Abstract
Phase behavior of carbon dioxide/water binary mixtures plays an important role in various CO2-based industry processes. This work aims to screen a thermodynamic model out of a number of promising candidate models to capture the vapor–liquid equilibria, liquid–liquid equilibria, and phase densities of CO2/H2O mixtures. A comprehensive analysis reveals that Peng–Robinson equation of state (PR EOS) (Peng and Robinson 1976), Twu α function (Twu et al. 1991), Huron–Vidal mixing rule (Huron and Vidal 1979), and Abudour et al. (2013) volume translation model (Abudour et al. 2013) is the best model among the ones examined; it yields average absolute percentage errors of 5.49% and 2.90% in reproducing the experimental phase composition data and density data collected in the literature. After achieving the reliable modeling of phase compositions and densities, a new IFT correlation based on the aforementioned PR EOS model is proposed through a nonlinear regression of the measured IFT data collected from the literature over 278.15–477.59 K and 1.00–1200.96 bar. Although the newly proposed IFT correlation only slightly improves the prediction accuracy yielded by the refitted Chen and Yang (2019)’s correlation (Chen and Yang 2019), the proposed correlation avoids the inconsistent predictions present in Chen and Yang (2019)’s correlation and yields smooth IFT predictions.
Similar content being viewed by others
1 Introduction
The interaction of CO2 with H2O is frequently seen in several subterranean processes (such as CO2-based enhanced recovery and CO2 storage). Phase behavior of CO2/H2O mixtures under subterranean conditions plays a great role in affecting the overall efficiency of these processes. Thus how to accurately model the phase behavior of CO2/H2O mixtures becomes drastically important. Overall, an appropriate combination of cubic equation of state (CEOS), mixing rule in CEOS, α function, volume translation, and interfacial tension (IFT) model should be determined to well capture the vapor–liquid equilibrium (VLE), liquid–liquid equilibrium (LLE), phase density, and IFT of CO2/H2O mixtures.
Due to their simplicity and good reliability, CEOSs such as SRK EOS (Soave 1972) and PR EOS (Peng and Robinson 1976) are the most widely used thermodynamic models for the phase behavior modeling of CO2/H2O binary mixtures (Aasen et al. 2017; Michelsen and Mollerup 2007). Numerous articles have addressed phase-composition modeling of CO2/H2O mixtures. Two types of methods, ϕ–ϕ (fugacity–fugacity) approach and γ–ϕ (activity–fugacity) approach (Trusler 2017; Zhao and Lvov 2016), are often applied in such modeling processes. Because γ-ϕ approach has a discontinuity issue in the phase diagram near the critical region (Zhao and Lvov 2016), this work focuses on ϕ–ϕ based methods. Valtz et al. (2004) found that the most accurate model is PR EOS (Peng and Robinson 1976), Mathias–Copeman α function (Mathias and Copeman 1983), and Wong–Sandler mixing rule (Wong and Sandler 1992) with average absolute percentage deviation (AAD) of 5.4% in reproducing the measured phase composition data for CO2/H2O mixtures. However, the temperature and pressure ranges used by Valtz et al. (2004) were narrow (278.2–318.2 K and 4.64–79.63 bar, respectively). In addition, the parameters in the Wong–Sandler mixing rule (Wong and Sandler 1992) are given as discrete values at different isotherms. Zhao and Lvov (2016) applied PRSV EOS (Stryjek and Vera 1986) and the Wong–Sandler mixing rule to calculate phase compositions, obtaining an AAD of 7.12% in reproducing the measured phase composition data of CO2/H2O mixtures over a wide range of temperatures and pressures. Similar to Valtz et al. (2004)’s study, in the study by Zhao and Lvov (2016), the parameters in the Wong–Sandler mixing rule are provided as discrete values at different isotherms, instead of generalized correlations; their model is inconvenient to use since one has to make extrapolations based on the provided values when making predictions at conditions different from those given by Zhao and Lvov (2016). Abudour et al. (2012a) applied van der Waals (1873) one-fluid (vdW) mixing rule with several temperature-dependent BIP correlations in PR EOS in determining phase compositions of CO2/H2O mixtures. With the tuned BIPs, their model yielded good accuracy (i.e., AAD of 5.0%) in aqueous phase-composition predictions but lower accuracy (i.e., AAD of 13.0%) in CO2-rich phase-composition predictions, respectively.
A recent comprehensive study by Aasen et al. (2017) revealed that the most accurate thermodynamic model (among the ones examined by them) in phase-composition and phase-density predictions for CO2/H2O mixtures is PR EOS, Twu α function (Twu et al. 1991), Huron-Vidal mixing rule, and constant volume translation. This model only yields an AAD of 4.5% in phase-composition calculations and an AAD of 2.8% in phase-density calculations for CO2/H2O mixtures. Aasen et al. (2017), Valtz et al. (2004), and Zhao and Lvov (2016) also pointed out that more advanced models [e.g., the Cubic-Plus-Association (CPA) EOS (Kontogeorgis et al. 1996)] do not guarantee an improvement in the phase-composition predictions for CO2/H2O mixtures.
With regards to phase-density calculations, CEOS based methods tend to overestimate liquid-phase molar volumes. A detailed discussion of this issue can be found in the studies by Matheis et al. (2016) and Young et al. (2017). In order to address this problem, Martin (1979) introduced the volume translation concept in CEOS to improve liquid-phase volumetric predictions. Peneloux et al. (1982) developed volume translation schemes in SRK EOS for pure substances. Jhaveri and Youngren (1988) applied volume translation into PR EOS, leading to the improvement of liquid phase-density predictions. A thorough comparison of different types of volume translation methods can be found in Young et al. (2017)’s work. According to the study by Young et al. (2017), the temperature-dependent volume translation method developed by Abudour et al. (2012b, 2013) provides the most accurate estimates on liquid-phase densities without thermodynamic inconsistencies (e.g., crossover of pressure–volume isotherms). Aasen et al. (2017) applied constant volume translation to phase-density calculations of CO2/H2O mixtures and achieved a significant improvement in density-prediction accuracies. However, a more accurate volume translation function, the one proposed by Abudour et al. (2012b, 2013), was not applied in Aasen et al. (2017)’s study; furthermore, it should be noted that Aasen et al. (2017) used GERG-2008 (Kunz and Wagner 2012) and EOS-CG (Gernert and Span 2016) calculated densities as reference densities instead of experimental data. In this study, we apply the volume translation method by Abudour et al. (2012b, 2013) to see if the use of this model can further improve phase-density predictions for CO2/H2O mixtures; these predictions are compared to the measured density data documented in the literature.
Parachor model (Sugden 1930) is one of the most widely applied models in predicting mixtures’ IFT (Schechter and Guo 1998). However, its accuracy heavily relies on the density difference between the two coexisting phases in a VLE or an LLE. Our experience in using Parachor model to calculate IFT of CO2/H2O mixtures shows that Parachor model is generally appropriate for the IFT estimation for VLE of CO2/H2O systems, but less suitable for the IFT estimation for LLE of CO2/H2O systems. This is primarily because an LLE of a CO2/H2O mixture has a smaller density difference than a VLE. Several empirical IFT correlations for CO2/H2O mixtures have been proposed in the literature. However, most of these correlations are only applicable to a limited temperature and pressure range (Zhang et al. 2016). Hebach et al. (2002) proposed a new correlation which correlated IFT with phase densities. Hebach et al. (2002)’s model is suitable over a wide range of temperature and pressure conditions, although the prediction accuracy decreases with an increase in temperature or pressure. Chen and Yang (2019) proposed a new empirical IFT correlation for CO2/CH4/H2O ternary systems based on mutual solubility, and this model performs well for CO2/H2O binary mixtures. However, our experience in applying Chen and Yang (2019)’s model shows that some breaking points can be observed in the predicted IFT curves under some conditions, hampering its ability in providing consistent and smooth IFT predictions. In addition, using two sets of BIPs (as applied in Chen and Yang (2019)’s study) in the aqueous phase and non-aqueous phase can lead to thermodynamic inconsistency issue near the critical region as demonstrated by Li and Li (2019).
