- Email: [email protected]
Adsorption of natural organic matter and disinfection byproduct precursors from surface water onto TiO2 nanoparticles: pH effects, isotherm modelling and implications for using TiO2 for drinking water treatment
Chemosphere 174 (2017) 363e370
Contents lists available at ScienceDirect
Chemosphere journal homepage: www.elsevier.com/locate/chemosphere
Adsorption of natural organic matter and disinfection byproduct precursors from surface water onto TiO2 nanoparticles: pH effects, isotherm modelling and implications for using TiO2 for drinking water treatment Stephanie L. Gora*, Susan A. Andrews Department of Civil Engineering, University of Toronto, 35 St. George Street, Toronto, Ontario M5S 1A4, Canada
h i g h l i g h t s TiO2 photocatalysis increased the DBPfp of surface water at short treatment times but reduced it at longer treatment times. In the absence of irradiation, TiO2 nanoparticles adsorbed NOM and DBP precursors from natural surface water. The adsorption of NOM, THM precursors and HAA precursors by TiO2 nanoparticles was modelled using a simple isotherm model. The adsorption of DOC, UV254, humic substances and DBP precursors occurred more readily at pH 4 than at pH 6 or pH 8.
a r t i c l e i n f o
a b s t r a c t
Article history: Received 4 November 2016 Received in revised form 22 January 2017 Accepted 24 January 2017 Available online 2 February 2017
Titanium dioxide is a photocatalyst that can remove organic contaminants of interest to the drinking water treatment industry, including natural organic matter (NOM) and disinfection byproduct (DBP) precursors. The photocatalytic reaction occurs in two steps: adsorption of the contaminant followed by degradation of the adsorbed contaminant upon irradiation with UV light. The second part of this process can lead to the formation of reactive intermediates and negative impacts on treated water quality, such as increased DBP formation potential (DBPfp). Adsorption alone does not result in the formation of reactive intermediates and thus may prove to be a safe way to incorporate TiO2 into drinking water treatment processes. The goal of this study was to expand on the current understanding of NOM adsorption on TiO2 and examine it in a drinking water context by observing NOM adsorption from real water sources and evaluating the effects of the resulting reductions on the DBPfp of the treated water. Bottle point isotherm tests were conducted with raw water from two Canadian water treatment plants adjusted to pH 4, pH 6 and pH 8 and dosed with TiO2 nanoparticles. The DOC results were a good fit to a modified Freundlich isotherm. DBP precursors and liquid chromatography with organic carbon detection NOM fractions associated with DBP formation were removed to some extent at all pHs, but most effectively at pH 4. © 2017 Elsevier Ltd. All rights reserved.
Handling Editor: W. Mitch Keywords: Titanium dioxide Advanced oxidation Natural organic matter Disinfection byproducts Adsorption Drinking water
1. Introduction Nanomaterials, defined as materials with any dimension in the nanoscale or having internal or surface structure in the nanoscale (ISO, 2010), are increasingly being used in the fields of electronics, computing and medicine. Some nanomaterials, including titanium
* Corresponding author. E-mail addresses: [email protected] (S.L. Gora), sandrews@civ. utoronto.ca (S.A. Andrews). http://dx.doi.org/10.1016/j.chemosphere.2017.01.125 0045-6535/© 2017 Elsevier Ltd. All rights reserved.
dioxide (TiO2) nanoparticles, may also prove to be useful in environmental applications, including the treatment of drinking water and wastewater.
1.1. Titanium dioxide for drinking water treatment The use of TiO2 for drinking water purification has been explored by many researchers, and at least two companies have developed small-scale systems on the basis of TiO2 photocatalysis, but it has yet to be widely applied for municipal drinking water
364
S.L. Gora, S.A. Andrews / Chemosphere 174 (2017) 363e370
treatment. Photocatalytic degradation of aqueous contaminants by TiO2 is generally thought to occur in two steps: adsorption and degradation. Adsorption can occur in the absence of light, but degradation occurs only when TiO2 is irradiated. Upon irradiation, reactive oxygen species (ROS), including the hydroxyl radical, and oxidative and reducing centres on the surface of the nanoparticle degrade contaminants that have been adsorbed on the surface of the photocatalyst. This two-step reaction is often described using the LangmuireHinshelwood mechanism (Malato et al., 2009). The two-stage nature of photocatalytic degradation differentiates it from other advanced oxidation processes such as UV/H2O2 and O3/ H2O2. Like other AOPs, however, many ROS formed when TiO2 is irradiated with UVA light are non-specific oxidants; hence, TiO2 photocatalysis has the potential to provide concurrent disinfection and degradation of organic drinking water contaminants, including taste and odour compounds and cyanotoxins (Fotiou et al., 2015), various pharmaceuticals (Avisar et al., 2013; Kanakaraju et al., 2014), and disinfection byproduct (DBP) precursors such as natural organic matter (NOM) (Liu et al., 2008; Huang et al., 2008; Philippe et al., 2010). The photocatalytic degradation process preferentially targets large aromatic NOM compounds, breaking them down into smaller ones (Huang et al., 2008; Philippe et al., 2010). This can result in decreased membrane fouling (Huang et al., 2008) and changes in DBP formation potential (DBPfp). The latter is of particular concern because although some of the degradation products of photocatalysis may be more likely to form DBPs than the original compounds, others may have an equal or greater DBPfp, particularly at shorter treatment times (Liu et al., 2008). As a result, an adsorption-based process may prove to be a safer and more effective option for the removal of NOM from drinking water using TiO2.
1.3. Adsorption of NOM to TiO2 Studies by Mwaanga et al. (2014), Erhayem and Sohn (2014) and Kim and Shon (2007) have established that pH and ionic strength affect the adsorption of NOM on TiO2 and have noted that larger, more aromatic NOM compounds are adsorbed preferentially. The presence of bicarbonate, phosphate and nitrate has been shown to decrease NOM adsorption on TiO2 while the presence of magnesium and calcium increases the likelihood of NOM adsorption (Erhayem and Sohn, 2014; Sun and Lee, 2012). Other studies have described the effects of individual ions on the agglomeration of TiO2 nanomaterials. Agglomeration decreases the overall surface area available for adsorption and as such is likely to have an impact on the ability of TiO2 nanomaterials to adsorb NOM. Liu et al. (2013) reported that three types of TiO2 nanomaterials were more likely to agglomerate under high-ionic strength conditions than at low-ionic strength conditions. This finding is corroborated by those of Erhayem and Sohn (2014). The type of ions present in solution may also have an effect. Liu et al. (2013) observed greater increases in agglomerate size when calcium was added to the water rather than sodium. They hypothesized that this was because of the greater ability of Ca2þ to compress the electrical double layer surrounding the nanomaterials relative to Naþ. Greater compression of the electrical double layer results in less repulsion between individual nanoparticles and thus, greater agglomeration. The presence of NOM increases the stability of nanomaterials in solution, though this effect is less pronounced in the presence of ions such as calcium (Liu et al., 2013; Zhang et al., 2009) and at high NOM concentrations (Erhayem and Sohn, 2014). According to Zhang et al. (2009), NOM inhibits agglomeration by increasing the overall negative charge of the particles, thus increasing the repulsive forces that keep them dispersed in solution.
