Grantee Research Project Results
Final Report: Graft Polymerization as a Route to Control Nanofiltration Membrane Surface Properties to Manage Risk of EPA Candidate Contaminants and Reduce NOM Fouling
EPA Grant Number: R830909Title: Graft Polymerization as a Route to Control Nanofiltration Membrane Surface Properties to Manage Risk of EPA Candidate Contaminants and Reduce NOM Fouling
Investigators: Kilduff, James E. , Belfort, Georges
Institution: Rensselaer Polytechnic Institute
EPA Project Officer: Aja, Hayley
Project Period: August 25, 2003 through August 25, 2007
Project Amount: $349,000
RFA: Environmental Futures Research in Nanoscale Science Engineering and Technology (2002) RFA Text | Recipients Lists
Research Category: Nanotechnology , Safer Chemicals
Objective:
The objectives of our research are to develop new nanofiltration membranes by modifying the surface structure of commercial membranes at the molecular level via UV-assisted graft polymerization of hydrophilic monomers using our patented method. Toward this end, we developed, validated and tested a novel high throughput method for synthesis and screening of customized membrane surfaces. The method combines, for the first time, a high throughput platform (HTP) approach together with photo-induced graft polymerization (PGP) for facile modification of commercial poly(aryl sulfone) membranes. Our goal is to develop new materials that offer high flux compared to commercial membranes (by improving membrane porosity), enhanced rejection of inorganic anions and ionizable organic compounds (by controlling membrane pore size distribution and surface charge), and enhanced ability to resist fouling by natural organic matter (NOM) by reducing adhesion. In addition, we seek to understand the characteristics of natural organic matter accumulated on membrane surfaces both in terms of resistance to flow (which can influence the cost of membrane processes), and its affects on the transport and rejection of charged solutes.
The findings of this research are presented in six sections. First, the general materials and methods used in the research are presented. In some cases, further detail is provided in subsequent sections. The next section deals with an analysis of approaches to compare membrane performance. This is especially important for our work since optimizing membrane performance is a primary goal. The next section highlights our development of a high throughput platform for membrane surface modification. This approach promises to revolutionize how membrane surface modification is done. The next section considers surface modification to optimize rejection of charged contaminants. Because membrane modification can affect properties like surface roughness, which can in turn affect membrane performance in terms of fouling, the next section details our investigation into the effects of surface roughness. Finally, we consider a novel approach to modeling contaminant breakthrough behavior. Such behavior was observed for some organic contaminants that exhibit time-dependent rejection as a result of solute adsorption.
Summary/Accomplishments (Outputs/Outcomes):
Materials and Methods
All membranes used in this research are thin-film composite membranes. Poly(ether sulfone) (PES) membranes were applied in surface modification and the effect of roughness on fouling and cleaning study. One sulfonated poly(sulfone) (PSf) and two aromatic polyamide (PA) membranes were used in the organic chemical filtration, adsorption and breakthrough curve modeling. All the monomers applied in graft polymerization were commercially available vinyl monomers. The monomers were dissolved either in DI water or ethanol depending on their solubility. A soil humic acid (Elliott Humic Acid) from IHSS (International Humic Substance Society) as a surrogate was applied as NOM source. This type of humic acid was chosen because it has a relatively high fouling potential, and therefore provides a rigorous test of anti-fouling surfaces.
Membranes were modified and tested at the bench scale (membrane area greater than 3.8 cm2) and at high-throughput scale, in a 96-well format (effective membrane area 0.19 cm2). Membranes were photochemically modified using a photochemical reactor system. Attenuated total reflection Fourier transform infrared spectroscopy (ATR/FTIR) was used to characterize the surfaces of as-received and modified membranes to quantify the amounts of polymers grafted to the membrane surface. Sessile contact angles of as-received and modified membranes were measured using the captive bubble technique. Contact angles were used to assess wettability of the modified surfaces, which is a measure of surface polarity. The surface charge or electrical potential properties of as-received and modified membranes was measured in terms of streaming potential by using an Electro Kinetic Analyzer. Streaming potential was then used to calculate zeta potential (ζ). Atomic force microscopy (AFM) was used to characterize surface morphology properties and to measure membrane roughness.
Two types of dead-end stirred cell systems were applied in the filtration experiments depending on the applied membranes and the required operating pressure for filtration. Both feed and permeate samples were collected to vials at scheduled times and the weights were measured with an electrical balance.
