Ind. Eng. C h e m . R e s . 1994,33,3217-3223
Sorption and Extraction of Lactic and Succinic Acids at pH 1. Factors Governing Equilibria
3217
> PKal.
Lisa A. Tung?and C. Judson King' Department of Chemical Engineering and Lawrence Berkeley Laboratory, University of California, Berkeley, California 94720
Many fermentations to produce carboxylic acids operate most effectively a t pH above pKal of the acid product, under which conditions the acid is largely in the carboxylate form. One approach to acid recovery from such solutions is to use solid sorbents or liquid extractants that are strongly enough basic to provide substantial capacity even at moderately high values of pH. Data are presented for sorption of lactic and succinic acids by several commercially available basic polymeric sorbents. Performance at pH > PKal is a function of sorbent basicity, and apparent pKa or monomer PKa can be used to predict sorbent performance. Data are also presented for the extraction of the acids by two commercial amine extractants, Alamine 336 and Amberlite LA-2,in various diluents. The extractants sustain capacity to higher pH in diluents that stabilize the acid-amine complex. Secondary amines provide higher capacities than do tertiary amines in diluents t h a t solvate the additional proton. Competitive uptakes of sulfate, phosphate, and carboxylate were also measured.
Introduction Recovery of carboxylic acids from fermentation broths presents a challenging separation problem. The dilute, complex nature of these broths makes potentially selective recovery methods, such as adsorption and extraction, highly attractive. Furthermore, if these methods may be used i n situ in the fermenter or in external recycle loops, higher overall yields can be achieved when product inhibition normally occurs (Davison and Thompson, 1992; RoMer et al., 1991). An important aspect of the recovery concerns the pH of the fermentation broth. Many fermentations operate best, or work only at all, at values of solution pH greater than pK,l of the acid being produced. For example, lactic acid 3.86),is typically produced at pH 5-6 (Jain et al., 1989; Vickroy, 1985). Succinic acid, a diacid = 4.2; pKd = 5.6), is produced at pH 5-7 (Glassner and Datta, 1992; Kaneuchi e t al., 1988). Citric acid, a notable exception, is produced at low pH, typically less than 3.5 (Kirk-Othmer, 1979). For most acids, then, if the recovery method is to be used without pH change, it must function well at pH > pKa1. Important aqueous waste streams can also have this characteristic. The conventional method for recovering low-volatility carboxylic acids from fermentation broths involves formation of the calcium carboxylate. Lactic and citric acids have been recovered this way industrially (Vickroy, 1985; Kirk-Othmer, 1979). The major disadvantage of the calcium salt process, however, is that for every mole of carboxylic acid produced, 1mol each of an acid (e.g., sulfuric acid) and the calcium base are consumed, and 1 mol of waste calcium salt (e.g., calcium sulfate) is created. This calcium sulfate poses a major disposal problem, especially as environmental controls tighten. A more efficient process would avoid both the consumption of chemicals and the production of high-volume waste streams. The recovery method investigated in this research t Present address: Rohm and Haas Co., 727 Norristown Rd., Spring House, PA 19477. *To whom correspondence should be addressed a t the Department of Chemical Engineering.
0888-5885/94/2633-3217$04.50/0
focuses on the use of solid sorbents or liquid extractants that are strongly enough basic to sustain capacity to several pH units above the PKal of the carboxylic acid. The advantage of using such basic sorbents and extractants is that recovery occurs at the pH of the fermentation broth. Thus, the recovery process can be carried out directly in the fermenter, or it can occur in an external loop or in downstream processing without consumption of chemicals for pH adjustment. Strongly basic sorbents and extractants require correspondingly powerful regeneration methods. Regeneration is explored in part 2 of this series. The primary goals of the current research were to identify sorbents and extractants that sustain capacity to two pH units or more above pKal and t o understand the equilibria involved. A number of sorbents and extractants were investigated, and differences in performance a t pH > PKal were related to differences in basicity stemming from structural differences.
Experimental Section Materials. All aqueous solutions were prepared from distilled water which had been passed through a Milli-Q water purification system (Millipore Corp.). Lactic acid (Aldrich, 85+% in water, ACS reagent grade) was diluted with water to approximately 20 wt % and boiled under total reflux for at least 12 h to hydrolyze any lactic acid polymers. Succinic acid (Fluka, >99.5% assay) was used as received. The liquid extractants, Alamine 336 (Henkel Corp.) and Amberlite LA-2 (Rohm and Haas Co.), were used as received. The solid sorbents used and their manufacturers are listed in Table 1. All of the sorbents are macroporous. Bio-Rad AG3-X4 was used without further purification. Note from Table l that Bio-Rad AG3X4 does contain some quaternized groups. The other unquaternized, or free-base, resins (Dowex MWA-1, Reillex 425, Duolite A7, and Amberlite IRA-35) were washed repeatedly with water, methanol, aqueous NaOH, and aqueous HC1, then further purified by Soxhlet extraction with methanol for 24 h. They were then dried t o constant weight in a vacuum oven at 60 "C and 50-60 kPa. With the exception of Reillex 425, 0 1994 American Chemical Society
3218 Ind. Eng. Chem. Res., Vol. 33, No. 12, 1994 Table 1. Basic Sorbents Usedo sorbent Bio-Rad AG3-X4 (Dowex XUS40091) Dowex MWA-1 Reillex 425 Duolite A7 Amberlite IRA910 Amberlite IRA-35 a
manufacture Dow Chemical Co.
