\hat{\sigma }^{\ast }$| will dominate at low frequencies (Olhoeft 1979). σ* relates to the magnitude |σ*| and the phase angle φ, thus \begin{equation} \sigma ^{\ast }(\omega ) = \left| \sigma ^{\ast } \right|e^{i\varphi } \end{equation} (5) with |$\left| \sigma ^{\ast } \right|=\sqrt{(\sigma ^{\prime })^2+(\sigma ^{\prime \prime })^2}$| and tan(φ) = σ''/σ'. Complex conductivity is related to complex resistivity by |$\rho ^{\ast }= \frac{1}{\sigma^{\ast }}$|⁠. 2.2 Complex conductivity of water-bearing clean sands 2.2.1 Full saturation Numerous mechanisms contribute to σ* in a clean, water-saturated siliceous porous medium. In the frequency range from 1 mHz to 1 kHz, which is the focus of this study, the effective conductivity may be described by (modified from Vinegar & Waxman 1984): \begin{equation} \sigma ^{\ast }(\omega ) = \sigma _{\rm el} + \sigma ^{\ast }_{\rm {surf}}(\omega ) + \sigma ^{\ast }_{\rm hf}(\omega ), \end{equation} (6) where σel denotes electrolytic conduction, |$\sigma ^{\ast }_{\rm {surf}}$| is surface conduction, and |$\sigma ^{\ast }_{\rm hf}$| includes high frequency effects. Electrolytic conduction σel takes place in the free and interconnected pore water phase. This is a purely ohmic and therefore real and frequency-independent conduction caused by the migration of ions in an electric field. It is directly proportional to the electric conductivity of the pore water σw and typically described by Archie's law (Archie 1942): \begin{equation} \sigma _{\rm el}(T,c) = \frac{1}{F}\sigma _{\rm w}(T,c) = \Phi ^m\sigma _{\rm w}(T,c) \end{equation} (7) where F = 1/Φm is the formation factor, Φ is porosity and m is the cementation exponent, respectively. Since σw depends on temperature T and the concentration c of dissociated salts, σel depends on T and c as well. Surface conduction |$\sigma ^{\ast }_{\rm {surf}}$| occurs at the grain–water boundary. Caused by the specific structure of the silica surface, an interface layer [described by electrical double (EDL) or triple layer models] is formed (e.g. Rink & Schopper 1974; Sharma & Yen 1984; Revil & Glover 1997; Duval et al.2002; Leroy et al.2008). The silica surface develops a negative surface charge when in contact with water. To balance this surface charge, cations from the electrolyte accumulate at the interface. When an electric field is applied the layer of reduced cation mobility causes a cation barrier at the pore throats. This diffusion-controlled polarization adds a frequency-dependent real and imaginary part to conductivity: \begin{equation} \sigma ^{\ast }_{\rm {surf}}(\omega ,T,c) = \sigma ^{\prime }_{\rm {surf}}(\omega ,T,c) + i\sigma ^{\prime \prime }_{\rm {surf}}(\omega ,T,c) \end{equation} (8) Just like σel, |$\sigma ^{\ast }_{\rm {surf}}$| shows dependencies on temperature and salinity although in a much weaker form. |$\sigma ^{\prime }_{\rm {surf}}$| and |$\sigma ^{\prime \prime }_{\rm {surf}}$| are widely considered to be linearly related, since they originate from the same mechanism (Börner 1992; Revil & Skold 2011; Weller & Slater 2012): \begin{equation} \sigma ^{\prime \prime }_{\rm {surf}} = l\sigma ^{\prime }_{\rm {surf}} \end{equation} (9) where l is considered salinity-independent and varies between approximately 0.01 and 0.15 depending on the rock type. In addition to the increasing relevance of the contribution from |$\hat{\epsilon }^{\ast }$| (cf. eqs 1 and 2) strong conductivity contributions may appear at high frequencies, which originate from different polarization and electronic coupling mechanisms (e.g. Wagner 1914; Olhoeft 1985; Leroy et al.2008). These effects are summarized in |$\sigma ^{\ast }_{\rm hf}$|⁠. Since we investigate the very low frequency range we do not consider |$\sigma ^{\ast }_{\rm hf}$| in the following but concentrate on σel and σsurf. Consequently, the imaginary part of any measured effective conductivity solely contains contributions originating from |$\sigma ^{\ast }_{\rm {surf}}$|⁠: \begin{equation} \sigma ^{\prime \prime }(\omega ,T,c) = \sigma ^{\prime \prime }_{\rm {surf}}(\omega ,T,c). \end{equation} (10) As long as the pore space geometry remains the same, σel and σsurf do not depend on pressure. A slight pressure dependence is included in c due to the compressibility of the fluid. 2.2.2 Partial saturation Both real and imaginary part of conductivity depend on saturation (Vinegar & Waxman 1984). Since electrolytic conduction is a volume effect it is affected the strongest by a reduction of water saturation. The dependence is given by Archie (1942) and is introduced in eq. (7) \begin{equation} \sigma _{\rm el}(T,c,S_{\rm w}) = \frac{1}{F}S_{\rm w}^{n}\sigma _{\rm w}, \end{equation} (11) where Sw refers to water saturation and n is the saturation exponent. The surface conductivity |$\sigma ^{\ast }_{\rm {surf}}$| also depends on saturation although less dominantly than the electrolytic contribution (e.g. Waxman & Smits 1968; Vinegar & Waxman 1984; Ulrich & Slater 2004; Breede et al.2012): \begin{equation} \sigma ^{\ast }_{\rm {surf}}(T,c,S_{\rm w}) = S_{\rm w}^{k}\sigma ^{\ast }_{\rm {surf},S_{w=1}}. \end{equation} (12) Here, |$\sigma ^{\ast }_{\rm {surf},S_{w=1}}$| is the surface conductivity at full saturation and k denotes the true saturation exponent of the imaginary part of conductivity. Note that k < n. Usually, k is determined by evaluating σ″ (see e.g. Vinegar & Waxman 1984; Ulrich & Slater 2004). Applying the same saturation exponent to both |$\sigma ^{\prime }_{\rm {surf}}$| and |$\sigma ^{\prime \prime }_{\rm {surf}}$| is not necessarily correct but proved to be a working assumption (Vinegar & Waxman 1984). So far, no dependence on frequency is included in the formulations. Eqs (11) and (12) are supposed to be valid in an approximate frequency range from 1 Hz to 1 kHz (Vinegar & Waxman 1984). Few full spectra have been published for partially saturated sands. The general decrease of σ″ with decreasing saturation is mostly found for the whole frequency range (Ulrich & Slater 2004). Some exceptions at low frequencies are documented, where σ″ increases at high saturation and then decreases as drainage proceeds (Breede et al.2012). 2.2.3 Salinity and pH For the electrolytic conductivity, changes in salinity and pH are reflected in pore water conductivity σw and are therefore already included in eqs (7) and (11). The salinity dependence of |$\sigma ^{\prime }_{\rm {surf}}$| is very difficult to determine experimentally due to the dominating σel at high salinities. Usually it is assumed to be directly proportional to |$\sigma ^{\prime \prime }_{\rm {surf}}$| (Revil & Skold 2011). Numerous publications deal with the behaviour of |$\sigma ^{\prime \prime }_{\rm {surf}}$| at varying salinity so far. Generally, an increase in salinity (and therefore in σw) results in a change in surface speciation and a reduced thickness of the electrical double layer (Debye & Hückel 1923; Schön 1996; Revil & Glover 1997; Schön 2011). Due to the capacitive nature of the EDL this effect increases the capacitance of the interface layer, which causes an overall increase of |$\sigma ^{\prime \prime }_{\rm {surf}}$|⁠. Since this increase is much weaker than the strong increase of σel with increasing salinity, the phase shift decreases with increasing salinity (cf. eq. 5 and definitions there). The increase of |$\sigma ^{\prime \prime }_{\rm {surf}}$| with salinity has widely been observed experimentally and described by both empirical and mechanistic models (Waxman & Smits 1968; Revil & Glover 1997; Revil & Skold 2011; Weller & Slater 2012). For NaCl solutions Revil & Skold (2011) derived the following equation based on considerations regarding the occurrence of surface species on the quartz–electrolyte interface, which has been extended by Skold et al. (2011): \begin{equation} \sigma ^{\prime \prime }_{\rm {surf}}=\left(\frac{F-1}{2d_oF}\right)\left(\frac{e\beta \Gamma Kc_{\rm NaCl}}{10^{\rm -pH}+Kc_{\rm NaCl}} + \Sigma _{\rm H^{+}}\right) \end{equation} (13) where F is the formation factor, e is the elementary charge, β is the cation mobility in the pore water, Γ is the total density of exchange sites at the grain surface and do is the grain diameter. Furthermore, K denotes the equilibrium constant for the sorption of the cations to the quartz surface, cNaCl is the concentration of the electrolyte and pH denotes the pH-value of the pore water. The extension |$\Sigma _{\rm H^{+}}$| describes a proton conduction mechanism, which seems to become important at low salinities and low pH (Skold et al.2011). Following Weller & Slater (2012), eq. (13) is generally equivalent to the simplified formulation \begin{equation} \sigma ^{\prime \prime }_{\rm {surf}}=a_s\frac{\sigma _{\rm w}}{10^{\rm -pH}b_s+\sigma _{\rm w}} + c_s, \end{equation} (14) where as, bs and cs are either fitting parameters or directly relate to the geometry, mobility and sorption properties from eq. (13). For numerous samples a decrease in |$\sigma ^{\prime \prime }_{\rm {surf}}$| with increasing salinity has been observed at very high salinities (σel >1 S m−1). This decrease is associated with a decrease in ion mobility at high salinities. An extension of the model in eq. (13) has recently been suggested by Weller et al. (2015). Very few data and models are available for the impact of pH on |$\sigma ^{\prime \prime }_{\rm {surf}}$|⁠. It is known from investigations of the surface speciation that the negative net surface charge, which is present at neutral pH due to an excess of >SiO− groups at the silica surface, is diminished with decreasing pH and is replaced by a positive net surface charge originating from the then dominating >SiOH2+ groups (Duval et al.2002). The isoelectric point, where the net charge of the silica surface equals to zero, is approximately located at a pH of 2.5 for a quartz–NaCl system (Skold et al.2011). The decreased surface charge at low pH-values is supposed to result in a decreased |$\sigma ^{\prime \prime }_{\rm {surf}}$| as the EDL is weakened. First data presented by Skold et al. (2011) confirm this model conception. The model given in eqs (13) and (14) includes a pH-dependence and is valid in this form for pH-values ranging from 5 to 8 (Revil & Skold 2011). The contribution from |$\Sigma _{\rm H^{+}}$| and an additional contribution from the anions (⁠|$\Sigma _{\rm Cl^{-}}$|⁠) are required to explain data at very low pH and below the isoelectric point. As already mentioned for the impact of partial saturation, information on the frequency dependence of salinity and pH effects is rare or non-existent. According to Weller & Slater (2012), the available relationships should hold for a moderate frequency range around 1 Hz due to the generally weak frequency dependence of sands and sandstones. 