All NOM samples were quantified using a total organic carbon analyzer. A dynamic light scattering technique was used to measure the size of NOM aggregates. A UV-visible spectrophotometer was used to measure the concentration of proteins, silica colloids and some NOM samples. An HPLC system equipped with an ultraviolet-visible (UV-Vis) detector was used to determine the concentration of phenol, 2,4-DNP and metolachlor samples. The concentrations of the unknown samples were then calculated based on the calibration curve obtained from the external standards. A reagent-free ion chromatography (IC) system was applied to determine the concentration of perchlorate and arsenate samples.
Evaluating Different Approaches to Membrane Comparison Based on Fouling Model Sensitivity Analysis
Introduction
A common goal in membrane process research is to minimize the effects of fouling, which presents the need to compare the performance of different membranes for the same feed, or compare performance for different feed solutions, or both. Flux decline is often used to assess filtration performance. The membranes (or feeds) with smaller flux decline are generally considered as having lower fouling potential and hence higher performance; few literature reports interpret such data in the context of the membrane hydraulic permeability. Such comparisons can be misleading when membranes having different permeability are evaluated, because flux decline is smaller for membranes having greater resistance, especially when the feed solution and membrane-foulant interactions are similar. We have employed the combined pore blockage and cake filtration model to analyze constant pressure dead-end filtration to identify the best (least biased) approaches to compare membrane performance.
Results
The sensitivity analysis was done with parameters determined from fitting representative data sets of natural organic matter (NOM) filtration by 1 kDa PES membranes. The effects of membrane resistance, Rm, on solution flux, keeping pore blockage and cake formation parameters constant, are shown in Figure 1. The sensitivity analysis is shown in terms of flux plotted both as a function of time and volume throughput, V. Even though the fouling parameters are identical, the flux decline curves are quite sensitive to membrane resistance. Therefore, flux decline (or normalized flux decline) plotted versus either operating time or volume throughput, is not a good comparative indicator of fouling – this plot tends to underestimate the fouling potential of membranes having higher resistance. This is true regardless of whether experiments use the same initial flux or the same initial transmembrane pressure. This is of particular relevance for membrane modification, which tends to increase membrane resistance. Therefore, modified membranes, can exhibit lower flux decline solely because membrane resistance increased.
Figure 1. Effect of membrane resistance on flux vs. operating time (A) and volume throughput (B).
A better approach to comparing the fouling of different membranes, especially when the membrane resistance is different, is to plot the fouling resistance, RF, versus volume throughput, V. As shown in Figure 2, when fouling parameters are held constant, RF vs. V data collapses onto one curve even for membranes having different resistance. This is also true regardless of whether membranes are compared in experiments using the same initial flux or the same initial transmembrane pressure. The mass accumulation for different membranes is not the same at the same operating time; therefore, the RF vs. t plot should not be used to compare membranes because this plot tends to overestimate the fouling potential of membranes having higher resistance.
Figure 2. Effect of membrane resistance on RF vs. operating time (A) and volume throughput (B).
Conclusions
Although we generally seek to minimize flux decline during membrane filtration, we should not compare membrane performance by comparing flux decline directly. This is because the membrane resistance has a significant effect on flux decline, both as a function of time and volume throughput. At both constant initial flux and constant pressure, the flux decline is smaller for membranes having greater resistance, even when the feed solution and membrane-foulant interactions are the same. This is relevant to the comparison of membranes having different pore sizes or MWCO’s resulting from either manufacture or modification, but is also relevant to comparing the same membrane because of inherent variability in membrane properties.
Two comparisons are recommended. To compare resistance due to fouling, the fouling resistance should be plotted as a function of volume throughput. Such curves are independent of membrane resistance, for all fouling mechanisms. It should be noted that fouling resistance as a function of time is not independent of membrane resistance. In addition, it is often desirable to compare total resistance, the sum of fouling resistance and membrane resistance, because this parameter relates to permeate production and energy cost. It is best to also plot total resistance versus volume throughput. Although total resistance clearly depends on the membrane resistance, plotting it in this way eliminates the effects of membrane resistance on the fouling component of the total resistance.
High Throughput Membrane Surface Modification
Introduction
The search for customized surfaces has been limited due to (i) the expense and time needed to develop and find new materials with optimal interfacial characteristics, and (ii) the diverse characteristics (i.e. multiple interactive forces acting simultaneously) of natural waters. More importantly, surface science has not yet developed to the point that allows prediction of the surface or functional characteristics needed to minimize undesirable interactions with solution components, and thus to control fouling.
