- Email: [email protected]
Polydisperse adsorbability composition of several natural and synthetic organic matrices
Journal of Colloid and Interface Science 265 (2003) 265–275 www.elsevier.com/locate/jcis
Polydisperse adsorbability composition of several natural and synthetic organic matrices Fusheng Li,a,∗ Akira Yuasa,b Hajime Chiharada,c and Yoshihiko Matsui a a Environmental Engineering Division, Department of Civil Engineering, Gifu University, 1-1 Yanagido, Gifu 501-1193, Japan b River Basin Research Center, Gifu University, 1-1 Yanagido, Gifu 501-1193, Japan c Graduate School of Engineering, Gifu University, Yanagido, Gifu 501-1193, Japan
Received 4 March 2003; accepted 7 May 2003
Abstract The polydisperse composition of nine dissolved organic materials (DOMs) from two river water sources, one ground water source, two biologically treated wastewater sources, and two commercial sources was analyzed based on their adsorbabilities by activated carbon. For each DOM, batch adsorption isotherms measured for both TOC and UV260 were analyzed using an overall isotherm model derived from the IAST–Freundlich expression. By accounting for the heterogeneity of each DOM with a log-normal distribution of the Freundlich parameter (K), its adsorption behavior was characterized with only four parameters (including three fitting ones). The average adsorptive strength (KM ) and heterogeneity (σ ) determined for all DOMs, which were defined by the mean value and the standard deviation of the log-normal distribution of the Freundlich K, changed over the ranges 2.5–62.2 and 0.22–0.97 (mg/g)/(mg/l)1/n , respectively, when the TOC index was used. Among all DOMs studied, a river water DOM at the upper stream was found least heterogeneous: the Freundlich K of its organic constituents varied in the range 10.8–190 (mg/g)/(mg/l)1/n , as compared to a commercial humic acid that exhibited the broadest Freundlich K distribution of 0.01–1494.3 (mg/g)/(mg/l)1/n . KM and σ , along with other two parameters (the Freundlich exponent 1/n and the nonadsorbable organic fraction parameter Cnon /CT 0 ), changed with both indices of TOC and UV260 in a regular manner, indicating that UV-absorbing organic molecules possessed adsorbabilities different from non-UV-absorbing ones. Also based on HPSEC chromatograms measured for solutions before adsorption, the molecular weight composition of all DOMs was also assessed and the molecular size impacts on adsorption characteristics of DOMs were briefly discussed. 2003 Elsevier Inc. All rights reserved. Keywords: Dissolved organic matter; Humic acid; Activated carbon; Adsorption; Molecular weight; Distributed fictive component method
1. Introduction A significant portion of dissolved organic materials (DOMs) present in the solution matrices of aqueous water systems, about 40–90% of total dissolved organic carbon, is composed of humic and fulvic acids [1,2]. Humic substances are expected to be removed from drinking water supplies as they react with chlorine to produce carcinogenic disinfection byproducts (DBPs). Furthermore, even if not targeted for removal, their presence can adversely affect the performance of water treatment processes in removing specifically targeted synthetic organic chemicals (SOCs). For instance, humic substances are capable of competing with small SOCs for adsorption sites and of hindering * Corresponding author.
E-mail address: [email protected] (F. Li). 0021-9797/$ – see front matter 2003 Elsevier Inc. All rights reserved. doi:10.1016/S0021-9797(03)00526-5
the access of SOCs to small pores where their adsorption takes place more efficiently [2–7]. DOMs are mixtures of organic compounds having different physicochemical properties. Conventional water treatment processes can generally remove about 10–90%, with an average of approximately 30%, of the constituting compounds having large molecular weight (MW) or size [8,9]. To remove the remaining small compounds, the 1986 Amendments to the Safe Drinking Water Act of the USA have identified activated carbon adsorption as one of the best available technologies. The performance of activated carbons (ACs) in removing polydisperse DOMs is affected by a variety of factors. Using natural and synthetic organic materials, several researchers have found that molecular size and chemical characteristics (including the functional group, charge density, aromaticity) are important features of organics that influence their adsorption capacities on ACs [1,2,10–15]. Based on the adsorption
266
F. Li et al. / Journal of Colloid and Interface Science 265 (2003) 265–275
results of several organic macromolecules extracted from such materials as soil, peat, and coal, a few researchers have further pointed out that molecular size effects may dominate the competitive adsorption among constituting compounds and chemical effects may play only minor roles [2,6,12]. In addition to the inherent physicochemical features of organic matters, solution chemistry, represented by variables such as pH and ionic strength [10,11,16–20], as well as the surface and pore characteristics of ACs [5,7,10,18,20], also affect their uptake from aqueous systems. All these findings have significant practical implications for the design and operation of AC and pretreatment processes. However, due to the likely involvement of so many factors, adsorption of polydisperse organic matters is generally considered much more complicated than adsorption in single-solute systems. Batch adsorption isotherms of natural and synthetic organic materials, which are usually assessed using lumped water quality indices such as total dissolved organic carbon (TOC) and ultraviolet absorbance (UV absorbance), exhibit multicomponent adsorption behavior [2,11,12,19,21–33]. Therefore, characterization of their composition in terms of constituting components’ adsorbabilities is of significant interest as it could provide much information for assessing their removal from water and wastewater solutions as well as for assessing their competitive impacts on the uptake of SOCs. Accordingly, the main objective of the present study was to investigate the adsorbability composition of several aqueous DOMs from different water and wastewater sources. For comparison, the adsorption behaviors of two humic products from commercial sources were also examined. A distributed fictive component method was used to analyze isotherm data assessed by both DOC and UV absorbance, respectively. By doing so, the adsorption characteristics of each organic material were reflected by only four parameters. Comparisons among all organic materials were thus made based on these four parameters determined. A brief description of the method is to be given later. Moreover, using a high-performance size exclusion chromatography (HPSEC) system, the molecular weight (MW) polydispersity of all organic materials before adsorption was examined. The dependency of the overall adsorption extent on the molecular weight of organic materials from different sources was also briefly discussed. An expectation that guided this discussion was that some parametric trends probably existed between parameters devised to reflect the overall MW and adsorptive characteristics of all organic materials used, if physical size was the predominant factor that controlled the adsorption behaviors of organic constituents, as suggested in [2,6,12].
2. Experimental materials and methods 2.1. Dissolved organic materials (DOMs) Seven dissolved organic materials contained in four river water samples, a ground water, and two biological process
effluents of a wastewater treatment plant were used along with two commercial humic acids. Four river water samples were collected from Tokoro River water (TRW) and Nagara River water (NRW) at the upper, middle, and lower streams (NRW-US, NRW-MS, NRW-DS), respectively. TRW was chosen because it is representative of most drinking water sources in Hokkaido, the second largest island of Japan, in the respect that water sources there generally contain relatively higher levels of color-causing humic substances. Nagara River water was used because it supplies sufficient water for social and industrial activities in the Gifu and Aichi prefectures of central Japan. The ground water (KGW) was collected from a shallow well at Kitamura village, Hokkaido. KGW was used because the organic matter contained is mainly from peat and such water samples have frequently been used to characterize the behavior of naturally occurring humic substances during physicochemical water treatment processes [2,6,12,13]. Two wastewater samples after biological treatment (WBT-1 and WBT-2) were collected from a wastewater treatment plant in Gifu prefecture. Although the origin of treated wastewater DOMs differs from that of natural organic materials, they generally exhibit many features of the latter in aspects such as the MW, the charge potential, and physicochemical treatment behavior. As model organic macromolecules used to characterize the behaviors of natural organic materials in water environmental systems [1, 2,11,12,17–19], Wako humic acid (WHA), a commercial product extracted from coal, and Aldrich humic acid sodium salt (AHA) were taken directly as received from Wako Pure Chemicals Co., Japan, and Aldrich Co. Ltd., USA, respectively. Except for WHA and AHA, the DOM stock solutions were obtained by filtering collected water samples through 0.2-µm membrane filters (Toyo Roshi, Japan) followed by adjustment of pH to 7.0 as necessary. The stock solution of WHA was prepared by dissolving the material in Milli-Q water raised to pH 11 with 0.5 M NaOH, lowering water pH to 7.0 with 0.5 M HCl, and filtering the solution through 0.2-µm membrane filters. The stock solution of AHA was obtained by dissolving the material in Milli-Q water, adjusting water pH to 7.0, and filtering the solution through 0.2-µm membrane filters. All stock solutions were stored refrigerated at 5 ◦ C in the dark to prevent biological activity. 2.2. Activated carbon The adsorbent used in this research was Filtrasorb-400 granular activated carbon (Calgon Co., USA), which has an average pore radius of 12.0 Å, a total pore volume of 0.566 cm3 /g, and a total surface area of 950 m2 /g [2]. This activated carbon (AC) was chosen because it is widely used in advanced water treatment applications and laboratory research. The granular AC was pulverized and sieved to particles with sizes below 45 µm, washed and rinsed with distilled water to remove fines, dried at 105 ◦ C overnight, and finally stored in a desiccator till use. Before being subjected
F. Li et al. / Journal of Colloid and Interface Science 265 (2003) 265–275
267
Table 1 The concentrations of DOMs in stock solutions with molecular weight information Source of DOM TRW NRW-US NRW-MS NRW-DS KGW AHA WHA WBT-1 WBT-2
UV260
TOC
SUVA
(m−1 )
(mg/l)
(m−1 /(mg/l))
4.3 1.2 1.7 3.2 30.0 33.6 23.5 230.0 106.0
1.9 0.7 0.8 1.3 9.1 4.1 2.0 105.7 43.8
2.30 1.76 2.13 2.51 3.30 8.12 11.85 2.18 2.42
to pulverizing, the granular AC was washed several times with and subsequently boiled in Milli-Q water for about 1 h to remove fines, washed and rinsed again several times, and finally dried at 105 ◦ C to minimize likely impacts of ash impurities on the adsorbent surface. 2.3. Isotherm experiments Batch isotherm experiments were conducted according to the bottle-point method of variable AC doses in 500-ml flasks sealed with Teflon-lined rubber septa. For each water sample, to measure the isotherm dependency on DOM’s initial concentrations, two or three working solutions with variable initial concentrations were prepared by diluting the stock solution with Milli-Q water as necessary. To minimize the influence of changes in solution chemistry caused by dilution on adsorption [10,11,16–20], for working solutions of each organic material, pH was controlled at 7.0 and the ionic strength was adjusted with 0.5 M NaCl based on the electrical conductivity of the working solution prepared with the smallest dilution ratio. After reaching equilibration by shaking the flasks containing 200 ml of working solutions placed on a rotation shaker at 200 rpm in a temperature-controlled room (20 ◦ C) for 7 days, solutions were filtered through 0.45-µm membrane filters (Toyo Roshi, Japan) rinsed beforehand with Milli-Q water. A preliminary adsorption rate study showed no measurable changes in liquid-phase concentrations after agitation for up to 6 days, thus ensuring that equilibration was reached by shaking for 7 days. DOM concentrations in the filtrates were quantified with TOC and UV absorbance at 260 nm (UV260), for which a TOC analyzer (Model TOC810, Sievers, USA) and a UV–visible spectrophotometer (Model U-3210, Hitachi Co., Japan) were used, respectively. Based on the results of control tests conducted without the addition of AC, there were no measurable losses of DOMs and thus all changes in the concentration were attributed to adsorption. The measurement of UV260 was made because, like UV absorbance at 254 nm (UV254), UV260 is capable of reflecting the presence of humic molecules present in water environment systems, and the specific ultraviolet absorption (SUVA), a calculated parameter of UV260 or
Range of MW
Mw
Mn
Polydispersity
3174 2972 2928 3033 2603 3185 2510 2026 1977
1.089 1.141 1.136 1.118 1.101 1.152 1.259 1.170 1.187
(g/mol of PSS) 842–6637 855–7365 805–7365 771–7365 1003–5244 1013–6797 908–7668 500–4449 450–4500
3456 3391 3326 3392 2865 3667 3161 2371 2347
(–)
UV254 divided by TOC, has been thus used as an indicator of the humic content. The US EPA [34] reports that water with a small SUVA contains primarily nonhumic organic compounds and is not amenable to enhanced coagulation. The values of TOC, UV260, and SUVA for all stock solutions are summarized in Table 1. Throughout the study, the values of UV260 are given in units of m−1 , which were converted from measured values with a regular 10-, 50-, or 100-mm cell (a longer cell was used for solutions with lower DOM concentrations). In addition to ease of analysis, the measurement of UV absorbance also has the advantage of good reproducibility: the detector response variance for repeated UV260 measurements was within ±0.0005. Apparent SUVA differences shown in Table 1 indicated that the humic content changed with the sources and the types of DOMs. 2.4. High performance size-exclusion chromatography (HPSEC) HPSEC was carried out to measure the MW distributions of all DOMs. A packed silica column (GL-W520-X 10.7 × 450 mm, Hitachi Co., Japan) and a UV–visible detector (Model LC-10AV, Shimadzu Co., Japan) set at 260 nm (UV260) were used. The mobile phase solution made up of Milli-Q water was buffered with phosphorous salts (0.02 M Na2 HPO4 + 0.02 M KH2 PO4 ) and was supplied to the chromatographic column at 0.5 ml/min. Three polystyrene sulfonates (PSS) with known molecular weights (MWs) of 1430, 4950, and 6530 g/mol were used to calibrate the HPSEC system and a good correlation (R 2 = 0.998) described by log10 (MW) = −0.1096t + 5.029 was obtained (t is the elution time). Semilog linear correlations between log10 (MW) and t were also confirmed by others [2,12]. 3. Results and discussion 3.1. Molecular weight characteristics of DOMs HPSEC chromatograms of all DOMs used in the study are displayed in Figs. 1a–1c. As shown, the MW distributions differed apparently with DOM sources. For river water
268
F. Li et al. / Journal of Colloid and Interface Science 265 (2003) 265–275
those of aqueous DOMs. Kilduff et al. [2] conducted HPSEC measurements for humic acids extracted from Laurentian soil (LaHA) and peat (PHA), respectively. Chromatograms that resembled each other (each had two peaks and the first peak eluted earlier from the column was much smaller than the second peak eluted later) but differed obviously from that of the Huron River water DOM also measured by them were obtained. The results we obtained in the present study, together with previous ones, indicate that humic products extracted from such natural materials as soil, peat, and coal following various procedures possess MW characteristics different from those of aqueous ones present in water and wastewater sources. Based on the chromatograms obtained, the weight-averaged (Mw ) and number-averaged (Mn ) molecular weights were determined according to two expressions appearing in the literature [35], N N Mw = (1) MWi (t)hi (t)t hi (t)t, i=1
Mn =
N i=1
Fig. 1. Size-exclusion chromatograms of DOMs. Values in parentheses are DOM concentrations of working solutions used for HPSEC measurements.
