Exchange in Biosorption - ACS Publications - American Chemical

Department of Chemical Engineering, McGill University, 3480. University Street, Montreal, Quebec, Canada H3A 2A7. Biosorption of the heavy metal ions ...
1 downloads 0 Views 1MB Size
Environ. Sci. Techno/. 1995, 29, 3049-3058

Modelins of the Proton-Metal Ion Exchange in Biosorption SILKE SCHIEWER AND

B O H U M I L VOLESKY* Department of Chemical Engineering, McGill University, 3480 University Street, Montreal, Quebec, Canada H3A 2A7

Biosorption of the heavy metal ions Cd2+, Cu2+, and Zn2+ by previously protonated nonliving biomass of the marine alga Sargassum fluitans was observed t o be coupled with a release of protons. Metal ion binding experiments with continuously controlled pH were performed. The metal ion and proton binding at equilibrium were modeled as a function of pH and metal ion concentration using a modified multicomponent Langmuir sorption model. Both the exchange of metal ions for protons from functional groups in their acidic form and the sorption of metal ions on ionized groups were considered. The model is applicable to adsorption by biomass with free or protonated metal binding sites as well as to metal ion desorption with acids since the direction of the reaction depends simply on the given initial conditions. The model parameters were incorporated into the M l N E Q L f equilibrium program, leading to a prediction of the equilibrium, e.!., of metal ion laden biosorbent desorption performance for given initial conditions.

Introduction The property of nonliving biomass to accumulate heavy metal ions, called biosorption, has been known for several decades. Presently, there is a growing focus on using biosorption for the elimination of toxic heavy metal ions (such as those considered in this study, i.e., Cd, Cu, and Zn) from industrial wastewaters. Biosorption could be employed most effectively in a concentration range below about 100 mg/L, where other techniques are ineffective or costly. Certain types of algal biomass have been found to be particularly effective in metal ion binding. Gold accumulation by the ubiquitous brown alga Sargassum can constitute up to 40% of its dry weight (I). The metal ion binding in biosorption has been attributed to different metal sequestering mechanisms such as ion exchange, complexation, electrostatic attraction, and microprecipitation (the latter as metal or metal salt). There have been some indications that ion exchange plays an important role in metal sorption by algal biomass (2, 3). For marine algae, the active molecular entities involved are believed to be carboxyl and sulfate groups (4). According * Corresponding author telephone: (514) 398-4494; fax: (514) 398e-mail address: [email protected].

6678;

0013-936x195/0929-3049$09,00/0

0 1995 American Chemical Society

to Buffle (18), the functional groups in marine algae, other than OH groups, are the carboxyl groups of alginic acid and the carboxyl and sulfate groups of fucoidan. Alginic acid is a polymer composed of mannuronic and guluronic acids, both of which contain carboxyl groups. Fucoidan contains L-fucose residues with sulfate groups on C4. Both alginic acid and fucoidan occur in the cell wall of marine algae and as extracellular polysaccharides (5). Since both groups are acidic, the availability of free sites depends on pH. This corresponds to an increased metal cation binding with the increasing pH of the sorption system (6, 7). In addition to the role of protons in changing the state of active metal ion binding sites, there are also two other ways in which pH influences sorption: first, the speciation of the metal ion in solution is dependent upon the decreasing solubilityof the metal complexes with increasing pH. This may impose limitations on the pH range suitable for studying biosorption. Since adsorption depends not only on the attraction of the sorbate to the solid surface but also on its lyophobic behavior (sorption increases with decreasing solubility), for most metals that means adsorption increases with increasing pH. Specifically, in the narrow pH range where the metal ions are hydrolyzed, sorption is enhanced so that even a charge reversal of the surface to positive values can occur (8). On the other hand, too high pH values, which cause precipitation of metal complexes, should be avoided during sorption experiments where distinguishing between sorption and precipitation metal removal becomes difficult. Second, extreme pH values, as they are employed in the regeneration (desorption) of the sorbent, may damage the structure of the biosorbent material. Microscopicobservations have shown distorted cells; significant weight loss and decrease in sorption capacityhave been observed (9). In the pH range 2-7, no serious damage is expected to occur. Although the crucial role of protons in biosorption is generallyknown, it is usuallyneglected in the mathematical description of the process. The complex nature of biosorbent materials makes the application of methodologies used in studyingthe behavior of chemicallysimple synthetic ion exchange resins difficult. As a consequence, it has been a common practice to determine a separate isotherm for each pH value or for each initial biomass saturation state (Le., protonated or loaded with ions from seawater). This has been necessary because the most frequently used Langmuir or Freundlich sorption models do not take into account the fact that metal ion biosorption is largely an ion exchange phenomenon. They do not allow the prediction of the remaining binding of those ions (e.g., protons or sodium) that were initially loaded onto the biosorbent, and neither do they incorporate the concentration of the exchangedspecies (e.g.,protons) as aparameter. As aresult, the conventional and tedious determination of the effects of these parameters has been necessary, normally not allowing any calculated predictions of the biosorbent performance. This work presents an approach to modeling the binding of heavy metal ions (Cd, Cu, and Zn chosen as examples) and protons, as a function of metal ion concentration and pH, for a wide range of initial sorbent loadings with the heavy metal ion and/or protons. The model presented here

VOL. 29, NO. 12, 1995 /ENVIRONMENTAL SCIENCE & TECHNOLOGY

3049

enables the prediction of the effect ofprotons as exchanged species on the metal ion binding. Although it has been recognized that the use of protonated biosorbent in sorption columns would tend to lower the pH, which in turn might adversely affect the sorption performance, this material was chosen for the present study because it enables us to study a simple sorption system with only one metal ion and protons. Using biosorbent initially saturatedwith other ions would introduce another species into the system since protons would always be present. Investigation of the proton-metal ion exchange is a prerequisite for further expansion of the model considering eventually also other ions.