In this research we have adapted – for the first time – high throughput platform (HTP) approaches successfully used in chemistry to the facile modification of PES, using a HTP together with photo-induced graft polymerization (PGP). In the PGP method, depicted schematically in Figure 3, poly(aryl sulfone) membranes are UV-irradiated, cleaving trunk polymer chains and forming reactive radical sites. Either water or methanol-soluble vinyl monomers covalently bond to these radical sites and undergo free-radical polymerization; our initial library of 66 monomers is also shown in Figure 3. The novel method developed here is an inexpensive, fast, simple, reproducible and scalable procedure to synthesize and screen fouling-resistant surfaces. Using the HTP-PGP process we first (i) validate the method with monomers previously investigated in large-scale, (ii) identify the best-grafted monomers from a large library of candidates, and (iii) challenge these grafted monomers with filtration and sorption assays to assess performance.
Figure 3. In the PGP method, poly(aryl sulfone) membranes are UV-irradiated (λ ≈ 300 nm), cleaving trunk polymer chains and forming reactive radical sites. Either water or methanol-soluble vinyl monomers chemically bond to these radical sites and undergo free-radical polymerization. Initial monomer library consists of 66 monomers in nine groups.
Methods and Materials
Polypropylene 96-well filter plates were used in HTP-PGP experiments. A 100 kDa cut-off PES membrane coupon (effective area 19.35 mm2) was mounted by the manufacturer on the bottom of each 400-μL well. Commercial vinyl monomers (66 total, Figure 3) were either dissolved in reagent grade water or ethanol depending on their solubility. Six of these monomers were used for HTP-PGP validation: N-vinyl pyrrolidinone (NVP), 2-hydroxyethyl methacrylate (HEMA); acrylic acid (AA), 2-acrylamidoglycolic acid (AAG), 3-Sulfopropyl methacrylate (SPMA) and 2-Acrylamido-2-methyl-1-propanesulfonic acid (AMPS). UV irradiation was done in a chamber containing an electrodeless microwave lamp.
The resistance of the 96 membrane coupons during solution filtration was measured simultaneously by mounting the filter plate on a vacuum manifold. The permeate from each membrane was collected into a corresponding well in the receiver plate, and was analyzed for solute concentration and volume to calculate the rejection and flux properties of each membrane. The resistance of modified and control membranes was also evaluated using a static adsorption protocol. The amount of NOM adsorbed and the water permeability after adsorption were measured as criteria to evaluate membrane performance.
Validation of HTP-PGP technique
Two separate experiments were conducted with 60 monomers to evaluate the reproducibility of the HT platform. The data from these experiments, expressed in terms of the resistance after modification, correlated with a correlation coefficient of 0.95, indicating excellent reproducibility. The HTP-PGP approach was validated by comparing results with those obtained from a conventional large-scale low-throughput (LT) approach using the six monomers (NVP, HEMA, AA, AAG, SPMA and AMPS) described by Taniguchi and Belfort (2004). The foulant used in these experiments was bovine serum albumin (BSA). The HTP-PGP technique identified the same trends as the large-scale experimental approach and is therefore scalable in terms of membrane resistance developed by interactions between the surface and feed components.
Resistance measured during the protein solution filtration assay was compared with resistance during filtration of a protein-free feed after static adsorption from a protein solution. Comparison of these data confirmed that the filtration and static adsorption evaluation protocols exhibit similar trends in measured resistance. The general agreement of the results validates the scalability of the static adsorption protocol.
HT discovery of new surfaces
Data analysis included calculating the membrane resistance after modification, Rmod, relative to the resistance of the as-received membrane, RAR, washed with water only. The ratio of these resistances, Rmod/RAR, represents the factor by which membrane resistance increased after modification, and is a rough indicator of the amount of grafted material. A fouling index was calculated as the resistance increase of grafted membranes caused by fouling (foulant adsorption) (Rfouled – Rinit)mod normalized by that of ungrafted membrane control, (Rfouled – Rinit)control. That is,
= (Rfouled – Rinit)mod / (Rfouled – Rinit)control. The control is the as-received membrane treated with either water or ethanol, depending on which was used to dissolve the monomer. For monomers dissolved in water, Rcontrol is the same as RAR. It is desirable for the increase in resistance after fouling to be lower for the modified membrane than the control, and for the membrane resistance after modification to be near that of the as-received membrane (Rmod ≤ RAR), although a higher resistance may be favorable when it correlates with increasing rejection.