1.2. Natural organic matter The removal of NOM, a heterogeneous mixture of organic compounds found in most surface water sources, is an important objective of drinking water treatment because of its aesthetic, operational and health effects. The latter are primarily linked to the role of NOM as a precursor to form both regulated and unregulated DBPs. Operational concerns related to NOM include membrane fouling, competitive adsorption to adsorbent materials intended for the removal of taste and odour compounds, interference with UV disinfection and consumption of chlorine during disinfection. NOM is commonly quantified as total organic carbon (TOC) or dissolved organic carbon (DOC), bulk parameters that measure the total mass of organic carbon compounds present in a water sample without differentiating them from one another. UVeVis absorbance at 254 nm (UV254) and fluorescence are also used to characterise NOM, though these methods are specific to NOM compounds containing aromatic chromophores or fluorophores. More complex methods such as liquid chromatography with organic carbon detection (LC-OCD) have been developed that allow researchers to separate NOM into different fractions based on size or chemical characteristics. The resulting fractions are then quantified using DOC or UV254. LC-OCD separates NOM into five fractions based on size and/or chemical characteristics as follows: biopolymers, associated with membrane fouling (Wray et al., 2013); humic substances, which have been linked to DBP formation (Wassink et al., 2011); building blocks; low-molecular-weight acids (LMWAs); and low-molecular-weight neutrals (LMWNs) (Huber et al., 2011). A given NOM sample's potential to form regulated and unregulated DBPs can also be assessed more directly by measuring its DBPfp by chlorinating water samples and measuring the DBPs formed under standardised conditions.
1.4. Adsorption models Adsorption processes are usually evaluated in the laboratory using adsorption isotherm models. The Freundlich isotherm often fits well to empirical data and can be used to model heterogeneous systems such as the adsorption of organic molecules to activated carbon (Summers and Roberts 1988). It can be used to describe multilayer adsorption, reversible adsorption and adsorbents with non-uniform adsorption sites (Shahbeig et al., 2013). Summers and Roberts (1988) found that a modified version of the Freundlich isotherm could be used to describe the adsorption of NOM to activated carbon when experiments were conducted with a constant initial concentration of NOM and changing doses (D) of activated carbon. They also developed the following equation (referred as the SR model in this paper) to express this relationship:
qe ¼ KSR
Ce D
1 nSR
(1)
where D has units of mg L1 or g L1. The SR model was used and extended upon by numerous researchers including Karanfil et al. (1999), Li et al. (2002), Hyung and Kim (2008) and Qi and Schideman (2008) to characterise the adsorption of NOM to activated carbon and carbon nanotubes. Erhayem and Sohn (2014) modelled the adsorption of Suwannee River humic acids, fulvic acids, and NOM to P25 TiO2 nanoparticles using the SR isotherm model. They noted that the SR adsorption constant (and thus the extent of adsorption) increased at lower pH and at higher ionic strength.
S.L. Gora, S.A. Andrews / Chemosphere 174 (2017) 363e370
1.5. Potential risks and opportunities associated with the use of TiO2 nanoparticles for water treatment Conventional water treatment technologies such as coagulation and activated carbon are effective for NOM removal but, like all treatment technologies, have limitations. In North America, coagulation is widely used to remove NOM and turbidity from drinking water, but it does not remove some recalcitrant organics. It also creates a substantial amount of waste, often referred to as coagulation residuals. Activated carbon readily removes NOM and other organic compounds but eventually becomes exhausted, and must undergo an expensive and energy intensive regeneration process if it is to be reused. Although TiO2 is more well known for its photocatalytic properties, it also adsorbs NOM as a part of that process, and is potentially regenerable onsite (Liu et al., 2014). As such, it might prove to be a useful alternative to the existing treatment options. TiO2 nanoparticles are not without health and environmental concerns. Inhalation is the route of exposure of the greatest concern for human health (Shi et al., 2013), but the transport of nanoparticles in the environment is coming under increasing scrutiny (Yang and Westerhoff, 2014). The adsorption of various water components, including NOM, on TiO2 can facilitate the latter's transport through water systems. To date, most of the studies on the interactions between NOM and TiO2 have aimed to elucidate these effects to predict and minimise the impact of nanomaterials on the natural environment (e.g. Mwaanga et al., 2014; Erhayem and Sohn, 2014; Kim and Shon, 2007; and Liu et al., 2013)). This study aims to expand on the existing research into the adsorption of NOM by TiO2, most of which have been conducted in a contaminant transport context. This study also explores the potential application of TiO2 as an adsorbent in drinking water treatment by studying its behaviour in natural surface water sources and measuring its ability to remove DOC, UV254 and DBP precursors.
2. Materials and methods 2.1. Materials Evonik Aeroxide P25 TiO2 nanoparticles were purchased from Sigma Aldrich (Canada) and used without further modification. THM and HAA standards were also purchased from Sigma Aldrich. DOC standards were prepared by dissolving potassium hydrogen phthalate into milliQ water to create a 1000-mg L1 stock solution, which was then diluted as required. Raw water was obtained from the inlets of two water treatment plants (WTPs) in Ontario, Canada, both of which use surface water supplies. The Peterborough WTP is supplied by the Otonabee River, while the R.C. Harris WTP in Toronto draws its water from Lake Ontario, one of the largest lakes in North America, which provides water to over 9 million people in Canada and the United States. All water samples were obtained ahead of any pre-chlorination point at the WTP and characterised upon return to the laboratory. Measured ranges for five relevant water parameters are provided in Table S1 in the supplemental file. The water sources varied primarily in terms of their NOM content and aromaticity. The DOC of the Otonabee River water was approximately 3e4 times higher than that of the Lake Ontario water, while its UV254 was approximately five times higher. The SUVA of the Otonabee River water ranged from 2.0 to 2.4 L/mg.m, while that of the Lake Ontario water ranged from 0.8 to 1.0 L/mg.m, indicating that the NOM in the former is more aromatic than that in the latter.
