Application of Silver Nanoparticles in Drinking Water Purification

Nanotechnology is an emerging and fast-growing technology. Currently, there are more than 1,317 nanotechnology-based products on the market. Silver nanoparticles account for more than 23% of all nano-products. The extensive application of the silver nanoparticle (AgNP) results in their inevitable release into the environment. Silver nanoparticles are known as excellent antimicrobial agents, and therefore they could be used as alternative disinfectant agents. On the other hand, released silver nanoparticles could pose a threat to naturally occurring microorganisms. In Chapter 1, we introduce the background information on the environmental fate, toxicological effects, and application of AgNP and review the current knowledge on the physicochemical and antimicrobial properties of AgNP in different aqueous solutions, as well as their application as alternative disinfectants in water-treatment systems. In Chapter 2 of this dissertation, we discuss the evaluation of AgNP’s antimicrobial properties at different water chemistry conditions. It was found that the aggregation of silver nanoparticles depends on the properties of the background ions, such as Na + and Ca 2+ , at different water chemistry conditions. Divalent cations can significantly enhance the aggregation, while monovalent cations and anions do not have such a significant influence. A saturation-type fitting curve was established, showing the survival of bacteria under different water chemistry conditions as a function of the size of the nanoparticles. In Chapter 3, we talk about the evaluation of the antimicrobial properties of AgNP when coated with different organic compounds using natural water conditions. The results obtained showed that silver nanoparticles in surface water, ground water, and brackish water are stable. However, in seawater conditions, AgNP tend to aggregate. This study also shows that the antimicrobial activity of AgNP can be impaired by the presence of a humic substance and high concentrations of divalent cations. These results are helpful in explaining how discharged AgNP behave in natural aquatic systems as well as their environmental toxicological effects on naturally occurring microorganisms. In Chapter 4, we discuss the investigation of the effect of silver nanoparticles on a model virus-MS2 bacteriophage. A negligible deactivation effect on MS2 phage was found regardless of the type of AgNP and the water chemistry conditions used. In Chapter 5, we talk about the comparison of AgNP-impregnated point-of-use ceramic water filters and ceramic filters impregnated with silver nitrate. This study was performed using different water chemistry conditions and different manufacturing materials. The results showed that AgNP-impregnated ceramic water filters are more appropriate for this application due to the lesser amount of silver desorbed compared with silver nitrate-treated filters. The bacterial removal performance of the silvertreated ceramic filters and concentration of viable bacteria in the filters are dosedependent on the amount of silver applied. However, the data showed that influent water chemistry conditions did not have a significant effect on the performance of the filters. This study established evidence-based silver application guidelines for the ceramic water filter manufacturers around the world. In Chapter 6, the discussion centers in the comparison of a polymer-based quaternary amine functionalized silsesquioxanes compound and AgNP. The results showed that the quaternary ammonium functionalized silsesquioxanes-treated ceramic water filter desorbed less from the filters and achieved higher bacteria removal than the filters impregnated with AgNP. This indicates that the quaternary ammonium functionalized silsesquioxanes compound could be considered as a substitute for silver nanoparticles due to its lower price and higher performance. However, more information regarding the possible chronic health effects of the silsesquioxanes compound is needed. In Chapter 7, we present the main conclusions and recommended future work based on the dissertation results.

nanoparticles could pose a threat to naturally occurring microorganisms.
In Chapter 1, we introduce the background information on the environmental fate, toxicological effects, and application of AgNP and review the current knowledge on the physicochemical and antimicrobial properties of AgNP in different aqueous solutions, as well as their application as alternative disinfectants in water-treatment systems.
In Chapter 2 of this dissertation, we discuss the evaluation of AgNP's antimicrobial properties at different water chemistry conditions. It was found that the aggregation of silver nanoparticles depends on the properties of the background ions, such as Na + and Ca 2+ , at different water chemistry conditions. Divalent cations can significantly enhance the aggregation, while monovalent cations and anions do not have such a significant influence. A saturation-type fitting curve was established, showing the survival of bacteria under different water chemistry conditions as a function of the size of the nanoparticles.
In Chapter 3, we talk about the evaluation of the antimicrobial properties of AgNP when coated with different organic compounds using natural water conditions.
The results obtained showed that silver nanoparticles in surface water, ground water, and brackish water are stable. However, in seawater conditions, AgNP tend to aggregate. This study also shows that the antimicrobial activity of AgNP can be impaired by the presence of a humic substance and high concentrations of divalent cations. These results are helpful in explaining how discharged AgNP behave in natural aquatic systems as well as their environmental toxicological effects on naturally occurring microorganisms.
In Chapter 4, we discuss the investigation of the effect of silver nanoparticles on a model virus-MS2 bacteriophage. A negligible deactivation effect on MS2 phage was found regardless of the type of AgNP and the water chemistry conditions used.
In Chapter 5, we talk about the comparison of AgNP-impregnated point-of-use ceramic water filters and ceramic filters impregnated with silver nitrate. This study was performed using different water chemistry conditions and different manufacturing materials. The results showed that AgNP-impregnated ceramic water filters are more appropriate for this application due to the lesser amount of silver desorbed compared with silver nitrate-treated filters. The bacterial removal performance of the silvertreated ceramic filters and concentration of viable bacteria in the filters are dosedependent on the amount of silver applied. However, the data showed that influent water chemistry conditions did not have a significant effect on the performance of the filters. This study established evidence-based silver application guidelines for the ceramic water filter manufacturers around the world.
In Chapter 6, the discussion centers in the comparison of a polymer-based quaternary amine functionalized silsesquioxanes compound and AgNP. The results showed that the quaternary ammonium functionalized silsesquioxanes-treated ceramic water filter desorbed less from the filters and achieved higher bacteria removal than the filters impregnated with AgNP. This indicates that the quaternary ammonium functionalized silsesquioxanes compound could be considered as a substitute for silver nanoparticles due to its lower price and higher performance. However, more information regarding the possible chronic health effects of the silsesquioxanes compound is needed.
In Chapter 7, we present the main conclusions and recommended future work based on the dissertation results.       AgNP by reduction of the Ag(NH 3 ) 2 + complex with two monosaccharides, glucose and galactose, and two disaccharides, maltose and lactose, and found the average particle size ranged from 25 to 450 nm at various ammonia concentrations (0.005-0.2 M) and pH conditions (11.5-13.0) [4].
Aggregation during synthesis can hinder the production of AgNP with small and uniform sizes. For antimicrobial purposes, formation of aggregates can reduce the antimicrobial ability of AgNP [7][8][9]. Stabilizers are incorporated in the AgNP manufacturing process to ensure their stability in aqueous solutions. Absorption of the stabilizing molecules onto the nanoparticle surface depends on the molecular weight, ionization, and charge density of the stabilizing molecules [3,4,10,11]. Stabilizing layers can increase the electrostatic and steric repulsion between nanoparticles and therefore enhance the stability of the nanosuspension [3,12]. Commonly used stabilizing agents include different surfactants (such as sodium dodecyl sulfate (SDS) and Tween) and polymers including Polyvinylpyrrolidone (PVP) [13], Polyvinyl Achohol (PVA) [12], starch [8,14], and various proteins [9,15] .

Environmental fate and antimicrobial properties of AgNP in different water chemistry conditions
The use of AgNP in a wide variety of consumer products will inevitably lead to the release of the nanoparticles into natural water, which are the final receptacles [16].
Therefore, knowing the fate and reactivity of AgNP under environmentally relevant conditions is essential to prevent possible negative impacts on microorganisms commonly found in aquatic ecosystems.

Aggregation of AgNP in different water chemistry conditions
Previous studies have shown that different water chemistry conditions affect the toxicity of AgNP on microorganism communities [17].
It is widely accepted that the aggregation of AgNP follows the Derjaguin-Landau-Verwey-Overbeek (DLVO) theory [18][19][20], which combines the effects of the van der Waals attraction force and the electrostatic repulsion force created by the double layer of counterions [8,20,21]. In an aqueous solution, AgNP are electrostatically stabilized when the energy barrier is formed, because the electrostatic repulsion force is in excess of the van der Waals attraction force. AgNP are stable under this condition. However, when electrolytes are introduced into the system, the counterions in the aqueous solution neutralize the surface charges and disrupt the energy barrier, leading to aggregation [19,[21][22][23][24][25]. Figure 1-1 is an example of AgNP aggregation at increasing concentrations of NaCl [13]. When addition of the electrolyte solution results in complete removal of the energy barrier, fast aggregation occurs and the cluster size increases regardless of the electrolyte concentration [19,20,26]. ξ-potential is used to quantify the stability of the colloidal systems. Its value (negative or positive) indicates the degree of repulsion between charged particles in a nanosuspension. For AgNP suspension, high ξ-potential indicates the AgNP are electrically stabilized, while AgNP with low ξ-potential tend to aggregate. AgNP, in aerobic aqueous systems, carry negative charges because Ag(OH) 2 species at the surface can be formed due to the oxidation of metallic silver in the presence of O 2 in an aqueous solution [9,19,20]. In an aqueous solution containing ions, adsorption of anions can impart negative surface charges. More negative ξ-potential values indicate stronger electrostatic repulsion force between nanoparticles. Previous studies have also examined the effects of pH on the ξ-potential of AgNP. Li et al. [20] found that increasing pH (from 4-10) can better help stabilize the uncoated nanoparticles. Other studies [22] found similar trends using citrate coated AgNP across a wide pH range (2)(3)(4)(5)(6)(7)(8)(9)(10). As the silver atoms at the surface of AgNP are coordinately unsaturated, the OH -5 group can donate a pair of electrons. Therefore, when pH increases from 2-10, the concentration of OHincreases, thus allowing the OHto more effectively compete for surface sits, which generates a negative surface charge in alkaline pH conditions [22].
As the commonly used AgNP are negatively charged, increasing electrolyte concentration can increase the neutralization by the cations present in the electrolytes, which results in a decreasing ξ-potential (less stable) colloidal system [8].
Another mechanism of stabilization is steric repulsion [8,9,17,18,27]. Neutral organic coatings such as PVP and starch can also sterically prevent AgNP from aggregating [17,20,28]. Similarly, in water solutions containing natural organic matter (NOM), the NOM can absorb onto the AgNP surface, creating a physical barrier that hinders the contact between nanoparticles, and thus sterically stabilizing the nanoparticles.
Numerous studies have focused on the aggregation behavior of AgNP and have been summarized in Table 1-1.