DOMs, TRW, NRW-US, NRW-MS, and NRW-DS exhibited chromatograms similar to one another, with the magnitude of detector responses changing with DOM concentrations (Fig. 1a). Ground water (KGW) showed a chromatogram (Fig. 1b) that resembled those of four river water DOMs and was also in close similarity to the documented one measured for Huron River water DOM [2]. For treated wastewater DOMs, similar chromatographic shapes were obtained for samples WBT-1 and WBT-2 (Fig. 1b). However, the number of chromatographic peaks they possessed was larger than river water and the ground water DOMs. Among all organic materials studied, the commercial humic products AHA and WHA had the fewest of chromatographic peaks, two and three, respectively (Fig. 1c), and exhibited chromatographic shapes markedly different from
hi (t)t
i=1 N
hi (t)t MWi (t) ,
(2)
i=1
where MWi (t) is the MW as a function of elution time t, hi (t) is the UV detector response and t is the time interval. Then, based on determined values of Mw and Mn , the MW polydispersity (= Mw /Mn ), a parameter that has been commonly used [2,12,35] to assess the MW heterogeneities of organic matrices, was also computed. As summarized in Table 1, the values of Mw varied with organic materials, following a general trend that Aldrich humic acid (AHA) > river water DOM > Wako humic acid (WHA) > ground water DOM > treated wastewater DOM. A similar trend was also found for the number-averaged MW, Mn . However, compared to the differences in Mw and Mn , the differences in polydispersity were less obvious, with its values following the decreasing order Wako humic acid (WHA) > treated wastewater DOM > Aldrich humic acid (AHA) > river water DOM ∼ = ground water DOM. The trend observed supports a previous finding that surface water organic materials was less disperse than organic materials extracted from soil and peat [2]. For all three samples collected from the Nagara River (NRW-US, NRW-MS, and NRW-DS), Mw , Mn , and polydispersity of the river water DOM changed only slightly as the water flowed from the upper to the down streams, even if the content of organic matter assessed by TOC was increased by approximately 90% (as shown in Table 1). Besides the commonly used definition of MW polydispersity as Mw /Mn , the MW range may also serve as an informative parameter in reflective of the MW characteristics of organic mixtures. The results summarized in Table 1 displayed that, compared to the differences reflected by the parameter Mw /Mn , those reflected by the MW range were somewhat more obvious among all DOMs investigated. However, it has to be noted that the MW range
F. Li et al. / Journal of Colloid and Interface Science 265 (2003) 265–275
parameter includes no information with respect to the chromatographic shapes as well as the presence levels of DOM constituents. 3.2. Adsorption characteristics of DOMs 3.2.1. Observed adsorption isotherms The adsorption isotherms of TRW, NRW-MS, AHA, and WBT-1 assessed by TOC are displayed in Fig. 2 as examples. As illustrated by symbols, reflecting the polydisperse adsorbabilities of organic constituents, the overall adsorption capacity (qT ) changed with the initial TOC concentrations (CT 0 ). The isotherm shifted its position upward as CT 0 decreased, indicating larger adsorption capacities for diluted working solutions. Besides, due to preferential adsorption of components with strong adsorbabilities, qT at low AC doses (CAC ), where the adsorption sites were limited, was extremely high. These features have been reported by numerous researchers for a variety of natural and synthetic organic matrices [2,5,6,11,12,19,21–33] and were also observed for TOC-based isotherms of KGW, NRW-US, NRWDS, and WBT-2, as well as for UV260-based isotherms of all organic materials used in this study (data not shown). 3.2.2. Model description of observed isotherms To describe the adsorption isotherms of polydisperse organic matters, a fictive component method (FCM) that uses a series of hypothetical components to account for competi-
269
tions among organic constituents with the aid of multicomponent adsorption models [3,30], in many cases the ideal adsorbed solution theory (IAST), has been applied [21–23, 27–29]. However, this method has the drawback of requiring a search for a large number of variables, namely the equilibrium parameters and initial concentrations of all hypothetical components. Thus, the values found are not always unique. Harrington and DiGiano [21] discussed the limitations of the FCM. They found that, even using a small number of hypothetical components (two adsorbable ones and a nonadsorbable one), which involved totally seven parameters, the description of isotherms observed for a natural organic material had severe computational and statistical problems. Besides this, the effects of preozonation on adsorption equilibrium were found not to be interpretable from the parameters determined. They thus suggested that, to promote the method’s application in practice, modifications that could lead to a significant reduction of adjustable variables were appealing. Different from the common practice of using a discrete number of fictive components, we proposed a distributed fictive component method (DFCM), in which a straightforward equation derived from the IAST–Freundlich model and a distribution function for the Freundlich K of all adsorbable DOM components were used [14]. Requiring searching for only three parameters, this method satisfactorily described and predicted the isotherms [14,31] of DOMs in wastewater and peat water sources. The equilibrium parameters
Fig. 2. The TOC-based overall isotherms of DOMs. Symbols are observed data. Lines are model fits to the observed data. Values in parentheses show the DOM concentration in working solutions used.
270
F. Li et al. / Journal of Colloid and Interface Science 265 (2003) 265–275
determined accordingly were then successfully applied to describe the breakthrough behavior of the total organic material in a fixed-bed adsorber [32]. Pretreatment effects of coagulation, ozonation, and chlorination on adsorption capacities of these two organic materials were also reasonably explained [36]. Hence, in this study, this method was employed to analyze all isotherm data obtained. A brief description of the method is given below. By assuming that the Freundlich exponent (1/ni ) and the conversion ratio (αi ) of molar concentration to the concentration of a lumped quality index (TOC or UV260) are identical for all adsorbable components (1/ni = 1/n, αi = α), the overall isotherms of an organic matrix at variable initial concentrations can be described by N i=1
Ci0 /(CT 0 −Cnon ) [qTn /(CT 0 −Cnon )]Ki−n +1−(CT −Cnon )/(CT 0 −Cnon )
= 1,
(3)
where N is the total number of components, Ki is the Freundlich parameter and Ci0 is the initial concentration of component i, Cnon is the nonadsorbable component concentration, CT 0 is the overall initial concentration, CT is the overall liquid phase equilibrium concentration, and qT is the overall solid phase equilibrium concentration. This equation is an extended version of and can be directly derived from the documented IAST–Freundlich model [26]. The appropriateness of assuming identical values of 1/ni and αi for all adsorbable components has been tested by several researchers and is thus applied as a relatively common practice in modeling the adsorption isotherms of organic materials with unknown composition [21–24,27]. With such assumptions, the heterogeneity of organic constituents in adsorbabilities is mainly accounted for by the Freundlich parameter Ki and the initial concentration Ci0 to be searched. When Freundlich K of all adsorbable constituents is described by a distribution function, f (K), the item Ci0 / (CT 0 − Cnon ) in Eq. (3) can be determined as Ci0 /(CT 0 − Ki+K/2 Cnon ) = Ki−K/2 f (K) dK, where K is the interval used to divide the whole distribution range of Freundlich K based on the total number of hypothetical components (N). Thus, a distributed overall isotherm equation can be obtained: ∞ f (K) dK [qTn /(CT 0 −Cnon )]K −n +1−(CT −Cnon )/(CT 0 −Cnon )
= 1.
