Remediation of soils contaminated with total petroleum hydrocarbons through soil washing with surfactant solutions

ABSTRACT Soil fulfils vital functions for life on Earth and so, just like water and air, its protection from all sources of contamination is a major concern. However, the extensive use of petroleum derived products, either as energy sources or as commodities, leads to important environmental liabilities. Ex situ soil washing is a technology to concentrate contaminants, allowing soil cleaning and the reuse of extracted petroleum derived products. This work focuses on the optimization of ex situ soil washing process using surfactants, introducing an evaluation of the washing solution recycling and its after use safe disposal, promoting the reduction of raw materials, energy and water resources costs. Two surfactants, sodium dodecyl sulphate (SDS) and polyoxyethylene sorbitan monooleate (Tween 80), were tested in the decontamination of an artificially contaminated soil with engine lubricant oil waste. The optimization of the washing conditions, such as stirring speed, liquid–solid ratio, number of washing stages, and surfactant concentration, was carried out using a design of experiments (DOE) software, so that the maximum extraction efficiency of total petroleum hydrocarbons (TPHs) was achieved. A TPH removal efficiency of (80.7 ± 3.2)% was obtained with Tween 80 after 5 h of washing and (90.7 ± 2.8)% with SDS after 2 h at 200 rpm on an orbital shaker with a liquid to solid ratio (L/S) of 15. The potential for reuse of the washing solutions was evaluated. Finally, the discharge of the washing solution was considered using activated carbon to remove the surfactants and ensure its safe disposal. GRAPHICAL ABSTRACT