365
2.2. Analytical methods DOC was determined using an O.I. Analytical Aurora 1030W TOC analyzer operating in the persulphate oxidation mode and UV absorption at 254 nm (UV254) was analysed on an Agilent 8453 UVeVis analyzer. SUVA was calculated by normalizing UV254 by DOC. Alkalinity was measured using Standard Method 2320 (APHA, 2005). Duplicate raw and treated water samples were analysed using size exclusion LC-OCD as described by Huber et al. (2011). The results of the analyses were processed using proprietary software (ChromCalc, DOC-LABOR, Karlsruhe, Germany). The isoelectric point of the nanoparticles was determined by measuring the zeta potential of the nanoparticles at different pH values. A series of samples containing 0.1 g L1 TiO2 in 10 mM NaCl were adjusted to pHs ranging from 2 to 9 using 0.1 N HCl or 0.1 N NaOH as per the method outlined by the Nanotechnology Characterization Laboratory (2009). The zeta potential of the samples was measured using a Horiba Scientific Nanopartica SZ100 Nanoparticle Analyzer. The size of the nanoparticle agglomerates formed at different pHs was determined using a Malvern MasterSizer 3000. The uniform formation conditions (UFCs) method as described by Summers et al. (1996) was used to assess the chlorine demand and DBPfp of the raw water and the water that had been treated with TiO2. The UFC test was designed to mimic the conditions commonly found in distribution systems in North America. The chlorine demand and DBPfp tests were conducted on samples buffered with a borate solution and adjusted to pH 8 with 1 N HCl or 1 N NaOH. Samples were stored in the dark at 20 C for 24 h, after which the free chlorine residual was measured using Standard Method 4500-G (APHA, 2005). The DBPfp samples were dosed with sufficient sodium hypochlorite to ensure that they would have a chlorine residual of 1 ± 0.4 mg L-1 after the 24 h holding time. After 24 h, the trihalomethanes and haloacetic acids formed during the UFC tests were extracted according to Standard Method 6232 B and Standard Method 6251 B (APHA, 2005) and analysed on an Agilent 7890B GC-ECD [Standard Method 6232 B and Standard Method 6251 B (APHA, 2005)]. Blanks and 20 mg L-1 check standards were analysed after every 10 samples. All statistical analyses, including Tukey's method for multiple comparisons to establish a point of practical equilibrium and linear regression of the adsorption data to determine KF and KSR, were conducted at the 95% confidence level. Reported error values represent half of the calculated confidence interval unless otherwise specified. 2.3. Sample preparation NOM degradation studies were conducted in a high-intensity UV reactor equipped with UV LED lamps emitting UVA light at 365 cm with an average irradiance of 6.25 mW/cm2. Unchlorinated raw water from the Peterborough Water Treatment Plant (Otonabee River water) was dispensed into three 50-mL batch reactors, dosed with 0.25 g L-1 of P25 nanoparticles, allowed to mix in the dark for 1 min, and then exposed to the LED light for times ranging from 0 to 60 min. The treated samples were chlorinated according to the UFC method and analysed for THMfp and HAAfp as described in Section 2.2. The time required to reach a stable adsorption equilibrium between NOM and TiO2 nanoparticles was determined by adding 75 mL of raw unchlorinated water to duplicate 125 mL amber bottles and, when necessary, adjusting the pH to 4, 6, or 8 with 1 N HCl or 1 N NaOH. The bottles were dosed with 0.5 g L-1 of Evonik P25 TiO2 nanoparticles and mixed end-over-end in a box mixer for
366
S.L. Gora, S.A. Andrews / Chemosphere 174 (2017) 363e370
times ranging from 1 to 8 h. This time range was chosen because previous experiments (results not shown) had indicated that most NOM adsorption occurred within 5 min and that equilibrium likely occurred between 1 and 4 h. For the bottle point isotherm tests, eight 250 mL amber bottles were filled with 150 mL of raw water; adjusted to pH 4, 6, or 8; dosed with 0, 0.01, 0.025, 0.05, 0.1, 0.25, 0.5, or 1 g L-1 of Evonik P25 TiO2 nanoparticles; and then mixed continuously in the dark for 4 h in an end-over-end box mixer. All experiments were run in triplicate. All replicate samples were analysed for DOC, UV254 and SUVA. The samples from one replicate experiment were used to establish chlorine demand, and the remaining samples were analysed for THMfp and HAAfp. The DOC, THMfp and HAAfp results of the bottle point isotherm tests were evaluated for fit against the linearised Freundlich and SR models. All raw and treated samples were filtered through a 0.45-mm polyethersulphone (Pall) laboratory filter in a standard vacuum filtration apparatus to remove particulate matter and TiO2 ahead of DOC, UV254, DBPfp and LC-OCD analysis. 3. Results and discussion 3.1. DBP formation during photocatalysis During the irradiation tests, the THMfp and HAAfp of the raw Otonabee River water were modestly reduced by adsorption alone (15% and 10%) but both increased upon irradiation (Fig. 1). The impact on THMfp was particularly dramatic. After only 5 min of irradiation, THMfp increased by 61% relative to the control. After 30 min, THMfp began to decrease and after 60 min the THMfp of the treated water matched that of the control. The increase in HAAfp upon irradiation was smaller than that of THMfp and was reversed after 30 min of irradiation. After 60 min of irradiation, HAAfp was reduced by 35% relative to the control. These results, which are similar to those obtained by Liu et al. (2008) and Philippe et al. (2010), illustrate the risk associated with the use of photocatalysis for NOM and DBPfp reduction. Although both THMfp and HAAfp were eventually reduced by the treatment, both increased at treatment times between 0 and 15 min. Shorter treatment times are desirable as they minimise the amount of space and energy required at full scale. The modest reductions of THMfp and HAAfp through adsorption, however, suggest an alternative treatment optiondcould higher concentrations of TiO2 adsorb a sufficient amount of DBP precursors to provide a viable reduction in overall DBPfp and do certain water quality conditions (e.g. pH) favour the adsorption of DBP precursors to TiO2?
Fig. 1. THMfp and HAAfp of Otonabee River water treated with 0.25 g L-1 and irradiated by high intensity UVA-LED light.
3.2. NOM removal through adsorption e time series experiments The practical adsorption equilibrium, defined as the point at which the 95% confidence of neighbouring means began to overlap and the slope of the line of mean concentration vs. time could no longer be distinguished from zero, was determined on the basis of the results of the DOC and UV254 time series experiments conducted in each water source. Irrespective of the water type used or the parameter observed, the results indicated that the majority of NOM adsorption to the P25 TiO2 particles occurred within minutes and that a practical adsorption equilibrium was reached within 1 h (see Figs. S1eS3 in the supplemental file). This is similar to results obtained by some researchers working with TiO2 materials, including P25 nanoparticles (Kim and Shon, 2007; Ng et al., 2014), though others have suggested that a longer period of time might be required to reach full equilibrium (Mwaanga et al., 2014; Erhayem, 2013). On the basis of the results of the time series experiments conducted in this study, all subsequent equilibrium experiments were conducted with a 4-h adsorption period to ensure that all data were gathered at a point well beyond the practical point of equilibrium. The results of the time series experiments also indicated that adsorption was most effective on pH 4, and that UV254 and SUVA were more strongly affected by the treatment than DOC, hinting that aromatic NOM may have been preferentially adsorbed by the nanoparticles. The latter effect was more apparent in the Otonabee River water because it had an initial raw water SUVA that was two times that of the Lake Ontario water. 3.3. Effects of pH and TiO2 dose on adsorption Bottle point isotherm tests were conducted to further characterise the adsorption behaviour of the nanoparticles at the three pH conditions as the dose of TiO2 was varied from 0.01 to 1 g L-1. As shown in Fig. 2 and Fig. S4 in the supplemental file, at equilibrium, the DOC and UV254 removals observed in the Otonabee River and Lake Ontario samples were found to be pH dependent and in all cases, more NOM was removed at pH 4 than at pH 6 and pH 8. Irrespective of pH, increasing the dose of TiO2 added to the water resulted in a decrease in the amount of DOC remaining in the treated water. In both the Otonabee River (Fig. 2A) and Lake Ontario water trials (Fig. 2B), DOC removal increased quickly as the TiO2 dose was increased from 0.01 to 0.25 g L-1, but slowed thereafter, though no definitive plateau was reached at any pH, suggesting that further increases in TiO2 dose beyond the maximum applied in this study (1 g L-1) may have improved DOC removal even further. At pH 4 and 1 g L-1 of TiO2, the DOC of the Otonabee River water was reduced from 4.69 ± 0.12 to 1.10 ± 0.12 mg L-1 whereas that of the Lake Ontario water was reduced from 1.64 ± 0.05 to 0.69 ± 0.04 mg L-1. DOC removal from both water sources was statistically significantly lower at pH 6 and pH 8 than at pH 4. The UV254 results followed similar trends as DOC and are shown in Fig. 4. UV254 was removed more effectively than DOC, particularly from the Otonabee River water, where UV254 was reduced from 0.112 ± 0.002 to 0.013 ± 0.002 cm1, approximately 88%, by a 1 g L-1 dose of P25 nanoparticles at pH 4. UV254 is a measure of the aromaticity of the NOM present in water; hence, the results presented here suggest that aromatic NOM was preferentially removed over non-aromatic NOM during the adsorption process. As has been suggested by other researchers (Mwaanga et al., 2014), this pH dependence may be partially explained through charge interactions. Zeta potential measurements conducted on the P25 nanoparticles indicated that they had an isoelectric point (IEP) between pH 6 and pH 6.5, consistent with the literature (Kosmulski,
S.L. Gora, S.A. Andrews / Chemosphere 174 (2017) 363e370
367
Fig. 3. Size distribution of agglomerates of P25 nanoparticles in Otonabee River water adjusted to pH 4, pH 6 and pH 8.
