Ind. Eng. Chem. Res. 1997, 36, 803-807
803
Solvent Selection for Cyclohexane-Cyclohexene-Benzene Separation by Extractive Distillation Using Non-Steady-State Gas Chromatography Aurelio Vega,* Fernando Dı´ez, Ricardo Esteban, and Jose´ Coca Department of Chemical and Environmental Engineering, University of Oviedo, 33071 Oviedo, Spain
The infinite-dilution activity coefficients of cyclohexane, cyclohexene, and benzene in N,Ndimethylformamide, N-methylpyrrolidone, N,N-dimethylacetamide, phenyl acetate, and dimethyl malonate have been determined at temperatures ranging from 40 to 80 °C, by non-steady-state gas chromatography. From these data, the limiting selectivity-solvency properties for cyclohexane-benzene, cyclohexene-benzene, and cyclohexane-cyclohexene, in the presence of the aforementioned solvents, are studied, and the solvents tested are considered for the cyclohexanecyclohexene-benzene separation by extractive distillation. According to the results, N,Ndimethylacetamide seems to be an adequate solvent for the cyclohexane-benzene and cyclohexene-benzene separations. The separation of cyclohexane-cyclohexene is the most difficult, in spite of the difference of boiling points, much higher than for cyclohexane-benzene. Introduction The chemical market for high-purity products has generated a need for improvement in both extraction and extractive distillation separations and also has stimulated the search for better solvents. Extractive distillation is an important separation process, used to separate components with close boiling points, and is often considered as an alternative to conventional distillation. In extractive distillation, a high boiling point solvent is introduced into the column in order to produce an alteration of the key component’s relative volatility. Owing to large deviations from ideality of the extractive distillation systems, the selection of a proper solvent cannot be made, in general, from pure-component properties only. In certain cases the selection can be guided by correlations (Tassios, 1969, 1972), but because of limitations in accuracy and applicability, experimentation is often necessary. The relative volatility of any two components i and j in a multicomponent mixture, aij, can be defined as
φsat i φjγipi°POYi
(yi/xi) Rij )
(yj/xj)
)
φsat j φiγjpj°POYj
(1)
where y and x are the mole fractions of the components in the vapor and liquid phases, respectively, γ is the activity coefficient, φ is the fugacity coefficient in the vapor phase, φsat is the fugacity coefficient at saturation, POY is the Poynting correction factor, p° is the vapor pressure, and subscripts correspond to components i and j, respectively. At low or moderate pressures the relative volatility is given by:
Rij )
γipi° γjpj°
(2)
In the presence of a solvent a new value of the relative volatility is obtained, and eq 2 can be rewritten as * Author to whom correspondence is addressed. E-mail:
[email protected]. S0888-5885(96)00426-5 CCC: $14.00
Rsij )
(ysi /xsi ) (ysj /xsj )
(ysi /pos i ) )
(ysj pos j )
(3)
where superscript s indicates in the presence of the solvent. As the ratio of the vapor pressures of the pure components is not usually significantly affected by the changes in the boiling point temperature at constant pressure due to the presence of the solvent, the influence of the solvent is usually quantified in terms of the socalled selectivity, Sij, which is defined as the ratio of the activity coefficients of the two key components in the presence of the solvent:
Sij ) γsi /γsj
(4)
As the activity coefficients depend on the phase composition and the solvent effect tends to increase as its concentration increases, it is a common practice to consider, at least in a preliminary solvent selection, the situation of infinite dilution. The selectivity at infinite dilution is defined as the ratio of the activity coefficients of both components at infinite dilution in the solvent s∞ S∞ij ) γs∞ i /γj
(5)