Introduction
The growth of industrial activity motivated by population growth led to an increase in oil exploration and commercialization [1], and concomitantly to an increasing number of environmental liabilities.In Europe, the main sources of contamination are related to the inappropriate use and disposal of waste (38.1%), industrial and commercial activities (34.0%), poor storage (10.7%) and accidental spills (7.9%) [2].According to the European Environment Agency, the total petroleum hydrocarbons (TPHs) contamination represents more than half of the environmental liabilities and includes, among others, mineral oils (23.8%), polycyclic aromatic hydrocarbon (10.9%), monoaromatic hydrocarbons (10.2%), halogenated hydrocarbons (8.3%), and heavy metals (35%) [2].These sources of contamination create imbalances in the environment, which have negative economic and social impacts, devaluing the affected areas [3,4].On the other hand, due to their infiltration in the soil, contaminants are able to reach the aquifers used as population water supply sources, causing health problems [5,6].
Several biological, chemical, physical, and thermal techniques, both in situ and ex situ, have been developed to treat contaminated soils [7].Biological remediation techniques include bioremediation, which involves pollutants degradation through microbial/enzymatic activity, and phytoremediation, which is based on the natural capacities of plants to extract, filter, stabilize, degrade and volatilize organic pollutants from the affected land [7,8].These methods are suitable for soils with high levels of organic matter, such as peat soils, which serves as food for the plants and microorganisms involved.Generally, these techniques are favoured by the mechanisms of bioaugmentation, with the increase of the autochthonous or exogenous microbial population, biostimulation, with the addition of nutrients, and bioventilation.Despite their simplicity and low cost, these methods do not allow satisfactory efficiencies in the remediation of soils with high levels of hydrocarbon contaminants, since they are toxic and even lethal to most microorganisms [7].Furthermore, these biological techniques are affected by environmental conditions.In fact, as demonstrated by Venosa et al. [9], the degradation of crude oil of the Delaware coast in nutrient-enriched water, with and without the addition of microorganisms on site, led to similar results to those obtained with the natural attenuation of the soil.These results were due to the high levels of nitrogen in the soil itself, which allowed some biodegradation of the pollutants.
Physical-chemical and thermal methods are more intensive for the soil matrix, requiring greater amounts of energy and higher costs than the biological ones, but allowing for better remediation efficiencies.Generally, physical-chemical methods can be divided into processes of immobilization, decomposition or extraction.Thermal processes such as thermal desorption, pyrolysis, gasification and incineration involve heating the soil using different temperatures and oxygen ratios [10].These processes allow a quick treatment, with a significant reduction in residues volume and with high efficiencies in the removal of pollutants from heterogeneous soils with high levels of hydrocarbons [11].However, these processes are energy intensive, which leads to high costs, and high greenhouse gases emissions [12].Immobilization techniques such as sanitary landfills are very attractive due to their simplicity, but there is usually a devaluation of the surrounding areas and also the risk of toxic substances leaching.Furthermore, in a circular economy perspective, landfill disposal should be the last alternative to be considered because it does not allow waste reuse [13,14].In this way, other alternatives have been developed to solve the problem of soil contamination [15].
Washing methods, such as soil flushing and soil washing, have shown satisfactory results in the removal of hydrocarbon contaminants from soils with the use of surfactants [6,[16][17][18][19][20][21][22].Moreover, soil washing is quite versatile since it allows to overcome soil heterogeneity problems that greatly influence several parameters like the permeability and contaminant distribution [23,24].Generally, non-ionic and/or anionic surfactants are chosen since cationic surfactants are strongly adsorbed by clay minerals such as aluminosilicates and soil organic matter [6,25].On the other hand, parameters such as the toxicity and biodegradability of the surfactants also play a key role [6,7,26].This justifies the choice of Tween 80 and sodium dodecyl sulphate (SDS), which are more than 96% biodegradable, when compared to other surfactants like Tween 20 and sodium dodecylbenzene sulfonate (SDBS) [21,27].Nevertheless, regarding toxicity, Tween 80 is slightly toxic, with an EC 50 of 70 mg/L, while SDS is quite toxic with an EC 50 of 1.0-13.9mg/L [6,21].Thus, this requires special care in the disposal of SDS solutions in the environment.
Several studies aiming at the remediation of soils contaminated with TPHs have been carried out using the washing soil technique.According to Zhao et al., removal efficiencies of 99% of phenanthrene (100 mg/ kg) from sandy soil were obtained using concentrated aqueous solutions of Tween 80 (720-4000 mg/L), 13 times above their critical micellar concentration (CMC), after 72 h of washing at 25°C [20].When comparing these results with those obtained by other authors [18,28], there is a reduction in TPH removal efficiencies from 99% to 85% and 74% as the clay and silt content increases from 10.3% to 25% and 40%, respectively, for similar operating conditions [18,20,28].Despite being more toxic, the use of SDS in soil remediation has allowed higher TPH removal efficiencies.Gitipour showed that it was possible to achieve efficiencies of 78.6% with an SDS solution in a sandy soil contaminated with 9000 mg/L cresols and according to Khalladi, a removal efficiency of 97% of diesel was achieved at 3450 mg/kg in a soil with 94% silt with a solution of 576 mg/L of SDS [22,29].
Despite the undeniable efficiency of soil washing, the high costs of this process require the optimization of the operating conditions as a crucial step for its implementation [30].Generally, the most relevant parameters to be optimized are the surfactant concentration, liquidsolid ratio, washing time, mixing conditions and the number of successive washing steps [17,31,32].For example, in a soil washing experiment for polycyclic aromatic hydrocarbons removal using Tween 80 at 156 mg/ L, the change in L/S from 1 to 4 and 15 allowed an increase in efficiency from 1% to 40% and 75%, respectively [32].In the same study, a fast kinetics was shown to be highly desirable since the washing efficiency changed from 45% to 75% for 30 and 180 min [32].
An aspect that is hardly ever studied is the possibility of regeneration, recycling and/or recovery of the solvents and/or surfactants used in the washing step, allowing for a closed recirculation of the washing solution, and thus more sustainable processes.From an industrial point of view, the solvent recirculation is extremely important as it will allow to decrease the total cost of the process and the impact on the environment.On the other hand, the treatment of the aqueous washing solution before its discard involves separation processes, and if necessary, a chemical or biological degradation process [33].One of the most efficient and thus widely used separation process is liquid-liquid extraction.However, this process typically uses hydrophobic solvents that are generally highly toxic and require expensive distillation processes to regenerate the solvents and recover the extracted contaminants [33].Also, the presence of surfactants hinders the final efficiency due to emulsification.The use of adsorption processes in wastewater treatment has grown enormously in recent years, where activated carbon is the most used adsorbent media due to its advantageous balance between cost and removal efficiency [34,35].Nevertheless, the surfactants are also retained in its pores, preventing their recovery for further use.
In this work, a circular process based on the use of soil washing to remove engine lubricant oil from contaminated soils is proposed.For that purpose, two different surfactants, Tween 80 and SDS, were chosen due to their different chemical structures which confer them different properties, namely their critical concentration (CMC) values.The fact that the CMC of SDS (8.05 × 10 −3 M) is 3 orders of magnitude lower than that of Tween 80 (9.92 × 10 −6 M) allows to study the importance of different extraction mechanisms, such as mobilization and/or solubilization, depending on if the surfactant concentration is below or above the CMC, respectively [36].
They are both cost-effective widely used surfactants, both biodegradable but, as mentioned above, SDS is much more toxic (EC 50 = 1.0-13.9mg/L) than Tween 80 (EC 50 = 70.0mg/L) [21,27].Also, the fact that SDS is an anionic surfactant and Tween 80 a neutral surfactant also brings different behaviours in solution.In fact, the first repelling negatively charged particles from the soil, which are thus little adsorbed by soil, whereas the opposite behaviour is observed for Tween 80.
Important process variables, such as residence time, stirring speed and liquid-solid ratio, were optimized using a design of experiments STATISTICA software, so that the conditions that maximize the TPH extractions are determined.To convey sustainability to the process, the possibility of reusing the washing solutions and also the implementation of multiple stages versus single stage process are evaluated.Finally, the cleaning of these washing solutions using activated carbon to remove the organic matter prior to their discard is also studied.