Fig. 2. Adsorption of DOC from raw unchlorinated water from Otonabee River water (A) and Lake Ontario water (B) adjusted to pH 4, pH 6 and pH 8 and mixed with 0.5 g L1 of P25 TiO2 nanoparticles for 4 h.
2009). Thus, at pH 4, they were positively charged; at pH 8, they were negatively charged; and at pH 6, they were approximately neutral. At pH 4, most NOM compounds would have been neutral or slightly negatively charged and both charge and hydrophobic interactions likely contributed to adsorption. At pH 6, hydrophobic interactions between NOM compounds were likely the main contributors to adsorption. At pH 8, both the nanoparticles and the NOM were negatively charged and thus charge interactions would result in repulsion, rather than attraction, and any adsorption that occurred would have been attributable to hydrophobic interactions. Nanoparticle agglomeration and its effects on surface area may also have contributed to the changes in adsorption efficiency observed at different pHs. Agglomeration is most likely to occur when the pH is near the isoelectric point/point of zero charge of the material in question because at this pH repulsive forces between individual particles are at a minimum (Liu et al., 2013). P25 has an IEP of approximately 6.5, which means that the nanoparticles were more likely to agglomerate at pH 6 than at pH 4 or pH 8. Indeed, the particle size distributions presented in Fig. 3 indicate that the agglomerates formed by the nanoparticles were larger in Otonabee River water adjusted to pH 6 than in Otonabee River water adjusted to pH 4 or pH 8. In this study, better adsorption was observed at pH 4 than at pH 6. Agglomeration and the subsequent reduction in available surface area at pH 6 versus at pH 4 may have contributed to the poorer adsorption observed at pH 6, whereas charge repulsion between NOM and the TiO2 nanoparticles was more of a driver at pH 8. 3.4. Modelling of adsorption isotherms The DOC results of the bottle point isotherm tests were
Fig. 4. DOC data from Otonabee River water tests (A) and Lake Ontario water tests (B) fitted to the Summers and Roberts (SR) model.
modelled using the linearised forms of the Freundlich and SR models, as shown in Fig. 4 (SR model) and Fig. S5 of the supplement (Freundlich model). The isotherm parameters are summarised in Table 1 and Table S2 in the supplemental file. At all pHs and in both water sources, the SR model was a better fit to the data, defined as a higher R2 value, than the Freundlich model. The SR model is generally thought to provide a more accurate fit for data from highly heterodisperse systems and when the isotherms are developed using variable doses of adsorbent; hence, this result was not surprising. Other researchers have observed that aromatic NOM and humic acids are preferentially adsorbed to P25 nanoparticles over other types of NOM (Erhayem and Sohn, 2014). Given that the two water
368
S.L. Gora, S.A. Andrews / Chemosphere 174 (2017) 363e370
Table 1 Summary of isotherm parameters for the adsorption of NOM from Otonabee River water onto P25 TiO2 nanoparticles at pH 4, pH 6 and pH 8. Parameter
DOC 1/nSR KSR (mg DOC/g TiO2)11/n R2 THMfp 1/nSR KSR (mg THMfp/g TiO2)11/n R2 HAAfp 1/nSR KSR (mg HAAfp/g TiO2)11/n R2
Otonabee River
Lake Ontario
pH 4
pH 6
pH 8
pH 4
pH 6
pH 8
0.37 ± 0.03a 3.7 ± 1.1 0.98
0.49 ± 0.04 1.3 ± 1.2 0.97
0.47 ± 0.06 1.0 ± 1.2 0.94
0.58 ± 0.05 1.5 ± 1.1 0.98
0.57 ± 0.07 0.8 ± 1.2 0.95
0.59 ± 0.10 0.5 ± 1.3 0.90
0.39 ± 0.09 27 ± 2 0.91
0.42 ± 0.10 13 ± 2 0.90
0.42 ± 0.20 8±4 0.71
eb e e
e e e
e e e
0.39 ± 0.15 8±2 0.81
0.48 ± 0.22 4±3 0.76
0.69 ± 0.29 1±5 0.76
e e e
e e e
e e e
a
Error values represent half of the 95% confidence interval for each parameter. Since the concentrations of THMs and HAAs detected in the raw Lake Ontario water after chlorination at UFC conditions were near the analytical detection limit, THMfp and HAAfp isotherm experiments were not conducted with this water source. b
sources differ mainly in terms of their NOM concentration and aromaticity, it is not surprising that the KSR values obtained from the Lake Ontario tests were lower than those obtained from the Otonabee River tests. The higher 1/nSR values in the Lake Ontario tests also indicate that adsorption was less favourable in this water than in the Otonabee River water. Within each water source, 1/nSR was nearly constant irrespective of pH and KSR was larger at pH 4 than at pH 6 and pH 8, which once again indicates that the adsorption of NOM to P25 TiO2 nanoparticles was more effective at pH 4 than at pH 6 or pH 8 in both water sources. The pH 6 and pH 8 confidence intervals for KSR overlapped in both water sources, perhaps suggesting that pH became less of a driver of adsorption capacity when the pH of the water was equal to or higher than the IEP. These findings are in agreement with those of other researchers (Mwaanga et al., 2014; Erhayem and Sohn, 2014; Sun and Lee, 2012), which have demonstrated that NOM adsorption to P25 TiO2 nanomaterials occurs more readily at low pH than at high pH. They are also in agreement with studies that have demonstrated that NOM adsorption to TiO2 can be modelled using the Freundlich (Wiszniowski et al., 2002) and SR models (Erhayem and Sohn, 2014), though it should be noted that all the aforementioned studies made use of standardised NOM or humic acid isolates (e.g. IHSS) in synthetic water matrices rather than natural water sources. The KSR and 1/nSR results of this study and those of other TiO2 researchers are lower than those achieved by other groups working with activated carbon and nanoscale carbon adsorbents, but not dramatically so. In their original study, which was conducted with four NOM isolates and GAC doses similar to the TiO2 doses used in this study, Summers and Roberts (1988) observed KSR values ranging from 4.22 to 11.4 (mg C/g GAC)11/n and 1/nSR values ranging from 0.254 to 0.347. Karanfil et al. (1999) evaluated the adsorption of commercially available NOM isolates and NOM from natural water onto a series of commercially available and modified activated carbons. They observed KSR values ranging from 1.754 to 10.695 (mg C/g GAC)11/n in the natural water matrices. Hyung and Kim calculated KSR values ranging from 5.471 to 13.088 (mg C/g MWNT)11/n and 1/nSR values ranging from 0.212 to 0.384 when they evaluated the adsorption of commercially available NOM isolates onto multi-walled carbon nanotubes.