matrix polystyrene-divinylbenzene
Dow Chemical Co. Reilly Ind., Inc. Rohm & Haas Corp. Rohm & Haas Corp. Rohm & Haas Corp.
polystyrene- divinylbenzene poly(4-vinylpyridine) phenol-formaldehyde-polyamine polystyrene-divinylbenzene acrylic
functional groups 90% 3" amine/ 10% 4" ammonium 3" amine pyridine polyamine 4" ammonium (type 11) 3" amine
mesh size 20-50
20-50 18-50 16-50 16-50 16-50
Sources: Reilly Ind., 1989; Rohm & Haas Co., 1981; Dow Chemical Co., 1983.
the free base resins were used initially in the dry form. Reillex 425 proved difficult to wet with water and was therefore prewet by stirring in methanol under vacuum and then displacing the methanol with water. Amberlite IRA-910 was washed as above and then converted to hydroxide form by contacting with 1 N NaOH in a column until no chloride was detected in the effluent. The resin was then rinsed with water until the pH of the effluent was less than 9 and was used in the water-wet form. The wet resin was centrifuged before use to remove interstitial water. Small samples of the water-wet resins were dried to obtain the moisture content gravimetrically. Methods. Sorption isotherms were generated by contacting known weights of sorbent (typically 1g) and acid solution (typically 10 g) in 20-mL scintillation vials sealed with foil-lined caps. The pH of the solution was not adjusted, and buffering agents were not used. The vials were placed on a shaker bath at 25 "C and 80 rpm for approximately 48 h. Studies with lactic acid and Dowex MWA-1 showed that equilibrium was reached within the experimental error within 1 h. Aqueousphase acid concentrations were determined by potentiometric titration with standardized NaOH or HCl(O.01 or 0.10 N). For measurements of the effect of pH upon sorbent capacity, 0.45 M aqueous solutions, corresponding to 4.0% and 5.3% (w/w) lactic acid or succinic acid, respectively, were adjusted to a range of pH values using aqueous NaOH. Eleven grams of each solution were contacted with 1.0 g of dry sorbent in a sealed vial which was placed on a shaker bath at 25 "C for 48 h. After equilibration, the pH of the solution was measured. The aqueous-phase acid concentration was determined by high-performanceliquid chromatography (HPLC)using a Shodex DE-613 polymethacrylate column, a 0.01 N H2S04 mobile phase, and a refractive index detector. In other experiments, aqueous solutions containing lactic acid and sodium sulfate or sodium phosphate were contacted with sorbent in sealed vials, as in the sorption isotherm experiments. Initial and final concentrations of lactic acid were determined via HPLC, and concentrations of sulfate and phosphate were determined by ion chromatography using a Dionex DX 300 gradient chromatograph equipped with an AS4A column and an aqueous Na2COaaHC03 mobile phase. The extraction experiments were analogous to the sorption experiments. Aqueous solutions of 0.45 M lactic or succinic acid were adjusted to various pH values using aqueous NaOH. The solutions were then contacted with equal volumes of 0.3 M extractant in the appropriate diluent in 40-mL vials equipped with Teflon-lined caps. (Molarities are stated per liter of total solution.) The vials were placed on a shaker bath at 25 "C and 80 rpm for approximately 48 h. After equilibration, the phases were separated by centrifugation for 8 min a t 2000 rpm. There were no detectable changes in the volumes of the phases, and no major phase separation problems were encountered. Aqueous-phase acid
cn,cnoncn,. I
cn,monc&. I
N
... .cn,cnoncn,- N. cn,cnoncn,.
.cnponcn, .N - ... N.
Ic - 0 I
.I.
cn,cnoncn,.
(4
... cn .cn, ..,
...
-. 'I
... (0
Figure 1. Chemical structures of basic polymeric sorbents: (a) Reillex 425,(b) Amberlite IRA-910, (c) Dowex MWA-1, (d) Duolite A7, (e) Bio-Rad AG3-X4, (f) Amberlite IRA-35.
concentrations were determined by HPLC. Organicphase acid concentrations were determined by twophase titration with aqueous NaOH. Mass balances closed within 1-5%. Sulfate and phosphate uptake experiments for extraction systems were entirely analogous to those for sorption systems. Additional details on the experiments are available elsewhere (Tung, 1993).
