1206
Ind. Eng. Chem. Res. 1996, 35, 1206-1214
Recovery of Propylene Glycol from Dilute Aqueous Solutions by Complexation with Organoboronates in Ion-Pair Extractants Robert R. Broekhuis, Scott Lynn, and C. Judson King* Department of Chemical Engineering and Lawrence Berkeley Laboratory, University of California, Berkeley, California 94720
Recovery of propylene glycol (1,2-PD) from aqueous solution was studied in batch experiments using extractants consisting of ion pairs of Aliquat 336 and phenylboronate in 2-ethylhexanol, toluene, o-xylene, or diisobutyl ketone. The heterogeneous complexation constant β11 calculated from the results at 25 °C was highest in 2-ethylhexanol (49-100 (mol/L)-1). The equilibrium water concentration in the extractants was 8-12 wt % and decreased with increasing uptake of 1,2-PD. Nearly all extractant/diluent systems exhibited stoichiometric overloading. Evidence for aggregation of the ion-pair extractant in the organic phase was found from water solubilization studies and 1H NMR spectroscopy studies. The complexation constant decreased with increasing temperature. Up to 80% of the extracted 1,2-PD was backextracted into water after acidification with CO2. The extractant could then be regenerated by stripping CO2 from solution at temperatures exceeding 110 °C. However, at these temperatures the color of the extractant changes, and the extraction capacity is reduced to about 60% of its original value. Regeneration by contacting with aqueous solutions of Na2CO3 did not cause extractant degradation; regeneration effectiveness increased with increasing pH. Introduction Recovery of compounds bearing multiple hydroxyl groups from aqueous solutions is important in many process industries. Typical examples are recovery of fermentation chemicals from complex and dilute broths, removal of these chemicals from waste streams generated by a variety of industrial processes, and recovery of glycols from aqueous solutions in which they are produced as petrochemicals. Recovering such chemicals with low volatility and strongly hydrophilic characteristics from dilute aqueous solutions is a difficult separation problem, one that in many fermentation processes causes downstream processing to account for 40% or more of the final product cost. The heat load per amount of recovered chemical for multiple-effect evaporation of dilute solutions is high, since all the water must be evaporated. Also, the chemical of interest is often just one of many nonvolatile chemicals in the solution; fractionation among these can be difficult and inefficient. To be more economical, separation methods need to capitalize on specific properties of the chemicals to be recovered to remove them selectively and efficiently from aqueous solution. Chemicals of interest include glycerol and diols (which have boiling points well above that of water) and saccharides (which are nonvolatile). The research described here focuses on propylene glycol (1,2-propanediol, 1,2-PD), which is a large-scale petrochemical and a potential commercial fermentation chemical. Propylene glycol is a petrochemical produced in about a 20% aqueous solution from propylene oxide. It can also be formed by fermentation of glucose using the bacterium Clostridium thermosaccharolyticum (Cameron and Cooney, 1986) at a concentration of about 0.8% by weight in the fermentation broth. Other glycols that can be produced by fermentation include 1,3-propanediol, using glycerol as a substrate (Cameron and Cooney, 1986), and 2,3-butylene glycol from glucose (Senkus, 1946). Broekhuis et al. (1994) investigated separation processes based on the reversible chemical reaction of 0888-5885/96/2635-1206$12.00/0
glycols with aldehydes. The research detailed here investigates an alternative separation scheme, based on the complexation of cis-vicinal diols with organoboronates. The complexation reaction between phenylboronate and 1,2-PD is shown below:
Shinbo et al. (1986) transported monosaccharides up a concentration gradient through an organic liquid membrane containing phenylboronate. Randel et al. (1994) prepared ion-pair extractants containing (3nitrophenyl)boronate, with the counterion provided by Aliquat 336 (trialkylmethylammonium cation, the hydrocarbon chains being predominantly octyl and decyl). They found that such extractants complex strongly with many multiple-hydroxyl compounds, including saccharides, propylene glycol, and glycerol. After extraction of such compounds into organoboronate solutions, the complexation reaction must be reversed to effect backextraction. Since the complexation reaction occurs only with the organoboronate anion, protonation of organoboronate to the corresponding organoboronic acid results in the release of 1,2-PD from the complex. Such a pH shift can be achieved by contacting the extractant with strong aqueous acids or, preferably, with a recoverable substance such as carbon dioxide, since with CO2 no acids and bases are consumed in the process and no salts are formed as byproducts. Carbon dioxide would need to be stripped from the extractant solution or extracted into a high-pH aqueous solution to regenerate the extractant for recycle. The research described here continues the work of Randel et al. (1994). Phenylboronic acid (HPB) was used as a less costly, albeit more water-soluble, alternative to (3-nitrophenyl)boronic acid. Extraction equilibria were studied in more qualitative and quantitative detail, and backextraction and extractant regeneration proce© 1996 American Chemical Society
Ind. Eng. Chem. Res., Vol. 35, No. 4, 1996 1207 Table 1. Extractant Characteristics and Complexation Parametersa symbol
a
diluent
(A366+PB-) (mol/L)
(A366+OH-) (mol/L)
(A366+Cl-) (mol/L)
β11 (mol/L)-1
β21 (mol/L)-2
75.1 99.6 49.6 36.5 54.4 34.8 37.6 14.4
32.8 62.2 18.3 6.0 (60 °C) 22.2 8.4 23.7 -0.3
T V X
2-ethylhexanol 2-ethylhexanol 2-ethylhexanol
0.284 0.252 0.212
0.069 0.069 0.044
0.201 0.228 0.194
XX TOL XYL DBK
2-ethylhexanol toluene o-xylene diisobutyl ketone
0.042 0.321 0.295 0.353
0.009 0.034 0.050
0.038 0.197 0.188 0.233
Values for β11 and β21 were determined at 25 °C, unless otherwise noted.