method; hence, the adsorption of DBP precursors to the P25 nanoparticles was not explored for this water source. The raw water and treated samples prepared during the Otonabee River adsorption tests were analysed for THMfp and HAAfp, and the results are shown in Fig. 5. As observed with DOC and UV254 removal, THMfp reduction through adsorption was pH dependent. Maximum THMfp reduction, 147 ± 16 mg L-1 to 38 ± 16 mg L-1 (74% reduction), was achieved at pH 4 and a TiO2 dose of 1 g L-1. Less removal was achieved at this dose at pH 6 (153 ± 11 mg L-1 to 75 ± 11 mg L-1, 51% reduction) and pH 8 (154 ± 15 mg L-1 to 104 ± 15 mg L-1, 34% reduction). HAA precursors were also removed through adsorption and this removal was pH dependent, though less so than for THM precursors. Fig. 5B shows that at the highest concentration of TiO2
3.5. Adsorption of DBP precursors Negligible amounts of THMs and HAAs were formed in the raw Lake Ontario water when it was chlorinated according to the UFC
Fig. 5. THMfp (A) and HAAfp (B) of Otonabee River water treated with increasing concentrations of P25 TiO2 nanoparticles at pH 4, pH 6 and pH 8.
S.L. Gora, S.A. Andrews / Chemosphere 174 (2017) 363e370
(1 g L-1) HAAfp was reduced from 39 ± 7 mg L-1 to 19 ± 5 mg L-1 (50% reduction) at pH 4 and from 36 ± 2 mg L-1 to 23 ± 3 mg L-1 (40% reduction) at pH 6. HAAfp reduction at pH 8 was not statistically significant at the 95% confidence level. Some of the variability in the HAA results can be explained by the fact that the UFC test, which was used to evaluate the THMfp and HAAfp of the raw and TiO2treated samples in this study, is conducted at pH 8, which does not favour the formation of HAAs. As a result, all the raw and TiO2treated samples had low HAAfp, making it difficult to isolate the effects of TiO2 adsorption on HAAfp removal, particularly at pH 8. The agreement between the THMfp and HAAfp datasets and the SR model are shown in Fig. S6, and the isotherm parameters are summarised in Table 1. Although the R2 values of the THMfp and HAAfp isotherms were lower than those of the DOC isotherms, the general trends indicate that, with the exception of HAAfp at pH 8, TiO2 could remove significant amounts of THM and HAA precursors from Otonabee River water through adsorption and that this adsorption could be modelled using the SR isotherm model. Nonetheless, as a whole, the isotherm parameters for the two classes of DBPs should be approached with caution because they were developed using a small dataset that contained substantial variation at low TiO2 doses. Additional experiments at higher TiO2 doses, using water sources with higher concentrations of DBP precursors, and/or employing chlorination regimes more likely to result in THM and HAA formation may help clarify the how well the SR model can predict the removal of DBP precursors from drinking water by TiO2 as well as the suitability of the model at different pHs.
3.6. Effects of pH on adsorption of LC-OCD fractions A selection of raw and TiO2-treated water samples from the Otonabee River experiment was analysed using LC-OCD to determine whether any specific fractions were being removed during adsorption and whether pH impacted the fractions adsorbed. As shown in Fig. 6, the biopolymers and humic substances fractions were targeted for adsorption at all pHs but most effectively removed at pH 4. The building blocks fraction was also removed to some degree at pH 4 but was essentially unaffected at pH 6 and pH 8. The LMWA and LMWN fractions were not adsorbed at any pH. These results indicate that, consistent with the findings of other researchers (Erhayem and Sohn, 2014), large and aromatic NOM compounds were preferentially adsorbed by TiO2 nanoparticles and help explain why U254, which is associated with the humic substances fraction, was sometimes removed more effectively than overall DOC.
369
4. Conclusions The results of this study show that during adsorption aromatic NOM (as measured by UV254) is preferentially removed over nonaromatic NOM and that the efficiency of NOM adsorption to TiO2 can vary by water source. They also demonstrate that TiO2 nanoparticles preferentially adsorb larger NOM molecules including the biopolymers and humic substances fractions. pH was shown to have a strong impact on the removal of NOM, including DBP precursors, from surface water by TiO2 nanoparticles. Specifically, more adsorption occurred at low pH than at higher pH. The poorer adsorption observed at pH 6 and pH 8 may be related to both agglomeration and charge repulsion at higher pH, with the former dominating at pH 6 and the latter at pH 8. A modified version of the Freundlich isotherm model (SR model) was found to provide an excellent fit to the DOC data gathered in this study. The resulting isotherm parameters were within but at the low end of the range usually observed during NOM adsorption to GAC and carbon nanomaterials, indicating that, particularly at neutral pH, the TiO2 nanoparticles were less effective than the adsorbents currently used in drinking water plants. Unlike TiO2, however, GAC is generally expensive and energy intensive to regenerate and the regeneration must usually be conducted offsite, whereas TiO2 is potentially regenerable and reusable in place. The THMfp and HAAfp datasets were also fitted to the SR model, with generally positive results. The results presented in this paper show that TiO2 adsorption is a viable way to remove NOM and DBP precursors from drinking water and that this removal can be modelled using simple isotherm models. The results also suggest that researchers hoping to design adsorption-based TiO2 processes should keep in mind that pH adjustment might be required to optimise performance. Acknowledgements The authors would like to acknowledge the training provided by Jim Wang, the laboratory assistance provided by Yijun (Jessie) Gai and Michelli Park, and the support of the Drinking Water Research Group at the University of Toronto. The authors are also grateful to Dr. Monica Tudorancea and Dr. Sigrid Peldzsus (University of Waterloo) for performing LC-OCD analyses. Funding was provided by the Natural Sciences and Engineering Research Council of Canada and the Ontario Ministry of Training, Colleges, and Universities. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.chemosphere.2017.01.125. References
Fig. 6. LC-OCD fractions present in raw unchlorinated Otonabee River water and water adjusted to pH 4, pH 6 and pH 8 and mixed with 0.5 g L-1 of P25 TiO2 nanoparticles for 4 h.