Antimicrobial property of AgNP in different water chemistry conditions
Antimicrobial ability is a well-known property of AgNP. AgNP can inactivate a   however, sulfide appeared to be more effective to reduce AgNP toxicity by 80%.
Natural organic matter has a negative impact of the antimicrobial performance of AgNP. Studies have shown that the organic matter can adsorb on the surface of AgNP and reduce the physical contact between AgNP and bacterial cells [8,9,17,51]. In addition, the adsorption of organic matters can also inhibit their dissolution, resulting in a decreasing antimicrobial property [28].
The abovementioned studies focused on investigating the isolated water compositions, such as individual electrolytes or humic substances. However, more comprehensive studies on the antimicrobial properties of AgNP in natural water conditions are needed for their environmental risk evaluations.  .

Figure 1-4 Bacteria trapped in CWF impregnated with AgNP or Ag +
Numerous studies have investigated the pathogen removal performance of silverimpregnated CWFs. Table 1-2 summarizes these studies, including the types of silver and pathogens as well as the removal performances.   The increase in the price of silver is threatening the sustainability of CWFs.
Therefore, alternative disinfectants are needed to ensure the antimicrobial efficacy of the CWF system. One promising disinfectant agent candidate is 3-(trihydroxysilyl) propyldimethyloctadecyl ammonium chloride (TPA), which is a quaternary amine functionalized silsesquioxane compound. Getman

Dissertation objectives
Evaluating the reactivity of AgNP is essential to support their application in water treatment technologies. Reactivity of nanoparticles is also important in order to address their environmental impacts once released in natural systems. Therefore, considering the gaps in the literature, the following objectives were developed:  Determine the effect of natural water composition on the bactericidal and antiviral activity of AgNP stabilized with different coatings. This goal is important, since few studies have focused on studying their antimicrobial properties (especially their antiviral properties) [17] and physicochemical properties [17,72] in natural water conditions. This study will provide new evidence of how natural waters affect the abovementioned properties of AgNP, which is helpful in explaining their environmental fate.
 Determine the performance of silver nanoparticles applied to ceramic water filters used for water treatment in developing communities. This goal is  of silver nanoparticles in different water chemistry conditions. J. Environ.  lead to a correlation between survival rate of of E. coli and average size of AgNP. We found a strong correlation between this two parameters tha can be fitted to a saturation type curve, reaching a survival plateau around 20% survival at an average particle size of 200 nm for all the water chemistry conditions tested.

Introduction
AgNP are commonly used in a wide range of applications, including solar energy absorption and chemical catalysis and disinfection [1][2][3][4][5]. Consumer products containing AgNP accounted for more than 25% of the 1,015 nanotechnology-based consumer products available on the market in 2009 [6]. AgNP have large surface areas per volume ratio and high reactivity compared with the bulk solid. This feature gives AgNP antimicrobial properties.
Three possible antimicrobial mechanisms of AgNP have been raised: (1) AgNP can damage cell membrane and intracellular components [7,8], (2) silver ions released from AgNP can be sorbed into the cell wall and cause lysis and death [7,9,10], and (3) reactive oxygen species (ROS) can be formed in AgNP solution [11][12][13][14]. While the antibacterial properties of AgNP have been extensively demonstrated [15][16][17][18],their performance at different water chemistries have not been fully understood yet. Some evidence shows that the disinfection effectiveness of AgNP is size dependent [19], and that the process of aggregation reduces their surface area, reducing the cell-particle interaction, membrane penetration, and the rate of silver ion release [6].
Studies measuring the rate of silver ions release at different dilutions of seawater showed that the salt concentration did not affect the AgNP' oxidation kinetics; however, high ionic strength increased the size of the particles from 1.9 to 200 nm after 24 hours [6]. Gao et al. [20] also found that fresh water samples with higher ionic strength produced large AgNP [20]. Jin et al. [21] studied the effect of different water matrices on the AgNP size, silver ions release, and antimicrobial activity using a fixed concentration of Ca 2+ and Mg 2+ . The study revealed that Ca 2+ and Mg 2+ increased the AgNP aggregation in different electrolyte solutions with the same ionic strength in comparison with mono-valent ions. The antimicrobial test showed that Gram-negative bacteria Pseudomonas putida was more resistant to AgNP compared to Gram-positive bacteria Bacillus subtilis [21].
In this work, we used seven different electrolyte solutions to study systematically the influence of different cations and anions on physico-chemical characteristic of the particles and disinfection performance. One objective of this study was to establish a correlation between survival rate of bacteria and particle characteristics that could easily predict the disinfection performance of AgNP at different water chemistry conditions.

Synthetic water solutions
Synthetic water solutions were prepared using eight different solutions, four mono-valent and three divalent salts in addition to one solution containing humic acids  anions on the disinfection performance and average particle size of AgNP was evaluated using a statistical test using a general linear model in PASW SPSS 18.0. All salts and other reagents were ACS reagent grade and used as received.

Preparation and characterization of AgNP
AgNP (70.37% w/w Ag 0 ) stabilized with casein were obtained from Argenol laboratories. The nanoparticles are proposed to bind to casein polymers surface via complexation with the carboxylate or amino group of casein [24]. A fresh AgNP stock 33 solution of 4 mM was prepared immediately before testing using deinozed (DI) water and the respective electrolyte as presented in Table A

Microbial cultures
A non-pathogenic wild strain of E.coli provided by IDEXX laboratories was used for bacteria transport experiment through ceramic filter [26]. This organism was selected because of its use as specific indicator of fecal contamination in drinking water and its extensive use in several studies on AgNP, which will allow us to compare our results with previously published work. Bacteria were grown as described by Vigeant et al. [27]. Cells were re-suspended in a sterilized solution prepared with the respective electrolyte solution (Table A- : Where P c is the pressure drop after the injection of glucose, P t is the OUR after the injection of AgNP, P is the OUR during the endogenous respiration without the addition of carbon source or nanoparticles. Figure

Data analysis
The data acquired were fitted using the Origin 8 software, minimizing the

Results
The shape and sizes of AgNP in DI water determined by TEM are presented in   [20] showed that the toxicity of AgNP of bacteria and microinvertebrate decreased with increasing concentration of natural organic matter as total organic carbon (10 -4 -10 mg/L) [20].

Discussion
Our results showed that AgNP exhibited high antimicrobial activity across several water chemistry conditions (20% survival and below).
Where y is the survival rate of E. coli; x is the particle size of AgNP (nm); y max is the maximum survival rate of E. coli (%); k m is the particle size at which y is half of y max (nm).
46 The null hypothesis proposed that different ions have no significant effect on the AgNP particle size or survival rate of E.coli. Any p value smaller than 0.05 implies that the null hypothesis is false that different ions have significantly different effect on particle size and/or survival rate. The results of the statistical analysis indicated that anions can affect the particle size and survival rate if there are paired with a monovalent cation. When divalent cations are present, anions showed no significant effect on the particle size and survival rate, indicating the divalent cations are the dominant influencing factor.  The antimicrobial performance of AgNP has been related to the collision frequency between bacteria and AgNP [32]. Other studies suggested that the antimicrobial performance is size-dependent [16,29,33]. Smaller particles can more easily attach to and penetrate the cell membrane compared to larger particles, damaging the cell membrane by impairing its permeability and respiration [8,19,29,34]. Morones et al. [29] indicated that AgNP with a diameter smaller than 10 nm were able to interact with E. coli and Pseudomonas aeruginosa [29]. However, Xu et al.
[35] observed that AgNP with a particle size up to 80 nm could also accumulate inside Pseudomonas aeruginosa cells [35]. In this study, we proposed that the disinfection performance of AgNP is size-dependent as shown in Figure 2 can be fitted to a saturation type curve, reaching a survival plateau around 20% survival at average particle size of 200 nm. However, it should be noticed that low survival rates are achieved (about 20%) even at large particle sizes (200nm < X<1000nm). This is could be because the presence of Ca 2+ and Mg 2+ could promote the formation of ion bridges between the negatively charged AgNP and the lipopolysaccharide molecules on the cell membrane. This interaction could decrease the electrostatic repulsion and enhance the aggregation of AgNP with E. coli [21]. The AgNP-E. coli aggregates can impair the permeability of the cell membrane, causing membrane leakage and cell death. In addition, particle size distribution showed that the sizes of AgNP in divalent cationic solutions were monodispersed, with few small nanoparticles capable to penetrate the cell membrane. Therefore, our results showed that the AgNP-E. coli interaction played a major role in the antimicrobial activity in divalent cationic solutions.
In our study, the levels of silver ions detected did not significantly contribute to the overall disinfection performance. Low concentration of silver ions (5.9-18.8 ppb) and high concentration of E. coli (10 9 CFU/ml) could explain the observed phenomena. Other studies, such as Choi et al. [2] reported 100% inhibition of E. coli PHL628-gfp with concentration of silver ions of 0.5 mg/L. Jin et al. [21] reported an IC 50 of 0.1 mg/L to 50 mg/L for Gram-negative bacteria Pseudomonas putida (initial concentration: 10 8 cells/ml) [21]. Suresh et al.   This study demonstrated that the anti-bacterial performance of AgNP at selected natural water conditions decreases in the presence of dissolved natural organic matter or divalent ions, such as humic acid and calcium carbonate. These results may be helpful in understanding the toxicity of AgNP in various natural water conditions and in explaining the risk associated with discharging AgNP in natural aquatic systems.