Solid Sorbents Basicities. The structures of the sorbents studied are shown in Figure 1 (Tung, 1993). Bio-Rad AG3-X4 is a cleaned, screened version of Dowex XUS40091, which is closely related to Dowex WGR-2, an epoxyamine resin formed by condensation of ammonia and epichlorohydrin and containing primarily tertiary amine groups. Bio-Rad AG3-X4 has also been reacted with methyl chloride to add 10% quaternary ammonium groups. Several investigators (Gustafson et al., 1970; Garcia and King, 1989; Clifford and Weber, 1983) have determined apparent pK, values for weak-base ion-exchange resins. These results are summarized by Tung (1993). The values calculated by Gustafson et al. do not take into account the fact that the pH inside the resin is different from the pH of the external solution (Helfferich, 1962). Hence the measured pKa values are expected to be low estimates of the apparent PKa values. However, in the work of Gustafson et al., the aqueous titrant was very carefully buffered, so that pH values could be determined with high reproducibility. The titrant was unbuffered in the studies of Clifford and Weber and of
Ind. Eng. Chem. Res., Vol. 33, No. 12, 1994 3219 5-
IRA-35
4-
MWA-I
I4
8 3-
V
2-
425
8 +
+ Garcia and King, 1989
1Solution Concentration
0
0
2
4
6
1 8 1 0 1 2 1 4
(a)
Apparent pKa
Figure 3. Correlation between log K for lactic acid and apparent PK~. Table 2. K Values for Lactic Acid Isotherms sorbent Reillex 425 Duolite A7
PH (b)
Dowex MWA-1 Amberlite IRA-35 Amberlite IRA-910
K (g oisolution/ g of acid) 47.6 K1= 1780 Kz = 38 Kay= 1100 5960 28200 165000
qm (g of acid/ !E of sorbent)
0.26 qml = 0.25 qmz = 0.16
0.35 0.40 0.36
Figure 2. Effect of sorbent K value on (a) sorption isotherm and (b) uptake-pH curve.
Garcia and King, a factor that can lead to greater uncertainty. Thus the results of Gustafson et al. are probably more precise indicators of relative basicity. The results of Gustafson et al. indicate a wide range of basicities, with pKa values spanning 4.4 units of pH. The poly(viny1pyridine) resin is the most weakly basic, followed by the phenol-formaldehyde-amine resin. Although the two tertiary amine resins are the most strongly basic, the polystyrene-based resin (Dowex MWA-1) is significantly less basic than the acrylic-based resin (Amberlite IRA-351, probably because of the proximity of the amine group to the base-weakening styrene ring (Perrin et al., 1981; Gustafson et al., 1970). Sorption Isotherms. The ideal-exchange model is a simple model that can be used to describe sorption of an acid by a basic sorbent (Garcia and King, 1989).The resulting isotherm, shown in eq 1,is of the Langmuir
form, with parameters qm, a measure of sorbent capacity, and K, a measure of the strength of complexation. The value of K is given by the initial slope of the sorption isotherm if qm is known, and, for complexation with a given acid, should be higher for more strongly basic sorbents. The effect of the K value can be seen in Figure 2. As K increases, the initial slope of the sorption isotherm in Figure 2a becomes steeper, and the amount of acid sorbed reaches its maximum value at lower solution concentration. As shown in Figure 2b, uptake drops off with increasing pH as the amount of un-ionized acid decreases. However, as K becomes greater, the pH value at which this drop begins to occur becomes higher, because capacity is larger a t low concentrations of unionized acid. Therefore, sorbents with high K values sustain capacity t o higher values of pH. If there is reason to believe that the resin contains two types of functional groups or if only two types of groups are strongly enough basic to give significant uptake of acid, a two-site Langmuir model (eq 2) can be used to describe the data.