American Public Health Association, 2005. Standard Methods for the Examination of Water and Wastewater, twenty-first ed. APHA, Washington D.C. Avisar, D., Horovitz, I., Lozzi, L., Ruggieri, F., Baker, M., Abel, M.-L., Mamane, H., 2013. Impact of water quality on removal of carbamazepine in natural waters by Ndoped TiO2 photo-catalytic thin film surfaces. J. Hazard. Mater. 244e245, 463e471. Erhayem, M., 2013. Effect of Naturally Occurring Organic Matter (NOOM) Type and Source on NOOM Adsorption onto Titanium Dioxide Nanoparticles under Varying Environmental Conditions. Florida Institute of Technology, USA. Thesis. Erhayem, M., Sohn, M., 2014. Stability studies for titanium dioxide nanoparticles upon adsorption of Suwannee River humic and fulvic acids and natural organic matter. Sci. Total Environ. 468e469, 249e257. Fotiou, T., Triantis, T.M., Kaloudis, T., Hiskia, A., 2015. Evaluation of the photocatalytic activity of TiO2 based catalysts for the degradation and mineralization of cyanobacterial toxins and water off-odor compounds under UV-A, solar, and visible light. Chem. Eng. J. 261, 17e26. Huang, X., Leal, M., Li, Q., 2008. Degradation of natural organic matter by TiO2 photocatalytic oxidation and its effect on fouling of low-pressure membranes.
370
S.L. Gora, S.A. Andrews / Chemosphere 174 (2017) 363e370
Water Res. 42, 1142e1150. Huber, S.A., Balz, A., Abert, M., Pronk, W., 2011. Characterisation of aquatic humic and non-humic matter with size-exclusion chromatography-organic carbon detection and organic nitrogen detection (LC-OCD-OND). Water Res. 45, 879e888. Hyung, H., Kim, J.-H., 2008. Natural organic matter (NOM) adsorption to multiwalled carbon nanotubes: effect of NOM characteristics and water quality parameters. Environ. Sci. Technol. 42, 4416e4421. International Organization for Standardization, 2010. Nanotechnologies e Methodology for the Classification and Categorization of Nanomaterials. ISO/TR 11360:2010. Kanakaraju, D., Glass, B.D., Oelgemoller, M., 2014. Titanium dioxide photocatalysis for pharmaceutical wastewater treatment. Environ. Chem. Lett. 12, 27e47. Karanfil, T., Kitis, M., Kilduff, J.E., Wigton, A., 1999. Role of granular activated carbon surface chemistry on the adsorption of organic compounds 2. Environ. Sci. Technol. 33, 3225e3233. Kim, S.-H., Shon, H.K., 2007. Adsorption characterization for multi-component organic matters by titanium oxide (TiO2) in wastewater. Sep. Sci. Technol. 42, 1775e1792. Kosmulski, M., 2009. Compilation of PZC and IEP of sparingly soluble metal oxides and hydroxides from literature. Adv. Colloid Interface Sci. 152, 14e25. Li, F., Yuasa, A., Ebie, K., Azuma, Y., Hagishita, T., Matsui, Y., 2002. Factors affecting the adsorption capacity of dissolved organic matter onto activated carbon: modified isotherm analysis. Water Res. 36, 4994-4604. Liu, S., Lim, M., Fabris, R., Chow, C., Drikas, M., Amal, R., 2008. TiO2 photocatalysis of natural organic matter in surface water: impact on trihalomethane and haloacetic acid formation potential. Environ. Sci. Technol. 42, 6218e6223. Liu, W., Sun, W., Borthwick, A., Ni, J., 2013. Comparison on aggregation and sedimentation of titanium dioxide titanate nanotubes and titanate nanotubes-TiO2: influence of pH, ionic strength, and natural organic matter. Colloids Surfaces A Physicochem. Eng. Aspects 434, 319e328. Liu, S., Lim, M., Amal, R., 2014. TiO2-coated natural zeolite: rapid humic acid adsorption and effective photocatalytic regeneration. Chem. Eng. Sci. 105, 46e52. Malato, S., Fernandez-Ibanez, P., Maldonado, M.I., Blanco, J., Gernjak, W., 2009. Decontamination and disinfection of water by solar photocatalysis: recent overview and trends. Catal. Today 147, 1e59. Mwaanga, P., Carraway, E.R., Schlautman, M.A., 2014. Preferential sorption of some natural organic matter fractions to titanium dioxide nanoparticles: influence of
pH and ionic strength. Environ. Monit. Assess. 186, 8833e8844. Nanotechnology Characterization Laboratory, 2009. NCL Method PCC-2: Measuring Zeta Potential of Nanomaterials. National Cancer Institute, U.S. National Institutes of Health. Ng, M., Kho, E.T., Liu, S., Lim, M., Amal, R., 2014. Highly adsorptive and regenerative magnetic TiO2 for natural organic matter (NOM) removal in water. Chem. Eng. J. 246, 196e203. Philippe, K.K., Hans, C., MacAdam, J., Jefferson, B., Hart, J., Parsons, S.A., 2010. Photocatalytic oxidation, GAC, and biotreatment combinations: an alternative for the coagulation of hydrophilic rich waters? Environ. Technol. 31, 1423e1434. Qi, S., Schideman, L.C., 2008. An overall isotherm for activated carbon adsorption of dissolved organic matter in water. Water Res. 42, 3353e3360. Shahbeig, H., Bagheri, N., Ghorbanian, S., Hallajisani, A., Poorkarimi, S., 2013. A new adsorption isotherm model of aqueous solutions on granular activated carbon. World J. Model. Simul. 9, 243e254. Shi, H., Magaye, R., Castranova, V., Zhao, J., 2013. Titanium dioxide nanoparticles: a review of current toxicological data. Part. Fibre Toxicol. 10, 15. Summers, R., Roberts, P., 1988. Activated carbon adsorption of humic substances: heterodisperse mixtures and desorption. J. Colloid Interface Sci. 122, 367e381. Summers, R.S., Hooper, S.M., Shukairy, H.M., Solarik, G., Owen, D., 1996. Assessing DBP yield: uniform formation conditions. J. Am. Water Works Assoc. 88, 80e93. Sun, D.D., Lee, P.F., 2012. TiO2 microsphere for the removal of humic acid from water: complex adsorption mechanisms. Sep. Purif. Technol. 91, 30e37. Wassink, J.D., Andrews, R.C., Peiris, R.H., Legge, R.L., 2011. Evaluation of fluorescence excitation-emission and LC-OCD as methods of detecting removal of NOM and DBP precursors by enhanced coagulation. Water Sci. Technol. Water Supply 11, 621. Wiszniowski, J., Robert, D., Surmacz-Gorska, J., Miksch, K., Weber, J.-V., 2002. Photocatalytic decomposition of humic acids on TiO2, Part I: discussion of adsorption and mechanism. J. Photochem. Photobiol. A Chem. 153, 267e273. rube , P.R., 2013. Surface shear stress and membrane Wray, H.E., Andrews, R.C., Be fouling when considering natural water matrices. Desalination 330, 22e27. Yang, Y., Westerhoff, P., 2014. Presence in, and release of, nanomaterials from consumer products, nanomaterials. In: Capco, D.G., Chen, Y. (Eds.), Advances in Experimental Medicine and Biology, 811. Springer Science þ Business Media, Berlin. Zhang, Y., Chen, Y., Westerhoff, P., Crittenden, J., 2009. Impact of natural organic matter and divalent cations on the stability of aqueous nanoparticles. Water Res. 43, 4249e4257.
