Article pubs.acs.org/IECR
Prediction of Phase-Inversion Temperature of a Triglyceride Microemulsion Using Design of Experiments Zahra Jeirani,† Badrul Mohamed Jan,†,* Brahim Si Ali,† Ishenny Mohd Noor,† See Chun Hwa,‡ and Wasan Saphanuchart‡ †
Department of Chemical Engineering, Faculty of Engineering, University of Malaya, 50603 Kuala Lumpur, Malaysia BCI Chemical Corporation Sdn. Bhd., Lot 7, Jalan BS 7/22, Taman Perindustrian Bukit Serdang, Seksyen 7, 43300 Seri Kembangan, Selangor Darul Ehsan, Malaysia
‡
ABSTRACT: This paper presents the estimation of phase inversion temperature (PIT) of a nonionic microemulsion by studying its interfacial tension (IFT) as a function of temperature. In this work, one-factor design (OFD), which is an approach of Design of Experiments (DOE), was used to model the variation of IFT with temperature. A transformed quadratic equation with lambda of 0.87 was then fitted and validated. The model was optimized to predict the PIT, the temperature at which the IFT is a minimum. The IFT from the optimization of the model at the predicted PIT of 44.43 °C is 0.000154709 mN/m. The value shows an excellent agreement with the experimental value (0.00018 mN/m). Therefore, DOE is capable of predicting the PIT of a microemulsion with high accuracy using only seven design points.
1. INTRODUCTION Microemulsions are translucent and thermodynamically stable dispersions which are made spontaneously from two immiscible phases (water and oil) together with an effective surfactant.1−3 The importance of microemulsions has been extensively established in numerous applications in various scientific fields such as liquid membrane,4 biotechnology,5 enhanced petroleum recovery,6 corrosion,7 coatings and textile finishing,8 detergent,9 cosmetics,10 agrochemical,11 pharmaceutical,12 nanoparticle synthesis,13 analytical chemistry, and remediation,14 food science,15 and microreactor16 industries. Microemulsions have been classified into three main categories namely Winsor Type I, Winsor Type II, and Winsor Type III.17 In the Winsor Type I system, oil is dissolved in normal micelles in the water phase. Thus the lower phase microemulsion is in equilibrium with the excess oil. In the Winsor Type II system, water is dissolved in reverse micelles in the oil phase. Thus the upper phase microemulsion is in equilibrium with the excess water. Winsor Type III system or middle phase microemulsion can be envisioned as having both the water and oil swollen micelles in a bicontinuous structure. Thus the middle phase microemulsion is in equilibrium with both the excess water and excess oil phases.17 It is no secret that Winsor Type I can be changed to Type III and subsequently to Type II by changing some parameters, which are different for various types of the surfactant used in the structure of microemulsion.18 For ionic surfactants, some of these parameters which cause the phase inversion are salinity and concentration of cosurfactant, while for nonionic surfactants, temperature is the controlling parameter which leads to phase inversion.18 In a microemulsion which is stabilized with a nonionic surfactant, the hydrophilic−lipophilic balance (HLB) of the surfactant tends to change considerably with temperature.19 The HLB value of a surfactant is a concept, which characterizes the relative oil and water solubility of the surfactants.20 The © 2012 American Chemical Society
HLB value can be calculated on the basis of the information of the surfactant structure using methodology described in the literature.21 At low temperatures, the hydration forces between the hydrophilic moiety of the surfactant and water are stronger, thus the surfactant tends to be more soluble in water than oil and it establishes the Winsor Type I microemulsion.18 On the other hand, at high temperatures, the hydration of the hydrophilic moiety of the surfactant tends to reduce. Thus the surfactant is more soluble in oil than water. It leads to the Winsor Type II microemulsion.18 In between the low and high temperatures, there seems to be a temperature at which the surfactant tends to shift its solubility preferential from water to oil as temperature increases. This temperature is referred to as the phase inversion temperature (PIT) or HLB temperature.19 This is because the phase inversion occurs when the hydrophilic−lipophilic property of a nonionic surfactant is well balanced. Therefore, Winsor Type III or middle phase microemulsion appears with unique properties such as ultralow IFT and high oil solubility.18 At this condition, the microemulsion is said to be very promising for various applications, such as enhanced oil recovery (EOR).18,22,23 Apparently, it is essential to estimate the value of the PIT for a fixed composition of a microemulsion. The PIT values of specific nonionic microemulsions have extensively been determined experimentally using various techniques at increasing temperature. Measurements of the IFT,18,24 conductivity,25 and viscosity,26 are some of the experimental techniques of determining the values of the PIT. PIT values could also be determined by visual observation of phase separation.27,28 Received: Revised: Accepted: Published: 744
August 13, 2012 December 10, 2012 December 18, 2012 December 18, 2012 dx.doi.org/10.1021/ie302180j | Ind. Eng. Chem. Res. 2013, 52, 744−750
Industrial & Engineering Chemistry Research
Article
