Adsorption kinetics of synthetic organic contaminants onto superfine powdered activated carbon

Erin Partlan, Yiran Ren, Onur G. Apul, David A. Ladner, Tanju Karanfil

Citation: Erin Partlan, Yiran Ren, Onur G. Apul, David A. Ladner, Tanju Karanfil; Adsorption kinetics of synthetic organic contaminants onto superfine powdered activated carbon. https://doi.org/10.1016/j.chemosphere.2020.126628

Abstract

Superfine powdered activated carbon (S-PAC) is an adsorbent material with the promise of properties that allow for rapid adsorption of small molecule contaminants. To explore the potential for rapid adsorption among varying activated carbon types, seven commercially available activated carbons were obtained and pulverized to produce S-PAC particles less than 1 μm in diameter. The carbons were chosen to include several types of common carbons produced from coal precursors as well as a wood-based carbon and a coconut shell-based carbon. In this study, the S-PACs and their parent PACs were tested for the adsorption of three aromatic compounds—2-phenylphenol, biphenyl, and phenanthrene—with and without the presence of natural organic matter (NOM). Adsorption rates were increased for adsorption onto S-PAC as compared to PAC in all trials without NOM and in most trials with NOM. Faster adsorption onto S-PAC was found to be a result of a smaller particle size, lower surface oxygen content, larger pore diameters, and neutral pHPZC. Adsorption of a planar compound, phenanthrene, increased the most between PAC and S-PAC, while adsorption of 2-phenylphenol, a nonplanar compound, was impacted the least. Phenanthrene additionally was minimally impacted by the presence of NOM while 2-phenylphenol adsorption declined severely in the presence of NOM.

1. Introduction

Activated carbon adsorption is an established physicochemical process used to remove organic micropollutants, such as synthetic organic chemicals (SOCs) that contribute to anthropogenic water contamination. Powdered activated carbon (PAC), which is defined as activated carbon particles where the predominating fraction passes an 80-mesh sieve of 0.30 mm openings (ASTM Standard D5158-98, 2013) is preferred for its small particle size enabling faster adsorption kinetics due to shorter diffusion pathway. PAC is typically dosed as powder in suspension into completely mixed contactors. Faster adsorption, such as via a reduced particle size, is not only beneficial for contaminant removal under a shorter contact time, it is also useful to subvert competitive adsorption by larger non-target molecules, such as natural organic matter (NOM) during the removal of SOCs, and for better utilization of adsorption capacity (Fontecha-Cámara et al., 2008; Newcombe et al., 2002; Zhang et al., 2011).

It has been found that further reduction of the particle size, especially in the submicron range, can provide additional adsorption rate benefits (Ando et al., 2010; Ellerie et al., 2013; Matsui et al., 2008; Pelekani and Snoeyink, 2000). Superfine powdered activated carbon (S-PAC) is a potential substitute for PAC that can be applied to drinking water treatment systems; whereas PAC is removed in the sedimentation basin, microfiltration is employed to ensure that S-PAC is retained in the treatment system. Furthermore, S-PAC has been documented to have faster adsorption kinetics than its parent material PAC for odor and taste compounds, natural organic matters (NOM), disinfection byproduct (DBP) precursors, synthetic organic contaminants (SOCs) such as atrazine, pharmaceuticals and personal care products (PPCPs; such as carbamazepine and sulfamethoxazole) and viruses (Amaral et al., 2016; Bonvin et al., 2016; Li, 2014; Matsui et al., 2013; Matsushita et al., 2005). However, the increase in adsorption kinetics is not solely a function of particle size and depends on specific adsorbent-adsorbate interactions.

Activated carbon kinetic performance is a function of physical attributes of adsorbents, including adsorbent surface area and pore size distribution, and chemical attributes, including the surface oxygen content, pH at the point of zero charge, and surface functional groups (Abuzaid and Nakhla, 1996; Ebie et al., 2001; Wu and Pendleton, 2001). In addition, adsorbate properties (e.g. solubility, molecular size, planarity, polarizability, and aromaticity) directly relate to adsorption performance (Apul et al., 2015; Bakkaloglu, 2014; Li et al., 2002; S Zhang et al., 2010). Pulverization of PAC into S-PAC is a high intensity process, and besides a size reduction, it results in changes to pore size distribution as internal pores are opened and oxidation of newly revealed internal surfaces (Partlan et al., 2016). For aromatic compounds, oxidation of activated carbon surfaces leads to decreased adsorption, a result attributed to the role of π-π interactions between adsorbate and adsorbent (Coughlin and Ezra, 1968; Dabrowski et al., 2005; Karanfil; Kilduff, 1999).

This study examines S-PACs produced from seven different parent carbons of four different material types for adsorption of three aromatic compounds. Aromatic compounds remain a topical pollutant in drinking water systems due to their toxicity and difficulty of removal at trace levels (Stackelberg et al., 2007). While multiple S-PAC types and multiple adsorbates have been studied, synthetic contaminants have not been studied in relation to S-PAC type (Bakkaloglu, 2014; Matsui et al., 2015). This study aims to understand mechanistic differences for variability among S-PAC performance for specific adsorbate-adsorbent combinations.