dures were explored. Broekhuis et al. (1995) give a more detailed account of experimental procedures and results. Experimental Procedures Materials. Phenylboronic acid (97%), Aliquat 336 in the chloride form, and 1,2-propanediol (99%) were obtained from Aldrich Chemical Co. Elemental analysis and titration of the phenylboronic acid indicated a sufficiently high purity to allow use of the formula weight (121.93) directly in calculations. Elemental analysis of the Aliquat 336 in combination with KarlFischer titration yielded an effective molar weight of 483. Analyses. Elemental analyses for C, H, and N were made by the Microanalytical Laboratory of the University of California, Berkeley, College of Chemistry. Analyses for B and Cl were carried out by Desert Analytics, Inc. (Tucson, AZ) using the Schoniger flask method and flame atomic absorption spectroscopy, respectively. Propylene glycol in aqueous phases and some organic phases was analyzed by gas chromatography. Aqueous-phase concentration measurements included acid-base titration with hydrochloric acid for phenylboronate and hydroxide, UV spectroscopy for phenylboronate, and Mohr’s titration with silver nitrate for chloride. Concentrations of water in organic phases were measured by Karl-Fischer titration. Preparation of Extractants. HPB and Aliquat 336 (A336+Cl-) were dissolved in the diluent (2-ethylhexanol, toluene, o-xylene, or diisobutyl ketone) and then were washed two or three times with NaOH solutions of a molarity close to that of Aliquat 336 in the organic phase and finally with pure water. Anions exchange between the aqueous and extractant phases according to the following reaction (organic-phase compounds are shown with an overbar).
A336+Cl- + HPB + OH- f A336+PB- + Cl- (2) Aqueous washes were carried out in separatory funnels, with manual shaking, followed by centrifugation to effect phase separation. Some phenylboronate is taken up into the sodium hydroxide solutions. The wash solutions were analyzed for chloride, phenylboronate, and hydroxide. The concentrations of phenylboronate (PB-) and chloride in the extractant after the washes were determined by mass balance. It was assumed that when the amount of chloride removed from the extractant exceeded the amount of phenylboronate remaining in the extractant, all phenylboronate was present in the anionic form, paired with A336+. Most of the extractions were carried out with extractants ranging in concentration from 0.2 to 0.35 mol of PB-/L. The densities and water concentrations of each of the extractants were also determined. Table 1
summarizes the characteristics of the extractants discussed in this paper. In an alternative extractant preparation method, Aliquat 336 is added to a solution of sodium phenylboronate in methanol. Sodium chloride precipitates, driven by its low solubility in methanol, leaving a solution of Aliquat 336 phenylboronate. Methanol is replaced by a suitable (non-water-miscible) diluent by distillation and aqueous washes. Phenylboronate losses with this method are lower, but it is more difficult to determine the concentration of PB-, which prompted the use of the former method for quantitative extraction studies. Extraction Studies. Measured amounts of extractant solutions and standard aqueous 1,2-PD solutions were contacted in a temperature-controlled shaker bath for at least 16 h. In a separate extraction kinetics experiment (Broekhuis et al., 1995), it was determined that extraction equilibrium is attained in 6 h under the conditions of these experiments. The final concentrations of 1,2-PD in aqueous solution were measured, as well as the water concentration in the extractant phase. The concentration of 1,2-PD in the extractant phase was determined by mass balance, taking into account the changes in phase volumes due to the transfer of water and 1,2-PD. Measurement of the Distribution Ratio for 1,2PD. The distribution ratios of 1,2-PD between an aqueous phase and several organic phases were determined by similar extraction studies. Organic phases included pure 2-ethylhexanol, for which 1,2-PD was measured directly by GC, and solutions of Aliquat 336 (chloride form) in several diluents, for which 1,2-PD was determined by mass balance. Backextraction by Sparging with CO2. Measured amounts of water and a loaded extractant were dispersed by stirring and sparged with water-saturated CO2 at atmospheric pressure and ambient temperature. The uptake of CO2 into the extractant was determined by the increase in mass during the experiment. The degree to which 1,2-PD was released from its complex with PB- was calculated from the concentration of 1,2PD in the aqueous phase after the backextraction. The sparging was stopped when no further weight increase was observed, typically after about 2 h. In a separate experiment the kinetics of CO2 uptake and 1,2-PD release were measured. The two processes were found to occur in parallel, with no apparent lag time between the former and the latter. Extractant Regeneration by Stripping. Nitrogen was sparged through a measured amount of extractant in a vessel submerged in an ethylene glycol temperature bath. The temperature of the glycol bath was varied between 80 °C and 130 °C, and the duration of the sparging ranged from 13 min to 3 h. In routine regeneration experiments, the temperature was 115 °C,
1208
Ind. Eng. Chem. Res., Vol. 35, No. 4, 1996
Table 2. Partition Coefficients and Distribution Ratios of 1,2-PD between Aqueous and Organic Phasesa diluent 2-ethylhexanol
toluene o-xylene diisobutyl ketone a
P 0.078 0.078 0.121 (50 °C)
0.0056
(A366+Cl-) (vol %)
D
30.6 23.0
0.105 0.103
23.0 26.9 26.8 26.8
0.181 (60 °C) 0.095 0.112 0.124
Unless otherwise stated, data were obtained at 25 °C.
