You are on page 1of 8

Journal of Colloid and Interface Science 240, 18 (2001) doi:10.1006/jcis.2001.7616, available online at http://www.idealibrary.

com on

Microcystin-LR Adsorption by Activated Carbon


Phillip Pendleton,1 Russell Schumann, and Shiaw Hui Wong
Porous Materials Research Group, School of Chemical Technology, University of South Australia, Mawson Lakes, SA 5095, Australia Received April 17, 2000; accepted April 13, 2001; published online June 21, 2001

We use a selection of wood-based and coconut-based activated carbons to investigate the factors controlling the removal of the hepatotoxin microcystin-LR (m-LR) from aqueous solutions. The wood carbons contain both micropores and mesopores. The coconut carbons contain micropores only. Conrming previously published observations, we also nd that the wood-based carbons adsorb more microcystin than the coconut-based carbons. From a combination of a judicious modication of a wood-based carbons surface chemistry and of the solution chemistry, we demonstrate that both surface and solution chemistry play minor roles in the adsorption process, with the adsorbent surface chemistry exhibiting less inuence than the solution chemistry. Conformational changes at low solution pH probably contribute to the observed increase in adsorption by both classes of adsorbent. At the solution pH of 2.5, the coconut-based carbons exhibit a 400% increased afnity for m-LR compared with 100% increases for the wood-based carbons. In an analysis of the thermodynamics of adsorption, using multiple temperature adsorption chromatography methods, we indicate that mLR adsorption is an entropy-driven process for each of the carbons, except the most hydrophilic and mesoporous carbon, B1. In this case, exothermic enthalpy contributions to adsorption also exist. From our overall observations, since m-LR contains molecular dimensions in the secondary micropore width range, we demonstrate that it is important to consider both the secondary micropore and the mesopore volumes for the adsorption of m-LR from aqueous solutions. C 2001 Academic Press Key Words: microcystin-LR; adsorption; activated carbon; porous solids.

INTRODUCTION

Cyanobacteria (blue-green algae) produce high levels of hepatotoxic chemicals with LD50 values in the low micrograms per kilogram range in the drinking water supply (1). Some of these toxins lead to liver enlargement; prolonged exposure leads to death. The most frequently occurring cyanobacteria hepatotoxins consist of heptapeptides and pentapeptides. The species Microcystis produces the heptapeptide toxin microcystin (1). Cyclic heptapeptide microcystins consist of ve invariant amino acids and two variable amino acids. The in-

1 To whom correspondence should be addressed. Fax: +61-8-8302-3668. E-mail: p.pendleton@unisa.edu.au.

variant amino acids are in the D-conguration and the variable amino acids are in the L-conguration. The ve invariant amino acids are D-alanine (Ala), D-erythro- -methylasparic acid (Masp), D-glutamic acid (Glu), N-methyldehydroalanine (Mdha), and 3-amino-9-methoxy-2,6,8-trimethyl-10-phenyldeca-4,6-dienoic acid (Adda). The two variable amino acids may be leucine (L), arginine (R), tyrosine (Y), and analine (A) or methionine (M) (1, 2). Lanaras et al. suggest that when microcystin-LR (m-LR) is solvated in water the solvated volume is 2.63 nm3 and the solvated area is 1.8 nm2 . The longest molecular length in m-LR, shown in Fig. 1, is approximately 1.9 nm, as determined from molecular modeling using covalent bond length data (1, 2). Microcystin-LR is the most common Microcystis toxin found in Australian drinking water supplies. Activated carbon (AC) removes it effectively via adsorption (3). Generally, peptides exhibit hydrophobic properties in aqueous media, suggesting that the primary interaction between m-LR and an AC is via hydrophobic or dispersion forces. It is not unreasonable to expect that the three charged groups also experience electrostatic interactions with polar surface sites on the carbons during adsorption. Lahti and Hiisvirta (4) demonstrated that small doses of powdered AC do not signicantly affect m-LR solution concentrations. In contrast, Donati et al. (5) investigated the rate and effectiveness of m-LR adsorption by several different powdered ACs and presented its removal as a direct proportionality of adsorbent mesopore volume. The inclusion of this pore range is not surprising considering m-LRs molecular dimensions. The adsorption of relatively simple organic molecules from solvent or aqueous solution very much depends on the adsorbent surface oxide concentration (610). The solution pH also affects adsorption by ACs, due to adsorbent surface chemical group preferential polarization (11, 12). Recently, Jiang et al. (13) examined the effect of pH in the range 5 to 7 on humic acid adsorption by ACs. Essentially they demonstrated that the adsorption afnity for humic substances decreased with increasing pH, interpreting their results in terms of the complex structure and solution properties. Solubility limits leading to precipitate adsorbent interactions may be responsible for mass depletion in solution at lower pH. Additionally, at higher pH, humic acid functions more like polyelectrolytes in a good solvent, exhibiting expanded dimensions, thereby reducing the available adsorbent
1 0021-9797/01 $35.00
Copyright C 2001 by Academic Press All rights of reproduction in any form reserved.