s∞ where γs∞ i and γj are the activity coefficients at infinite dilution of components i and j in the solvent, respectively. A number of methods have been reported for the experimental determination of activity coefficients. A standard procedure is differential ebulliometry, in which the boiling point difference between a solution and the solvent is measured as a function of composition. It is a reliable but time-consuming technique. Gas-liquid chromatography (GLC) has been used extensively to obtain the activity coefficients at infinite dilution (Littlewood et al., 1955; Everett and Stoddart, 1961; Conder and Young, 1979; Arnold et al., 1987; Ferreira et al., 1987; Vega and Coca, 1991), but a limitation of the technique is that the liquid phase (solvent) must have a low volatility in order to avoid the entrainment of the stationary phase. This problem can be overcome, at least to some extent, by a combination of GLC and
© 1997 American Chemical Society
804 Ind. Eng. Chem. Res., Vol. 36, No. 3, 1997
liquid-liquid chromatography (LLC), using a highmolecular-weight compound as the liquid phase in both techniques (Locke, 1968; Vega and Coca, 1990). These chromatographic techniques have recently been reviewed by Vega and Coca (1993). Another method is the non-steady-state gas chromatography (NSGC), in which the solvent is itself being eluted from the column during the chromatographic experiment. NSGC has been used to determine activity coefficients of solutes in volatile solvents (Belfer and Locke, 1984; Belfer et al., 1990; Landau et al., 1991; Dallinga et al., 1993). Benzene, cyclohexane, and cyclohexene are hydrocarbons present in different streams in the petrochemical industry, e.g., in the products of partial catalytic hydrogenation of benzene. In order to separate the mixture into its components, an efficient separation method, such as extractive distillation, is needed because of their close boiling points (benzene, 80.10 °C; cyclohexane, 80.74 °C; cyclohexene, 82.98 °C). Different solvents have been used for the separation of aromatics from these types of streams: sulfones (Fu-Ming, 1986; Fu-Ming and Coombs, 1987, 1988), morpholines (Preusser et al., 1971; Lackner and Emmrich, 1971), amides (Mikitenko et al., 1973; Berg, 1991), esters (Mori and Moriya, 1987; Kodama and Yamashita, 1989), pyrrolidones (Ueno and Minaga, 1977; Lee et al., 1991; Brown and Lee, 1990), etc. The aim of this work was to evaluate different solvents for extractive distillation of the aforementioned mixture of hydrocarbons using the NSGC technique to determine activity coefficient and selectivities at infinite dilution. The five solvents studied in this work were N,N-dimethylformamide (DMF), N-methylpyrrolidone (NMP), N,N-dimethylacetamide (DMAC), phenyl acetate (PA) and dimethyl malonate (DMM). Theoretical Background In NSGC, the column either is initially packed with bare GC solid support material or is an uncoated fused silica capillary column. The liquid solvent is injected into the heated injection port of a gas chromatograph. The vapors are transported into the column by the carrier gas, and if the column temperature is lower than the solvent boiling point, the liquid solvent condenses in a more or less uniform film on the packing material or onto the walls of the capillary column. Once the solvent has equilibrated in the column and the excess vapors have been eluted, the gas chromatograph detector stabilizes and gives a high, flat baseline, recording the plateau of the eluting solvent vapors. As carrier gas flows through the column, it becomes saturated with solvent vapor. Solvent is thus steadily depleted from the wetted packing or walls of the capillary, starting at the inlet end of the column. Once a steady baseline is obtained, small amounts of a solute more volatile than the solvent are injected at small intervals. As the total weight of solvent in the column decreases with time, the retention volume of an injected solute decreases over the lifetime of the column. An analysis of the system (Belfer and Locke, 1984) leads to a simple equation relating the infinite-dilution activity coefficient of the solute (i), γs∞ i , in the volatile solvent (s) to the ratio of their vapor pressures, and the rate at which retention time decreases with time of injection. This relationship can be derived taking into account that, assuming ideal behavior of the vapor phase and that the solutions are sufficiently diluted so