Lactic Acid. For lactic acid data (Tung, 1993) all of the sorbents, with the exception of Duolite A7, follow a Langmuir isotherm. The Duolite A7 data are better described by a dual-site Langmuir isotherm. Table 2 lists the fitted values of K and q m for each sorbent. For comparison of Duolite A7 with other sorbents, an average of KI and Kz, weighted by the corresponding qm values, can be used (Tung, 1993). This weighted average is also listed in Table 2. The K values span a wide range and reflect the order of increasing basicity of the sorbents. Since the K value is both an equilibrium constant and a measure of sorbent basicity, it should be possible to correlate log K with pKa via a linear free-energy relationship (LFER) (Garcia and King, 1989). Figure 3 presents a plot of log K versus the values of apparent pKa reported by Gustafson et al. (1970) and Garcia and King (1989). The PKa values are very highly correlated with log K. This correlation is not surprising and can be predicted quantitatively. First, K corresponds to a complexation, or ion pairing, constant for the carboxylic acid and the base:
where overbars indicate sorbent-phase concentrations. PKal corresponds to an ionization constant, Kal, for the carboxylic acid:
K: = [H+][A-ll/[HAl
(4)
pKa for the base corresponds to a dissociation constant, Ka, for the protonated base:
If the association of the carboxylic acid and the base is exclusively by ion pairing, then combination of eqs 3-5 gives
3220 Ind. Eng. Chem. Res., Vol. 33, No. 12, 1994
Table 3. Values of K and qm for Succinic Acid Isotherms KI ( g of solution/ qml(g of acid/ sorbent g of acid) g of sorbent) Reillex 425 71 0.47 Duolite A7 1 830 0.30 Bio-Rad AG3-X4 21 300 0.26 Dawex MWA-1 38 400 0.30 Amberlite IRA-35 330 000 0.25 Amberlite IRA-910 199 000 0.28
.
6
IRA45
5
MWA-1 AG?-X4
o J . . 0
2
.
Kz (g of solution/
g of sorbent)
g of acid)
17.9 79.4 61.9 104 93.2
0.35 0.22 0.26 0.32 0.29
6
(g of solution/ g of acid)
853 11 600 20 600 145 000 96 600
1
.’./ . . . . . . . . II
4
K,
I
/
A1
Ks (g of solution/
8
I I
1 0 1 2 1 4
Apprent pKa
2
Figure 4. Correlation between log K for succinic acid and apparent PK,.
K = K,,IK,
4
8
6
10
12
PH
Figure 5. Effect of pH an normalized uptakes of lactic acid (initial acid concentration = 0.45 M solution-sorbent phase ratio = 11.0 P/pL
The relationship predicted by eq 6 is shown by the solid line in Figure 3. This analysis contains a number of simplifying assumptions. One is neglect of the difference between [H+lin solution and [H+lwithin the resin, brought about by Donnan exclusion and other factors. Another is neglect of effectsof complexation by hydrogen bonding and association due to hydrophobic exclusion of the carboxylate from aqueous solution. For Amberlite IRA-910 the pK. value is difficult to quantify. Strong-base ion exchangers typically have pK. 7 13 (Helfferich, 1962). From a comparison of the experimental K value in Table 2 with the correlation indicated in Figure 3, it is apparent that the pK. value corresponding to the experimental K value is much lower than 13. However, the experimental value of K for this resin is imprecise because the concentrations in the initial region of the isotherm are all extremely low. Succinic Acid. Sorption isotherms were also generated for systems containing succinic acid. As for lactic acid, the Reillex 425 data are well fit by a Langmuir isotherm, while the Duolite A7 data are described best by a dual-site Langmuir isotherm. For the other sorbents the Langmuir isotherm does not fit well (Tung, 19931, a result that is probably related to the fact that succinic acid is a diacid. For the purpose of comparing sorbent basicities, dual-site Langmuir curves were fitted to each isotherm other than that for Reillex 425, and a weighted-average K value was calculated as for lactic acid. The resulting values of K are given in Table 3. The order of increasing K values is the same as that displayed for lactic acid, except that for succinic acid Amberlite IRA-35 reflects a larger K value than does Amberlite IRA-910. This finding is inconsistent with the fact that Amberlite IRA810 should be a much stronger base than Amberlite IRA-35 and again probably relates to the error inherent in measuring extremely low concentrations of acid. The weighted-average K values correlate well with the apparent pK. values given by Gustafson et al. (1970) and Garcia and King (19891, as shown in Figure 4. Again, the solid line corresponds to eq 6.
Effect of pA upon Capacity. Lactic Acid. The uptakes of lactic acid by all five sorbents are presented as functions of the pH of the equilibrium aqueous solution in Figure 5. The cumes are derived from the models fitted to the sorption isotherms and thus contain no additional fitted parameters. As was noted, these uptakes were measured for a constant initial phase ratio (grams of solutiodgrams of sorbent) and a constant initial concentration of total acid (ionized and unionized). Using these results and appropriate models, it is possible to generate curves for uptake as a function of pH a t a fixed equilibrium concentration of total acid in the aqueous phase (Tung, 1993). The uptakes in Figure 5 are reported as fractions of sorbent capacity (q,,,), obtained from the sorption isotherms, in order to highlight the effects of resin basicity. The order of effectiveness for sustaining capacity with increasing pH is Amberlite IRA-910 7 Amberlite IRA35 > Dowex MWA-1 > Duolite A7 7 Reillex 425. This order reflects the relative basicities of the resins and is the same order followed by the experimental K values. For use with lactic acid fermentation, it is desirable that the sorbent demonstrate significant uptake of acid in the pH 5-6 range. The uptake in this range varies substantially among sorbents and is strongly dependent upon the basicity of the sorbent, as well as the capacity of the resin. Reillex 425 has virtually no selective uptake of acid in this range. Uptakes are 0.18-0.08 g of acidg of sorbent for Duolite A7,0.3-0.18 g of acid/g of sorbent for Dowex MWA-1, 0.35-0.29 g of acidg of sorbent for Amberlite IRA-35, and about 0.3 g of acidg of sorbent for Amberlite IRA-910. On this basis, Dowex MWA-1 and Amberlite IRA-35 are probably the best candidates for use for product recovery in a fermentation process at pH 5-6. Amberlite IRA-910, which displays good capacity, would probably be more difficult to regenerate by means of a pH-swing process, since capacity is sustained to very high values of pH. Succinic Acid. The uptakes of succinic acid as a fraction of sorbent capacity (q,) are presented as a function of pH in Figure 6. The order of choice for sustaining capacity with increasing pH is Amberlite
Ind. Eng. Chem. Res., Vol. 33, No. 12,1994 3221 l.W R-C’
-+ *E D
sz
-
~, O - - - - H - OR’
R-C’.
