The effect of free-phase NAPL on the spectral induced polarization signature of variably saturated soil


  • I. Shefer,

    1. Department of Civil and Environmental Engineering, Technion—Israel Institute of Technology, Haifa, Israel
    Search for more papers by this author
  • N. Schwartz,

    1. Department of Civil and Environmental Engineering, Technion—Israel Institute of Technology, Haifa, Israel
    Search for more papers by this author
  • A. Furman

    Corresponding author
    1. Department of Civil and Environmental Engineering, Technion—Israel Institute of Technology, Haifa, Israel
    • Corresponding author: A. Furman, Department of Civil and Environmental Engineering, Technion—Israel Institute of Technology. Technion City, Haifa 32000, Israel. (

    Search for more papers by this author


[1] In this study, the influence of a free phase nonaqueous phase liquid (NAPL), decane, on the soil's SIP signature was experimentally investigated. The complex electrical conductivity was determined using the SIP measurement system and compared between two main treatment types: clean and decane contaminated. Complementary chemical and temporal measurements were conducted. The results show a clear decrease in the imaginary part of the complex conductivity for the decane contaminated soil. Moreover, a shift of the relaxation frequency was observed for the contaminated soil. Our chemical analysis suggests that there was no change in the chemical composition of the Stern layer, and clearly, the grain size distribution did not change as well. Therefore, these results are attributed to membrane polarization. The decane addition to the unsaturated porous media changes the pore-scale liquid phase distribution, thus affecting membrane polarization. Further, the electrical signature is a time-related process associated with liquid phase arrangement time. The findings of this study can enable a better understanding of the SIP response for soils contaminated with free-phase organic compounds.

1. Introduction

[2] Contamination of soil and groundwater systems by organic compounds is considered a growing problem. The vast industrial use of organic pollutants and their toxic and carcinogenic nature makes the need for their removal a great concern [Schmoll, 2006]. Moreover, while the world's population is growing, the need for clean potable water is getting greater every day [Gleick, 1998]. Groundwater constitutes about one third of the world's freshwater supply [Schmoll, 2006], therefore they are of great importance alongside great vulnerability. The need to identify and locate organic pollutants in the vadose zone is an important component in groundwater protection and specifically important for achieving efficient remediation process. Of great concern are nonaqueous phase liquids (NAPLs). NAPLs do not mix readily with water and, depending on their density, can sink under or float above the water table. Complex NAPLs such as gasoline or diesel can contain many different functional groups. Some of these groups are polar and can dissolve in the water phase and as such, pose an additional threat on water sources.

[3] Traditional methods for detecting and monitoring soil contamination involve invasive soil drilling and sophisticated laboratory analysis. These techniques provide only local and time specific information of the pollution distribution and are relatively expensive [Fetter, 1999]. In recent years, there has been an increasing interest in using noninvasive geophysical techniques in general and geo-electrical methods in particular, to achieve a spatiotemporal characterization and monitoring of contaminated soils [Olhoeft, 1985; Vanhala, 1997; Titov et al., 2002; Schmutz et al., 2012; Deceuster and Kaufmann, 2012]. Spectral induced polarization (SIP) is one of these techniques and is considered as a promising method to identify organic contaminations. The reason for that is the relatively high sensitivity of this method to the pore shape and fluid content, to the mineral-fluid interfaces and to the chemical composition of the pore water [Weller and Borner, 1996]. The SIP method is based on injecting an alternating current through two electrodes and measuring the resultant potential through another pair of electrodes in the current's path. The SIP response of soils is sensitive to both energy dissipation and storage processes, often characterized by the real math formula and imaginary math formula part of the complex conductivity (often called quadrature conductivity), respectively. The soil's complex conductivity math formula can be written in Cartesian form using math formula and math formula, or in polar form using the magnitude of the complex conductivity and the phase ϕ (rad)

display math(1)

where math formula is the imaginary unit.

[4] At the frequency range typical to SIP measurements (mHz to few KHz), there are three major mechanisms that are believed to govern the electrical response of porous media [Revil and Florsch, 2010; Vaudelet et al., 2011]. Two of these mechanisms, namely Stern layer polarization and membrane polarization are related to the polarization of the electrical double layer (EDL) and the other (usually called interfacial or Maxwell-Wagner polarization) is related to polarization of the interfacial plains in a multiphase system (see section 2 for more details on polarization mechanisms).