2.2.4 Impact of CO2 A free CO2 phase is non-conducting as well as both colourless and odourless at normal conditions (25 °C, 0.1 MPa). It enters the supercritical state when the critical temperature of 30.98 °C and the critical pressure of 7.377 MPa are exceeded (cf. Fig. 8; Span & Wagner 1996). These conditions correspond to approximately 800 m depth, where carbon dioxide storage consequently becomes feasible (IPCC 2005; Marini 2007). In supercritical state, CO2 bears a high density (731.7 kg m−3 at 30 MPa and 80 °C) whereas its viscosity is still gas-like (4.55e-5 Pa s at 30 MPa and 80 °C). Besides the specific density-viscosity combination, supercritical CO2 shows solvent properties and sorption affinity (e.g. Raveendran et al.2005). The strongest impact on σ* is caused by the free and non-conducting CO2 phase in pore space (shown by Börner et al.2013, for σ΄, cf. process 1 in Fig. 1). In the system of quartz, water and CO2, water is the wetting, CO2 the non-wetting phase. However, with increasing pressure and temperature the strength of the water-wettability decreases leaving the system only weakly water-wet at supercritical conditions, which is supposed to have great impact on storage efficiency (Sarmadivaleh et al.2015). At the same time, the interfacial tension between water and CO2 strongly decreases (Sarmadivaleh et al.2015). Consequently, the maximum CO2 saturation in pore space is expected to decrease with increasing pressure and temperature. Figure 1. Open in new tabDownload slide Schematic representation of the processes acting on the electrical properties of a water-bearing porous medium during exposure to carbon dioxide. EDL is the electrical double layer. In contrast to systems with air, oil or natural gas as non-conducting phase, CO2 massively dissolves in the pore water and forms electrically neutral aqueous complexes (Li & Duan 2007; cf. process 2 in Fig. 1). How much CO2 dissolves in a specific pore water at thermodynamic equilibrium depends on pressure, temperature and salinity (e.g. Duan et al.2006). A small fraction of the dissolved CO2 forms carbonic acid (H2CO3), which dissociates due to its instability in the following two-stage dissociation: \begin{eqnarray} {\rm CO_{2(aq)} + H_{2}O} &\iff& {\rm H^+ + HCO_3^-} \end{eqnarray} (15) \begin{eqnarray} {\rm HCO_3^-} &\iff& {\rm H^+ + CO_3^{2-}}. \end{eqnarray} (16) This process adds charge carriers to the pore water (cf. process 3 in Fig. 1). The formation and dissociation of carbonic acid (eqs 15 and 16) cause a decrease in pH of the pore water (cf. process 4 in Fig. 1). For the pressure and temperature range investigated experimentally in the following (2–30 MPa, 15–80 °C) the software PHREEQC (Parkhurst & Appelo 1999), used with the approach described by Börner et al. (2013), predicts a pH ranging from 2.8 to 3.3, which is close to the isoelectric point (cf. Section 2.2.3). Experimental data measured with Raman spectroscopy indicate that pH-values predicted by geochemical modelling software are too low, especially at high salinities (Schaef et al.2003). We consequently assume that the experimental data presented in the following is fully in the range of negative net surface charge. The impact of dissolution and dissociation on pore water conductivity in the absence of a silica matrix has been studied by Börner et al. (2015a) and may be described by an extension of eq. (11): \begin{equation} \sigma _{\rm el}(p,T,c_{\rm NaCl},c_{\rm CO2},S_{\rm w}) = \frac{1}{F}S_{\rm w}^{n}\sigma ^{\rm norm}_{\rm w}\sigma _{\rm w}. \end{equation} (17) The dimensionless |$\sigma ^{\rm norm}_{\rm w}$| characterizes the change in pore water conductivity due to CO2 dissolution and dissociation and depends on pressure, temperature and salinity cNaCl (cf. red curve in Fig. 2). Note that the amount of dissolved and dissociated CO2 (cCO2) is determined by the pressure-temperature-salinity combination at the thermodynamic equilibrium. Depending mainly on salinity, the dissolution and dissociation of CO2 either results in an increase in σel due to the additional charge carriers provided to the solution (‘low-salinity regime’, |$\sigma ^{\rm norm}_{\rm w}>1$|⁠) or in a decrease of σel due to a decreased mobility of the soluted species (‘high-salinity regime’, |$\sigma ^{\rm norm}_{\rm w}<1$|⁠). The contrast in σel caused by CO2 at thermodynamic equilibrium is given by: \begin{eqnarray} &&{\sigma _{\rm w}^{{\rm norm}}(p,T,c_{\rm NaCl}) = \frac{\sigma _{\rm w,CO2}}{\sigma _{\rm w}}}\nonumber\\ &&{\quad\quad= \frac{\left({\color {black}{\Lambda _{\rm NaCl}\gamma _{\rm NaCl}c_{\rm NaCl}}} + {\color {black}{\Lambda _{\rm CO2,dis}\gamma _{\rm CO2,dis}c_{\rm CO2,dis}}}\right)}{\Lambda _{\rm NaCl}\gamma _{\rm NaCl}^{\rm o}c_{\rm NaCl}}} \end{eqnarray} (18) where σw, CO2 is the pore water conductivity with CO2, ΛNaCl and ΛCO2, dis denote the molar conductivities at infinite dilution of NaCl and dissociated CO2, respectively. Furthermore, cNaCl and cCO2, dis refer to the concentrations of NaCl and dissociated CO2. γNaCl and γCO2, dis reflect the inter-species interactions (for details, see Börner et al.2015a). Note that |$\sigma _{\rm w}^{{\rm norm}}$| fundamentally characterizes the change of the aqueous phase in the presence of CO2. |$\sigma _{\rm w}^{{\rm norm}}$| has to be expected to be present in surface conductivity as well, due to the dependence of |$\sigma ^{\ast }_{\rm {surf}}$| on σel (cf. eq. 14). Figure 2. Open in new tabDownload slide Bulk fluid CO2 correction term |$\sigma _{\rm w}^{\rm norm}$| and CO2 correction term for surface conductivity |$\sigma _{\rm {surf}}^{\rm norm}$| versus pore water salinity in terms of molality bNaCl at 5 MPa and 25 °C. |$\sigma _{\rm w}^{\rm norm}$| has been computed with eq. (18), |$\sigma _{\rm {surf}}^{\rm norm}$| is predicted from eq. (14) using the characteristics of the sand (cf. Table 1), |$\sigma _{\rm w}^{\rm norm}$|⁠, and assuming an initial pH of 6.5. As shown above, changes in pH and ion content have an impact on σ″. Consequently, changes in the imaginary part of conductivity have to be expected during the interaction with CO2. No data is available for the impact of CO2 on |$\sigma ^{\ast }_{\rm {surf}}$|⁠. By analogy with eq. (17) an extension of eq. (12) may be postulated: \begin{equation} \sigma ^{\ast }_{\rm {surf}}(p,T,c_{\rm NaCl},c_{\rm CO2},S_{\rm w}) = S_{\rm w}^{k}\sigma ^{\rm norm}_{\rm {surf}}\sigma ^{\ast }_{\rm {surf},S_{w=1}} \end{equation} (19) where |$\sigma ^{\rm norm}_{\rm {surf}}$| reflects the impact of CO2 on the surface conductivity. |$\sigma ^{\rm norm}_{\rm {surf}}$| is dimensionless and can generally depend on pressure, temperature, salinity, pH and mineral composition. Fig. 2 exemplarily shows the expected behaviour of |$\sigma ^{\rm norm}_{\rm {surf}}$| at 5 MPa and 25 °C (blue curve). The estimate is based on eq. (14), the characteristics of the sand used later on (cf. Table 1), and computed pH-values. Note that |$\sigma ^{\rm norm}_{\rm w}$| (red curve) is included to adapt σel to the presence of CO2. Fig. 2 clearly demonstrates that the decrease in pH due to CO2 is expected to dominate the behaviour of σ″. The influence of |$\sigma ^{\rm norm}_{\rm w}$| is reduced due to σel appearing in both numerator and denominator. Consequently, although the specific slope of |$\sigma ^{\rm norm}_{\rm {surf}}$| depends on the considered rock, we expect |$\sigma ^{\rm norm}_{\rm {surf}}$| to be generally smaller than one and to show no clear correlation with |$\sigma ^{\rm norm}_{\rm w}$|⁠. Table 1. Characteristics of the sand used for the experiments. Grain and pore space properties Porosity Φ 0.41 ± 0.05 Median grain diameter 185.0 μm Specific surface area Sma 0.61 ± 0.08 m2g−1 Sporb 2.3 μm−1 Median pore-throat diameterc 53.8 μm Electrical properties Formation factor F 3.27 ± 0.28 |$\sigma ^{\prime }_{\rm surf,S_{w=1}}$| at 25 °C d 7 × 10−4 ± 2 × 10−4 S m−1 |$\sigma ^{\prime \prime }_{\rm surf,S_{w=1}}$| at 25 °C, 0.003 M NaCl 3 × 10−5 ± 1 × 10−5 S m−1 n 1.866 ± 0.08 k 0.62 ± 0.07 Grain and pore space properties Porosity Φ 0.41 ± 0.05 Median grain diameter 185.0 μm Specific surface area Sma 0.61 ± 0.08 m2g−1 Sporb 2.3 μm−1 Median pore-throat diameterc 53.8 μm Electrical properties Formation factor F 3.27 ± 0.28 |$\sigma ^{\prime }_{\rm surf,S_{w=1}}$| at 25 °C d 7 × 10−4 ± 2 × 10−4 S m−1 |$\sigma ^{\prime \prime }_{\rm surf,S_{w=1}}$| at 25 °C, 0.003 M NaCl 3 × 10−5 ± 1 × 10−5 S m−1 n 1.866 ± 0.08 k 0.62 ± 0.07 a From repeated BET analysis. b |$S_{\rm por}=\frac{1-\Phi }{\Phi }d_{\rm s}S_{\rm m}$| with matrix density ds = 2.65 g cm−3. c From mercury intrusion porosimetry. d From separate multi-salinity measurements. Open in new tab Table 