The discussion above reveals that the previous studies of phase behavior modeling of the CO2/H2O mixtures tend to primarily focus on phase-composition modeling and pay less attention to phase-density calculations (especially for the CO2-rich phase). Whereas, phase-density is one important property in VLE and LLE since IFT calculations and flow simulations can heavily rely on such property. As for the IFT modeling, we are currently lacking a reliable IFT correlation that not only pays due tribute to the phase composition and density of CO2/H2O mixtures but also gives smooth and consistent IFT predictions over a wider range of temperature/pressure conditions.
In this study, we first conduct a thorough literature review to select the most promising thermodynamic models that can well capture the VLE and LLE of CO2/H2O mixtures. Then, we conduct phase-composition calculations by using PR EOS, Twu α function, and Huron-Vidal mixing rule [as suggested by Aasen et al. (2017)], and validate the accuracy of this model by comparing the calculated VLE and LLE phase compositions to the measured ones. Then, we introduce Abudour et al. (2012b, 2013) volume translation model in phase-density calculations to check if applying this model can further improve the density-prediction accuracies. A new empirical IFT correlation for CO2/H2O mixtures is then proposed based on the reliable thermodynamic model that incorporates the Huron-Vidal mixing rule and the Abudour et al. (2013) volume translation model.
2 Methodology
2.1 PR EOS and α functions
In this study, PR EOS (Peng and Robinson 1976) is implemented because of its simplicity and more accurate liquid-density predictions compared with SRK EOS (Aasen et al. 2017). The expression of PR EOS is detailed in “Appendix A”.
With regards to α functions in PR EOS, Twu α function and Gasem α function are used in this study. Compared with the other types of α functions, the Twu α function can describe the thermodynamic properties of polar compounds more accurately and perform better in the supercritical region (Young et al. 2016). In addition, according to the study by Aasen et al. (2017), Twu α function coupled with PR EOS and Huron-Vidal mixing rule yields the most accurate phase-composition estimations on CO2/H2O mixtures among the models evaluated by them. Therefore, we select Twu α function as one of the evaluated α functions in this study.
The Gasem α function provides more accurate representation of supercritical phase behavior (Gasem et al. 2001). Besides, based on the study by Abudour et al. (2013), Gasem α function coupled with PR EOS, vdW mixing rule and Abudour volume translation yields the most accurate liquid-phase-density predictions for the chemical compounds examined by their study. Therefore, we select the Twu α function and the Gasem α function in VLE/LLE and phase-density calculations. “Appendix A” shows the expressions of Twu α function and Gasem α function.
2.2 Mixing rules
Mixing rules have a great impact on phase equilibrium calculations. Huron and Vidal (1979) proposed a new expression by considering the excess Gibbs energy for CEOS, which made more accurate the phase-composition predictions for mixtures containing polar substances. Furthermore, according to the comprehensive study by Aasen et al. (2017), the most accurate thermodynamic model among the ones examined by them is PR EOS coupled with Twu α function and Huron-Vidal mixing rule, which provides an AAD of 4.5% in reproducing the phase-composition data measured for CO2/H2O mixtures. Hence, in the first part of this study, we collect more phase equilibria data for CO2/H2O mixtures to verify the performance of the model suggested by Aasen et al. (2017). These additional experimental data are not included in the study by Aasen et al. (2017).
The vdW mixing rule is one of the most commonly used mixing rules in petroleum industry (Pedersen et al. 2014). Although vdW mixing rule is originally developed for nonpolar systems, the vdW mixing rule coupled with the tuned BIPs can be reliably used for describing the phase behavior of mixtures containing polar components (e.g., water). Besides, based on the study by Abudour et al. (2012a), Gasem α function with vdW mixing rule and their temperature-dependent volume translation function provided a promising means to well reproduce the measured liquid-phase densities for CO2/H2O mixtures. Therefore, in this study, we also employ the model suggested by Abudour et al. (2012a) to test if it outperforms the model suggested by Aasen et al. (2017). The expressions of vdW mixing rule and Huron–Vidal mixing rule and their BIPs are detailed in “Appendices B and C”.
2.3 Volume translation models
Volume translation is used to overcome the inherent deficiency of CEOS in liquid-phase-density predictions. In order to improve liquid-phase-density calculations, Peneloux et al. (1982) developed a constant volume translation model in SRK EOS, while Jhaveri and Youngren (1988) developed a constant volume translation model in PR EOS. Abudour et al. (2012b, 2013) revised the temperature-dependent volume translation function to improve both saturated and single-phase liquid density calculations. Furthermore, unlike other temperature-dependent volume translation models, the volume translation model developed by Abudour et al. (2012b, 2013) does not yield thermodynamic inconsistency issues unless at extremely high pressures. Therefore, we select the constant and Abudour et al. volume translation models in this study for phase-density predictions. The expressions of these two volume translation models are detailed in “Appendix D”.
2.4 IFT correlations for CO2/H2O mixtures
In this study, we select Parachor model (Sugden 1930), Hebach et al. (2002)’s correlation, and Chen and Yang (2019)’s correlation to predict IFT of CO2/H2O mixtures. The Parachor model (Sugden 1930) is one of the most widely used methods in determining mixtures’ IFT. It correlates IFT with phase compositions and molar densities of each phase. Parachor is a component-dependent constant. The expression of the Parachor model is shown in “Appendix E”. Hebach et al. (2002)’s correlation correlates IFT with temperature, pressure, and phase densities. Phase compositions are not included in their correlation. To make fair comparison, we also refit coefficients in their correlation based on the IFT database employed in this study. Values of the original and refitted coefficients as well as the Hebach et al. (2002)’s correlation are shown in “Appendix E”.
Chen and Yang (2019)’s correlation correlates IFT with phase equilibrium ratios (K-values) and the reduced pressure of CO2. Unlike the Parachor model and the Hebach et al. (2002)’s correlation, the density of the two equilibrating phases is not one input in the Chen and Yang (2019)’s correlation. Chen and Yang (2019)’s correlation, they proposed four groups of coefficient sets, i.e., one coefficient set (using one coefficient set on the whole pressure range) with or without the reduced pressure term, and two coefficient sets (dedicated to the pressure ranges of p ≤ 73.8 and p > 73.8 bar) with or without the reduced pressure term. Since using the reduced pressure term can improve prediction accuracy (Chen and Yang 2019), we introduce the reduced pressure term in this study. Similarly, we refit these coefficients based on the IFT databased employed in this study to make fair comparison. Values of the original and refitted coefficients as well as the Chen and Yang (2019)’s correlation are detailed in “Appendix E”.