standard deviation (σK ) of log10 (K), the Freundlich exponent (1/n), the overall initial concentration (CT 0 ), and the concentration of the nonadsorbable fraction (Cnon ): [log10 (K/KM )]2 1 exp − . f (K) = √ (5) 2σ 2 2πσ Since CT 0 and Cnon in Eq. (4) can be measured independently (observed data at higher AC dosages are used for determination of Cnon ), parameters to be searched are limited to only three: KM , σ , and 1/n. These three parameters are separately sensitive to each portion of the isotherm’s curvatures; hence, a unique set of parameters can logically be obtained from the observed isotherm data according to the following objective function (Er) to be minimized. For a given DOM, isotherm data measured for two or more initial concentrations are simultaneously required in the search because 1/n is a characteristic parameter reflecting the isotherm dependency on initial concentrations [22–24,26]. The objective function is
ND
CT (obs,i) − CT (cal,i)
1
Er =
2ND CT (obsm) i=1
CT (obs,i) − CT (cal,i)
, +
(6)
CT (obs,i) where CT (cal,i) is the calculated liquid phase concentration, CT (obs,i) is the observed liquid phase concentration, CT (obsm) is the arithmetic average of the observed liquid phase concentration, and ND is the number of data points obtained for all initial concentrations of a given organic material. Simulation results that best fitted the observed TOCbased isotherms for TRW, NRW-MS, AHA, and WBT-1 are displayed by solid lines in Fig. 2 as examples. For all DOMs studied, the simulation results described the observed isotherms with reasonably high accuracies. The error defined by Eq. (6) was in the range of 2.97–10.7% for the TOC-based isotherms and of 2.48–8.09% for the UV260based ones, as summarized in Tables 2 and 3, respectively. The results obtained, together with our previous ones using the DFCM for description and prediction of the adsorption equilibrium and column breakthrough of aqueous
(4)
0
To find a distribution function suitable for application, Li [37] tested a large number of distribution patterns defined by functions such as the triangular function, the normal distribution function, and the logarithmic normal distribution function. Model simulations demonstrated clearly that the log-normal distribution function was most suitable, since it was capable of generating isotherm shapes most close to observed ones and involves only two explanatory variables. With this function, as described by Eq. (5), total parameters regulating the overall isotherms of DOMs described by Eq. (4) are further reduced to the mean value (KM ) and the
Table 2 Parameters determined from TOC-based isotherms DOM source
KM
σ
Range of K
((mg/g)/(mg/l)1/n ) TRW NRW-US NRW-MS NRW-DS KGW AHA WHA WBT-1 WBT-2
40.4 45.3 62.2 48.2 37.0 4.0 2.5 42.5 50.5
0.30 0.22 0.39 0.41 0.40 0.72 0.97 0.43 0.36
5.6–290.8 10.8–190.0 4.7–825.6 3.3–715.3 2.7–514.1 0.03–463.7 0.01–1494.3 2.5–728.9 4.9–525.4
Cnon /CT 0
Er
(–)
(–)
(%)
0.26 0.33 0.29 0.20 0.35 0.20 0.21 0.35 0.25
0.15 0.38 0.14 0.15 0.04 0.00 0.00 0.02 0.03
5.66 10.70 4.63 8.65 4.95 2.97 3.21 5.66 7.03
1/n
F. Li et al. / Journal of Colloid and Interface Science 265 (2003) 265–275
Table 3 Parameters determined from UV260-based isotherms DOM source
KM
σ
Range of K
(m1/n−1 /(g/l)) TRW NRW-US NRW-MS NRW-DS KGW AHA WHA WBT-1 WBT-2
125.2 94.6 141.1 115.2 123.6 22.2 20.0 122.0 134.5
0.22 0.11 0.25 0.27 0.22 0.60 0.79 0.42 0.34
29.8–525.4 47.1–190.0 27.8–716.7 19.1–694.27 29.5–518.7 0.42–1165.1 0.11–3717.5 7.7–1933.6 14.8–1226.7
1/n
Cnon /CT 0
Er
(–)
(–)
(%)
0.20 0.29 0.27 0.19 0.27 0.18 0.20 0.33 0.25
0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.02 0.00
6.95 5.39 8.09 6.70 2.48 5.43 5.31 7.94 4.03
DOMs [31,32], further indicate that this method is effective for characterizing aqueous organic matrices present in surface water and wastewater sources, as well as synthetic humic products. Further validations will be made using marine organic matter and organic matter from different latitudes. 3.2.3. Adsorbability distributions Four parameters (KM , σ , 1/n and Cnon ) determined for all DOMs based on the TOC and UV260 isotherms are summarized in Tables 2 and 3, respectively. As a parameter devised to reflect the mean adsorptive strength of a given organic material, KM differed considerably with respect to the sources and types of DOMs. The TOC-based KM varied in the range 2.5–62.2 (mg/g)/(mg/l)1/n, while the UV260based one varied in the range 20.0–141.1 m1/n−1 /(g/l). The largest KM difference was found between the river water NRW-MS and the commercial WHA, with the values of the former being approximately 25 and 7 times as large as the values of the latter for the indices of TOC and UV260, respectively. The larger difference noticed with the TOC index may imply that some organic compounds not detected by UV260 in the river water source NRW-MS were adsorbed to much greater extents than UV-absorbing humic macromolecules. Such an implication is considerable since UV260 (or UV254) is generally used as an index representing the content of humic substances or organic macromolecules having properties similar to humic ones. As an important parameter devised for assessing the polydispersity of organic materials in adsorbabilities, the TOC-based σ changed with DOMs in the decreasing order: WHA > AHA > WBT-1 > NRW-DS > PW > NRW-MS > WBT-2 > TRW > NRW-US. The much larger σ values noticed for WHA (0.97) and AHA (0.72) as compared to aqueous source DOMs (0.22–0.41) suggested that, among all organic materials studied, the commercial humic acids were most heterogeneous with respect to their constituting compounds’ adsorption behavior. In contrast, the smallest σ noticed for NRW-US (0.218) implied that this river water DOM was the least heterogeneous matrix. It was also noteworthy that for all organic materials studied, the TOC-based σ values (in the range 0.22–0.97) were about 1.1–2.1 times as large as the UV260-based ones (in the range 0.11–0.79).
271
Such a result is also expectable since, as mentioned earlier, while UV260 is an index reflective of only the portion of organic compounds having UV absorbing features, such as humic substances, the index TOC measures all organic molecules dissolved in water solutions, including those not detectable with a UV–visible spectrophotometer. For the Nagara River water DOM, the values of σ assessed by both TOC and UV260 varied with the sampling points in the ranges 0.22–0.41 and 0.11–0.27, respectively. However, it was interesting to see that, for both these indices used, the changes in the magnitude of σ followed the increasing order of NRW-US < NRW-MS < NRW-DS. This thus indicated that the adsorption polydispersity of the river water DOM was becoming expanded as the river water flowed downstream. The expansion might be caused by the merging of some organic compounds having strong (or much weak) adsorbabilities into the river water through some of its tributary streams. The ranges of Freundlich K determined for all DOMs based simply on the corresponding KM and σ values are tabulated in Tables 2 and 3. As could be seen from the results, the TOC-based ones in Table 2, for instance, along the water flow direction, the range of the Freundlich K of the river water DOM changed from 10.8–190.0 (NRW-US) through 4.7–825.6 (NRW-MS) to 3.3–715.3 (NRW-DS). For a given organic material with unknown composition, the Freundlich exponent 1/n, which is assumed to be identical for all constituting compounds to facilitate the modeling of its overall isotherm, is regarded as a characteristic parameter reflecting the overall isotherm’s dependency on initial concentrations [22–24,26]. Meanwhile, it is also a parameter indicative of the overall adsorption energy (i.e., the affinity) between organic molecules and the internal AC pore surfaces [5]. As summarized in Tables 2 and 3, although 1/n changed with DOM, its values were relatively small: 0.20–0.35 (for the TOC index) and 0.18–0.33 (for the UV260 index). This thus implies that all DOM studied in the present work had stronger affinities with the AC used. Moreover, it was also interesting to see that, except for the treated wastewater DOM in WBT-2, the UV260-based 1/n values were comparatively smaller than the TOC-based ones. This may thus indicate that, of the organic constituents of DOMs, UV-absorbing humic molecules had stronger affinities with the AC surfaces as compared to those coexistent organic species not possessing UV-absorbing features. It is possible that the adsorption capacity of organic compounds that do not possess UV-absorbing features is larger, as discussed earlier based on the KM values; e.g., if these compounds have access to AC pores from which larger UV-absorbing molecules are excluded. Nonetheless, these compounds may have weaker affinities for the AC surfaces because they might be more hydrophilic than UV-absorbing ones. In addition to KM , σ , and 1/n, the parameter Cnon /CT 0 devised to reflect the nonadsorbable organic fraction also demonstrated changes with the sources and types of DOMs, as tabulated in Tables 2 and 3. However, compared to the TOC-based Cnon /CT 0 values (in the range 0–0.38), the
272
F. Li et al. / Journal of Colloid and Interface Science 265 (2003) 265–275
Fig. 3. Freundlich K distributions of DOMs depicted using TOC-based values of KM and σ .
UV260-based ones for most DOMs were zero, suggesting that UV-absorbing organic compounds were completely adsorbable. It could be thus inferred that the nonadsorbable TOC fraction was coming from some compounds not possessing UV-absorbing features. For DOMs contained in samples of TRW, WBT-1 and WBT-2, although non-adsorbable organic fractions were also detected for the UV260 index, the Cnon /CT 0 values were smaller than TOC-based ones. The results shown above demonstrated that, by analyzing observed isotherm data with the DFCM, the polydisperse composition of organic materials could be assessed based on their constituents’ adsorption behaviors. Accordingly, as displayed in Fig. 3 as examples, the polydisperse feature of a given organic material could be explicitly evaluated according to the Freundlich K distribution depicted by simply introducing the determined KM and σ values into Eq. (5). Comparisons of all four characteristic parameters involved in the DFCM for all organic materials used in the present work indicated that the adsorption of polydisperse organic materials by activated carbon differed with their types and sources. However, the differences were relatively small among river water, ground water, and biologically treated wastewater DOMs. Commercial humic acids exhibited adsorption behaviors very different from aqueous DOMs. Some caution must be exercised when such humic products are used for characterization of the adsorption behaviors of natural organic materials.
Fig. 4. Effects of humic content on the adsorbability of DOMs: (a) the mean adsorptive strength (KM ) vs the humic content indicator (SUVA); and (b) the adsorptive strength polydispersity (σ ) vs SUVA. The dashed lines exhibit the general parametric trends.
3.3. Adsorption dependency on humic content and molecular weight To investigate the humic content effects on the adsorbability of total organic matrices, the adsorption parameters KM and σ determined for the total organic carbon index TOC were plotted against the values of SUVA shown earlier in Table 1. Although there were some exceptions, the parametric line shown in Fig. 4a demonstrated a decreasing trend of KM with increasing SUVA. This trend supported the indication given earlier that humic molecules were adsorbed to a less extent as compared to other coexisting organic compounds exhibiting non UV-absorbing features. Contrary to this trend, the parametric line in Fig. 4b displayed that the heterogeneity parameter (σ ) increased with increased SUVA, thus implying that humic molecules were more disperse with regard to their adsorption behaviors. However, it must be pointed out that both these parametric trends were heavily influenced by the results of the commercial humic acids (AHA and WHA), which behaved much differently from those aqueous DOMs with respective to their SUVA, KM , and σ values. If these two humic acids were excluded, i.e., when analyses were made based on the results of seven aqueous DOMs alone, both trends would become a bit vague.
F. Li et al. / Journal of Colloid and Interface Science 265 (2003) 265–275
Fig. 5. Weight-averaged molecular weight (Mw ) versus mean adsorptive strength (KM ) of all DOMs.
Fig. 6. Molecular weight polydispersity versus the Freundlich K heterogeneity (σ ) of all DOMs.