-
K.o....H* MR. I *
Bio-Rad AG3-X4
H ---.O.R’
..V. DmierMWA.1
0.60
!g
o---.H.oR’
k.o....H* NR*
0.80
$$
/,
I
-0- AmbeOite IRA45
H @)
(a)
0.40
Figure 7. Stabilization by oxygenated diluents: (a) tertiary
E?
amine; (b) secondary amine
1 0 0.20
1.1 0.W -0.10
0.9 2
4
6
8
10
12
14
0.7
PH
Figure 6. Effect of pH on normalized uptakes of suceinic acid (initial acid concentration = 0.45 M solution-sorbent phase ratio = 11.0 dg).
IRA-910 > Bio-Rad AG3-X4 > Amberlite IRA-35 > Dowex MWA-1 > Duolite A7 > Reillex 425. This order is consistent with the results for lactic acid and, in general, reflects the relative basicities of the resins. BioRad AG3-X4, predicted to be a slightly weaker base than Dowex MWA-1, sustains capacity to higher values of pH. This result is, at least in part, attributable to the 10% quaternary groups. For use with a succinic acid fermentation, a sorbent should demonstrate significant uptake of acid a t pH about 6. Reillex 425 demonstrates essentially no selective uptake of acid at pH 6, and Duolite A7 shows an uptake of 0.08 g of acid/g of sorbent. The stronger bases (Bio-Rad AG3-X4, Dowex MWA-1, Amberlite IRA-35, and Amberlite IRA-910) all have uptakes between 0.22 and 0.25 g of acid/g of sorbent. Extractants Long-chain, aliphatic amines are effective extractants for the recovery of carboxylic acids from aqueous solutions (Kertes and King, 1986 Tamada et al., 1990; Tamada and King, 1990a,b). The amine extractant is typically dissolved in an organic solvent, or diluent, which is useful for controlling the density and viscosity of the organic phase. The diluent can also have a profound effect on the extraction equilibria. Several researchers have examined the performance of amine extractants at values of pH typical of fermentation systems (Yabannavar and Wang, 1987, 1991; Yang et al., 1991). Yang et al. (1991) investigated the effect of pH on the extraction of several carboxylic acids by Alamine 336 and Aliquat 336 (a long-chain quaternary alkylammonium chloride) in kerosene and 2-octanol diluents. Higher complexation constants, and hence increased extraction power, can be achieved in systems in which the diluent is able to stabilize the acid-amine complex. Diluents can be classified as “inert” or “active” based upon their degree of interaction with the acid-amine complex. Inert diluents, such as alkanes o r aromatics, are relatively nonpolar and provide very little solvation of the acid-amine complex. Diluents such as ketones and nitrobenzene are polar and therefore provide a good solvating medium for the ammonium carboxylate ion pairs. Protic diluents, such as chloroform and alcohols, interact with the acid-amine complex through hydrogen bonding, as shown in Figure 7a, thus providing further stability (Tamada and King, 1990b; Barrow and Yerger, 1954; Yerger and Barrow, 1955a,b). Pearson and Vogelsong (1958) showed that primary and secondary, but not tertiary, amines are stronger
!g 3
0.5 0.3
0.1
-0.1 1
3
5
7
9
PH Figure 8. Effect of pH on the extraction of lactic acid by Alamine 336 in various diluents. Initial extractant concentration = 0.30 M, phase ratio = 1.0.