Soil preparation
A soil from Leiria, Portugal (39°41 ′ 41 ′′ N 8°55 ′ 32 ′′ W) was collected from uncontaminated site at the depth of 40-60 cm [37].Then, the soil was sieved to remove the coarsest fraction (d > 2mm) and then air dried and quartered.Some properties such as the porosity, humidity and the organic matter content were determined.Afterwards, the finest fraction (<2 mm) was artificially contaminated at 5% (m/m) with engine lubricant oil supplied by the F.S. Portugal car workshop through rigorous mixing.The necessary quantity of the contaminated soil was prepared as the tests were carried out.It was left for 4 days at room temperature in closed plastic container before use.The lubricant oil given by the supplier is a mixture of used motor oil, composed of mineral, semisynthetic and synthetic oil fractions.The concentration of contaminants applied is within the range of percentages commonly used in this type of soil remediation studies [15,21,31,38].

Surfactants
The surfactants used were polyoxyethylene (20) sorbitanmonooleate (Tween 80) of 99.98% purity supplied by PanReac AppliChem, and SDS of 98.5% purity supplied by Sigma-Aldrich.These surfactants were selected due to their physicochemical characteristics, namely molecular weight (MW), low critical micelle concentration (CMC), high biodegradability, low toxicity in case of Tween 80, a high hydrophilic-lipophilic balance (HLB) value and low cost [6,16,21,27,39].The relevant properties of the two surfactants chosen are shown in Table 1.

Soil washing experiments
Soil washing experiments were carried out by weighting 5 g of the contaminated soil samples that were transferred to Erlenmeyer flasks, where 50 mL of the surfactant solution (liquid/soil ratio of 10:1) were added.
Different concentrations of the surfactant's solutions were prepared.A blank assay with Milli-Q water was also carried out as a control experiment.The soil washing was carried out in an orbital shaker (Agitorb200, Aralab).
After the pre-established contacting time and decantation, the mixture was vacuum filtered and dried at 105°C until constant weight.The amount of TPH removed was analysed using a gravimetric method before and after the soil washing.Escosteguy et al. had a positive correlation for the determination of organic matter with the thermogravimetric method, allowing the substitution of the Walkley-Black standard-method, which uses dichromate [40].In addition to authors Kristl et al. and Bensharada et al., this method is quite feasible for the determination of organic matter [41,42].According to Ahn et al., the thermogravimetry method is quite reliable for the determination of lubricant oils in contaminations [43].In the gravimetric method, the soil sample was placed in an oven and submitted to the following optimized temperature protocol: heating from the room temperature to 400°C in 30 min, remaining at this temperature for 2 h, and then increasing the temperature to 700°C in 2 h, remaining at this temperature for 3 h.Triplicates were carried out for all the experiments to allow the calculation of standard deviations.
The content of TPH removed in the washing was determined by Equations ( 1)-( 3) where C i (g/g) is the initial concentration of TPH in the soil, C f (g/g) is the final concentration of TPH after soil washing, M OM (g) and OM(%) are the mass and the percentage, respectively, of organic matter in the soil, M AO (g) is the mass of dried soil, after being in the oven at 105°C (g) and M AM (g) is the mass of calcinated soil after the temperature treatment (g).The optimization of the relevant experimental parameters, such as surfactant concentration, from 9.92 × 10 −4 M (100 × CMC) to 6.95 × 10 −3 M (700 × CMC) for Tween 80 and from 5.75 × 10 −3 M (0.71 × CMC) to 6.95 × 10 −3 M (0.86 × CMC) for SDS, liquid-solid ratio at 5:1, 10:1 and 15:1, stirring speed at 100, 150, and 200 rpm and time from 1 h to 72 h, was carried out.The studied concentrations of both surfactants are similar in mass concentration but rather different in molar concentrations due to the high molar mass of Tween 80.The liquid-solid ratio and stirring speed were chosen based on literature results for TPH contaminants, and scalable at the industrial level [6,20,21].

Kinetic models
A kinetic study of the removal of TPH from soil was also carried out.For that purpose, analysis of TPH content in soil samples was carried out after different washing times (1, 2, 5, 16, 24, 36, 48 and 72 h).The results of the effect of the contact time on the TPH concentration in the soil were fitted to the pseudo-first-and -secondorder models, described by Equation ( 4), with n equal to 1 or 2, respectively: where k 1/2 represents the pseudo-first-or -second-order adsorption rate constant, q e and q t correspond to the amounts of contaminant adsorbed by the surfactant at equilibrium and time t, respectively.