1. Characteristics of the sand used for the experiments. Grain and pore space properties Porosity Φ 0.41 ± 0.05 Median grain diameter 185.0 μm Specific surface area Sma 0.61 ± 0.08 m2g−1 Sporb 2.3 μm−1 Median pore-throat diameterc 53.8 μm Electrical properties Formation factor F 3.27 ± 0.28 |$\sigma ^{\prime }_{\rm surf,S_{w=1}}$| at 25 °C d 7 × 10−4 ± 2 × 10−4 S m−1 |$\sigma ^{\prime \prime }_{\rm surf,S_{w=1}}$| at 25 °C, 0.003 M NaCl 3 × 10−5 ± 1 × 10−5 S m−1 n 1.866 ± 0.08 k 0.62 ± 0.07 Grain and pore space properties Porosity Φ 0.41 ± 0.05 Median grain diameter 185.0 μm Specific surface area Sma 0.61 ± 0.08 m2g−1 Sporb 2.3 μm−1 Median pore-throat diameterc 53.8 μm Electrical properties Formation factor F 3.27 ± 0.28 |$\sigma ^{\prime }_{\rm surf,S_{w=1}}$| at 25 °C d 7 × 10−4 ± 2 × 10−4 S m−1 |$\sigma ^{\prime \prime }_{\rm surf,S_{w=1}}$| at 25 °C, 0.003 M NaCl 3 × 10−5 ± 1 × 10−5 S m−1 n 1.866 ± 0.08 k 0.62 ± 0.07 a From repeated BET analysis. b |$S_{\rm por}=\frac{1-\Phi }{\Phi }d_{\rm s}S_{\rm m}$| with matrix density ds = 2.65 g cm−3. c From mercury intrusion porosimetry. d From separate multi-salinity measurements. Open in new tab The low pH environment can give rise to dissolution and precipitation reactions at the grain surfaces (Wigand et al.2008, cf. processes 5 and 6 in Fig. 1). Such processes will generally roughen the grain surfaces and thereby act on σ″, which is directly related to the inner surface area Spor (Börner & Schön 1991; Weller & Slater 2012). We can describe this effect by \begin{equation} \sigma ^{\ast }_{\rm {surf}}(p,T,c_{\rm NaCl},c_{\rm CO2},S_{\rm w}) = S_{\rm w}^{k}X\sigma ^{\rm norm}_{\rm {surf}}\sigma ^{\ast }_{\rm {surf},S_{w=1}} \end{equation} (20) where the dimensionless X describes the change in inner surface area due to mineral dissolution and precipitation. X would be expected to be larger than one in case of a reactive rock matrix since Spor increases. In the case of pure quartz sand, we do not expect any changes of the inner surface area. X should therefore remain equal to one. In summary, we expect σ΄ and σ″ to act in accordance with the following relationships: \begin{eqnarray} \sigma ^{\prime } &=& \frac{1}{F}S_{\rm w}^{n}\sigma ^{\rm norm}_{\rm w}\sigma _{\rm w} + S_{\rm w}^{k}X\sigma ^{\rm norm}_{\rm {surf}}\sigma ^{\prime }_{\rm {surf},S_{w=1}} \end{eqnarray} (21) \begin{eqnarray} \sigma ^{\prime \prime } &=& S_{\rm w}^{k}X\sigma ^{\rm norm}_{\rm {surf}}\sigma ^{\prime \prime }_{\rm {surf},S_{w=1}}. \end{eqnarray} (22) The special challenge of the experiments presented later on is that all processes, effects and dependencies described above occur simultaneously. The effects on SIP due to changes in saturation, ion content and pH cannot be separated during the experimental injection. Quantifying Sw and X would be the central aim of a monitoring with complex conductivity measurements. This should be possible if the undisturbed rock properties and the CO2-effect in terms of |$\sigma ^{\rm norm}_{\rm w}$| and |$\sigma ^{\rm norm}_{\rm {surf}}$| are known. 2.3 Debye decomposition One way of accessing the frequency dependent conductivity in terms of pore space characteristics is the Debye decomposition approach (Nordsiek & Weller 2008). The complex resistivity spectrum is represented by a series of Debye relaxations: \begin{equation} \rho ^{\ast }(\omega )=\frac{1}{\sigma ^{\ast }(\omega )}=\frac{1}{\sigma _0}\left(1-\sum _\ell {\nu _\ell \left(1-\frac{1}{1+{i}\omega \tau _\ell }\right)}\right) \end{equation} (23) with a direct current conductivity σ0, relaxation time τ and chargeability ν. Each relaxation time τℓ contributes with a peak whose magnitude is controlled by νℓ at a constant frequency to the conductivity spectrum: \begin{equation} f_\ell ^{\rm peak}=\frac{1}{2\pi \tau _\ell }. \end{equation} (24) Nordsiek & Weller (2008) propose the total normalized chargeability ∑νnorm and the mean relaxation time |$\bar{\tau }$| as indicative parameters to characterize a whole spectrum: \begin{eqnarray} \sum {\nu _{\rm norm}}&=&\sigma _0\sum _\ell {\nu _\ell } \end{eqnarray} (25) \begin{eqnarray} \bar{\tau }&=&{\rm exp}\left(\frac{\sum _\ell {(\nu _\ell {\rm ln}\tau _\ell )}}{\sum _\ell {\nu _\ell }}\right) \end{eqnarray} (26) Basically, the Debye decomposition is a transformation of the spectral data. The transformation helps to identify the dominant relaxation times, which can sometimes be related to pore space characteristics. For example, Leroy et al. (2008) deduce for a theoretical porous medium consisting of spherical grains of constant diameter d0: \begin{equation} \tau _0 = \frac{d_0^2}{2D^{\rm S}}, \end{equation} (27) where DS is the cation mobility in the Stern layer. Empirical relationships between τ and ∑νnorm on the one hand and Spor and dominant pore throat diameters on the other hand are, for example, given by Scott & Barker (2005) for some sandstones. These relationships are usually very individual for specific rocks or formations and difficult to generalize. Additionally, it is known that dominant relaxations may also be caused by chemical reactions. Olhoeft (1985), for example, shows the impact of redox reactions on σ* as phase maxima. Besides pore space characteristics, ∑νnorm also reflects saturation changes and the chemical behaviour of CO2. In order to demonstrate this, Fig. 3 shows the theoretical dependence of ∑νnorm on water saturation when a single relaxation (ℓ = 1) at f = f  peak is considered. It is assumed that σ* acts in accordance with eqs (17) and (19). The blue curve shows the behaviour when no CO2 is present (⁠|$\sigma ^{\rm norm}_{\rm w}=\sigma ^{\rm norm}_{\rm {surf}}=1$|⁠). The red curve shows the behaviour with CO2 assuming |$\sigma ^{\rm norm}_{\rm {surf}}$| is constant and <1. Note that due to the normalization νnorm is independent from σw. From Fig. 3 we expect decreasing normalized chargeabilities due to drainage. An additional offset may be expected when CO2 comes into play. Figure 3. Open in new tabDownload slide Theoretical behaviour of νnorm = σ0ν (eq. 23) for a single τ at fpeak assuming σ* acts in accordance with eqs (17) and (19) without CO2 (blue) and with CO2 presence (red). The properties from Table 1 were used. 3 MATERIALS AND METHODS 3.1 Materials A clean quartz sand served as unconsolidated rock matrix for all experiments presented in this study because sand is the most controllable natural rock available. It is well determined in terms of grain size, porosity, pore space geometry and mineral content. We therefore are able to use homogeneous samples with repeatable properties. Selected characteristics of the sample material are summarized in Table 1. The sample material is a natural clean quartz sand which has experienced only moderate geological transport and displacement (see also Fig. 4a). It shows a typical narrow and monodisperse grain size distribution with a median grain diameter of 185 μm and a median pore throat diameter of ca. 54 μm. The true saturation exponent, the real part of the surface conductivity and the saturation exponent of the imaginary part of conductivity have been derived independently at normal conditions without CO2 influence. Measurements of the inner surface area of the sand before and after CO2 exposure match within the measurement uncertainties (Sm after exposure to supercritical CO2 is 0.5 ± 0.07 m2 g−1, which corresponds to an Spor of 1.9 μm−1). The visual assessment of secondary electron microscope (SEM) images of the grain surfaces further proves that no alteration of surface roughness takes place during the presented experiments, which would act on σ″ (cf. Figs 1 and 4). It can consequently be concluded that changes of σ″ are due to saturation changes or the reactive nature of CO2. Figure 4. Open in new tabDownload slide Secondary electron microscope (SEM) pictures of two quartz grains before (top) and after (bottom) CO2 exposure. No changes in surface roughness occur. For all experiment dry, 99.5 per cent pure CO2 was used. Sodium chloride solutions were set up from deionized water and cross-checked with an independent conductivity measurement. 3.2 Experimental setup The experimental setup shown in Fig. 5 was used for all experiments presented in this study. The setup is based on the apparatus described by Börner et al. (2013) and Börner et al. (2015a) and now equipped with a self-designed measuring cell for flow-through experiments and simultaneous electrical measurement. A stable temperature is ensured by a hot air cabinet. Pressure is controlled with a two-staged feed pressure regulation. The autoclave (Fig. 6 , right) can operate at maximum pressures of 40 MPa and at a maximum temperature of 80 °C. All pressure values are expressed in terms of overpressure, that is, 0 MPa refer to ambient conditions. Pressure may be measured and controlled at the inlet (top of the autoclave) and the outlet (bottom right of the autoclave), thereby introducing a pressure gradient at arbitrary mean pressure if required. The accuracy of the pressure measurement is ±0.1 MPa in the operating range (equals 0.25 per cent of the measuring span). Figure 5. Open in new tabDownload slide Scheme of the laboratory setup used for the experiments. Figure 6. Open in new tabDownload slide Photograph and schematic drawing of the measuring cell (left and centre) and CAD drawing of the autoclave (right) used for the experiments described in this study. The measuring cell is fixed tightly to the autoclave by means of a thread. The interface between autoclave and measuring cell is sealed by an O-ring. Any CO2 flow is thereby forced to pass through the sample holder. The measuring cell shown in Fig. 6 holds an unconsolidated rock sample of ca. 0.42 L volume and is made of polyvinyl