that Henry's law applies, the solute net retention volume, VN, in steady-state (conventional) gas chromatography is related to the limiting activity coefficient by the expression:
VN ) KiVs ) RTns/pi°γs∞ i
(6)
where Ki and pi° are the distribution coefficient and vapor pressure of the solute, Vs and ns are the volume and the number of moles of solvent in the column, and T is the column temperature. In NSGC, the decrease in retention volume is proportional to the decrease in the amount of solvent (Landau et al., 1991):
∆VN ) Ki∆Vs ) (RT/pi°γs∞ i )∆ns
(7)
At constant carrier gas flow rate, F, and constant column temperature, the loss of solvent is directly proportional to time. Over a time interval ∆t ) t1 - t2, where t1 and t2 are two different times of injections of solute, the loss of solvent is
-(∆ns/∆t) ) nvF ) ps°F/RT
(8)
where nv is the number of moles of volatilized solvent per unit volume of the gas phase. Combining eqs 6-8 and taking into account that for a constant carrier gas flow rate ∆VN/F ) ∆tR:
( )/( )
γis∞ ) -
ps° pi°
∆tR ∆t
(9)
Equation 9 indicates that in order to determine the limiting activity coefficients the data needed are the vapor pressures and the solute retention times measured at different injection times. Although, in principle, only two injections are needed, the larger the number of solute injections made, the better the precision of the experimental measurement. Experimental Section A Varian Star 3400 gas chromatograph equipped with a flame ionization detector and septum-equipped programmable injector was used with a Supelco 30 m × 0.25 mm i.d. uncoated and deactivated fused silica capillary column. Measurements were made at column temperatures in the range 40-80 °C, using helium as the carrier gas. Chromatographic data were collected in a Varian Star chromatography workstation. In order to load the column with a uniform and relatively thick film of solvent, the column temperature was set to a value slightly greater than the solvent boiling point. Condensation of the solvent onto the walls of the capillary column was achieved by decreasing slowly the column temperature to the desired operating value and using a very low carrier gas flow rate. As soon as the baseline stabilization was observed, solutes were injected at different times without waiting for the complete elution of the sample before making a new injection. Indeed, the more solute injections made over the life of the column, the better the definition of the slope of the retention time vs injection time plot. The complete evaporation of solvent is detected by a sharp drop in the baseline, which indicates the need for a new fresh solvent injection. Solutes (benzene, cyclohexane, and cyclohexene) were reagent grade or HPLC grade from Aldrich (Aldrich
Ind. Eng. Chem. Res., Vol. 36, No. 3, 1997 805 Table 1. Activity Coefficients at Infinite Dilution for Cyclohexane, Cyclohexene, and Benzene Solutes in Several Solvents and Temperatures benzene temp (°C)
cyclohexene temp (°C)
cyclohexane temp (°C)
solventa
40
55
70
80
40
55
70
80
40
55
70
80
DMF NMP DMAC DMM PA
2.05 1.67 1.18 2.11 1.29
1.72 1.57 0.94 1.77 1.15
1.57 1.44 0.86 1.51 1.10
1.48 1.32 0.90 1.44 1.08
8.16 6.75 5.33 9.10 3.55
5.86 5.11 3.52 6.27 2.66
4.80 4.13 2.65 4.80 2.05
4.34 3.70 2.65 4.08 1.96
26.5 12.3 7.08 31.1 4.74
11.6 8.83 5.17 14.9 3.59
9.53 7.32 4.17 8.51 3.03
8.27 5.87 4.43 6.29 2.90
a DMF ) N,N-dimethylformamide. NMP ) N-methylpyrrolidone. DMAC ) dimethylacetamide. DMM ) dimethyl malonate. PA ) phenyl acetate.
