Investigating Polymeric Entrainers for Azeotropic Distillation of theEthanol/Water and MTBE/Methanol Systems

Adnan M. Al-Amer*,†

Department of Chemical Engineering, King Fahd University of Petroleum and Minerals,Dhahran 31261, Saudi Arabia

In this work, selected polymeric entrainers have been investigated to assess their capability ofbreaking the azeotrope of ethanol/water and MTBE/methanol systems. Solubility testing andgroup contribution model calculations were used to guide in the initial selection of potentialpolymers. Experimental VLE measurements were performed to determine whether the selectedpolymers are capable of breaking the azeotrope. We have found polymeric entrainers capable ofbreaking the azeotrope for the ethanol/water system. Poly(ethylene glycol) at 10 wt % and poly-(acrylic acid) at 0.45 wt % did break the azeotrope for the ethanol/water system. This conclusionis based on composition and temperature data. Other polymers used with the ethanol/watersystem might be capable of breaking the azeotrope, but we could not conclusively determinethis from the collected data. From the results obtained for the MTBE/methanol system, we werenot able to definitively identify such entrainers. This is because of the difficulty in finding apolymer that will substantially dissolve in both MTBE and methanol and, at the same time,will provide the required specific interaction with each component. The experimental VLE datafor selected systems were fit to the UNIQUAC model. A satisfactory fit was obtained, and theparameters are reported.

Introduction

Complete separation of azeotropic mixtures requireseither the coupling the distillation columns with otherseparation methods such as adsorption, membranes,and extraction or the use of more complex distillationschemes based on a modification of the equilibrium toeffect the complete separation. Although many newseparation techniques are being developed, distillationwill remain the method of choice for large-scale separa-tion of nonideal mixtures including azeotropic mix-tures.1 Separation of such mixtures is achieved by useof one of the enhanced distillation methods. Theseinclude extractive distillation, salt distillation, pressure-swing distillation, reactive distillation, and azeotropicdistillation. The latter method involves the use ofentrainers to alter the relative volatility of the compo-nents and break the azeotrope. The choice of separationmethod depends on the specific system and economics.2

Much research on entainer selection and distillationsystem configuration has been performed. In 1991,Laroche et al.3 presented practical solutions for choosingthe best entrainers required for separating binaryazeotropes into pure components. They also suggestedthe feasible flow sheet of separation sequences for eachentrainer. Their analysis showed that a good entraineris a component that breaks the azeotrope and yieldshigh relative volatilities between the two azeotropicconstituents. They used equivolatility curve diagramsto compare entrainers.

In 1996, Widago and Seider4 reviewed recent workon the separation of azeotropic mixtures. They examinedthe important considerations in the selection of entrain-

ers as success in azeotropic distillation is largelydetermined by the choice of entrainer. In the past,entrainers were selected using a trial-and-error proce-dure with much reliance on experimental data. Thisresults in a waste of time and resources. More recently,entrainers have been selected on the basis of theirpotential for producing feasible designs. Widago andSeider presented simple rules using maps of distillationlines in preference to residue curves for screening themany possible entrainers. These maps are useful inearly design stages, providing specifications for thedesired separations.

Bekiaris and Morari (1996)5 extended the multiplicityanalysis from ternary systems (two components plusand entrainer) to quaternary systems for the case ofinfinite flux and infinite number of trays. They showedhow their work could be useful for the selection ofentrainer.

Guttinger and Morari (1996)6 studied multiple steadystates in two or more columns typically used in ternaryazeotropic distillation. They found that, for the inter-mediate entrainer scheme, multiplicity could occur forall feed compositions.

Safrit and Westerberg (1997)7 examined the case ofa continuously flowing extractive agent to facilitate theseparation of azeotropic mixtures. They showed thesensitivity of the separation’s profile to the entrainerflow rate.

Group contribution calculations reported in the lit-erature revealed a number of suitable low-molecular-weight entrainers.8 One possible guide for polymericentrainers is to choose the polymeric counterparts ofthose low-molecular-weight entrainers. For example, thefollowing entrainers were found suitable for breakingthe azeotrope in the ethanol/water system: acetic acid,2-aminoethanol, N,N-dimethylformamide (DMF), eth-ylene glycol, meso-2,3-butanediol, and morpholine. Ace-

* Telephone: 8602353.Fax: 8602334.E-mail: alamer@kfupm.edu.sa.