chloride (PVC). The cell consists out of eight segments. The bottom and top covers are permeable filter plates to allow for flow through the sample. The cell bottom is furthermore equipped with a polytetrafluoroethylene membrane, which prevents untimely drainage as it only becomes permeable when a threshold pressure gradient is exceeded. Current for the measurements of the spectral complex conductivity is supplied by platinized platinum wire mesh electrodes at the top and bottom of the measuring cell. The slits at the contact of neighbouring cell segments harbour six platinum wire ring electrodes for voltage recording at different heights of the sample. The default measuring configuration uses the central pair of potential electrodes (P3 and P4 in Fig. 6). A temperature sensor (Pt 100, accuracy: ±0.16 °C at 5 °C and ±0.31 °C at 80 °C, see also Börner et al.2015a) is integrated in the cell wall at a height of 10.5 cm to measure temperature directly at the sample without any disturbance of the sand bulk. The main requirements and challenges for the design of the sample holder are to allow for measurements of the spectral complex conductivity during two-phase flow and partial saturation and at the same time to withstand pressure, pressure gradients, temperature and CO2 exposure during the experiments and to ensure a uniform specific feed distribution. The PVC material has been tested to be geometrically stable despite the occurring CO2 sorption. Still the material requires time to release the sorbed CO2 during depressurization and ages gradually. Consequently, the PVC components were completely replaced by identical replicas two times during the study. To ensure the permanent contact of the potential electrodes to the partially saturated sample, the slits harbouring the electrodes (width = 0.2 mm) were filled with kaolinite saturated with the same solution as used as pore water. The complex conductivity measurement was carried out with a SIP Fuchs III device (Radic Research), which is generally capable to measure 25 frequencies between 1.4 mHz and 20 kHz. The measurement is carried out in terms of resistivity magnitude and phase shift, which are translated into real and imaginary part of conductivity for interpretation (cf. Section 2.1). For a pure water sample, where no phase shift should occur, the amplitude and phase spectrum of the measuring cell equipped with kaolinite for contacting the potential electrodes is shown in Fig. 7. The maximum phase shift is 0.39 mrad, which is sufficiently small to allow for a quantitative interpretation of the measured spectra. Figure 7. Open in new tabDownload slide Impedance amplitude (left) and phase shift (right) for a pure sodium chloride solution (0.02 M NaCl) within the measuring cell used for the high-pressure and high-temperature experiments (cf. Figs 5 and 6 and Section 3.2). The maximum phase shift is 0.39 mrad. 3.3 Experimental procedure The fully water saturated sample is produced by step-wise trickling the sand into the water and compacting it by tapping on the side of the cell. This method has been identified as advantageous for sands by Bairlein et al. (2014). Then the cell is fixed within the autoclave and sealed against evaporation. Autoclave components and cell are adjusted to the experiment temperature during the next 15–20 hr within the hot-air cabinet. Temperature and conductivity are indicated and monitored from this point onwards throughout the whole experiment including the depressurization afterwards. When temperature reaches a constant level and the measured spectral complex conductivity is stable, the evaporation cover lid is removed and the autoclave is closed. Pressure is now monitored as well (see also Fig. 9a, top for an exemplary pressure, temperature and cumulative CO2 mass sequence). Pressure is adjusted to the experiment pressure while the bypass (Fig. 5) is open. By doing so, pressure rises uniformly on both sides of the cell and flow is generally prevented. After pressure adjustment the bypass is closed. Now, a CO2 mass flow is enforced though the sample by applying a pressure gradient and consequently portions of the pore water are driven out through the bottom of the cell. The pressure gradient is increased stepwise. During the first hour of flow a CO2 mass flow of ca. 25 g min−1 is applied, during the second hour of 45 g min−1 and in the final period of 70 g min−1. At the end of the experiment the autoclave is depressurized gradually over a time-span of up to 50 hr. After opening the autoclave the remaining water content of the sample is immediately measured with a moisture analyser at different heights within the sample. Overall, one flow experiment took between 2 and 5 d. Throughout the entire experiment the spectral complex conductivity is measured continuously. Due to the changing sample during flow, only frequencies down to 91 mHz were recorded. During periods without flow, lower frequencies down to 0.0014 Hz were measured as well. 3.4 Experimental agenda An overview of all experiments carried out in this study is given in Table 2. One flow experiment at 5 MPa and 25 °C has been carried out using nitrogen as streaming gas (see experiment 15 in Table 2). This experiment serves as a reference since it provides data for the case of an inert gas, which does not interact chemically with the pore water as carbon dioxide does. We then conducted 14 flow experiments with carbon dioxide. Pressure ranged from 2 to 30 MPa and temperature varied from 15 to 80 °C. At the same time salinities between 0.003 and 1 mol kg−1 sodium chloride were used. The chosen pressure–temperature–salinity combinations mimic the conditions at different depths within the subsurface. Considering hydrostatic pressure and a standard geothermal gradient, our experiments approximately cover the most important depth range between 200 and 3000 m (Fig. 8). Finally, two static experiments were carried out, where the sample was kept under pressure and CO2 for several days in order to access the impact of CO2 on the fully water-saturated sand (experiments 16 and 17 in Table 2). Figure 8. Open in new tabDownload slide Overview of flow experiments and their pressure and temperature conditions (coloured squares) plotted in the phase diagram of carbon dioxide after Span & Wagner (1996). Colour denotes initial brine salinity. Note that most pressure–temperature conditions were covered by repeated experiments (total number of flow experiments is 15, see also Table 2). Table 2. Overview of the flow experiments (1–15) and two static experiments without drainage (16,17) carried out for this study. bNaCl denotes molality of NaCl, mtotal is the total mass of CO2 and N2, respectively, forced through the sample. No. . Gas . Physical state . p [MPa] . T [°C] . bNaCl [mol kg−1] . σw at 25 °C [S m−1] . Sres [] . mtotal [kg] . 1 CO2 gaseous 2 15 3.0e-3 3.59e-2 0.20 6.89 2 CO2 gaseous 2 15 3.0e-3 3.75e-2 0.17 5.36 3 CO2 gaseous 2 15 3.0e-3 3.73e-2 0.17 8.03 4 CO2 gaseous 2 15 1.0e-1 1.06e+0 0.17 9.72 5 CO2 gaseous 4 50 3.0e-3 3.70e-2 0.19 6.96 6 CO2 gaseous 4 50 3.0e-3 3.65e-2 n.m.a n.m.a 7 CO2 gaseous 5 25 3.0e-3 3.67e-2 0.15 9.68 8 CO2 gaseous 5 25 3.0e-3 3.67e-2 0.18 7.57 9 CO2 supercritical 12 40 1.0e-2 1.17e-1 0.44 7.83 10 CO2 supercritical 12 40 1.0e-2 1.17e-1 0.35 8.42 11 CO2 supercritical 18 60 1.0e-1 1.07e+0 0.24 8.49 12 CO2 supercritical 18 60 1.0e-1 1.07e+0 0.39 7.85 13 CO2 supercritical 25 80 1.0e+0 9.23e+0 0.22 8.49 14 CO2 supercritical 30 80 1.0e-1 1.06e+0 0.19 5.89 15 N2 supercritical 5 25 3.0e-3 3.67e-2 0.17 3.99 16 CO2 gaseous 4 15 1.0e-3 1.29e-2 – b 0 17 CO2 gaseous 4 15 1.0e-3 1.36e-2 – b 0 No. . Gas . Physical state . p [MPa] . T [°C] . bNaCl [mol kg−1] . σw at 25 °C [S m−1] . Sres [] . mtotal [kg] . 1 CO2 gaseous 2 15 3.0e-3 3.59e-2 0.20 6.89 2 CO2 gaseous 2 15 3.0e-3 3.75e-2 0.17 5.36 3 CO2 gaseous 2 15 3.0e-3 3.73e-2 0.17 8.03 4 CO2 gaseous 2 15 1.0e-1 1.06e+0 0.17 9.72 5 CO2 gaseous 4 50 3.0e-3 3.70e-2 0.19 6.96 6 CO2 gaseous 4 50 3.0e-3 3.65e-2 n.m.a n.m.a 7 CO2 gaseous 5 25 3.0e-3 3.67e-2 0.15 9.68 8 CO2 gaseous 5 25 3.0e-3 3.67e-2 0.18 7.57 9 CO2 supercritical 12 40 1.0e-2 1.17e-1 0.44 7.83 10 CO2 supercritical 12 40 1.0e-2 1.17e-1 0.35 8.42 11 CO2 supercritical 18 60 1.0e-1 1.07e+0 0.24 8.49 12 CO2 supercritical 18 60 1.0e-1 1.07e+0 0.39 7.85 13 CO2 supercritical 25 80 1.0e+0 9.23e+0 0.22 8.49 14 CO2 supercritical 30 80 1.0e-1 1.06e+0 0.19 5.89 15 N2 supercritical 5 25 3.0e-3 3.67e-2 0.17 3.99 16 CO2 gaseous 4 15 1.0e-3 1.29e-2 – b 0 17 CO2 gaseous 4 15 1.0e-3 1.36e-2 – b 0 a Not measured. b Steady-state experiment, Sw = 1 throughout the whole experiment. Open in new tab Table 2. Overview of the flow experiments (1–15) and two static experiments without drainage (16,17) carried out for this study. bNaCl denotes molality of NaCl, mtotal is the total mass of CO2 and N2, respectively, forced through the sample. No. . Gas . Physical state . p [MPa] . T [°C] . bNaCl [mol kg−1] . σw at 25 °C [S m−1] . Sres [] . mtotal [kg] . 1 CO2 gaseous 2 15 3.0e-3 3.59e-2 0.20 6.89 2 CO2 gaseous 2 15 3.0e-3 3.75e-2 0.17 5.36 3 CO2 gaseous 2 15 3.0e-3 3.73e-2 0.17 8.03 4 CO2 gaseous 2 15 1.0e-1 1.06e+0 0.17 9.72 5 CO2 gaseous 4 50 3.0e-3 3.70e-2 0.19 6.96 6 CO2 gaseous 4 50 3.0e-3 3.65e-2 n.m.a n.m.a 7 CO2 gaseous 5 25 3.0e-3 3.67e-2 0.15 9.68 8 CO2 gaseous 5 25 3.0e-3 3.67e-2 0.18 7.57 9 CO2 supercritical 12 40 1.0e-2 1.17e-1 0.44 7.83 10 CO2 supercritical 12 40 1.0e-2 1.17e-1 0.35 8.42 11 CO2 supercritical 18 60 1.0e-1 1.07e+0 0.24 8.49 12 CO2 supercritical 18 60 1.0e-1 1.07e+0 0.39 7.85 13 CO2 supercritical 25 80 1.0e+0 9.23e+0 0.22 8.49 14 CO2 supercritical 30 80 1.0e-1 1.06e+0 0.19 5.89 15 N2 supercritical 5 25 3.0e-3 3.67e-2 0.17 3.99 16 CO2 gaseous 4 15 1.0e-3 1.29e-2 – b 0 17 CO2 gaseous 4 15 1.0e-3 1.36e-2 – b 0 No. . Gas . Physical state . p [MPa] . T [°C] . bNaCl [mol kg−1] . σw at 25 °C [S m−1] . Sres [] . mtotal [kg] . 