The fouling index was rated into 7 classes to facilitate evaluation of filtration performance. The designation “---” means the membrane had among the highest resistance or amount adsorbed after surface modification, which indicates unfavorable surface chemistry; in contrast, the designation “++++” means the membrane had among the lowest resistance or amount adsorbed after surface modification, which indicates favorable surface chemistry. NOM results in terms of resistance increase after adsorption vs. resistance change after modification is shown in Error! Reference source not found., and the optimized selection of NOM-resistant surfaces relative to the as-received PES membrane without considering resistance increase after modification is shown in Figure.
For NOM, there were 20 monomers which had Rmod/RAR less than 2, and fouling index less than unmodified membrane. The PEGs performed well, and the performance improved with increasing molecular weight, i.e., the repeating unit n increased. However, compared to PEGs, two monomers (monomer #53, diacetone acrylamide, and #60, [2-(Methacryloyloxy)ethyl]dimethyl-(3-sulfopropyl) ammonium hydroxide) had a negative resistance increase value, indicating their extraordinary-NOM fouling resistant properties. In addition, several amines and acid monomers were also promising at mitigating PES surface adsorption fouling by NOM.
Figure 4. HT screening of NOM/surface interactions in terms of DI filtration resistance after NOM adsorption by grafted membranes, and DI filtration resistance by grafted membranes in the absence of NOM. Selected surfaces (of 66 total) are shown; numbers refer to Figure 3. Other surfaces exhibit either fouling greater than the as-received membrane, or high membrane resistance. The values shown are the mean of four separate measurements.
Figure 5. Optimized selection of NOM-resistant surfaces from a total of 66 commercial monomers relative to the as-received PES membrane using the HTP-PGP method; numbers refer to Figure 3. Success is measured in terms of fouling index, .
Conclusions
A novel high throughput method for synthesis of customized surfaces by photo-induced graft polymerization was developed in our work. This method offers inexpensive, fast, simple, reproducible and scalable properties. This technique was validated against previous results of six monomers obtained using a large scale method. A library of 66 commercial vinyl monomers were screened and the best performing monomers and feed-specific surfaces were identified for six feeds, including NOM. Two different assays were evaluated, foulant solution filtration, and filtration of a foulant-free feed after static adsorption from the foulant solution, and found to yield consistent results. Low fouling and feed-specific surfaces were identified.
The data reported here were obtained under well-defined experimental conditions of pH, temperature, and monomer concentration, and the grafting conditions were chosen based on our previous experience. It is likely that the grafting conditions (e.g., monomer concentration and UV irradiation time) for many previously untested monomers were sub-optimal. The HTP-PGP method should be used to optimize these conditions in future work.
Surface Modification to Optimize Rejection of Water Contaminants
Introduction
Nanofiltration (NF) is the filtration process between reverse osmosis (RO) and ultrafiltration. NF membranes are generally non-porous or have pores on the order of 1 nm, and the membranes are usually charged. These properties of NF membranes make them capable of removing charged and/or large chemicals from water at lower operating pressure than RO filtration process. The mechanisms of chemical rejection by NF membranes include the steric (size) exclusion, electrostatic interaction (charge exclusion), and the affinity between chemical and the membrane materials, such as the adsorption of the organic compounds to membrane matrix. The chemical removal efficiency by NF membranes can by manipulated by tailoring surface properties including pore size, pore size distribution and surface charge. In this chapter, graft polymerization of vinyl monomers was used to modify NF membrane surface properties as an effort to manipulate the solute transport through membrane pores and to improve rejection.
Materials and methods
The membranes used in this research was OS-001 (PES, 1 kDa MWCO) obtained from Pall Corp. East Hills, NY. Membranes coupons were pre-cleaned by soaking in DI water overnight before modification. Vinyl monomers evaluated included neutral NHMA and strongly (sulfonic) acidic SPMA. Four compounds were selected as surrogate contaminants, including arsenate (arsenic V), perchlorate, 2,4-dinitrophenol (2,4-DNP) and metolachlor. Arsenic is regulated and the rest three are on the EPA Contaminant Candidate List. Perchlorate and arsenic represent inorganic anions, 2,4-DNP represents ionizable organic chemical, and metolachlor is a fairly large, neutral compound. Solutions containing single solute (0.1 mM 2,4-DNP, 0.01 mM metolachlor, 10 mg/L arsenate) and binary solutes (0.1 mM 2,4-DNP and 0.1 mM perchlorate) at different pHs were used in filtration experiments to evaluate the ability of surface graft-polymerized membranes to reject chemicals in single chemical and binary chemical solution, and to elucidate the rejection mechanism. Membrane modification was conducted using the large scale method described previously. The as-received and modified membranes were characterized with DG and zeta potential using the methods described previously.