The relationship between KM and Mw is plotted in Fig. 5. Similarly, the relationship between σ and the MW polydispersity is plotted in Fig. 6. Since all MW-related information (Mw , Mw , and polydispersity shown in Table 1) was generated from HPSEC chromatograms measured with a UV–visible detector, for better correspondency, the values of KM and σ determined for the UV260 index (shown in Table 3) were used. This investigation was guided by the expectation that some parametric trends probably existed, if, as previously reported [2,6,12], molecular sizes dominated the adsorption of organic materials and the chemical effects played only minor roles. As illustrated in Fig. 5, a parametric trend revealing either increasing or decreasing adsorptive strength (KM ) with increasing molecular weight (Mw ) was not exhibited. Such a trend was also confirmed not to exist when the number-averaged molecular weight (Mn ) was used instead of the weight-averaged molecular weight (Mw ). Regarding the relationship of the adsorption polydispersity (σ ) with the MW polydispersity, the data plotted in Fig. 6 did not reveal an expected parametric trend of expanding the distribution of Freundlich K with the increase of MW polydispersity. An analysis based on TOC-based KM and σ values was also conducted; however, no trend was obtained (results not shown). In previous studies [2,35,38], several researchers investigated the relationships between
273
UV-absorptivity (SUVA) of humic substances and HPSEC weight-averaged molecular weight (Mw ). Combined analysis of our results shown in Figs. 4–6 indicated no trend between these two parameters no matter if plots for two humic acids were excluded or not (direct Mw –SUVA plotting showed the same result). The indication coincided with the result shown in the literature [38], but was in contradiction to the findings of others who showed either an increase [35] or a decrease in UV-absorptivity with increased molecular weight [2]. For a given type of organic material, preferential adsorption of low-MW constituents might be a general feature of adsorption from polydisperse organic material solutions [2,6,11,12], although some exceptions were also observed [10,35,39]. For organic materials from different sources, however, besides molecular size effects, the chemical nature of organic materials as well as the solution chemistry may also play significant roles. Karanfil et al. [12] investigated the relationship between the adsorption capacities and the Mw values (measured by HPSEC) of nine commercial and synthetic macromolecular organic materials prepared from such materials as soil, peat, and coal. Even if isotherm experiments were performed under identical solution chemistry, data from four organic materials were found to deviate totally from the parametric order of decreasing adsorption capacities with increasing molecular weights. These deviations do not support and are in confrontation with the idea that molecular size is the only, or even the most important parameter controlling adsorption of organic macromolecules. With respect to the impacts of solution chemistry, numerous researches have demonstrated increased adsorption capacity with decreasing water pH [10,17,18,20] and with increasing concentrations of cationic ions, especially Ca2+ and Mg2+ [17–19]. Humic substances and ACs generally contain functional groups that can be ionized to exhibit anionic characteristics, with the extent of ionization varying according to the inorganic chemistry of solutions. Taking this feature into account, three categories of interactions have been proposed [17] to explain the potential mechanisms involved: (i) interactions between solution cationic ions (including H+ ) and ACs to neutralize the repulsive forces between anionic adsorbates and ACs; (ii) interactions between solution cationic ions and adsorbates to reduce the charge potentials of anionic adsorbates and hence to form aggregates more favorable for adsorption; and (iii) interactions between adsorbed adsorbates and solution cationic ions that possibly bring about changes in the packing, spacing, or alignment of adsorbed molecules within the AC pores. In the present study, since all solutions were adjusted to pH 7.0 before being subjected to isotherm experiments, pH could be excluded as a likely factor that attributed to different uptake behavior among all DOMs investigated, especially between aqueous ones and commercial ones. It was thus conceivable that likely affecting factors were probably relative to DOMs’ physicochemical features and the inor-
274
F. Li et al. / Journal of Colloid and Interface Science 265 (2003) 265–275
ganic composition of solutions, which need further investigations. Results we obtained in the present study do not necessarily and are not sufficient to indicate that the size effects are not significant. However, they further stress the issue related to potential impacts of chemical characteristics of organic materials (including the functional group, charge density, and aromaticity) and the chemistry of solutions. In a coming paper, the issue of molecular size effects for aqueous organic matters will be addressed in details based on HPSEC chromatograms measured for a larger number of organic matrices (including all those used in the present study) after reaching equilibrium at variable AC dosages. Physicochemical properties of ACs reflected in such aspects as pore size distribution, surface area, and surface chemistry are also important factors that influence the adsorption capacity of DOMs [1,5,7,11] as well as the competitive impacts of DOMs on the uptake of SOCs [2–7]. In this study, a single coal-based AC, which is widely used in advanced water treatment applications and laboratory research works, was used. However, taking into consideration that competitive adsorption is a general feature of adsorption from multicomponent mixtures, the trends and findings obtained were considered generic in reflecting the polydisperse adsorbabilities of aqueous and commercial organic matrices.
4. Conclusions The composition of nine organic materials from different sources was investigated based on their activated carbon adsorbabilities. For each organic material, the constituents’ heterogeneity in adsorbabilities was accounted for using a log-normal distribution of the Freundlich parameter (K). An overall isotherm expression derived based on the IAST– Freundlich model successfully described the isotherms of all organic materials measured by both TOC and UV260, respectively. The adsorption behavior of organic materials differed with their sources and types: the average adsorptive strength defined by the mean value of the log-normal distribution of Freundlich K (KM ) changed over the ranges of 2.5–62.2 (mg/g)/(mg/l)1/n and 20.0–141.1 m1/n−1 /(g/l); and the adsorptive heterogeneity defined by the standard deviation of the corresponded distribution (σ ) changed over the ranges of 0.22–0.97 (mg/g)/(mg/l)1/n and 0.11–0.79 m1/n−1 /(g/l), for the indices of TOC and UV260, respectively. Compared to organic materials from river water, ground water, and treated wastewater sources, the overall adsorption capacities of commercial humic acids were much smaller and their polydispersity in adsorbabilities was much broader. Besides, indicating strong affinities to AC surfaces, the Freundlich exponent 1/n values determined for all organic materials were relatively small and differed within relatively narrow ranges of 0.20–0.35 and 0.18–0.33 for TOC and UV260, respectively.
The weight-average molecular weight (Mw ) varied with organic matters (Mw = 2347–3667) by following a general trend that Aldrich humic acid > river water DOM > Wako humic acid > ground water DOM > treated wastewater DOM. The polydispersity defined by Mw /Mn for all organic materials changed in a narrow range of 1.09–1.26 by following the declining order: Wako humic acid > treated wastewater DOM > Aldrich humic acid > river water DOM ∼ = ground water DOM. A decreasing trend of the average adsorptive strength of total organic carbon (KM ) and an increasing trend of the corresponding adsorption polydispersity (σ ) with increasing humic content (SUVA) were observed. However, these parametric trends became vague when data related to commercial humic acids, which behaved very differently from seven aqueous organic materials, were excluded. The different behavior exhibited by commercial humic acids was attributed to their inherit chemical features (including the functional group, charge density, and aromaticity) and also to differences between the inorganic chemistry of humic acid and aqueous DOM solutions, which require further elucidation. Moreover, analysis of the results obtained in this study showed no trends between the molecular weights and the adsorbabilities of organic matters. This finding does not necessarily and maybe less sufficient to imply that the molecular size impacts are not significant. However, it may serve as supporting information to further stress the issue relates to the likely impacts of organic materials’ chemical characteristics and the chemistry of solutions.
Acknowledgments The authors sincerely thank Professor Alexander P. Mathews of Kansas State University for his valuable discussions and comments on this manuscript. This work was supported in part by Grants-in-Aid for Scientific Research 13750520 from the Japan Society for the Promotion of Science (JSPS).
References [1] M.C. Lee, V.L. Snoeyink, J. Crittenden, J. Am. Water Works Assoc. 73 (1981) 440. [2] J.E. Kilduff, T. Karanfil, Y. Chin, W.J. Weber Jr., Environ. Sci. Technol. 30 (1996) 1336. [3] J.C. Crittenden, P. Luft, D.W. Hand, Water Res. 19 (1985) 1537. [4] I.N. Najm, V.L. Snoeyink, Y. Richard, J. Am. Water Works Assoc. 83 (1991) 57. [5] C. Pelekani, V.L. Snoeyink, Water Res. 33 (1999) 1209. [6] J.E. Kilduff, A. Wigton, Environ. Sci. Technol. 33 (1999) 250. [7] K. Ebie, F.S. Li, Y. Azuma, A. Yuasa, T. Hagishita, Water Res. 35 (2001) 167. [8] S.J. Randtke, J. Am. Water Works Assoc. 80 (5) (1988) 40. [9] K. Ebie, F.S. Li, A. Yuasa, Environ. Eng. Res. Jpn. Soc. Civ. Eng. 32 (1995) 9.
F. Li et al. / Journal of Colloid and Interface Science 265 (2003) 265–275
[10] J. McCreary, V.L. Snoeyink, Water Res. 14 (1980) 151. [11] R.S. Summers, P.V. Roberts, J. Colloid Interface Sci. 122 (1988) 382. [12] T. Karanfil, J.E. Kilduff, M.A. Schlautman, W.J. Weber Jr., Environ. Sci. Technol. 30 (1996) 2187. [13] Y. Matsui, T. Kamei, N. Tambo, K. Taniguchi, J. Jpn. Soc. Water Environ. 16 (1993) 497. [14] A. Yuasa, F.S. Li, Y. Matsui, K. Ebie, Water Sci. Technol. 36 (12) (1997) 231. [15] A.M. El-Rehaili, W.J. Weber Jr., Water Res. 21 (1987) 573. [16] W.J. Weber Jr., T.C. Voice, A.M. Jodellah, J. Am. Water Works Assoc. 75 (1983) 612. [17] S.J. Randtke, C.P. Jepsen, J. Am. Water Works Assoc. 74 (1982) 84. [18] F.S. Li, A. Yuasa, K. Ebie, Y. Azuma, T. Hagishita, Y. Matsui, Water Res. 36 (2002) 4592. [19] R.S. Summers, P.V. Roberts, J. Colloid Interface Sci. 122 (1988) 367. [20] F.S. Li, A. Yuasa, K. Ebie, Y. Azuma, J. Colloid Interface Sci. 262 (2003) 331. [21] G.W. Harrington, F.A. DiGiano, J. Am. Water Works Assoc. 81 (1989) 93. [22] J. Fettig, H. Sontheimer, J. Environ. Eng. ASCE 113 (1987) 795. [23] H. Sontheimer, J. Crittenden, R.S. Summers, Activated Carbon for Water Treatment, 2nd ed., DVGW-Forschungsstelle, Karlsruhe, 1988 [in English]. [24] A. Yuasa, J. Jpn. Water Works Assoc. 62 (7) (1993) 38. [25] E.H. Smith, Water Res. 28 (1994) 1693.
275
[26] A. Yuasa, F.S. Li, Y. Matsui, K. Ebie, Environ. Eng. Res. Jpn. Soc. Civ. Eng. 33 (1996) 123. [27] B.R. Frick, H. Sontheimer, in: I.H. Suffet, M.J. McGuire (Eds.), Treatment of Water by Granular Activated Carbon, American Chemical Society, Washington, DC, 1983, p. 247. [28] J.C. Crittenden, K. Vaitheeswaran, D.W. Hand, E.W. Howe, E.M. Aieta, C.H. Tate, M.J. McGuire, M.K. Davis, Water Res. 27 (1993) 715. [29] C.L. Warta, S.P. Papadimas, G.A. Sorial, M.T. Suidan, T.F. Speth, Water Res. 29 (1995) 551. [30] F.A. DiGiano, G. Baldauf, B. Frick, H. Sontheimer, Chem. Eng. Sci. 33 (1978) 1667. [31] Y. Matsui, A. Yuasa, F.S. Li, J. Environ. Eng. ASCE 124 (1998) 1099. [32] F.S. Li, A. Yuasa, Y. Matsui, in: F. Meunier (Ed.), Fundamentals of Adsorption 6, Elsevier, Amsterdam, 1998, p. 1041. [33] R.L. Malcom, P. MacCarthy, Environ. Sci. Technol. 20 (1986) 904. [34] US-EPA, Office of Water (4607), EPA 815-R-99-012, Environmental Protection Agency, Washington, DC, 1999. [35] Y.P. Chin, G. Aiken, E. O’Loughlin, Environ. Sci. Technol. 28 (1994) 1853. [36] Y. Matsui, A. Yuasa, F.S. Li, Water Sci. Technol. 40 (9) (1999) 223. [37] F.S. Li, Ph.D. dissertation, Gifu University, Japan, 1998 [in Japanese]. [38] J. Buffle, P. Deladoey, W. Haerdi, Anal. Chim. Acta 101 (1978) 339. [39] A. Yuasa, F.S. Li, Y. Matsui, in: F. Meunier (Ed.), Fundamentals of Adsorption 6, Elsevier, Amsterdam, 1998, p. 407.