bases in oxygenated solvents than in nonoxygenated solvents with the same dielectric constant. They suggest that oxygenated solvents stabilize the primary and secondary amine complexes by forming a hydrogen bond with the amine protons, as in Figure 7b. Tertiary amines, however, have no additional protons with which an oxygenated diluent can form hydrogen bonds (Figure 7a). Alamine 336 (Henkel Corp.) is a straight-chain tertiary amine with a mixture of about 67% Cs and 33% Cl0 side chains, while Amberlite LA-2 (Rohm and Haas Co.) is an asymmetric alkylamine with a straight Clz chain and a highly branched chain of 12-15 carbon atoms (Ricker et al., 1979). Three different diluentsmethyl isobutyl ketone (MiBK), chloroform, and l-oct a n o l w e r e used with these amines. Effect of pH upon Capacity. Lactic Acid. The stoichiometric loadings of Alamine 336 (moles of carboxylic acid/mole of extractant) with lactic acid in three different diluents are presented as functions of equilibrium aqueous-phase pH in Figure 8. Smooth curves have been drawn through the data. Equilibrium distribution ratios can readily be calculated from these loadings through knowledge of the phase ratio and the total amounts of carboxylic acid and the extractant present. The stoichiometric loading is used here because it is less dependent upon these other parameters. At low pH, 100%loading of the amine is achieved for all diluents. Chloroform and 1-octanol sustain significant capacity to two or more pH units above the pK,1 of the acid (3.86). In particular, a t pH 5.5, where a typical fermentation might occur, the loading is 0.61 and 0.50 for chloroform and 1-octanol, respectively, as diluents. The corresponding loading with MiBK diluent is only 0.15. The observed differences in performance among dilueuts reflect their relative abilities to stabilize the acid-amine complex. MiBK is polar and thus provides general solvation of the acid-amine ion pair. Chloroform and 1-octanolhydrogen bond with the acid-amine complex, providing additional stabilization.
3222 Ind. Eng. Chem. Res., Vol. 33, No. 12,1994 1.1 1
0.9
0.8
0.7 M
M
2
.-
0.5
0.6
3
3 0.3
0.4
0.1
0.2 0
-0.1 1
3
5
7
1
9
PH Figure 9. Effect of pH on the extraction of lactic acid by Amhrlite LA-2in various diluents. Initial extractant concentration = 0.30 M,phase ratio = 1.0. 1.1
I
3
PH 11. Effect of pH on the extraction of succinic acid by Amberlite LA-2 in various diluents. Initial extractant concentration = 0.30 M,phase ratio = 1.0.
Table 4. a Values for SulfaWLactic Acid wt%
Dowex MWA-1 Amberlite LA-WdiBK Amberlite LA-7Jl-odanol Alamine 336MiBK Alamine 336/1-actanol
4 1
I 3
5
7
9
PH Effect of pH on the extraction of succinic acid by Figure 10. Alamine 336 in various diluenta. Initial extractant concentration = 0.30 M,phase ratio = 1.0.
Figure 9 presents the corresponding results for lactic acid and Amberlite LA-2. At low pH, the amine is again a t least 100% loaded. With the MiBK diluent, a small degree of overloading is observed. This result is consistent with the findings of Tamada et al. (1990), who showed that ketones tend to promote overloading. At pH 5.5 Amberlite LA-2 retains loadings of 0.72, 0.61, and 0.32 in diluents of 1-octanol,chloroform, and MiBK, respectively. Comparison of Figures 8 and 9 reveals that the MiBK and 1-octanol curves are shifted toward significantly higher pH values for the secondary amine, whereas the chloroform curve is essentially the same. The shifts in the MiBK and 1-octanolcurves are expected, since LA-2 is a secondary amine and therefore has an additional proton available for solvation by an appropriate diluent. Succinic Acid. The loading of Alamine 336 with succinic acid is presented as a function of pH in Figure 10. The loadings at low pH are consistent with the work of Tamada et al. (1990), who observed promotion of overloading with MiBK. Otherwise, the results are similar to those for lactic acid, in that chloroform and 1-octanol sustain capacity to higher pH than does MiBK. The relative dropoff in loading with increasing pH is less steep in 1-octanolthan in chloroform. This behavior relates to the relative tendencies to form (1,2) and (2,2) complexes (Tamada et al., 1990; Tamada and King, 1990a). Overall, the performance of the extractants at a given pH is similar to that for lactic acid. The fact that succinic acid has a higher p& than lactic acid is offset
9
7
Fi-
sorbenffextractant
.O.l
5
lactic acid 2
wt% SO4
0.1
2 2
0.2 0.2
2
0.2 0.2
2
a 124 5.4
11.1 4.2 13.2
by the correspondingly lower equilibrium constant for complexation with the amine (eq 6). Finally, Figure 11presents the results for extraction of succinic acid with Amberlite LA-2. Even a t low pH, the loading of Amberlite LA-2 in 1-octanol is significantly lower than the loading of Alamine 336 in the same diluent. Amberlite LA-2 may have a greater tendency to form complexes with more than one amine per acid. Several researchers (Yerger and Barrow, 1955a,b; Guskova et al., 1986) have reported that primary and secondary amines are more likely to form multiamine complexes than are tertiary amines. As a result of the unusual shape of the 1-octanol curve, chloroform as diluent exhibits higher loadings than 1-octanol in the pH 5-5.5 range, but 1-octanol shows higher loadings the pH 6-7 range. As with lactic acid, the curves for MiBK and 1-octanol are shifted toward higher pH values relative to those for Alamine 336, whereas the chloroform curve is almost the same.