Data analy sis and optimization using STATISTICA software
An optimization of the experimental variables that lead to the maximization of the TPH removal efficiencies was carried out using the STATISTICA for Windows version 10 software.The results were obtained using an experimental design (DOE) in central composite, nonfactorial, surface designs.
It was also possible to find a relationship between the response variable (y), that is, the TPH removal efficiency, and the independent experimental variables (x i ), time, stirring speed and L/S through a pseudo-second-order polynomial model given by where k is the number of independent variables, b 0 is the intercept parameter and b i , b ii and b ij are the regression parameters for linear, quadratic and interaction effects between variables.

Reuse of the washing solution
The reuse of the contaminated aqueous solutions of SDS and Tween 80 was investigated.These solutions went through a liquid-liquid separation in a decantation funnel.After separating the aqueous solution from the oil phase, the aqueous solution was centrifuged at 3000g for 10 min by the Sigma Centrifuge 4-16S.The reuse of the solution was tested, with and without the addition of more surfactant.

Treatment of the disposal solution
The effluent resulting from the washing processes was collected and proceeded to a batch adsorption test with Arkema's Acticarbone® 830WLP Chemical Powders granular activated carbon.The adsorption experiment with activated carbon was carried out under orbital stirring at 200 rpm and a liquid-solid ratio of 10 for 5 h.At the end, the pulp was vacuum filtered and the evaluation of the efficiency of the process was carried out by measuring the chemical oxygen demand (COD).

Soil classification
According to the granulometry of the soil grains on the Atterberg scale, the collected soil sample is classified as sandy soil, given its high coarse sand constitution as presented in Table 2.

Washing efficiencies and experimental parameter optimization
The TPH present in the soil after the washings was evaluated through the previously described thermogravimetric method.The percentage of TPH recovered using this method of analysis was 4.44 ± 0.01% (m/m) of the 5% (m/m) present in the contaminated soil.
Figure 1 shows the TPH removal efficiencies for two different concentrations of each surfactant, together with the control essay with milli-Q water, while the respective data are in Table S1 in Supplementary Information.It can be observed that there was only a slight removal (3.4 ± 2.1)% of TPH after washing with milli-Q water (control sample) at 100 rpm, even for a long period of time, which is probably due to mechanical mechanisms, as stirring and friction between soil particles, given the insoluble character of TPH in water.However, the importance of these mechanisms is well illustrated when using Milli-Q water at 200 rpm for 16 h, where a substantial increase in the removal of the pollutants to (12.8 ± 3.7)% was observed.
For soil washing under standard conditions (L/S = 10, 72 h of washing, 150 rpm), a TPH removal efficiency of 68.9% ± 2.7% for Tween 80 and 92.2% ± 1.6% for SDS was obtained.Please note that these comparisons were carried out at similar molar concentrations but correspond to concentrations above CMC for Tween 80 (9.92 × 10 −6 M) and below CMC for SDS (8.05 × 10 −3 M).The high efficiency of TPH removal using SDS solutions with concentrations below the CMC shows that the mobilization mechanism plays an important role.However, since SDS in solution are close to CMC, the removal of hydrocarbons from the soil can also be due to the reduction in the interfacial tension between the soil surface and the contaminants [44].As for Tween 80, which was used in concentrations much higher than the CMC, mobilization and solubilization mechanisms act simultaneously and are both responsible for the TPH removal.To confirm this, an extra test was performed at a Tween 80 concentration equal to 5 × CMC, where a removal efficiency of only 16.6 ± 1.4% is obtained, very close to that obtained for the control.Note that even though this concentration is still above CMC, it is a very low concentration.Thus, it can be concluded that for Tween 80 to be effective in the removal of TPH from soil, it is crucial to have concentrations much higher than the CMC.The difference in TPH removal efficiencies reached with the two surfactants can be related to the different behaviour they have in solution.Anionic surfactants, by negatively charging oil droplets, attract water molecules and positively charged counter-ions from the aqueous solution and repel the negatively charged particles from the soil, which are little adsorbed by soil.This last phenomenon occurs due to the predominance of silicates, SiO 2 , in the earth's crust [45], where clay silicates, essentially consisting of hydrated aluminosilicates, dominate as a result of partial replacement of silicon atoms, Si 4+ , of SiO 2 by Al 3+ in its tetrahedral structure, causing a charge deficit for each substitution, and thus generating an electrostatically unbalanced structure leading to a preferential cation adsorption capacity [45].In the case of non-ionic surfactants, such as Tween 80, the soil particles are not repelled and can be easily adsorbed by surfactant.Therefore, a greater amount of Tween 80 is required to achieve similar TPH removal efficiency obtained with SDS.
It can also be seen in Figure 1 that when using the same concentration of both surfactants' solutions, 6.95 × 10 −3 M, which corresponds to 0.86 CMC of SDS and 700 CMC of Tween 80, the SDS solution has a significantly faster kinetics, reaching equilibrium (>90%) around 2-5 h, whereas for the solution of Tween 80 a removal of 45% only was achieved in 5 h, increasing slowly to 90% after 72 h of washing.It should be noted that the high removal values are also promoted by the type of soil used in the tests (sandy soil), because according to other authors, the remediation of sandy soils presents better TPH removal efficiencies than clay soil [46,47].This difference can be due to the different porosity of the soils.In fact, clay soils have smaller particle size, a larger pore volume and surface area thus presenting greater retention capacities in comparison with sandy soil [6,48].This difference explains why clay traps impurities more effectively than sandier soil with bigger particles [47].As the contaminant penetrates deeper into the clay soil and forms a strong bond with it, it becomes more difficult to remove these contaminants from this soil using the surfactant-mediated method [6,46].
Two kinetic models, a pseudo-first-and -second-order models, were tested to describe the kinetics of the washing processes with both surfactants.As it can be seen in Table S1 in Supplementary Information, the pseudo-second-order model gave the best fitting traduced by higher values of the correlation coefficient for all the experiments.Another important finding is that the increase in the concentration of both surfactants leads to an increase in contaminants adsorption at the equilibrium (q e ) and in the adsorption rate of the contaminant in the surfactant (k 2 ).