2. Materials and methods

2.1. Selection and pulverization of parent PACs to produce S-PACs

Seven commercial activated carbons were used as parent material: three of bituminous coal origin (Filtrasorb 400, Calgon; 20B, Norit; Watercarb-800, Standard Purification), two of lignite coal origin (Hydrodarco B, Norit; Hydrodarco 3000, Norit), one of coconut shell origin (Aquacarb 1230C, Siemens), and one of wood origin (Aqua Nuchar, Mead Westvaco). For convenience, the activated carbons will be referred to, in their respective order, as: F400, 20B, WC800, AC, HDB, HD3000, and Nuchar. HDB, F400, and AC were received as granular activated carbon (GAC). 20B, WC800, HDB, and Nuchar were received as PAC. Carbons received as GAC were processed into PAC form by grinding in a coffee grinder and dry sifting through a #100 mesh sieve (150 μm openings).

S-PAC was produced from these parent carbons via pulverization in a bead mill (MiniCer, Netzsch Premier Technologies, Exton, PA, USA) containing 0.3–0.5 mm yttrium-stabilized zirconium oxide ceramic beads. Milling was performed in distilled and deionized water (DDI) without dispersant and produced S-PACs were stored in slurry form. The S-PAC products analyzed in this study were produced with 4–6 h of recirculated milling. Additional details of superfine pulverization are provided in Partlan et al. (2016).

2.2. Adsorbate molecules selected as representative SOC

Three aromatic synthetic organic contaminants were selected as adsorbates representing common pollutants or pollutant precursors: phenanthrene (PNT) (99.5%, Fluka), biphenyl (BP) (99%, Sigma Aldrich), and 2-phenylphenol (2PP) (99+%, Sigma Aldrich). The selected adsorbates differ in molecular size, hydrophobicity, number of double bonds, and polarizability. Their abbreviations, molecular configurations and some physicochemical properties are provided in Table 1.

a = Molecular weight.

b = Octanol-water partitioning constant.

c = Water solubility at 25 °C.

d = Obtained from ACDLABS11.0 (ChemSketch and ACD/3D Viewer).

e = Obtained from ACD/ADME Suite 5.0.

2.3. Physicochemical characterization of adsorbents

The PACs and S-PACs were characterized for physical and chemical attributes. Nitrogen adsorption at 77 K was performed with a physisorption analyzer (Micromeritics ASAP 2020) and the nitrogen isotherms were used to determine the specific surface area via the Brunauer–Emmett–Teller (BET) equation. Pore volume (PV) was determined via nitrogen absorbed near the saturation point (P/P₀ = 0.99), and pore size and surface area distributions were computed via Density Functional Theory (DFT). Pore channels were categorized as micropores (<2 nm diameter), mesopores (2–50 nm diameter), and macropores (>50 nm diameter). A Flash Elemental Analyzer 1112 series (Thermo Electron Corporation) was used to analyze the oxygen content of the adsorbents. The details of the characterization procedures and results can be found in previous works (Partlan et al., 2016).

2.4. Evaluating adsorption kinetics

Adsorption kinetics experiments were performed in 255 mL amber bottles with teflon-lined screw caps using 4 mg/L of adsorbent. All experiments were performed at room temperature (20 ± 2 °C) without the addition of a buffer. Bottles were fully filled with background solution to minimize headspace. Trials were performed in distilled and deionized water (DDI) with and without background natural organic matter (NOM) (Suwanee River RO isolation, Humic Substances Society). The NOM solution, which was adjusted to pH 7 by addition of sodium bicarbonate, had a dissolved organic carbon (DOC) concentration of 4.2 mg DOC/L (±5%), as measured by a Shimadzu TOC-VCHS analyzer, and a SUVA₂₅₄ value of around 4.0 L/mg-m.

Adsorbate stock solutions were prepared in methanol and used to spike the bottles to produce an adsorbate concentration of 1 mg/L (a higher concentration than would typically be present in the environment, but this study focused on a comparison among adsorbents and therefore required sufficient adsorbate to measure effects.) The volume percentage of the methanol spiked per bottle was kept below 0.1% (v/v) in each case to minimize a co-solvent effect. After spiking, the bottles were placed into a rotary tumbler at a speed of 0.6 g (5.88 m/s²). Select bottles were sacrificed at intervals over 6 h to draw an aliquot. The aliquots were centrifuged at 3500 rpm for 30 min to separate adsorbents from solution and halt adsorption. Total adsorption time is reported as the time of sampling plus 30 min for centrifuging. Bottles prepared without any adsorbents served as blanks to measure adsorption to the bottle and cap, which were found to be negligible.

The supernatant of each aliquot was analyzed for adsorbate concentration by high-performance liquid chromatography (HPLC) with a UV detector using 4.6∗150 mm ZORBAX Extend-C18 Columns (Agilent). Solution pH was found to be stable around 6.6. More details about the analytical methods can be found in our previous study (Zhang et al., 2010).