† Previous publications by the author appeared under hissubfamily name of Al-Jarallah.

3901Ind. Eng. Chem. Res. 2000, 39, 3901-3906

10.1021/ie0000515 CCC: $19.00 © 2000 American Chemical SocietyPublished on Web 09/08/2000

tic acid and ethylene glycol are already used in theindustry. Therefore, in this research, we tested theirpolymeric counterparts: poly(acrylic acid) and poly(vinylalcohol). However, one should keep in mind that ter-minal functional groups are sometimes lost in thepolymerization, such as in ethylene glycol, or packedvery closely along the main chain, which might reducethe effectiveness of these polymers.

The advantage of using such polymers is that theyremain in the liquid phase. They can be separated byultrafiltration. Hence, the separation can be completedin a smaller number of columns.

Thus, in this work, selected polymeric entrainers havebeen investigated to assess their capability of breakingthe azeotrope of the ethanol/water and MTBE/methanolsystems. Solubility testing and a group contributionmodel calculations were used to guide in the initial se-lection of potential polymers. Experimental VLE mea-surements were performed to determine whether theselected polymers are capable of breaking the azeotrope.The VLE data were fit using the UNIQUAC model.

Criteria for Entrainers SelectionBecause there were a very large number of polymers

that might have been applicable, the following criteriawere used to guide the selection of the polymers:

(1) Polymer availability and cost. In general, thepolymer should be available commercially at reasonablecost.

(2) Polymer solubility. The polymer should be solublein the system. The solubilities of the selected polymersmust be large enough to allow for sufficient specificinteractions to attract one of the components and thusbreak the azeotrope. For the ethanol/water system, thepolymers must be polar in order to dissolve in water.

(3) Group contribution calculations. To obtain a rela-tive ranking of the ability of the potential polymers tobreak the azeotrope, one can use a group contributionmodel. Group contribution models such as UNIFAC donot require a set of fitted parameters. It is sufficient toknow the structure of the components to obtain the VLEbehavior (assuming that the group parameters areavailable). These methods are approximate and areexpected to only give a qualitative indication as to whichpolymer is the best, which is the second best, and soon.

Group contribution models allow for the calculationof activity coefficients based on groups rather thanwhole molecules. Calculations for a polymer can be doneby selecting the proper groups for that polymer inaddition to the groups for the two volatile components.The criterion for a useful polymer is the relativevolatility, R, at the azeotropic temperature at a totalpressure of 1 atm.

The MTBE/methanol system forms an azeotrope at atemperature of 55 °C and atmospheric pressure with xi) yi ≈ 0.7.9 The relative volatility is equal to unity forthe azeotrope. Using a group contribution model, theactivity coefficient γ1 is calculated for different addedamounts of the selected polymers. For this system, theobjective is to have a polymer with strong interactionswith MTBE and weak interactions with methanol. Thiswill lead to a value of R12 lower than unity. This is thedesired behavior for this system because the feed to thedistillation column, which comes from the reactor, hasan MTBE mole fraction close to unity. For the ethanol/water system, the objective is to increase R12 aboveunity.

The details of the initial list of polymers and theresults of the solubility tests and the group contributioncalculations are given in ref 10. On the basis of thesolubility tests and the relative volatility calculationsmentioned above, we selected 11 systems with entrain-ers to run the VLE experiments. These are given inTable 1.

The selected polymeric entrainers for ethanol/waterinclude nonionic and ionic polymers. For the methanol/MTBE system, it was difficult to find a polymer solublein both components. This fact limited our study to onepolymer [poly(ethylene glycol)]. However, we have in-cluded the olegomeric counterparts of this polymer,tetraethylene glycol and triethylene glycol. These com-pounds have very low vapor pressures at the tempera-ture of the VLE and, therefore, were assumed to benonvolatile in our calculations.

Experimental Section

Analar-grade MTBE and methanol were used for theMTBE/methanol system.