[5] Several studies have investigated the influence of organic contaminations on a soil's electrical signature. Vinegar and Waxman [1984] investigated the influence of oil saturation on the imaginary conductivity component of shaly sand and compared the laboratory results to their model of math formula; one of their model predictions was that the imaginary conductivity component is directly proportional to the shaliness (the soil's clay content). Olhoeft [1985] studied the SIP response of soil contaminated with organic waste and reported that the measured resistivity (the reciprocal of math formula) and phase are sensitive to NAPL presence. He suggested that the organic molecules are adsorbed to the clay surface, therefore inhibiting the cation exchange process. Vanhala et al. [1992] showed an increase of the resistivity when glacial till was contaminated by toluene or heptane and a decrease of the resistivity with the addition of ethylene glycol. Borner et al. [1993] observed a decrease in the imaginary part of the complex conductivity with the addition of oil to shaly sandstone.

[6] More recently, Cassiani et al. [2009] investigated the effect of NAPL (octanol and benzene) on the SIP response of unsaturated soil with low clay content. They observed a decrease of the resistivity when air was replaced with NAPL (while water saturation was constant). They attributed the observed results to heterogeneous distribution of the NAPL due to density differences and concluded that the SIP effects of air and NAPL in their system are due to pore obstruction. Schmutz et al. [2010] reported an increase in resistivity and a decrease in the imaginary part of the complex conductivity with increasing saturation of nonwetting oil in a saturated system. In a continuing research, Revil et al. [2011] showed that for a wetting oil, both the resistivity and the phase decrease, and the normalized imaginary conductivity decreases. They explained the decrease in resistivity due to a possible increase of the cation exchange capacity (CEC) attributed to oil surface in the presence of polar molecules, which form a very conductive gel. Schmutz et al. [2012] investigated the dependence of the complex conductivity on two types of oils: strongly water repellent oils and complex oils (e.g., engine oil). They observed an increase of math formula with decreasing oil saturation of both types. However, math formula increased with the strong water repellent oil (paraffin) saturation and decreased with the complex oil saturation. They assumed that the different effect of the water repellent oils on math formula might be due to an increase of water-oil surface area. Schwartz et al. [2012] investigated the SIP signature of unsaturated porous media contaminated with diesel fuel and motor oil. They observed an increase of the real part of the complex conductivity with NAPL addition for both diesel and motor oil. Furthermore, their measurements showed a decrease of the imaginary part of the conductivity with increasing NAPL saturation. They explained these results by an exchange process between the polar organic compounds that can be found in the NAPL, and the inorganic ions originally adsorbed to the mineral surface. In their conceptual model, these inorganic ions are released to the pore water, and by that increasing their conductivity, while the organic ions are adsorbed to the mineral surfaces. Schwartz and Furman [2012] further investigate the effect of polar organic compounds on the SIP signature of soil. They were able to show that the decrease in the soil polarization is due to the low mobility of the adsorbed organic compounds on the soil mineral surface.

[7] Earlier studies show inconsistent response of SIP to organic contamination. Clearly, this is due to the existence of multiple processes that act simultaneously. Our objective in this work is to isolate the influence of free-phase NAPL on the SIP signature of unsaturated soil. Specifically, our goal is to understand how the presence of a nonconductive organic phase affects the different processes taking place in unsaturated porous media, and primarily pore-scale geometrical arrangement of the different phases, and how these processes affect the electrical signature of the soil. In order to do so, we used pure decane (a nonpolar hydrocarbon with very low solubility) to contaminate an unsaturated soil with a constant water saturation, thus creating a multiphase system where nonconductive NAPL replaces nonconductive air. This enabled the comparison between the influence of air and free-phase NAPL, on the electrical signature of soil.

2. SIP Mechanisms

[8] As mentioned earlier, the two main mechanisms that govern the low frequency polarization of soils are electrochemical and interfacial polarization. Because electrochemical polarization is related to the polarization of the soil's electrical double layer (EDL), we start with a short introduction of the soil's EDL followed by a brief discussion on electrochemical polarization. As in this work, we are mainly interested in the processes associated with electrochemical polarization, only a very short review of interfacial polarization is given at the end of this section.