Results and Discussions
Permeability, DG and Zeta Potential
The membrane permeability was reduced dramatically after grafting, especially when NHMA was used, consistent with DG data. The permeability of RO membranes are usually in the range of 0.08 to 0.2 LMH/psi, which was lower than that of the membranes modified with 5 wt % SPMA even at 60 s UV, but higher than most of the NHMA grafted membranes. Therefore, lower concentration and shorter UV times should be used for NHMA to ensure the modified membranes have sufficiently high flux. Zeta potential measurement indicated that neutral HEMA did not change membrane zeta potential, but negatively charged SPMA was very effective at tailoring membrane surface charge. With grafting of SPMA, membrane zeta potential decreased from almost zero to -15 mV at 20 and 40 s of UV irradiation.
Effect of SPMA on 2,4-DNP filtration
At pH 3, the 2,4-DNP rejection by all the membranes was nearly 100 %, with the value decreasing slightly with filtration time. The high rejection at pH 3 was likely due to the adsorption of 2,4-DNP by membrane material, i.e., the rejection was dominated by solute adsorption. At pH 10, all the 2,4-DNP molecules were ionized and negatively charged. Therefore, the charge exclusion was believed to play important role in the solute rejection, in addition to sieving effect, if any. The rejection result was quite consistent with membrane surface charge measurement. Therefore, it is likely that the high rejection of 2,4-DNP by membranes with 40 and 60 s UV shown in Figure 6 was due to the combination effects of charge repulsion and sieving.
Effect of SPMA on metolachlor filtration
During the filtration results of 0.01 mM metolachlor at pH 5.4 by as-received and SPMA modified OS-001 membranes breakthrough behavior was observed. At the end of the filtration, the concentration of the permeate nearly reached that of the feed, lead to a low rejection of only 5 %. Time-dependent rejection was due primarily to adsorption, and grafting the membrane surface with SPMA did not improve metolachlor rejection.
Effect of SPMA on arsenate filtration
The membranes modified with 5 wt % SPMA were also evaluated by arsenate (10 mg/L at pH 7.5) filtration, with the results shown in Error! Reference source not found.. Rejection of arsenate was increased from 80% by unmodified membranes to 94% by grafted membranes at 40 s UV irradiation. The concentration profiles of feed solution indicate that adsorption was not the major reason for rejection. Considering the size of arsenate, it is likely that charge repulsion played a more important role than sieving.
Figure 6. Filtration results of 0.12 mM 2.4-DNP at pH 10 by as-received and 5 wt% SPMA modified OS-001 membranes with UV irradiation times of 5 - 60 s.
Figure 7. Filtration results of 10 mg/L arsenate at pH 7.5 by as-received and 5 wt% SPMA modified OS-001 membranes with UV irradiation times of 5 - 60 s.
Effect of SPMA on binary mixture filtration
In addition to single solute filtration, solutions containing two solutes (2,4-DNP and perchlorate) were also employed to examine the performance of grafted membranes, and to study the effect of solute interaction on their transport. At pH 4.3, 33% of 2,4-DNP molecules were protonated and uncharged. Therefore, both adsorption and charge exclusion should play important roles in solute rejection. The 2,4-DNP rejection results demonstrate that the modified membranes had a slightly higher rejection than the as-received membrane. The 2,4-DNP rejection at pH 3 was higher than that at pH 4.3, indicating that adsorption was more important than charge exclusion in 2,4-DNP rejection.
The perchlorate rejection at pH 4.3 shown in Error! Reference source not found. indicates that grafting of SPMA to OS-001 membranes could significantly improve perchlorate rejection at UV irradiation times of 20, 40 and 60 seconds. This result is consistent with the membrane zeta potential measurement, which demonstrates that at these three UV irradiation times, the membrane surface negative charge increased greatly. Therefore, the increase of the perchlorate rejection was due to the influences of enhanced charge repulsion of the modified membrane, as observed for single solute filtration.