Introduction
Nanotechnology is an emerging and fast-developing technology. There are more than 1,300 nanotechnology based consumer products according to the Woodrow Wilson center in 2011 [1]. Among these nanoproducts, silver nanoparticles (AgNP) accounted for more than 23% of total applications of nanomaterials.
AgNP are of great interest because of their chemical catalytic ability, stability and antimicrobial properties [2,3]. AgNP are introduced into consumer products, it is unavoidable that AgNP will be released into the natural water bodies [4]. Therefore, the fate and reactivity of these nanoparticles at different water chemistry conditions is essential to elucidate their impacts in microorganism commonly found in different water ecosystems.
AgNP can be synthesized using physical or chemical methods. Sharma et al.
introduced the Tollens method, which involves the reduction of Ag(NH 3 ) 2 + in the aqueous phase by an aldehyde (usually saccharides). Different stabilizing agents, such as sodium citrate, some surfactant and polymers are used to create a stable dispersion of AgNP in liquid solutions [2].
Critical coagulation concentration (CCC) is defined as the minimum amount of electrolyte that is required to completely destabilize a suspension. This parameter is frequently used to measure the stability of nanosuspensions since it quantify the minimum amount of electrolyte solutions that is required to destabilize the nanoparticle suspension [5]. Compared to -naked‖ AgNP, AgNP coated with 59 polymers have shown higher stability [5,6]. Huynh and Chen, [5] found that the CCC value of AgNP coated with PVP is more than two folds and four folds higher than the value obtained for AgNP coated with sodium citrate [5] and -naked‖ AgNP, respectively [6]. While many researchers have manufactured and used AgNP stabilized with different compounds, only few research studies have compared their disinfection performance and stability (in terms of particle size and ξ-potential) using natural water conditions [7,8].
For the purposes of this work, we selected casein, dextrin, and PVP-stabilized AgNP. Commercially available AgNP coated with casein was selected because casein because of their extensive application in point-of-use ceramic water filters [9,10], therefore their performance and stability at natural water conditions is of great interest.
PVP coating was selected because they are considered to be environmentally-friendly stabilizers and because other researchers have used them to conduct AgNP characterization and anti-microbial activity tests [2,11]. Few studies have synthesized silver nanoparticles using dextrin as a coating agent and studied the antibacterial activity and physicochemical properties of the coated nanoparticles [12]. In this study, polyvinylpyrrolidone, (PVP, average molecular weight: 29,000 g/mol, Sigma Aldrich) (AgNP-PVP) were prepared via the Tollens method described by Kvitek et al. [15] with minor modifications, i.e., 0.35 wt% for both dextrin and PVP were added as stabilizers, respectively [15].

Characterization of AgNP
The shape and sizes of AgNP in DI water were determined by TEM. All three AgNP were spherical, which has been reported in other studies using the same manufacturing methods [15]. AgNP aggregates were observed in all TEM images.
Average particle sizes of AgNP-casein, AgNP-dextrin, and AgNP-PVP measured using DLS were larger than those determined by TEM (Table B-2). The discrepancy is likely due to the formation of AgNP aggregates, as shown in the TEM images.In addition, DLS measurements are volume-squared weighted distributions. Compared to TEM size measurements, they can be influenced by nanoparticle aggregates in different water chemistries. Table B [5,6]. However, Li et al. [6] using fulvic acid did not observe this trend.

Figures 3-2 and 3-3 and
The discrepancy could be due to the fact that fulvic acid molecules are smaller than humic acid molecules [6].
Our study showed that the CCC values of AgNP with different coatings in all water conditions followed the same trend i.e., AgNP-PVP>AgNP-casein>AgNPdextrin. It is proposed that the binding force of different atoms in the stabilizers on the surface of the AgNP plays an important role in the stability of AgNP [8]. Nitrogen atoms in PVP are more strongly bonded on the surface of AgNP in comparison with oxygen atoms in dextrin [8]. Casein includes both strong nitrogen atom bonds and weak oxygen atom bonds onto the nanoparticles [16]. Stronger binding forces could ensure a stronger attachment of the stabilizers on the surfaces of the AgNP, precluding substitution by other ions or polymers [15].
In addition, a steric repulsion effect can also influence the aggregation behavior.
Long polymer chains may form complex steric configurations, thus increasing the repulsion force between nanoparticles [17,18]. In our study, the length of the stabilizer chains were in the following order: PVP (MW polymer: 29,000; MW of monomer: 111) > dextrin (MW polymer: 1,670; MW monomer: 162). Therefore, the length of the stabilized chain could have an important effect on the stability of the AgNP for all of the natural and synthetic water conditions tested. Casein has a complicated molecular formula, and it is likely that the intricate steric configurations and the electrostatic effect of the casein proteins contributed to its stabilizing effect [19].

Silver ion release in different water conditions
The release of silver ions (Figure 3-5) measured during the tests was less than 0.5% of the total mass of the silver added as AgNP (3.6 -48.2 ppb). Discrepancies in the silver ion release have been found in previous studies. Liu and Hurt, [20] reported that the silver ion release of AgNP (initial concentration of 0.05 mg/L) in seawater 69 (ionic strength: 0.7 M) and deionized (DI) water after 24 h was 20 wt% and 50 wt%, respectively [20]. However, percentages below 1% also have been obtained under conditions similar to those used in this study [7,21]. In this study, the silver ion release was sensitive to different water chemistries. It was previously reported that Clmay co-precipitate Ag + by forming AgCl, and NOM could coat the AgNP surface rapidly and inhibit AgNP dissociation [7,21]. This could explain why silver ion release in water conditions with high Cland NOM content is small. The lowest silver ion release was found in the humic acid solution. Since the TOC concentration in the humic acid solution was similar to the TOC concentrations found in natural water, the low ion release in humic acid solution may be attributed to the difference between humic acid and other NOM species in natural waters. To determine the contribution of the released silver ions to the overall disinfection performance of the AgNP, several AgNO 3 solutions were prepared, ranging from 1 to 50 ppb (as Ag + ). The survival rate of E coli was determined by the manometric respirometric method mentioned earlier.
At all the concentrations tested, no significant disinfection efficacy was observed ( Figure A2). Similarly, Suresh et al. [21] indicated that a concentration of 0.48 mg Ag + /L released from AgNP (initial concentration: 100 mg/L) had no significant effect on E. coli when he used an initial concentration of 10 5 CFU/mL [21]. Fabraga et al. [22] indicated that 2-20 ppb of silver ion showed negligible antimicrobial property.
The results we obtained likely can be attributed to the low silver ion concentrations at these water conditions and the high E. coli concentration used (10 10 CFU/mL) [22].
The fact that silver ion concentrations were low and that they made a negligible contribution to the overall anti-microbial effect do not imply that silver release did not 70 occur. Most likely, due to the high concentration of Cl -, silver released from the nanoparticles formed AgCl, which was precipitated. Additionally, although some previous studies reported high bactericidal properties for silver ions [13,15], however, microbial growth-based methods were use the antibacterial properties of the AgNP.
The growth-based methods measure the bacteriostatic effect of the nanoparticles, while the respirometric respiration method (used in this study) measures the total deactivation of bacteria. Therefore, it is likely that at similar concentration of AgNP, respirometry based methods will show a lower activity than growth-based methods since the bacteria could have lost their capacity to replicate, while maintaining some (or full) metabolic activity.

Effect of different water conditions on the disinfection performance of AgNP
Prior to assessing the disinfection performance of AgNP, the toxicities of the three stabilizers were measured using the same respirometric method. The results indicated that no toxicity was associated with the three stabilizers (data not shown). water conditions. For natural water conditions, we found a similar disinfection behavior of AgNP in surface and brackish water. This likely occurred because of the presence of NOM that could be adsorbed on the surface of AgNP, reducing their toxicity by creating physical barriers between the nanoparticles and E. coli [8].
Anionic ligands, such as Cl -, also could impair the toxicity of the AgNP by reacting with their surfaces to form AgCl, thus reducing their toxicity. Different from surface and brackish water, AgNP in ground water had greater disinfection performance, which likely was due to the lower NOM content. It is noteworthy that the disinfection performance of AgNP in seawater was the lowest among all the water conditions. In addition to the above-mentioned factors that reduce the toxicity of AgNP, the formation of the large aggregates that were observed in seawater (Figure 3-2) may impair their toxicity.
In synthetic electrolyte solutions, the order of the Clcontent in NaCl, CaCl 2 , and MgCl 2 solution was MgCl 2 >CaCl 2 >NaCl. The toxicity of AgNP should follow the order of MgCl 2 <CaCl 2 <NaCl since Clcontent could plays a role in reducing the toxicity (due to silver precipitation). However, the order of measured toxicity of AgNP in the electrolyte solutions was NaCl>CaCl 2 ≈ MgCl 2 . This is likely because the bactericidal property of the AgNP could be size-dependent ( Figure B-3). It is noteworthy that the sizes of the AgNP aggregates in the CaCl 2 and MgCl 2 solutions were much larger than that in the NaCl solution (Figure 3-2). This result agrees with previously published data, suggesting that the toxicity of AgNP is size-dependent [14,23,24]. Smaller particles can attach to and penetrate cell membranes more easily than 73 larger particles and therefore damage the cell membrane by impairing its permeability and interfering the respiration process [14,23,24]. In this study, HA solution exhibited lower toxicity than other synthetic electrolyte solutions despite the fact that the sizes of the AgNP in the HA solution were similar to those in the NaCl solution.
Previous studies have proposed that the low toxicity is due to the rapid coating of the surface of AgNP by humic acid, which creates a physical barrier that impedes AgNPbacteria interactions and thus reduces their toxicity [8,22].
74 The statistical analysis showed that Clconcentration, divalent cation concentration, NOM concentration and silver ion release are significantly correlated with the antibacterial activity of AgNP (Table B- 4). We propose that a combination of the above-mentioned factors is responsible for the disinfection performance of AgNP. Table B-4 and Table B-5 show the correlation analysis for all the conditions tested and water conditions without NOM, respectively. When NOM is included in the analysis no correlation between size and antibacterial activity was obtained.
However, when NOM-containing water conditions were removed from the analysis, the bactericidal effect of AgNP is significantly correlated with particle size (Table B-5).