Effect of the stirring speed
In Figure 2, the effect of changing the stirring speed from 100 rpm to 200 rpm for the two surfactant solutions is presented.It can be immediately observed that increasing the stirring speed from 100 rpm to 200 rpm has a significantly higher impact on the TPH removal efficiencies for Tween 80 than for SDS solutions.In fact, after 72 h of washing with Tween 80 aqueous solutions at 100 rpm, an efficiency of 24.2 ± 2.3% was obtained, while at 150 and 200 rpm, efficiencies of 68.9 ± 2.7% and 95.6 ± 2.3% were obtained, respectively.The efficiency of TPH removal obtained with SDS aqueous solutions at 100 rpm is already high (76.8%).However, an increase in the stirring speed allows reaching excellent removal efficiencies (96.9% for 200 rpm).These differences in the TPH extraction efficiency due to stirring speed are probably due to the different removal mechanisms previously proposed for the two surfactants, mobilization and solubilization for Tween 80 and mobilization for SDS, being solubilization much more dependent on mixing that mobilization.On the other hand, the adsorption of Tween 80 in the soil matrix can be favoured by the absence of negative charges.However, this phenomenon can be attenuated increasing the stirring speed, which promotes greater availability of the surfactant in solution, increasing its  extraction efficiency.In SDS, the presence of negative charges hinders its adsorption on the soil, allowing high efficiencies to be obtained in a shorter period of time than Tween 80.
The fitting parameters and the correlation coefficient of the pseudo-first-and -second-order kinetic models to the experimental data are presented in Table S2 in Supplementary Information.When comparing the R 2 of the fittings of the two models, again the pseudo-secondorder model provides better fittings than pseudo-firstorder.For both surfactants, the increase in the stirring speed increases the amount of contaminant adsorbed in equilibrium (q e ).However, for both surfactants, the kinetic constant decreases when the stirring speed is increased from 100 rpm to 150 rpm but increases when the stirring speed is further increased to 200 rpm.Despite these oscillations in the kinetic constants with the stirring speed, the overall TPH removal efficiency increases with stirring speed for both Tween 80 and SDS, which leads to the conclusion that q e plays a dominant role in the extraction process.This phenomenon can be explained by the better fit of the mathematical model for longer washing times, but the less accurate fit for the experimental points for initial times caused some variations in the prediction of kinetic constant with the stirring speed.

Effect of the liquid-solid ratio
In Figure 3, the effect of liquid-solid ratio on the TPH removal for the two surfactant solutions is presented.As expected, the use of high quantity of surfactant solution facilitates the contaminants removal.In fact, high L/S ratios allow to have high TPH concentration gradients between the two phases, thus increasing the contaminant mass transfer from the solid to the aqueous solution.Nevertheless, the cost of the process also greatly increases with the increase of the L/S ratio and thus it is very important to determine the optimal value for this parameter.
Regarding the kinetic model that best describes the results, Table S3 in Supplementary Information shows that the pseudo-second-order model provides a better description of the experimental data for both surfactant aqueous solutions.It can also be observed that the increase in the L/S ratio increases the amount of contaminant adsorbed in equilibrium (q e ) and the adsorption rate k 2 .