1 CO2 gaseous 2 15 3.0e-3 3.59e-2 0.20 6.89 2 CO2 gaseous 2 15 3.0e-3 3.75e-2 0.17 5.36 3 CO2 gaseous 2 15 3.0e-3 3.73e-2 0.17 8.03 4 CO2 gaseous 2 15 1.0e-1 1.06e+0 0.17 9.72 5 CO2 gaseous 4 50 3.0e-3 3.70e-2 0.19 6.96 6 CO2 gaseous 4 50 3.0e-3 3.65e-2 n.m.a n.m.a 7 CO2 gaseous 5 25 3.0e-3 3.67e-2 0.15 9.68 8 CO2 gaseous 5 25 3.0e-3 3.67e-2 0.18 7.57 9 CO2 supercritical 12 40 1.0e-2 1.17e-1 0.44 7.83 10 CO2 supercritical 12 40 1.0e-2 1.17e-1 0.35 8.42 11 CO2 supercritical 18 60 1.0e-1 1.07e+0 0.24 8.49 12 CO2 supercritical 18 60 1.0e-1 1.07e+0 0.39 7.85 13 CO2 supercritical 25 80 1.0e+0 9.23e+0 0.22 8.49 14 CO2 supercritical 30 80 1.0e-1 1.06e+0 0.19 5.89 15 N2 supercritical 5 25 3.0e-3 3.67e-2 0.17 3.99 16 CO2 gaseous 4 15 1.0e-3 1.29e-2 – b 0 17 CO2 gaseous 4 15 1.0e-3 1.36e-2 – b 0 a Not measured. b Steady-state experiment, Sw = 1 throughout the whole experiment. Open in new tab The experiments were designed and conducted with respect to several aspects. Understanding the behaviour of SIP at elevated pressure/temperature and during CO2 exposure allows to evaluate the potential of complex conductivity measurements for geoelectrical monitoring applications. We consequently address the following questions: What is the impact of the chemical interaction on surface conductivity and is it possible to quantify the effect? How accurate can saturation be derived from electrical measurements when chemical interaction interferes? Do real and imaginary part of conductivity reflect different processes, do they provide more detailed information about these processes and does this justify measuring them during monitoring? How does the conductivity spectrum of a sand sample appear at pressures up to 30 MPa, at temperatures up to 80 °C and the presence of a reactive gas? Which is the best frequency range for in-situ SIP in reactive environments regarding stability and interpretability? Following the questions formulated in this section, we present and interpret our data in the next sections. 4 RESULTS AND DISCUSSION 4.1 Laboratory data set Measurements of σ* for frequencies down to 0.0014 Hz under pressure and exposure to CO2 were successfully carried out at multiple temperatures and salinities. Due to the rapidly changing sample properties during flow, measurements were restricted to the frequencies above 0.091 Hz in the periods of active flow. The frequency range was extended during the flow breaks. Fig. 9 shows data of two flow-through experiments in terms of experimental conditions (Fig. 9a top, for experiment 8), conductivity at 11.7 Hz (Fig. 9a bottom) and spectra for experiments 15 (with N2, Fig. 9b) and 8 (with CO2, Fig. 9c). Note that the data for experiment 15 has been shifted slightly to achieve coinciding starting points for the flow periods. Five characteristic stages of the experiments are marked by vertical coloured lines. At the initial state of the sample at the set temperature (dark blue), no pressure is applied to the system and the sample is fully water saturated. The following three stages (light blue, green and yellow) mark the flow interruptions after the periods of active flow (cf. Section 3.3). Pressure is present at these stages, the sample is partially water saturated. The final stage (red) represents the partially water saturated sample after pressure is released. In Figs 9(b) and (c) (bottom), the full spectra for the five stages are plotted using the same colours. Figure 9. Open in new tabDownload slide Time-series and full spectra for experiments 8 and 15 (cf. Table 2) at 5 MPa, 25 °C and with 0.003 mol kg−1 NaCl pore water. The spectra in panels (b) and (c) and their colours correspond to the five experimental stages marked by vertical coloured lines in panel (a). Experiment 15 with N2 at 5 MPa, 25 °C and 0.003 M NaCl serves as reference in order to distinguish between changes of σ* due to drainage on the one hand and chemical interaction on the other hand. σ΄ (grey circles in Fig. 9a) shows very stable conditions at full saturation. The flow has the expected strong impact on σ΄. The strongest drop in σ΄ happens during the first flow period (17–18.25 hr), which is also expected due to the easily drainable sand. σ΄ decreases further during the second and third flow period, though less prominently than during the first period. This is also expected since drainage becomes less effective when saturation is closer to the irreducible phase content (e.g. van Genuchten 1980; Busch et al.1993). During depressurization (25–44 hr) σ΄ remains stable. For the flow experiment with CO2 at identical p, T and cNaCl (experiment 8, black squares) the initial and final conditions are very well comparable to the N2 case. The periods of flow are equally long for both experiments, the residual water saturations are very similar and because of that the initial and final conductivity coincide. During the period of elevated pressure (17–43 hr) σ΄ for CO2 follows the same general characteristics as for N2. However, both curves deviate significantly. In the case of CO2 the σ΄ reduction during elevated pressure is less pronounced. During pressure release (24–44 hr) σ΄ experiences a further significant drop, which cannot be associated with a change in saturation due to the technical implementation of the pressure release. The conductivity after the experiment (stage 5, red) is the same for both N2 and CO2. In Fig. 9(a) the imaginary part for experiments 15 and 8 is given as well (grey diamonds and black triangles, respectively). Generally, σ″ is small compared to σ΄, which is a typical observation for a clean, unconsolidated sediment. At the beginning of both experiments σ″ is stable. As expected from eq. (12), σ″ reacts to the drainage and the resulting partial saturation with a decrease. Again, the largest decrease occurs during the first flow period. This decrease is smaller than in the case of σ΄, which generally is in accordance with the saturation exponent for the imaginary part k being smaller than n (cf. Section 2.2.2). Compared to the N2 case, σ″ shows a much more significant drop during drainage. When the system is depressurized (between stages 4 and 5) the imaginary part of conductivity is stable for N2 but reacts with an increase in the case of CO2, which is the reverse behaviour compared to σ΄ for CO2 (cf. black squares in Fig. 9a). At stage 5 of both experiments the measured σ″ at 11 Hz coincide very well again. The full frequency characteristics of σ″ is shown in Figs 9(b) and (c). The complete time-lapse data for the frequency range between 0.1 Hz and 100 Hz, which has been measured continuously, is plotted for experiments 8 and 15 in Fig. 10. Additional cases are shown in Fig. 11. The black curve in Fig. 11 shows σ″ for the initial state of experiment 16 at normal conditions and full saturation. The green curve refers to experiment 16 in the equilibrium state after 5.5 d under pressure. The light grey curve is for a low saturation of 0.11 at normal conditions (20 °C, no pressure). Figure 10. Open in new tabDownload slide 3-D time-series of the spectra between 0.1 and 100 Hz for experiments 8 and 15 (cf. Table 2) at 5 MPa, 25 °C and with 0.003 mol kg−1 NaCl pore water. Bullets represent measured data points. The complex conductivity spectrum at a fixed time during flow may be derived by interpolation of the time-series of each frequency. By doing this, effects of the rapidly changing sample are excluded from the spectrum. At static conditions, interpolation is not required (cf. Fig. 9). Figure 11. Open in new tabDownload slide Full spectra for experiment 16 (black: starting conditions of experiment 16 with Sw = 1; green: experiment 16 after 5.5 d under pressure) and a separate measurement at normal conditions and Sw = 0.11 (light grey) for comparison. Comparing the spectra under pressure and CO2 and/or partial saturation with those at normal conditions reveals an unusual behaviour. While the fully saturated cases (black and dark blue, respectively) rapidly decrease towards approximately 1e-6 S m−1 at 1 mHz, the data measured under pressure show significantly larger values (up to 1e-5 S m−1) for a frequency 0.02 Hz and lower. The effect is particularly pronounced in the case of the full saturation with CO2 (Fig. 11, green), where another maximum appears at low frequencies. It seems to be more dominant for CO2 than for N2. Usually, partial saturation mainly causes a shift in magnitude of σ″, which is similar for all frequencies (e.g. Ulrich & Slater 2004), although one sample from Breede et al. (2012) at normal conditions shows a remotely similar effect at low frequencies. The following considerations concentrate on the frequency range between 0.1 and 100 Hz due to the consistent behaviour of the spectra and the possibility to robustly measure in said frequency range during flow (cf. Fig. 10). Since the frequency dependence in this range is small we choose 11.7 Hz as a reliable and representative frequency for the evaluation of chemical interaction and saturations. In the following sections the impact of CO2 on both electrolytic and surface conductivity is investigated. Based on that, the suitability of σ΄ and σ″ for the estimation of water saturation from SIP measurements is evaluated. Finally, the behaviour of σ″ below 0.1 Hz is discussed. 