[9] Most minerals in contact with water develop a net charge on their surface due to the presence of reactive surface sites or isomorphic substitutions such as in clay minerals [Stumm, 1987]. An ionic atmosphere of countercharges counterbalances this surface charge. Usually, the electrical structure near a charged surface is considered to consist of two layers (in what is known as the electrical double layer (EDL)). The first layer, known as the Stern layer, exists in the closest vicinity to the surface and consists of counter-ions (often with a hydration shell), which are strongly adsorbed to the charged surface through electrostatic and specific interactions (i.e., chemical bonding) [Adamson, 1990]. The second layer, known as the diffuse layer, extends between the Stern layer and the bulk solution where the electrical potential can be considered as neutral. The diffuse layer consists of both co-ions and counter-ions but due to the electrostatic attraction between the counter-ions and the surface, the concentration of the counter-ions is larger than the co-ions concentration [Sposito, 1984]. Overall, both the Stern and the diffuse layer compensate the surface charge so the soil remains electrically neutral.

[10] Electrochemical polarization of soils is related to the processes of migration and back diffusion of ions in response to an applied electric field [Chelidze et al., 1999]. In general, two approaches are used to describe the electrochemical polarization of a soil: (1) Stern layer polarization and (2) membrane polarization. Simplistically put, according to the Stern layer polarization model, application of an external electrical field leads to a tangential movement of the mobile counter-ions in the Stern layer [Schwarz, 1962; Revil and Florsch, 2010]. Furthermore, according to the Stern layer polarization model there is continuity of the diffuse layer and discontinuity of the Stern layer in a porous media with finite contiguity between the grains [Leroy et al., 2008]. Therefore, polarization of the diffuse layer is assumed to be infeasible and the only layer being polarized is the Stern layer [Leroy et al., 2008; Revil and Florsch, 2010]. In addition, according to Vaudelet et al. [2011] charge accumulation mainly occurs at the Stern layer because the lowest frequency applied (using the SIP), is not enough for the time required to induce polarization at the pore scale.

[11] The complex conductivity of the Stern layer model is dependent on the particle's size (radius for spherical particles), and the specific surface conductivity of both the Stern and the diffuse layer [Revil and Florsch, 2010]. The specific surface conductivity is related to the concentration and valances of the ionic species in the Stern layer alongside their mobility [Leroy and Revil, 2009; Vaudelet et al., 2011]. Thus, the Stern layer model is strongly related to the soil's solution ion composition and to the soil's solid surface physical and physicochemical properties.

[12] Membrane polarization, as opposed to the Stern layer mechanism, is associated with the diffuse layer and the pore space. It describes the accumulation of charge in the diffuse layer or in pore throats (i.e., “bottlenecks”). As a result of this accumulation, back diffusion through the pore space can be formed [Revil and Florsch, 2010]. The first concept of membrane polarization influence on the IP signature was introduced by Marshall and Madden [1959], who referred to the existence of cation and anion selective zones in the porous media. These zones can be created through changes in the pore scale (radius in simplified models) or the presence of clay minerals. As the transference of cations and anions at the different zones is different, a charge buildup at the end of each zone develops, increasing polarization. Later, Vinegar and Waxman [1984] presented a membrane polarization model for shaly sands and investigated two mechanisms that cause ion concentration gradients (thus influencing the imaginary part of the complex conductivity). The first one relates to the formation of concentration gradients at clay rich zones by counter-ion displacement. The second mechanism is electrolyte blockage in the form of cation-selective membrane in the clay-rich zones. They showed that the membrane effect is more dominant when the clay content is higher and the brine salinity is lower than 0.4M. Titov et al. [2002] have presented a membrane polarization model which takes into account the geometry of the pore space. According to this model, polarization occurs at the conductive liquid bottlenecks (contact area between the grains), between large and narrow pores that have different ion transport numbers. This model is based on the idea that ion-selective narrow passages (active zones) are considered to be much shorter in length than the large passages (passive zones), and is called short narrow pores (SNP). Their model predicts growth in the relaxation time with increasing length of ion selective zones (since the relaxation time is dependent on the square of the short pore length). More recently, Bucker and Hordt [2013] extended the membrane polarization model to account for the pore radii (in addition to the pore length) and to the effect of the Stern layer chemistry. Membrane polarization relates to transport of anions and cations in narrow and large capillaries. Therefore, polarization is strongly influenced by the ions transport number and the pore geometry.