Optimization of experimental parameters using STATISTICA software
The previous results were obtained using the classic methodology, varying one factor at a time while keeping the other variables constant.This methodology  is useful to obtain a preliminary knowledge of the system behaviour.However, it only allows to detect the individual impact of the different variables on the system response but not the interaction between these variables.On the contrary, factorial designs, which are based on the simultaneous variation of all factors in a limited number of levels, allow the determination of interactions between factors.Furthermore, in the response surface designs where the factors are varied in three levels, this procedure also allows to find a model equation to describe the system response as a function of the variables [49].
The STATISTICA software was used to generate a mathematical model that correlates the experimental data and allows to estimate the optimal washing conditions for each surfactant used.Using the leastsquares method, it was possible to estimate the polynomial equation of the TPH removal efficiencies as a function of time, stirring speed and liquid-solid ratio, allowing the evaluation of the linear and quadratic effect of each of the variables and the interactions between two variables.
Thus, using this software, it is possible to predict the values of the TPH removal efficiencies for the washing experiments with Tween 80, according to Equation ( 6): where X 1 is the time parameter in hours, X 2 is the stirring speed in rpm and X 3 is L/S in g/g.The mathematical model fits well with a correlation coefficient (R 2 ) of 0.9556 and a root mean square error (RMSE), calculated by Equation ( 7), of 0.0568.
where y i and ŷi are the experimental and predicted values of TPH removal efficiencies and n is the number of experiments.However, the generated equation has a high number of coefficients, totalling 10 terms of different orders of magnitude.In this way, a refinement of the model is desirable to identify the most significant terms, avoiding over-fitting of the data while keeping the model's performance.Thus, the Pareto diagram shown in Figure S1 in Supplementary Information was analysed, and the least significant terms, L/S-time and L/S-stirring speed, were removed, one by one, evaluating their impact by analysing the R 2 and the RMSE shown in Figure S2(a) in Supplementary Information.It can be concluded that it is possible to predict the values of washing efficiencies with Tween 80, using an equation with only four coefficients, plus the independent term, according to Equation (8).
where X 1 is the time parameter in hours, X 2 is the stirring speed in rpm and X 3 is L/S in g/g.
The refined model provides a good description of the experimental results, with R 2 of 0.9422 and an RMSE of 0.0648.For the validation of the model, some experimental points were carried out.The performance of the simplified model for Tween 80 washes can be seen in Figure 4(a) where the predicted versus experimental data were plotted.
Using Equation ( 8), it was possible to obtain the contour plots presented in Figure 5, where the influence of the three variables on the washing efficiency of TPH from soil with Tween 80 solutions is presented.These plots show the effect of changing two variables when the third is fixed at its standard conditions value.Thus, for L/S = 10 (Figure 5(a)), removal efficiencies greater than 80% can be obtained only for contact times higher than 20 h and for stirring speeds higher than 178 rpm.The surface plot of L/S versus time (Figure 5(b)) shows that TPH removal efficiencies greater than 80% can be achieved for stirring speeds of 150 rpm after 38 h of washing with a L/S higher than 12. Finally, it is possible to remove more that 80% of the contaminants in 24 h for L/S above 10 and for stirring speeds greater than 170 rpm (Figure 5(c)).
The maximum TPH removal efficiency predicted by the STATISTICA software for the Tween 80 solutions was (77.4 ± 4.3)%, achieved in a washing time of 24 h, stirring speed of 175 rpm and L/S of 12.5.
The STATISTICA software was also used to generate a mathematical model that correlates the experimental data for TPH removal using SDS washes, obtaining Equation (9).
where X 1 is the time parameter in hours, X 2 is the stirring speed in rpm and X 3 is L/S in g/g.The mathematical model provides a good description of the experimental results, with R 2 of 0.9175 and RMSE of 0.04853.Again, a study of the Pareto diagram, represented in Figure S3 in Supplementary Information, allows for a refinement of the model.It can be observed that the term L/S (quadratic) is the least relevant, followed by stirring speed-L/S.In this way, the least significant terms from the model were removed one by one, and their effect on the model is analysed through the values of the correlation coefficient (R 2 ) and the RMSE is shown in Figure S2(b) in Supplementary Information.In conclusion, it is possible to predict the values of SDS washing efficiencies, using an equation with four coefficients, plus the independent term, according to Equation (10).
where X 1 is the time parameter in hours, X 2 is the stirring speed in rpm and X 3 is L/S in g/g.This refined model provides a good description of the experimental results, with R 2 of 0.9044 and an RMSE of 0.0523.Again, for the validation of the model, some experimental points were measured.The performance of the refined, simplified model for the SDS can be seen in Figure 4(b) where the predicted versus experimental data were plotted.
The contour plots presented in Figure 6 show the influence of the three variables on the washing efficiency of TPH from soil with SDS solutions.It can be observed that for L/S = 10 (Figure 6(a)), removal efficiencies greater than 80% can be obtained only for contact times higher than 10 h and for stirring speeds higher than 100 rpm.The contour plot of L/S versus time (Figure 6(b)) shows that TPH removal efficiencies greater than 80% can be achieved for stirring speeds of 150 rpm after 9 h of washing with a L/S higher than 5.1.Finally, it is possible to remove more that 80% of the contaminants in 24 h for a L/S above 6 and for stirring speeds greater than 100 rpm (Figure 6(c)).
The optimized point predicted by STATISTICA software for the SDS washes was (82.1 ± 3.2)%, achieved for a wash time of 23 h, stirring speed of 150 rpm and L/S of 10.
Through the prediction of results provided by STATIS-TICA software, it can be concluded that the stirring speed is the most relevant parameter in obtaining better TPH removal efficiencies using Tween 80, while the time parameter is the most relevant for SDS washes, which is according to the conclusions drawn by the adjustment  of the pseudo-second-order equation performed.Overall, it can be concluded that the models obtained by STATISTICA software allow not only the identification of the most significant experimental parameters in soil washing using both SDS and Tween 80 surfactants but also the confident prediction of the experimental points and the optimal experimental conditions for the best performance.