PENDLETON, SCHUMANN, AND WONG

EXPERIMENTAL

Materials PICA (Australia) supplied ve ACs as three wood-based carbons and two coconut-based carbons. The wood-based carbons are phosphoric acid-activated while the coconut-based carbons are steam-activated. We also used a nonporous carbon sold as a specic surface area standard for calibration purposes (from Micromeritics, Australia). Prior to their use all of the carbons were washed with organic molecule-free Milli-Q (18 M resistance) quality water. The wash water was monitored until the efuent conductivity matched the inuent water conductivity. The carbons were then ltered and dried overnight in an oven at 125 C. The three wood-based carbons used are coded B1, F, and N; the coconut-based carbons are coded A1 and P. Calbiochem-Novabiochem Corp. (San Diego, CA) supplied the m-LR, and BDH (Australia) the HPLC-grade acetonitrile. Both were used without further purication. All other chemicals used were analytical grade. All aqueous solutions were made using organic molecule-free water with a resistance of 18 M . Methods
FIG. 1. The structure of microcystin-LR: (a) structural formula; (b) typical structural dimensions.

pore volume and, consequently, leading to a decrease in observed amount adsorbed. All drinking water treatment processes involve competitive adsorption between the adsorptive of interest and many other dissolved species classied as humic substances or dissolved organic matter (DOM). Such solution mixtures represent complex systems that are difcult to interpret unequivocally. Previous literature (4, 5) discussing m-LR adsorption from solution focuses on adsorbent porosity only, with the authors expressing satisfaction with a correlation between mesopore volume and the maximum amount adsorbed by each adsorbent. Unfortunately, in the latter work, the isotherms are drawn as curve-tted lines with no actual adsorption data to indicate the precision of the measurements. Although much research has been published on m-LR adsorption, we believe that it is important to develop a comprehensive understanding of the m-LR adsorption process from a relatively simple solution prior to attempting to understand adsorption from solutions that begin to more closely mimic actual drinking water solutions. In the present work we contribute to the body of knowledge about m-LR adsorption by ACs by investigating the effects of adsorbent surface chemistry and porosity as well as solutionphase chemistry as separate variables. Additionally, to interpret the adsorption process we use a novel solvent system to dene approximate various thermodynamics parameters for m-LR adsorbed by selected ACs. From these data, we interpret the adsorption process in terms of the role of the adsorbent surface chemistry and porosity.

Adsorbent characterizationsurface area and porosity. We used nitrogen gas adsorption at 77 K to dene the pore size distribution of each adsorbent. The stainless-steel apparatus employed for these measurements exploits the volumetric adsorption principle (14). The apparatus and methods employed are described in detail by Badalyan et al. (15). We used a nonporous carbon black specic surface area standard (from Micromeritics, Norcross, GA) to calibrate the precision of the overall equipment and calculated data. We degassed each adsorbent prior to nitrogen adsorption by heating at 1 C/min from room temperature to 125 C. At this temperature the sample was heated for 16 h, resulting in a background pressure over the sample of <1 mPa. Two high-precision, calibrated, MKS Baratron differential manometers with maximum differential pressures of 0.133 and 133.3 kPa ensured high-precision pressure readings over the entire adsorption pressure range. The liquid nitrogen level surrounding the sample was held constant 0.1 mm during the entire adsorption measurements for each sample. Adsorbent characterizationsurface chemistry. The Australian Water Quality Center (SA, Australia) performed surface titration measurements for each carbon. The method employed is well documented by Newcombe et al. (16). The Chemical and Micro Analytical Services Pty. Ltd (Belmont, VIC) performed the oxygen analyses using hydrogen reduction techniques. Solution adsorption isothermspure water solutions. 1.65mL amount of pure water added to a glass vial containing asreceived m-LR, 500 g, gave a stock solution with a concentration of 300 mg/l. Ultraviolet light absorbance at 238 nm combined with an extinction coefcient of 104.6 (17) yielded the exact adsorptive concentration.