Contents lists available at ScienceDirect
Chemosphere journal homepage: www.elsevier.com/locate/chemosphere
Adsorption of natural organic matter and disinfection byproduct precursors from surface water onto TiO2 nanoparticles: pH effects, isotherm modelling and implications for using TiO2 for drinking water treatment Stephanie L. Gora*, Susan A. Andrews Department of Civil Engineering, University of Toronto, 35 St. George Street, Toronto, Ontario M5S 1A4, Canada
h i g h l i g h t s TiO2 photocatalysis increased the DBPfp of surface water at short treatment times but reduced it at longer treatment times. In the absence of irradiation, TiO2 nanoparticles adsorbed NOM and DBP precursors from natural surface water. The adsorption of NOM, THM precursors and HAA precursors by TiO2 nanoparticles was modelled using a simple isotherm model. The adsorption of DOC, UV254, humic substances and DBP precursors occurred more readily at pH 4 than at pH 6 or pH 8.
a r t i c l e i n f o
a b s t r a c t
Article history: Received 4 November 2016 Received in revised form 22 January 2017 Accepted 24 January 2017 Available online 2 February 2017
Titanium dioxide is a photocatalyst that can remove organic contaminants of interest to the drinking water treatment industry, including natural organic matter (NOM) and disinfection byproduct (DBP) precursors. The photocatalytic reaction occurs in two steps: adsorption of the contaminant followed by degradation of the adsorbed contaminant upon irradiation with UV light. The second part of this process can lead to the formation of reactive intermediates and negative impacts on treated water quality, such as increased DBP formation potential (DBPfp). Adsorption alone does not result in the formation of reactive intermediates and thus may prove to be a safe way to incorporate TiO2 into drinking water treatment processes. The goal of this study was to expand on the current understanding of NOM adsorption on TiO2 and examine it in a drinking water context by observing NOM adsorption from real water sources and evaluating the effects of the resulting reductions on the DBPfp of the treated water. Bottle point isotherm tests were conducted with raw water from two Canadian water treatment plants adjusted to pH 4, pH 6 and pH 8 and dosed with TiO2 nanoparticles. The DOC results were a good fit to a modified Freundlich isotherm. DBP precursors and liquid chromatography with organic carbon detection NOM fractions associated with DBP formation were removed to some extent at all pHs, but most effectively at pH 4. © 2017 Elsevier Ltd. All rights reserved.
Handling Editor: W. Mitch Keywords: Titanium dioxide Advanced oxidation Natural organic matter Disinfection byproducts Adsorption Drinking water
1. Introduction Nanomaterials, defined as materials with any dimension in the nanoscale or having internal or surface structure in the nanoscale (ISO, 2010), are increasingly being used in the fields of electronics, computing and medicine. Some nanomaterials, including titanium
* Corresponding author. E-mail addresses: [email protected] (S.L. Gora), sandrews@civ. utoronto.ca (S.A. Andrews). http://dx.doi.org/10.1016/j.chemosphere.2017.01.125 0045-6535/© 2017 Elsevier Ltd. All rights reserved.
dioxide (TiO2) nanoparticles, may also prove to be useful in environmental applications, including the treatment of drinking water and wastewater.
1.1. Titanium dioxide for drinking water treatment The use of TiO2 for drinking water purification has been explored by many researchers, and at least two companies have developed small-scale systems on the basis of TiO2 photocatalysis, but it has yet to be widely applied for municipal drinking water
364
S.L. Gora, S.A. Andrews / Chemosphere 174 (2017) 363e370
treatment. Photocatalytic degradation of aqueous contaminants by TiO2 is generally thought to occur in two steps: adsorption and degradation. Adsorption can occur in the absence of light, but degradation occurs only when TiO2 is irradiated. Upon irradiation, reactive oxygen species (ROS), including the hydroxyl radical, and oxidative and reducing centres on the surface of the nanoparticle degrade contaminants that have been adsorbed on the surface of the photocatalyst. This two-step reaction is often described using the LangmuireHinshelwood mechanism (Malato et al., 2009). The two-stage nature of photocatalytic degradation differentiates it from other advanced oxidation processes such as UV/H2O2 and O3/ H2O2. Like other AOPs, however, many ROS formed when TiO2 is irradiated with UVA light are non-specific oxidants; hence, TiO2 photocatalysis has the potential to provide concurrent disinfection and degradation of organic drinking water contaminants, including taste and odour compounds and cyanotoxins (Fotiou et al., 2015), various pharmaceuticals (Avisar et al., 2013; Kanakaraju et al., 2014), and disinfection byproduct (DBP) precursors such as natural organic matter (NOM) (Liu et al., 2008; Huang et al., 2008; Philippe et al., 2010). The photocatalytic degradation process preferentially targets large aromatic NOM compounds, breaking them down into smaller ones (Huang et al., 2008; Philippe et al., 2010). This can result in decreased membrane fouling (Huang et al., 2008) and changes in DBP formation potential (DBPfp). The latter is of particular concern because although some of the degradation products of photocatalysis may be more likely to form DBPs than the original compounds, others may have an equal or greater DBPfp, particularly at shorter treatment times (Liu et al., 2008). As a result, an adsorption-based process may prove to be a safer and more effective option for the removal of NOM from drinking water using TiO2.
1.3. Adsorption of NOM to TiO2 Studies by Mwaanga et al. (2014), Erhayem and Sohn (2014) and Kim and Shon (2007) have established that pH and ionic strength affect the adsorption of NOM on TiO2 and have noted that larger, more aromatic NOM compounds are adsorbed preferentially. The presence of bicarbonate, phosphate and nitrate has been shown to decrease NOM adsorption on TiO2 while the presence of magnesium and calcium increases the likelihood of NOM adsorption (Erhayem and Sohn, 2014; Sun and Lee, 2012). Other studies have described the effects of individual ions on the agglomeration of TiO2 nanomaterials. Agglomeration decreases the overall surface area available for adsorption and as such is likely to have an impact on the ability of TiO2 nanomaterials to adsorb NOM. Liu et al. (2013) reported that three types of TiO2 nanomaterials were more likely to agglomerate under high-ionic strength conditions than at low-ionic strength conditions. This finding is corroborated by those of Erhayem and Sohn (2014). The type of ions present in solution may also have an effect. Liu et al. (2013) observed greater increases in agglomerate size when calcium was added to the water rather than sodium. They hypothesized that this was because of the greater ability of Ca2þ to compress the electrical double layer surrounding the nanomaterials relative to Naþ. Greater compression of the electrical double layer results in less repulsion between individual nanoparticles and thus, greater agglomeration. The presence of NOM increases the stability of nanomaterials in solution, though this effect is less pronounced in the presence of ions such as calcium (Liu et al., 2013; Zhang et al., 2009) and at high NOM concentrations (Erhayem and Sohn, 2014). According to Zhang et al. (2009), NOM inhibits agglomeration by increasing the overall negative charge of the particles, thus increasing the repulsive forces that keep them dispersed in solution.