Conclusions
Several conclusions can be drawn from the results of this study, i.e., (i) the antimicrobial performance of AgNP is mainly particle size-dependent in absence of NOM,

Introduction
Silver nanoparticles (AgNP) have a wide range of applications especially as an antimicrobial agent. It is estimated that AgNP based consumer products account for more than 23% of the total nanomaterials impregnated products. Due to their extensive application in industry and households, it is inevitable that AgNP will be released into the natural aquatic systems and interact with the local microorganisms [1].
Previous studies have evaluated the toxicity of AgNP towards different microorganisms in aquatic system. Gao et al. [2] have reported that the LC 50 (µg/L) of AgNP against Escherichia coli (E. coli) and Ceriodaphnia dubia (C. dubia) are less than 112.14 and 6.18 µg/L, respectively. Zhang et al. [3] showed that the antibacterial performance of AgNP (11.2 mg/L) ranges from 67.9-81.8% E. coli inhibition in surface, ground, brackish, and sea water. Studies on toxicity of AgNP against other microorganisms such as Staphylococcus aureus [4], Leuconostoc mesenteroides [5], Bacillus subtilis [6], Pseudomonas aeruginosa [7] were also reported. However, little is known about the antiviral effect of AgNP. A recent study reported that AgNP Three possible mechanism of antiviral effect have been proposed. Studies have reported that the silver ion dissolved from AgNP can cause toxicity by interacting with viral DNA or RNA and the thiol groups in capsule proteins [9] [10]. Lara et al. [9] proposed that AgNP can inhibit the virus binding onto host cells by direct interacting with viral particles with similar sizes. Liga et al. [10] showed virus can be inactivated photocatalytically by AgNP doped TiO 2 nanoparticles due to the generation of reactive oxygen species (ROS).
The mechanisms proposed have associated the antiviral property of AgNP with their physicochemical properties such as particle size and dissolution of AgNP. These physicochemical properties vary in different water chemistry conditions. For example, it has been widely reported that AgNP tend to aggregate in aqueous solutions containing high concentration of divalent cations such as Ca 2+ and Mg 2+ [11][12][13][14][15]. Ionic strength and the natural organic matter content in water can also alter the dissolution behavior of AgNP [3,11,16,17].

83
As viruses are responsible for a wide spectrum of diseases in bacteria, plants, and animals and they play important role in aquatic food webs as active constituents of the microbial loop [18], it is important to investigate the antiviral effect of discharged AgNP in different water conditions to elucidate their environmental risks. However, to the author's knowledge, few studies have focused on this issue [8].
In this study, we selected sodium citrate and PVP stabilized AgNP because these two stabilizers are environmentally friendly and have been used in previous research

Synthetic water solutions preparation
Synthetic water solutions were prepared using NaCl, CaCl 2 , and HA with cations concentrations ranging from 10-10,000 mg/L (ionic strength for NaCl: 0.44-440 mM; ionic strength for CaCl 2 : 0.75-750 mM) and humic acid ranging from 1-10 mg/L as total organic carbon (TOC). Natural water samples were collected from in Rhode Island from Thirty Acre Pond (surface water) and Narragansett Bay (seawater).
Ground water was collected in Kingston campus, University of Rhode Island.

AgNP preparation and characterization
AgNP stabilized with sodium citrate (Sigma Aldrich) (AgNP-Cit) and PVP (average molecular weight: 29,000 g/mol, Sigma Aldrich) (AgNP-PVP) were prepared via the Tollens method described by Kvitek et al. [20] with minor modifications. 0.035 w/w% for both sodium citrate and PVP were added respectively. The obtained AgNP 85 suspensions were cleaned using DI water by ultrafiltration using a 10k kDa membrane (Whatman) in an ultrafiltration cell (Millipore). Concentrations of the AgNP were measured as total silver using ICP-AES (Thermo Elemental).

Bacteriophage inactivation by AgNP
AgNP were added into the synthetic solutions and natural water samples. Then, 100 µl of MS2 stock solution was added into the suspension to obtain the final silver concentration of 1 mg/L. After 2-h exposure to AgNP in dark condition at room temperature, the AgNP were exposed to 400 mg/L of sodium thiosulfate for 2 min to neutralize any AgNP or silver ions remaining in solution. MS2 in these water samples was served as the controls. The phage samples were serially diluted in 1 mM NaCl solution during DAL assay. The plates were incubated at 37 o C overnight and only samples with the number of PFU from 0 to 300 were enumerated. Silver nitrate was used to compare the antiviral activity between AgNP and silver ions.

Particle size, aggregation kinetics, and zeta potential measurement
Particle size, aggregation kinetics, and zeta potential were measured for both

Silver ion release experiment
Ion release from AgNP samples was quantified by ultrafiltration using a 3 kDa molecular weight cut-off membrane as described previously [22]. Nitric acid was added to the percolate (HNO 3 concentration: 2%). ICP-AES was used to quantify the total silver concentration.   , which is responsible for the increased dissolved silver concentration. Our results agree with previous literatures. Ken and Vikesland. [1] reported the dissolved silver concentration increased with increasing NaCl concentration (10-550 mM) for after approximately 2 weeks experiment. However, dissolved silver was only detected in CaCl 2 solution at the highest electrolyte concentration (10 4 mg/L Ca 2+ ). It is proposed that the AgNP dissolution in CaCl 2 could be inhibited due to the formation of AgNP aggregates, which could diminish the exposed surface area of AgNP and hinder the mass transport of reactants to active sites. Soluble silver chloride complex can still be formed in the presence of excessive Cl -.

Dissolution of AgNP in different water conditions
No dissolved silver was observed in presence of HA. Previous studies suggested that natural organic compound such as HA could inhibit AgNP dissolution by several mechanisms including surface adsorption to block AgNP oxidation sites, reversible reaction of released Ag + to Ag 0 with HA as a reductant [2][3][4][5]. In natural water samples, less AgNP dissolution was found in seawater compared with surface and ground water. This could be due to the combining effects of the two abovementioned dissolution inhibition mechanisms: aggregation and presence of natural organic compounds. Similarly, Liu and Hurt. [2] compared the dissolution kinetics of AgNP in DI, low salt seawater buffer, and seawater. It was found that the silver release is much slower in seawater than in DI water. reported [3,[6][7][8][9][10][11]. Previous published work suggested that these environmental factors affect the physicochemical properties of AgNP which is closely associated to their toxicity [9]. Ionic strength and presence of divalent or multivalent cations result in aggregation of AgNP. Large aggregates exhibited lower toxicity compared with monodispersed nanoparticles [3,[11][12][13]. Specific ions such as Cland SO 4 2presence in water can form complexes with released silver ions. The complexes are usually less toxic than Ag + [7,8]. Natural organic compounds can also reduce toxicity of AgNP by inhibiting AgNP dissolution and adsorbing on the nanoparticle surface to create physical barriers between nanoparticles and microorganism [2,6]. The combination of these environmental factors diminishes the antiviral performance of AgNP. In addition, compared to other microorganisms such as bacteria, MS2 bacteriophage is more resistant to AgNP [14]. Our result is consistent with previous studies. You et al. [14] reported that AgNP (coated with polyvinyl alcohol) failed to inactivate MS2 bacteriophage at the highest concentration (5 mg/L Ag) tested, which is related to the availability of dissolved Ag + and their preferable binding to virus structures. Other study found that 5 mg/L uncoated and polysaccharide coated AgNP (particle size 10 nm) had no effect on virus propagation [15]. Interestingly, it is also noteworthy that a slight decrease in virus survival was observed in aqueous solutions containing high concentration of Ca 2+ . As zeta potential deceases with increasing ionic strength, 98 divalent cations enhanced the interaction between nanoparticles and MS2 phages by forming aggregates, which may result in a slight increase in the antiviral activity of AgNP [10]. Previous studies have reported similar effect on bacteria [10,16]. In addition, AgNP can facilitate MS2 phage to infect the bacteria host by changing bacteria cell membrane structures and permeability [14]. However, aggregation of AgNP can reduce the membrane disruption effect [3], which may consequently interfere with the ability of the virus to infect the E. coli bacteria resulting in less plaque forming units found in the agar plates.

Abstract
Locally produced ceramic water filters (CWF) are an effective technology to treat pathogen-contaminated drinking water at the household level. CWF manufacturers apply silver to filters during production; although the silver type and concentration vary and evidence-based silver application guidelines have not been established. We evaluated the effects of three concentrations of two silver species on effluent silver concentration, E. coli removal, and biofilm formation inside ceramic disks manufactured with clay imported from three CWF factories using sawdust as the burn-out material. Additionally, we evaluated performance using water with three chemistry characteristics on disks made from the clays using either sawdust or rice husk as burn-out material. Results showed: 1) desorption of silver nitrate (Ag + ) was higher than desorption of silver nanoparticle (AgNP) for all disks; 2) effluent concentration, E. coli removal, and biofilm formation inside the disks were dosedependent on the amount of silver applied; and, 3) neither water chemistry conditions nor burn-out material showed an effect on any of the parameters evaluated at the silver concentration tested. Recommendations for filter manufacturers to use only silver nanoparticles at a higher concentration than currently recommended are discussed.

Introduction
Worldwide, an estimated 783 million people do not have access to an improved water source [1] and hundreds of millions more drink water that is contaminated at the source or during collection, transport or storage [2]. Drinking water contaminated by pathogenic microorganisms causes gastrointestinal infections, which account for 1.87 million childhood deaths each year, mostly in developing countries [3]. Potters for Peace (PfP) style ceramic water filters (CWF) are a low-cost technology produced locally in developing countries by pressing a mixture of clay and an organic (burn-out) material into the filter shape and then firing it to a ceramic state. Combustion of the burn-out material during the firing process creates the filter structure. CWFs remove pathogens from water by retaining them on the surface or trapping them within the filters pores.
CWFs are effective at removing more than 99% of protozoan [4,5]and 90-99.99% of bacterial organisms from drinking water [6],however, the removal of viruses remains a challenge. In the field, water treated by CWFs is often improved to the World Health Organization's (WHO) low-risk standard [7] of fewer than 10 CFU (colony forming unit) E. coli /100 mL [6], and filter use has been associated with a reduction in diarrheal disease among users [8].
Silver nanoparticles (AgNP) and silver nitrate (AgNO 3 , Ag + ) are known antimicrobial agents, and are added to filters, mostly after the firing process [9]. Reported log reduction values (LRVs) of E. coli by CWFs coated with AgNP range from 2.5 to 4.56 [9,10]. LRVs of 2.1 to 2.4 of E. coli have been measured using filters coated with Ag + [6]; however, in the same study similar LRVs were also measured in CWFs without Ag + application [6]. In production, 83% of factories use AgNP and 17% use Ag + [11]. Factories use Ag + because it is cheaper than AgNP and/or it is locally available. The concentration of silver applied at each factory varies. Reported AgNP concentrations range from 107 to 288 ppm [11], excluding probable outliers. The silver solution is applied to fired filters by brushing or dipping. When silver solution is applied by brushing, factories reported applying from 32 to 96 mg of AgNP per filter.