MICROCYSTIN-LR ADSORPTION BY ACTIVATED CARBON

Adsorption by the wood-based carbons occurred from initial adsorptive concentrations ranging from 0.07 to 3.03 mg/l. For the coconut-based carbons the concentration range was lower, 7 to 288 g/l. In each case, the solids concentration was 8.0 0.1 mg/l. In all adsorption experiments the total solution volume was 5 ml. We sealed the vials containing the carbon and m-LR and mixed the contents thoroughly by rotating the vials at 25 C. After 72 h contact, we separated the solid and liquid phases by ltration through a 0.22-m nylon lter. We used a Millipore HPLC with tunable absorbance and conductivity detectors to determine accurately the equilibrium m-LR solution concentration. The conditions for adsorptive analysis were as follow: column, Supelcosil LC-AB2 15 cm 4.6 mm (from Supelco); solvent, 30% acetonitrile and 70% potassium phosphate buffer (pH 7.2); ow rate, 1 ml/min. Solution adsorption isothermsadsorbent surface chemistry effects. We modied the surface chemistry of carbon B1 by cautious thermal reduction in an argon environment. The details of this treatment are described elsewhere (7). The reduced samples, coded B1-600H1 and B1-1000H1, signify thermal reduction at 600 and 1000 C for 1 h. The m-LR adsorption procedures and analyses were the same as those above. As a further test of surface chemistry effects, we also measured solution adsorption isotherms from solutions at pH 2.5. The adsorption procedure reproduced the pure water method except that the solution phase also contained sufcient 12 M hydrochloric acid to give a solution phase pH of 2.5. Solution adsorption isothermsthermodynamics of adsorption. (i) Column preparation. The activated carbon HPLC columns used in this work consist of 2555 mg of AC packed into stainless-steel columns measuring 15 mm 3.2 mm (nominal i.d.). After being weighed, the dry, packed column is wetted by a ow of pure methanol with a volumetric ow rate of 0.4 ml/min, over a period of 30 min. The total void volume, V0 , may be determined from the difference in mass of wetted and dry powder. We also followed Yuns method to determine V0 using CD3 OD as the wetting uid (18). Dead volume corrections were also made using a zero dead volume union. Retention time analysis of the series of alcohols ranging from ethanol to n -octanol yields the mobile phase volume, Vm . The stationary phase volume, Vs , the volume of solvent retained or adsorbed by the AC surface, is determined as V0 Vm . Column phase ratios, , are then readily determined from the ratio Vs / Vm . (ii) Measurement of microcystin-LR retention as a function of temperature. Injection of 10 l of a 75 mg/l solution of m-LR in methanol allowed us to determine the adsorption (as a retention) of m-LR at various column temperatures. For all carbons except B1, the solvent system consisted of methanol and 10% dichloromethane (DCM) and 0.1% triuoroacetic acid. The B1 surface adsorbed m-LR too strongly to recover the adsorptive in 10% DCM. This problem was overcome by measuring adsorption from solutions containing 50, 40, 30, and 20% DCM. Linear extrapolation of log capacity factorDCM concentrations at each temperature gave retention data for 10% DCM solutions. The

plots are linear for DCM concentrations below 40%. Retention times for 50% DCM approached mobile phase retention times, greatly reducing the accuracy of the capacity factor data. Four separate temperatures were employed at 10 C intervals over the range 535 or 1040 C. A UV detector tuned to 238 nm gave adequate sensitivity to detect m-LR. An average of three retention times raised the precision of the adsorption data.
RESULTS AND DISCUSSIONS