[13] A third polarization mechanism, mostly associated with higher frequencies (roughly 100 Hz and above), is the interfacial polarization, also called Maxwell-Wagner (MW) polarization, and refers to charge buildup at interfaces that separate phases with different electrical properties (e.g., dielectric constant, electrical conductivity) [Chen and Or, 2006]. The change in electrical properties forms discontinuity of the displacement current (the current created in a charging dielectric medium) at the interfaces, thus affecting the measured soil conductivity [Revil and Florsch, 2010]. As noted above, this mechanism is relevant for high frequencies and therefore less relevant to this study.

3. Materials and Methods

[14] The soil that was used in the following experiments was red-sandy-loam (“Hamra”) which is typical to the coastal plain of Israel. This soil is composed of 93.1% sand, 5% clay, and 1.9% silt, with 0.58% organic matter content, as was tested by a hydrometer. The soil's cation exchange capacity (CEC) was obtained by a standard method [Sumner and Miller, 1996] and was found to be 3.8 meq per math formula. The air dried soil was sieved with a 4 mm sieve for all the experiments. Two major sets of experiments were held, equilibrium and kinetic. In all experiments, we used decane (density of math formula), a hydrocarbon molecule, due to its hydrophobic nature and its relatively comfortable working conditions, i.e., very low solubility ( math formula) and volatility ( math formula).

[15] In the equilibrium experiment the soil had undergone four types of treatments. In each treatment, 10 kg of soil were first mixed with 1 L of tap water (electrical conductivity of 1139 math formula). One should note that tap water has the advantage of being closer to the chemistry of the soil water and therefore, unlike a synthetic brine, will not lead to significant exchange of ions. Nevertheless, it may include small amounts of elements that may bias the results; however, as can be seen in our results, this is not the case. For completeness, the major cation concentration in tap water was analyzed and found to be 10.94, 7.55, 6.50, and 21.94 math formula for math formula, math formula, math formula, and math formula, respectively. The treatments were as follows: (a) clean—no decane added; (b) addition of math formula decane (decane to dry soil weight) to the soil; (c) addition of math formula decane to the soil; and (d) addition of math formula decane to the soil. In all cases, the water saturation was held constant and only the decane saturation varied according to the treatment.

[16] For each treatment, the soil was mixed using a large kitchen mixer for 6 min and then was hand stirred. After mixing, the soil was placed in a sealed container and set at room temperature ( math formula) for five days. This time is needed to achieve pore-scale equilibrium of the different phases. After five days, the soil was packed in PVC columns ( math formula diameter, math formula long; see Figure 1). Similar PVC sample holders were used in previous works [Ulrich and Slater, 2004; Schwartz and Furman, 2012].

Figure 1.

A sketch of the measurement system and the soil column (sample holder). In black are the measurements of the distance between the electrode in the equilibrium experiment and in red the measurements for the kinetics experiment.

[17] Four electrode ports were installed in a straight line along the column. Brass point electrodes (cylindrical with flat edge) were used; two current electrodes ( math formula long) that cross the whole column diameter and two potential electrodes ( math formula) placed slightly above the soil interface. Electrical contact between the potential electrodes and the soil was achieved by filling the empty space between the electrodes and the soil with the same soil that was used to pack the column. Packing was done by compressing small and constant portions of soil to the column. Five replicates were taken for every treatment. The electrical measurements (10 mHz to 45 KHz) were taken right after the end of the packing for each column using the ZEL-SIP04 impedance spectrometer, designed specifically for SIP measurement in porous media [Zimmermann et al., 2008]. Each column was weighed before and after packing in order to keep track of the samples porosity (see Table 1). As can be seen, the packing was highly consistent and porosity variation is limited to 1% between treatments and 0.2% between replicates. To test the accuracy of the measurement system, the electrical properties of a solution, close in conductivity to the bulk conductivity of the sample, were measured (using the proceeding describe above) and the results showed a very low error ( math formula mrad at math formula for a solution of 232 math formula).