107
The current guideline, which is experiential rather than evidence based, is 64 mg of AgNP per filter [12].
A variety of water sources are used at factories to prepare silver solutions, from untreated surface water to treated water [11].Water characteristics at the filter user's home also vary with location. Previous studies have reported a reduction in antibacterial properties of AgNP with increased size of the nanoparticle clusters due to aggregation in the presence of divalent ions such as Ca 2+ and Mg 2+ [13,14]. In addition, water can contain organic compounds, such as humic acids (HA). These can rapidly coat the nanoparticle surfaces, creating a physical barrier that prevents interaction between nanoparticles and bacteria [13][14][15].While previous studies have reported that different water chemistry conditions can impact the disinfection performance of AgNP in the aqueous phase, these parameters have not been evaluated on CWFs either in the field or in laboratory tests.
Desorption of silver from coated CWFs has been reported during the first flushes of water [9]. A study using phosphate buffer as influent solution reported a decrease in silver concentration (as total silver) in effluent from AgNP-impregnated CWFs to below the United States Environmental Protection Agency (USEPA) maximum contaminant level (MCL) for silver [16] in drinking water (0.1 mg/L or 100 ppb) within few flushes [9]. To our knowledge, no comprehensive study has evaluated the desorption of either AgNP or Ag + from CWFs using different pottery clays and water chemistry conditions.
In this study, we evaluated the performance of ceramic disks manufactured with clays from three different factories and two types of burn-out material, sawdust and rice husks. In Phase I, disks manufactured with the different clays and sawdust were coated with three different concentrations of either AgNP or Ag + and evaluated for: 1) effluent silver concentration and silver retention; 2) E. coli removal; and, 3) biofilm formation. In Phase II, the influence of three water chemistries (Na + -NaCl, Ca 2+ -CaCl 2 , and humic acid as natural organic matter) on AgNP and Ag + were evaluated on disks manufactured with each of the clays and each of the burn-out materials against the same outcome parameters.

Disk manufacturing and pretreatment
While PFP-style filters are a 10-liter capacity filter pot, in this study 10-cm diameter disks were manufactured to simplify transport and testing. Disks were manufactured at Advanced Ceramics Manufacturing (Tucson, AZ) with clay imported from filter factories in Indonesia (Indo), Tanzania (Tanz) and Nicaragua (Nica). The burn-out material, processed between U.S. sieve numbers 16 and 30 (1.19-mm and 0.595-mm openings, respectively), comprised 15% (burn-out:clay ratio by weight) of the filter mixture. Disks were pressed at 3.58 PSI and air-dried. The different clays required different firing temperatures in order to achieve sufficient strength for testing.
Fired disk thickness was approximately 1.5 cm. At the manufacturing facility, disks were boiled in water for one hour and the percent porosity of each disk was calculated by dividing the difference between saturated weight and dry weight by the geometric disk volume.
Disks were then shipped to the University of Rhode Island (URI) where they were cut to 3.8 cm diameter to fit existing filter holders. To eliminate any possible microbiological contamination, disks were heat treated to 550°C for 30 min, then allowed to cool at room temperature. The sides of the disks were sealed with silicone, allowed to dry, and then sealed in the filter holders with silicone.

Disk characterization
Tracer experiments were conducted to determine the intrinsic characteristics of the disks and to identify possible anomalies such as preferential channels in the porous matrix. Tracer tests and the subsequent determination of the advection and dispersion coefficients were performed using the procedure described in Oyanedel-Craver et al.
[9] but with NaCl instead of tritiated water.

Silver release and retention
Suspended silver nanoparticles (AgNP) (coated with casein; 70 w/w% AgNP) was purchased from Laboratorios Argenol and dissolved silver (Ag + ) from Sigma Aldrich. Silver was applied by brushing each disk with a specific concentration of either Ag + or AgNP in the appropriate electrolyte solution according to the procedure described in Oyanedel-Craver et al. [9]. Disks were then flushed with a bacteria-free solution for 24 hours. Effluent silver concentration was measured after preserving samples by adding 2% nitric acid, using inductively coupled plasma atomic emission spectroscopy (ICP-AES) after 100 minutes, 200 minutes, 300 minutes, and at 24 hours. Percent retention of silver (R) was calculated by dividing the difference between the initial mass of silver added (m 0 ) and the total mass of silver released over the 24 hours period (m t ), by the initial amount of silver added to the disk:

Bacterial removal performance
The two phases of microbiological removal testing included: 1) the determination of the optimal amount of silver required to achieve maximum bacterial deactivation (only on sawdust disks); and, 2) evaluation of the bacterial removal efficacy under different influent water chemistry conditions (using both sawdust and rice-husk disks) using current recommended silver concentration [12] (Table 5-1).
Phase I and Phase II tests were conducted in duplicate, using two disks of each clay. In Phase I, only disks manufactured with sawdust were tested due to not having sufficient disks with rice husks for both phases of the study. In Phase II, disks manufactured with sawdust or rice husk and each of the clays were tested. 150 mg/L Na + -NaCl 0.003 150 mg/L Ca 2+ -NaCl 0.003 5 mg/L humic acid as total organic carbon After 24 hours of flushing with a bacteria free solution, a concentration of 10 6 CFU/mL E. coli in water of the same chemical composition as used during the flushing stage (i.e., deionized water with a buffer solution, electrolytes, or humic acid) was prepared and continuously fed to the disks at a flow rate of 0.5 mL/min using a peristaltic pump. The concentration of bacteria in the influent and effluent were measured using the method described by Vigeant et al. [17]. Samples (10 mL) were taken daily for 10 days, and LRVs were calculated. For Phase I testing, a phosphate buffer solution was selected to minimize natural decay of bacteria during the test period. A fresh solution of bacteria was prepared daily for Phase I and Phase II testing.
At a feed of 0.5 mL/min for 10 days, the total throughput for each disk over the study period was ~7.2 L, which equates to ~1300 L through a full-sized filter. This was calculated by multiplying the flow rate per cm 2 of the filter disk by the area of a full-sized Nicaraguan filter using filter dimensions presented in van Halem. [5] Using this calculation, the test period simulated approximately four months of a filter treating 10 L of water per day.

Bacteria retention
After completing the bacterial removal tests, the concentration of biofilm (viable bacteria) contained in the pores of the disks was determined. The disks were ground and 10 grams were transferred to a 50-ml flask. The bacteria were dispersed in the buffer solution by gentle sonication (20% amplitude) for 15 minutes to detach the bacteria from the ceramic material. The concentration of bacteria was determined using Vigeant et al. [17] as above.

Disks characterization
A total of 144 disks were tested, including 30 each of Indo-sawdust, Tanzsawdust, and Nica-sawdust and 18 each of Indo-rice husk, Tanz-rice husk, and Nicarice husk. The average advection (v) (directly proportional to the fluid velocity) and dispersion (D) (directly proportional to the effective porosity) coefficients and geometric porosity values for the ceramic disks manufactured from the same recipe were similar. Results from disks manufactured with Indonesian and Tanzanian clays were also similar; however, disks manufactured with the Nicaraguan clay had higher advection and dispersion coefficients, indicating that the solute spread fastest through the Nicaraguan disks. For each of the clay groups, disks manufactured with rice husk had slightly lower porosities than disks manufactured with sawdust. Values are presented in Table B-1.

Phase I Silver release and retention
For both types of silver and regardless of clay type, a higher concentration of silver was measured in effluent from disks coated with higher concentrations of silver ( Figure 5-1).
With the exception of 0.003 mg/g Ag + , for each silver concentration Ag + resulted in a higher effluent silver concentration in comparison with AgNP. Silver concentration in the effluent reduced with solution throughput regardless of silver type.
Effluent silver concentration from AgNP-coated disks was below the USEPA MCL after 24 hours in all but one case (disks made with Nicaraguan clay and impregnated with 0.3 mg/g AgNP) (Table C-2). Effluent concentration from disks impregnated with 0.3 mg/g Ag + exceeded the USEPA's MCL in all cases and ranged from 797 ppb to 2,697 ppb after 24 hours.
An increased concentration of silver resulted in increased silver retention in disks coated with AgNP regardless of clay type ( Figure 5-2). AgNP retention did not vary 113 widely between disks made with different clays. A greater percentage of AgNP was retained in disks in comparison with Ag + , most notably in disks made with Nicaraguan clay. An increase in Ag + concentration from 0.003 mg/g to 0.03 mg/g resulted in increased retention; however, the highest concentration of Ag + 0.3 mg/g resulted in the lowest percent retention.

Bacterial Removal Performance
In all samples, a sharp reduction in LRV was observed from day one to five ( Figure 5-3); however, the LRV leveled off from day five to ten. Thus, the LRV performance comparison is based on the average of the results from the last five days of testing.
Disks made with Indonesian and Tanzanian clay resulted in an increased LRV with increased silver concentration, regardless of species applied ( Figure 5-3). No change in terms of LRV was measured from Nicaraguan disks regardless of silver species or concentration applied. LRV was comparable between disks coated with AgNP and Ag + , with the exception of Ag + applied at 0.3 mg/g which achieved the highest LRV in both Tanzanian and Indonesian disks.

116
Disks made with either Indonesian or Tanzanian clay and coated with 0.3 mg/g AgNP achieved >4 LRV on the 10 th day of testing (1-1.7 LRV improvement over control disks without silver). A small improvement in LRV over the control disks was measured with 0.03 mg/g of either silver (<1 LRV) but disks coated with 0.003 mg/g of silver showed little or no improvement in LRV in comparison with the control disks.