Materials and Methods Preparation of Sorbent. Beach-dried Sargassumpuitans, collected in Naples, FL, was ground in a homogenizer and sieved. To eliminate weight loss and leaching of soluble biomass components, such as alginate, the size fraction (0.84- 1 mm) was crosslinkedwith formaldehyde (10).After being washed with distilled deionized water, the biomass was protonated in 0.1 N HC1 (10 g of biomass/L),washed 10 times in the same volume of distilled deionized water, and dried in an oven at 60-80 "C. The protonation of the biomass was performed to eliminate any other exchangeable ions that were present on the raw biomass, thereby enabling the study of a simple sorption system, one involving only two cations. In contrast to the nonprotonated biomass, analysis of the solution after heavy metal ion binding showed that no other ions such as Mg, Ca, Na, or K were released even at high heavy metal ion concentrations. Metal Ion Binding Experiments. The sulfate salts 3CdS04.8H20(AESAR),CuS04.5H20(ACP Chemicals),and ZnSO4.7H200.T. Baker Chemical)were dissolved in distilled deionized H20. A total of 0.1 g of biomass was contacted with 50 mL of metal ion-containing solution in 125-mLErlenmeyer flasks on a gyrotory shaker (New Brunswick Scientific,Model G2) at 2 Hz for 12 h. The pH was constantly controlled using a pH controller (ColeParmer, Model 5997-20). Biosorption isotherms for each of the three metal ions (Cd, Cu, and Zn) were established at pH 2.5, 3.0, and 4.5, respectively. The amount of 0.1 M HCl or NaOH consumed was monitored. Blanks, controls, and duplicates were run as appropriate. Some experiments were performed with [MI, = [Mlf,so that a desired final concentration was obtained. This was achieved by choosing a liquid volume of 1 L. Since the metal ion binding could not be determined from the concentration difference, the metal ion laden biomass was separated by filtration, dried, and subsequently desorbed in 50 mL (vdes) of 1 M HCl overnight on a rotary shaker. The desorption solution was analyzed for the metal ion concentration ([M]des). A correction was made for the metal ion in the water volume (Vad) adhering to the biomass ( m ) after filtering:

Determination of Binding. Initial and final metal ion concentrations were determined using an atomic absorption spectrometer (Thermo JarrelAsh, Model Smith-Hieftje 111, calibrated with standard solutions prepared from 1000 ppm certified reference solutions (Fischer). The concentration range used for the determination of the model 3050

ENVIRONMENTAL SCIENCE & T E C H N O L O G Y ' V O L 29 NO. 12, 1995

3

0

.

'

_ - - - _pH 4 5 1 site model - - - - - p H 2 5 1 site model pH 4 5 2 site model

-

0.01

0.1

1

10

Cd Concentration (mM)

FIGURE 1. Biosorption isothermfor Cd binding by initially protonated biomass at pH 4.5 and 2.5 (experimental data, the one-site model and the two-site model).

parameters was 0-10 mmol/L. The metal ion binding during biosorption was calculated as

with Vas the liquid volume, m as the mass of sorbent, [Mli and [Mlf as the initial and final metal concentrations, respectively. The change in the proton binding AqH was calculated as the difference between the amount of protons added ([HIaddVadd)and the amount of protons that accumulated inthesolution(([H]f- [H]i)V).ForpHnear 7, anadditional term for water dissociation may be added that was, however, negligible under the conditions imposed.

EPM Measurements. For measurements of the electrophoretic mobility (EPM),the binding experiments were performed as usual except that more finely ground biomass was used. The pH was adjusted to different values. For each sample, 10runs were performed on an electrophoresis apparatus (Rank Brothers). The time ( t ) of the particle to cross the distance (d = 100pm) was measured during each run. The average value (f) was used to calculate the electrophoretic mobility (particle velocitylpotential gradient) for a cell length L = 7.13 cm at an applied potential U=96V

Results and Discussion Influence of pH on Sorption. A trend of increasing metal ion binding with increasing pH could be observed for all three metal ions examined in this work. Figure 1 shows the results for cadmium. The same trend was noted by other researchers, e.g., for the sorption of cobalt on Ascophyllum (6) or uranium with Streptomyces niveus or ion exchange resins (7), while other types of microbial biomass did not show a large pH dependence of sorption above pH 3.5 (7). Differing results were obtained for the sorption of the soft ions Au3+, Hg2-, and Ag' by algal biomass, where the binding was independent of pH in the pH range 2-7 (11). The influence of pH-dependent uranium solution speciation was suspected to be responsible for the variation of uranium binding with pH. While uranium occurs as U022' at pH 2, it is present as a negatively charged hydrolyzed species at pH 4 and above (7). This

can explain the lower sorption of uranium at low pH for anionic resins. Since Cd, Cu, and Zn feature no hydrolyzed species at pH 9 4.5, the pH influence on their binding by the biosorbent is an indication of the interaction of biomass active sites with protons. This means that protons and metal ions compete for the same binding sites, with more sites being available for metal ion sorption at higher pH values. It has been recognized by Crist et al. (12)that the main effect of pH on metal ion binding consists of a reduction in the number of binding sites available with decreasing pH. Change of pH. It was noticed in this work that the pH dropped from an initial value of about 4-5 to around 3 during sorption with protonated biomass, if no pH adjustment was undertaken. This means that proton release took place, because no reaction in the solution could account for the pH change. Raw, unprotonated biomass exhibited the reverse phenomenon: the pH rose from an initial pH of 3.5 (biomass and metal ion-free solution, adjusted to pH 3.5 with HC1) to pH 6 after addition of the biomass (data not shown). After readjusting the pH to 3.5, a release of light metal ions that had been bound from the seawater (0.28 mequiv of Ca, 0.32 mequiv of Mg, 0.15 mequiv of K, and 0.22 mequiv of Na) was measured. Since no reactions that involve protons should occur in the solution, the increase of pH must have been due to a binding of protons from the solution by the biomass. The observations, therefore, can be explained by an exchange between the light metal ions initially present and the protons. Since a higher ionic strength and also higher pH characterize seawater, it is plausible that the equilibrium in seawater is shifted toward the sorption of these light metalions; whereas in distilled water, the equilibrium tends toward lowering the amount of these sequestered metal ions, which are replaced by protons. An increase of pH during sorption was reported by Tsezos and Volesky ( 7 ) , who explained it as a process of reverse hydrolysis of the metal ion that released OH- groups. Kuyucak and Volesky (6) claimed that the dissolution of cytoplasmic components or the release of carbonate ions from Ascophyllum biomass could perhaps be responsible for the observed increase in solution pH. The similarity of Ascophyllum and Sargassum makes it more likely that this was caused by a binding of the protons from solution during the course of a release of the ionic species that initially were present on the algal biomass. Holan et al. (10) observed a decrease from pH 4.9 to pH 3.5 in sorption experiments with crosslinked Ascophyllum nodosum, without giving an explanation for that phenomenon. Since the biomass was crosslinked under acidic conditions, it appears probable that an excess of protons was released. Crist et al. (13) observed a decrease of pH during sorption that was explained by proton release. The freshwater algae used had been washed at pH 3-3.5 for partial protonation. When formulating a model to describe the sorption behavior, it appears important to incorporate these differences in behavior that are due to the different initial states of the biomass. Otherwise, the sorption model for the equilibrium, which considers the final concentrations of the sorbed metal ions, is still valid only for a specific initial state reflecting the ion exchange capabilities of the biosorbent.