and the duration about 1 h. The effectiveness of the regeneration was determined by reloading the extractant with 1,2-PD and determining its extraction capacity. In several experiments, the same batch of extractant was repeatedly cycled through loading, backextraction, and regeneration steps. Extractant Regeneration with Na2CO3 Solutions. Sodium carbonate solutions were prepared using analytical-grade Na2CO3‚H2O. Measured quantities of carbonated extractant solutions and sodium carbonate solutions were contacted on a shaker bath at 25 °C for a minimum of 16 h. The final pH of the aqueous phase was measured using a pH probe. The effectiveness of the regeneration was determined by reloading the extractant with 1,2-PD and determining its extraction capacity. In one experiment the extractant was regenerated in several batch stages; the regeneration effectiveness was determined after each stage. Results and Discussion Distribution Ratio for 1,2-PD. The uptake of 1,2PD from an aqueous phase into an organic phase may be attributed to both physical distribution of 1,2-PD between the two phases and chemical complexation of 1,2-PD within the extractant phase. Physical distribution may be approximated by a simple proportionality between the concentrations in both phases, where the constant of proportionality is the partition coefficient P (when both phases are otherwise pure solvents) or the distribution ratio D (when one or both phases are mixtures). Both are defined as the ratio of the organicphase concentration and the aqueous-phase concentration at equilibrium. The partition coefficient for 1,2PD between water and 2-ethylhexanol was determined to be 0.078 ( 0.004. However, since extractants are concentrated solutions of A336+PB- and A336+Cl-, the pure diluent does not closely resemble actual extractants. Therefore, distribution ratios were measured between aqueous solutions and solutions of A336+Clin several diluents, at concentrations comparable to those of the extractants. It was assumed that the distribution ratio is similar into solutions of A336+Cland solutions of A336+PB-. To verify this assumption, the distribution of ethanol between aqueous and organic phases containing A336+ Cl- and A336+ PB- was measured. Ethanol, which would be expected to behave similarly to 1,2-PD except for its inability to participate in the complexation reaction, did not appear to be taken up more strongly by either organic phase (Broekhuis, et al., 1995). For all diluents the distribution ratio was higher than the corresponding partition coefficient. Experimental values of D are given in Table 2. Complexation Calculations. The expected stoichiometry of complexation in a glycol-organoboronate system is 1:1, i.e., one molecule of 1,2-PD complexes with
one A336+PB- ion pair. Randel et al. (1994) observed no higher stoichiometry. The 1:1 complexation may be described by a heterogeneous complexation constant β11, defined as
β11 )
(PB-‚1,2-PD) (PB-)[1,2-PD]
(3)
in which (PB-‚1,2-PD) and (PB-) are the organic-phase concentrations of the complex and the uncomplexed phenylboronate, respectively, and [1,2-PD] is the aqueous-phase 1,2-PD concentration. Another useful quantity is the extractant loading Z, defined in eq 4.
Z)
(PB-‚1,2-PD) (PB-‚1,2-PD) + (PB-)
(4)
If each boronate anion can take up no more than one glycol molecule, then Z will take on values between 0 and 1. The expression in the denominator is the total organic-phase phenylboronate concentration, (PB-)tot. Equations 3 and 4 may be combined to find the relationship between Z and [1,2-PD].