Polydisperse adsorbability composition of several natural and synthetic organic matrices Fusheng Li,a,∗ Akira Yuasa,b Hajime Chiharada,c and Yoshihiko Matsui a a Environmental Engineering Division, Department of Civil Engineering, Gifu University, 1-1 Yanagido, Gifu 501-1193, Japan b River Basin Research Center, Gifu University, 1-1 Yanagido, Gifu 501-1193, Japan c Graduate School of Engineering, Gifu University, Yanagido, Gifu 501-1193, Japan
Received 4 March 2003; accepted 7 May 2003
Abstract The polydisperse composition of nine dissolved organic materials (DOMs) from two river water sources, one ground water source, two biologically treated wastewater sources, and two commercial sources was analyzed based on their adsorbabilities by activated carbon. For each DOM, batch adsorption isotherms measured for both TOC and UV260 were analyzed using an overall isotherm model derived from the IAST–Freundlich expression. By accounting for the heterogeneity of each DOM with a log-normal distribution of the Freundlich parameter (K), its adsorption behavior was characterized with only four parameters (including three fitting ones). The average adsorptive strength (KM ) and heterogeneity (σ ) determined for all DOMs, which were defined by the mean value and the standard deviation of the log-normal distribution of the Freundlich K, changed over the ranges 2.5–62.2 and 0.22–0.97 (mg/g)/(mg/l)1/n , respectively, when the TOC index was used. Among all DOMs studied, a river water DOM at the upper stream was found least heterogeneous: the Freundlich K of its organic constituents varied in the range 10.8–190 (mg/g)/(mg/l)1/n , as compared to a commercial humic acid that exhibited the broadest Freundlich K distribution of 0.01–1494.3 (mg/g)/(mg/l)1/n . KM and σ , along with other two parameters (the Freundlich exponent 1/n and the nonadsorbable organic fraction parameter Cnon /CT 0 ), changed with both indices of TOC and UV260 in a regular manner, indicating that UV-absorbing organic molecules possessed adsorbabilities different from non-UV-absorbing ones. Also based on HPSEC chromatograms measured for solutions before adsorption, the molecular weight composition of all DOMs was also assessed and the molecular size impacts on adsorption characteristics of DOMs were briefly discussed. 2003 Elsevier Inc. All rights reserved. Keywords: Dissolved organic matter; Humic acid; Activated carbon; Adsorption; Molecular weight; Distributed fictive component method
1. Introduction A significant portion of dissolved organic materials (DOMs) present in the solution matrices of aqueous water systems, about 40–90% of total dissolved organic carbon, is composed of humic and fulvic acids [1,2]. Humic substances are expected to be removed from drinking water supplies as they react with chlorine to produce carcinogenic disinfection byproducts (DBPs). Furthermore, even if not targeted for removal, their presence can adversely affect the performance of water treatment processes in removing specifically targeted synthetic organic chemicals (SOCs). For instance, humic substances are capable of competing with small SOCs for adsorption sites and of hindering * Corresponding author.
E-mail address: [email protected] (F. Li). 0021-9797/$ – see front matter 2003 Elsevier Inc. All rights reserved. doi:10.1016/S0021-9797(03)00526-5
the access of SOCs to small pores where their adsorption takes place more efficiently [2–7]. DOMs are mixtures of organic compounds having different physicochemical properties. Conventional water treatment processes can generally remove about 10–90%, with an average of approximately 30%, of the constituting compounds having large molecular weight (MW) or size [8,9]. To remove the remaining small compounds, the 1986 Amendments to the Safe Drinking Water Act of the USA have identified activated carbon adsorption as one of the best available technologies. The performance of activated carbons (ACs) in removing polydisperse DOMs is affected by a variety of factors. Using natural and synthetic organic materials, several researchers have found that molecular size and chemical characteristics (including the functional group, charge density, aromaticity) are important features of organics that influence their adsorption capacities on ACs [1,2,10–15]. Based on the adsorption
266
F. Li et al. / Journal of Colloid and Interface Science 265 (2003) 265–275
results of several organic macromolecules extracted from such materials as soil, peat, and coal, a few researchers have further pointed out that molecular size effects may dominate the competitive adsorption among constituting compounds and chemical effects may play only minor roles [2,6,12]. In addition to the inherent physicochemical features of organic matters, solution chemistry, represented by variables such as pH and ionic strength [10,11,16–20], as well as the surface and pore characteristics of ACs [5,7,10,18,20], also affect their uptake from aqueous systems. All these findings have significant practical implications for the design and operation of AC and pretreatment processes. However, due to the likely involvement of so many factors, adsorption of polydisperse organic matters is generally considered much more complicated than adsorption in single-solute systems. Batch adsorption isotherms of natural and synthetic organic materials, which are usually assessed using lumped water quality indices such as total dissolved organic carbon (TOC) and ultraviolet absorbance (UV absorbance), exhibit multicomponent adsorption behavior [2,11,12,19,21–33]. Therefore, characterization of their composition in terms of constituting components’ adsorbabilities is of significant interest as it could provide much information for assessing their removal from water and wastewater solutions as well as for assessing their competitive impacts on the uptake of SOCs. Accordingly, the main objective of the present study was to investigate the adsorbability composition of several aqueous DOMs from different water and wastewater sources. For comparison, the adsorption behaviors of two humic products from commercial sources were also examined. A distributed fictive component method was used to analyze isotherm data assessed by both DOC and UV absorbance, respectively. By doing so, the adsorption characteristics of each organic material were reflected by only four parameters. Comparisons among all organic materials were thus made based on these four parameters determined. A brief description of the method is to be given later. Moreover, using a high-performance size exclusion chromatography (HPSEC) system, the molecular weight (MW) polydispersity of all organic materials before adsorption was examined. The dependency of the overall adsorption extent on the molecular weight of organic materials from different sources was also briefly discussed. An expectation that guided this discussion was that some parametric trends probably existed between parameters devised to reflect the overall MW and adsorptive characteristics of all organic materials used, if physical size was the predominant factor that controlled the adsorption behaviors of organic constituents, as suggested in [2,6,12].
2. Experimental materials and methods 2.1. Dissolved organic materials (DOMs) Seven dissolved organic materials contained in four river water samples, a ground water, and two biological process
effluents of a wastewater treatment plant were used along with two commercial humic acids. Four river water samples were collected from Tokoro River water (TRW) and Nagara River water (NRW) at the upper, middle, and lower streams (NRW-US, NRW-MS, NRW-DS), respectively. TRW was chosen because it is representative of most drinking water sources in Hokkaido, the second largest island of Japan, in the respect that water sources there generally contain relatively higher levels of color-causing humic substances. Nagara River water was used because it supplies sufficient water for social and industrial activities in the Gifu and Aichi prefectures of central Japan. The ground water (KGW) was collected from a shallow well at Kitamura village, Hokkaido. KGW was used because the organic matter contained is mainly from peat and such water samples have frequently been used to characterize the behavior of naturally occurring humic substances during physicochemical water treatment processes [2,6,12,13]. Two wastewater samples after biological treatment (WBT-1 and WBT-2) were collected from a wastewater treatment plant in Gifu prefecture. Although the origin of treated wastewater DOMs differs from that of natural organic materials, they generally exhibit many features of the latter in aspects such as the MW, the charge potential, and physicochemical treatment behavior. As model organic macromolecules used to characterize the behaviors of natural organic materials in water environmental systems [1, 2,11,12,17–19], Wako humic acid (WHA), a commercial product extracted from coal, and Aldrich humic acid sodium salt (AHA) were taken directly as received from Wako Pure Chemicals Co., Japan, and Aldrich Co. Ltd., USA, respectively. Except for WHA and AHA, the DOM stock solutions were obtained by filtering collected water samples through 0.2-µm membrane filters (Toyo Roshi, Japan) followed by adjustment of pH to 7.0 as necessary. The stock solution of WHA was prepared by dissolving the material in Milli-Q water raised to pH 11 with 0.5 M NaOH, lowering water pH to 7.0 with 0.5 M HCl, and filtering the solution through 0.2-µm membrane filters. The stock solution of AHA was obtained by dissolving the material in Milli-Q water, adjusting water pH to 7.0, and filtering the solution through 0.2-µm membrane filters. All stock solutions were stored refrigerated at 5 ◦ C in the dark to prevent biological activity. 2.2. Activated carbon The adsorbent used in this research was Filtrasorb-400 granular activated carbon (Calgon Co., USA), which has an average pore radius of 12.0 Å, a total pore volume of 0.566 cm3 /g, and a total surface area of 950 m2 /g [2]. This activated carbon (AC) was chosen because it is widely used in advanced water treatment applications and laboratory research. The granular AC was pulverized and sieved to particles with sizes below 45 µm, washed and rinsed with distilled water to remove fines, dried at 105 ◦ C overnight, and finally stored in a desiccator till use. Before being subjected
F. Li et al. / Journal of Colloid and Interface Science 265 (2003) 265–275
267
Table 1 The concentrations of DOMs in stock solutions with molecular weight information Source of DOM TRW NRW-US NRW-MS NRW-DS KGW AHA WHA WBT-1 WBT-2
UV260
TOC
SUVA
(m−1 )
(mg/l)
(m−1 /(mg/l))
4.3 1.2 1.7 3.2 30.0 33.6 23.5 230.0 106.0
1.9 0.7 0.8 1.3 9.1 4.1 2.0 105.7 43.8
2.30 1.76 2.13 2.51 3.30 8.12 11.85 2.18 2.42
to pulverizing, the granular AC was washed several times with and subsequently boiled in Milli-Q water for about 1 h to remove fines, washed and rinsed again several times, and finally dried at 105 ◦ C to minimize likely impacts of ash impurities on the adsorbent surface. 2.3. Isotherm experiments Batch isotherm experiments were conducted according to the bottle-point method of variable AC doses in 500-ml flasks sealed with Teflon-lined rubber septa. For each water sample, to measure the isotherm dependency on DOM’s initial concentrations, two or three working solutions with variable initial concentrations were prepared by diluting the stock solution with Milli-Q water as necessary. To minimize the influence of changes in solution chemistry caused by dilution on adsorption [10,11,16–20], for working solutions of each organic material, pH was controlled at 7.0 and the ionic strength was adjusted with 0.5 M NaCl based on the electrical conductivity of the working solution prepared with the smallest dilution ratio. After reaching equilibration by shaking the flasks containing 200 ml of working solutions placed on a rotation shaker at 200 rpm in a temperature-controlled room (20 ◦ C) for 7 days, solutions were filtered through 0.45-µm membrane filters (Toyo Roshi, Japan) rinsed beforehand with Milli-Q water. A preliminary adsorption rate study showed no measurable changes in liquid-phase concentrations after agitation for up to 6 days, thus ensuring that equilibration was reached by shaking for 7 days. DOM concentrations in the filtrates were quantified with TOC and UV absorbance at 260 nm (UV260), for which a TOC analyzer (Model TOC810, Sievers, USA) and a UV–visible spectrophotometer (Model U-3210, Hitachi Co., Japan) were used, respectively. Based on the results of control tests conducted without the addition of AC, there were no measurable losses of DOMs and thus all changes in the concentration were attributed to adsorption. The measurement of UV260 was made because, like UV absorbance at 254 nm (UV254), UV260 is capable of reflecting the presence of humic molecules present in water environment systems, and the specific ultraviolet absorption (SUVA), a calculated parameter of UV260 or
Range of MW
Mw
Mn
Polydispersity
3174 2972 2928 3033 2603 3185 2510 2026 1977
1.089 1.141 1.136 1.118 1.101 1.152 1.259 1.170 1.187
(g/mol of PSS) 842–6637 855–7365 805–7365 771–7365 1003–5244 1013–6797 908–7668 500–4449 450–4500
3456 3391 3326 3392 2865 3667 3161 2371 2347
(–)
UV254 divided by TOC, has been thus used as an indicator of the humic content. The US EPA [34] reports that water with a small SUVA contains primarily nonhumic organic compounds and is not amenable to enhanced coagulation. The values of TOC, UV260, and SUVA for all stock solutions are summarized in Table 1. Throughout the study, the values of UV260 are given in units of m−1 , which were converted from measured values with a regular 10-, 50-, or 100-mm cell (a longer cell was used for solutions with lower DOM concentrations). In addition to ease of analysis, the measurement of UV absorbance also has the advantage of good reproducibility: the detector response variance for repeated UV260 measurements was within ±0.0005. Apparent SUVA differences shown in Table 1 indicated that the humic content changed with the sources and the types of DOMs. 2.4. High performance size-exclusion chromatography (HPSEC) HPSEC was carried out to measure the MW distributions of all DOMs. A packed silica column (GL-W520-X 10.7 × 450 mm, Hitachi Co., Japan) and a UV–visible detector (Model LC-10AV, Shimadzu Co., Japan) set at 260 nm (UV260) were used. The mobile phase solution made up of Milli-Q water was buffered with phosphorous salts (0.02 M Na2 HPO4 + 0.02 M KH2 PO4 ) and was supplied to the chromatographic column at 0.5 ml/min. Three polystyrene sulfonates (PSS) with known molecular weights (MWs) of 1430, 4950, and 6530 g/mol were used to calibrate the HPSEC system and a good correlation (R 2 = 0.998) described by log10 (MW) = −0.1096t + 5.029 was obtained (t is the elution time). Semilog linear correlations between log10 (MW) and t were also confirmed by others [2,12]. 3. Results and discussion 3.1. Molecular weight characteristics of DOMs HPSEC chromatograms of all DOMs used in the study are displayed in Figs. 1a–1c. As shown, the MW distributions differed apparently with DOM sources. For river water
268
F. Li et al. / Journal of Colloid and Interface Science 265 (2003) 265–275
those of aqueous DOMs. Kilduff et al. [2] conducted HPSEC measurements for humic acids extracted from Laurentian soil (LaHA) and peat (PHA), respectively. Chromatograms that resembled each other (each had two peaks and the first peak eluted earlier from the column was much smaller than the second peak eluted later) but differed obviously from that of the Huron River water DOM also measured by them were obtained. The results we obtained in the present study, together with previous ones, indicate that humic products extracted from such natural materials as soil, peat, and coal following various procedures possess MW characteristics different from those of aqueous ones present in water and wastewater sources. Based on the chromatograms obtained, the weight-averaged (Mw ) and number-averaged (Mn ) molecular weights were determined according to two expressions appearing in the literature [35], N N Mw = (1) MWi (t)hi (t)t hi (t)t, i=1
Mn =
N i=1
Fig. 1. Size-exclusion chromatograms of DOMs. Values in parentheses are DOM concentrations of working solutions used for HPSEC measurements.