4.2 Quantification of chemical interaction 4.2.1 Electrolytic conductivity The electrolytic conductivity σel dominates the real part of σ* and may consequently be addressed by evaluating σ΄. The difference between σ΄ for the CO2 and σ΄ for the N2 experiment in Fig. 9 may be explained with the impact of CO2 on pore water conductivity. The 0.003 M NaCl solution, which served as pore water for both experiments reacts in accordance with the low-salinity regime as expected from eq. (18). This means the dissolution of CO2 results in an increase of σel. This conductivity increasing effect dampens the σ΄ reduction due to the partial saturation. To check whether all flow experiments act in accordance with the predicted |$\sigma _{\rm w}^{\rm norm}$|⁠, we evaluate σ΄ during depressurization at the end of the experiments. During pressure release the CO2 degasses from the remaining pore water and the CO2-affected σel predominantly approaches its initial value again. The pressure release is an indicator for the chemical impact of CO2, therefore. Fig. 12 shows this effect for all experiments. The ratio of σ΄ at stage 4 (⁠|$\sigma ^{\prime }_{\rm end,p}$|⁠, yellow in Fig. 9) and σ΄ at stage 5 (⁠|$\sigma ^{\prime }_{\rm end}$|⁠, red in Fig. 9) is plotted versus the |$\sigma _{\rm w}^{\rm norm}$| predicted for each of the p/T/salinity conditions by eq. (18). Colours denote salinity and the grey areas show the two σel regimes. If the data points lie inside the grey areas the pressure release recovers the predicted salinity regime. Figure 12. Open in new tabDownload slide Impact of pressure release on σ΄. The impact of chemical interaction on σ΄ is dominated by the two-regime behaviour of σel (cf. Section 2.2.4). The deviation from perfect correlation (cf. dashed black line) is due to the interplay of electrolytic and surface conductivity (cf. eqs 21 and 22). The data coincides very well with the predicted behaviour. Experiments with low saline waters (blue and cyan dots, |$\sigma _{\rm w}^{\rm norm}$| >1) react to the pressure release with a conductivity reduction (⁠|$\sigma ^{\prime }_{\rm end,p}$|/|$\sigma ^{\prime }_{\rm end}$| > 1). Experiments with highly saline water (yellow and red dots, |$\sigma _{\rm w}^{\rm norm}$| < 1) react with a conductivity increase (⁠|$\sigma ^{\prime }_{\rm end,p}$|/|$\sigma ^{\prime }_{\rm end}$| < 1). For comparison, the N2 experiment does not react at all to the pressure release (open red circle, |$\sigma ^{\prime }_{\rm end,p}$|/|$\sigma ^{\prime }_{\rm end}$| = 1) since N2 is an inert gas. From these observations we conclude that |$\sigma _{\rm w}^{\rm norm}$| behaves as expected in the sand-water-CO2 system and that eq. (17) is valid. 4.2.2 Surface conductivity Due to the advantage that surface conductivity alone contributes to the measured imaginary part of conductivity we can access the impact of CO2 on |$\sigma ^{\ast }_{\rm {surf}}$| directly. In analogy to the approach for σel we evaluate the period of pressure release because according to eq. (22) the ratio |$\sigma ^{\prime \prime }_{\rm end,p}/\sigma ^{\prime \prime }_{\rm end}$| should yield a direct measure for |$\sigma ^{\rm norm}_{\rm {surf}}$|⁠. The ratio is plotted versus |$\sigma ^{\rm norm}_{\rm w}$| in Fig. 13 in order to address a possible correlation with the bulk fluid effect described by |$\sigma ^{\rm norm}_{\rm w}$|⁠. Figure 13. Open in new tabDownload slide Impact of pressure release on σ″. The impact of chemical interaction on surface conductivity is independent on salinity, pressure and temperature for the presented data set. The overall estimate |$\sigma ^{\rm norm}_{\rm {surf}}=0.81$| is indicated as dashed black line. The data basically shows the behaviour expected from the theoretical considerations in Section 2.2.4. |$\sigma ^{\rm norm}_{\rm {surf}}$| is smaller than one for all experiments. The two-regime behaviour, which is present in |$\sigma ^{\rm norm}_{\rm w}$|⁠, cannot be found for |$\sigma ^{\rm norm}_{\rm {surf}}$|⁠. We consequently attribute the drop in σ″ to the change in pH due to the dissociation of carbonic acid. Starting with a neutral pH, the presence of CO2 causes a drop in pH, which is similar for all pressure/temperature/salinity conditions (cf. Section 2.2.3). As described in the theory section, a low pH causes a reduced imaginary conductivity. Data presented by Skold et al. (2011) shows a ratio of approximately 0.55 between low and neutral pH σ″ for fully saturated samples with 0.01 M NaCl, which is in a similar range as our observations. From the data presented in Fig. 13 we deduce the following first quantification of the impact of CO2 on the surface conductivity valid for our data set by averaging: \begin{equation} \sigma ^{\rm norm}_{\rm {surf}}=0.81\pm 0.11. \end{equation} (28) We use the given lump-sum value in the following considerations where it proves to be a good overall estimate. 4.3 Application to reservoir monitoring At this point the chemical behaviour of CO2 and its impact on both σel and |$\sigma ^{\ast }_{\rm {surf}}$| are quantified. Furthermore, the capabilities of monitoring the saturation according to eqs (21) and (22) are evaluated. Measuring both σ΄ and σ″ allows for monitoring the following reservoir characteristics: The robust computation of fluid saturation Sw indicates the fate of the injected CO2 in the target formation and enables a better balancing of the amount of stored CO2. Sw may be computed from eqs (21) and (22) with the help of eq. (9) by \begin{equation} S_{\rm w}=\left(\frac{\sigma ^{\prime }-\frac{1}{l}\sigma ^{\prime \prime }}{\frac{1}{F}\sigma ^{\rm norm}_{\rm w}\sigma _{\rm w}}\right)^{\frac{1}{n}}. \end{equation} (29) Monitoring the change of the inner surface area X provides an independent indicator for mineral dissolution and precipitation processes. Based on the computation of Sw, X may be derived from eq. (22) by \begin{equation} X = \frac{\sigma ^{\prime \prime }}{S_{\rm w}^k\sigma ^{\rm norm}_{\rm {surf}}\sigma ^{\prime \prime }_{\rm {surf},S_{w=1}}}. \end{equation} (30) In eqs (29) and (30), σ΄ and σ″ are the measured data at each point in time of the monitoring and |$\sigma ^{\prime \prime }_{\rm {surf},S_{w=1}}$| is the measured imaginary part of the fully water-saturated, undisturbed formation. All other quantities required may be derived from independent baseline investigations and are given in Table 1 for our sample material. The proportionality factor l of our sand may be estimated from |$\sigma ^{\prime }_{\rm {surf},S_{w=1}}$| and |$\sigma ^{\prime \prime }_{\rm {surf},S_{w=1}}$| in Table 1 to be approximately equal to 0.04 ± 0.02 and cross-checked by applying eq. (21) to the initial data of the experiment. l = 0.045 at 11.7 Hz recovers the full saturation stage of experiment 8 and is used for the whole analysis. Note that the concept of the proportionality factor l is originally associated with the constant-phase-angle model (CPA) from Börner (1992). In our analysis (eq. 29) the parameter l is valid only for the chosen evaluation frequency, which is in accordance with the approach from Revil & Skold (2011) and Weller & Slater (2012). If the spectral characteristics of the formation rock significantly deviates from the CPA behaviour, l is frequency-dependent. The processing of the measured σ* in terms of Sw and X for experiment 8 is plotted in Fig. 14. In order to demonstrate the effect of the chemical interaction both Sw and X are computed without consideration of CO2 (⁠|$\sigma ^{\rm norm}_{\rm w}=\sigma ^{\rm norm}_{\rm {surf}}=1$|⁠, blank symbols) and with CO2 considered (black symbols). The 5 stages of the experiment are again marked by the coloured lines for orientation. Fluid saturation is known at the beginning of the experiment (full saturation) and at the end (measured partial saturation). Both values are plotted in grey for comparison. The analysis shows that saturation at the beginning and the end (dark blue and red stage) are recovered very well by our approach. Including |$\sigma ^{\rm norm}_{\rm w}$| and |$\sigma ^{\rm norm}_{\rm {surf}}$| during the period of elevated pressure yields reasonable and well interpretable results. The final fluid saturation is reached within small deviations after approximately 23 hr, which is in accordance with the experimental procedure, whereas Sw is significantly overestimated when the CO2 effect is neglected. At the same time X is stably around one throughout the whole experiment, which is in accordance with the unaffected quartz surface. When the CO2 effect is neglected, a reduction of the inner surface Spor (i.e. a smoothing of the mineral grains due to X < 1) is recovered, which is unlikely. Figure 14. Open in new tabDownload slide Time-lapse reconstruction of Sw and X from the σ*-data at 11.7 Hz of experiment 8. The vertical coloured lines refer to the five stages from Fig. 9(a). The computed saturations at stage 4 are plotted for all experiments in Fig. 15 and compared to the saturation measured after pressure release. Saturation is generally recovered well for most experiments. The deviation between calculated and measured saturation is smaller for subcritical conditions. The source of the deviations can be a slight variation of the sand characteristics between the individual samples, inevitable uncertainties during the moisture measurement and—especially at high pressures and temperature—the procedure of depressurization. Occasionally, pressure is released relatively fast due to technical requirements and some drying out of the sample can take place. In this case the measured saturation is lower than the original saturation at the end of active flow. Usually pressure is released very slowly. The whole procedure can take up to 2 d. In this time span the residual water might redistribute according to capillary pressure, water might rise within the sample. In this case the measured saturation in the sample centre would be larger than the saturation under pressure. Generally, saturations are reconstructed satisfactorily, as long as the CO2 dissolution is taken into account and the sample material is well known. Figure 15. Open in new tabDownload slide Measured fluid saturation versus fluid saturation calculated from measurements of σ΄ and σ″ at stage 4 by means of eq. (29). 