Z)
β11[1,2-PD] 1 + β11[1,2-PD]
(5)
The extraction equilibrium is determined by measurements of [1,2-PD] before and after extraction. These concentrations are related through the mass balance in equation 6,
Vaq,0[1,2-PD]0 - Vaq[1,2-PD] ) VorgD[1,2-PD] + Vorg,0Z(PB-)tot (6) in which Vaq and Vorg are the volumes of the aqueous and organic phases, respectively. The subscript 0 denotes initial, preextraction quantities. In this equation D is the value for the physical distribution ratio determined for solutions of Aliquat 336. The simplification that may be made by assuming Vaq ) Vaq,0 and Vorg ) Vorg,0 is not strictly correct, since the transfer of 1,2-PD and water between the aqueous and organic phases changes the volumes of both phases, most notably at high extractant concentrations. The volume change was not measured directly but could be inferred from the change in the aqueous-phase 1,2-PD concentration and the organic-phase water concentration. In practice, the effects of the transfer of 1,2-PD into and water out of the organic phase cancel to a large extent so that the phase-volume correction is generally not an important one. The extractant loading Z may be calculated from the experimental extraction results and eq 6. The error in the calculation of Z results primarily from either error in the estimate of D, or error in the measurement of [1,2-PD]. An error analysis shows that both sources of error become important at high ratios of [1,2-PD] to (PB-)tot. To obtain precise results even at high [1,2PD], high extractant concentrations were used in nearly all the extraction studies (Table 1). Loading Curves. Batch extractions were carried out in series in which the aqueous-phase concentration was varied at constant extractant composition and temperature. Unless otherwise stated, all extraction experiments were carried out at 25 °C. The results of one such extraction series are shown in Figure 1, in the form of
Ind. Eng. Chem. Res., Vol. 35, No. 4, 1996 1209
Figure 1. Loading curve for extractant V.
a loading curve. The high slope at low 1,2-PD concentrations is indicative of a large complexation constant. When two extractants differ in complexation strength, the difference between their loading curves will be most pronounced at low [1,2-PD]. For this reason, when extractants are compared, the loading is plotted vs the logarithm of the concentration. Overloading. In Figure 1 the loading exceeds unity for concentrations greater than 0.15 mol/L, which is in conflict with the notion of simple 1:1 complexation stoichiometry. This overloading phenomenon was observed for all extractants with 2-ethylhexanol, toluene or o-xylene as a diluent (Figure 4). Note also that the loading does not appear to level off to a constant value but continues to increase with concentration. Arguments that incorrect values for (PB-) or D may have been used or that the assumption that A336+PBexhibits the same physical affinity toward 1,2-PD as A336+Cl- is invalid cannot credibly explain the high calculated values for Z (Broekhuis et al., 1995). One mode in which 1,2-PD could be taken up into the extractant in excess of stoichiometric loading would be by solubilization into reverse micelles or similar aggregate structures formed by amphiphilic Aliquat 336 ion pairs. The results of experiments to test for the presence of such structures are presented later. Alternatively, the A336+PB- ion pair may, in fact, exhibit overloading. After the first complexation step, additional 1,2-PD might be taken up either by further complexation with the remaining hydroxyl functionality of the boronate or by solvation around the complex. In the former case, it would be appropriate to define a second complexation constant, β21, to describe the equilibrium of 2:1 complexation. This results in a twoparameter model. The curves drawn in Figures 1-4 are fits of such a model to the experimental data. The fact that the curves appear to fit the data adequately does not provide proof for the hypothesis of 2:1 complexation. Other modes of interaction between the Aliquat 336/ phenylboronate ion pair and 1,2-PD, or any combination of such interactions, could probably explain the observed behavior as well. The distinction between these modes cannot be made from phenomenological data such as loading curves. However, it is convenient to fit the data with the two-parameter model, since it yields parameters β11, which is a good measure of the extraction affinity at low glycol concentrations, and β21, a measure
Figure 2. Loading curves for two extractant concentrations (extractants X and XX), at 25 °C.
Figure 3. Loading curves for extractant X at 25 and 60 °C.
of the degree to which overloading occurs. Fitted parameter values for several extractants are included in Table 1. Effect of Extractant Concentration. Extractant X (0.212 mol/L of PB-) was diluted with 2-ethylhexanol to prepare extractant XX (0.042 mol/L of PB-). The results of two extraction series, one with extractant X and the other with XX, are shown in Figure 2. The loading curves nearly overlap, and the calculated complexation constants differ from each other by less than the standard errors; i.e., no concentration dependence of extraction behavior was observed in 2-ethylhexanol as a diluent. Effect of Temperature. Figure 3 shows the results of series of extraction experiments at 25 and 60 °C. The complexation is somewhat stronger at 25 °C. From the fitted values of β11 at both temperatures (Table 1) the heat of complexation was calculated by the van’t Hoff equation, and found to be ca. 7 kJ/mol. Although these results indicate that there is a temperature effect, backextraction following a temperature increase could not provide enough driving force for such a process to achieve a significant concentrating effect. Effect of the Diluent. Results of series of extractions carried out with extractants in which toluene, o-xylene, and diisobutyl ketone replaced 2-ethylhexanol as the diluent are shown in Figure 4. The experimental
1210
Ind. Eng. Chem. Res., Vol. 35, No. 4, 1996
Figure 4. Loading curves for extractants prepared in four different diluents, at 25 °C.