Activated carbon is the adsorbent of choice in most commercial adsorption separation processes due to its performance relative to its cost. Most suppliers appreciate that a high specic surface area available for adsorption is important for removal of a particular adsorptive. Based on recent 2-methylisoborneol (MIB) adsorption analyses (6, 7), we suggest that this assumption must be made with caution. An adsorbent may exhibit a high specic surface area from nitrogen adsorption; however, the surface area available for adsorption of a particular adsorptive may be quite small due to so-called molecular sieve effects. For solution adsorption processes, the solvent is generally regarded as passive. This assumption must also be made cautiously (7). Adsorption of the relatively hydrophobic molecule MIB (6) shows that the solvent (water) molecules are more readily displaced from low oxygen content ACs than from high oxygen content ACs. For m-LR adsorption it is important to recognize rst that m-LR is a large molecule (Fig. 1) and second that it is a complex aggregate of amino acids rendering hydrophobic character to its aqueous solution properties. Consequently, the correct selection of an AC for m-LR removal from an aqueous solution, prior to any adsorption measurements, requires an appreciation of these two properties combined with a detailed knowledge of the adsorbents physical and surface chemical properties. Adsorbent Characterization The nitrogen adsorption isotherms for a sample of wood(B1) and coconut-based (A1) ACs are reproduced in Fig. 2. Both are typical of the class of samples used in this work. All the ACs exhibit the classical enhanced adsorption afnity for nitrogen at pressures well below 0.01 p 0 , indicating the presence of micropores. The wood-based carbon isotherms are Type IV, exhibiting desorption hysteresis over the entire mesopore pressure range, 0.4 p / p 0 < 1.0. Although the coconut-based carbons exhibit very slight desorption hysteresis, the isotherms are classied as Type I, with the mesopore volumes determined to be just within the pore volume analysis precision, and regarded as a minor contribution to the total pore volume. An s analysis of each adsorbent rendered micropore volumes (19). The use of high-precision adsorption data allows us to classify the micropore volume into primary micropores (of width, w 0.7 nm) and secondary micropores (0.7 nm w 1.6 nm) (20). The mesopore volume results from subtracting

PENDLETON, SCHUMANN, AND WONG

FIG. 2. Nitrogen adsorption isotherms on coconut (A1, 3) and wood (B1, ) activated carbons. Open symbols represent adsorption.

the micropore volume adsorbed (as a liquid volume) from the total volume adsorbed at p / p 0 1.0. Table 1 summarizes these volume data for each solid. Table 1 includes the bulk oxygen content for each carbon. Clearly the wood-based carbons contain more oxygen than the coconut-based carbons. The heat-treated samples of B1 show a decrease in oxygen content with increasing reduction temperature, reaching, at 1000 C, an oxygen content similar to that of the most hydrophobic carbon, sample A1. Interestingly, the reduction conditions are sufciently mild that no further activation nor pore volume changes occur with increasing reduction temperature (7). Infrared spectroscopy analyses of these carbons (21) show that the oxygen in the wood-based ACs exists as hydroxyl, phenol, or lactol groups as well as the expected ethers, carbonyls, quinones, and lactones (22). Since these wood-based ACs are activated with phosphoric acid, a proportion of the oxygen is also attributed to surface phosphate groups. In contrast, the coconut-based ACs contain only the latter functional groups, and no surface OH groups.
TABLE 1 Summary of Activated Carbon Physical Properties
Micropore vol, ml(liq)/g Carbon B1 F N A1 P B1-T600H1 B1-T1000H1 Primary vol 0.52 0.46 0.42 0.68 0.29 0.54 0.51 Secondary vol 0.33 0.50 0.29 0.13 0.10 0.33 0.31 Mesopore vol, ml(liq)/g 0.40 0.38 0.26 0.05 0.07 0.41 0.40 pHpzc 3.4 3.4 3.4 6.2 6.2 Oxygen content, % 14.9 9.2 11.4 3.1 5.7 11.2 3.5

Surface charge titration gives us an important indication of how any (surface) ionizable groups respond to pH changes (23). The point of zero charge (pzc) represents a pH condition where the surface ionic groups are neutralized, to give an effectively uncharged surface. The pHpzc for each AC is given in Table 1. The pHpzc for the two classes of carbon may be grouped into two approximately constant pH values. We suggest that this is a consequence of the method and conditions of activation. The pHpzc for the wood-based carbons is 3.4 and that for the coconutbased carbons is 6.2. From these data we conclude that the type of ionizable groups on each class of carbon is constant. The change in surface charge with pH indicates that the quantity of such groups is also quite different for each class of carbon. The change for the wood-based carbons is considerably stronger than that for the coconut-based carbons. The surface charge for carbon F ranges around 0.27 0.28 mequiv/g while those for carbons B1 and N are approximately 0.55 0.75 mequiv/g over the pH range 39.5. In conjunction with the infrared spectroscopy analyses, we conclude that the latter carbons contain a higher surface concentration of hydroxyl or phenol groups than carbon F. In contrast, the surface charge is 0 0.2 mequiv/g for both coconut-based carbons. A solution pH decrease results in surface hydroxyl group protonation for the wood-based ACs. Infrared spectroscopy indicates that no such groups exist on the coconut-based AC surface. Consequently, protonation of carbonyl, lactone, and quinone groups must occur. These same groups exist on the wood-based ACs, which will experience similar pH effects. In contrast, the increase in solution pH above 3.4 for the woodbased ACs, up to pH 9.3, results in a relatively strong deprotonation or hydroxyl group adsorption, or both, to render a surface charge of 11.3 mequiv/g. The lower oxygen content of the coconut-based ACs offers fewer deprotonation opportunities, manifest as a reduced effect on surface charge decrease. Up to similar basic conditions, the surface charge only decreases to