4.4 Low frequency characteristics So far, we have concentrated on the frequency range between 0.1 and 100 Hz and one representative frequency of 11.7 Hz, which describes the effects and quantifications for the whole frequency range. The described effects and quantifications hold for this frequency range. Since the full spectra including the low frequency range may only be measured during the flow breaks, a limited amount of spectra are available. A quantitative and extensively conclusive interpretation is not possible at this stage, therefore. As a first step in assessing the behaviour at frequencies below 0.1 Hz, we apply a Debye decomposition to our data (cf. Section 2.3). Our implementation of the Debye decomposition computes the direct current conductivity and chargeabilities for a given logarithmic equidistant number of relaxation times with an iterative least squares approach. Data weighting and a regularization, which includes a penalty for the model norm and a smoothness constraint, are used. For the decomposition presented here we include the data from 0.0014–200 Hz and discretize τ with 3 relaxation times per decade ranging from 1e-5 s to 1e3 s. Since each relaxation contributes to σ* in a whole frequency range (not only at fpeak), there is also sensitivity for relaxation times with an fpeak outside of the data range. The decompositions of six spectra are plotted in Fig. 16 for the relevant range of relaxation times in terms of normalized chargeability. The fully saturated case without and with CO2 is given in the top row, partial saturation without CO2 in the central row, and cases with partial saturation during and after exposure to CO2 are plotted in the bottom row. The total normalized chargeability and mean relaxation time are given for all cases. The general behaviour of ∑νnorm is as expected from Fig. 3. ∑νnorm decreases with decreasing saturation. Since |$\sigma ^{\rm norm}_{\rm {surf}}$| is smaller than one, the presence of CO2 further reduces ∑νnorm. Figure 16. Open in new tabDownload slide Debye decompositions of selected spectra from Figs 9 and 11. Looking now at the distribution of ∑νnorm we can see that the general characteristics of the spectra are well recovered by the decomposition. The flat spectra are represented by broad distributions of relaxation times. All spectra show a more or less dominant relaxation around 0.03 s. In the case of full saturation at normal conditions (black) this is the only feature of the decomposition. According to eq. (27) this τ might be related to a characteristic diameter of approximately 12.5 μm, which could loosely be associated with the surface roughness of the quartz grains (cf. Fig. 4; Leroy et al.2008). A second feature at longer relaxation times appears for all spectra except the first one (black), which also causes |$\bar{\tau }$| to increase for all decompositions compared to the fully saturated case at normal conditions. The dominant relaxation is most pronounced for full saturation with CO2 (green), where the corresponding peak is clearly visible in the σ″ spectrum (cf. Fig. 11). However, the feature is also robustly recovered for the other CO2 experiments. For partial saturation without CO2 (dark and light grey) a similar feature is detected though broader and located at shorter relaxation times. Applying eq. (27) again, this feature would correlate to a characteristic diameter of 270 μm. Although this is roughly in the order of magnitude as the mean grain diameter, we believe a connection to be very unlikely. If the correlation existed, we would expect it to show up under normal conditions as well. We rather suspect an electrochemical process to cause this relaxation for the CO2 experiments, which might also be connected with pressure itself, since the relaxation appears in a weakened form for N2 as well. A more detailed interpretation, for example, whether the effect is controlled by kinetics or diffusion (cf. e.g. Olhoeft 1985), requires further investigations. 5 CONCLUSIONS AND OUTLOOK In our laboratory study we show measurements of the spectral complex conductivity of water-bearing sand during exposure to and flow-through by CO2. Conductivity spectra were measured successfully at pressures up to 30 MPa and temperatures up to 80 °C during active flow and at stationary conditions. Spectra could stably be measured for the frequency range between 0.0014 and 20 000 Hz. The frequency range between 0.1 and 100 Hz showed the highest measurement precision and a very consistent behaviour and is consequently most indicative for potential monitoring applications. The spectral complex conductivity during interaction with CO2 is influenced by partial saturation due to the replacement of conductive pore water with CO2 and by chemical interaction of the reactive CO2 with the bulk fluid and the grain-water interface. We could show that the impact of CO2 on the bulk fluid within the pore space is covered by the CO2-correction |$\sigma ^{\rm norm}_{\rm w}$| from an earlier study, which depends on salinity, pressure and temperature. The new data further shows that chemical interaction causes a reduction of surface conductivity, which could be related to the low pH in the acidic environment. We consider this effect with the correction term |$\sigma ^{\rm norm}_{\rm {surf}}$|⁠, which is equal to 0.81 ± 0.11 as a first approximation. The quantification of the CO2 effect may then be used as a correction during saturation monitoring. We have shown that, when the impact of CO2 is taken into account, a correct reconstruction of fluid saturation from electrical measurements is possible. In addition, we can get access to an indicator for changes of the inner surface area, which is related to mineral dissolution or precipitation processes. Changes of the inner surface area may be relevant during CO2 storage monitoring since they generally relate to changes of the hydraulic permeability. The low frequency range between 0.0014 and 0.1 Hz showed additional characteristics, which deviate from the behaviour at higher frequencies. A Debye decomposition approach is applied to isolate the feature dominating the data at low frequencies. We conclude from our study that electrical conductivity is not only a highly sensitive indicator for CO2 saturation in pore space. When it is measured in its full spectral and complex form it contains much more information on the chemical state of the system, which holds the potential of getting access to both saturation and surface properties with one monitoring method. Our results may be included in simulation and field studies concerned with the feasibility and the optimal design of geoelectric and electromagnetic monitoring methods (e.g. Spitzer 1998; Börner et al.2015b). Future laboratory work at additional pressure and temperature conditions is needed to reveal the dependence of |$\sigma ^{\rm norm}_{\rm {surf}}$| on pressure, temperature and salinity. The processes causing the behaviour at low frequencies should be investigated more intensively. Investigating the influence of other reactive gases, such as, for example, SO2, as well as experiments including a reactive rock matrix will enhance the general understanding in the future. Acknowledgments This work has been funded by the German Research Foundation (DFG) (grant numbers SP 356/12-1 and RE 1705/9-1). The authors thank Dr Franziska Wilke from GFZ Potsdam for the SEM pictures. We thank Ziqiu Xue and one anonymous reviewer for helping improve this manuscript. Additionally, Jana Börner thanks the Christiane Nüsslein-Volhard-Foundation for their support. REFERENCES Archie G.E. , 1942 , The electrical resistivity log as an aid in determining some reservoir characteristics . Trans. AIME , 146 ( 146 ), 54 – 62 Google Scholar Crossref Search ADS WorldCat Bairlein K. , Hördt A., Nordsiek S., 2014 , The influence of sample preparation on spectral induced polarization of unconsolidated sediments . Near Surf. Geophys. , 12 , 667 – 677 Google Scholar Crossref Search ADS WorldCat Börner F.D. , 1992 , Complex conductivity measurements of reservoir properties . Advances in Core Evaluation III (Reservoir Management) , 359 – 386 , Worthington P. Gordon and Breach Science Publishers Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Börner F.D. , Schön J.H., 1991 , A relation between the quadrature component of electrical conductivity and the specific surface area of sedimentary rocks . Log Anal. , 32 , 612 – 613 Google Scholar OpenURL Placeholder Text WorldCat Börner J.H. , Herdegen V., Repke J.-U., Spitzer K., 2013 , The impact of CO2 on the electrical properties of water bearing porous media—laboratory experiments with respect to carbon capture and storage . Geophys. Prospect. , 61 , 446 – 460 Google Scholar Crossref Search ADS WorldCat Börner J.H. , Herdegen V., Repke J.-U., Spitzer K., 2015a , The electrical conductivity of CO2–bearing pore waters at elevated pressure and temperature: a laboratory study and its implications in CO2 storage monitoring and leakage detection . Geophys. J. Int. , 203 , 1072 – 1084 Google Scholar Crossref Search ADS WorldCat Börner J.H. , Wang F., Weißflog J., Bär M., Görz I., Spitzer K., 2015b , Multi-method virtual electromagnetic experiments for developing suitable monitoring designs: A fictitious CO2 sequestration scenario in Northern Germany . Geophys. Prospect. , 63 , 1430 – 1449 Google Scholar Crossref Search ADS WorldCat Breede K. , Kemna A., Esser O., Zimmermann E., Vereecken H., Huisman J.A., 2012 , Spectral induced polarization measurements on variably saturated sand-clay mixtures . Near Surf. Geophys. , 10 , 479 – 489 Google Scholar Crossref Search ADS WorldCat Busch K.