3. Results and discussion

3.1. Comparing adsorption kinetics between PAC and S-PAC in DDI

Analysis of supernatant from centrifuged aliquots was used to determine the amount of adsorbate removed by PAC and S-PAC under each trial condition. Adsorbed quantities ranged from 36–174 mg/g for 2PP, 63–214 mg/g for BP, and 46–256 mg/g for PNT across different carbons, S-PAC vs. PAC, and different contact times. Adsorption capacities were proportional to the log KOW values—2PP = 2.94, BP = 3.98, PNT = 4.68 (Table 1)—indicating the notable role of surface interactions in adsorption capacity. 2PP adsorbed 30.2% less (σ = 8.1%) than BP on PACs and 24.4% less (σ = 7.0%) on S-PACs. PNT adsorbed 27.0% more (σ = 24.4%) than BP on PACs and 56.2% more (σ = 16.9%) on S-PACs.

An indicator of comparative adsorption kinetics and the impact of wet milling on adsorption was defined by taking the ratio of S-PAC adsorption capacities to corresponding PAC adsorption capacities at the same time point. The S-PAC:PAC adsorption ratio as well as individual adsorption kinetics for each SOC is presented in Fig. 1. At 1 h of contact time, all adsorbate-adsorbent pairings resulted in S-PAC to PAC capacity ratios of greater than one at the 1-h mark, indicating more adsorption onto S-PAC than PAC when time is limited. However, the adsorption ratios considerably decline with increased contact time in general. This indicates that the rate of adsorption onto S-PACs slowed significantly after the 1-h point, while adsorption onto PACs continued at higher rates over the course of the 6-h experiment. Since the adsorption of SOCs onto PAC usually approached the adsorption onto S-PAC at the end of the experiment, the effect of pulverization to improve S-PAC performance is not a result of increased adsorption capacity. The observations are congruent with the theory of a shortened diffusion pathway that permits adsorbates to reach their adsorption capacity onto S-PAC rapidly while taking longer to adsorb to full capacity on PAC. In a study of S-PAC for adsorption of 10 representative pharmaceutical micropollutants, Bonvin et al. (2016) found that S-PAC reached adsorption capacity in about 10 min, while parent PACs reached adsorption capacity only after 12 or more hours, aligning well with our observations.

Fig. 1. Comparative adsorption kinetics at 1, 2, 4, and 6 h of contact time in DDI for all carbons and adsorbates. Points: Adsorption onto PAC and S-PAC at each time point. Lines connecting the points are to guide the eye. Bars: Ratio of S-PAC adsorption to PAC adsorption at each time point.

Of all the carbons, the largest increase in adsorption rate occurred with AC pulverization, where AC S-PAC adsorbed ∼5 times more PNT, four times more 2PP and three times more BP than the parent PAC at the 1-h mark. Furthermore, AC S-PAC stands apart from other adsorbents for outperforming its parent PAC even at the 6-h mark over all three adsorbates. All other adsorbents showed increased performance on S-PAC relative to PAC at the 1-h mark, but similar performance at the 6-h mark, a result of many S-PACs reaching their adsorption capacity for the adsorbent within 6 h. The pairing where pulverization was least helpful is 2PP and WC800 where adsorption on both PAC and S-PAC is nearly identical S-PAC and PAC at all time points. Furthermore, some adsorbate-adsorbent pairs showed higher adsorption on PAC than S-PAC at later time points—as seen by an S-PAC to PAC ratio less than one—despite increased adsorption onto S-PAC at earlier time points. Where PAC capacity exceeds S-PAC capacity, such as in the adsorption of BP onto WC800, S-PAC is beneficial under short contact times and PAC is beneficial under longer contact times.

3.2. Comparison of empirical adsorption rates

The adsorption kinetics can be modeled using a pseudo-second order rate model comparing the amount adsorbed with the equilibrium capacity (Equation (1)) (Ho and McKay, 1999; Plazinski et al., 2013). Using time t and the mass of adsorbate per mass carbon at time t (qt), an adsorption capacity qe and rate coefficient k2 are found. Comparison of an empirical rate coefficient allows for evaluation of adsorbate-adsorbent pairs based on adsorption kinetics alone—in essence, disregarding adsorption capacity.

$\frac{t}{q_t} = \frac{1}{k_2 q^2_e} + \frac{t}{q_e}$ —- eq(1)

The model of Ho and McKay is most accurate when data encompasses both steady state and transient behavior. When transient data is minimal or absent, rate coefficients skew higher, and when steady state data is absent, rate coefficients skew lower. Therefore, pairs that reached steady state before 1 h or did not approach steady state after 6 h have insufficient data for determination of a rate coefficient. The rate coefficient k2 is plotted for PAC and S-PAC over all adsorbents and adsorbates in Fig. 2; pairs with insufficient data are marked in red.