Chemie AG, Steinheim, Germany) or Fluka (Fluka Chemie AG, Buchs, Switzerland) and a minimum purity of 99.5%. The solvents used were also reagent grade from Aldrich or Riedel-de Hae¨n (Riedel-de Hae¨n AG, Seelze, Germany) and a minimum purity of 99.0%. Results and Discussion The activity coefficient of the solute i at infinite dilution in the solvent s, γs∞ i , can be calculated with eq 9 from the slope of a plot of retention time vs injection time and the ratio of solute-to-solvent vapor pressures. Vapor pressures for all compounds were calculated from Solvents Guide (1963) data fitted to the Antoine equation. In order to test the experimental technique, the n-hexane activity coefficient at infinite dilution in toluene at 40 °C was determined. The value of 1.64 obtained agrees very well with the value of 1.62 obtained by Landau et al. (1991). Activity coefficients at infinite dilution obtained in this work are listed in Table 1 for cyclohexane, cyclohexene, and benzene as solutes in the solvents tested at different temperatures. An efficient extractive distillation solvent must have high selectivity but also sufficient capacity or solvency, so that it results in a good performance in the distillation process. The pattern of selectivity-solvency can aid in the reliable choice of solvents having a superior balance of these properties, although other factors, such as solvent cost and availability, solvent thermal and chemical stability, corrosivity, favorable physical properties, toxicity, etc., must be taken into account. The infinite-dilution activity coefficient for a given solute can be used to characterize the solvent capacity or solvency (Deal and Derr, 1964). In the case of very high activity coefficients at infinite dilution, the mole fraction solubility xi, is given by
xi ) 1/γi
Figure 1. Selectivity-solvency for cyclohexane-benzene.
Figure 2. Selectivity-solvency for cyclohexene-benzene.
(10)
and can be used to approximate solubilities out to several mole percent. At lower γi and higher solubilities, the relationship between γi and concentration is less straightforward but can be estimated in terms of the Van Laar relationships by defining changes in γi with concentration. Usually, a γi less than 10 corresponds to a hydrocarbon solubility of over 20 mol %. Deal and Derr (1964) proposed a logarithmic plot of selectivity against the solvency parameter, γi, as a basis for indicating limits in the selectivity to be expected at a given desired solvency and for showing how such selectivity varies with solvency. This information is useful in selecting solvents for either extractive distillation or extraction. In extractive distillation processes, it is required that the solvent has a relatively high solvency for the
components which concentrate in the upper section of the distillation column. For common solvent levels of 60-70 mol %, the usable limiting activity coefficient can be set in the 8-10 region (Deal and Derr, 1964); however, no exact limits can be fixed because of complicated interrelations among feed, recovery-purity requirements, reflux, and equilibrium stages available, plus the selectivity of the solvents itself. Figures 1-3 show the selectivity-solvency logarithmic plots for cyclohexane-benzene, cyclohexene-benzene, and cyclohexane-cyclohexene, respectively. Each segment corresponds to the selectivity-solvency properties of a particular solvent over the 40-80 °C range, with values at 80 °C being given by the left terminals of the segments (low value of the limiting activity coefficient) and those at 40 °C for the right or high value of limiting activity coefficient terminals.
806 Ind. Eng. Chem. Res., Vol. 36, No. 3, 1997 n ) number of moles R ) gas constant (atm‚L/(K‚mol)) S ) selectivity t ) time (min) T ) absolute temperature (K) p° ) vapor pressure (atm) POY ) Poynting factor V ) retention volume (L) x ) mole fraction in the liquid phase y ) mole fraction in the vapor phase Greek Letters R ) relative volatility γ ) activity coefficient φ ) fugacity coefficient Subscripts Figure 3. Selectivity-solvency for cyclohexane-cyclohexene.