MICROCYSTIN-LR ADSORPTION BY ACTIVATED CARBON

FIG. 3. Adsorption isotherms of m-LR on selected activated carbons at pH 6.5: carbon B1 (), carbon N (), and carbon A1 ().

0.2 mequiv/g. The adsorption isotherms dened at normal conditions occurs at pH 6.5, where the coconut-based ACs are, essentially, of neutral charge, with the wood-based adsorbents displaying a net negative charge. Solution Adsorption Isotherms Figure 3 shows how the wood-based carbons exhibit a higher afnity for m-LR during adsorption from pure water at pH 6.5 than the coconut-based carbon. This result is similar to recent reports (5, 24). The isotherms were reasonably well tted using the Langmuir adsorption isotherm model. The calculated monolayer equivalent amount adsorbed and the average maximum adsorbed by each carbon are summarized in Table 2. Generally, a good agreement exists between each set of data. In fact, it is interesting to note that the amount of m-LR adsorbed by the microporous coconut-based carbons is very similar to that for the nonporous carbon black, suggesting that the micropores in these carbons offer only a nominal internal surface for adsorption. The effect of surface chemistry on m-LR adsorption by AC is probed by reducing the oxygen content of carbon B1, without affecting its various pore volumes (shown in Table 1). In these measurements, m-LR is exposed to ACs with increasing hydrophobic character, but a constant volume space available for adsorption. It is reasonable to presume that the inuence of surface chemistry applies at low adsorptive concentrations;

FIG. 4. Adsorption of m-LR by carbon B1 and surface chemistrymodied B1: carbon B1 (), carbon B1-600H1 (), and carbon B1-1000H1 ().

hence, the equilibrium concentrations in Fig. 4 are considerably lower than the equilibrium data shown in Fig. 3. The average amount adsorbed supported by an equilibrium concentration of 0.15 mg/l m-LR is essentially constant at 177 12 g/mg. We suggest that the adsorption isotherms are indistinguishable with respect to the oxygen content of the adsorbent. In contrast with the oxygen content of the adsorbents, the solution chemistry affects the m-LR adsorption process considerably. In Fig. 5, we present the adsorption isotherms at pH 2.5. Comparing these data with those in Fig. 3, obtained at pH 6.5, we see that the equilibrium amount adsorbed, supported by an equilibrium solution concentration of 0.1 mg/l, increases by 100% for the carbons B1 and N, and by 400% for the coconut-based A1. From the surface charge and pzc data in Table 1, we note that by reducing the pH from 6.5 to 2.5 we render all of the carbons surface positive; i.e., proton adsorption occurs on the available surface hydroxyl groups. The infrared spectra of these carbons indicate (21) that the wood-based carbons contain both hydroxyl groups and carboxylic acid groups. The coconut-based carbons

TABLE 2 Average Maximum Amount m-LR Adsorbed and Monolayer Amount Adsorbed
Carbon B1 F N A1 P Carbon black Max ads, g/mg 189 10 200 16 161 12 22 2 72 12 7 Monolayer amt ads, g/mg 196 204 172 23 9 9

FIG. 5. Adsorption isotherms of m-LR on selected activated carbons at pH 2.5: carbon B1 (), carbon N (), and carbon A1 ().