results were processed in the same way as for the 2-ethylhexanol diluent, using the physical distribution ratios of propylene glycol given in Table 2. The parameter values (β11, β21) obtained from a curve fit are included in Table 1. The results indicate that complexation in the aromatic solvents is not as strong as in a 2-ethylhexanol environment and is even weaker in diisobutyl ketone. Overloading occurred in the aromatic solvents but was not observed in diisobutyl ketone. The smaller fitted value of β21 for the aromatic solvents may be an artifact of the strong interdependence of the two complexation parameters and thereby may not reflect actual observed differences in complexation behavior. Extractants prepared in diisobutyl ketone and o-xylene were found to undergo spontaneous chemical changes at room temperature over the course of several weeks, as evidenced by a change in aroma as well as a decrease in extraction capacity. Extractants with 2-ethylhexanol as diluent did not appear to degrade in any way, even when stored for considerable lengths of time. Apparently the Aliquat 336/phenylboronate ion pair is preferentially stabilized in 2-ethylhexanol solution, due to either the higher polarity or the hydrogen-bonding capability of the diluent. Water Content of Extractants. The water concentration of each extractant, after the final aqueous wash, was measured by Karl-Fischer titration. Additionally, the final water contents of the extractant phases were measured in batch extraction experiments. The water concentration was found to decrease as the 1,2-PD concentration increases. In Figure 5 the concentration of water is plotted against the concentration of 1,2-PD for a variety of extractants. Linear first-order regressions to the data are also shown. The slope of the lines suggests that for each mole of glycol taken up into the extractant 3-4 mol of water is displaced from 2-ethylhexanol-based extractants and 4-5 mol of water from extractants with aromatic diluents. The water concentration was also measured in the experiments to determine the physical distribution of 1,2-PD into A336+Cl-containing extractants. The trend is reversed in this case without boronate present, as shown in Figure 5. The same experiment was also carried out with just 2-ethylhexanol as the organic phase, with a similar result. These results suggest that the displacement of water from the extractant is due to the presence of phenylboronate in the organic phase. Either many more
Figure 5. Concentration of water in extractants vs concentration of 1,2-PD.
Figure 6. Molar solubilization ratio of water as a function of Aliquat 336 concentration in o-xylene solutions, for solutions containing only A336+Cl- and solutions containing A336+Cl- and A336+PB-.
water molecules are involved in the solvation of the uncomplexed Aliquat 336/phenylboronate ion pair than with its complex with propylene glycol or 1,2-PD physically displaces water when it complexes with phenylboronate, e.g., from the interior of an aggregate structure. Water concentration results for the lowconcentration extractant (XX) are also included in Figure 5. As evidenced by the near-zero slope of the line, the displacement of water by glycol as a result of extraction did not occur significantly in the lowconcentration extractant. Aggregation in the Extractant Phase. The concentration of water in the extractant phase far exceeds the solubility of water in the diluents. Expressed in terms of moles of water per mole of Aliquat 336, the equilibrium concentration of water in o-xylene extractants can be as much as 10 times the concentration of Aliquat 336 (Figure 6). Such high ratios cannot be explained by general solvation, especially since Aliquat 336 itself is largely hydrophobic. This fact suggests that the high concentration of water may be due to solubilization inside aggregates of the amphiphilic ion pairs formed by Aliquat 336. Additional evidence for such a phenomenon was found in a series of experiments in which the equilibrium water concentration was mea-
Ind. Eng. Chem. Res., Vol. 35, No. 4, 1996 1211
Figure 7. Molar solubilization ratio of water as a function of A336+Cl- concentration in 2-ethylhexanol solutions.