2.5r I

'

0.51

,

A

A,-......'...---

A .,... ,.-.

1

a ' * A

0.1 1 Cd concentration (mM)

0.01

lo

FIGURE 2. Biosorption isotherm for Cd and H binding by initially protonated biomass at pH 3.0 (experimental data 81 the two-site model). 3.51

91

0

4

I

1 ' .8.

0.5

I --e--H"+liQht

I

metals

1 1.5 2 2.5 3 Cd Concentration (mM)

3.5

FIGURE 3. Binding of Cd, H, and light metal ions (Na, K, Mg, Ca) by crude biomass as a function of the Cd concentration.

Relation between Proton Release and Metal Ion Binding. In order to examine whether heavy metal ion binding and corresponding proton release by protonated biomass can be modeled as an ion exchange process, the proton release (eq 3) during the binding of Cd, Cu, or Zn at pH 2.5, 3, or 4.5, respectively, was quantified. The consumption or release of protons due to reactions in the aqueous phase is negligible because modeling with MINEQL+ (14) showed that no hydrolyzed species occur until pH 7 or higher, except for Cu-containing solutions where precipitation may begin at around pH 5. The change in proton binding and Cd binding with increasing metal ion concentration at pH 3 is plotted in Figure 2. It was observed that while proton binding decreases and metal ion binding increases with an increasing metal ion concentration, the total binding on a charge basis ( 4+~q M ) ofthe biomass stays approximatelyconstant. This observation confirms that ion exchange does take place with a metal ion to proton ratio close to 1:2. The same conclusion applies for sorption of Cd byraw biomass, where the release of ions bound from the seawater (Na,Ca, K, Mg) is balanced by the binding of protons and heavy metal ions (Figure 3). The fact that the total metal ion binding by the untreated biomass (Figure 3) was higher than that of protonated biomass (Figures 1 and 2) may be due to a loss of binding sites during crosslinking and acid washing. If all biomass binding sites were initially protonated, it could be expected that any binding of a divalent ion would be associated with the release of exactly two protons. The experimental data, however, do not show such a perfect exchange behavior: a slight increase in the total binding ( q +~2 q C d ) during Cd sorption by protonated biomass was VOL. 29, NO. 12. 1995 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

3051

MODEL 3

e

al

n

-1

2'5

m

-----

H, without Cu in solution H, for total Cu conc of 400 mg/L Cu, for total Cu conc of 400 mg/L

g

2

0

-2

5

1.5

0

n

-3 m -4 -5'

2

3

4

5

6

1

-s

0.5

' -5

PH FIGURE 4. Electrophoretic mobility and the change in binding of Cd and H by initially protonated biomass as a function of pH (the total Cd concentration 0 or 3OOO mg/L).

noticeable (Figure 2). The ratio of proton release to metal ion binding was slightly smaller than 2. This can be explained by the presence of a small number of unprotonated groups that bound metal ions but did not release protons. In the biomass studied, these are likely to be sulfate groups, which are strongly acidic. At pH 4.5, the increase of the total binding ( q ~qCd) with metal ion concentration was more pronounced: it rose from 1.5 mequivlg for [Cd] = 0 mmol/L to 2 mequivlg for ICdl = 6 mmol/L (data not shown). This can be explained by the availabilityof a larger number of sites that were not protonated but ionized or occupied by other ions such as Na. Alginate is partially dissociated at pH 4.5. The amount of protons displaced per metal ion sorbed has been reported to be less than 2. The proton release increased with increasing binding strength of the metal ion to freshwater algal biomass (13). There was no proton release during the binding of sodium, which was interpreted as pointing to electrostatic attraction as the sole sequestering mechanism. The contribution of covalent binding increased with the increasing ability to displace protons. Crist et al. (3)showed that the total charge of the released ions Ca, Mg, and H equaled the charge of the metal ion taken up, demonstrating a perfect ion exchange behavior. This result was obtained for freshwater algae where no sulfate groups were present in their biomass. Therefore, no free groups contributed to sorption through simple adsorption instead of ion exchange. Similar observations were made by Kuyucak and Volesky (2)for cobalt sorption by A. nodosum. Treen-Sears et al. (15) reported that two protons were released for each uranyl ion sorbed by Rhizopus arrhizus in a flow-through packed biosorption bed. Surface Charge. In order to investigate the effect of metal ion binding on the overall charge of the biomass, the electrophoretic mobility (EPM) was studied as a function of pH and metal ion concentration in solution (Figure 4). Between pH 2 and pH 6 , the EPM and therefore the charge of the biomass was always negative, as the direction of the movement indicated. In the metal ion-free solution, the EPM decreased and the quantity of protons released increased with increasing pH. At higher concentrations of a heavy metal ion, however, the EPM remained constant at a small negative value while the metal ion binding increased, and the proton binding decreased with pH. The increase of the magnitude of the EPM with rising pH in a metal ion-free solution can be explained by an increasing number of negatively charged deprotonated sites. The constancy of EPM as a function of pH in the presence of

+

3052 rn ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 29, NO. 12, 1995

n 2

3

4

5

6

7

PH

FIGURE 5. Binding of Cu and H for titration of protonated biomass with 0.1 M NaOH (total Cu concentration 0 or 4 W m g k experimental data and two-site model).