Figure 8. Perchlorate filtration by as-received and 5 wt% SPMA modified OS-001 membranes with UV irradiation times of 5 - 60 s. Feed solution condition: mixture of 0.1 mM 2.4-DNP and 0.1 mM NaClO4, pH 4.3.
Conclusions
Four chemicals including 2,4-dinitrophenol (2,4-DNP), perchlorate, arsenic and metolachlor were chosen as model compounds to evaluate modified membranes. Zeta potential measurements indicate that grafting SPMA to the surface of a 1 kDa PES membrane significantly increased the negative surface charge. Whether this increase in surface charge enhanced membrane performance depended on contaminant charge, as expected. Two solutes, neutral metlaochlor and partially dissociated 2,4-DNP at pH 4.3 (pKa = 4) exhibited breakthrough behavior. Whereas rejection was initially high, it decreased with time as the solute was transported through the membrane. The initially high rejection was attributed to adsorption to the membrane matrix. In contrast, rejection of perchlorate, arsenate, and ionized 2,4-DNP was significantly increased by surface modification, indicating the important role of charge repulsion. The rejection of perchlorate by the as-received membrane was near zero, but modification was able to increase rejection up to 90%. Rejection of arsenate was increased from 80% by unmodified membranes to 94% by UV grafted membranes at pH 7.5. These results illustrate that graft polymerization of charged monomers on commercial membrane surfaces represents a promising approach to enhance the removal of charged contaminants.
Effects of Surface Roughness
Introduction
Membrane surface roughness is an important but often neglected characteristic with respect to macromolecular adsorption and membrane fouling, and one that may be changed during surface modification. Therefore, we are interested in the effects of surface roughness on fouling and cleaning. We have studied as-received membranes which are made of the same material (PES), but having a wide MWCO range in an effort to study the effect of membrane pore size and roughness. These membranes were fouled by humic acid, and subsequently cleaned with water and sodium hydroxide solution. Calcium was added to selected experiments to study the effect of foulant charge and size on fouling. The results were analyzed and presented using resistance increase to rule out the influence of membrane resistance on fouling.
Experimental
The PES membranes with a wide range of MWCO were used in this study. These membranes were OMEGA series membranes obtained from Pall Corp. (East Hills, NY), and could be divided into two types named OM and OT according to the manufacturer. Elliott soil humic acid (HA) from the International Humic Substance Society was used as a surrogate for natural organic matter. Stock solutions were prepared in the presence and absence of calcium ion.
Results and Discussion
Membrane Properties
ATR/FTIR spectra of studied membranes indicated that these membranes have essentially the same surface chemistry because the spectra for all the membranes in the region of 700-4000 cm-1 are identical. Therefore, the wettability and surface charge of these membranes is expected to be the same too. Therefore, we attribute the different fouling and cleaning behavior of these membranes to factors other than membrane surface chemistry. Roughness was characterized in terms of the root-mean-square (RMS) value. The OM membranes studied in this research had a much rougher surface than OT membranes.
Humic Acid Fouling and Cleaning
For both OT and OM membranes, the humic acid rejection monotonically decreased with increasing MWCO in both the presence and absence of calcium, although the rate of decrease was greater when calcium was absent. When membrane MWCO was less than 100 kDa, humic acid rejection was higher when calcium was absent from the feed solution. A higher rejection in the absence of calcium is consistent with charge repulsion between the humic acid and the membrane surface. Although aggregate size was smaller in the absence of calcium, it was still large enough to be rejected by the <100 kDa membranes. For the larger MWCO membranes, the smaller aggregates were not rejected, while the larger ones were rejected; therefore, rejection increased when calcium was present and aggregate size was increased. Charge does not seem to play a significant role for the larger MWCO membranes.
When calcium was absent, the initial layer of deposit imposed a more significant resistance than the later deposit. This suggests the importance of pore blockage and constriction during the initial stages of filtration. During later stages of filtration, the fouling mechanism transitions to cake formation. The low increase in the resistance during this time suggests that charge repulsion forces between charged molecules resulted in a cake with a loose, permeable structure.
The trends in resistance observed for the OM membranes during humic acid filtration were similar to those observed for the OT membranes. However, the rougher OM membranes had a higher fouling than smoother OT membranes. This trend was especially evident when Ca2+ was present in the feed. The membrane characterization revealed that all the membranes have similar surface chemistry, but that the OM membranes have a rougher surface. Therefore, the difference in fouling by these two types of membrane appears to be caused by difference in roughness, which seems to have the largest effect on membrane fouling when calcium was present in the feed.