Fig. 2. Pseudo-second order rate coefficient for PAC and S-PAC over all adsorbates and adsorbents. Red X markers represent datasets where data is insufficient, and are either the fastest or slowest adsorption pairs. Lines connecting the data points are to guide the eye. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

As expected, all of the S-PAC rate coefficients are faster than the PAC rate coefficients, except the adsorption of 2PP onto WC800. Conversion of WC800 from PAC to S-PAC appears to have had no impact on the adsorption rate of 2PP. S-PAC rates were also more scattered than PAC rates—S-PAC rate coefficients have a standard deviation of 73% the average rate, compared to 60% for PAC rates. F400 S-PAC and WC800 S-PAC both had high scatter among adsorption rates for the three adsorbates; Nuchar PAC had the highest scatter among PACs and less than the highest S-PACs. Thus, for use in targeted applications, S-PAC should be tested for suitability.

Among PAC, AC exhibited the slowest adsorption rates overall. The slow AC PAC rates account for the high increased performance of AC S-PAC, despite performance similar to other S-PACs. The rate data shows that PNT and 2PP were generally adsorbed faster by the S-PACs than BP, except for WC800 S-PAC, which adsorbed BP the fastest and had no increase in 2PP adsorption rate. Overall, 20B S-PAC exhibited the fastest adsorption rates over all three adsorbates, and thus the best performance, while AC S-PAC had the lowest average rate, despite the largest increase in performance from PAC.

3.3. Effects of adsorbate and adsorbent characteristics on adsorption performance

3.3.1. Normalizing performance by specific surface area

While measuring absolute performance is useful for evaluating the utility of a material, normalizing performance to surface area is more useful for comparing materials on their fundamental adsorption behavior. Pulverization results in shifts in specific surface area between PAC and S-PAC, perhaps through revealing internal pores and increasing BET specific surface area, or crushing pores located closer to the outer shell and detracting from BET surface area. The combination of these changes means that pulverization can result in an increase, decrease, or no net change to the BET surface area of an S-PAC as compared to PAC. Three carbons decreased in specific surface area (20B, F400, Nuchar), while the remaining increased in specific surface area (Table 2). By normalizing to surface area, effects of net changes in BET surface area are removed.

Nuchar lost the most surface area (36%) as a result of pulverization. In terms of absolute performance, Nuchar S-PAC only performed slightly better than Nuchar PAC. However, when considering the significant loss of surface area, it is clear that Nuchar S-PAC actually adsorbed much faster than Nuchar PAC (Fig. 3). The comparable performances without normalization indicate that the lost surface areas in Nuchar are not the primary SOC sorption sites. AC gained the most surface area (13%) as a result of pulverization, but even with surface area normalization, AC S-PAC still greatly outperforms AC PAC, indicating that AC S-PAC performance benefitted from both a reduced particle size and increased surface area.

Fig. 3. Specific adsorption for each adsorbate-adsorbent pair shows adsorption in DDI normalized to surface area.

3.3.2. Effects of pore size distribution

As structural changes from pulverization can alter the specific surface area between PAC and S-PAC, the distribution of pore sizes can also shift. Adsorbent pores are categorized by the diameter of pore channels into macroporous (>50 nm), mesoporous (2–50 nm), and microporous (<2 nm) channels, and the volumes summed into pore fractions. In general, carbons tended to gain in the mesoporous fraction after milling (Table 2). Notably, the coconut-shell carbon AC, which is known for being a microporous material, did not have mesopores as PAC and developed a mesoporous fraction after pulverization. Pore fractions as a percentage of total pore volume had little correlation with adsorption performance overall, nor with PAC adsorption performance alone; however, the trade-off between micropores and mesopores has an effect on the adsorption of specific compounds on S-PAC (Fig. 4). For the adsorption of 2PP and BP, an increase in micropores and a decrease in mesopores resulted in higher rates of adsorption. For PNT, the opposite is true—a smaller microporous fraction and a larger mesoporous fraction achieved faster adsorption. Thus, effects of pore size distribution—and adjacently, effects of carbon precursor material—on adsorption are present but not straightforward.

Fig. 4. The distribution of internal pores in mesopore and micropore fractions is a stronger factor for surface area normalized adsorption after 1 h on S-PAC (b,d,f) than for PAC (a,c,e).

3.3.3. Effects of adsorbate configuration

The behavior with regards to micropores and mesopores alludes to an effect of planarity, since PNT is planar while BP and 2PP are not. Planar molecules can be excluded from pores even if their overall size is small due to random molecule orientation, though utilization of pores can still be high as its planar orientation allows for tight packing into pores via molecular stacking (Zhang et al., 2010). PNT, as a rigid and planar molecule with a large longest dimension, has a lower probability of entering micropores and is more likely to utilize mesopores for accumulation.

3.3.4. Effects of external surface area

A key feature of a reduced particle diameter, in addition to a shortened internal diffusion pathway, is the increase in external surface area per mass of carbon. Increased external surface area allows adsorbate molecules to access internal pores in parallel and additionally, enhances probabilistic outcomes of adsorbate orientation. For all carbons tested, improvement in adsorption between PAC and S-PAC was highest for rigid and planar PNT. In contrast, BP and 2PP are more flexible, non-planar molecules that are not particularly affected by stochastic molecule orientation and are governed instead by conventional size exclusion. Furthermore, external surface area is an important consideration for factors that relate to the surface of the particle, especially in governing the transfer of molecules to the particle surface, as will be seen in the discussion of oxygen content below.