Cd indicated that the charge of the particle remained constant. This means that, at high metal ion concentrations, the charge of protons released with increasing pH equals the charge of metal ions taken up. These results correlate with marine algal biomass findings of Kuyucak and Volesky (6) according to which a strong negative surface charge was present at pH 4-5. The magnitude of the surface charge decreased sharply when the pH was lowered to 3. Collins and Stotzky (16) found for bacteria and yeasts that with a higher pH (around 7) a surface charge reversal occurred in the presence of heavy metal ions (Cd, Cr, Cu, Ni, Zn) that did not take place in the presence of light metal ions (Na, Mg). This phenomenon was explained by a change in the solution speciation from the free hydrated ion to the still positively charged hydrolyzed species, which apparently was sorbed well. With a further increase in the pH, when the dominant hydrolyzed species in solution were neutral or negative, the sorption decreased, and a surface charge reversal back to negative values occurred. In the pH range investigated in this study, hydrolysis was negligible, and no such surface charge reversal could be observed. Titrations. The curve representing the protonation state of the biomass during titration with 0.1 M NaOH in the metal ion-free solution (Figure 5) showed an inflection at pH -4.8. This indicated the presence of an acidic group with an apparent pK-4.8, which constitutes approximately 2.0 mmollg of biomass. The titration of protonated Sargassum biomass in the presence of 400 mg/L Cu (Figure 5 ) showed the protonation of the same number of sites (2.0 mequiv/g) as in the metal ion-free solution. However, in the presence of metal ions, the inflection point shifted from pH 4.8 to pH 3.2. This indicated clearly the competition between metal ions and protons for the same binding sites. As these sites became occupied by metal ions, a higher proton concentration was needed to displace them and to achieve the same binding of protons, Le., the apparent pK is lowered. The change in the metal ion binding with changing pH also equals approximately 2.0 mequivlg. Therefore, it appears that all sites with the pK = 4.8 could be used for metal ion binding, with a ratio of two sites per divalent metal ion, thus conserving the charge of the biomass. This was further demonstrated by the mirror symmetry of the curves for the metal ion binding and proton release. The curve for the metal ion binding does, however, not reach zero at pH 2. Again, this could be explained by metal ion binding to strong acidic groups that do not become

protonated at pH 2 (the pH at which all carboxyl groups should be expected to be protonated). The proportion of this group was estimated from the biomass binding capacity at low pH to be 0.25 mequiv/g, as seen from the titration curve (Figure 5). In addition, this equals the amount of protons released when protonated biomass is equilibrated in distilled water, yielding pH 3.3 (data not shown). An increase in proton concentration in solution by 10-3.3mol/L corresponds to a decrease in proton binding by0.25 mmollg At pH 3.3, strongly acidic sites should be mostly ionized. However, the degree of ionization of the weakly acidic groups (pK 4.8, 2 mequivlg) was according to the model predictions only 0.06 mequivlg (data not shown). Therefore, the amount of protons released is predominatelythat which was bound to sulfate sites. The apparent pK = 4.8 exhibited in the titration of Sargassum biomass is in the range of carboxylicgroup pKs. Crist et al. (4) reported a pKbetween 5 and 7 for the carboxyl groups of marine algae. The pK of pure alginate is lower however: 3.38 for polymannuronic acid and 3.65 for polyguluronic acid, respectively (I 7). Buffle (18) reported pK values between 2.6 and 4.7 for carboxyl groups in different organic compounds. This deviation of the pKs in different compounds indicates the relevance of secondary effects such as different molecular environments. The other, more strongly acidic group, which has a capacity of -0.25 mequiv/g, is likely to be a sulfate group. Crist et al. (4) reported the pKof biomass sulfate groups to be between 1 and 2.5. Availability of Carboxyl and Sulfate Groups. Alginic acid makes up 14-40% of the dry weight of brown algae (1 7). The amount of carboxyl groups in extractable alginic acid in Sargassum was determined as 2.25 mequivlg (19), which corresponds to the number of titrable sites (Figure 5).

Sulfate groups are known to be present in the fucoidan of brown algae, which makes up between 5 and 20% of the dry weight of species belonging to the order fucales to which Sargassum belongs (20). Fourest (19)determined the total number of sulfate groups in Sargassum as 0.27 mequivlg. This is very close to the amount of strong acidic groups that was postulated from the titration and ion binding experiments (0.25 mequiv/g). Relevance of Carboxyl and Sulfate Groups for Metal Ion Binding. Gardea et al. (21) observed that after the blocking of carboxyl groups of algal species by esterification, the binding capacityfor Cu and Al decreased. This decrease was correlated to the degree of esterification. Fourest et al. (22) used the same technique to demonstrate that the Zn binding by fungi was to at least 30-70% achieved by carboxyl groups. Recent results of Fourest (19) show that the blocking with propylene oxide of weakly acidic groups (pK near 5) in S. fluituns, which are likely to be carboxyl groups, was 90% effective and resulted in 80% reduction of the metal ion binding. This demonstrated the responsibility of weakly acidic groups for most of the metal ion binding. The sorption of Pb and Cd to sulfate groups has been described by Veroy et al. (23). The metal ion binding capacity of carrageenan was correlated to the degree of sulfation. The binding is believed to be mainly due to electrostatic attraction between the sulfate ester groups and the metal cation, possiblywith involvement of hydroxyl groups.