The fouling removal after DI and NaOH cleaning of the humic acid fouled membrane reached almost 100 % (resistance dropped to the initial value) for the membranes fouled with humic acid in absence of calcium. In contrast, fouling removal was much lower when calcium was present in the feed solution. This is consistent with the research results that the calcium neutralized NOM aggregates have a higher adsorption for the membrane surface. An important finding is that the lower roughness membranes were easier to clean by both DI water and NaOH solution, although differences in fouling removal due to differences in fouling mechanism cannot be ruled out.
Conclusions
Results demonstrated that larger roughness caused greater flux decline for organic colloids. However, although membrane roughness varied with MWCO, it was found that in general, fouling decreased with increasing membrane MWCO; therefore, the effect of pore size and rejection on flux was more important than roughness.
As for cleaning, membranes exhibiting lower roughness were easier to clean when affinity of foulant with membrane surface was high (HA-calcium experiment), and/or when the foulant size was similar to membrane surface roughness value. The membrane surface properties have effect only on the initial fouling, which can more easily be seen from cleanability of the membranes than from fouling extent due to the quick transition of fouling mechanisms.
Modeling Permeate Breakthrough Curves
Introduction
During organic compound filtration, solutes may be rejected by steric (size) exclusion, electrostatic interaction (charge exclusion), and by adsorption to the membrane matrix. This sorption effect is significant especially for hydrophobic compounds, which exhibit breakthrough behavior. A model was developed to describe breakthrough curves of organic solutes under both neutral and charged conditions, using a one-dimensional porous medium model coupled with hindered transport theory. Sensitivity analysis on the model parameters was performed to gain more insight into the model and the solute transport mechanisms.
Modeling
The well-known mass conservation equation for one-dimensional solute transport through porous media, the convection-dispersion equation, was employed. To solve this equation analytically, two assumptions were made: local equilibrium, and linear adsorption uptake. The local equilibrium assumption implies that no mass transfer limitations control the rate of adsorption. Under the above assumptions, the advection dispersion equation can be written in terms of a retardation factor, which accounts for linear and reversible equilibrium adsorption. Using flux-type boundary conditions, an analytical solution to the advection dispersion equation is available. The effects of retardation caused by solute adsorption, and hindrance caused by solute-pore steric interactions, are accounted for by applying the retardation factor in combination with solute hindered transport theory.
Hindered Transport Theory
In the pressure-driven nanoporous membrane filtration processes, pore size and solute size are on the same order, and convective and diffusive transport coexists when solutes pass through membrane active layer. Because of the comparable size of the pores and solutes, both solute diffusion and convection are influenced greatly by the pore walls. This phenomenon is called hindered transport. Hindered transport theory was used to quantify the hindrance effect on solute transport assuming uncharged, solid, spherical solute diffusion and convection in the membrane pores.
Experimental
Three NF membranes, NTR 7450, NF 90, and NF 270, were used in this research. NTR 7450 was sulfonated polysulfone membrane obtained from Nitto Denko Japan, Inc.; NF 90 and NF 270 were aromatic polyamide thin-film composite membranes donated by Dow-FilmTec. The organic chemicals applied were phenol and 2,4-DNP. In filtration experiments, 2 mM phenol solution with pH 6.10 and 0.1 mM 2,4-DNP solution with pH 4.30 were used as feed solutions. Batch isotherms were measured to test the effect of pH on adsorption and to study adsorption mechanisms.
Results and Discussion
Adsorption Isotherms
Batch sorption experiments, confirmed that solute dissociation had a significant effect on adsorption, decreasing as the solute becomes more ionized. The sorption uptake of NTR 7450 was higher than that of NF 90 and NF 270, which were similar.
Filtration Performance
In all the filtration experiments in this study, the concentration of feed solution remained constant during the entire filtration process, and concentration of permeate increased rapidly with time, approaching that of the feed solution, in the form of a classic breakthrough curve. Because the feed solution did not increase during filtration, solute removal was not due to size or charge exclusion; therefore, we attribute solute removal to adsorption. Solutes adsorbed inside the pores will narrow the free pathway of the water flow, leading to pore blockage when adsorption is strong and when the solute and membrane pore sizes are similar (λ approach to 1). Some flux decline was observed here, increasing with solute uptake by the membranes. However, the flux decline was modest in comparison to that caused by organic colloids such as natural organic matter.