Figures 1 and 2 show a similar behavior for cyclohexane-benzene and cyclohexene-benzene. All the solvents studied exhibit the expected increase in solvency (decrease in limiting activity coefficient) and decrease in selectivity with increasing temperature. It is interesting to note that the solvent’s relative position (selectivity at given solvency) is not altered with rather marked changes in temperature. It can be observed that the selectivity-solvency logarithmic plots for DMM, DMF, NMP, and PA lie on a common straight line, with PA yielding the highest capacity but lowest selectivity, followed, in order of decreasing capacity and increasing selectivity, by NMP, DMM, and DMF. The plot for DMAC is another straight line in a higher position. As far as the solvency-selectivity properties are concerned, the best solvents are fairly clearly those which lie along the upper left region of the plot. According to Figure 1, NMP, DMM, DMF, and DMAC are good solvents for the cyclohexane-benzene separation, with selectivities in the range 4.5-5.0 for a 80 °C operation and slightly lower, 2.7, for PA. DMAC yields the highest selectivity at a given solvency, so it would require a lower solvent level and circulation. It can be observed in Figure 2 that the trends for cyclohexene-benzene are the same as those for the cyclohexane-benzene separation, but selectivities are lower (about 3.0 for a 80 °C operation, except for PA, 1.8), while solvencies are higher. Again, the better solvent seems to be DMAC. Figure 3 shows the selectivity-solvency logarithmic plot for cyclohexane-cyclohexene. The selectivities for this system are lower than that in previous cases. DMM gives the highest selectivity at 40 °C (3.5) and DMF at 80 °C (1.9), while NMP, DMAC, and PA show very low selectivities (smaller than 2.0 in all the temperature range), with DMAC and PA presenting an anomalous behavior of increasing solvency and selectivity for increasing temperatures. According to these results, DMAC seems to be an adequate solvent for the cyclohexane-benzene and cyclohexene-benzene separations. The separation of cyclohexane-cyclohexene is the most difficult, in spite of the difference of boiling points (2.24 °C), much higher than for cyclohexane-benzene (0.64 °C). Nomenclature F ) carrier gas flow rate (L/min) K ) distribution coefficient
i ) component i j ) component j N ) net R ) retention s ) solvent Superscripts s ) in the presence of the solvent sat ) at saturation ∞ ) at infinite dilution
Literature Cited Arnold, D. W.; Greenkorn, R. A.; Chao, K. Infinite-Dilution Activity Coefficients for Alkanals, Alkanoates, Alkanes, and Alkanones in 4-Methyl-2-pentanone. J. Chem. Eng. Data 1987, 32 (1), 103. Belfer, A. I.; Locke, D. Non-Steady-State Gas Chromatography for Activity Coefficient Measurement. Anal. Chem. 1984, 56, 2485. Belfer, A. I.; Locke, D.; Landau, I. Non-Steady-State Gas Chromatography Using Capillary Columns. Anal. Chem. 1990, 62, 347. Berg, Ll. Separation of Cyclohexane from Cyclohexene by Azeotropic or Extractive Distillation. U.S. Patent 5,069,756, 1991. Brown, R. E.; Lee, F. M. Extractive Distillation of Cycloalkanes Employing a Mixed Solvent System. U.S. 4,954,224, 1990. Conder, J. R.; Young, C. L. Physicochemical Measurement by Gas Chromatography; Wiley: New York, 1979. Dallinga, L.; Schiller, M.; Gmehling, J. Measurement of Activity Coefficients at Infinite Dilution Using Differential Ebulliometry and Non-Steady-State Gas-Liquid