sured in solutions of A336+Cl- in o-xylene and similarly in dilutions of the o-xylene-based extractant. The results are shown in Figure 6 as the molar ratio of water to Aliquat 336 (the molar solubilization ratio, or MSR) vs the concentration of Aliquat 336. The MSR is corrected for physical dissolution in the xylene diluent. It is clear that the MSR for water increases with Aliquat 336 concentration, from a value close to unity at infinite dilution to values exceeding 10 above 0.5 mol/L of Aliquat 336 for the extractants containing phenylboronate. This dependence of MSR on concentration suggests that some form of aggregation plays a role in the solubilization. Extractants containing Aliquat 336 in the phenylboronate form exhibit a higher MSR than simple solutions of Aliquat 336/chloride. Possible explanations would be that when coupled with phenylboronate, Aliquat 336 can more readily aggregate or form larger aggregates, leading to more water solubilized in each aggregate structure. In addition, there appear to be discontinuities in the slopes, most notably for extractants containing phenylboronate, at around 0.08 mol/L of Aliquat 336, possibly indicating a change in aggregation behavior at that concentration. Similar experiments were carried out using an extractant with 2-ethylhexanol as diluent. The results are shown in Figure 7. Here, the MSR does not seem to depend on the concentration of Aliquat 336, instead remaining at a high level even at low concentrations. This is in agreement with the fact that in 2-ethylhexanol no concentration dependence of the loading curve was observed. However, the water concentration data (Figure 5) suggest that the mechanism of glycol complexation and solubilization is different at low extractant concentrations. Since 2-ethylhexanol is less hydrophobic than o-xylene, one would expect the tendency to form reverse micelles to be lower. On the other hand, a slightly amphiphilic character can be ascribed to 2-ethylhexanol itself, so that the diluent itself could perhaps play a role in determining the aggregate structure. This may also explain why 2-ethylhexanol-based extractants, while exhibiting a smaller MSR for water, exhibit about the same degree of overloading as the aromatic extractants. Neutron magnetic resonance (NMR) spectroscopy was also used to obtain information on the structures existing in extractant solutions. Chemical shifts of water hydrogens in several chemical environments were observed using 1H NMR on a Bruker AMX-400 instru-
ment. In benzene, these hydrogen nuclei absorbed at a chemical shift of 0.4 ppm. In solutions of Aliquat 336 in benzene, a peak around 4.2 ppm could be attributed to water. This is close to 4.6 ppm, the chemical shift observed for bulk water. In this state, the water molecules are part of a hydrogen-bonded structure. From these results it may be inferred that the water in extractant solutions more closely resembles bulk water and must therefore be present in clustered structures, such as reverse micelles. Backextraction by Acidification with Carbon Dioxide. Each batch sparging experiment for backextraction of 1,2-PD yielded two equilibrium resultssthe amount of CO2 taken up into the extractant solution and the amount of propylene glycol displaced into the aqueous phase. The amount of CO2 absorbed was found to be roughly equal to the amount of Aliquat 336 in the extractant that is not in the chloride form (i.e., either in the hydroxide or the boronate form). Most of the propylene glycol was displaced from its complex with boronate into the aqueous solution. This suggests that the following reactions take place:
A336+‚PB-‚1,2-PD + CO2 + H2O f A336+‚HCO3- + HPB + 1,2-PD (7) A336+‚OH- + CO2 f A336+‚HCO3-
(8)
The effectiveness of backextraction, calculated from the amount of 1,2-PD extracted into the aqueous phase, may be expressed as Zbe, the loading after backextraction. About 20 experiments were carried out with 2-ethylhexanol-based extractants. There was considerable scatter in the experimental results, due largely to the compounding and leveraging of experimental and analytical errors. The values of Zbe varied from -0.050.24, but most values were in the range 0.02-0.10. The fact that the calculation of Zbe depends on both extraction and backextraction calculations, with the experimental and analytical errors of both, explains some of the scatter. Also, inaccuracies due to evaporation of diluent and water from the extractant and aqueous phases during backextraction limited the time that CO2 was sparged through, possibly leading to nonequilibrium results in some cases. Most backextraction experiments were carried out with extractants made using 2-ethylhexanol as a diluent. With each of the other extractants (using toluene, o-xylene, and diisobutyl ketone as diluents) one or two experiments were carried out. The degree of backextraction was high in the aromatic diluents and somewhat lower in diisobutyl ketone. Regeneration of the Extractant by Stripping. Reactions 7 and 8 show that A336+OH- and A336+PBare converted to the bicarbonate form in the backextraction step. In order to reuse the extractant in an extraction mode, the Aliquat 336 must be reconverted to the phenylboronate form. Because the hydration of carbon dioxide is an exothermic reaction, reactions 7 and 8 may be reversed by an increase in temperature. This suggests regeneration of the extractant by stripping out the CO2 at elevated temperatures. Stripping with nitrogen at 80 °C was unsuccessful, as was an attempt to regenerate by the steam stripping effect from a boiling water phase underneath the extractant phase. In these cases, the degree to which the extractants reloaded upon extraction after regeneration was less than 30% of the original loading. To effect
1212
Ind. Eng. Chem. Res., Vol. 35, No. 4, 1996
Figure 8. Extractant loading capacity as a function of stripping temperature.
Figure 9. Extractant capacity as a function of the duration of stripping at 116 °C.
more complete regeneration, higher temperatures were used. After nitrogen stripping at temperatures between 110 and 120 °C, the degree of reloading was higher, up to 60%. Regeneration of the extractants with aromatic diluents was largely unsuccessful, degrees of reloading not exceeding 15%. In all experiments with temperatures above 110 °C the extractant changed color from a light yellowish brown to darker shades of brown; the higher the temperature, the darker was the resulting extractant color. To investigate whether the chemical change evident from the discoloration was accompanied by a decrease in extraction capacity, several batches of extractant were heated to temperatures ranging from 80-130 °C for 1 h, without having been contacted with 1,2-PD solution. Figure 8 shows the effect of this treatment on the resultant extraction capacity. In a similar experiment, the effect of the duration of the heat treatment was investigated by leaving samples at 116 °C for times ranging from 13 min to 3 h. Figure 9 shows the extraction capacities of these extractants as a function of duration of regeneration. From these experiments it may be concluded that there is no appreciable degeneration at temperatures below 100 °C and that the capacity of the extractant reaches a level of about 60% of the original capacity in about 2 h at 116 °C, after which no further degeneration is observed. Solutions of HPB, A336+Cl-, or A336+OH- in 2-ethylhexanol did not discolor with treatment at 117 °C.