Table 1. Mean and Standard Deviation (SD) of the Porosity (ϕ), Water Saturation (Sw), and Decane Saturation (Sd) for Both Equilibrium and Kinetic Experiments
 Equilibrium ExperimentKinetic Experiment
Treatment(a) clean(b) 0.8%(c) 1.6%(d) 3.16%(a.1) clean(b.1) 1.6%
Mean  ϕ0.3430.3420.3450.3530.3280.314
SD ϕ0.00130.00140.00370.00180.01010.0022
Mean Sw0.5080.5130.5100.4990.5430.595
SD Sw0.0030.0030.0090.0040.0250.006
Mean Sd0.0610.1200.2310.140
SD Sd0.0000.0020.0020.001

[18] To investigate the temporal change in the electrical signature of soil contaminated by NAPL, the electrical properties of soil contaminated by decane were recorded over time. Two treatments were used for this experiment. In each treatment, 10 kg of soil were first mixed with 1 L of tap water (electrical conductivity math formula). The following treatments were used: (a.1) clean—no decane added; and (b.1) addition of math formula decane to the soil (decane to dry soil weight). In this experiment the soil was mixed as described above, but was packed in the columns right after mixing (with no equilibrium time). The same type of PVC columns and the same electrodes (with a slight difference in electrode separation see Figure 1) as for the equilibrium experiment were used. Three replicates of each treatment were prepared. For each column the porosity was calculated (see Table 1). After packing, the columns were placed horizontally in order to minimize gravitational gradients, and periodic electrical measurements (at 0, 1, 2, 4, 24, 48, 72, 96, 120, and 167 h.) were taken. Laboratory temperature was held constant in that period (at math formula). All electrical measurements were taken at a frequency range of 10 mHz to 45 KHz.

[19] At the end of the experiment, soil solution was extracted from the soil by repacking small portions of the soil in a 45 ml tubes and using rhizons (Rhizon MOM 10, Rhizosphere, Wageningen, Netherlands) attached to 20 ml plastic syringe. After approximately 12–18 h, the extracted soil solutions were taken out of the syringes and measurements of both pH and EC values were taken (approximately pH of 7.6 and EC of 1450 ( math formula) were measured for the two treatments). Later, the samples' solutions were filtered (with a 0.45 math formula filter) and diluted (1:5) in order to investigate the major ion composition (K, Mg, Ca, Na, and Fe) of the solutions (using ICP-OES Spectrometer; iCAP 6300 DUE, Thermo Scientific).

[20] A separate batch experiment was held in order to confirm the column experiment chemical results. Twenty milliliters of tap water (EC = 1113 math formula) was added to 10 g of the soil in two sets of treatments: (a) clean—no decane; and (b) 1.6% decane addition. Three replicates for each treatment were shaken (at 110 rpm) for 24 h. Later, the supernatant was removed, filtered, and measurements of EC and major ion composition were taken (using ICP-OES, as mentioned above).

4. Results and Discussion

[21] In Figure 2, the real part of the complex conductivity for the different treatments of the equilibrium experiment is presented. We focus on the relatively low frequency and only results in the range of 0.01–100 Hz are presented. Seemingly, the mean conductivity of the clean soil is lower than the mean conductivity of the contaminated soil (averaging the repetitions), though the maximum difference in conductivity between clean and contaminated treatments is small ( math formula, math formula). Moreover, the standard deviation for each treatment (not shown) is of the same order of magnitude as the difference between clean and contaminated treatments (i.e., the difference in math formula between clean and contaminated treatment is insignificant). Similar observations were found by Borner et al. [1993], who investigated the influence of different organic contaminations on the measured low frequency complex conductivity of clay and shaly sandstone.

Figure 2.

The real component of the complex conductivity ( math formula) as a function of frequency for the clean and contaminated treatments examined in the equilibrium experiment. Standard deviations of the measurements are relatively large (order of 4–15 math formula) and are not presented.