2.5 IFT correlation proposed in this study
Before we finalize our IFT correlation, we tried several scenarios to find the optimal one to correlate the IFT of CO2/H2O mixtures. Since the Parachor model is one of the most widely used models in mixtures’ IFT predictions, we revise the original Parachor model by introducing a component-dependent correction term αi; furthermore, we replace the constant exponential term in the original Parachor model by correlating it with several physical properties (e.g., equilibrium ratios). The new IFT correlation can be expressed as follows:
where σ in the interfacial tension; αi is the introduced component-dependent correction term; xi and yi are the mole fractions of component i in liquid and vapor phases, respectively; Pi is the Parachor value of component i (\(P_{{{\text{H}}_{2} {\text{O}}}} = 52\), \(P_{{{\text{CO}}_{2} }} = 78\)) (Liu et al. 2016); \(\rho_{{\text{L}}}^{{\text{M}}}\) is the molar density of liquid phase in mol/cm3; \(\rho_{{\text{V}}}^{{\text{M}}}\) is the molar density of vapor phase in mol/cm3. N is the number of component; n is the exponent.
First, the component-dependent correction term αi is set as a constant for each component, and the exponential term n can be expressed by the equilibrium ratios of CO2-rich phase and aqueous phase:
where C1, C2, and C3 are empirical coefficients; \(K_{{{\text{CO}}_{2} }}\) and \(K_{{{\text{H}}_{2} {\text{O}}}}\) are the equilibrium ratios (as known as K-values) of CO2 and H2O:
Since using one coefficient set for both αi and n on the whole CO2-rich-phse density range cannot converge after reaching the maximum iterations, we use two coefficient sets based on CO2-rich-phase densities. Table 1 listed the values of these coefficients and αi determined by fitting the proposed correlation (abbreviated as Scenario #1) to the IFT training dataset.
Since using constants to represent αi leads to a larger AAD compared with the refitted Chen and Yang (2019)’s correlation (i.e., 8.8746% vs. 7.8520%, respectively), we correlate equilibrium ratios to αi to see if it can improve the IFT predictions. The expression of n in this scenario (abbreviated as Scenario #2) is the same as that in Scenario #1. The expression for αi is given as:
Specifically, when the CO2-rich-phase density is greater than 0.2 g/cm3, \(\alpha_{{{\text{H}}_{2} {\text{O}}}}\) can be simplified as:
Table 2 listed the values of these coefficients determined by fitting the proposed correlation to the IFT training dataset. Since using one coefficient set for αi and n on the whole CO2-rich-phse density range in Scenario #2 cannot converge after reaching the maximum iterations, we use two coefficients based on CO2-rich-phase densities (the same as Scenario #1).
We find that using correlations to represent αi can slightly improve the IFT predictions (i.e., AAD of 8.31% in Scenario #2 vs. AAD of 8.87% in Scenario #1). Besides, we find that the value of n is around 4 over a wide range of temperature/pressure conditions in all scenarios (i.e., its value only slightly changes with the change of equilibrium ratios); therefore, we set the value of n as 4 for simplicity.
However, our experience in applying Scenarios #1 and #2 shows that some breaking points can be observed in the correlated IFT curves due to the fact that two different sets of coefficients are adopted under the conditions of \(\rho_{{{\text{CO}}_{2} {\text{-rich}}}} < 0.2\) g/cm3 and \(\rho_{{{\text{H}}_{{2}} {\text{O }\text{-rich}}}} \ge 0.2\) g/cm3, respectively. In addition, based on the study by Chen and Yang (2019), introducing the reduced pressure of CO2 can improve IFT predictions. Thus, we introduce the reduced pressure of CO2 in the expressions of αi and use one coefficient set to see if these settings can further improve the prediction accuracies without yielding inconsistent IFT predictions. Based on the calculation results, the following IFT correlation yields the lowest AAD among the ones examined in this study:
where the αi term in the new correlation can be expressed as:
where pr is the reduced pressure of CO2.
Table 3 lists the values of these coefficients determined by fitting the proposed correlation to the IFT training dataset.
3 Results and discussion
The values of critical pressure (pc), critical temperature (Tc), acentric factor (ω), molecular weight (M), critical compressibility factor (Zc) used in this study are retrieved from the NIST database (Lemmon et al. 2011).
3.1 Performance comparison of thermodynamic models in phase equilibrium calculations
Table 4 summarizes the measured phase equilibrium data of CO2/H2O mixtures over 278–378.15 K and 0.92–709.3 bar reported in the literature. Note that these experimental data were not included in the study by Aasen et al. (2017). Comparison between the measured and calculated phase-composition results is evaluated by the average absolute percentage deviation (AAD) defined as:
where AAD is the average absolute percentage deviation; N is the number of data points; xCAL and xEXP are the calculated and measured mole fraction of CO2 or H2O in the aqueous phase (or the CO2-rich phase), respectively.
Table 5 details the settings of the four thermodynamic models examined in this work. Table 6 summarizes the performance of different thermodynamic models in phase-composition predictions.
As shown in Table 6, although AAD for \(x_{{{\text{CO}}_{2} }}\) of Case 3 (Twu + HV) is slightly higher than that of Case 4 (Gasem + HV), i.e., 4.73% of Case 3 vs. 3.64% of Case 4, Case 3 (Twu + HV) significantly outperforms the other models in \(y_{{{\text{H}}_{{2}} {\text{O}}}}\) predictions, i.e., AAD of 10.37% of Case 3 vs. > 16% of other cases. Thus, given the overall performance, Case 3 (Twu + HV) is found to be the best model in phase-composition predictions. Figure 1 compares the performance of different models at T = 323.15 K and T = 348.15 K. As can be seen from these two figures, the thermodynamic model Case 3 (Twu + HV) can well capture the trend exhibited by the measured solubility data over a wide pressure range.
3.2 Evaluation of thermodynamic models in density calculations
Table 7 summarizes the experimental aqueous-phase and CO2-rich-phase densities of CO2/H2O mixtures over 278–478.35 K and 2.5–1291.1 bar documented in the literature. The pressure–temperature coverage of the phase density data collected from the literature are shown in “Appendix F”.
Since Case 3 (Twu + HV) outperforms other thermodynamic models in phase-composition predictions for CO2/H2O mixtures, we only focus on the performance of Case 3 coupled with volume translation in phase-density predictions. Table 8 summarizes the performance of different volume translation models in both aqueous-phase and CO2-rich-phase density calculations.
As shown in Table 8, incorporation of VT into the thermodynamic framework can generally improve the phase-density prediction accuracy. Case 3–1 (Twu + HV + Abudour VT) provides the most accurate estimates of both aqueous-phase and CO2-rich-phase density, yielding AAD of 2.90% in reproducing the measured phase-density data. Figure 2 further visualizes some of the calculation results by these three different models at different pressure/temperature conditions.
It can be seen from Fig. 2 that, regarding aqueous-phase density predictions, the performance of Case 3–2 (Twu + HV + Constant VT) improves dramatically as temperature rises. As shown in Fig. 2e, f, at high temperature conditions, Cases 3–2 yields the most accurate aqueous-phase density predictions; however, it fails to accurately predict CO2-rich-phase densities. As a lighter phase, CO2-rich-phase density can be accurately predicted without the use of volume translation functions. Applying Abudour VT method is able to only slightly improve the prediction accuracy (i.e., AAD of 2.62%). In contrast, applying constant VT in CO2-rich-phase density predictions can lead to larger errors than the case without the use of VT.