DOMs, TRW, NRW-US, NRW-MS, and NRW-DS exhibited chromatograms similar to one another, with the magnitude of detector responses changing with DOM concentrations (Fig. 1a). Ground water (KGW) showed a chromatogram (Fig. 1b) that resembled those of four river water DOMs and was also in close similarity to the documented one measured for Huron River water DOM [2]. For treated wastewater DOMs, similar chromatographic shapes were obtained for samples WBT-1 and WBT-2 (Fig. 1b). However, the number of chromatographic peaks they possessed was larger than river water and the ground water DOMs. Among all organic materials studied, the commercial humic products AHA and WHA had the fewest of chromatographic peaks, two and three, respectively (Fig. 1c), and exhibited chromatographic shapes markedly different from
hi (t)t
i=1 N
hi (t)t MWi (t) ,
(2)
i=1
where MWi (t) is the MW as a function of elution time t, hi (t) is the UV detector response and t is the time interval. Then, based on determined values of Mw and Mn , the MW polydispersity (= Mw /Mn ), a parameter that has been commonly used [2,12,35] to assess the MW heterogeneities of organic matrices, was also computed. As summarized in Table 1, the values of Mw varied with organic materials, following a general trend that Aldrich humic acid (AHA) > river water DOM > Wako humic acid (WHA) > ground water DOM > treated wastewater DOM. A similar trend was also found for the number-averaged MW, Mn . However, compared to the differences in Mw and Mn , the differences in polydispersity were less obvious, with its values following the decreasing order Wako humic acid (WHA) > treated wastewater DOM > Aldrich humic acid (AHA) > river water DOM ∼ = ground water DOM. The trend observed supports a previous finding that surface water organic materials was less disperse than organic materials extracted from soil and peat [2]. For all three samples collected from the Nagara River (NRW-US, NRW-MS, and NRW-DS), Mw , Mn , and polydispersity of the river water DOM changed only slightly as the water flowed from the upper to the down streams, even if the content of organic matter assessed by TOC was increased by approximately 90% (as shown in Table 1). Besides the commonly used definition of MW polydispersity as Mw /Mn , the MW range may also serve as an informative parameter in reflective of the MW characteristics of organic mixtures. The results summarized in Table 1 displayed that, compared to the differences reflected by the parameter Mw /Mn , those reflected by the MW range were somewhat more obvious among all DOMs investigated. However, it has to be noted that the MW range
F. Li et al. / Journal of Colloid and Interface Science 265 (2003) 265–275
parameter includes no information with respect to the chromatographic shapes as well as the presence levels of DOM constituents. 3.2. Adsorption characteristics of DOMs 3.2.1. Observed adsorption isotherms The adsorption isotherms of TRW, NRW-MS, AHA, and WBT-1 assessed by TOC are displayed in Fig. 2 as examples. As illustrated by symbols, reflecting the polydisperse adsorbabilities of organic constituents, the overall adsorption capacity (qT ) changed with the initial TOC concentrations (CT 0 ). The isotherm shifted its position upward as CT 0 decreased, indicating larger adsorption capacities for diluted working solutions. Besides, due to preferential adsorption of components with strong adsorbabilities, qT at low AC doses (CAC ), where the adsorption sites were limited, was extremely high. These features have been reported by numerous researchers for a variety of natural and synthetic organic matrices [2,5,6,11,12,19,21–33] and were also observed for TOC-based isotherms of KGW, NRW-US, NRWDS, and WBT-2, as well as for UV260-based isotherms of all organic materials used in this study (data not shown). 3.2.2. Model description of observed isotherms To describe the adsorption isotherms of polydisperse organic matters, a fictive component method (FCM) that uses a series of hypothetical components to account for competi-
269
tions among organic constituents with the aid of multicomponent adsorption models [3,30], in many cases the ideal adsorbed solution theory (IAST), has been applied [21–23, 27–29]. However, this method has the drawback of requiring a search for a large number of variables, namely the equilibrium parameters and initial concentrations of all hypothetical components. Thus, the values found are not always unique. Harrington and DiGiano [21] discussed the limitations of the FCM. They found that, even using a small number of hypothetical components (two adsorbable ones and a nonadsorbable one), which involved totally seven parameters, the description of isotherms observed for a natural organic material had severe computational and statistical problems. Besides this, the effects of preozonation on adsorption equilibrium were found not to be interpretable from the parameters determined. They thus suggested that, to promote the method’s application in practice, modifications that could lead to a significant reduction of adjustable variables were appealing. Different from the common practice of using a discrete number of fictive components, we proposed a distributed fictive component method (DFCM), in which a straightforward equation derived from the IAST–Freundlich model and a distribution function for the Freundlich K of all adsorbable DOM components were used [14]. Requiring searching for only three parameters, this method satisfactorily described and predicted the isotherms [14,31] of DOMs in wastewater and peat water sources. The equilibrium parameters
Fig. 2. The TOC-based overall isotherms of DOMs. Symbols are observed data. Lines are model fits to the observed data. Values in parentheses show the DOM concentration in working solutions used.
270
F. Li et al. / Journal of Colloid and Interface Science 265 (2003) 265–275
determined accordingly were then successfully applied to describe the breakthrough behavior of the total organic material in a fixed-bed adsorber [32]. Pretreatment effects of coagulation, ozonation, and chlorination on adsorption capacities of these two organic materials were also reasonably explained [36]. Hence, in this study, this method was employed to analyze all isotherm data obtained. A brief description of the method is given below. By assuming that the Freundlich exponent (1/ni ) and the conversion ratio (αi ) of molar concentration to the concentration of a lumped quality index (TOC or UV260) are identical for all adsorbable components (1/ni = 1/n, αi = α), the overall isotherms of an organic matrix at variable initial concentrations can be described by N i=1
Ci0 /(CT 0 −Cnon ) [qTn /(CT 0 −Cnon )]Ki−n +1−(CT −Cnon )/(CT 0 −Cnon )
= 1,
(3)
where N is the total number of components, Ki is the Freundlich parameter and Ci0 is the initial concentration of component i, Cnon is the nonadsorbable component concentration, CT 0 is the overall initial concentration, CT is the overall liquid phase equilibrium concentration, and qT is the overall solid phase equilibrium concentration. This equation is an extended version of and can be directly derived from the documented IAST–Freundlich model [26]. The appropriateness of assuming identical values of 1/ni and αi for all adsorbable components has been tested by several researchers and is thus applied as a relatively common practice in modeling the adsorption isotherms of organic materials with unknown composition [21–24,27]. With such assumptions, the heterogeneity of organic constituents in adsorbabilities is mainly accounted for by the Freundlich parameter Ki and the initial concentration Ci0 to be searched. When Freundlich K of all adsorbable constituents is described by a distribution function, f (K), the item Ci0 / (CT 0 − Cnon ) in Eq. (3) can be determined as Ci0 /(CT 0 − Ki+K/2 Cnon ) = Ki−K/2 f (K) dK, where K is the interval used to divide the whole distribution range of Freundlich K based on the total number of hypothetical components (N). Thus, a distributed overall isotherm equation can be obtained: ∞ f (K) dK [qTn /(CT 0 −Cnon )]K −n +1−(CT −Cnon )/(CT 0 −Cnon )
= 1.