3.3.5. Effects of oxygen content

Pulverization also resulted in higher fractions of oxygen per mass of adsorbent, and is theorized to be a result of oxidation on newly exposed external surfaces (Partlan et al., 2016). Increased oxygen at the external surface creates a higher negative charge that results in an increase in bound water molecules around a particle, thus effecting a barrier to adsorption for hydrophobic compounds (Li et al., 2002). Thus, S-PAC performance appears to be in contrast to expected performance reduction as a result of oxidation. However, pulverization necessarily produces more surface area per mass by nature of a decreased particle diameter. Assuming approximately spherical particles and normalizing the oxygen content to the available external surface area per mass, S-PAC actually has less oxygen on its surfaces than PAC, and therefore the S-PACs are more suitable to the adsorption of PNT and BP that have low solubilities in contrast to the higher solubility of 2PP (Fig. 5). Changes in oxygen content support observed results of the large increases in adsorption capacity for PNT especially on all pulverized adsorbents.

Fig. 5. Adsorption capacity at 1 h is increased with less oxygen per area of external particle surfaces per gram of adsorbent.

The degree of oxidation on activated carbon surfaces is a function of multiple parameters, one of them being time as environmental oxygen migrates towards newly exposed surfaces. Thus, it is preferable that carbon be milled shortly before use, and milled and stored in deoxygenated water when possible. S-PAC may also be deoxygenated after pulverization via heat treatment in an oxygen-free environment, though this may prove more costly due to the need to fully dry and then re-suspend the S-PAC slurry.

3.3.6. Effects of pH of the point-of-zero-charge

A significant factor in adsorbate-adsorbent interactions is electrochemical behavior. On the adsorbent side, stronger electrostatic forces within pores can increase the strength of adsorption and result in higher adsorption capacities, but these same electrostatic forces result in drag during intrapore diffusion and negatively impact adsorption kinetics. pH is used to evaluate surface charge, whether more strongly negative (acidic) or positive (basic). Previously, oxidation of the external surface was determined by a difference between the pH of the external surface (measured as the pH of the isoelectric point, pH_IEP) and the average pH of all surfaces (measured as the pH of the point of zero charge, pH_PZC) (Partlan et al., 2016). The oxidation resulted in all S-PACs having a pH_PZC more acidic than their parent (Table 2).

Looking at the basic carbons (20B, WC800, AC, HDB, F400), adsorption capacity correlates with pH_PZC, even when both PAC and S-PAC are included; the trend is stronger at 1 h than at 6 h as a result of several S-PACs having reached adsorption capacity (Fig. 6). The more neutral PACs and S-PACs tend to adsorb faster than carbons with a more extreme pH, indicating the effect of surface charge on retarding adsorbate movement. Additionally, the correlation was much stronger for PNT and weakest for 2PP, which is explained by the higher adsorption observed as a function of lower oxygen density for the less soluble PNT (see Section 3.3.5).

Fig. 6. Adsorption uptake at 1 h (blue) and 6 h (orange) for basic carbons (20B, WC800, AC, HDB, F400). Adsorption is highest for carbons with more neutral pHpzc. Goodness of fit: 2PP 1 h, R² = 0.59; 6 h, R² = 0.29. BP 1 h, R² = 0.66; 6 h, R² = 0.24. PNT 1 h, R² = 0.91; 6 h, R² = 0.71. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

3.3.7. Effects of adsorbate polarizability

On the adsorbate side, molecular charge contributes to electrochemical drag in pore channels; an acidic and negatively charged surface is typically correlated with a lower adsorption capacity, and is theorized to be a result of interactions with charged adsorbates (Dabrowski et al., 2005). For uncharged molecules, the ability to form dipole moments, known as polarizability, can also result in electrochemical drag, though to a lesser extent. In membrane filtration, polarizability has been shown to result in decreased passage and improved rejection—the opposite of desired intrapore passage in carbon adsorption (Jeffery-Black and Duranceau, 2016).

Aromatics, generally with a high pK_a, primarily exist in an uncharged state so polarizability is dominant over charge effects. Polarizability is affected by molecule size, where larger molecules exhibit more polarizability. Previous research on S-PAC examining the adsorption of geosmin, a highly polarizable compound with a polarizability of 22.2 Å³, found pulverized carbon to have improved adsorption kinetics (Matsui et al., 2009). The tested aromatic adsorbates have lower polarizability than geosmin with values of approximately 1–1.4 Å³, but higher than other solutes, such as a number of aliphatic compounds that range from 0.3 to 0.8 Å³. Such aliphatics have been shown to exhibit increased adsorption capacity with increased polarizability (Apul et al., 2015). As polarizability is independent of pH, a relationship between polarizability and the ease of adsorption can therefore be attributed to a change in surface charges on the carbons.