water interface, and the other interfacial tension is in between the oil phase and the surfactant monolayer. The oil−water interfacial tension, which is equal to the sum of the two mentioned interfacial tensions, reaches its minimum in the middle of the temperature range of the three phase region, where both interfacial tensions are equal.18,33 In other words, the minimum of oil−water interfacial tension with temperature represents the PIT of the microemulsion. Therefore, this paper attempts to model the oil−microemulsion interfacial tension with temperature and predict the PIT using the design of experiment (DOE) technique. Response surface methodology (RSM) is a statistical method of DOE that uses quantitative data from appropriate experiments to determine its corresponding regression model equations and operating conditions.34 Thus RSM is used to find an approximating function in predicting future responses and determine the factor values to optimize the response function. RSM is a collection of mathematical and statistical techniques for modeling and analysis of problems. A simple RSM design called OFD was applied in this study to model and optimize the trend of IFT with temperature in determining the PIT. It was assumed that the temperature is the only factor that affects the IFT (response) for a fixed composition of microemulsion. Thus OFD is a suitable technique of RSM in this specific modeling case. Design Expert software version 8.0.6 (STAT-EASE Inc., Minneapolis, USA) was used to carry out OFD. In OFD, a quadratic model was developed to predict the single-response as a function of the single-variable. The coefficients of the model for the response were estimated using the regression analysis technique included in the OFD. The adequacy of the model was justified through analysis of variance (ANOVA). The final model was minimized to estimate the temperature at which the IFT is minimum. The predicted temperature corresponds to the PIT of the microemulsion.
DOE is an approximation technique, which has been applied to model microemulsion issues in the past.29,30 However, its application in estimating the PIT of a microemulsion is very scarce. In this paper, an attempt was made to investigate the potential of using DOE in the determination of the PIT by comparing the experimental and predicted values of the PIT obtained from the model for a specific nonionic microemulsion. Design Expert software was used for the DOE modeling.
2. MATERIALS AND METHODOLOGY 2.1. Materials. The surfactant used in the formulation of the microemulsion in this study was Glucopon 650EC, which is a mixture of alkyl polyglycosides (APG) having an average alkyl chain length of 11, HLB of 11.9, and critical micelle concentration (CMC) of 0.073 g/L at 37 °C.31 It was supplied by Cognis (Malaysia) Sdn. Bhd. which is currently part of BASF Chemical Company. Palm kernel oil, which was used as the oil phase of the triglyceride microemulsion, was purchased from Delima Oil Products Sdn. Bhd. in Malaysia. In addition, sodium chloride (NaCl, A.R. grade) used in the microemulsion was supplied by LGC Scientific, Malaysia. Compared to other vegetable oils, palm oil is an abundant and widely available vegetable oil in Southeast Asia, particularly in Malaysia. In comparison with other palm oils, palm kernel oil is much cheaper and it is preferred for industrial and nonfood applications. Finally, n-octane (free of olefins) and glyceryl monooleate (GM) were supplied by Sigma-Aldrich Sdn. Bhd., Malaysia. 2.2. Sample Preparation. A triglyceride microemulsion was prepared by mixing an equal mass of the oil and aqueous phases. The aqueous phase consists of 1 wt % APG, 3 wt % NaCl, 3 wt % GM, and 93 wt % deionized water. The oil phase was pure palm oil. After mixing all of the components at their desired fractions, the sample was shaken for a minimum duration of an hour to ensure a well-mixed and uniform mixture. Then the sample was poured into a separating funnel and left undisturbed for about one week to make sure equilibrium was reached. The Winsor Type I microemulsion phase was then separated from the excess oil phase and was used in the next IFT measurements. The triglyceride microemulsion is an effective chemical slug for the purpose of chemical flooding to enhance oil recovery.22,23 This is because the measured IFT values between the microemulsion and the oil remaining in the reservoir were ultralow.22,23 2.3. IFT Measurements. The IFTs between the heavy and light phases were measured and recorded over the temperature range of 35 to 55 °C at increments of 2.5 °C. In all of the IFT measurements, the heavy phase was the triglyceride microemulsion phase previously prepared, while the light phase was n-octane, which represents the crude oil in reservoirs.32 IFTs were measured using a spinning drop tensiometer site 4 (KRUSS, Germany) with the same procedure explained in our previous work.30 2.4. Application of DOE in the PIT Estimation. Phase inversion is a process whereby the internal and external phases of a microemulsion abruptly change from oil-in-water to waterin-oil microemulsion and vice versa. It takes place when the hydrophilic and lipophilic properties are balanced. The temperature at which this phenomenon takes place is called the PIT. The phase inversion phenomenon is related to the balance of the two interfacial tensions in the three phase region;18,33 one of the interfacial tensions is in between the water phase and the surfactant monolayer adsorbed at the oil−
3. RESULTS AND DISCUSSION Figure 1 shows the measured IFTs between the microemulsion and oil phases at various temperatures. It can be seen in Figure
Figure 1. Experimental IFTs between the microemulsion and oil phase at various temperatures.