Successive washes
To check if the removal efficiency with time increases using fresh washing solutions, successive washing stages with 2 h under standard conditions were carried out, using L/S = 10, 150 rpm and 100 CMC for the Tween 80 and 0.71 CMC for the SDS.The experimental procedure used in these washing stages consisted in washing for 2 h with the surfactant solution, and after filtration, a fresh surfactant solution with the same concentration was added to the soil for another 2 h of contact at 150 rpm.As shown in Figure 7, for the washing in several stages with Tween 80 aqueous solution, a TPH removal efficiency of TPH of 37.8 ± 1.8% after two washing stages was obtained, which surpasses the efficiency of 30.6 ± 1.6% obtained with one stage after 5 h.On the other hand, in three stages a TPH removal efficiency of 44.0 ± 4.3% was obtained, which is similar to that obtained in one stage after 24 h of washing (49.4 ± 3.1%).
In what concerns the washing with SDS aqueous solutions, a TPH removal efficiency of 81.6 ± 3.3% was obtained after two washing stages of 2 h each.This value is close to that obtained when using only one washing stage for 24 h (82.5 ± 5.3%), as shown in Figure 7.When three washing stages are used, a TPH removal efficiency of 90.8 ± 4.3% was obtained, a value that is close to that obtained for 72 h of one single stage, 92.2 ± 1.6%.
These results indicate that the introduction of washing stages, with a shorter duration time, can be a viable answer to reduce the total washing time, while achieving the same TPH extraction efficiency.However, it will still be necessary to carry out an economic evaluation of the entire process.

Reuse of washing solutions
To evaluate if a more cost-effective and sustainable strategy could be used, the reuse of the washing solutions was studied.Thus, to check the effect of reusing the washing solution on the performance of the decontamination process, experiments were carried out where after a defined contact time (24 h) [50], the surfactant solutions were decanted to separate the extracted oil and solid particles, centrifuged to break the existing surfactant micelles and release the oil and the oily phase was separated from water phase through decantation.The extracted oil can be used for energy production, such as in the industrial burners of cement or lime kilns [51].The recycled aqueous solution can be put in contact with a new contaminated soil.In this way, the recycling of the  aqueous solution was studied under the same conditions used throughout this study (0.71 CMC for SDS and 100 CMC for Tween 80, at 150 rpm and L/S of 10) and the results are presented in Figure 8.
Unfortunately, the reuse of the solution of Tween 80 leads to a very weak TPH removal efficiency of only 6.1 ± 1.9%, close to that achieved with the blank.These results indicate that probably most of the surfactant was lost during the first wash, possibly due to its adsorption on the soil particles and/or in the decantation and centrifugation stages during oil separation.Thus, an amount equal to the initial of surfactant was added to the washing solution.For Tween 80, despite the increase in the TPH removal efficiency to 41.4 ± 3.8%, the value is still below that obtained using a fresh solution (49.4 ± 3.1%).This reduction in the efficiency of the reused solution with total replacement of surfactant in relation to the fresh one can be probably explained by the inhibition caused by the soluble organic matter already present in the solution.
In what concerns the use of SDS washing solutions, Figure 8 shows that there is still a potential to reuse the washing solution.In fact, even with the soluble organic matter present in solution, the reuse of the washing solution allows an efficiency of 38.8 ± 2.7% in 24 h.This circular approach allows to reuse about 10,000 L of water and 11.75 kg of SDS for each ton of contaminated soil.When fresh surfactant was added, a TPH removal efficiency of 95.2 ± 2.2% is obtained, which is higher than the one obtained with the fresh solution.