[22] In Figure 3, the imaginary component of the complex conductivity for the different treatments at equilibrium time is shown. The shape of the spectra is typical for sandy soils [Leroy et al., 2008; Cassiani et al., 2009; Revil et al., 2011]: local peak value (or critical frequency, which relates to the relaxation time) at lower frequencies (due to polarization of the EDL) and increasing values with the higher frequencies (due to Maxwell Wagner polarization and capacitive coupling typical of four electrodes setup). Interestingly, the imaginary part of the complex conductivity for the contaminated samples is significantly lower ( math formula at 0.5 Hz) than that of the clean samples at all frequencies presented, and especially in the low frequency range, associated with electrochemical polarization. It is important to note that in contrast to the real part of the conductivity (Figure 2) where the standard deviation was of the same order of magnitude as of the difference between the clean and contaminated treatments, the standard deviation in this case is significantly lower than the difference. In addition, the difference in the imaginary part of the complex conductivity between the different contaminated treatments is very low, especially compared to the clean-contaminated difference, allowing clear distinction between clean and contaminated soils.

Figure 3.

The imaginary part of the complex conductivity ( math formula) for clean and contaminated treatments in the equilibrium experiment. The error bars are the standard deviation values of replicates at each treatment.

[23] Another prominent difference between the contaminated samples and the clean samples is the critical frequency, approximately 1 Hz and 0.5 Hz, respectively (these values were obtained by visual analysis of the spectra). All contaminated samples have the same critical frequency of 1 Hz (see Figure 3). The same phenomenon is also observed in the kinetic experiment (see further).

[24] In Figure 4, statistical analyses of both the real and imaginary compounds for the equilibrium experiment are shown. The p values, calculated with “Student-t test” (these values indicate whether the two treatments are likely to have come from the same population and have the same mean, i.e., if the difference between their mean values are significant or not), for math formula of the equilibrium experiment are much larger than 0.05 (significance level), while the p values for math formula are few orders of magnitude smaller (up to 10 Hz). Thus, statistically, with a 0.05 significance level, there is no difference between the mean values of the clean and 1.6% decane contaminated treatments for the real part of the complex conductivity along the whole frequency spectrum, while a significant difference between the mean of math formula between the clean and contaminated sample is verified.

Figure 4.

Statistical analysis of the (a) real and (b) imaginary parts of the complex conductivity. The blue lines represent the mean of the clean treatments while the red lines stand for the mean of all decane saturation levels. The confidence interval range for the clean treatment (blue), and for the contaminated treatments (light gray) is calculated from the standard error (1.96 × SE) and indicates 95% confidence. The p values were calculated from t-test on the mean difference between the clean and contaminated (1.6%) treatment.

[25] To further investigate the effect of free-phase NAPL on the polarization of porous media we measured the soil's solution EC and the major cation concentration (Na, K, Ca, and Mg) for both the clean and contaminated samples (see Figure 5). The EC values obtained by extracting the soil's solution at the end of the kinetic experiment showed a relatively small difference ( math formula with 6.5% standard deviation) between the clean and the contaminated treatments. Moreover, for all chemical species measured, the difference between the cation concentration of the clean and contaminated samples is negligible. Because the concentration of the different species at the EDL is related to their bulk concentration [Appelo and Postma, 2005], we argue that the difference in the chemical species concentration and composition between the clean and contaminated samples at the EDL is also negligible.

Figure 5.

Major ions composition for the clean (blue) and 1.6% decane contaminated (red) soil solutions. (a) Samples taken from the end of the kinetic experiment, Fe ion was under detection level for all the samples and the K results had a large error, apparently from measurement fault, therefore not displayed. (b) Batch experiment with the same type of treatments, ferrous was not tested this time. Note that in all cases the difference between the concentration of the different species for the clean and contaminated treatment is negligible.

[26] This observation is of great importance as for both the membrane polarization model and the Stern layer polarization model, the composition and concentration of the chemical species at the EDL play a significant role in the polarization of the soil [Vinegar and Waxman, 1984; Titov et al., 2002; Leroy et al., 2008; Vaudelet et al., 2011; Bucker and Hordt, 2013]. In the Stern layer model, polarization is dependent on the chemical composition of the Stern layer, which can be predicted using a chemical complexation model [Davis et al., 1978; Goldberg and Sposito, 1984]. In the membrane polarization model, polarization is dependent on the active zone efficiency (relates to the chemistry of the soil mineral-fluid interface), and on the cross section area of the conductive phase (i.e., water) in the pore spaces [Titov et al., 2002, 2004].