Figure 3 compares the performance of different models in terms of their accuracy in phase-density predictions over 382.14–478.35 K and 35.3–1291.9 bar. Note that the results of CPA EOS model from the work by Tabasinejad et al. (2010) focuses on the same pressure and temperature ranges. As can be seen from Fig. 3, although the CPA EOS model can accurately predict the aqueous-phase density, it tends to be less accurate in determining the CO2-rich-phase density. Overall, the thermodynamic model Cases 3–1 (Twu + HV + Abudour VT) give an accuracy comparable to the more complex CPA EOS model.
In addition, according to the study by Aasen et al. (2017), CPA EOS model yields higher percentage errors (AAD) in reproducing phase-composition data for CO2/H2O mixtures compared with Case 3 (PR EOS + Twu + HV), i.e., 9.5% vs. 4.5% (Aasen et al. 2017). Therefore, overall, Case 3–1 (Twu + HV + Abudour VT) is a more accurate model in both phase-composition and phase-density predictions for CO2/H2O mixtures.
3.3 Evaluation of the newly proposed IFT correlation
Table 9 summarizes the experimental IFT data of CO2/H2O mixtures over 278.15–477.59 K and 1–1200.96 bar documented in the literature. Ideally, phase densities should be directly measured; however, only Chiquet et al. (2007), Kvamme et al. (2007), Bikkina et al. (2011), Bachu and Bennion (2009), and Shariat et al. (2012) applied measured phase densities in IFT calculations. In order to expand our IFT database, IFT data with precisely determined phase densities are also included in our IFT database. The collected IFT data are randomly placed into two bins: a training dataset (including 589 data points) and a validation dataset (including 189 data points).
Results in Sects. 3.1 and 3.2 reveal that the thermodynamic model using PR EOS, Twu α function, Huron-Vidal mixing rule, and Abudour et al. (2013) VT yields the most accurate estimates on both phase compositions and densities. Therefore, the aforementioned thermodynamic model provides reliable phase-composition and phase-density predictions that can be fed into the proposed IFT correlation.
Mean absolute errors (MAE), AAD, and coefficient of determination (R2) are used as performance measures. The expressions of MAE and R2 are as follows:
where σEXP is the measured IFT data in mN/m; σCAL is the calculated IFT in mN/m by different correlations; \(\overline{\sigma }_{{{\text{EXP}}}}\) is the average of the measured IFTs in mN/m.
3.3.1 Performance of different IFT correlations
Table 10 shows the details of the different IFT models examined in this study. Table 11 summarize the performance of different correlations in IFT estimations. As can be seen, the most accurate IFT model is Model 3 proposed in this study, although it only shows a marginal edge over Model 2.
Figure 4 visually compares the measured IFTs vs. pressure and the calculated ones by different IFT models at selected temperatures. As shown in these plots, in general, Model 3 (this study) outperforms other empirical correlations over a wide range of temperatures and pressures. It can be also observed from these plots that breaking points appear in the predicted IFT curves at p = 73.8 bar by Model 2 (Refitted Chen and Yang (2019)’s correlation with two coefficient sets). Such discontinuous IFT prediction can be attributed to the fact that two different sets of coefficients are adopted under the conditions of p ≤ 73.8 and p > 73.8 bar, respectively, in Chen and Yang (2019)’s correlation. Although using one coefficient in Chen and Yang (2019)’s correlation (e.g., Model 5) can avoid such discontinuous IFT predictions, it yields larger percentage errors. Therefore, Model 3 (this study) is the best model in IFT predictions for CO2/H2O mixtures over a wide range of temperatures and pressures.
Figure 5 illustrates how the IFTs predicted by Model 3 (this study) vary with pressure at different temperatures. It can be observed from Fig. 5 that the new IFT correlation provides smooth and consistent IFT predictions at different pressures and temperatures. Overall, Model 3 proposed in this study yields accurate and consistent IFT predictions over the wide range of temperatures and pressures, although it yields relatively higher percentage errors at higher temperature conditions (e.g., T = 478 K) compared with that at lower temperature conditions (i.e., T < 378 K). It is interesting to observe from Fig. 5a that when the pressure is less than around 15 bar and the temperature is between 278.15 and 368.15 K, an increase in temperature leads to a decrease in the predicted IFT under the same pressure. In comparison, when the pressure is larger than around 15 bar, an increase in temperature leads to an increase in the predicted IFT. At higher temperatures of 378.15–478.15 K, an increase in temperature always results in a decline in the predicted IFT under the same pressure, as seen in Fig. 5b. Most of the measured IFTs documented in the literature follow this trend (Akutsu et al. 2007; Chalbaud et al. 2009; Chiquet et al. 2007; Chun and Wilkinson 1995; Da Rocha et al. 1999; Georgiadis et al. 2010; Hebach et al. 2002; Heuer 1957; Hough et al. 1959; Kvamme et al. 2007; Khosharay and Varaminian 2014; Liu et al. 2016; Park et al. 2005; Pereira et al. 2016; Shariat et al. 2012), except for the studies by Bachu and Bennion (2009) and Bikkina et al. (2011), i.e., an increase in temperature leads to an increase in IFT at a temperature range of 373.15–398.15 K in the study by Bachu and Bennion (2009), and an increase in temperature leads to an increase in IFT over 298.15–333.15 K in the study by Bikkina et al. (2011). Again, the sharp drops in the IFT curves at lower temperatures (where CO2 remains subcritical) are due to the transformation of VLE to LLE.
3.3.2 Statistical significance tests of IFT correlations
As shown in Table 11, the AADs yielded by Model 2 (refitted Chen and Yang (2019)’s correlation) and Model 3 (this study) are on the same scale. Therefore, it is necessary to conduct statistical significance tests to check if the marginal edge of Model 3 over Model 2 is statistically significant. Figure 1 shows the frequency distribution of the differences between the measured IFT data (i.e., whole dataset including 778 data points) and calculated ones by Model 2, while Fig. 7 shows the same information for Model 3. As can be seen from Figs. 6 and 7, the distribution of the deviations generated by the two models can be considered to follow Gaussian distributions. As such, paired one-tailed t-tests are applied as the statistical significance test method (Japkowicz and Shah 2011).
P-value is used to check if one model is better than another one. Typically, the significance threshold α is 0.05; when \(P > \alpha\), two models have the same performance. In contrast, when \(P \le \alpha\), it is reasonable to say that one model is significantly better than another one (Japkowicz and Shah 2011).
P-value of Model 2 (refitted original Chen and Yang (2019)’s correlation with two coefficient sets) and Model 3 (this study) is 0.0069, \(P < \alpha\) (\(\alpha = 0.05\)); therefore, it is reasonable to say that Model 3 statistically outperforms Model 2. In addition, the new model does not give discontinuous IFT predictions, while Chen and Yang (2019)’s IFT model bears such issue.
4 Conclusions
The objective of this study is to screen and develop reliable models for describing the VLE, LLE, phase density, and IFT of CO2/H2O mixtures. Based on the comparison between the experimental data and the calculated ones from different models, we can reach the following conclusions:
-
1.
The most accurate method to represent CO2/H2O VLE and LLE is PR EOS, Twu α function, and Huron-Vidal mixing rule, which only yields AAD of 5.49% and 2.90% in reproducing measured CO2/H2O phase-composition data and phase-density data over a temperature range of 278–378.15 and 278–478.35 K and over a pressure range of 6.9–709.3 and 2.5–1291.1 bar, respectively.
-
2.