Viable Bacteria Retention Inside the Disks
The concentration of viable bacteria inside disks decreased with increased silver concentration of either AgNP or Ag + , with the exception of disks made with Nicaraguan clay which showed little difference in viable bacteria regardless of silver concentration ( Figure 5-4). Results between AgNP and Ag+ were comparable, although disks coated with 0.3 mg/g of AgNP had fewer viable bacteria than Ag + .
Small changes were detected in the amount of viable bacteria remaining in disks coated with 0.003 mg/g of either silver species and the control groups (without silver application), regardless of clay type.

Effect of Influent Chemical Composition and Burn-out Material on Silver
Effluent concentration of silver from disks manufactured with rice husk or sawdust with 0.003 mg/g of either silver was below the MCL value of 100 µg/L after two hours of influent throughput with each water chemistry used (Figure B-1).
Variation in influent water characteristics resulted in little difference in silver retention 118 among disks treated with AgNP ( Figure B-2). In disks coated with Ag + , there was some variability in silver retention in disks manufactured with Tanzanian and Nicaraguan clays when HA was used as the influent solution. A difference in silver retention was not observed between disks manufactured with the same clay but different burn-out materials.
For each clay, LRVs were similar regardless of influent water chemistry applied, the silver type or the burn-out material used (Figure B-3). The amount of viable bacteria retained in disks coated with AgNP or Ag + was similar regardless of the influent water chemistry conditions or burn-out material ( Figure B-3). Fewer viable bacteria were retained in disks manufactured with clay from Nicaragua than disks made with either Indonesian or Tanzanian clays regardless of silver type or influent water chemistry.

Discussion and Recommendations
In this study, we evaluated AgNP and Ag + performance at varying concentrations in disks manufactured with filter material manufactured from clays from different countries. Additionally, silver was evaluated under water chemistries in disks manufactured with different clays and different burn-out materials. Disks were used as a model for full-sized filters due to space, time and laboratory constraints. The main difference between our methods and full-sized filters is the manufactured procedure.
At CWF manufacturing facilities, filter mixture recipes are established by selecting a ratio and firing temperature that, using the locally available materials, result in filters that meet specific quality criteria such as flow rate, LRV and strength. In this study, rather than manufacture filter disks that would have met factory quality criteria, the ratio of clay to burn-out material was held constant regardless of the clay origin or burn-out type in order to keep all variables except silver application constant. Firing temperature was thus adjusted, depending on the clay, to achieve enough disk strength for testing. A variation in pore structure likely resulted from using the same amount of burn-out material regardless of burn-out material type or clay properties and also as a result of the variation in firing temperature.
Disks manufactured from Indonesian and Tanzanian clays had comparable advection and dispersion parameters and their porosity and mechanical strength were suitable for testing. The higher advection and dispersion coefficients in disks manufactured with Nicaraguan clay suggested the solute spread faster, possibly due to larger or more interconnected pores. The Nicaraguan clay was exceptionally challenging to work with during manufacturing and the firing temperature (1085°C) likely resulted in a level of vitrification (over-fired) not found in filters at the factory level. A separate particle size analysis carried out on the raw clays used in this research found a very low percentage of clay (<2 micron) in the material from Nicaragua in comparison with the Tanzanian and Indonesians clays (0.5%, 28.5% and 31%, respectively) [18]. Results from the Nicaraguan disks should therefore be interpreted with caution in this study. In future research, we recommend holding pore structure constantas opposed to mix ratio and manufacturing variablesto account for differences in manufacturing needed for different raw materials.
The results of our study showed: 1) increased desorption of Ag + compared with AgNP; 2) a difference between effectiveness of AgNP and Ag + (Ag + is more effective at high concentration) ; 3) variation in LRV of E. coli depending upon silver concentration; 4) variation in biofilm formation depending upon silver concentration; and, 4) at the concentrations tested, no impact of water chemistry on the efficacy of silver. These results are discusssed in the following paragraphs.
Using phosphate buffer influent water, disks retained AgNP more efficiently that Ag + . Desorption of AgNP ranged from 5% to 10% for all disks tested, while for Ag + , 10% to 30% desorbed from Tanzanian and Indonesian clay disks and 30% to 40% from Nicaraguan clay disks. This effect has been reported by other authors who evaluated sorption of silver species on unfired clays [19]. Ag + can be displaced by cations with higher valence or higher charge density, while AgNP are trapped in the nano-and micro-porous structure of the filter allowing a slow release of silver ions as the surface of the nanoparticles is oxidized by the dissolved oxygen in water [20].
In the Indonesian and Tanzanian clays, for both silver species, a dose-response relationship was observed: an increased concentration of silver resulted in increased LRV of E. coli.
Disks coated with 0.3 mg/g Ag + resulted in the highest LRV; however, this was likely the effect of the high concentration of silver in the effluent (one order of magnitude above the EPA MCL) and therefore bacteria deactivation is achieved via a different mechanism of action than the contact of the silver sorbed/trapped in the porous structure of the disks. With lower concentrations of silver, comparable bacterial reduction was achieved between AgNP and Ag + ; however, less silver was measured in the effluent of disks coated with AgNP. The application of high concentrations of Ag + in filters causes concern about: 1) the time that Ag + remains in the filter material, thus having implications on long-term of silver efficacy; and, 2) potential health consequences associated with ingestion of elevated concentrations of Ag + by filter users. The Nicaraguan disks showed little change in LRV regardless of the type or concentration of silver applied.
In disks manufactured with Indonesian and Tanzanian  The application of 0.003 mg/g of either silver did not demonstrate improved LRV over the control disks (without silver) after 10 days of testing (equivalent to 1300 L throughput in a full-sized filter). This data is consistent with another study that compared CWF performance with and without Ag + application [6].
The biofilm quantification inside the disks supports and expands upon the LRV results. An increase in silver concentration resulted in reduced biofilm formation in disks; and, disks with a higher concentration of AgNP had less biofilm formation than disks with a higher concentration of Ag + . The results of this study demonstrated that a silver coating reduces the biofilm formation (up to two orders of magnitude) compared 122 with disks without silver. To our knowledge this is the first study providing quantitative information about the antibiofouling properties of silver in CWF.
Phase II of our study focused on evaluating the impact of: 1) inorganic and organic compounds present in natural water; and, 2) burn-out materials, on silver sorption, bacterial removal, and biofilm formation. The silver concentration (0.003 mg/g) used in this phase was selected to minimize the impact of residual silver from either AgNP or Ag + on bacteria deactivation. At the selected test conditions, a difference was not observed between the clays, burn-out material, or the silver species with the water chemistries evaluated in terms of silver retention, LRV or concentration of viable bacteria remaining in the disks. This could be due to the low concentration of silver used, as little impact of this low concentration was also seen using phosphate buffered water, and several others studies have shown the influence of the chemical characteristics of the solution on AgNP aggregate size [14,20]. Based on current knowledge about the aggregation of nanoparticles in different electrolyte solutions, water containing a low concentration of divalent ions should be used to prepare the silver solution used to coat filters.
Recommendations resulting from this research include: 1) factories should use AgNP rather than Ag + due to higher retention within the filters: Ag + is not recommended for filter application as it can lead to a silver concentration exceeding health standards in the filtered water; 2) factories could increase the AgNP concentration to 0.3 mg/g (approximately 640 g/filter) to achieve improved microbiological performance without compromising the quality of the effluent from the filter; and, 3) although this study did not show significant differences in terms of performance with water chemistry, there is evidence from other studies that organic and inorganic compounds present in natural water can affect AgNP performance.
We recognize that these results and recommendations will have an impact on factories. Ag + is locally available and significantly cheaper in some countries (and in some cases information about quality or concentration may be unavailable); and, importing AgNP can be a challenge. Therefore, the recommendation to use only AgNP and to increase AgNP concentration by 10 times the current recommendation will add a cost burden to the manufacturers. While the manufacture of high quality filter material will remain important in achieving high performing filters, silver application improves filter effectiveness; however, silver application at lower concentrations does not appear to have lasting effectiveness and therefore is not cost effective.
We also note that previous research has not always documented sufficient detail required to compare research results including type of silver, silver concentration, dilution and throughput water characteristics. In some cases this may be attributable to a lack of documentation or information about silver type or concentration. Previous research should therefore be compared with caution.
The limitations in this study include the use of a controlled 5.4 L/hr flow rate, which is about 2-3 times the flow rate used in the field. This flow rate was selected to achieve throughput equivalent to represent long-term operation in a short period of time. The results likely overestimate biofilm formation and underestimate microbiological performance due to faster water velocity, constant pressure, and reduced contact time between silver and bacteria. This study design was such to 124 challenge the materials and therefore, silver would likely be more effective at the household level than measured in this study.
This study identified several key parameters that require more detailed studies, such as silver concentration and the effects of various influent water characteristics and the nature of clay and other manufacturing variables. Further research recommendations include: 1) evaluation of higher concentrations of AgNP under a selection of water chemistry conditions to evaluate nanoparticle aggregation and silver particle size distribution in filter effluent; 2) evaluate effects of water characteristics both on silver dilution and filter use (influent solutions); and, 3) evaluate physicochemical interaction between clay and AgNP or Ag + on silver sorption and LRV and the influence of these properties and the pore size distribution of the porous matrix.
for TPA should be evaluated before their future application in ceramic water filters is considered.