Model. For quantitative description of biosorption, it would be very useful to develop a mathematical model capable of reflecting biosorption as a process that involves ion exchange between metal ions and protons as well as binding to unprotonated sites. Specifically,the influence of pH on the active sites has to be incorporated. In order to predict the final metal ion binding from the initial conditions, this model has to allow the calculation of the change of binding of the initially sorbed species. Most mathematical biosorption models used in the literature describe simple Langmuir (10) or Freundlich (7) sorption isotherms, where the metal ion binding is determined as a function of the equilibrium metal ion concentration in the solution, without reference to pH or other ions in the same solution system. Some authors (12, 24) have recognized that ion exchange is an important sorption mechanism, concluding that it is inadequate to assume only binding on free sites. Therefore, the authors used ion exchange models with equilibrium constants that take into account the reversibility of the ion exchange reactionswhere the ion that is displaced into the solution can compete with the sorbed metal ion for the sites. The Langmuir equation and the ion exchange constant for the binding of metal ion M (for simplicity here a monovalent ion) by replacing a proton H on a complexing site C are related as follows: Langmuir C+M=+CM

q = [CMI = K[M][C],/(l + x"[M])

(5)

ion exchange CH

+M

--L

CM

+H

therefore

x" = K/[H]

(7)

The difference between the two approaches is that the first one assumes that all sites are initially free and does not consider any reverse reaction of a displaced ion, in this case a proton, with the site. The second model assumes that all sites to which metal ions are sorbed are initially occupied, Le., the number of free sites stays constant. Crist (12) compared the fit of the Langmuir sorption isotherm model and the one using ion exchange constants. The differences between the two models were especially pronounced at low metal ion concentrations because of the effect of the reverse reaction involving the displaced ion. Correspondingly,it has been postulated that the Langmuir model applies only at higher metal ion concentrationswhere binding of the displaced ion is low. This deficiency of the simple Langmuir model, however, loses significance for the multicomponent Langmuir model used here (eqs 12 and 13) since the effect of competing ions, which is high at low metal ion concentrations, is taken into consideration. This makes the model presented below applicable even at low metal ion concentrations as could also be seen from Figures 1 , 2 , and 7 . The ion exchange approach is perhaps somewhat closer to the reality than the simple Langmuir VOL. 29, NO. 12, 1995 / ENVIRONMENTAL SCIENCE &TECHNOLOGY

3053

model, but it is not completely satisfymg either: While the constant number of free sites may be a reasonable working assumption for a constant pH system, it c6rtainly does not hold for systems with changing pH values, as demonstrated above. According to Stumm (23, the cation exchange capacity increases with increasing pH above the isoelectric point. Therefore, one cannot simply model the competitive binding of metal ions and protons by a metal-proton ion exchange constant. Treen-Sears et al. (15) formulated a theoretical model that included metal ion and proton binding to the active sites. However, the model assumes a 1:l exchange ratio between metal ions and protons, which is clearly not the case. Also, it was only used in avery limited version for the reaction of one metal ion with one or two sites as an empirical tool. Constants were fitted arbitrarily without taking into account the number and nature of groups that were actually present on the biomass (26). One outcome of this approach is that the number of sites postulated in order to fit the model for different metal ions varied from one metal ion to the other. No modeling of pH influence or of the proton binding has been attempted. Sometimes the relationship between the charge of the biomass and pH is described as a linear one, which would lead to a linear relationship between the metal ion binding and pH (13). Only one attempt has been made to model the biosorptive binding of metal ions as a function of pH in this way (27). The author assumed a simplified linear relationship between the metal ion binding and pH difference. This only can be used as a purely empirical model that does not convey information about the real processes occurring, such as the involvement of different groups. This approach is only applicable within a rather limited pH range (comparetitration curve in Figure 5). The model used (27)also cannot represent the change of proton binding in the presence of metal ions. In addition, no indication is given as to how well the model fits the data it is based on. In the modeling of industrial ion exchangers, protons have been considered (28). However,these synthetic resins are chemically rather simple compounds that mostly employ only one active group, making mathematical models describing their performance not applicable in the case of more complex materials used as biosorbents. BuMe (29) systematically described several possible models for complexation equilibria including site protonation, multiple sites, and secondary effects such as variations in site properties, conformation changes with degree of occupation, and influence of the electric field in polyelectrolytes on the local concentration. For simple polymers, the effect of the electric potential has been modeled (30)by assuming that the negative charges of the molecule are uniformly distributed over a spherical particle surface. However, it can be assumed that for algal biomass the charge distribution is neither uniform (since different parts of the cell have a different molecular composition) nor confined to the surface. This makes it very difficult to model the influence of a surface charge on the local ion concentration in the alga. A model is proposed in the present work that is simple and adequately accurate. It is capable of representing the relevant findings about the metal ion biosorption mechanism pertaining to brown marine algae materials. Theoretically, the modeling of adsorption could include terms for electrostatic attraction, specific chemical bonds (com3054

ENVIRONMENTAL SCIENCE &TECHNOLOGY / VOL. 29. NO. 12, 1995

plexation of metal ions, acid base reactions), and solvation energies (8). Although it is likely that at least electrostatic attraction and chemical bonds do contribute to metal ion binding, both phenomenawill be lumped into one constant. Since acid-base reactions that are not electrostatic play an important role and since sorption is known to be specific, the choice is to express the binding through overall chemical equilibrium constants. Additionally, the number of free sites is small compared to the total number of binding sites for pH 54.5, and therefore the surface charge density (and thereby the electrical potential) is small, especially since the charges are not concentrated on the particle surface but distributed throughout the particle volume. It has to be noticed, however, that these chemical equilibrium constants are not thermodynamicallywell defined and that they may deviate from constancy when the activity of an ion in the matrix differs from its activity in the bulk solution (25). For further improvement of the model, a separate term for electrostatic attraction may be considered. Although this also is a simplification, it is assumed that the biomass contains two homogeneous groups that are solely responsible for the binding. From the experimental results for the Sargassum biosorbent, it follows that one site largely consists of carboxyl groups (in the following called group C), while the second group (called S) is referred to a sulfate group (see above). One divalent metal ion reacts with two such monovalent groups. The reactions considered are as follows: H H

+C

CH

KCH = [CHI/ ([HI[Cl)

(8)

+ S * SH

KSH= [SHl/([HI[Sl)

(9)

--L

+

KCM= [CMo,512/([Ml[C12)(10)

+ 2s -- 2SM0,,

KSM= [SM,,,l2/([MI[Sl2)

M2+ 2C * 2CMo,5 M2+

(11)

The formulation ~ C M Ois .chosen ~ instead of C2M in order to stress that, for desorption, two bonds between the metal ion and biomass have to be broken: the case expected for a specific competitive binding (18) where not only electrostatic attraction but also complexation is relevant. The formation constants easily can be transformed into ion exchange constants or selectivity coefficients by dividing eq 10 by the 2-fold of eq 8. It is convenient to use sorption isotherm equations where the binding of one species can be calculated from the final concentrations of all species directly. The above equations can be reformulated into multicomponent Langmuir sorption isotherms as adapted from Hill (31):