Applying either constant or Abudour et al. (2013) VT method can significantly improve aqueous-phase density calculations. In addition, when the temperature is higher than 373 K, constant VT method can yield lower error in reproducing measured phase-density data than Abudour et al. (2013) VT method;
-
3.
Constant VT method cannot improve the prediction accuracy of CO2-rich-phase density. Abudour et al. (2013) VT method can slightly improve CO2-rich-phase density predictions, but such improvement is more obvious at low to moderate temperature conditions.
-
4.
The new IFT correlation based on the aforementioned PR EOS model outperforms other empirical correlations with an overall AAD of 7.77% in reproducing measured IFT data of CO2/H2O mixtures. The new IFT correlation is only slightly more accurate than the refitted Chen and Yang (2019)’s correlation with two coefficient sets. But the new correlation yields smooth IFT predictions, avoiding the issue of discontinuous IFT predictions yielded by Chen and Yang (2019)’s correlation.
References
Aalto M, Keskinen KI, Aittamaa J, Liukkonen S. An improved correlation for compressed liquid densities of hydrocarbons. Part 2. Mixtures. Fluid Phase Equilib. 1996;114:21–35. https://doi.org/10.1016/0378-3812(95)02824-2.
Aasen A, Hammer M, Skaugen G, Jakobsen JP, Wilhelmsen Ø. Thermodynamic models to accurately describe the PVTxy-behavior of water/carbon dioxide mixtures. Fluid Phase Equilib. 2017;442:125–39. https://doi.org/10.1016/j.fluid.2017.02.006.
Abudour AM, Mohammad SA, Gasem KAM. Modeling high-pressure phase equilibria of coalbed gases/water mixtures with the Peng–Robinson equation of state. Fluid Phase Equilib. 2012;319:77–89. https://doi.org/10.1016/j.fluid.2012.01.030.
Abudour AM, Mohammad SA, Robinson R, Gasem KAM. Volume-translated Peng–Robinson equation of state for saturated and single-phase liquid densities. Fluid Phase Equilib. 2012;335:74–87. https://doi.org/10.1016/j.ijrefrig.2019.04.014.
Abudour AM, Mohammad SA, Robinson RL, Gasem KAM. Volume-translated Peng–Robinson equation of state for liquid densities of diverse binary mixtures. Fluid Phase Equilib. 2013;349:37–55. https://doi.org/10.1016/j.fluid.2013.04.002.
Akutsu T, Yamaji Y, Yamaguchi H, Watanabe M, Smith R, Inomata H. Interfacial tension between water and high pressure CO2 in the presence of hydrocarbon surfactants. Fluid Phase Equilib. 2007;257:163–8. https://doi.org/10.1016/j.fluid.2007.01.040.
Bachu S, Bennion DB. Interfacial tension between CO2, freshwater, and brine in the range of pressure from (2 to 27) MPa, temperature from (20 to 125) °C, and water salinity from (0 to 334000) mg·L−1. J Chem Eng Data. 2009;54:765–75. https://doi.org/10.1021/je800529x.
Bikkina PK, Shoham O, Uppaluri R. Equilibrated interfacial tension data of the CO2-water system at high pressures and moderate temperatures. J Chem Eng Data. 2011;56:3725–33. https://doi.org/10.1021/je200302h.
Briones JA, Mullins JC, Thies MC. Ternary phase equilibria for acetic acid-water mixtures with supercritical carbon dioxide. Fluid Phase Equilib. 1987;36:235–46. https://doi.org/10.1016/0378-3812(87)85026-4.
Chalbaud C, Robin M, Lombard JM, Martin F, Egermann P, Bertin H. Interfacial tension measurements and wettability evaluation for geological CO2 storage. Adv Water Resour. 2009;32:98–109. https://doi.org/10.1016/j.advwatres.2008.10.012.
Chen Z, Yang D. Correlations/estimation of equilibrium interfacial tension for methane/CO2-water/brine systems based on mutual solubility. Fluid Phase Equilib. 2019;483:197–208. https://doi.org/10.1016/j.fluid.2018.11.037.
Chiquet P, Daridon J, Broseta D, Thibeau S. CO2/water interfacial tensions under pressure and temperature conditions of CO2 geological storage. Energy Conversat Manag. 2007;48:736–44. https://doi.org/10.1016/j.enconman.2006.09.011.
Chun BS, Wilkinson GT. Interfacial tension in high-pressure carbon dioxide mixtures. Ind Eng Chem Res. 1995;34:4371–7. https://doi.org/10.1021/ie00039a029.
Dalmolin I, Skovroinski E, Biasi A, Corazza M, Dariva C, Oliverira JV. Solubility of carbon dioxide in binary and ternary mixtures with ethanol and water. Fluid Phase Equilib. 2006;245:193–200. https://doi.org/10.1016/j.fluid.2006.04.017.
Da Rocha SRP, Harrison KL, Johnston KP. Effect of surfactants on the interfacial tension and emulsion formation between water and carbon dioxide. Langmuir. 1999;15:419–28. https://doi.org/10.1021/la980844k.
Dohrn R, Bünz AP, Devlieghere F, Thelen D. Experimental measurements of phase equilibria for ternary and quaternary systems of glucose, water, CO2 and ethanol with a novel apparatus. Fluid Phase Equilib. 1993;83:149–58. https://doi.org/10.1016/0378-3812(93)87017-U.
D’Souza R, Patrick JR, Teja AS. High pressure phase equilibria in the carbon dioxide—n-hexadecane and carbon dioxide-water systems. Can J Chem Eng. 1988;66:319–23. https://doi.org/10.1002/cjce.5450660221.
Efika EC, Hoballah R, Li X, May EF, Nania M, Vicente YS, Trusler JPM. Saturated phase densities of (CO2 + H2O) at temperatures from (293 to 450) K and pressures up to 64 MPa. J Chem Thermodyn. 2016;93:347–59. https://doi.org/10.1016/j.jct.2015.06.034.
Gasem KAM, Gao W, Pan Z, Robinson RL. A modified temperature dependence for the Peng–Robinson equation of state. Fluid Phase Equilib. 2001;181:113–25. https://doi.org/10.1016/S0378-3812(01)00488-5.
Georgiadis A, Maitland G, Trusler JP, Bismarck A. Interfacial tension measurements of the (H2O + CO2) system at elevated pressures and temperatures. J Chem Eng Data. 2010;55:4168–75. https://doi.org/10.1021/je100198g.
Gernert J, Span R. EOS-CG: a Helmholtz energy mixture model for humid gases and CCS mixtures. J Chem Thermodyn. 2016;93:274–93. https://doi.org/10.1016/j.jct.2015.05.015.
Gillepsie PC, Wilson GM. Vapor-liquid and liquid–liquid equilibria: water-methane, water-carbon dioxide, water-hydrogen sulfide, water-n-pentane, water-methane-n-pentane. Tulsa.: Gas Processors Association; 1982.
Guo H, Chen Y, Hu Q, Lu W, Ou W. Quantitative Raman spectroscopic investigation of geo-fluids high-pressure phase equilibria: part I. Accurate calibration and determination of CO2 solubility in water from 273.15 to 573.15 K and from 10 to 120 MPa. Fluid Phase Equilib. 2014;382:70–9. https://doi.org/10.1016/j.fluid.2014.08.032.
Han JN, Shin HY, Min BM, Han KH, Cho A. Measurement and correlation of high-pressure phase behavior of carbon dioxide + water system. J Ind Eng Chem. 2009;15:212–6. https://doi.org/10.1016/j.jiec.2008.09.012.