[27] The fact that the bulk solution (and with it the EDL) chemical composition do not change is evident from Figure 5, which presents major ion composition of both the kinetic experiment (Figure 5a) and the batch experiment (Figure 5b). In both experiments and for all tested cations in the soil's solution, the difference between clean and contaminated samples is negligible. Moreover, the EC (and pH) values measured at the end of the kinetic experiment showed a relatively small difference between the clean and the decane-contaminated treatments. As was shown by Schwartz and Furman [2012] and by Schmutz et al. [2010], change in the composition of the Stern layer affects polarization. Clearly, the Stern layer ionic composition did not change when decane was added to the soil.

[28] As mentioned above, a clear difference in relaxation time between clean and contaminated samples (critical frequency of 0.5 Hz and 1 Hz, respectively) was observed. In the Stern layer model, assuming spherical particles, the relaxation time is dependent on the grain radius math formula and the diffusion coefficient math formula of the ions adsorbed on the mineral surface, math formula [Leroy et al., 2008]. Nevertheless in a recent research, Revil et al. [2012] concluded that the relaxation time of sand and sandstone might be dominated by a characteristic pore size (dynamic pore radius) rather than by grain diameter. In the membrane polarization model, relaxation time is associated with the length of the short pores math formula through math formula [Titov et al., 2002]. Because in our experiments, the chemical composition of the EDL did not change (see Figure 5), we can safely assume that the diffusion coefficient at the EDL, of both the clean and contaminated samples is the same. Clearly, the grain size did not change between the different treatments (as we used the same soil). Therefore, the shift in relaxation time between the clean and contaminated treatments is likely related to changes in the characteristics of the pore length caused by the addition of NAPL, and could be explained using the membrane polarization model.

[29] Following all the above, the decrease in polarization and relaxation time with the addition of a nonpolar hydrophobic oil (decane), should in our perspective be attributed to membrane polarization due to transformations of the pore length characteristics. Addition of NAPL (such as decane) might change the “effective” pore shape and cause an obstruction of certain areas in the pore space [Titov et al., 2004; Cassiani et al., 2009]. This can affect membrane polarization and the relaxation time. We argue that in our experiment, the addition of decane to the unsaturated media alters the fluid phase distribution (primarily the conductive fluid, water) and blocks (fully or partially) narrow passages in the pore space (Figure 6), thus shortening the length of the ion selective zone causing a decrease in the relaxation time. This effect is likely to occur since the decane-water surface tension value is much higher than that of decane-air interface, 51.98 and 23.57 math formula, respectively [Zeppieri et al., 2001; Rolo et al., 2002]. This means that the water phase geometry (the conductive phase) will change at the presence of NAPL and cause pore obstruction or narrowing due to its higher surface tension with water compared to air. The decane might have changed the anion/cation selective zones location or size, which can cause a change in the concentration gradients and therefore decrease in polarization. Figure 6 schematically depicts this idea. Similar shape of the interface can be found at Chapman et al. [2013] for decane-air interface, and at Aveyard et al. [2000] for octane-water interface. We note that in addition to the difference in the interfacial properties of the different fluids, there is also a difference in the density between water and decane, and that this difference may have some effect on the fluid distribution. While quantification of the geometrical properties of the different interfaces is not in the scope of this work, we would like to emphasize that regardless of the density difference between decane and water, the fluid hierarchy at the pore scale is determined by the interfacial properties of the different fluids and that for our system the hierarchy is solid-water-decane-air [see Parker et al, 1987; Aveyard et al., 2000; Chapman et al., 2013].

Figure 6.

A sketch of a wide to narrow pore throat (passage) in an unsaturated media. (a) Before the addition of decane and (b) after decane addition (in red).

[30] Another support for the proposed mechanism comes from the insensitivity of math formula and the relaxation time to further addition of decane above the minimum amount added in the experiment (see Figure 3). According to our conceptual model, changes in polarization and relaxation time results from changes in the pore-scale geometry of the conductive phase (water), i.e., pore throats, which is due to the replacement of the solid-water-air interfaces by a solid-water-decane-air interfaces. The physical parameters that determine the geometry of the conductive phase (i.e., surface tension and contact angle) are not related to the bulk volume of the different phases, but only to the interfacial area between the phases [e.g., Martys and Hudong, 1996]. Simplistically, this means that as soon as the water-air interface is completely replaced by water-decane interface, further addition of decane will not cause further changes in the geometry of the water phase and therefore no change in the electrical spectra is expected.