1 that the IFT tends to reduce to a minimum (approximately 45 °C) before it starts to increase. The temperature at which the IFT is minimum corresponds to the PIT. 3.1. OFD Modeling. OFD can be developed up to a cubic model for one numeric factor. In this OFD modeling, the factor of the model is temperature, and the response is IFT. Table 1 presents the design summary. The low and high values of temperature set for the construction of adequate and reliable 745
dx.doi.org/10.1021/ie302180j | Ind. Eng. Chem. Res. 2013, 52, 744−750
Industrial & Engineering Chemistry Research
Article
Table 1. Design Summary study type: response surface design type: one factor design model: quadratic runs: 7 actual
coded
factor
name
units
low
high
A response
temperature name
°C
35
−1.000 maximum
1.000 ratio
0.01
47.8469
Y1
IFT
units
analysis
55 minimum
mN/m
polynomial
0.000209
runs were 35 and 55 °C, respectively. This led to the development of a mathematical model with the best experimental data fitting. Based on this information, the software automatically suggested the conditions of seven experimental runs to be performed. User then conducted the tests at the determined conditions proposed by the OFD model. The conditions of the seven experimental runs determined by the model and the corresponding experimental response values are presented in Table 2.
factor 1A: temperature °C
response 1: IFT mn/m
1 2 3 4 5 6 7
35 40 35 55 50 55 45
0.007824 0.001378 0.007812 0.01000 0.002596 0.00990 0.000209
The quadratic model was selected as suggested by the Design Expert software. Regression analysis was performed to fit the response function of the IFT. The final empirical model in terms of actual factors can be expressed as follows: IFT = + 0.17578 − 7.92347 × 10−3T + 8.92417 × 10−5T 2
high
the model is significant. In addition, the ANOVA table also shows that all of the model parameters are significant because their p-values are less than 0.05. Lack-of-fit is another statistical parameter in an ANOVA table, which is displayed when unnecessary additional design points were used for replication to provide an estimate of pure error. Lack-of-fit compares the residual error to the pure error. It is desirable that the F value of lack-of-fit to be small and its pvalue to be greater than 0.05. In this modeling, the p-value of lack-of-fit is 0.0323, which is less than 0.05. On the other hand, the F-value of lack-of-fit is 29.95, which is very high. Therefore, the lack-of-fit is significant and the model has not fitted the design points properly. Thus the OFD model of eq 1 is not accurate and cannot be applied to predict the response precisely. The incapability of the model was also confirmed by graphical analysis. Figure 2 shows the OFD model graph of eq 1. As illustrated in Figure 2, at a temperature of 44.49 °C the IFT reaches a minimum value of −9.48622 × 10−5. The negative value of the predicted IFT is not practical and indicates the inefficiency of the model. To improve the model, an appropriate transformation can be applied on the response data. Power transformation is one of the available transformations in OFD of Design Expert software. Design Expert software recommends the most appropriate lambda value for the power transformation from the location of the minimum in the Box−Cox plot. Lambda is the power raised by the response in power transformation analysis. The Box−Cox plot for the model in eq 1 is shown in Figure 3. The suggested lambda for the power transformation was 0.87. Thus the power transformation with lambda value of 0.87 was imposed on the model in eq 1 and its effect on the improvement of the model was investigated. Therefore, a new quadratic model with power transformation was achieved. The transformed model in terms of the actual factor can be expressed as follows:
Table 2. The Conditions of the 7 Runs Determined by the Model and Their Experimental Response Values run
low
(1)
where T is temperature in °C and IFT is interfacial tension between the microemulsion and n-octane in mN/m. The coefficients of the model for the response were estimated using multiple regression analysis technique included in the OFD. ANOVA, one of the powerful statistical analyses, was used to check the adequacy of the model. ANOVA results are given in Table 3. The model F-value of 1317.59 implies the model is significant. There is only 0.01% chance that a “Model F-Value” in this magnitude could occur due to noise. The p-value (or Prob > F) of the model is less than 0.0001, which confirms that
(IFT)0.87 = + 0.31367 − 0.014096T + 1.58611 × 10−4T 2 (2)
Table 3. ANOVA Table for the Polynomial Model source model A-temperature A2 residual lack-of-fit pure error cor total
sum of squares 1.034 5.277 9.813 1.570 1.519 5.072 1.036
× × × × × × ×
−4
10 10−6 10−5 10−7 10−7 10−9 10−4
df 2 1 1 4 2 2 6
mean square 5.170 5.277 9.813 3.924 7.594 2.536
× × × × × ×
−5
10 10−6 10−5 10−8 10−8 10−9
746
F value
p-value prob > F
remarks
15786.08 7029.39 2475.08