PENDLETON, SCHUMANN, AND WONG

TABLE 3 Summary of Linear Correlation Coefcients for Pore Volume Analysis


Pore volume type Primary micropore Secondary micropore Total micropore Mesopore 2 micropore + mesopore Total pores Linear correlation coefcient (intercept set at 12.03 g/mg) 0.21 0.84 0.54 0.93 0.92 0.70

tainly milligram) quantities of adsorptive. Milligram quantities of m-LR are prohibitively expensive. HPLC techniques offer alternative methods for enthalpy (and entropy) evaluation. Adsorptive retention on a particular stationary phase (in this case the adsorbent ACs) determined over a range of temperatures gives adsorption enthalpy data for microgram quantities of adsorptive. The equilibrium between the solution phase and carbonadsorbed m-LR is described by the material balance equation below, leading to an expression for the equilibrium constant, K , m-LR (solution) + activated carbon m-LR (Adsorbed)AC. [1] In examining the equilibrium described by Eq. [1], it is necessary to detect the adsorptive in the solution phase to dene the equilibrium constant K and hence the thermodynamic parameters H and S . For adsorption from an aqueous solution, the equilibrium lies so far to the right that m-LR concentrations are below the lower detection limits of our equipment. Consequently, K cannot be adequately determined. To overcome this problem, we chose a mixed solvent system offering increased solubility in the solution phase, thereby increasing the equilibrium solution concentration. Methanol offers sufcient polarity that it reasonably models water and adequate solubility, allowing elution of the adsorbed phase. The small quantity of DCM added to the methanol gives increased elution strength of the solvent. The addition of triuoroacetic acid, typically used as a buffer for peptide chromatography, gave readily detectable elution peaks. While this solvent system is very different from a pure aqueous solution, analysis of the adsorption of m-LR by the same ACs will still be informative in our goal to understand m-LR adsorption from aqueous solutions. Table 4 shows the m-LR adsorption enthalpy and entropy, and other state variables, for adsorption by the various ACs at 25 C. Conversion of S to T S allows a direct analysis of the relative contributions of enthalpy and entropy to the adsorption free energy, G , leading to the adsorption equilibrium coefcient and an indication of the propensity for m-LR adsorption. An important observation from these data is that the adsorption
TABLE 4 Thermodynamic Values for m-LR Adsorption by Various ACs (Solvent System Used, 10% DCM:90% MeOH:0.1% TFA)
Carbon B1b F A1 P
a

contain only the latter functional groups. Hence, decreasing the pH ensures that these sites are positively charged. Of course, the adsorptive is also affected by pH changes. During a decrease in pH, several scenarios may be considered, each leading to an increased amount adsorbed. When the pH is lowered from 6.5 to 2.5, the free -carboxyl groups in the glutamic acid (Glu) and -methylasparic acid (Masp) residues of m-LR will, to a large extent, be protonated, causing a decrease in the m-LR water solubility. Subsequently, we expect an increase in m-LR afnity for the carbon surface. A pH change causes molecular shape to change (2). Let us consider that m-LR behaves as a polyelectrolyte in solution. At low pH, it may behave as a lament while at high pH, the structure may become an open net, offering larger dimensions or cross-sectional area. We suggest that at low pH, the intramolecular forces within the molecule strengthen, reducing overall molecular dimensions. From the discussion of adsorbent surface chemistry effects on m-LR adsorption, we suggest that physical rather than chemical effects affect the adsorption process. This conclusion is drawn from the large increase in amount adsorbed by the coconut-based carbons, which essentially offer considerably smaller pore dimensions and volumes for adsorption. Weak ionic interactions may participate in the adsorption of m-LR by the AC surfaces in the form of an association of the positively charged arginine side chain of the toxin with the negatively charged carbon surface. The surface charge titration data indicate that the carbons move from a negatively charged surface at pH 6.5 to a neutral charge (pzc) at 3.4, indicating an increase in positive charge. On lowering the pH to 2.5, a decrease of the ionic interaction should be observed; however, the overall effect of lowering the pH is an increase in adsorption. Therefore, it is suggested that the ionic interactions play a minor role in m-LR adsorption by these carbons. These observations again indicate that surface chemistry has a negligible effect on m-LR adsorption by these particular ACs. We need to re-examine the physical effects that control their removal from solution, in terms of the pore volume distribution and its suitability for adsorption. Displacement enthalpy measurements help to distinguish the role of adsorptive and solvent effects on adsorption (610). Unfortunately, such measurements usually require reasonable (cer-

H, kJ mol1 8.9 1.0 16.0 0.8 2.4 0.1 0.3 0.9

S, kJ mol1 18 17 88 27 23 1 20 14

T S ,a kJ mol1 5.4 5.0 26.2 8.0 6.8 0.1 6.0 4.2

G ,a kJ mol1 14.3 6.0 10.2 8.8 4.4 0.2 5.7 5.1

Ka 320 134 61 53 61 10 9

At 298 K. Values extrapolated from measurements made at 50, 40, 30, and 20 parts DCM.
b