Although the mathematical form of these isotherm equations is similar to that where the Langmuir type adsorption of several competing metal ions to free sites is modeled (e.g., ref 32), a very different principle lies behind the formulation. In the present case, the isotherm describes

TABLE 1

Model Parameters for One-Site and Two-Site Sorption Models: Equilibrium Constants K and Total Number of Carboxyl and Sulfate Sites (Ct and SJ8 1 site (case a )

Cd cu

Zn 1 site (case b)

Cd cu

Zn 2 sites

Cd cu

Zn

104.0 105.3 103,9 105 105 105 I 04.8 I 04.8 I 04.8

2.7 2.4 3 2.7 2.7 2.7 2 2 2

3.3 x 103 5.1 105 7.4 x 102 2.0 105 2.9 105 4.5 x 104 8 x IO4 20 x 104 I x 104

0 0 0 0

0 102.5

I 02.5 102.5

0.8 x 103 3.7 x 103 0.5 x 103

0 0.25 0.25 0.25

0.1 1 0.14 0.33 0.14 0.17 0.17 0.046 0.1 1 0.086

Initially protonated biomass of Sargassum fluitamwas used. Model parameters were determined from five to six experimental points for each pH (2.5, 3.0, 4.5) for metal ions Cd, Cu,and Zn, respectively. Standard deviations refer to qu. a

exchange reactions, not a simple competition for free binding sites. The main difference is the consideration of the reverse reaction, which depends on the final concentration of the displaced ion (in this case the proton) in solution (or if pH is controlled it affects the amount of acid or base added),which in turn depends on the initialloading of the biomass with exchangeable species. With the above equations, it is possible to determine the equilibrium bindings of metal ions and protons as a function of the final pH and metal ion concentration. These parameters themselves are, however, not independent variables (although they are often treated as such in isotherm models, since they can be determined directly), but they depend on the initial loading of the sorbent, pH, and concentrations. In order to predict the complete final state of the sorption system (the cation concentrations in the solution as well as bindings) from the known initial state, it is necessary to include the mass balances of all species as additional equations: [MI, = [MI + 0.5[CMo,,l

+ 0.5[SMo,,l

(14)

Since there is no explicit analytical solution to these equations, the calculations are conveniently done by mathematical equilibrium programs that perform the necessary iterations. In this case, the program MINEQLS. (14)has been used for this purpose. The necessary input parameters are the total concentrations of all species (Ht, M,, C,, S,) as well as the equilibrium constants. Determination of Model Parameters. For calculating the sorption model parameters that best fit the data, the average absolute error of all metal ion binding values for n experiments was minimized. In order to verify whether it is necessary t o use a twosite model, or whether a one-site model would be sufficiently accurate, the data were also fitted to a one-site model, (i.e., St = 0). For case a, all three constants were determined individually for each metal ion. In case b, the constants were determined such that the total number of sites Ct and the equilibrium constant KCHare the same for all three metal ions. For both one-site models, the average

error of metal ion and proton binding was minimized, since the KCHvalues still had to be determined. For the two-site model, the parameters Ct, St, and KCHwere fured at those values obtained in the titration experiments. By this process, instabilities in the optimization process could be avoided that might occur if too many parameters are optimized simultaneously. Using this approach, the parameters that characterize the biomass are the same for all three metal ions. KSHwas assumed to be lo2., in order to fit the desorption data of Aldor et al. (33). The parameters for the one-site models and for the two-site model are summarized in Table 1. ModehgofExperimentalData. For the one-site model (Table 1,case a), it was possible to find parameters Ct, KCH, and KCM for each metal ion in such a way that the experimental data could be well represented. However, the values determined for KCH and Ct were obtained by curve fitting, and they did not correspond to the values that were determined experimentally by titration (Figure 5). Furthermore, they were different for each metal ion, although the different metal ions should be using the same sites. When the same values for KCHand C, are used for all three metal ions (Table 1,case b), the accuracy of the onesite model decreased. Specifically,the predictions of the one-site model were too high for pH 4.5 and too low for the low pH values (pH 2.51, because the availability of unprotonated, easily accessible S sites is not given in that model (Figure 1). Therefore, a two-site model should be used, especially if the behavior of different metal ions is to be described by the same model. The two-site model uses only two parameters, KCMand KSM,to fit the metal ion and proton binding for each of the three different pH values (C,, KCH, and St were fixed at the values obtained by titration). Considering the relation between the variety of modeled conditions and the number of parameters, the fit is quite good (Table 11,both for the binding of the heavy metal ions at different pH values (Figures 1 and 2) and for the release ofprotons (Figure2): the standart deviation of0.08 mmol/g corresponds to 7%of the maximum uptake. This is a similar magnitude as the experimental errors (34). The fact that the titration curves are modeled well (Figure 5) shows that the px"s of weakly acidic groups do not differ widely. They can be approximated by a single pKvalue of 4.8. It is therefore not necessary to use affinity spectra for the binding groups. It can also be seen from Figure 5 that the model can be extrapolated over the entire region of pH 2-6.

VOL. 29. NO. 12, 1995 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

3055

IRELEASE

140

120L

A

A

--

0016Ndata 0 0 8 N data - 0 0 1 6 N model

---

0 08 N, model

PH 0

5

0016N,data 45 OOEN, data 0016 N model 08 N, model

-0

w

35

TABLE 2

Variation of Langmuir Sorption Parameters with pHa metal parameter Cd

K

Zn

qmax

[C,][mol/Ll = C,[mol/gl mlV [g/Ll

(18)

[S,l [mollLI = S,[mol/gl ml V [g/Ll

(19)

[MI][mollLl = q,,[mol/gI mlVIgiL1 + [Mli[mollLl

(20)

[H,l[mol/Ll = q,[mol/gl mlV[g/Ll + [Hl,[mollLl + [Hl,dd[mollLl V&IV[-I

(21)

An example of such desorption modeling is shown in Figure 6 for different initial acid concentrations and for solid-to-

liquid ratios as independent variables. The experimental points are from the recent results by Aldor et al. (33). Since the initial proton binding was unknown, it was assumed that it was 1.4mequivlg, which is the value obtained from the model for q~ = 0.296 mmol/g at pH 3.8, as obtained by Aldor (33). The model is able to /