Uptakes of Sulfate and Phosphate In real fermentation systems, competitive uptake of other species by the extractant or sorbent may be an issue. In particular, anions of strong acids such as phosphate and sulfate may be taken up competitively. Batch equilibrium experiments were used to determine separation factors for sulfateflactic acid and phosphate/ lactic acid in sorbent and extractant systems. The separation factor, a, is a measure of the relative partitioning of the two species between the aqueous and organic, or sorbent, phases and is given by the ratio of the equilibrium distribution ratio for sulfate or phosphate to that for the carboxylic acid. For sorbent systems, the concentration in the organic phase is given by the composite uptake. Tables 4 and 5 present experimentally determined values of a for sulfate- and phosphate-containing systems, respectively, at equilibrium lactic acid concentrations of 2 w t %. One sorbent (Dowex MWA-1) and two extractants (Amberlite LA-2 and Alamine 336) were examined. In these experiments sulfate and phosphate
Ind. Eng. Chem. Res., Vol. 33, No. 12, 1994 3223 Table 5. a Values for PhosphateLactic Acid wt%
wt%
sorbenffextractant
lactic acid
PO4
a
Dowex MWA-1 Amberlite LA-2/MiBK Amberlite LA-Wl-octanol Alamine 336iMiBK Alamine 33611-octanol
2 2 2 2 2
0.5 0.6 0.6 0.6 0.6
3.8 0 0 0 0
were supplied as sodium salts, and the resultant pH values of the solutions were not adjusted. The concentrations of sulfate and phosphate are somewhat higher than would typically be found in fermentation broths. The results indicate that sulfate is taken up more strongly than phosphate, despite the fact that phosphate was present a t higher concentrations. Furthermore, uptake of both anions is much greater for sorbents than for extractants, presumably because the sorbent provides an aqueous environment whereas the extractant provides a more organic one, which would accommodate the more organic carboxylic acids preferentially. Correspondingly, the preferential uptake for sulfate is less with MiBK, the less polar diluent, than with 1-octanol.
Acknowledgment This research was supported by the Biological and Chemical Technology Research Program, Advanced Industrial Concepts Division, Office of Industrial Technologies, U.S.Department of Energy, and by a National Science Foundation Graduate Fellowship.
Nomenclature C m = concentration of un-ionized carboxylic acid (grams of acidgrams of solution) K = complexation constant for sorption (g residg acid) K1 = complexation constant for sites of class 1 K2 = complexation constant for sites of class 2 K2, Kal, Ka2 = see PKa, PKal and ~Ka2 pKa = negative logarithm (base 10) of the dissociation constant (used for deprotonation of a basic sorbent or extractant) pKal = negative logarithm (base 10) of the first, or only, ionization constant (used for carboxylic acid solutes) pKe = negative logarithm (base 10)of the second ionization constant (used for dicarboxylic acid solutes) q = amount of sorbed carboxylic acid, glg of dry sorbent qm = maximum value of q , at high concentration qml = qm for sites of class 1 q m 2 = qm for sites of class 2
Literature Cited Barrow, G. M.; Yerger, E. A. Acid-Base Reactions in Nondissociating Solvents: Acetic Acid and Triethylamine in Carbon Tetrachloride and Chloroform. J . Am. Chem. SOC.1954, 76, 5211. Clifford, D.; Weber, W. J., Jr. The Determinants of Divalent' Monovalent Selectivity in Anion Exchangers. React. Polym. 1993, I , 77. Davison, B. H.; Thompson, J. E. Simultaneous Fermentation and Separation of Lactic Acid in a Biparticle Fluidized-Bed Bioreactor. Appl. Biochem. Biotechnol. 1992,34135, 431. Dow Chemical Co.: DOWEX Zon Exchange Resins Engineers' Handbook; Midland, MI, 1983. Garcia, A.; King, C. J. The Use of Basic Polymer Sorbents for the Recovery of Acetic Acid from Dilute Aqueous Solution. Znd. Eng. Chem. Res. 1989,28, 204.