MICROCYSTIN-LR ADSORPTION BY ACTIVATED CARBON

enthalpy for each AC, except for carbon B1, is positive. Thus, the driving force for adsorption is not enthalpic. Second, the adsorption entropy for each AC, except for carbon B1, is positive with T S > H . We conclude from these observations that m-LR adsorption from this particular solvent system is entropically driven. Carbon B1 exhibits an exothermic enthalpy component, as well as a positive entropy component, with both favoring adsorption. This result also reects a considerably larger equilibrium constant, K , for carbon B1 than the remaining ACs. Generally, adsorption is an exothermic process. Certainly, it is not unreasonable to expect that the adsorption of a relatively hydrophobic adsorptive by a hydrophobic adsorbent from an aqueous solution will be exothermic. Negative adsorption enthalpies would indicate a relatively strong afnity between the hydrophobic m-LR and the hydrophobic stationary phase. The positive adsorption enthalpies reect the inuence of the solvent on adsorption. The equilibrium described by Eq. [1] greatly simplies the actual system. The displacement of solvent from the carbon surface on adsorption of m-LR should be considered (6, 7, 19). In the present investigation it is clear that for most of the carbons the heat liberated during adsorption is insufcient to compensate for the heat required to displace the solvent. Adsorbent B1 is the exception. The inuence of solvent is also evident in the adsorption entropies. In general, one expects adsorption to raise order in a system, giving net negative entropy. This is indeed the case for the adsorption of many organics (25) and for microcystins from aqueous solvent systems by chromatographic adsorbents (26). Positive S values, on the other hand, indicate a decrease in the order of the system during adsorption, and are considered to be either processes driven by hydrophobic interactions (27), or the result of the displacement of adsorbed solvent on adsorptive adsorption. In this work, the former explanation for the positive S values is unlikely, as the solvent system is not aqueous. To gain a better understanding of the adsorption parameters shown in Table 4, the data were analyzed in terms of the carbon properties, in particular, carbon porosity and surface chemistry. Since enhanced adsorption can be expected within pores of dimensions similar to the molecular dimension of the adsorptive (28), carbon pore size distributions were examined. Donati et al. (5) investigated the efciency of m-LR adsorption by powdered activated carbons and related its removal only to the mesoporosity of the ACs considered. Although their published data suggest a good correlation, they gave no correlation analysis. In contrast with this and other previously published data examining pore volume inuence on m-LR adsorption (24), we included in this study adsorption by a nonporous carbon black. We show in Fig. 6 a correlation of the maximum average amount of m-LR adsorbed with the combined secondary micropore and mesopore volumes for each carbon. We include this smaller pore range because each AC contains these pores and components of m-LR should (energetically (28)) orient themselves to adsorb in the secondary micropores. Since there is an appreciable amount of m-LR adsorbed by the nonporous carbon, it is also

FIG. 6. Comparison of maximum m-LR adsorbed with secondary micropore and mesopore volumes.

appropriate to make this amount adsorbed the intercept of any linear correlation analysis. A summary of the linear analyses is given in Table 3. The structural formula calculations of Lanaras et al. (2) suggest that m-LR has dimensions ranging from 1.1 to 1.9 nm, as shown in Fig. 1. Clearly such molecular dimensions cannot t into primary micropores and a correlation between the maximum amount adsorbed and this micropore volume is not expected, r 2 = 0.21. Considering the correlation suggested by the previous authors, however, one cannot ignore the fact that the secondary micropores contribute to the adsorption process where an 84% linear correlation is achieved. The combination of these and the mesopores give a 92% correlation. In light of the precision of these results, it is appropriate to suggest that any selection of an AC for m-LR removal should consider both the secondary micropore volume and the mesopore volume.
CONCLUSIONS

Both the adsorbent surface chemistry and the primary micropore volume have virtually no inuence on the amount of m-LR adsorbed from an aqueous (pure water) solution. An adjustment of the solution pH conditions, to low pH, results in an enhanced adsorption of m-LR due either to a precipitation or to a reduced solvency effect. M-LR adsorption is an entropy-driven process for each carbon except carbon B1, where both exothermic processes also contribute to the partition between the adsorbed and dissolved states. To achieve effective m-LR removal from an aqueous (pure water) solution, the combination of secondary micropore and mesopore volumes should be considered as the principal criterion for adsorbent selection.
ACKNOWLEDGMENTS
We thank the Australian Research Council for nancial support through the Collaborative Research Grant scheme and PICA for their generous nancial and in-kind support throughout this research program. We also thank the Australian Water Quality Center for their support and performing the surface titration analyses.