3.3.8. Ranking importance of characteristics for adsorption in DDI

To determine the parameters that have the strongest effect on adsorption performance, the discussed parameters were evaluated using the method “Relative Importance of Regressors in Linear Models” (implemented in R as the package relaimpo) for their percent contribution to explaining variance among the adsorption dataset according to a linear correlation (Grömping, 2006). At the outset of the study, the effect of particle size was the most obvious, having been demonstrated by a number of researchers, and is the clear differentiator between PAC and S-PAC performance. However, particle size is not a significant determinant among S-PAC alone when carbon type and adsorbate are varied. Considering additional parameters, planarity becomes the biggest contributor to prediction of adsorption performance, which implies that behavior for planar molecules is markedly different from non-planar molecules (Table 3). The next largest parameter contribution is particle size, as expected, followed by the normalized surface oxygen content. Like particle size, the normalized surface oxygen content captures the boundary between PAC and S-PAC well, but also captures pH information as more oxygen content is associated with a more acidic pH. The smallest contributors are the percentage of pores as micropores and mesopores, and adsorbate polarizability.

3.4. Effects of background NOM on adsorption kinetics

In trials with background NOM, the relative adsorption between S-PAC and PAC as a function of contact time decreased with increasing time, similar to the results from trials in DDI. However, unlike the increased performance in DDI trials, in many cases with NOM, adsorption onto S-PAC was similar to and in some cases even less than adsorption onto PAC at the same time point (Fig. 7). In these cases, particularly so for the adsorption of 2PP but less so for PNT and with mixed results for BP, S-PAC still performed as good or better than PAC at the 1-h mark but is eclipsed with increasing contact time. The initial improvement indicates that a reduced particle size creates a slight kinetic advantage initially, allowing small adsorbates quick access to internal pores, but the impact is dulled by S-PAC quickly reaching its adsorption capacity, especially in solution with NOM that can block internal pores by adsorbing to external and macropore adsorption sites (Newcombe et al., 2002).

Fig. 7. Comparative adsorption kinetics at 1 h, 2 h, 4 h, and 6 h of contact time in NOM for all carbons and adsorbates. Points: Adsorption onto PAC and S-PAC at each time point. Bars: Ratio of S-PAC adsorption to PAC adsorption at each time point.

3.4.1. Comparing adsorption performance with and without NOM

The adsorption of PNT was the least unchanged of the three adsorbates in background NOM as compared to DDI, with PACs adsorbing at almost exactly the same rate and S-PACs only adsorbing slightly slower. PNT adsorbed fast and in high amounts while encountering almost no competitive effects from NOM, while 2PP was the most affected by NOM presence (Fig. 8). Similar results are reported by Zhang et al. (2011) in an examination of adsorption of 2PP and PNT in the presence of NOM onto carbon nanotubes, activated carbon fiber, and GAC.

Fig. 8. Ratio of adsorbed amount in DDI to adsorbed amount in NOM at 1 h for PAC (hashed) and S-PAC (solid).

2PP, a non-planar and hydrophilic molecule, adsorbed much more quickly onto S-PACs in DDI than in NOM, and somewhat more quickly onto PACs in DDI than in NOM. In fact, NOM has the effect of reducing 2PP adsorption such that the benefit of adsorbing onto a smaller particle size in DDI is almost completely lost; in a background NOM solution, PAC performs better than S-PAC except in the shortest time frames. In contrast, the adsorption of PNT, a planar and hydrophobic molecule, is minimally impacted by the presence of NOM, especially with regards to adsorption onto PAC. In five out of seven carbons, PNT adsorbed to PAC at the exact same rate regardless of background solution; S-PAC adsorption of PNT was diminished slightly in the presence of NOM. By mass, PNT adsorbed more across PAC and S-PAC trials than BP or 2PP. Due to nearly identical molecular weights, the molar concentration of PNT and 2PP in these trials is almost nearly identical.

Zhang et al. (2011) found that 2PP adsorption was impacted whereas PNT adsorption was not, indicating that NOM shares adsorption sites in common with 2PP more than PNT. Perhaps due to the hydrophobic and planar nature of PNT, its adsorption occurs preferentially in sites that are not impacted by initial NOM adsorption—that is, adsorption within the first hour—either from site adsorption competition or from pore blockage.

Adsorption of Suwanee River NOM is useful to evaluate competition between small and large molecules at the bench scale, especially in conjunction with higher concentrations of adsorbate. However, in a realistic setting, water additionally has monovalent and divalent ionic content that can affect adsorption behavior. Monovalent ions reduce the ability for S-PAC to remain colloidally dispersed and induces aggregation; aggregated S-PAC results in reduced adsorption kinetics and capacity due to reduced access to internal pores (Partlan, 2017). Divalent ions have the effect of bridging compounds such as NOM, and a larger effective size prevents adsorption to S-PAC macropores, though Amaral et al. also found that atrazine adsorption was reduced in the presence of calcium (Amaral et al., 2016). These two ionic strength effects—in addition to trace contaminants often present in very low concentrations of ng/L to μg/L levels—point to the need for further testing in natural waters to bridge the gap between the laboratory and the field.