Chromatography. J. Chem. Eng. Data 1993, 38, 147. Deal, C. H.; Derr, E. L. Selectivity and Solvency in Aromatics Recovery. Ind. Eng. Chem. Process Des. Dev. 1964, 3 (4), 394. Everett, D. H.; Stoddart, C. T. H. The Thermodynamics of Hydrocarbon Solutions from G. L. C. Measurement. Part 1. Solutions in Dinonyl Phthalate. Trans. Faraday Soc. 1961, 57, 746. Ferreira, P. O.; Bastos, J. C.; Medina, A. G. Infinite-Dilution Activity Coefficients for Aromatic and Nonaromatic Compounds in N-methylpyrrolidone, Ethylene Glycol, and Mixtures of the Two Solvents. J. Chem. Eng. Data 1987, 32, 25. Fu-Ming, L. Use of Organic Sulfones as the Extractive Distillation Solvent for Aromatics Recovery. Ind. Eng. Chem. Process Des. Dev. 1986, 25, 949. Fu-Ming, L.; Coombs, D. M. Two-Liquid-Phase Extractive Distillation for Aromatic Recovery. Ind. Eng. Chem. Res. 1987, 26, 564. Fu-Ming, L.; Coombs, D. M. Two-Liquid-Phase Extractive Distillation for Upgrading the Octane Number of the Catalytically Cracked Gasoline. Ind. Eng. Chem. Res. 1988, 27, 118. Kodama, S.; Yamashita, K. Separation of Cyclohexene by Mixed Solvents. JP 01,135,730, 1989. Lackner, K.; Emmrich, G. Higher Octanes, less Benzene. HighOctane Blending Stream low in Benzene is the Product of the Octenar Process. Pure Benzene is a Valuable Byproduct. Hydrocarbon Process. 1971, 14, 67. Landau, I.; Belfer, A. J.; Locke D. C. Measurement of Limiting Activity Coefficients Using Non-Steady-State Gas Chromatography. Ind. Eng. Chem. Res. 1991, 30, 1900.
Ind. Eng. Chem. Res., Vol. 36, No. 3, 1997 807 Lee, F. M.; Brown, R. E.; Johnson, M. M. Extractive Distillation of Hydrocarbons Mixtures. U. S. 5,032,232, 1991. Littlewood, A. B.; Phillips, C. S. G.; Price, D. T. The Chromatography of Gases and Vapors. Part V. Partition Analyses with Columns of Silicone 702 and Tritolyl Phosphate. J. Chem. Soc. 1955, 1480. Locke, D. Chromatographic Study of Solutions of Hydrocarbons in Acetonitrile. J. Chromatogr. 1968, 35, 24. Mikitenko, P.; Cohen, G.; Asselineau, L. Benzene and toluene purification processes by azeotropic-extractive distillation. Fr. Demande 2,176,488, 1973. Mori, Y.; Moriya, O. Separation of Monocyclic Hydrocarbons by Extractive Distillation. JP 62,123,135, 1987. Preusser, G.; Ritchter, K.; Schulze, M. Extractive Recovery of Highly Pure Aromatics from Hydrocarbon Mixtures Containing High Amounts of Nonaromatics. Ger. Offen. 2,013,298, 1970. Solvents Guide; Marsden, C., Ed.; Interscience: New York, 1963. Tassios, D. P. Choosing Solvents for Extractive Distillation. Chem. Eng. 1969, 10, 118. Tassios, D. P. Rapid Screening of Extractive Distillation Solvents. Adv. Chem. Ser. 1972, No. 115, 46.
Ueno, K.; Minaga, M. Separation and Purification of Cyclohexene. JP 77,05,733, 1977. Vega, A.; Coca, J. Activity Coefficients at Infinite Dilution of Organic Compounds in Acetonitrile and Methanol by Liquid Chromatography. J. Liq. Chromatogr. 1990, 13 (4), 789. Vega, A.; Coca, J. Activity Coefficients at Infinite Dilution Determined by Gas-Liquid Chromatography: Organic Solvents in Apiezon L. J. Chromatogr. 1991, 586, 303. Vega, A.; Coca, J. Physico-Chemical Measurements by Chromatography of Relevance in Separation Processes. In Trends in Chemical Engineering, Vol. 1; Menon, J., Alexander, J. C., Eds.; Research Trends: Trivandrum, 1993.
Received for review July 19, 1996 Revised manuscript received November 12, 1996 Accepted November 14, 1996X IE960426F X Abstract published in Advance ACS Abstracts, January 15, 1997.