Refractometer model RFM 340, Bellingham & StanleyLtd. (BS), was used for the measurement of refractiveindex. This instrument was calibrated and used for theanalysis of methanol/MTBE mixtures. This analysistechnique gave better results than gas chromatography.

For the MTBE/methanol systems with polymers, wealso used the refractive index. We prepared a calibrationcurve for each polymer weight fraction.

Analar-grade ethanol and distilled water were usedfor the ethanol/water system.

Karl Fischer titration techniques using Mettler KFtitrator model DL30 were used for the analysis ofethanol/water mixtures as we found that this methodgives the most satisfactory results compared with gaschromotographic and refractive index methods.

The VLE measurements were made using the FischerVLE model 0602 constant-pressure still operated atatmospheric pressure. More details on the experimentalprocedure are provided in ref 10

Results and Discussion

Experimental VLE Data. First, only mixtures of thevolatile components (ethanol/water and MTBE/metha-nol) without the entrainers were used. The purpose ofthis step was to test the VLE equipment and verify itsproper functioning by comparison to literature data fromPerry11 and to verify the reproducibility of the measure-ments. A sample result from this test is shown in Figure1 for the ethanol/water system without entrainer. Fromthis figure, good agreement with published data isverified. The agreement ensures us about the good

Table 1. Systems with Entrainers

system entrainer wt %

ethanol/water poly(ethylene glycol) (MW ) 1000) 35

10poly(ethylenimine) (BDH cat#15047) 3

515

poly(acrylic acid) (MW ) 2000) 0.45poly(2-acrylamido-2-methyl-1-propanesulfonic acid) (Aldrich cat#19197-3)

5

MTBE/methanol poly(ethylene glycol) (MW ) 1000) 3.3tetraethylene glycol 3.3triethylene glycol 3.3

3902 Ind. Eng. Chem. Res., Vol. 39, No. 10, 2000

functioning of the still and the reliability of the experi-mental procedure and analysis technique.

The presence of an azeotrope is judged from thecomposition data crossing the x ) y line and from thepresence of a minimum value in temperature. Thus, thecriteria for the success of a polymer entrainer inbreaking the azeotrope is the absence of this behavior,from both curves to be conservative.

For all the systems with entrainers in Table 1, theVLE data were plottted as xy and Txy curves. Selectedresults of VLE data with the entrainers for both systemsare plotted in Figures 2-10.

The criteria for a successful entrainer were not metfor all of the MTBE/methanol systems investigated.Figure 2 shows a sample of a typical result for thissystem with 3.3 wt % poly(ethylene glycol) in which theVLE data crossed the x ) y line. Figure 3 confirms this

result as a minimum in temperature is clearly shown.Thus, none of the polymers tested for this system brokethe azeotrope. The failure to break the azeotrope in themethanol/MTBE system is due to the difficulty infinding a polymer that will substantially dissolve in bothMTBE and methanol and, at the same time, will providethe required specific interaction with each component.The search for such a polymer requires an extensivestudy.

For the ethanol/water system, poly(ethylene glycol)at 10 wt % and poly(acrylic acid) at 0.45 wt % did breakthe azeotrope. This is concluded from the compositionand from the temperature data. Even if there is someerror in composition date, the absence of a minimum intemperature confirms the absence of an azeotrope.These results are shown in Figures 4-7 for thesesystems. The rest of the polymer systems tested for the

Figure 1. Ethanol/water xy diagram from the literature and fromthis work.

Figure 2. xy diagram for methanol/MTBE with 3.3 wt % poly-(ethylene glycol). The curve is the UNIQUAC fit.

Figure 3. Txy diagram for methanol/MTBE with and without 3.3wt % poly(ethylene glycol). The curve is the UNIQUAC fit for theternary mixture.

Figure 4. xy diagram for ethanol/water with 10 wt % poly-(ethylene glycol). The curve is the UNIQUAC fit.

Ind. Eng. Chem. Res., Vol. 39, No. 10, 2000 3903

ethanol/water system came very close to the x ) yline, but we are not sure if they crossed this line. Inother words, it is not certain that they broke the azeo-trope. For example, the results for 5% and 15% poly-ethyleneimine are shown in Figures 8 and 9. The Txyplot for the latter system is shown in Figure 10; as onecan see it is not clear that a minimum in temperatureis not present. Therefore, a conservative conclusion canbe made that these polymers did not break the azeo-trope.