Modeling Breakthrough Curves
Representative model fitting of the breakthrough curves using the 1-D porous medium model coupled with hindered transport theory are shown as solid lines in Figure 9 for phenol; similar results were obtained for 2,4-DNP. We found that the model fitting was influenced greatly by the size of the solute relative to the size of the pore. In general the model fit improved as the solute size decreased relative to the pore size. A possible explanation of this observation was that when λ approaches 1, effects such as van der Waals forces and charge repulsion between solute and pore wall play important roles on solute transport, and these effects are not considered in the model. The fitted retardation coefficient values from the permeate concentration profiles and the batch isotherm data are consistent – the value of retardation factor, which represents the adsorption uptake capacity of solutes by membranes, increased with increasing batch adsorption uptake.
Model Sensitivity Analysis
To examine the influence of solute velocity and diffusivity on the shape of the breakthrough curves and to shed light on the directions of further study, model sensitivity analysis were performed for 2,4-DNP filtration with NF 90 and NF 270 membranes, the two sets of experiments which the model fitting was not good. Model fitting of experimental data improved when the solute convective transport decreased and/or diffusive transport increased. This suggested that solute convective flux was overestimated, and diffusion transport was underestimated by the model. Of the two parameters, the solute convective velocity had a greater effect than diffusion coefficient on breakthrough curve shape, especially at longer operating times.
Conclusions
During filtration of organic chemicals, the sorption of solutes by membrane could lead to a permeate breakthrough behavior. The breakthrough curve was modeled using a 1-D solute transport model, treating the membrane as a porous medium, coupled with hindered diffusive and convective transport theory to account for the pore wall effect on solute transport. This new modeling approach accurately fitted experimental data, and is a promising approach for predicting membrane performance. More work is needed to incorporate other effects (such as charge repulsion) on hydrodynamic hindrance factors. In addition, isotherm nonlinearity and nonequilibrium should also be considered.
Figure 9. Permeate and feed concentrations of phenol as a function of filtration time for (a) NTR 7450; (b) NF 90. The symbols represent experimental data, solid lines represent the pore transport model fitting.
Journal Articles on this Report : 4 Displayed | Download in RIS Format
| Other project views: | All 24 publications | 5 publications in selected types | All 4 journal articles |
|---|
| Type | Citation | ||
|---|---|---|---|
|
|
Kilduff JE, Mattaraj S, Zhou M, Belfort G. Kinetics of membrane flux decline: the role of natural colloids and mitigation via membrane surface modification. Journal of Nanoparticle Research 2005;7(4-5):525-544. |
R830909 (2005) R830909 (Final) |
Exit Exit |
|
|
Zhou M, Liu H, Kilduff JE, Langer R, Anderson DG, Belfort G. High-throughput membrane surface modification to control NOM fouling. Environmental Science & Technology 2009;43(10):3865-3871. |
R830909 (Final) |
Exit |
|
|
Zhou M, Liu H, Kilduff JE, Langer R, Anderson DG, Belfort G. High throughput synthesis and screening of new protein resistant surfaces for membrane filtration. AIChE Journal 2010;56(7):1932-1945. |
R830909 (Final) |
Exit |
|
|
Zhou M, Liu H, Venkiteshwaran A, Kilduff J, Anderson DG, Langer R, Belfort G. High throughput discovery of new fouling-resistant surfaces. Journal of Materials Chemistry 2011;21(3):693-704. |
R830909 (Final) |
Exit |
Supplemental Keywords:
RFA, Scientific Discipline, Water, POLLUTANTS/TOXICS, Sustainable Industry/Business, Sustainable Environment, Arsenic, Technology for Sustainable Environment, Environmental Monitoring, Water Pollutants, Engineering, Chemistry, & Physics, Drinking Water, Environmental Engineering, monitoring, public water systems, Safe Drinking Water, graft polymerization, nonocomposite filter, natural organic material, membranes, nanotechnology, chemical contaminants, community water system, treatment, nanofiltration, nanoporous membranes, contaminant removal, drinking water contaminants, drinking water treatment, water treatment, nanocomposite filter, nanofiltration membranes, green chemistry, drinking water systemProgress and Final Reports:
Original AbstractThe perspectives, information and conclusions conveyed in research project abstracts, progress reports, final reports, journal abstracts and journal publications convey the viewpoints of the principal investigator and may not represent the views and policies of ORD and EPA. Conclusions drawn by the principal investigators have not been reviewed by the Agency.