Introduction
Ceramic filters impregnated with silver nanoparticles (AgNP) are a promising point-of-use water treatment technology in the developing world that can be applied with local materials and labor. Currently, ceramic water filters (CWFs) are manufactured by pressing and firing a mixture of clay and a combustible material such as flour, rice husks, or sawdust prior to treatment with AgNP. The filter is formed using a filter press, after which it is air-dried and fired in a flat-top kiln, increasing the temperature gradually to about 900 ˚C during an 8-h period [1]. This forms the ceramic material and burns off the sawdust, flour, or rice husk in the filters, making it porous and permeable to water. After firing, the filters are cooled and impregnated with a silver solution (either AgNP or silver nitrate) by either dipping the filters in the silver solution or painting the silver solution onto the filters.It has been demonstrated that the silver compounds add disinfection properties to the CWFs [1,2], increasing the bacteria removal and therefore increasing the water quality [3]. Also, it has been hypothesized that the use of the silver solution extends the useful life of the ceramic filters by preventing the formation of biofilm inside the ceramic matrix [1][2][3].
The silver compounds are added to the CWFs as the disinfectant in three different ways: 1) they can be painted on the CWFs, 2) the CWFs can be dipped in the silver solution, and 3) the silver compounds, in powder form (AgNP or AgNO 3 ) can be mixed with clay, sawdust and water. Rayner [4] estimated that 56% of the factories painted the silver solution onto the CWF, 33% dipped the CWFs into the silver solution, and the remaining 11% mixed the silver in powdered form with the clay and sawdust. About 83% of factories use AgNP, while other 17% use silver nitrate.
The prices of noble metals have increased significantly in the past few years. The price of silver has increased by almost a factor of three since 2005 [5], threatening the sustainability of the current application of colloidal silver in CWFs. Therefore, alternative disinfectant compounds are needed to ensure the efficacy of these systems.
One candidate compound is 3-(trihydroxysilyl) propyldimethyloctadecyl ammonium chloride (TPA), a polymer that compounds silane with a quaternary ammonium species. Similar to AgNP, TPA exhibits excellent antimicrobial properties in dissolved solutions and has long-term durability. A recent report indicated that 0.25 w/w% TPA exhibited a 99.99% inhibitory effect on Escherichia coli while 1% silver zeolite can only achieve 41.18% inhibition [6].TPAis usually applied as an antibacterial or antimold reagent. Current TPA applications include impregnation into thermoplastics or thermoset resins; dissolution in water and other solvents for use in coating, as caulk or as adhesive formulations; and application asa surface treatment for disinfection purposes [6]. The price of pure TPA powder is about $222/kg, which is much lower than the price of silver (about $1024/kg) [5,7]. The high disinfection performance and lower price makes TPA a potential disinfectant alternative to AgNP.
To the author's knowledge, no study has compared the disinfection performance of TPA and AgNP in CWFs. In this work, the potential application of TPA as a disinfectant in CWFs was evaluated. TPA was chosen because it is a non-toxic, nonirritating, biocompatible, thermally stable, and environmental friendly polymer.
Compared with similar quaternary ammonian silane compounds, TPA has the 132 advantageous property of non-leaching because reactive silanol groups within the TPA can either react with the treated materials to form covalent bonds or form strong hydrogen bonds [6,8]. The deactivation efficacies of AgNP and TPA were measured using a manometric technique that determines the overall activity of active bacteria and has advantage of avoiding miscounting compared to the traditional plate-count method. Bacteria were transported through the CWFs to compare the initial disinfectant performance of CWFs impregnated with AgNP and TPA. The release of AgNP and TPA from ceramic materials was also measured so that potential manufacturers could be informed regarding the best management practices to minimize the exposure of workers to the hazardous materials.

Ceramic disks manufacturing
The composition of the ceramic disks consisted of 40% Guatemalan clay, 10% flour and 50% (grog) pottery residues by weight. This selected combination of materials was the optimum combination because CWFs with higher percentages of flour were weak and can be broken easily, while CWFs with higher clay content has low hydraulic conductivity resulting in low flow rates [1]. The clay, flour and grog were homogeneously (244 g) mixed after which 75 mL of DI water were added [1].
The mixture was molded using a cylindrical mold (with a diameter of 6.5 cm) and compressed for 1 min at a pressure of 1000 psi. The disks were air-dried at room temperature for three days and then fired in a furnace with the temperature increasing at a rate of 150 0 C/h until a temperature of 600 0 C was reached. Then, the rate was increased to 300 0 C/h until the temperature reached 900 0 C, where it was maintained for 3 h. The obtained CWFs were cut into small water filter disks of 1.5 inches in diameter. The porous structure of the disks was confirmed by scanning electron microscopy (SEM).

Preparation and characterization of AgNP and TPA
Commercial AgNP (70.37% w/w Ag 0 ) stabilized with casein was obtained from

Microbial cultures
A non-pathogenic wild strain of E.coli provided by IDEXX laboratories (Westbrook, USA) was used for the bacteria-transport experiments. This organism was selected because of its use as a specific indicator of fecal contamination in drinking water and its extensive use in several studies of AgNP [1,9,10], which will allowed us to compare our results with previously published work. The bacteria were grown as described by Vigeant et al. [11]. Cells were re-suspended in a sterilized solution prepared with the respective water samples and synthetic water solution to a concentration of 10 10 -10 11 CFU/ml. Determination of the E. coli concentration was performed using the membrane filtration technique, applying m-FC with Rosolic Acid Broth (Millipore) and incubation at 44.5 0 C for 24 h [11].

Evaluation of the antibacterial activity
Batch tests of the deactivation of the bacteria by AgNP and TPA (concentration range: 6-45 mg/L) were performed in duplicates during 20 h using the manometric respirometric technique described by Zhang et al. [9,10].

Tracer and bacteria transport experiment
Transport experiments were performed with the CWFs under two conditions.1) without AgNP/TPA, and 2) after being painted with the two compounds. The CWFs were placed in holders connected to a Masterflex pump that maintained a flow rate of 0.5 mL/min. NaCl was used as the conservative tracer. The flow rate of 0.5 mL/min through a filter disk is estimated at approximately 5.4 L/h using the dimension of a full-sized water filter. One milliliter of NaCl (10 g/L) was injected, and DI water was used as the inflow solution. Samples were collected over time, and an electric conductivity (EC) meter was used to determine the NaCl concentrations. Before the bacteria transport experiment, the CWFs were saturated with 10% PBS inflow solution (consist of 1.12 g/L K 2 HPO 4 , 0.48 g/L KH 2 PO 4 , and 2 mg/L Ethylenediaminetetraacetic acid (EDTA); pH: 7.3) for 12 h. Then, the CWFs were dried in an incubator for 12 h. One milliliter of E. coli (10 10 -10 11 CFU/mL) was passed through the CWFs, and effluent samples were collected over time to determine the breakthrough of the E. coli. Then, the CWFs were heated in an oven at 500 0 C for 30 min and then cooled to room temperature. The CWFs were painted with AgNP and TPA using brushes according to the total mass applied as described by Oyanedel-Craver and Smith [1]. After drying for 24 h, one milliliter of E. coli solution was passed through the CWFs, and the concentration of the bacteria in the effluent was determined. Meanwhile, the rate of decay of the bacteria in the 10% PBS also was measured as a control using the same bacterial concentration and inflow conditions. The CXTFIT program (Riverside, USA) was used to simulate the obtained breakthrough curves. The advection-dispersion equation with first-order decay is: The initial and boundary conditions [1] were: Where R is the retardation coefficient, c is the concentration of NaCl or E. coli (CFU/ml), t is time (min), t 0 is the tracer or bacteria pulse injection time (min), D is the dispersion coefficient (cm 2 /min), µ is the first-order decay coefficient (min -1 ), v is the linear velocity (cm/min), and L is the thickness of the CWF. The model assumes local equilibrium sorption.
The concentration of AgNP in the effluent was measured as total silver (ICP-MS, X series, Thermo Elemental, Waltham, USA). TPA concentration was measured as quaternary ammoniumcompounds (QAC) concentration using a QAC test kit (HACH, Loveland, USA) and a DR 2100 spectrophotometer (HACH).   and ζ-potential of AgNP and TPA. The average particle size of the AgNP (69.5 ± 2.6 nm) was smaller than that of TPA (247 ± 27.8 nm). Further analysis was performed to compare the size distribution of AgNP and TPA. The comparison displayed that the 138 TPA exhibited a wider size distribution than AgNP. However, it was noticed that the ζ-potential of the AgNP was much smaller than that of TPA. AgNP shows negative ζpotential value (-25.6±0.75 mV). In DI water condition, previous studies showed that AgNP has negative ζ-potential values because an Ag(OH) 2 like species at the surface can be formed due to the oxidation of metallic silver by O 2 presence in water. In water matrices containing ions, adsorption of anions can also impart a negative surface charge. On the contrary, the positive ζ-potential value of TPA (27 ± 0.95 mV) was due mainly to the positively-charged quaternary ammonia groups in the TPA compound.

Characterization of the manufactured ceramic and the disinfectants
Similarly, Cumberland and Lead [12] found that AgNP show a ζ-potential value of -25.8 ± 5 mVin solutions containing HNO 3 or NaOH (pH in the range of 5 to 8). A ζpotential value of 28.29 mV was found for 3-(trimethoxysilyl) propyldimethyloctadecyl ammonium chloride that was very similarto that of TPA in the DI water condition [13]. Since most of the observed ζ-potentials of ceramic surfaces are negative [14], the ζ-potential values implied that the surface of the CWFs could adsorb TPA preferentially over the AgNP, which was supported by the results from the AgNP/TPA release experiment presented in Table 6-2 and Figure 6-4. Our study also investigated the ζ-potential of E. coli (-6.46 ± 0.43 mV) in a 10% PBS, which indicates that E. coli may preferentially contact TPA rather than the AgNP due to the adverse ζ-potential values for TPA and E. coli. The mechanism of the antimicrobial activity of AgNP has been attributed to the damage of bacteria by pitting the cell membrane, lysis of cells caused by silver ion release, or damage of the cell by the reactive oxygen species formed on the surface of the nanoparticles [15,16]. The toxicity of TPA is determined mainly by the positively charged ammonium groups and octadecyl groups. Kim et al [17] suggested that the interactions between TPA and the cytoplasmic membrane of E. coli can be enhanced by the positively charged ammonium groups. In addition, octadecyl groups can also increase the hydrophobic interaction with the cytoplasmic membrane and then cause cell disruption and leakage of the membrane. The value of ζ-potential for E. coli in 10% PBS was also determined (-6.46 ± 0.43 mV), which indicates that 1) attraction forces could exist between the negatively charged E. coli membrane and the positively charged TPA, and 2) electrostatic repulsion forces could exist between the negatively charged E. coli membrane and the negatively charged AgNP. This may explain why TPA has better disinfection performance than AgNP at similar low concentrations. In addition, high ionic strength in PBS may also account for the lower antimicrobial property of AgNP. Liu and Hurt [18] reported a decreased concentration of silver ion released from AgNP in solutions with high ionic strength. As silver ion release was proposed as an important factor that could affect the antimicrobial properties of AgNP, decreasing silver ion release could result in a lower antimicrobial property. Zhang and Oyanedel-Craver [9] observed an increasing particle size of AgNP with increasing ionic strength and indicated a negative correlation between particle size and 140 antimicrobial property.  The CXTFIT program was used to provide optimum fits for the tracer and bacteria transport data. Linear velocity v and dispersion coefficient D were determined with R = 1 and µ = 0 using the conservative tracer transport experiment data. R and µ were determined from the bacteria transport experiment. The fitted parameters and the percentages of bacteria removal are listed in Table 6    143 Similar results were observed in earlier work [1]. The main mechanism for the removal of bacteria by CWFs before applying AgNP or TPA was size exclusion [1].