Hebach A, Oberhof A, Dahmen N, Kögel A, Ederer H, Dinjus E. Interfacial tension at elevated pressures—measurements and correlations in the water + carbon dioxide system. J Chem Eng Data. 2002;47:1540–6. https://doi.org/10.1021/je025569p.
Hebach A, Oberhof A, Dahmen N. Density of water plus carbon dioxide at elevated pressures: measurements and correlation. J Chem Eng Data. 2004;49:950–3. https://doi.org/10.1021/je034260i.
Heuer GJ. Interfacial tension of water against hydrocarbons and other gases and adsorption of methane on solids at reservoir temperatures and pressures. PhD Dissertation, Univ. of Texas at Austin, Austin, Texas; 1957.
Hough EW, Henuer GJ, Walker JW. An improved pendant drop, interfacial tension apparatus and data for carbon dioxide and water. Pet Trans AIME. 1959;216:469–72.
Huron MJ, Vidal J. New mixing rules in simple equations of state for representing vapor-liquid equilibria of strongly nonideal mixtures. Fluid Phase Equilib. 1979;3:255–71. https://doi.org/10.1016/0378-3812(79)80001-1.
Japkowicz N, Shah M. Evaluating learning algorithms—a classification perspective. Cambridge: Cambridge University Press; 2011. https://doi.org/10.1017/CBO9780511921803.
Jhaveri BS, Youngren GK. Three-parameter modification of the Peng–Robinson equation of state to improve volumetric predictions. SPE Res Eng. 1988;8:1033–40. https://doi.org/10.2118/13118-PA.
Khosharay S, Varaminian F. Experimental and modeling investigation on surface tension and surface properties of (CH4 + H2O), (C2H6 + H2O), (CO2 + H2O) and (C3H8 + H2O) from 284.15 K to 312.15 K and pressures up to 60 bar. Int J Refrig. 2014;47:26–35. https://doi.org/10.1016/j.ijrefrig.2014.08.003.
King MB, Mubarak A, Kim JD, Bott TR. The mutual solubilities of water with supercritical and liquid carbon-dioxide. J Supercrit Fluids. 1992;5:296–302. https://doi.org/10.1016/0896-8446(92)90021-B.
Kontogeorgis GM, Voutsas EC, Yakoumis IV, Tassios DP. An equation of state for associating fluids. Ind Eng Chem Res. 1996;35:4310–8. https://doi.org/10.1021/ie9600203.
Kvamme B, Kuznetsova T, Hebach A, Oberhof A, Lunde E. Measurements and modelling of interfacial tension for water + carbon dioxide systems at elevated pressures. Comput Mater Sci. 2007;38:506–13. https://doi.org/10.1016/j.commatsci.2006.01.020.
Kunz O, Wagner W. The GERG-2008 wide-range equation of state for natural gases and other mixtures: an expansion of GERG-2004. J Chem Eng Data. 2012;57:3032–91. https://doi.org/10.1021/je300655b.
Lemmon EW, McLinden MO, Friend DG. Thermophysical properties of fluid systems. In: Linstrom PJ, Mallard WG, editors. NIST chemistry WebBook, NIST standard reference database number 69. Gaithersburg: National Institute of Standards and Technology; 2011, p. 20899 https://doi.org/10.18434/T4D303
Li R, Li H. Improved three-phase equilibrium calculation algorithm for water/hydrocarbon mixtures. Fuel. 2019;244:517–27. https://doi.org/10.1016/j.fuel.2019.02.026.
Li Z, Dong M, Li S, Dai L. Densities and solubilities for binary systems of carbon dioxide plus water and carbon dioxide plus brine at 59 °C and pressures to 29 MPa. J Chem Eng Data. 2004;49:1026–31. https://doi.org/10.1021/je049945c.
Liu Y, Li H, Okuno R. Measurements and modeling of interfacial tension for CO2/CH4/brine systems under reservoir conditions. Ind Eng Chem Res. 2016;55:12358–75. https://doi.org/10.1021/acs.iecr.6b02446.
Martin JJ. Cubic equations of state—which? Ind Eng Chem Fundam. 1979;18:81–97. https://doi.org/10.1021/i160070a001.
Martinez A, Guennec YL, Privat R, Jaubert JN, Mathias PM. Analysis of the combinations of property data that are suitable for a safe estimation of consistent Twu α-function parameters: updated parameter values for the translated-consistent tc-PR and tc-RK cubic equation of state. J Chem Eng Data. 2018;3:3980–8. https://doi.org/10.1021/acs.jced.8b00640.
Matheis J, Müller H, Lenz C, Pfitzner M, Hickel S. Volume translation methods for real-gas computational fluid dynamics simulations. J Supercrit Fluids. 2016;107:422–32. https://doi.org/10.1016/j.supflu.2015.10.004.
Mathias PM, Copeman TW. Extension of the Peng–Robinson equation of state to complex mixtures: evaluation of the various forms of the local composition concept. Fluid Phase Equilib. 1983;13:91–108. https://doi.org/10.1016/0378-3812(83)80084-3.
Mathias PM, Naheiri T, Oh EM. A density correction for the Peng–Robinson equation of state. Fluid Phase Equilib. 1989;47:77–87. https://doi.org/10.1016/0378-3812(89)80051-2.
Michelsen ML, Mollerup JM. Thermodynamic models: fundamentals & computational aspects. 2nd ed. Holte: Tie-Line Publications; 2007.
Nighswander JA, Kalogerakis N, Mehrotra AK. Solubilities of carbon dioxide in water and 1 wt% NaCl solution at pressures up to 10 MPa and temperatures from 80 to 200 °C. J Chem Eng Data. 1989;34:355–60. https://doi.org/10.1021/je00057a027.
Park JY, Lim JS, Yoon CH, Lee CH, Park KP. Effect of a fluorinated sodium bis(2-ethylhexyl) sulfosuccinate (Aerosol-OT, AOT) analogue surfactant on the interfacial tension of CO2 + water and CO2 + Ni-plating solution in near- and supercritical CO2. J Chem Eng Data. 2005;50:299–308. https://doi.org/10.1021/je0499667.
Pedersen KS, Christensen PL, Shaikh JA. Phase behavior of petroleum reservoir fluids. Boca Raton: CRC Press; 2014. https://doi.org/10.1201/b17887.
Peneloux A, Rauzy E, Freze R. A consistent correction for Redlich–Kwong–Soave volumes. Fluid Phase Equilib. 1982;8:7–23. https://doi.org/10.1016/0378-3812(82)80002-2.
Peng DY, Robinson DB. A new two-constant equation of state. Ind Eng Chem Fundam. 1976;15:59–64. https://doi.org/10.1021/i160057a011.
Pereira L, Chapoy A, Burgass R, Oliveira MB, Coutinho J, Tohidi B. Study of the impact of high temperatures and pressures on the equilibrium densities and interfacial tension of the carbon dioxide/water system. J Chem Thernodyn. 2016;93:404–15. https://doi.org/10.1016/j.jct.2015.05.005.
Sako T, Sugata T, Nakazawa N, Obuko T, Sato M, Taguchi T, Hiaki T. Phase equilibrium study of extraction and concentration of furfural produced in reactor using supercritical carbon dioxide. J Chem Eng Jpn. 1991;24:449–54. https://doi.org/10.1252/jcej.24.449.