Glassner, D. A,; Datta, R. Process for the Production and Purification of Succinic Acid. U. S Patent 5,143,834, Sept 1992 . Gusakova, G. V.; Denisov, G. S.; Smolyanskii, A. L. Infrared Spectra and Structures of Complexes Formed for Dibasic Carboxylic Acids with Amines in Nonaqueous Solvents: Malonic Acid. J. Gen. Chem. USSR (Engl. Transl.) 1986,56, 531. Gustafson, R. L.; Filius, H. F.; Kunin, R. Basicities of Weak Base Ion Exchange Resins. Znd. Eng. Chem. Fundam. 1970,9,221. Helfferich, F. Zon Exchange; McGraw-Hill: New York, 1962. Jain, M. K.; Datta, R.; Zeikus, J. G. High-Value Organic Acids Fermentation-Emerging Processes and Products. In Bioprocess Engineering: The First Generation; Ghose, T. K., Ed.; Ellis Horwood: Chichester, 1989; pp 366-389. Kaneuchi, C.; Seki, M.; Komagata, K. Production of Succinic Acid from Citric Acid and Related Acids by Lactobacillus Strains. Appl. Environ. Microbiol. 1988, 54, 3053. Kertes, A. S.; King, C. J. Extraction Chemistry of Fermentation Product Carboxylic Acids. Biotechnol. Bioeng. 1986,28, 269. Kirk-Othmer Encyclopedia of Chemical Technology, 3rd ed.; John Wiley & Sons: New York, 1979. Citric Acid, Vol. 6; Hydroxycarboxylic Acids, Vol. 13. Pearson, R. G.; Vogelson, D. C. Acid-Base Equilibrium Constants for 2,4-Dinitropherol and Some Aliphatic Amines in Nonaqueous Solvents. J. Am. Chem. SOC.1958,80,1038. Perrin, D. D.; Dempsey, G.; Serjeant, E. P. pK. Prediction of Organic Acids and Buses; Chapman & Hall: London, 1981. Reilly Industries, Inc., Reillex: A New Family of Crosslinked Polyvinylpyridines from Reilly, Indianapolis, 1989. Ricker, N. L.; Michaels, J . N.; King, C. J. Solvent Properties of Organic Bases for Extraction of Acetic Acid From Water. J . Sep. Process Technol. 1979, I , 36. RoMer, S. R.; Randolph, T. W.; Miller, D. A.; Blanch, H. W.; Prausnitz, J. M. Extractive Bioconversions with Nonaqueous Solvents. In Extractive Bioconversions, Mattiasson, B., Holst, O., Eds., Marcel Dekker, New York, 1991; pp 133-171. Rohm & Haas Company, AmberlitelDuolite Zon Exchange Resins: Technical Notes; Philadelphia, 1981. Tamada, J. A.; Kertes, A. S.; King, C. J. Extraction of Carboxylic Acids with Amine Extractants. 1.Equilibria and Law of Mass Action Modeling. Znd. Eng. Chem. Res. 1990,29, 1319. Tamada, J. A.; King, C. J. Extraction of Carboxylic Acids with Amine Extractants. 2. Chemical Interactions and Interpretation of Data. Znd. Eng. Chem. Res. 1990a, 29, 1327. Tamada, J . A.; King, C. J. Extraction of Carboxylic Acids with Amine Extractants. 3. Effect of Temperature, Water Coextraction, and Process Considerations. Znd. Eng. Chem. Res. 199Ob, 29, 1333. Tung, L. A. Recovery of Carboxylic Acid at pH Greater than pKa. Report No. LBL-34669, Lawrence Berkeley Laboratory, Berkeley CA, August, 1993. Vickroy, T. B. Lactic Acid. In Comprehensive Biotechnology, Blanch, H. W., Drew, S., Wang, D. I. C., Eds.; Pergamon Press: New York, 1985; Vol. 3., pp 761-776. Yabannavar, V. M.; Wang, D. I. C. Extractive Fermentation for Lactic Acid Production. Biotechnol. Bioeng. 1991, 37, 1095. Yabannavar, V. M.; Wang, D. I. C. Bioreador System with Solvent Extraction for Organic Acid Production. Ann. N.Y. Acad. Sci. 1991,506, 523. Yang, S.-T.; White, S. A.; Hsu, S.-T. Extraction of Carboxylic Acids with Tertiary and Quaterary Amines: Effect of pH. Znd. Eng. Chem. Res. 1991,30, 1335. Yerger, E. A.; Barrow, G. M. Acid-Base Reactions in Nondissociating Solvents Acetic Acid and Diethylamine in Carbon Tetrachloride and Chloroform. J . Am. Chem. SOC.1955a, 77, 4474. Yerger, E. A.; Barrow, G. M. Acid-Base Reactions in Nondissociating Solvents: n-Butylamine and Acetic Acid in Carbon Tetrachloride. J . Am. Chem. SOC.1955b, 77, 6206. Received for review January 27, 1994 Revised manuscript received August 12, 1994 Accepted August 29, 1994* Abstract published in Advance ACS Abstracts, November 1, 1994. @