3.4.2. Ranking importance of characteristics for adsorption in DDI

Once again looking at the relative contribution of parameters to explaining variance among adsorption measurements, planarity stands out as the factor explaining the most variance (Table 4). The finding is unsurprising as the discussion above indicates differences in adsorption patterns between planar PNT and non-planar 2PP and BP. In fact, such a large contribution from a binary categorical variable suggests that separate models are warranted to describe adsorption of planar vs. non-planar molecules. Another difference in factors for adsorption in NOM vs. DDI is the equal contribution of particle size and normalized oxygen content. Since the two variables are related via the normalization by external surface area, the contribution most likely stems from a PAC vs. S-PAC dichotomy. Lastly, polarizability, pH in units from neutral, and pore size fractions have a minor contribution. Thus, NOM competition is solved best by using S-PAC in place of PAC and competition is the least concern when targeting planar molecules.

4. Conclusions

S-PAC exhibited faster adsorption of 2-phenylphenol, biphenyl, and phenanthrene than PAC overwhelmingly in DDI and primarily even in background NOM. The advantage of S-PAC over PAC in kinetics diminished with increased time as a result of decreased S-PAC adsorption uptake over time indicating that S-PAC reached adsorption capacity within the 6-h measurement timeframe.
Particle size reduction and the related increase in external surface area, which allows for increased access to adsorbent pores plus a shortened diffusion pathway, are the primary differentiators between PAC and S-PAC performance.
The primary differentiator to explain adsorption rate variance was the planarity of the adsorbate molecule. The difficulty of planar molecules to enter small pore passages plus the ability of planar molecules to stack for high adsorption density likely led to the improved performance of phenanthrene on carbons with less micropores and in competitive NOM solution.
The lower density of oxygen per external surface area on S-PAC compared to PAC, and the related shift away from basic pH, explains why increased oxygen content did not hinder S-PAC adsorption. Further study should be done to determine if S-PAC oxidized to the same level as PAC suffers from worsened adsorption performance to accurately evaluate the importance of maintaining S-PAC in a low oxygen environment.

6. CRediT authorship contribution statement

Erin Partlan: Conceptualization, Methodology, Visualization, Writing – original draft.

Yiran Ren: Conceptualization, Methodology, Investigation, Writing – original draft.

Onur G. Apul: Conceptualization, Methodology, Writing – original draft.

David A. Ladner: Supervision, Writing – review & editing.

Tanju Karanfil: Supervision, Writing – review & editing.

7. Acknowledgements

This work was supported by a research grant from the National Science Foundation (Award No.1236070). However, the manuscript has not been subjected to the peer and policy review of the agency and therefore does not necessarily reflect it views.