standard deviation (σK ) of log10 (K), the Freundlich exponent (1/n), the overall initial concentration (CT 0 ), and the concentration of the nonadsorbable fraction (Cnon ): [log10 (K/KM )]2 1 exp − . f (K) = √ (5) 2σ 2 2πσ Since CT 0 and Cnon in Eq. (4) can be measured independently (observed data at higher AC dosages are used for determination of Cnon ), parameters to be searched are limited to only three: KM , σ , and 1/n. These three parameters are separately sensitive to each portion of the isotherm’s curvatures; hence, a unique set of parameters can logically be obtained from the observed isotherm data according to the following objective function (Er) to be minimized. For a given DOM, isotherm data measured for two or more initial concentrations are simultaneously required in the search because 1/n is a characteristic parameter reflecting the isotherm dependency on initial concentrations [22–24,26]. The objective function is
ND
CT (obs,i) − CT (cal,i)
1
Er =
2ND CT (obsm) i=1
CT (obs,i) − CT (cal,i)
, +
(6)
CT (obs,i) where CT (cal,i) is the calculated liquid phase concentration, CT (obs,i) is the observed liquid phase concentration, CT (obsm) is the arithmetic average of the observed liquid phase concentration, and ND is the number of data points obtained for all initial concentrations of a given organic material. Simulation results that best fitted the observed TOCbased isotherms for TRW, NRW-MS, AHA, and WBT-1 are displayed by solid lines in Fig. 2 as examples. For all DOMs studied, the simulation results described the observed isotherms with reasonably high accuracies. The error defined by Eq. (6) was in the range of 2.97–10.7% for the TOC-based isotherms and of 2.48–8.09% for the UV260based ones, as summarized in Tables 2 and 3, respectively. The results obtained, together with our previous ones using the DFCM for description and prediction of the adsorption equilibrium and column breakthrough of aqueous
(4)
0
To find a distribution function suitable for application, Li [37] tested a large number of distribution patterns defined by functions such as the triangular function, the normal distribution function, and the logarithmic normal distribution function. Model simulations demonstrated clearly that the log-normal distribution function was most suitable, since it was capable of generating isotherm shapes most close to observed ones and involves only two explanatory variables. With this function, as described by Eq. (5), total parameters regulating the overall isotherms of DOMs described by Eq. (4) are further reduced to the mean value (KM ) and the
Table 2 Parameters determined from TOC-based isotherms DOM source
KM
σ
Range of K
((mg/g)/(mg/l)1/n ) TRW NRW-US NRW-MS NRW-DS KGW AHA WHA WBT-1 WBT-2
40.4 45.3 62.2 48.2 37.0 4.0 2.5 42.5 50.5
0.30 0.22 0.39 0.41 0.40 0.72 0.97 0.43 0.36
5.6–290.8 10.8–190.0 4.7–825.6 3.3–715.3 2.7–514.1 0.03–463.7 0.01–1494.3 2.5–728.9 4.9–525.4
Cnon /CT 0
Er
(–)
(–)
(%)
0.26 0.33 0.29 0.20 0.35 0.20 0.21 0.35 0.25
0.15 0.38 0.14 0.15 0.04 0.00 0.00 0.02 0.03
5.66 10.70 4.63 8.65 4.95 2.97 3.21 5.66 7.03
1/n
F. Li et al. / Journal of Colloid and Interface Science 265 (2003) 265–275
Table 3 Parameters determined from UV260-based isotherms DOM source
KM
σ
Range of K
(m1/n−1 /(g/l)) TRW NRW-US NRW-MS NRW-DS KGW AHA WHA WBT-1 WBT-2
125.2 94.6 141.1 115.2 123.6 22.2 20.0 122.0 134.5
0.22 0.11 0.25 0.27 0.22 0.60 0.79 0.42 0.34
29.8–525.4 47.1–190.0 27.8–716.7 19.1–694.27 29.5–518.7 0.42–1165.1 0.11–3717.5 7.7–1933.6 14.8–1226.7
1/n
Cnon /CT 0
Er
(–)
(–)
(%)
0.20 0.29 0.27 0.19 0.27 0.18 0.20 0.33 0.25
0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.02 0.00
6.95 5.39 8.09 6.70 2.48 5.43 5.31 7.94 4.03
DOMs [31,32], further indicate that this method is effective for characterizing aqueous organic matrices present in surface water and wastewater sources, as well as synthetic humic products. Further validations will be made using marine organic matter and organic matter from different latitudes. 3.2.3. Adsorbability distributions Four parameters (KM , σ , 1/n and Cnon ) determined for all DOMs based on the TOC and UV260 isotherms are summarized in Tables 2 and 3, respectively. As a parameter devised to reflect the mean adsorptive strength of a given organic material, KM differed considerably with respect to the sources and types of DOMs. The TOC-based KM varied in the range 2.5–62.2 (mg/g)/(mg/l)1/n, while the UV260based one varied in the range 20.0–141.1 m1/n−1 /(g/l). The largest KM difference was found between the river water NRW-MS and the commercial WHA, with the values of the former being approximately 25 and 7 times as large as the values of the latter for the indices of TOC and UV260, respectively. The larger difference noticed with the TOC index may imply that some organic compounds not detected by UV260 in the river water source NRW-MS were adsorbed to much greater extents than UV-absorbing humic macromolecules. Such an implication is considerable since UV260 (or UV254) is generally used as an index representing the content of humic substances or organic macromolecules having properties similar to humic ones. As an important parameter devised for assessing the polydispersity of organic materials in adsorbabilities, the TOC-based σ changed with DOMs in the decreasing order: WHA > AHA > WBT-1 > NRW-DS > PW > NRW-MS > WBT-2 > TRW > NRW-US. The much larger σ values noticed for WHA (0.97) and AHA (0.72) as compared to aqueous source DOMs (0.22–0.41) suggested that, among all organic materials studied, the commercial humic acids were most heterogeneous with respect to their constituting compounds’ adsorption behavior. In contrast, the smallest σ noticed for NRW-US (0.218) implied that this river water DOM was the least heterogeneous matrix. It was also noteworthy that for all organic materials studied, the TOC-based σ values (in the range 0.22–0.97) were about 1.1–2.1 times as large as the UV260-based ones (in the range 0.11–0.79).
271
Such a result is also expectable since, as mentioned earlier, while UV260 is an index reflective of only the portion of organic compounds having UV absorbing features, such as humic substances, the index TOC measures all organic molecules dissolved in water solutions, including those not detectable with a UV–visible spectrophotometer. For the Nagara River water DOM, the values of σ assessed by both TOC and UV260 varied with the sampling points in the ranges 0.22–0.41 and 0.11–0.27, respectively. However, it was interesting to see that, for both these indices used, the changes in the magnitude of σ followed the increasing order of NRW-US < NRW-MS < NRW-DS. This thus indicated that the adsorption polydispersity of the river water DOM was becoming expanded as the river water flowed downstream. The expansion might be caused by the merging of some organic compounds having strong (or much weak) adsorbabilities into the river water through some of its tributary streams. The ranges of Freundlich K determined for all DOMs based simply on the corresponding KM and σ values are tabulated in Tables 2 and 3. As could be seen from the results, the TOC-based ones in Table 2, for instance, along the water flow direction, the range of the Freundlich K of the river water DOM changed from 10.8–190.0 (NRW-US) through 4.7–825.6 (NRW-MS) to 3.3–715.3 (NRW-DS). For a given organic material with unknown composition, the Freundlich exponent 1/n, which is assumed to be identical for all constituting compounds to facilitate the modeling of its overall isotherm, is regarded as a characteristic parameter reflecting the overall isotherm’s dependency on initial concentrations [22–24,26]. Meanwhile, it is also a parameter indicative of the overall adsorption energy (i.e., the affinity) between organic molecules and the internal AC pore surfaces [5]. As summarized in Tables 2 and 3, although 1/n changed with DOM, its values were relatively small: 0.20–0.35 (for the TOC index) and 0.18–0.33 (for the UV260 index). This thus implies that all DOM studied in the present work had stronger affinities with the AC used. Moreover, it was also interesting to see that, except for the treated wastewater DOM in WBT-2, the UV260-based 1/n values were comparatively smaller than the TOC-based ones. This may thus indicate that, of the organic constituents of DOMs, UV-absorbing humic molecules had stronger affinities with the AC surfaces as compared to those coexistent organic species not possessing UV-absorbing features. It is possible that the adsorption capacity of organic compounds that do not possess UV-absorbing features is larger, as discussed earlier based on the KM values; e.g., if these compounds have access to AC pores from which larger UV-absorbing molecules are excluded. Nonetheless, these compounds may have weaker affinities for the AC surfaces because they might be more hydrophilic than UV-absorbing ones. In addition to KM , σ , and 1/n, the parameter Cnon /CT 0 devised to reflect the nonadsorbable organic fraction also demonstrated changes with the sources and types of DOMs, as tabulated in Tables 2 and 3. However, compared to the TOC-based Cnon /CT 0 values (in the range 0–0.38), the
272
F. Li et al. / Journal of Colloid and Interface Science 265 (2003) 265–275
Fig. 3. Freundlich K distributions of DOMs depicted using TOC-based values of KM and σ .
UV260-based ones for most DOMs were zero, suggesting that UV-absorbing organic compounds were completely adsorbable. It could be thus inferred that the nonadsorbable TOC fraction was coming from some compounds not possessing UV-absorbing features. For DOMs contained in samples of TRW, WBT-1 and WBT-2, although non-adsorbable organic fractions were also detected for the UV260 index, the Cnon /CT 0 values were smaller than TOC-based ones. The results shown above demonstrated that, by analyzing observed isotherm data with the DFCM, the polydisperse composition of organic materials could be assessed based on their constituents’ adsorption behaviors. Accordingly, as displayed in Fig. 3 as examples, the polydisperse feature of a given organic material could be explicitly evaluated according to the Freundlich K distribution depicted by simply introducing the determined KM and σ values into Eq. (5). Comparisons of all four characteristic parameters involved in the DFCM for all organic materials used in the present work indicated that the adsorption of polydisperse organic materials by activated carbon differed with their types and sources. However, the differences were relatively small among river water, ground water, and biologically treated wastewater DOMs. Commercial humic acids exhibited adsorption behaviors very different from aqueous DOMs. Some caution must be exercised when such humic products are used for characterization of the adsorption behaviors of natural organic materials.
Fig. 4. Effects of humic content on the adsorbability of DOMs: (a) the mean adsorptive strength (KM ) vs the humic content indicator (SUVA); and (b) the adsorptive strength polydispersity (σ ) vs SUVA. The dashed lines exhibit the general parametric trends.
3.3. Adsorption dependency on humic content and molecular weight To investigate the humic content effects on the adsorbability of total organic matrices, the adsorption parameters KM and σ determined for the total organic carbon index TOC were plotted against the values of SUVA shown earlier in Table 1. Although there were some exceptions, the parametric line shown in Fig. 4a demonstrated a decreasing trend of KM with increasing SUVA. This trend supported the indication given earlier that humic molecules were adsorbed to a less extent as compared to other coexisting organic compounds exhibiting non UV-absorbing features. Contrary to this trend, the parametric line in Fig. 4b displayed that the heterogeneity parameter (σ ) increased with increased SUVA, thus implying that humic molecules were more disperse with regard to their adsorption behaviors. However, it must be pointed out that both these parametric trends were heavily influenced by the results of the commercial humic acids (AHA and WHA), which behaved much differently from those aqueous DOMs with respective to their SUVA, KM , and σ values. If these two humic acids were excluded, i.e., when analyses were made based on the results of seven aqueous DOMs alone, both trends would become a bit vague.
F. Li et al. / Journal of Colloid and Interface Science 265 (2003) 265–275
Fig. 5. Weight-averaged molecular weight (Mw ) versus mean adsorptive strength (KM ) of all DOMs.
Fig. 6. Molecular weight polydispersity versus the Freundlich K heterogeneity (σ ) of all DOMs.