Tracer and bacteria transport
Bacteria passing through CWFs can be retained by the small pores of the CWFs and pass through the large pores preferentially. However, there is an alternative mechanism that suggests that the bacteria possibly could be retained in the CWF by a reversible sorption process [1]. However, if a reversible sorption process is occurring, the breakthrough of bacteria should appear after the breakthrough of NaCl. It is noteworthy that the peak of the observed NaCl breakthrough occurred after the peak of the bacteria breakthrough when comparing Figure6-3 (a) with Figures 6-3 (b) and 6-3 (c). Similar results were observed for bacteria transport in CWFs and natural soils [1,[19][20][21]. It was also observed that effluent bacteria-concentration peaks occurred before the tracer peaks.    Similarly, the concentration of TPA in the effluent decreased over time. Currently, no specific regulatory information has been listed for TPA. As comparison, a similar quaternary ammonium silane product, 3-(trimethoxysilyl) propyldimethyloctadecyl ammonium chloride (Oral LD 50 >5000 mg/kg; Dermal LD 50 >2000 mg/kg; Inhalation 146 LC 50 >2.0 mg/L), has been documented by the U.S. EPA which shows a lack of toxicological effects at dose level up to and including a limit dose (i.e. 1000 mg/kg/day) [22]. We recommend that CWFs should be washed at least 300 min to minimize any possible adverse health effect.
One of the adverse health effects associated with exposure to silver is Argyria, an irreversible skin condition. The Occupational Safety and Health Administration (OSHA Washington, USA) and the National Institute for Occupational Safety and Health (NIOSH, Atlanta, USA) have established an exposure limit of 0.01 mg/m 3 (in air) for metallic and soluble silver compounds. However, although the exposure limit and health effects of silver have been studied extensively [23], the threshold exposure limit of TPA still is not available. Studies related to the exposure threshold and health effects of TPA on animals and humans are needed in future studies.  [27]. Table 6-3 shows some of the toxicological data that have been published for silver and TPA. These data indicate that silver has a higher acute oral toxicity than TPA.
However, Table 6-3 suggests that acute dermal toxicity and acute inhalation toxicity are low for both AgNP and TPA. The oral reference dose (RfD) of AgNP has been 147 determined. However, the RfD for TPA is not available because it is an emerging disinfectant. Future studies are needed to determine the RfD of TPA to assess its chronic health effects. Conclusively, this study showed that TPA is a viable alternative to AgNP in ceramic disks due to its high antimicrobial properties. Additionally, due to its lower price compared with AgNP, the application of TPA lowers the cost of CWFs. This could benefit to manufactures that will be using a product with a market value less 148 variable and could economically benefit the CWF users due to a lower cost of the product.
TPA can achieve higher bacterial reduction than AgNP in both aqueous solution and ceramic disks, suggesting that TPA could be an alternative disinfectant agent for this particular technology. The concentration of both AgNP and TPA released to the effluent are similar and decrease over time, indicating that similar amount of AgNP and TPA was retained inside disks. Considering the current toxicological information of TPA, the release of TPA may not have an acute impact on human health. However, more information regarding the long-term health effects of TPA is still needed to support the application of this product to ceramic water filters.

Conclusions
To achieve the three main goals presented in Chapter 1, a systematic investigation on the physicochemical properties and antimicrobial performance of AgNP in electrolyte solutions and natural water conditions was conducted. Studies on the application of AgNP in CWFs manufactured with different clay and burn-out materials as well as application of TPA in CWFs were also conducted.
Various techniques, such as DLS, membrane ultrafiltration and ultrafiltration, and ICP were applied to quantify the particle size, aggregation kinetics, surface charges, and dissolution of AgNP. Antimicrobial performance on E. coli and MS2 bacteriophage were evaluated with a nano-metric respirometric technique and double layer technique, respectively. Our results showed that AgNP can achieve up to 90% antibacterial activity. However, negligible antiviral effects of AgNP were observed.
The stability of AgNP varies due to the different stabilizers applied. PVP-coated AgNP are most stable compared with casein-and dextrin-coated AgNP. The abovementioned physicochemical properties and antibacterial properties change in different water chemistry conditions.
To apply the antimicrobial property of AgNP in drinking-water treatment, AgNP have been coated on CWFs to purify drinking water. This research also focused on establishing the silver application guidelines for CWF manufacturing. Silver concentration in the effluent of CWFs has been measured to make sure the effluent 154 concentration is below the U.S. EPA MCL (0.1 ppm) to minimize the negative health impacts of silver. Simple POU CWFs manufactured with different materials and different concentrations of two types of silver (AgNP and Ag + ) were studied. Bacterial removal and biofilm formation inside the disks were dose-dependent on the amount of silver applied. However, neither water chemistry conditions nor burn-out material showed an effect on any of the abovementioned parameters evaluated at the silver concentration tested. This study recommends applying a high concentration of AgNP (≥0.3 mg/g) on CWFs to achieve improved antimicrobial performance.
Investigation on the antimicrobial performance of AgNP and TPA showed that TPA-coated CWFs outperformed AgNP in a phosphate buffer solution. Amounts of AgNP or TPA released from CWFs are similar. Our results also showed that the disinfectant performance of TPA-applied CWFs exceeded AgNP-applied CWFs, indicating that TPA can be considered as a good alternative to AgNP.

Future work
1. Our research on the environmental fate of AgNP has elucidated the dissolution behavior of AgNP in different water chemistry conditions. However, the chronic environmental ecotoxicity of AgNP is still unknown. Future studies are recommended to cover the knowledge gap of their long-term (such as experimental duration in months or longer) toxicological effect on different microorganisms (such as algae and protozoa) and higher organisms (such as small fish embryos) in different water chemistry conditions. 2. In the application of AgNP on CWFs, instead of using CWFs applied with 0.003 mg/g AgNP, we recommend to conduct studies on higher concentrations of AgNP 155 (such as 0.3 mg/g) under a selection of water chemistry conditions to evaluate nanoparticle aggregation and silver particle size distribution in filter effluence. It is also recommended to evaluate physicochemical interaction between clay and AgNP or Ag + on silver adsorption and LRV and the influence of these properties and the pore size distribution of the porous matrix.
3. Although our study indicated that TPA can be considered as a good alternative to AgNP in CWF applications, more information regarding the long-term health effects of TPA is still needed. As our study focused on bacterial removal performances using a phosphate buffer solution, natural water conditions are recommended to evaluate the practicality of TPA-applied CWFs. Finally, we recommend the spectrum of waterborne pathogens (such as viruses and protozoa) that contribute to waterborne diseases be broadened. Future research on other quaternary ammonia functionalized silane species is also suggested. n/a n/a 6.5 10 n/a n/a 6.65 2 n/a n/a 6.66 1 n/a n/a 6.66 0.2 n/a n/a 6.65
Total alkalinity was determined by phenolphthalein and methyl orange titration.
Dissolved natural organic matter (NOM) was measured as total organic carbon (TOC) via a TOC analyzer (Apollo 9000, Tekmar Dohrman). Table S1  Mg 2+ in seawater, in which the AgNP solution exhibited the lowest toxicity [1]. A humic acid solution (TOC: 5 mg/L) was also prepared to study the influence of NOM on the disinfection properties of AgNP. All salts and other reagents were ACS reagent grade and used as received.

Microbial cultures
A non-pathogenic, wild strain of E. coli provided by IDEXX laboratories was used for bacteria disinfection experiments. E. coli was selected because of its use as a specific indicator of fecal contamination in drinking water and because of its extensive use in several studies on AgNP [2][3][4].This allows the comparison of our results with the results of previously published work. Bacteria were grown as described by  [5]. Cells were re-suspended in a sterilized solution prepared with the respective water samples and synthetic water solution to a concentration of (1±0.4)×10 10 CFU/mL.
Determination of the E. coli concentration was performed using the membrane filtration technique, applying m-FC with Rosolic Acid Broth (Millipore) and 24-h incubation at 44.5 0 C.

Determination of critical coagulation concentration
In brief, the initial aggregation rate of AgNP in different electrolyte solutions is proportional to the change of particle radius (nm) over time (s) ( dr dt ) 11 1 dr k N dt  (eq.1) Where k 11 is the initial aggregation rate of AgNP; N is the initial nanoparticle concentration.
When the electrolyte concentration in AgNP solution is low, the strong repulsion forces between AgNP will cause a low degree of aggregation, which is termed as slow aggregation. After the added electrolyte solution reaches a certain concentration, the 160 aggregation of AgNP falls into the rapid aggregation regime, where the aggregation energy barriers are completely removed and the aggregation rate is independent of the increasing electrolyte concentration [6].
The attachment efficiency -α‖, also defined as the inverse stability ratio (1/W), was determined by calculating the ratio of a given k 11 over the (k 11 ) rapid , which is the normalized k 11 value in the rapid aggregation regime. Where P c is the OUR after the injection of glucose, P t is the OUR after the injection of AgNP, and P is the OUR during the endogenous respiration without the addition of carbon source or nanoparticles. Figure