Although we found two polymers that were capableof breaking the azeotrope for the ethanol/water system,the gap between the equilibrium curve and the x ) yline is small. This means that a large number of trayswill be needed in distillation columns. More work isneeded to increase the difference between x and y inthe ethanol-rich region.

Fitting VLE Data. A number of models were devel-oped recently to specifically represent the activitycoefficients of polymers in solutions. In this work, wechoose not to use these models because of the lowmolecular weights of our polymers and the fact that weare more concerned with the behavior of the volatilecomponents. The collected VLE data were fit to theUNIQUAC model. This model was chosen because of itswide acceptance in the literature and its accuracy inrepresenting VLE data for a wide range of systems.

Equilibrium was represented by the equation

For binary systems, there are two parameters, but forternary systems, there are six parameters. The param-eters obtained in this work are listed in Table 2. We fit

Figure 5. Txy diagram for ethanol/water with and without 10wt % poly(ethylene glycol). The curve is the UNIQUAC fit for theternary system.

Figure 6. xy diagram for ethanol/water with 0.45 wt % poly-(acrylic acid). The curve is the UNIQUAC fit.

Figure 7. Txy diagram for ethanol/water with and without 0.45wt % poly(acrylic acid). The curve is the UNIQUAC fit for theternary mixture.

Figure 8. xy diagram for ethanol/water with 5% polyethylene-imine.

yiP ) xiγiPisat (1)

3904 Ind. Eng. Chem. Res., Vol. 39, No. 10, 2000

the vapor composition to the experimental data via

where γi is the activity coefficient of component icalculated from the UNIQUAC equation, with τij )exp(-Rij/T) and τij ) 1 for all i ) j.

For the two-component systems, i.e., methanol/MTBEand ethanol/water, a12 ) θ1 and a21 ) θ2, whereas forthe three-component systems (methanol/MTBE/poly-mer or ethanol/water/polymer), a12 ) θ1, a13 ) θ2, a21 )θ3, a23 ) θ4, a31 ) θ5, and a32 ) θ6. The experimentaldata for the various systems were fit to the UNIQUACequation and the resulting values of θ are summarizedin the Table 2. The table also shows the sum of squareerror in the fitted yi value and the maximum absoluteerror. The fit is satisfactory, except for ethanol/waterwith 5% PEG. The large maximum error for this weightfraction is probably due to experimental error. The fitsto the other two weight fractions, 3% and 10%, aresatisfactory. We perform the fit for PEI in ethanol/waterfor only the highest weight fraction (15%).

The VLE data are plotted in Figures 1-10, along withthe UNIQUAC prediction in most cases.

Conclusions

No polymer among the polymers studied was foundto break the azeotrope for the MTBE/methanol system.For the ethanol/water system, at least two polymers,10 wt % poly(ethylene glycol) and 0.45 wt % poly(acrylicacid), have been found to break the azeotrope for thissystem. The UNIQUAC fit for the VLE data has beenfound to be satisfactory. The UNIFAC group contribu-tion prediction was not satisfactory in predicting therelative volatility. More accurate group contributionmethods would make the polymer-design part morereliable.

Acknowledgment

The support of the King Fahd University of Petroleumand Minerals and SABIC for Project ChE/SABIC/96-4is duly acknowledged. The help of my colleague and co-investigator, Professor E. Z. Hamad, is greatly appreci-ated.