Schechter DS, Guo B. Parachor based on modern physics and their uses in IFT prediction of reservoir fluids. SPE Res Eval Eng. 1998;1:207–17. https://doi.org/10.2118/30785-PA.
Shariat A, Moore RG, Mehta SA, Fraassen KC, Newsham KE, Rushing JA. Laboratory measurements of CO2–H2O interfacial tension at HP/HT conditions: implications for CO2 sequestration in deep aquifers. In: Carbon management technology conference, 7–9 Februrary, Orlando, Florida, USA; 2012. https://doi.org/10.7122/150010-MS
Soave G. Equilibrium constants from a modified Redlich–Kwong equation of state. Chem Eng Sci. 1972;27:1197–203. https://doi.org/10.1016/0009-2509(72)80096-4.
Song KY, Kobayashi R. Water content of CO2 in equilibrium with liquid water and/or hydrates. SPE Form Eval. 1987;2:500–8. https://doi.org/10.2118/15905-PA.
Spycher N, Pruess K, King JE. CO2–H2O mixtures in the geological sequestration of CO2. I. Assessment and calculation of mutual solubilities from 12 to 100°and up to 600 bar. Geochim Cosmochim Acta. 2003;67:3015–31. https://doi.org/10.1016/S0016-7037(03)00273-4.
Stryjek R, Vera JH. PRSV: an improved Peng–Robinson equation of state for pure compounds and mixtures. Can J Chem Eng. 1986;64:323–33. https://doi.org/10.1002/cjce.5450640224.
Sugden S. The Parachors and valency. London: G. Routledge; 1930.
Tabasinejad F, Barzin Y, Moore GR, Mehta SA, Fraassen KCV, Rushing JA, Newsham KE. Water/CO2 system at high pressure and temperature conditions: measurement and modeling of density in equilibrium liquid and vapor phases. In: SPE EUROPEC/EAGE annual conference and exhibition, 14–17 June, Barcelona, Spain; 2010. https://doi.org/10.2118/131636-MS.
Teng H, Yamasaki A, Chun MK, Lee H. Solubility of liquid CO2 in water at temperatures from 278 K to 293 K and pressures from 6.44 MPa to 29.49 MPa and densities of the corresponding aqueous solutions. J Chem Thermodyn. 1997;29:1301–10. https://doi.org/10.1006/jcht.1997.0249.
Tödheide K, Frank EU. Das Zweiphasengebiet und die kritische Kurve im system Kohlendioxid–Wasser bis zu Drucken von 3500 bar. Z Phys Chem. 1962;37:387–401. https://doi.org/10.1524/ZPCH.1963.37.5_6.387.
Trusler JPM. Thermophysical properties and phase behavior of fluids for application in carbon capture and storage processes. Ann Rev Chem Biomol Eng. 2017;8:381–402. https://doi.org/10.1146/annurev-chembioeng-060816-101426.
Twu CH, Bluck D, Cunningham JR, Coon JE. A cubic equation of state with a new alpha function and a new mixing rule. Fluid Phase Equilib. 1991;69:33–50. https://doi.org/10.1016/0378-3812(91)90024-2.
Valtz A, Chaopy A, Coquelet C, Paricaud P, Richon D. Vapor-liquid equilibria in the carbon dioxide-water system, measurement and modeling from 278.2 to 318.2 K. Fluid Phase Equilib. 2004;226:333–44. https://doi.org/10.1016/j.fluid.2004.10.013.
van der Waals JD. Continuity of the gaseous and liquid state of matter. PhD Dissertation, Univ. of Leiden; 1873.
Weibe R, Gaddy VL. The solubility in water of carbon dioxide at 50, 75, and 100 °C, at pressures to 700 atmospheres. J Am Chem Soc. 1939;61:315–8. https://doi.org/10.1021/ja01871a025.
Weibe R, Gaddy VL. The solubility of carbon dioxide in water at various temperatures from 12 to 40 °C and at pressures to 500 atmospheres: critical phenomena. J Am Chem Soc. 1940;62:815–7. https://doi.org/10.1021/ja01861a033.
Wong DSH, Sandler SI. A theoretically correct mixing rule for cubic equations of state. AIChE J. 1992;38:671–80. https://doi.org/10.1002/aic.690380505.
Yaginuma R, Sato Y, Kodama D, Tanaka H, Kato M. Saturated densities of carbon dioxide + water mixture at 304.1 K and pressures to 10 MPa. J Jpn Inst Energy. 2000;79:144–6. https://doi.org/10.3775/jie.79.144.
Young AF, Pessoa FLP, Ahón VRR. Comparison of 20 alpha functions applied in the Peng–Robinson equation of state for vapor pressure estimation. Ind Eng Chem Res. 2016;55:6506–16. https://doi.org/10.1021/acs.iecr.6b00721.
Young AF, Pessoa FLP, Ahón VRR. Comparison of volume translation and co-volume functions applied in the Peng–Robinson EOS for volumetric corrections. Fluid Phase Equilib. 2017;435:73–87. https://doi.org/10.1016/j.fluid.2016.12.016.
Zhang J, Feng Q, Wang S, Zhang X, Wang S. Estimation of CO2-brine interfacial tension using an artificial neural network. J Supercrit Fluids. 2016;107:31–7. https://doi.org/10.1016/j.supflu.2015.08.010.
Zhao H, Fedkin MV, Dilmore RM, Lvov SN. Carbon dioxide solubility in aqueous solutions of sodium chloride at geological conditions: experimental results at 323.15, 373.15, and 423.15 K and 150 bar and modeling up to 573.15 K and 2000 bar. Geochim Cosmochim Acta. 2015;149:165–89. https://doi.org/10.1016/j.gca.2014.11.004.
Zhao H, Lvov SN. Phase behavior of the CO2–H2O system at temperatures of 273–623 K and pressures of 0.1–200 MPa using Peng–Robinson–Stryjek–Vera equation of state with a modified Wong–Sandler mixing rule: an extension to the CO2–CH4–H2O system. Fluid Phase Equilib. 2016;417:96–10. https://doi.org/10.1016/j.fluid.2016.0.027.
Author information
Authors and Affiliations
Corresponding author
Additional information
Edited by Yan-Hua Sun
Appendices
Appendix A: PR EOS and α functions
The PR EOS (Peng and Robinson 1976) can be expressed as:
where p is the pressure in bar; v stands for the molar volume in cm3/mol; T is the temperature in K; a is the attraction parameter with unit of bar cm6/mol2, and b is the repulsion parameter with unit of cm3/mol; a and b can be determined by Eqs. (12) and (13):
where R is the universal gas constant in J/(mol K); Tc is the critical temperature in K; pc is the critical pressure in bar; and α is the so-called alpha function.
The expression of Twu α function can be written as (Twu et al. 1991):
where Tr is the reduced temperature; L, M and N are compound-specific parameters. The values of these parameters regressed by Martinez et al. (2018) are used in this study.
Gasem α function can be expressed by (Gasem et al. 2001):
where the values of correlation parameters A through E are 2.0, 0.836, 0.134, 0.508 and −0.0467, respectively.
Appendix B: Summary of van der Waals one-fluid mixing rule and its BIPs
The van der Waals one-fluid mixing rule can be expressed as (van der Waals 1873):
where zi is the molar fraction of the ith component in the mixture; ai and bi can be calculated by Eqs. (12) and (13); kcij and kdij are the BIPs that need to be fitted. In this study, the linear temperature-dependent BIP correlations from the study by Abudour et al. (2012a, b) are applied for CO2/H2O mixtures. Table
12 lists the BIP correlations obtained by Abudour et al. (2012a, b).