2

3

(mmolig) (Lig) (mmolig) (Lig) (mmol/g) (Lig)

0.3 1.1 0.3 1.5 0.2 0.7

0.6 1.5 0.9 2.3 0.3 1.0

4

PH 5

6

7

1.9 2.5 2.6 2.6 5.0 14 17 18 2.2 2.6 2.7 2.7 7.8 18 21 21 1.3 2.2 2.4 2.4 2.8 9.1 13 13

Initially p r o t o n a t e d b i o m a s s of Sargassum fluitans w a s used

200

Prediction of Final Conditions on a Desorption Example. Modeling attempts of competitive metal ion desorption have been rare. Desorption is frequently achieved by acid washing, in which case it can be represented by this model. The description of the desorption efficiency is complicated by the use of a number of the following characteristic process parameters that are commonly employed: the solid-to-liquid ratio (S/L),the initial metal ion loading of the sorbent, the concentration of the desorbing acid, the equilibrium pH value, the percent of elution, the overall process concentration ratio, and the types of elutants, biomass, and metal ions (9, 33). Given the type of biomass, metal ion, and eluent (acid), it only is necessary to specify KCMand KSMfor the model developed here in order to describe the desorption behavior for any combination of independent parameters. All of the other parameters either are specific, arbitrarily chosen initial conditions or dependent variables that can be calculated from the initial ones by using the model. For any biomass that can be described using this model, the complete desorption behavior is very conveniently characterized by two constants. Thus, it is superfluous to mention the details concerning the initial and final conditions. To calculate the parameters for the equilibrium state (pH, binding, concentrations), a model like MINEQL+ requires only the information concerning the equilibrium constants as well as the total concentrations:

ENVIRONMENTAL SCIENCE &TECHNOLOGY

qmax Kqmax

a

100 150 Solid to Liquid Ratio S/L (g/L)

50

FIGURE 6. Cd release and the final pH during desorption with HCI (0.08 and 0.016 N HCI) as a function of the solid to liquid (SA.) ratio (initial Cd binding 33 mg/g; experimental data of Aldor (33)and the two-site model).

3056

Cu

Kqmax

--0

qmax Kqmax

60

unit

VOL 29, NO. 12. 1995

predict the influence of S/L ratio and initial acid concentration of elution efficiency and final pH. Influence of pH: Comparison of Model to Langmuir Approach. A major advantage of the model presented here in contrast with simple isotherm models is that both the effect of pH on sorption as well as the change of pH during the sorption process can be predicted. Neither the Langmuir nor the Freundlich isotherm model in their simple form include the pH value as one of the variables. If these modeling approaches are used, it is therefore not possible to predict changes in proton binding. Also, the respective constants have to vary with pH since binding changes with pH (35). To illustrate this by an example, the variations of the Langmuir constants q m u (correspondingto the number of sites C,) and K (related to the equilibrium constant) are summarized in Table 2 for different pH values. Since the simple Langmuir model and the modified Langmuir model described in this work give different sorption isotherm shapes, the Langmuir parameters were determined such that the Langmuir sorption isotherm equals the predictions of the presented model both at [MI = 0.1 mmol/L (representing the initial slope region) and at [MI = 10 mmol/L (representing the maximum saturation region). Note that although an accurate experimental determination of the Cu binding above pH 5 is not possible due to precipitation, the equilibrium relationship of eq 12 between free Cu*+in solution and sorbed Cu is still valid. The respective reaction constants are unchanged, only the concentration of free ions in the solution is lowered due to precipitation. The Langmuir parameters in Table 2 have been derived by approximating eq 12, not the experimental data. Consequently, they are also valid at the higher pH values where they could not be determined experimentally. It is often stated in the literature that there is little difference between biosorption at pH 4 and 5 (6, 7, 36). Indeed, qmaxdoes not change much above pH 4.5. The product (qmuK),which corresponds to the initial slope of the isotherm, however increases up to about pH 6. This means that for high metal ion concentrations pH 4.5 is already optimal. For low metal ion concentrations, however, a significant increase in the metal ion binding can be achieved by raising the pH to 5.5 or 6, which is shown in Figure 7. The modeling exercise confirmed that sorption of Cd and Zn in the low concentration range can be optimized by increasing the pH to 6, where still no hydrolysis or precipitation occurs for these metal ions. For Cu, pH 4.5 can be maintained as optimal because of the danger of precipitation for pH '5. When studying the biosorption process, it is essential to maintain pH values below the metal ion precipitation range in order to avoid its unpredictable effects on the metal ion removal from solution.

initial total Molecular Species carboxyl and sulfate binding sites c, s H protons M metal ion CH, SH protonated sites CM, CMo.5, metal ion-biomass complexes SM0.5 i t

-1

f Y

or00 1

0.01

0.1

1

Cd Concentration (mM)

FIGURE 7. Isotherms of Cd for different pH values at low concentrations (experimental data and the two-site model).

However, under process conditions where the maximum metal ion removal is desired, precipitation may increase the overall metal ion removal efficiency,which is reflected in the elevated overallprocess metal ion concentration ratio values achievable.

Conclusion The exchange between metal ions and protons was the major binding mechanism in biosorption by protonated biomass. Experiments showed the relevance of two specific biomass binding sites in metal ion sequestering. One weakly acidic site, with a pK around 4.8,was responsible for the main part of the binding, and a strongly acidic site was present to a smaller extent. The contribution of the strongly acidic group to metal ion binding was significant at low pH. A model that considers reversible binding of metal ions and protons to two types of binding sites was used to describe metal ion-proton exchange. The binding of metal ions and protons could be predicted as a function of metal ion concentration and pH. It was possible to use the same model for both adsorption and desorption. It is recommended to use the two-site model instead of Langmuir or Freundlich isotherm models in order to accommodate the significance of ion exchange in biosorption. Although the version of the multicomponent Langmuir model presented here (eqs 12 and 13) applies for the sorption of divalent ions or protons onto sites that are free or occupied by the other respective ion, the model can be easily modified to account for different competing ions. In that form, it can be even suited for other than protonated biosorbents where ion exchange is predominant.