Figure 10. Results of 51/2 cycles of extraction, backextraction, and regeneration. The plan and hatched bars show Zn - Zbe,n-1 and Zb - Zbe,n, respectively, and the symbols above represent the uptake of CO2 in the extractant during backextraction.
Apparently, only the combination of Aliquat 336 with a phenylboronate anion leads to the degeneration. Repeated Extraction/Backextraction/Regeneration Cycles. In several experiments, one batch of extractant was repeatedly cycled through the loading, backextraction, and regeneration steps. The uptake and release of propylene glycol was measured as usual. Figure 10 shows the results of one such cycle experiment. For each cycle, the changes in loading in the extraction and backextraction steps, Zn - Zbe,n-1 and Zn - Zbe,n, respectively, are shown as two bars, and the uptake of CO2 is shown as the connected symbols. Since the degree of loading for each step is calculated from the results of all previous steps, the error in the absolute value of Z is considerable after several cycles. However, the change in Z in each step, i.e., the uptake and release of propylene glycol, is determined directly from the corresponding concentration measurements. After the initial degradation of the extractant in the first regeneration step, no significant further deterioration of extraction performance is evident from these experiments. The uptake of CO2 in the backextraction steps follows a similar pattern: The CO2 uptake is high in the first cycle, drops considerably after the first regeneration step, but does not significantly decrease in subsequent cycles. The trend observed in Figure 10, with an increase in CO2 uptake after the third cycle, is peculiar to this cycled experiment and was not observed in other such experiments. The loss of extraction capacity after regeneration and the incomplete backextraction combine with the loss of overall efficiency due to the fact that the physical distribution of glycol is not influenced by the uptake of CO2. The result is that the extraction characteristics in the second and subsequent steps are significantly less attractive than what might be expected from the initial loading capacity. Only about 30% of the original extraction capacity of the extractant in Figure 10 is effectively used toward uptake and release of propylene glycol in the second and subsequent cycles. Extractant losses to the aqueous phase are also potentially important. In a UV analysis of a postextraction aqueous phase, the concentration of PB- was less than 10-3 mol/L. HPB, which is at this time not
Ind. Eng. Chem. Res., Vol. 35, No. 4, 1996 1213
Figure 11. Schematic of a process utilizing CO2 and K2CO3 for regeneration of the ion-pair extractant.
manufactured in bulk quantities, is likely to be an expensive chemical, and the cost of process losses of HPB should be a consideration in an overall process evaluation. Regeneration of the Extractant with Na2CO3 Solutions. An alternative to stripping as a method of removing carbon dioxide from the extractant solution could be contacting with a high-pH aqueous phase. Carbon dioxide can be taken up into a sodium carbonate solution by the following reaction:
extractant was subjected to several batch stages of regeneration, showed both the pH and the regeneration effectiveness increasing after each stage. The postextraction pH would need to exceed 11 for a high level of regeneration, which requires a very high carbonate-tobicarbonate ratio. To achieve such high effectiveness, a regenerative absorption process might be carried out countercurrently. Figure 11 shows a schematic of a process that might be used for regeneration, shown using K2CO3 rather than Na2 CO3.