Table 2. UNIQAC Fit of Experimental Data

exptl systema θ1 θ2 θ3 θ4 θ5 θ6 SSEb |err ymax|cMeOH-MTBE -57.062 412.822 0.0042 0.042MeOH-MTBE-T3G

3.3 wt % -66.223 -335.967 532.446 146.72 -231.677 -567.662 0.0067 0.013MeOH-MTBE-T4G

3.3 wt % -74.578 -341.708 580.129 172.605 -233.72 -606.714 0.0006 0.015MeOH-MTBE-PEG

3.3 wt % -68.843 -176.808 461.326 -28.7472 29.3171 419.911 0.0014 0.020EtOH-water 123.993 -31.511 0.0036 0.037EtOH-water-PAA

0.45 wt % 297.008 3197.53 -113.489 -3024.71 6221.23 543.105 0.0048 0.042EtOH-water-PEI

15 wt % 114.868 -136.413 -10.264 -209.725 -174.236 -153.013 0.0052 0.038EtOH-water-PEG

10 wt % PEG -47.7105 -1123.99 171.571 1270.69 -8.43557 -481.231 0.0018 0.0195 wt % PEG -120.099 269.347 423.963 -199.267 -868.854 -801.525 0.0394 0.1843 wt % PEG -11.1501 -2278.72 110.854 5461.57 -581.793 -633.953 0.0033 0.043

a T3G ) triethylene glycol; T4G ) tetraethylene glycol; PEG ) poly(ethylene glycol); PAA ) poly(acrylic acid); and PEI ) poly(ethyleneimine). b SSE ) sum of square error in fitted yi. c |err ymax| ) maximum absolute fit residual.

Figure 9. xy diagram for ethanol/water with 15 wt % polyethyl-eneimine.

yi ) xiγiPisat/P (2)

Figure 10. Txy diagram for ethanol/water with and without 15wt % polyethyleneimine.

Ind. Eng. Chem. Res., Vol. 39, No. 10, 2000 3905

Nomenclature

Psat ) Saturation vapor pressureT ) Temperaturexi ) Mole fraction of component i in liquid phaseyi ) Mole fraction of component i in vapor phaseaij ) binary interaction parameterγi ) Activity coefficient of component i

Literature Cited

(1) Fair, J. R. Distillation: Whither, not whether. Inst. Chem.Eng. Symp. Ser. 1987, 104, 613.

(2) Seader, J. D.; Henley, E. J. Separation Process Principles;John Wiley & Sons: New York, 1998.

(3) Laroche, L.; Bekiaris, N.; Andersen, H. W.; Morari, M.Homogeneous Azeotropic Distillation: Comparing Entrainers.Can. J. Chem. Eng. 1991, 69, 1302.

(4) Widagdo, S.; Seider, W. D. Azeotropic Distillation. AlChEJ. 1996, 42(1), 96.

(5) Bekiaris, N.; Morari, M. Multiple Steady State in Distilla-tion -∞/∞ Predictions, Extensions and Implications for De-sign, Synthesis and Simulation. Ind. Eng. Chem. Res. 1996, 35,4264.

(6) Guttinger, T. E.; Morari, M. Multiple Steady States inHomogeneous Separation Sequences. Ind Eng. Chem. Res. 1996,35, 4597.

(7) Safrit, B. T.; Westerberg, A. W. Improved OperationalPolicies for Batch Extractive Distillation Columns. Ind. Eng.Chem. Res. 1997, 36, 436.

(8) Foucher, E.; Doherty, M.; Malone, M. Automatic Screeningof Entrainers for Homogeneous Azeotropic Distillation. Ind. Eng.Chem. Res. 1991, 30, 760.

(9) Alm, K.; Ciprian, M. Vapor Pressures, Refractive Index at20.0 °C, and Vapor-Liquid Equilibrium at 101.325 kPa in Methyltert-Butyl Ether-Methanol System. J. Chem. Eng. Data 1980, 25,100.

(10) Al-Amer, A. M.; Hamad, E. Z. Investigating PolymericEntrainers for Azeotropic Distillation of MTBE/Methanol andEthanol/Water Systems. Final Report, Project CHE/SABIC/96-4,Deanship of Research and Graduate Studies, King Fahd Univer-sity of Petroleum & Minerals, Saudi Arabia, 1999.

(11) Seader, J. D.; Zdzislaw, M. K. Distillation. In ChemicalEngineer’s Handbook; Perry, R. B., Green, D., Eds.; McGraw-Hill: New York, 1984.

Received for review January 11, 2000Accepted July 22, 2000

IE0000515

3906 Ind. Eng. Chem. Res., Vol. 39, No. 10, 2000