The relationship between KM and Mw is plotted in Fig. 5. Similarly, the relationship between σ and the MW polydispersity is plotted in Fig. 6. Since all MW-related information (Mw , Mw , and polydispersity shown in Table 1) was generated from HPSEC chromatograms measured with a UV–visible detector, for better correspondency, the values of KM and σ determined for the UV260 index (shown in Table 3) were used. This investigation was guided by the expectation that some parametric trends probably existed, if, as previously reported [2,6,12], molecular sizes dominated the adsorption of organic materials and the chemical effects played only minor roles. As illustrated in Fig. 5, a parametric trend revealing either increasing or decreasing adsorptive strength (KM ) with increasing molecular weight (Mw ) was not exhibited. Such a trend was also confirmed not to exist when the number-averaged molecular weight (Mn ) was used instead of the weight-averaged molecular weight (Mw ). Regarding the relationship of the adsorption polydispersity (σ ) with the MW polydispersity, the data plotted in Fig. 6 did not reveal an expected parametric trend of expanding the distribution of Freundlich K with the increase of MW polydispersity. An analysis based on TOC-based KM and σ values was also conducted; however, no trend was obtained (results not shown). In previous studies [2,35,38], several researchers investigated the relationships between
273
UV-absorptivity (SUVA) of humic substances and HPSEC weight-averaged molecular weight (Mw ). Combined analysis of our results shown in Figs. 4–6 indicated no trend between these two parameters no matter if plots for two humic acids were excluded or not (direct Mw –SUVA plotting showed the same result). The indication coincided with the result shown in the literature [38], but was in contradiction to the findings of others who showed either an increase [35] or a decrease in UV-absorptivity with increased molecular weight [2]. For a given type of organic material, preferential adsorption of low-MW constituents might be a general feature of adsorption from polydisperse organic material solutions [2,6,11,12], although some exceptions were also observed [10,35,39]. For organic materials from different sources, however, besides molecular size effects, the chemical nature of organic materials as well as the solution chemistry may also play significant roles. Karanfil et al. [12] investigated the relationship between the adsorption capacities and the Mw values (measured by HPSEC) of nine commercial and synthetic macromolecular organic materials prepared from such materials as soil, peat, and coal. Even if isotherm experiments were performed under identical solution chemistry, data from four organic materials were found to deviate totally from the parametric order of decreasing adsorption capacities with increasing molecular weights. These deviations do not support and are in confrontation with the idea that molecular size is the only, or even the most important parameter controlling adsorption of organic macromolecules. With respect to the impacts of solution chemistry, numerous researches have demonstrated increased adsorption capacity with decreasing water pH [10,17,18,20] and with increasing concentrations of cationic ions, especially Ca2+ and Mg2+ [17–19]. Humic substances and ACs generally contain functional groups that can be ionized to exhibit anionic characteristics, with the extent of ionization varying according to the inorganic chemistry of solutions. Taking this feature into account, three categories of interactions have been proposed [17] to explain the potential mechanisms involved: (i) interactions between solution cationic ions (including H+ ) and ACs to neutralize the repulsive forces between anionic adsorbates and ACs; (ii) interactions between solution cationic ions and adsorbates to reduce the charge potentials of anionic adsorbates and hence to form aggregates more favorable for adsorption; and (iii) interactions between adsorbed adsorbates and solution cationic ions that possibly bring about changes in the packing, spacing, or alignment of adsorbed molecules within the AC pores. In the present study, since all solutions were adjusted to pH 7.0 before being subjected to isotherm experiments, pH could be excluded as a likely factor that attributed to different uptake behavior among all DOMs investigated, especially between aqueous ones and commercial ones. It was thus conceivable that likely affecting factors were probably relative to DOMs’ physicochemical features and the inor-
274
F. Li et al. / Journal of Colloid and Interface Science 265 (2003) 265–275
ganic composition of solutions, which need further investigations. Results we obtained in the present study do not necessarily and are not sufficient to indicate that the size effects are not significant. However, they further stress the issue related to potential impacts of chemical characteristics of organic materials (including the functional group, charge density, and aromaticity) and the chemistry of solutions. In a coming paper, the issue of molecular size effects for aqueous organic matters will be addressed in details based on HPSEC chromatograms measured for a larger number of organic matrices (including all those used in the present study) after reaching equilibrium at variable AC dosages. Physicochemical properties of ACs reflected in such aspects as pore size distribution, surface area, and surface chemistry are also important factors that influence the adsorption capacity of DOMs [1,5,7,11] as well as the competitive impacts of DOMs on the uptake of SOCs [2–7]. In this study, a single coal-based AC, which is widely used in advanced water treatment applications and laboratory research works, was used. However, taking into consideration that competitive adsorption is a general feature of adsorption from multicomponent mixtures, the trends and findings obtained were considered generic in reflecting the polydisperse adsorbabilities of aqueous and commercial organic matrices.
4. Conclusions The composition of nine organic materials from different sources was investigated based on their activated carbon adsorbabilities. For each organic material, the constituents’ heterogeneity in adsorbabilities was accounted for using a log-normal distribution of the Freundlich parameter (K). An overall isotherm expression derived based on the IAST– Freundlich model successfully described the isotherms of all organic materials measured by both TOC and UV260, respectively. The adsorption behavior of organic materials differed with their sources and types: the average adsorptive strength defined by the mean value of the log-normal distribution of Freundlich K (KM ) changed over the ranges of 2.5–62.2 (mg/g)/(mg/l)1/n and 20.0–141.1 m1/n−1 /(g/l); and the adsorptive heterogeneity defined by the standard deviation of the corresponded distribution (σ ) changed over the ranges of 0.22–0.97 (mg/g)/(mg/l)1/n and 0.11–0.79 m1/n−1 /(g/l), for the indices of TOC and UV260, respectively. Compared to organic materials from river water, ground water, and treated wastewater sources, the overall adsorption capacities of commercial humic acids were much smaller and their polydispersity in adsorbabilities was much broader. Besides, indicating strong affinities to AC surfaces, the Freundlich exponent 1/n values determined for all organic materials were relatively small and differed within relatively narrow ranges of 0.20–0.35 and 0.18–0.33 for TOC and UV260, respectively.
The weight-average molecular weight (Mw ) varied with organic matters (Mw = 2347–3667) by following a general trend that Aldrich humic acid > river water DOM > Wako humic acid > ground water DOM > treated wastewater DOM. The polydispersity defined by Mw /Mn for all organic materials changed in a narrow range of 1.09–1.26 by following the declining order: Wako humic acid > treated wastewater DOM > Aldrich humic acid > river water DOM ∼ = ground water DOM. A decreasing trend of the average adsorptive strength of total organic carbon (KM ) and an increasing trend of the corresponding adsorption polydispersity (σ ) with increasing humic content (SUVA) were observed. However, these parametric trends became vague when data related to commercial humic acids, which behaved very differently from seven aqueous organic materials, were excluded. The different behavior exhibited by commercial humic acids was attributed to their inherit chemical features (including the functional group, charge density, and aromaticity) and also to differences between the inorganic chemistry of humic acid and aqueous DOM solutions, which require further elucidation. Moreover, analysis of the results obtained in this study showed no trends between the molecular weights and the adsorbabilities of organic matters. This finding does not necessarily and maybe less sufficient to imply that the molecular size impacts are not significant. However, it may serve as supporting information to further stress the issue relates to the likely impacts of organic materials’ chemical characteristics and the chemistry of solutions.
Acknowledgments The authors sincerely thank Professor Alexander P. Mathews of Kansas State University for his valuable discussions and comments on this manuscript. This work was supported in part by Grants-in-Aid for Scientific Research 13750520 from the Japan Society for the Promotion of Science (JSPS).
References [1] M.C. Lee, V.L. Snoeyink, J. Crittenden, J. Am. Water Works Assoc. 73 (1981) 440. [2] J.E. Kilduff, T. Karanfil, Y. Chin, W.J. Weber Jr., Environ. Sci. Technol. 30 (1996) 1336. [3] J.C. Crittenden, P. Luft, D.W. Hand, Water Res. 19 (1985) 1537. [4] I.N. Najm, V.L. Snoeyink, Y. Richard, J. Am. Water Works Assoc. 83 (1991) 57. [5] C. Pelekani, V.L. Snoeyink, Water Res. 33 (1999) 1209. [6] J.E. Kilduff, A. Wigton, Environ. Sci. Technol. 33 (1999) 250. [7] K. Ebie, F.S. Li, Y. Azuma, A. Yuasa, T. Hagishita, Water Res. 35 (2001) 167. [8] S.J. Randtke, J. Am. Water Works Assoc. 80 (5) (1988) 40. [9] K. Ebie, F.S. Li, A. Yuasa, Environ. Eng. Res. Jpn. Soc. Civ. Eng. 32 (1995) 9.
F. Li et al. / Journal of Colloid and Interface Science 265 (2003) 265–275
[10] J. McCreary, V.L. Snoeyink, Water Res. 14 (1980) 151. [11] R.S. Summers, P.V. Roberts, J. Colloid Interface Sci. 122 (1988) 382. [12] T. Karanfil, J.E. Kilduff, M.A. Schlautman, W.J. Weber Jr., Environ. Sci. Technol. 30 (1996) 2187. [13] Y. Matsui, T. Kamei, N. Tambo, K. Taniguchi, J. Jpn. Soc. Water Environ. 16 (1993) 497. [14] A. Yuasa, F.S. Li, Y. Matsui, K. Ebie, Water Sci. Technol. 36 (12) (1997) 231. [15] A.M. El-Rehaili, W.J. Weber Jr., Water Res. 21 (1987) 573. [16] W.J. Weber Jr., T.C. Voice, A.M. Jodellah, J. Am. Water Works Assoc. 75 (1983) 612. [17] S.J. Randtke, C.P. Jepsen, J. Am. Water Works Assoc. 74 (1982) 84. [18] F.S. Li, A. Yuasa, K. Ebie, Y. Azuma, T. Hagishita, Y. Matsui, Water Res. 36 (2002) 4592. [19] R.S. Summers, P.V. Roberts, J. Colloid Interface Sci. 122 (1988) 367. [20] F.S. Li, A. Yuasa, K. Ebie, Y. Azuma, J. Colloid Interface Sci. 262 (2003) 331. [21] G.W. Harrington, F.A. DiGiano, J. Am. Water Works Assoc. 81 (1989) 93. [22] J. Fettig, H. Sontheimer, J. Environ. Eng. ASCE 113 (1987) 795. [23] H. Sontheimer, J. Crittenden, R.S. Summers, Activated Carbon for Water Treatment, 2nd ed., DVGW-Forschungsstelle, Karlsruhe, 1988 [in English]. [24] A. Yuasa, J. Jpn. Water Works Assoc. 62 (7) (1993) 38. [25] E.H. Smith, Water Res. 28 (1994) 1693.
275
[26] A. Yuasa, F.S. Li, Y. Matsui, K. Ebie, Environ. Eng. Res. Jpn. Soc. Civ. Eng. 33 (1996) 123. [27] B.R. Frick, H. Sontheimer, in: I.H. Suffet, M.J. McGuire (Eds.), Treatment of Water by Granular Activated Carbon, American Chemical Society, Washington, DC, 1983, p. 247. [28] J.C. Crittenden, K. Vaitheeswaran, D.W. Hand, E.W. Howe, E.M. Aieta, C.H. Tate, M.J. McGuire, M.K. Davis, Water Res. 27 (1993) 715. [29] C.L. Warta, S.P. Papadimas, G.A. Sorial, M.T. Suidan, T.F. Speth, Water Res. 29 (1995) 551. [30] F.A. DiGiano, G. Baldauf, B. Frick, H. Sontheimer, Chem. Eng. Sci. 33 (1978) 1667. [31] Y. Matsui, A. Yuasa, F.S. Li, J. Environ. Eng. ASCE 124 (1998) 1099. [32] F.S. Li, A. Yuasa, Y. Matsui, in: F. Meunier (Ed.), Fundamentals of Adsorption 6, Elsevier, Amsterdam, 1998, p. 1041. [33] R.L. Malcom, P. MacCarthy, Environ. Sci. Technol. 20 (1986) 904. [34] US-EPA, Office of Water (4607), EPA 815-R-99-012, Environmental Protection Agency, Washington, DC, 1999. [35] Y.P. Chin, G. Aiken, E. O’Loughlin, Environ. Sci. Technol. 28 (1994) 1853. [36] Y. Matsui, A. Yuasa, F.S. Li, Water Sci. Technol. 40 (9) (1999) 223. [37] F.S. Li, Ph.D. dissertation, Gifu University, Japan, 1998 [in Japanese]. [38] J. Buffle, P. Deladoey, W. Haerdi, Anal. Chim. Acta 101 (1978) 339. [39] A. Yuasa, F.S. Li, Y. Matsui, in: F. Meunier (Ed.), Fundamentals of Adsorption 6, Elsevier, Amsterdam, 1998, p. 407.