When vdW mixing rule is used in PR EOS, the fugacity coefficient can be written as:
where
where Z is the compressibility factor. For PR EOS, Z can be calculated by Eq. (21).
where
Appendix C: Summary of Huron–Vidal mixing rule and its BIPs
In the Huron–Vidal mixing rule, the following equations are applied to calculate am and bm (Huron and Vidal 1979):
where \(G_{{\infty}}^{{\text{E}}}\) is the excess Gibbs energy at infinite pressure, and Λ is an EOS-dependent parameter. For PR EOS, Λ = 0.62323 (Huron and Vidal 1979).
The excess Gibbs energy corresponding to the non-random two-liquid (NRTL) (Zhao and Lvov 2016; Wong and Sandler 1992) model can be expressed by (Aasen et al. 2017; Huron and Vidal 1979):
where
The generalized BIP correlations for τij obtained by Aasen et al. (2017) are given below:
where T0 = 1000 K is the reference temperature.
When the Huron–Vidal mixing rule is used in PR EOS, the fugacity coefficient can be calculated by (Zhao and Lvov 2016):
where lnγi is the activity coefficient of component i and can be expressed as (Zhao and Lvov 2016):
The derivation of the expression of the activity coefficient in Huron-Vidal mixing rule is detailed in “Appendix H”.
Appendix D: Summary of volume translation models applied in this study
The constant volume translation can be expressed as (Peneloux et al. 1982; Jhaveri and Youngren 1988):
where vcorr is corrected molar volume in cm3/mol; vEOS stands for PR-EOS-calculated molar volume in cm3/mol; ci is the component-dependent volume shift parameter which can be determined by Eq. (35) (Young et al. 2017).
The values of si used by Liu et al. (2016) are applied in this study (\(s_{{{\text{H}}_{2} {\text{O}}}} = 0.23170\) and \(s_{{{\text{CO}}_{2} }} = - 0.15400\)).
Abudour volume translation model can be expressed as (Abudour et al. 2012a, b):
where \(\delta_{{\text{c}}}\) is volume correction at the critical temperature in cm3/mol; d is the dimensionless distance function given by (Mathias et al. 1989; Abudour et al. 2012b):
where ρ is the molar density in mol/cm3. The volume translation function proposed by Abudour et al. (2013) was extended to mixtures by the following equations (Abudour et al. 2013):
where (Abudour et al. 2013):
where Tcm, pcm and δcm are the critical temperature, critical pressure and volume correction of the mixture at the critical point, respectively. c1i is the specie-specific parameter of component i and has a linear relationship with critical compressibility (\(Z_{{\text{c}}}\)) (Abudour et al. 2012b):
The term dm can be derived using the original PR EOS (Matheis et al. 2016):
The volume correction of the given mixture at the critical point, δcm, can be determined by (Abudour et al. 2013):
where vci is the critical volume of component i; θi is the surface fraction of component i defined by (Abudour et al. 2013):
The critical temperature of the mixture can be calculated via the following mixing rule (Abudour et al. 2013).
The critical pressure of the mixture can be determined by the correlation proposed by Aalto et al. (1996):
where ωm is the acentric factor of the mixture (Abudour et al. 2013):
where ωi is the acentric factor of component i.
Appendix E: Summary of existing IFT correlations for CO2/H2O mixtures
Parachor model (Sugden 1930) can be expressed as below (Schechter and Guo 1998):
where xi and yi are the mole fractions of component i in liquid and vapor phases, respectively; Pi is the Parachor value of component i (\(P_{{{\text{H}}_{2} {\text{O}}}} = 52\), \(P_{{{\text{CO}}_{2} }} = 78\)) (Liu et al. 2016); \(\rho_{{\text{L}}}^{{\text{M}}}\) is the molar density of liquid phase in mol/cm3; \(\rho_{{\text{V}}}^{{\text{M}}}\) is the molar density of vapor phase in mol/cm3.
Hebach et al. (2002)’s correlation can be expressed as:
where (Hebach et al. 2002):
where \(\rho_{{{\text{CO}}_{2} }}\) is the CO2-rich-phase density in g/cm3; \(\rho_{{{\text{H}}_{2} {\text{O}}}}\) is the aqueous-phase density in g/cm3; k0 to k6 and b0 to b1 are empirical coefficients. The units of T, p, and dd are K, bar, and g2/cm6, respectively. Table
13 lists the values of original and refitted coefficients.
Chen and Yang (2019)’s correlation is given as:
where σ is IFT in mN/m; pr is the reduced pressure of CO2; C1 to C5 are empirical coefficients. To make fair comparisons, we refit these coefficients based on the IFT database employed in this study. Table
14 summarizes the values of these refitted coefficients.
Appendix F: Pressure–temperature coverage of phase-density and IFT data collected from the literature.
Figure
8 depicts the pressure–temperature coverage of phase-density and IFT data collected from the literature over 278–478.35 K and 2.5–1291.1 bar, and 278.15–477.59 K and 1–1200.96 bar, respectively.
Appendix G: Experimental data selection in IFT database employed in this study
The collected IFT data are further screened to remove any obvious outliers. Figure
9 shows the identification of the outliers from the collected data over 40–60 bar and 278.15–298.15 K, while Fig.
10 shows the identification of outliers from the collected data over 100–270 bar and 307.15–314.15 K.
As seen in Fig. 9a, the measured IFT data by Chun and Wilkinson (1995) and Park et al. (2005) fall into the range of 5–8 mN/m over 278.15–288.15 K and 40–60 bar, which are significantly lower than the measured IFT data (i.e., around 22–28 mN/m) obtained by other studies under similar conditions. Figure 9b indicates that the measured IFT data by Chun and Wilkinson (1995) and Park et al. (2005) fall into the range of 10–14 mN/m over 293.15–298.15 K and 50–70 bar, which are significantly lower than the measured IFT data (i.e., around 28–31 mN/m) obtained by other studies under similar conditions.
Figure 10 indicates that the measured data by Bachu and Bennion (2009) fall into the range of 16–19 mN/m over 307.15–314.15 K and 120–270 bar, which are significantly lower than the measured IFT values (i.e., around 30 mN/m) obtained by other studies under similar conditions. No outlier exists at other temperature and pressure conditions. We have removed these outliers in the IFT regression analysis.
Appendix H: Derivation of activity coefficient in the fugacity expression when Huron-Vidal mixing rule is used.
Similar to the approach used by Wong and Sandler (1992), the activity coefficient of component i can be expressed by the following formula:
where the excess Gibbs free energy can be expressed as (Huron and Vidal 1979):
To make the derivation process more intuitive, we can set i = 1 (the first component) and n = 2 (two compounds in the system). Then the excess Gibbs free energy can be expressed as:
Taking the partial derivative of the first term in the right hand side of Eq. (56) yields:
Taking the partial derivative of the second term in the right hand side of Eq. (56) yields:
As such, Eq. (54) can be expressed as:
Using letter i to replace number 1 leads to a general expression of activity coefficient (Huron and Vidal 1979):
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Cui, Z., Li, H. Toward accurate density and interfacial tension modeling for carbon dioxide/water mixtures. Pet. Sci. 18, 509–529 (2021). https://doi.org/10.1007/s12182-020-00526-x
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s12182-020-00526-x