8. References

Abuzaid, N.S., Nakhla, G.F., 1996. Effect of solution pH on the kinetics of phenolics uptake on granular activated carbon. J. Hazard. Mater. 49, 217e230.
Amaral, P., Partlan, E., Li, M., Lapolli, F., Mefford, O.T.T., Karanfil, T., Ladner, D.A.D.A., 2016 Superfine powdered activated carbon (S-PAC) coatings on microfiltration membranes: effects of milling time on contaminant removal and flux. Water Res. 100, 429e438.
Ando, N., Matsui, Y., Kurotobi, R., Nakano, Y., Matsushita, T., Ohno, K., 2010. Comparison of natural organic matter adsorption capacities of super-powdered activated carbon and powdered activated carbon. Water Res. 44 (14),4127e4136.
Apul, O.G., Zhou, Y., Karanfil, T., 2015. Mechanisms and modeling of halogenated aliphatic contaminant adsorption by carbon nanotubes. J. Hazard. Mater. 295, 138e144.
ASTM Standard D5158-98, 2013. Standard Test Method for Determination of Particle Size of Powdered Activated Carbon by Air Jet Sieving. ASTM International, West Conshohocken, PA.
Bakkaloglu, S., 2014. Adsorption of Synthetic Organic Chemicals: A Comparison of Superfine Powdered Activated Carbon with Powdered Activated Carbon. Clemson University. https://doi.org/10.1007/s13398-014-0173-7.2.
Bonvin, F., Jost, L., Randin, L., Bonvin, E., Kohn, T., 2016. Super-fine powdered activated carbon (SPAC) for efficient removal of micropollutants from wastewater treatment plant effluent. Water Res. 90, 90e99.
Coughlin, R.W., Ezra, F.S., 1968. Role of surface acidity in the adsorption of organic pollutants on the surface of carbon. Environ. Sci. Technol. 2 (4), 291e297.
Dabrowski, A., Podkoscielny, P., Hubicki, Z., Barczak, M., 2005. Adsorption of phenolic compounds by activated carbonda critical review. Chemosphere 58, 1049e1070.
Ebie, K., Li, F., Azuma, Y., Yuasa, a, Hagishita, T., 2001. Pore distribution effect of activated carbon in adsorbing organic micropollutants from natural water. Water Res. 35 (1), 167e179. Retrieved from. http://www.ncbi.nlm.nih.gov/pubmed/11257871.
Ellerie, J.R., Apul, O.G., Karanfil, T., Ladner, D.A., 2013. Comparing graphene, carbon nanotubes, and superfine powdered activated carbon as adsorptive coating materials for microfiltration membranes. J. Hazard. Mater. 261C, 91e98.
Fontecha-Camara, M. a, L opez-Ram on, M.V., Pastrana-Martínez, L.M., Moreno- Castilla, C., 2008. Kinetics of diuron and amitrole adsorption from aqueous solution on activated carbons. J. Hazard. Mater. 156 (1e3), 472e477.
Gromping, U., 2006. Relative importance for linear regression in R: the package €
relaimpo. J. Stat. Software 17 (1).
Ho, Y.S., McKay, G., 1999. Pseudo-second order model for sorption processes. Process Biochem. 34 (5), 451e465.
Jeffery-Black, S., Duranceau, S.J., 2016. The influence of solute polarizability and molecular volume on the rejection of trace organics in loose nanofiltration membrane processes. Desalination Water Treat. 57, 29059e29069.
Karanfil, T., Kilduff, J.E., 1999. Role of granular activated carbon surface chemistry on the adsorption of organic compounds. 1. Priority pollutants. Environ. Sci.Technol. 33 (18), 3217e3224.
Li, L., Quinlivan, P.A., Knappe, D.R.U., 2002. Effects of activated carbon surface chemistry and pore structure on the adsorption of organic contaminants from aqueous solution. Carbon 40, 2085e2100.
Li, M., 2014. Effects of Natural Organic Matter on Contaminant Removal by Superfine Powdered Activated Carbon Coupled with Microfiltration Membranes. Clemson University.
Matsui, Y., Ando, N., Sasaki, H., Matsushita, T., Ohno, K., 2009. Branched pore kinetic model analysis of geosmin adsorption on super-powdered activated carbon. Water Res. 43 (12), 3095e3103.
Matsui, Y., Murai, K., Sasaki, H., Ohno, K., Matsushita, T., 2008. Submicron-sized activated carbon particles for the rapid removal of chlorinous and earthy-musty compounds. J. Water Supply Res. Technol. – Aqua 57 (8), 577.
Matsui, Y., Nakao, S., Sakamoto, A., Taniguchi, T., Pan, L., Matsushita, T., Shirasaki, N., 2016 Adsorption capacities of activated carbons for geosmin and 2-methylisoborneol vary with activated carbon particle size: effects of adsorbent and adsorbate characteristics. Water Res. 85, 95e102.
Matsui, Y., Nakao, S., Taniguchi, T., Matsushita, T., 2013. Geosmin and 2-methylisoborneol removal using superfine powdered activated carbon: shell adsorption and branched-pore kinetic model analysis and optimal particle size. Water Res. 47 (8), 2873e2880.
Matsushita, T., Matsui, Y., Shirasaki, N., K, Y., 2005. Effect of membrane pore size , coagulation time , and coagulant dose on virus removal by a coagulationceramic microfiltration hybrid system. Desalination 178, 21e26. November2004.
Newcombe, G., Morrison, J., Hepplewhite, C., Knappe, D.R.U., 2002. Simultaneous adsorption of MIB and NOM onto activated carbon II. Competitive Effects 40, 2147e2156.
Partlan, E., 2017. Superfine Powdered Activated Carbon (S-PAC) Coupled with Microfiltration for the Removal of Trace Organics in Drinking Water Treatment. Clemson University.
Partlan, Erin, Davis, K., Ren, Y., Apul, O.G.O.G., Mefford, O.T., Karanfil, T., Ladner, D.A.D.A., 2016. Effect of bead milling on chemical and physical characteristics of activated carbons pulverized to superfine sizes. Water Res. 89,161e170.
Pelekani, C., Snoeyink, V.L., 2000. Competitive adsorption between atrazine and methylene blue on activated carbon: the importance of pore size distribution. Carbon 38 (10), 1423e1436.
Plazinski, W., Dziuba, J., Rudzinski, W., 2013. Modeling of sorption kinetics: the pseudo-second order equation and the sorbate intraparticle diffusivity. Adsorption 19 (5), 1055e1064.
Stackelberg, P.E., Gibs, J., Furlong, E.T., Meyer, M.T., Zaugg, S.D., Lippincott, R.L., 2007. Efficiency of conventional drinking-water-treatment processes in removal of pharmaceuticals and other organic compounds. Sci. Total Environ. 377 (2e3), 255e272.
Wu, S.H., Pendleton, P., 2001. Adsorption of anionic surfactant by activated carbon: effect of surface chemistry, ionic strength, and hydrophobicity. J. Colloid Interface Sci. 243, 306e315.
Zhang, S., Shao, T., Kose, H.S., Karanfil, T., 2010. Adsorption of aromatic compounds by carbonaceous adsorbents: a comparative study on granular activated carbon, activated carbon fiber, and carbon nanotubes. Environ. Sci. Technol. 12 (864), 6377e6383. Retrieved from. http://pubs.acs.org/doi/full/10.1021/es100874y.
Zhang, Shujuan, Shao, T., Karanfil, T., 2011. The effects of dissolved natural organic matter on the adsorption of synthetic organic chemicals by activated carbons and carbon nanotubes. Water Res. 45 (3), 1378e1386.