-F. , Luckner L., Tiemer K., 1993 , Geohydraulik , 3rd edn, vol. 3 of Lehrbuch der Hydrogeologie , Gebrüder Borntraeger Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Debye P. , Hückel E., 1923 , On the theory of electrolytes. I. Freezing point . Phys. Z. , 24 , 185 – 206 Google Scholar OpenURL Placeholder Text WorldCat Dissado L.A. , Hill R.M., 1984 , Anomalous low-frequency dispersion – near direct current conductivity in disordered low-dimensional materials . J. Chem. Soc. Faraday Trans. II , 80 ( 3 ), 291 – 319 Google Scholar Crossref Search ADS WorldCat Doetsch J. , Fiandaca G., Auken E., Christiansen A.V., Cahill A.G., Jakobsen R., 2015 , Field-scale time-domain spectral induced polarization monitoring of geochemical changes induced by injected CO2 in a shallow aquifer . Geophysics , 80 ( 2 ), WA113 – WA126 Google Scholar Crossref Search ADS WorldCat Duan Z. , Sun R., Zhu C., Chou I., 2006 , An improved model for the calculation of CO2 solubility in aqueous solutions containing Na+, K+, Ca2 +, Mg2 +, Cl− and |$\mathrm{SO_4^{2-}}$| . Mar. Chem. , 98 , 131 – 139 Google Scholar Crossref Search ADS WorldCat Duval Y. , Mielczarski J.A., Pokrovsky O.S., Mielczarski E., Ehrhardt J.J., 2002 , Evidence of the existence of three types of species at the quartz–aqueous solution interface at pH 0–10: XPS surface group quantification and surface complexation modeling . J. Phys. Chem. B , 106 ( 11 ), 2937 – 2945 Google Scholar Crossref Search ADS WorldCat IPCC, 2005 , IPCC Special Report on Carbon Dioxide Capture and Storage , Prepared by Working Group III of the Intergovernmental Panel on Climate Change , Cambridge Univ. Press Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Kim J. , Matsuoka T., Xue Z., 2011 , Monitoring and detecting CO2 injected into water-saturated sandstone with joint seismic and resistivity measurements . Explor. Geophys. , 42 , 58 – 68 Google Scholar Crossref Search ADS WorldCat Leroy P. , Revil A., Kemna A., Cosenza P., Ghorbani A., 2008 , Complex conductivity of water-saturated packs of glass beads . J. Colloid Interface Sci. , 321 ( 1 ), 103 – 117 Google Scholar Crossref Search ADS PubMed WorldCat Li D. , Duan Z., 2007 , The speciation equilibrium coupling with phase equilibrium in the H2O–CO2–NaCl system from 0 to 250 °C, from 0 to 1000 bar, and from 0 to 5 molality of NaCl . Chem. Geol. , 244 , 730 – 751 Google Scholar Crossref Search ADS WorldCat Marini L. , 2007 , Geological Sequestration of Carbon Dioxide —Thermodynamics, Kinetics and Reaction Path Modelling , vol. 11 of Developments in Geochemistry , Elsevier Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Maxwell J.C. , 1881 , A Treatise on Electricity and Magnetism , vol. 1 , Clarendon Press Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Nabighian M. , 1989 , Electromagnetic Methods in Applied Geophysics —Theory , vol. 1 of Investigations in Geophysics , Society of Exploration Geophysicists Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Nordsiek S. , Weller A., 2008 , A new approach to fitting induced-polarization spectra . Geophysics , 73 ( 6 ), F235 – F245 Google Scholar Crossref Search ADS WorldCat Nover G. , von der Gönna J., Heikamp S., Köster J., 2013 , Changes of petrophysical properties of sandstones due to interaction with supercritical carbon dioxide—a laboratory study . Eur. J. Phys. , 25 , 317 – 329 Google Scholar OpenURL Placeholder Text WorldCat Olhoeft G.R. , 1979 , Electrical properties, initial report of the petrophysics laboratory, Circular 789 , U.S. Geological Survey Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Olhoeft G.R. , 1985 , Low-frequency electrical properties . Geophysics , 50 , 2492 – 2503 Google Scholar Crossref Search ADS WorldCat Parkhurst D. , Appelo C., 1999 , User's Guide to PHREEQC (version 2)—A Computer Program for Speciation, Batch-reaction, One-dimensional Transport and Inverse Geochemical Calculations , U.S. Geological Survey Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Raveendran R. , Ikushima Y., Wallen S., 2005 , Polar attributes of supercritical carbon dioxide . Acc. Chem. Res. , 38 ( 6 ), 478 – 485 Google Scholar Crossref Search ADS PubMed WorldCat Revil A. , Glover P.W.J., 1997 , Theory of ionic-surface electrical conduction in porous media . Phys. Rev. B , 55 , 1757 – 1773 Google Scholar Crossref Search ADS WorldCat Revil A. , Skold M., 2011 , Salinity dependence of spectral induced polarization in sands and sandstones . Geophys. J. Int. , 187 , 813 – 824 Google Scholar Crossref Search ADS WorldCat Rink M. , Schopper J., 1974 , Interface conductivity and its implications to electric logging . SPWLA 15th Annual Logging Symposium , 2–5 June, McAllen, Texas , Society of Petrophysicists and Well-Log Analysts Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Sarmadivaleh M. , Al-Yaseri A., Iglauer S., 2015 , Influence of temperature and pressure on quartz-water-CO2 contact angle and CO2-water interfacial tension . J. Colloid Interface Sci. , 441 , 59 – 64 Google Scholar Crossref Search ADS PubMed WorldCat Schaef H.T. , McGrail B.P., Martin P.F., 2003 , Direct measurements of pH and dissolved CO2 concentrations in H2O–CO2–NaCl mixtures to supercritical conditions . Carbon Sequestration Second Annual Conference , Alexandria, Virginia Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Schön J. , 1996 , Physical Properties of Rocks —Fundamentals and Principles of Petrophysics , 1st edn, vol. 18 of Handbook of Geophysical Exploration , Pergamon Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Schön J. , 2011 , Physical Properties of Rocks —A Workbook , Elsevier Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Scott J.B.T. , Barker R.D., 2005 , Characterization of sandstone by electrical spectroscopy for stratigraphical and hydrogeological investigations . Q. J. Eng. Geol. Hydrogeol. , 38 , 143 – 154 Google Scholar Crossref Search ADS WorldCat Sharma M.M. , Yen T.F., 1984 , Interfacial electrochemistry of oxide surfaces in oil-bearing sands and sandstones . J. Colloid Interface Sci. , 98 ( 1 ), 39 – 54 Google Scholar Crossref Search ADS WorldCat Skold M. , Revil A., Vaudelet P., 2011 , The pH dependence of spectral induced polarization of silica sands: experiment and modeling . Geophys. Res. Lett. , 38 , L12304 , 10.1029/2011GL047748 Google Scholar Crossref Search ADS WorldCat Span R. , Wagner W., 1996 , A new equation of state for carbon dioxide covering the fluid region from the triple-point temperature to 1100 K at pressures up to 800 MPa . J. Phys. Chem. Ref. Data , 25 ( 6 ), 1509 – 1596 Google Scholar Crossref Search ADS WorldCat Spitzer K. , 1998 , The three-dimensional DC sensitivity for surface and subsurface sources . Geophys. J. Int. , 134 , 736 – 746 Google Scholar Crossref Search ADS WorldCat Ulrich C. , Slater L., 2004 , Induced polarization measurements on unsaturated, unconsolidated sands . Geophysics , 69 , 762 – 771 Google Scholar Crossref Search ADS WorldCat van Genuchten M. , 1980 , A closed-form equation for predicting the hydraulic conductivity of unsaturated soils . Soil Sci. Soc. Am. J. , 44 ( 5 ), 892 – 898 Google Scholar Crossref Search ADS WorldCat Vinegar H.J. , Waxman M.H., 1984 , Induced polarization of shaly sands . Geophysics , 49 ( 8 ), 1267 – 1287 Google Scholar Crossref Search ADS WorldCat Wagner K.W. , 1914 , Erklärung der dielektrischen Nachwirkungsvorgänge auf Grund Maxwellscher Vorstellungen . Arch. Elektrotech. , 2 ( 9 ), 371 – 387 Google Scholar Crossref Search ADS WorldCat Waxman M.H. , Smits L.J.M., 1968 , Electrical conductivities in oil-bearing shaly sands . Soc. Pet. Eng. J. , 8 , 107 – 122 Google Scholar Crossref Search ADS WorldCat Weller A. , Slater L., 2012 , Salinity dependence of complex conductivity of unconsolidated and consolidated materials: comparisons with electrical double layer models . Geophysics , 77 ( 5 ), D185 – D198 Google Scholar Crossref Search ADS WorldCat Weller A. , Slater L., Nordsiek S., Ntarlagiannis D., 2010 , On the estimation of specific surface per unit pore volume from induced polarization: a robust empirical relation fits multiple data sets . Geophysics , 75 ( 4 ), WA105 – WA112 Google Scholar Crossref Search ADS WorldCat Weller A. , Zhang Z., Slater L., 2015 , High-salinity polarization of sandstones . Geophysics , 80 , D309 – D318 Google Scholar Crossref Search ADS WorldCat Wigand M. , Carey J., Schütt H., Spangenberg E., Erzinger J., 2008 , Geochemical effects of CO2 sequestration in sandstones under simulated in situ conditions of deep saline aquifers . Appl. Geochem. , 23 , 2735 – 2745 Google Scholar Crossref Search ADS WorldCat Zisser N. , Kemna A., Nover G., 2010 , Relationship between low-frequency electrical properties and hydraulic permeability of low-permeability sandstones . Geophysics , 75 ( 3 ), E131 – E141 Google Scholar Crossref Search ADS WorldCat © The Authors 2016. Published by Oxford University Press on behalf of The Royal Astronomical Society. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. © The Authors 2016. Published by Oxford University Press on behalf of The Royal Astronomical Society.
TI - Spectral induced polarization of the three-phase system CO2 – brine – sand under reservoir conditions
JF - Geophysical Journal International
DO - 10.1093/gji/ggw389
DA - 2017-01-01
UR - https://www.deepdyve.com/lp/oxford-university-press/spectral-induced-polarization-of-the-three-phase-system-co2-brine-sand-8gczC9sNTT
SP - 289
EP - 305
VL - 208
IS - 1
DP - DeepDyve
ER -