8
REFERENCES

PENDLETON, SCHUMANN, AND WONG 14. Young, D. M., and Crowell, A. D., Physical Adsorption of Gases, Butterworths, Washington, DC, 1962. 15. Badalyan, A., Pendleton, P., and Wu, H., Rev. Sci. Instrum. 72 (2001). 16. Newcombe, G., Drikas, M., and Hayes, R., Water Res. 31, 1065 (1997). 17. Harada, K., Matsuura, K., Suzukic, M., Watanabe, M. F., Oishi, S., Dahlem, A. M., Beasley, V. R., and Carmicheal, W., Toxicon 28, 55 (1990). 18. Yun, K. S., Zhu, C., and Parcher, J. F., Anal. Chem. 67, 613 (1995). 19. Rouquerol, F., Rouquerol, J., and Sing, K. S. W., Adsorption by Powders and Porous Solids, Academic Press, London, 1999. 20. Kaneko, K., J. Membr. Sci. 96, 59 (1994). 21. Schumann, R., Wong, S.-H., Pendleton, P., Levay, G., and Mulcahy, D. E., AWWA 17th Fed. Conv, pp. 574. Melbourne, Hallmark, Brighton, vic, 1996. 22. Rodr guez-Reinoso, F., and Molina-Sabio, M., Adv. Colloid Interface Sci. 7677, 271 (1998). 23. Wong, S.-H., Ph.D. thesis, University of South Australia, Adelaide, 1999. 24. Lambert, T. W., Holmes, C. F. B., and Hrudey, S. E., Water Res. 30, 1411 (1996). 25. Glod, B. K., Alexander, P. W., Chen, Z. L., and Haddad, P. R., Anal. Chim. Acta 306, 267 (1995). 26. Fox, R., B.Sc. (Hons), University of South Australia, Adelaide, 1995. 27. Cole, L. A., Dorsey, J. G., and Dill, K. A., Anal. Chem. 64, 1324 (1992). 28. Everett, D. H., and Powl, J. C., J. Chem. Soc., Faraday Trans. 72, 619 (1976).

1. Nicholson, B. C., and Rositano, G., Chem. Aust. 452 (1992). 2. Lanaras, T., Cook, C. M., Eriksson, J. E., Meriluoto, J. A. O., and Hotokka, M., Toxicon 29, 901 (1991). 3. Falconer, I. R., Runnegar, M. T. C., Buckley, T., Huyn, V. L., and Bradshaw, P., J. AWWA 81, 102 (1989). 4. Lahti, K., and Hiisvirta, L., Water Supply 7, 148 (1989). 5. Donati, C., Drikas, M., Hayes, R., and Newcombe, G., Water Res. 28, 1735 (1994). 6. Pendleton, P., Wong, S-H., Schumann, R., Levay, G., Denoyel, R., and Rouquerol, J., Carbon, 35, 1141 (1997). 7. Considine, R., Denoyel, R., Pendleton, P., Schumann, R., and Wong, S.-H., Colloids Surf. A: Physicochem. Eng. Aspects 179, 271 (2001). 8. Ogino, K., in Studies in Surface Science and Catalysis, Vol. 80, p. 491, Elsevier, New York, 1992. 9. Bansal, R. C., and Dhami, T. L., Carbon 15, 153 (1977). 10. Bansal, R. C., Bhatia, N., and Dhami, T. L., Carbon 16, 65 (1978). 11. Monneyron, P., Pendleton, P., and Schumann, R., J. Colloid Interface Sci. in press (2001). 12. M uller, G., Radke, C. J., and Prausnitz, J. M., J. Colloid Interface Sci. 103, 466 (1985). 13. Jiang, Z. P., Yang, Z. H., Yang, J. X., Zhu, W. P., and Wang, Z. S. in Inuence and Removal of Organics in Drinking Water (J. Mallevialle, I. H. Suffet, and U. S. Chen, Eds.), pp. 79, Lewis, Boca Raton, 1992.

You might also like