GIossary ct, st EPM K

m 4 S/L V

[I Indices ad add des f

amount of binding sites C and S (mol/g) electrophoretic mobility (m2s-l V-l) equilibrium constant (formation) (Llmol) dry weight of biomass (g) binding (mequiv/g) solid to liquid ratio (g/L) volume of solution (L) concentration (of species within the brackets) (mol/L) adhering added desorbed final

Acknowledgments The DAAD (German Academic Exchange Service) HSP2 scholarship for S.S. is gratefully acknowledged.

literature Cited (1) Kuyucak, N.; Volesky, B. Biorecouery 1989, 1, 189-204. (2) Kuyucak, N.; Volesky, B. Biotechnol. Bioeng. 1989,33,823-831. (31 Crist, R.H.; Martin, J. R.;Guptill, P. W.; Eslinger, J. M.; Crist, D. R. Environ. Sci. Technol. 1990, 24, 337-342. (4) Crist, R.H.; Oberholser, K.; McGarrity, J.; Crist, D. R.; Johnson, J. K.; Brittsan, J. M. Environ. Sci. Technol. 1992, 26, 496-502. (5) Mackie, W.; Preston, R.D. In Algal physiology and biochemistry; Stewart, W. D. P., Ed.; Blackwell Scientific Publications: Oxford, 1974; p p 40-85. (6) Kuyucak, N.; Volesky, B. Biotechnol. Bioeng. 1989,33,809-814. (7) Tsezos, M.; Volesky, B. Biotechnol. Bioeng. 1981, 23, 583-604. ( 8 ) Pagenkopf, G. K. Introduction to natural water chemistry;Marcel Dekker: New York, 1978; p p 161-167 and 220-230. (9) Kuyucak, N.; Volesky, B. Biotechnol. Bioeng. 1989,33,815-822. (10) Holan, 2. R.; Volesky, B.; Prasetyo, I. Biotechnol. Bioeng. 1993, 41, 619-8725. (111 Darnall, D. W.; Greene, B.; Henzl, M. T.; Hosea, J. M.; McPherson, R. A.; Sneddon, J.; Alexander, M. D. Environ. Sci. Techno[.1986, 20, 206-208. (12) Crist, R. H.; Martin, J. R.;Carr, D.; Watson, J. R.;Clarke, H. J.; Crist, D. R. Enuiron. Sci. Technol. 1994, 28, 1859-1866. (13) Crist, R. H.; Oberholser, K.; Shank, N.; Nguyen, M. Environ. Sci. Technol. 1981, 15, 1212-1217. (14) Schecher, W. D. MINEQLI: A Chemical Equilibrium Modelfor Personal Computers, UsersManual Version 2.22 Environmental Research Software, Inc.: Hallowell, ME, 1991. (15) Treen-Sears, M. E.; Volesky, B.; Neufeld, R.J. Biotechnol. Bioeng. 1984, 26, 123-129. (16) Collins, Y. E.; Stotzky, G. Appl. Enuiron. Microbiol. 1992, 58, 1592- 1600. (171 Percival, E.; McDowell, R. H. ChemistryandEnzymologyofMarine Algal Polysaccharides;Academic Press: London, U.K., 1967; p p 99-124. (18) Buffle, J. Complexation reactions inaquaticsystems: Ananalytical approach; Ellis Horwood Ltd.: Chichester, U.K., 1988; pp 156157, 284-286, and 323. (19) Fourest, E.; Volesky, B. Enuiron. Sci. Technol., in press. (20) Chapman, V. 1. Seaweeds and their uses; Chapman and Hall: London, 1980; pp 195-240. (21) Gardea-Torresday, J.; Becker-Hapak, M. K.; Hosea, J. M.; Darnall, D. W. Enuiron. Sci. Technol. 1990, 24, 1372-1378. (22) Fourest, E.; Serre, A.; R o w J.-C. Toxicol.Enuiron. Chem., in press. (231 Veroy, R. L.; Montano, N.; Guzrnan, M. L. B.; Laserna, E. L.; Cajipe, G. J. B. Bot. Mar. 1980, 22, 59-62. (241 Haug, A.; Srnidstod, 0. Acta Chem. Scand. 1970, 24, 843-854. (25) Stumm, W.; Morgan, 1. J. Aquatic Chemistry; J. Wiley & Sons: New York, 1970; pp 445-513. (26) Sears, M. E. Ph.D. Thesis, McGill University, Montreal, Canada, 1986. (27) Votruba, J. In Abstracts: IUMS Congress;IUMS: Prague, Czech Republic, 1994.

VOL. 29, NO. 12, 1995 /ENVIRONMENTAL SCIENCE & TECHNOLOGY

3057

(28) Marcus, Y. In Ion exchange, a series of advances;Marinsky, I. A,, Ed.; Marcel Dekker: New York, 1966; pp 101-138. (29) Buffle,J. Complexation reactions inaquaticsystems: Ananalytical approach; Ellis Horwood Ltd.: Chichester, IJ.K., 1988; pp 199200, 216-258, and 296-299. 1301 Marinskv, J. A. Coord. Chem. Rev. 1976. 19. 125-171. (31) Hill, C. G. J. An Introduction to Chemical Engineering Kinetics and Reactor Design; J. Wiley & Sons: New York, 1977; pp 167204.

(32) Chong, K. H.; Volesky, B. Biotechnol. Bioeng. 1995,47,451-460. (33) Aldor, I.; Fourest, E.; Volesky, B. Can. 1. Chem. Eng. 1995, 73, 516-522. (34) Schiewer, S.; Volesky, B. Biotechnol. Techniques, submitted for publication.

3058 a ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 29, NO. 12, 1995

(35) Schiewer, S.; Fourest, E.; Chong, K. H.; Volesky, B. InProceedings of the International Biohydrometallurgy Symposium: Vina-delMar, Chile, November 1995. (36) Chong, K. H.; Volesky, B. Biotechnol. Bioeng., submitted for publication.

Received for review April 18, 1995. Revised manuscript received August 4, 1995. Accepted August 4, 1995.@

ES950264Z Abstract published in AdvanceACSAbstracts, September 1, 1995.