[31] The temporal changes (the kinetics experiment) of the real and imaginary components of the complex conductivity at the critical frequency of 1 Hz are shown in Figures 7a and 7b, respectively. Note that the ending point of these experiments is not the data used for Figures 2 and 3. For both the real and imaginary parts of the complex conductivity, there is a notable difference between clean and contaminated samples. At the beginning of the experiment, for both the clean and contaminated samples, math formula increases with time. For the clean sample math formula increases up to 48 h from the beginning of the experiment, while for the contaminated sample stabilization starts at 72 h (see Figure 7a). From this moment, the difference between the bulk conductivity of the clean and contaminated samples is roughly constant ( math formula), except for the last measurement of the contaminated sample (at math formula). In Figure 7b, the temporal change in the imaginary part of the complex conductivity of the clean and contaminated samples is shown. For the clean treatment, the results indicate that math formula does not significantly change over time. In contrast, the imaginary part of the conductivity for the contaminated treatment starts to decrease at about 24 h after packing and continuously decreases for all the experiment period. The maximum difference between the imaginary part of the conductivity of the clean and the contaminated samples (32%) is observed at the end of the experiment ( math formula). It is important to note that at the early stages of the experiment ( math formula) the imaginary part of the conductivity (at 1 Hz) is roughly the same for both the clean and contaminated samples. However, as time passes, the spectra of the contaminated samples not only decrease relatively to the clean samples, but also the relaxation frequency of the contaminated sample shifts toward the higher frequencies.

Figure 7.

(a) The real and (b) the imaginary part of the complex conductivity along time in the critical frequency of 1 Hz (of decane contaminated sample). The error bars are the standard deviation values of replicates at each treatment.

[32] These dynamic results support our conclusion regarding the role of free-phase NAPL in membrane polarization as they provide a second evidence for the decrease in polarization with the addition of NAPL. That is as the time required for exchange process, such as in the Stern layer model, is in the order of several hours [Vaudelet et al., 2011; Schwartz and Furman, 2012] and not days as seen here. Moreover, these results suggest that the membrane polarization due to soil contamination is a time dependent process and, at least in time scales of days, temporal changes in the electrical signature should be considered. This is probably due to the time required for the different phases (water-NAPL-air) to reach stabilization at the pore scale [Frey et al., 2012].

5. Summary and Conclusions

[33] The main objective of this study was to examine the influence of free-phase NAPL on the SIP response of an unsaturated soil. Laboratory experiments including geophysical and chemical measurements were conducted in order to understand the different processes occurring in the clean and contaminated soil. The study's results show that addition of free-phase NAPL to the soil is involved in a substantial decrease of the imaginary part of the complex conductivity, i.e., polarization, and a clear increase in the relaxation time.

[34] We argue that the polarization mechanism causing these observations is membrane polarization. We support this argument with chemical evidences that assure that the alternative mechanism of Stern layer polarization cannot explain the observations. Since no change in the solution major ion composition was observed, we conclude that no ion exchange processes took place and there was no change in the Stern layer ion composition. Clearly there was also no change in grain size. Hence, the Stern layer polarization model that depends primarily on the chemical composition of the stern layer and on grain size distribution could not explain these findings. Moreover, we have witnessed a consistent decrease in the relaxation time between clean and contaminated treatments and explain it, as well, with the membrane polarization mechanism. We suggest that addition of free-phase NAPL to the porous media alters the pore space and changes the pore throat's characteristic length, thus affecting the membrane polarization and relaxation time. The membrane polarization, as presented in the results section, is a time-related process and is subjected to temporal changes. These are related to the relaxation time of the liquids rearrangement, following the addition of another phase to the system.

[35] This study emphasizes the characteristic influence of a free-phase organic contamination on the soil's SIP signature. It allows isolating the effect of free-phase from the effect of the other materials that present in most commercial NAPL. Nonetheless, further work is needed to understand the temporal process associated with the addition of NAPL to the soil and perhaps for a greater time scale.


[36] This research was supported in part by the Israel Science Foundation (grant 381/09). We would like to thank Andreas Hordt, Konstantin Titov, and another anonymous referee for their detailed and constrictive review. We also thank the Associate Editor Andrew Binley for his deep and useful comments.