A336+HCO3- + HPB + CO32- h
Conclusions
A336+PB- + 2HCO3- (9) On the basis of the difference in pKa between HPB and HCO3- (8.9 and 10.25, respectively) and the expectation that Aliquat 336 would exhibit a stronger affinity toward phenylboronate than toward bicarbonate due to the hydrophobic nature of the former, one would expect the regeneration effectiveness to be high, as long as a high pH is maintained in the aqueous phase. The advantage of this procedure for regeneration is that it could be operated at mild temperatures, avoiding heat degradation of Aliquat 336. A drawback of an absorptive regeneration procedure may be the coextraction of phenylboronic acid into the basic aqueous phase. In experiments to regenerate extractants by absorbing CO2 into aqueous sodium carbonate solutions the extractant was regenerated to an appreciable extent, the uptake in the reloading step corresponding to a change in loading value ∆Z of 0.32 to 0.64 (Broekhuis et al., 1995). The degree of reloading increases with increasing aqueous-phase pH (measured after regeneration). A staged experiment, in which a batch of carbonated
Extraction of propylene glycol from dilute aqueous solutions, which was previously observed by Randel et al. (1994) in extractants with (3-nitrophenyl)boronate, is also effective using phenylboronate. Most extractions were carried out with high-concentration extractants; no effect of extractant concentration on extraction results was observed. Because of superior extraction results and greater chemical stability, 2-ethylhexanol was found to be the most useful diluent of those studied for the ion-pair extractants. Overloading, the phenomenon in which more than one molecule of 1,2-PD is taken up for each mole of phenylboronate, was observed in nearly all extractants at concentrations exceeding 0.2 mol/L. Several possible explanations were offered to explain this phenomenon. The most likely explanation is the uptake of 1,2-PD into the interior aqueous-like phase of extractant aggregate structures, e.g., reverse micelles. With extractants prepared in 2-ethylhexanol the loading is 0.5 or higher at an aqueous concentration of 0.02 mol/L (0.15 wt %) or above and 0.33 or higher at 0.01 mol/L (0.08 wt %) or above. On the basis of such
1214
Ind. Eng. Chem. Res., Vol. 35, No. 4, 1996
strong complexation, it would be feasible to design a countercurrent extraction process for recovering propylene glycol from dilute aqueous streams. A high degree of concentration could be achieved using an acceptable organic-to-aqueous phase ratio. Backextraction of 1,2-PD into an aqueous phase can be achieved by sparging CO2 into a dispersion of water in a loaded extractant. Typically, around 80% of the extracted 1,2-PD can be backextracted by this procedure. Extractants must then be regenerated by stripping at temperatures around 115 °C, to release the bicarbonate anions in the form of CO2. At these elevated temperatures, extractants undergo some chemical change evidenced by a darkening of color and a loss of extraction capacity. The loss of capacity is limited to about 50% of the original capacity, but in combination with the capacity that remains unutilized due to incomplete backextraction, the effective extraction/backextraction capacity is reduced to about 30% of the initial value. In terms of a process for recovery of glycol from aqueous solutions, this would mean that more extractant would be needed to achieve the same degree of glycol recovery and that the overall concentrating effect between initial and final aqueous phases would be limited. Assuming that the extractant is effectively blocked by the reaction with CO2, the highest practical aqueous product concentration would be limited by the physical distribution of 1,2-PD into the organic phase. Extractant regeneration using aqueous sodium carbonate solutions is an effective alternative to stripping if a high enough carbonate-to-bicarbonate ratio can be maintained to keep the aqueous-phase pH above 11. Such conditions suggest operating a countercurrent contacting process for regeneration. Extractant degradation in the stripping step or incomplete regeneration in an extractive regeneration step poses a serious problem that limits the utility of the glycol recovery process. Finding optimum and effective conditions for extractant regeneration would be a priority in further research into a recovery process using organoboronate extractants. Acknowledgment This work was supported by the Chemical and Biochemical Technology Research (BCTR) Program, Advanced Industrial Concepts Division, Office of Industrial Processes, Asst. Secretary of Energy under Contract No. DE-AC03-76SF00098.
P ) partition coefficient T ) temperature [°C] V ) phase volume [mL] Z ) extractant loading Greek Letters β11 ) complexation constant for 1:1 complexation [(mol/ L)-1] β21 ) complexation constant for 2:1 complexation [(mol/ L)-2] Subscripts 0 ) initial aq ) aqueous phase be ) backextraction n, n - 1 ) cycle number org ) organic phase tot ) total Aqueous-phase concentrations are indicated by square brackets. Organic-phase concentrations are indicated by parentheses.
Literature Cited Arenson, D. R.; King, C. J. Separation of low molecular weight alcohols from dilute aqueous solutions by reversible chemical complexation. Report LBL-24944; Lawrence Berkeley Laboratory: Berkeley, CA, 1989. Broekhuis, R. R.; Lynn, S.; King, C. J. Recovery of propylene glycol from dilute aqueous solutions via reversible reaction with aldehydes. Ind. Eng. Chem. Res. 1994, 33, 3230. Broekhuis, R. R.; Lynn, S.; King, C. J. Recovery of propylene glycol from dilute aqueous solutions by reversible chemical complexation with organoboronates. Report LBL-36913, Lawrence Berkeley Laboratory: Berkeley, CA, March 1995. Cameron, D. C.; Cooney, C. L. A novel fermentation: The production of R-(-)-1,2-propanediol and acetol by clostridium thermosaccharolyticum. Biotechnology 1986, 4, 651. Randel, L. A.; Chow, T. K.-F.; King, C. J. Ion-pair extraction of multi-OH compounds by complexation with organoboronate. Solvent Extr. Ion Exch. 1994, 12, 765. Senkus, M. Recovery of 2,3-butanediol produced by fermentation. Ind. Eng. Chem. 1946, 38, 913. Shinbo, T.; Nishimura, K.; Yamaguchi, T.; Sugiura, M. Uphill transport of monosaccharides across an organic liquid membrane. J. Chem. Soc., Chem. Commun. 1986, 349.
Received for review August 14, 1995 Revised manuscript received February 6, 1996 Accepted February 8, 1996X IE950508O
Nomenclature D ) distribution ratio MSR ) molar solubilization ratio
X Abstract published in Advance ACS Abstracts, March 15, 1996.