Disposal of the washing solutions
After the soil washing cycles, the washing solutions need to be discarded and they are typically discharged in the municipal sewage system.Thus, to fulfil the legal requirements, the washing solutions must have a low COD value and so a low concentration of hydrocarbons, organic matter and surfactants.Discharge limits for process effluents vary from site to site but are generally in the range from 500 to 1500 mg/L O 2 , which is significantly lower than the COD of the Tween 80 and SDS washing solutions.Therefore, the solutions should be treated somehow to reduce the COD values down to the discharge limits.In this way, activated carbon was added to the washing fluid using a L/S ratio of 10 and stirred at 200 rpm for 24 h, and subsequently filtered.The COD results obtained before and after the activated carbon treatment of the washing solution are shown in Table 3.As seen, a reduction in COD values from 2800 mg/LO 2 to 22 mg/LO 2 was obtained for Tween 80, while for the SDS washing solution a decrease from 3400 mg/LO 2 to 22 mg/LO 2 was obtained, allowing the discharge of both solutions.In addition to advantages of the safe disposal of the effluent, this water resource can re-enter the process, requiring only the rectification of the surfactant concentration.

Conclusions
The results obtained in this work allow to conclude that soil washing using aqueous solutions of surfactants is a good strategy for remediating sandy soils contaminated  with TPH [52].Two very different surfactants were chosen SDS and Tween 80, the first widely used although with very high CMC, the second less toxic and with a very low CMC.The same molar concentration was used for both surfactants' solutions but since SDS has a much higher CMC than Tween 80, SDS aqueous solutions were evaluated at concentrations below CMC, while Tween 80 at concentration much larger than the CMC.This leads to different removal mechanisms which may also be responsible for the differences observed in the removal efficiencies.Although both surfactants followed a pseudosecond-order TPH removal kinetics, SDS surfactant solutions show a faster kinetics than Tween 80, for the same surfactant concentration, reaching equilibrium after 2 h for SDS, with TPH removal percentage of 88.6 ± 4.1%, while for the same period, a TPH removal efficiency of 25.9 ± 1.7% was obtained with Tween 80, which only reaches the equilibrium after 16 h with an efficiency of 71.5 ± 3.3%.An optimization of the experimental parameters relevant in the removal of TPH from soil, such as concentration of surfactant, L/ S ratios and stirring speed, was carried out.Through the prediction of results provided by the STATISTICA software, it was concluded that the stirring speed parameter is the most relevant in obtaining better TPH removal efficiencies in Tween 80 washes, and the time parameter for SDS washes, due to the structural differences of both surfactants and the removal mechanisms involved.In addition, it was verified that the mathematical model provided by the software allows predicting the experimental values and thus analysing the optimal conditions of the process, a crucial factor for reducing the costs of surfactants, water resources and time.
Alternative strategies, such as the reuse of washing solutions and the washing by stages, were also studied.The reuse of the washing solutions might be a viable strategy for SDS aqueous solutions, but not for Tween 80, which was lost in the washing process, probably by adsorption on the soil.The soil washing in stages is an excellent alternative, saving time while keeping high removal of TPH.Nevertheless, it requires the use of higher amounts of surfactants than a single stage process.As for the disposal of the washing solutions, activated carbon was found to be effective in removal of surfactants, thus allowing the washing solutions discharge into the sewage system.
The results obtained in this work show that the soil washing using aqueous surfactants solutions is a viable technology for the treatment of sandy soils contaminated with low volatile hydrocarbons, giving the conditions necessary to proceed with a less invasive process such as phytoremediation for finishing soil treatment.

Figure 4 .
Figure 4. Predicted versus experimental TPH removal efficiencies for the refined model.(a) Tween 80 washes and (b) SDS washes.

Figure 5 .
Figure 5. Contour plots/surface graphs for TPH removal efficiencies for Tween 80: (a) relating the stirring speed to the washing time, keeping the L/S at 10; (b) relating L/S to the washing time, keeping the stirring speed at 150 rpm and (c) relating L/S with stirring speed maintaining a wash time of 24 h.

Figure 6 .
Figure 6.Contour plots/surface graphs for TPH removal efficiencies for SDS: (a) relating the stirring speed to the washing time, keeping the L/S at 10, (b) relating L/S to the washing time, keeping the stirring speed at 150 rpm and (c) relating L/S with stirring speed maintaining a wash time of 24 h.

Figure 8 .
Figure 8. Efficiencies of TPH removal by reusing the wash solution of the Tween 80 ( ) and of the SDS ( ) with and without full surfactant replacement and a fresh surfactant wash solution after 24 h of wash.

Table 2 .
Characterization of the soil sample used in this work.

Table 3 .
COD values of Tween 80 and SDS washing solutions before and after treatment with activated carbon.