1.2. Natural organic matter The removal of NOM, a heterogeneous mixture of organic compounds found in most surface water sources, is an important objective of drinking water treatment because of its aesthetic, operational and health effects. The latter are primarily linked to the role of NOM as a precursor to form both regulated and unregulated DBPs. Operational concerns related to NOM include membrane fouling, competitive adsorption to adsorbent materials intended for the removal of taste and odour compounds, interference with UV disinfection and consumption of chlorine during disinfection. NOM is commonly quantified as total organic carbon (TOC) or dissolved organic carbon (DOC), bulk parameters that measure the total mass of organic carbon compounds present in a water sample without differentiating them from one another. UVeVis absorbance at 254 nm (UV254) and fluorescence are also used to characterise NOM, though these methods are specific to NOM compounds containing aromatic chromophores or fluorophores. More complex methods such as liquid chromatography with organic carbon detection (LC-OCD) have been developed that allow researchers to separate NOM into different fractions based on size or chemical characteristics. The resulting fractions are then quantified using DOC or UV254. LC-OCD separates NOM into five fractions based on size and/or chemical characteristics as follows: biopolymers, associated with membrane fouling (Wray et al., 2013); humic substances, which have been linked to DBP formation (Wassink et al., 2011); building blocks; low-molecular-weight acids (LMWAs); and low-molecular-weight neutrals (LMWNs) (Huber et al., 2011). A given NOM sample's potential to form regulated and unregulated DBPs can also be assessed more directly by measuring its DBPfp by chlorinating water samples and measuring the DBPs formed under standardised conditions.
1.4. Adsorption models Adsorption processes are usually evaluated in the laboratory using adsorption isotherm models. The Freundlich isotherm often fits well to empirical data and can be used to model heterogeneous systems such as the adsorption of organic molecules to activated carbon (Summers and Roberts 1988). It can be used to describe multilayer adsorption, reversible adsorption and adsorbents with non-uniform adsorption sites (Shahbeig et al., 2013). Summers and Roberts (1988) found that a modified version of the Freundlich isotherm could be used to describe the adsorption of NOM to activated carbon when experiments were conducted with a constant initial concentration of NOM and changing doses (D) of activated carbon. They also developed the following equation (referred as the SR model in this paper) to express this relationship:
qe ¼ KSR
Ce D
1 nSR
(1)
where D has units of mg L1 or g L1. The SR model was used and extended upon by numerous researchers including Karanfil et al. (1999), Li et al. (2002), Hyung and Kim (2008) and Qi and Schideman (2008) to characterise the adsorption of NOM to activated carbon and carbon nanotubes. Erhayem and Sohn (2014) modelled the adsorption of Suwannee River humic acids, fulvic acids, and NOM to P25 TiO2 nanoparticles using the SR isotherm model. They noted that the SR adsorption constant (and thus the extent of adsorption) increased at lower pH and at higher ionic strength.
S.L. Gora, S.A. Andrews / Chemosphere 174 (2017) 363e370
1.5. Potential risks and opportunities associated with the use of TiO2 nanoparticles for water treatment Conventional water treatment technologies such as coagulation and activated carbon are effective for NOM removal but, like all treatment technologies, have limitations. In North America, coagulation is widely used to remove NOM and turbidity from drinking water, but it does not remove some recalcitrant organics. It also creates a substantial amount of waste, often referred to as coagulation residuals. Activated carbon readily removes NOM and other organic compounds but eventually becomes exhausted, and must undergo an expensive and energy intensive regeneration process if it is to be reused. Although TiO2 is more well known for its photocatalytic properties, it also adsorbs NOM as a part of that process, and is potentially regenerable onsite (Liu et al., 2014). As such, it might prove to be a useful alternative to the existing treatment options. TiO2 nanoparticles are not without health and environmental concerns. Inhalation is the route of exposure of the greatest concern for human health (Shi et al., 2013), but the transport of nanoparticles in the environment is coming under increasing scrutiny (Yang and Westerhoff, 2014). The adsorption of various water components, including NOM, on TiO2 can facilitate the latter's transport through water systems. To date, most of the studies on the interactions between NOM and TiO2 have aimed to elucidate these effects to predict and minimise the impact of nanomaterials on the natural environment (e.g. Mwaanga et al., 2014; Erhayem and Sohn, 2014; Kim and Shon, 2007; and Liu et al., 2013)). This study aims to expand on the existing research into the adsorption of NOM by TiO2, most of which have been conducted in a contaminant transport context. This study also explores the potential application of TiO2 as an adsorbent in drinking water treatment by studying its behaviour in natural surface water sources and measuring its ability to remove DOC, UV254 and DBP precursors.
2. Materials and methods 2.1. Materials Evonik Aeroxide P25 TiO2 nanoparticles were purchased from Sigma Aldrich (Canada) and used without further modification. THM and HAA standards were also purchased from Sigma Aldrich. DOC standards were prepared by dissolving potassium hydrogen phthalate into milliQ water to create a 1000-mg L1 stock solution, which was then diluted as required. Raw water was obtained from the inlets of two water treatment plants (WTPs) in Ontario, Canada, both of which use surface water supplies. The Peterborough WTP is supplied by the Otonabee River, while the R.C. Harris WTP in Toronto draws its water from Lake Ontario, one of the largest lakes in North America, which provides water to over 9 million people in Canada and the United States. All water samples were obtained ahead of any pre-chlorination point at the WTP and characterised upon return to the laboratory. Measured ranges for five relevant water parameters are provided in Table S1 in the supplemental file. The water sources varied primarily in terms of their NOM content and aromaticity. The DOC of the Otonabee River water was approximately 3e4 times higher than that of the Lake Ontario water, while its UV254 was approximately five times higher. The SUVA of the Otonabee River water ranged from 2.0 to 2.4 L/mg.m, while that of the Lake Ontario water ranged from 0.8 to 1.0 L/mg.m, indicating that the NOM in the former is more aromatic than that in the latter.
365
2.2. Analytical methods DOC was determined using an O.I. Analytical Aurora 1030W TOC analyzer operating in the persulphate oxidation mode and UV absorption at 254 nm (UV254) was analysed on an Agilent 8453 UVeVis analyzer. SUVA was calculated by normalizing UV254 by DOC. Alkalinity was measured using Standard Method 2320 (APHA, 2005). Duplicate raw and treated water samples were analysed using size exclusion LC-OCD as described by Huber et al. (2011). The results of the analyses were processed using proprietary software (ChromCalc, DOC-LABOR, Karlsruhe, Germany). The isoelectric point of the nanoparticles was determined by measuring the zeta potential of the nanoparticles at different pH values. A series of samples containing 0.1 g L1 TiO2 in 10 mM NaCl were adjusted to pHs ranging from 2 to 9 using 0.1 N HCl or 0.1 N NaOH as per the method outlined by the Nanotechnology Characterization Laboratory (2009). The zeta potential of the samples was measured using a Horiba Scientific Nanopartica SZ100 Nanoparticle Analyzer. The size of the nanoparticle agglomerates formed at different pHs was determined using a Malvern MasterSizer 3000. The uniform formation conditions (UFCs) method as described by Summers et al. (1996) was used to assess the chlorine demand and DBPfp of the raw water and the water that had been treated with TiO2. The UFC test was designed to mimic the conditions commonly found in distribution systems in North America. The chlorine demand and DBPfp tests were conducted on samples buffered with a borate solution and adjusted to pH 8 with 1 N HCl or 1 N NaOH. Samples were stored in the dark at 20 C for 24 h, after which the free chlorine residual was measured using Standard Method 4500-G (APHA, 2005). The DBPfp samples were dosed with sufficient sodium hypochlorite to ensure that they would have a chlorine residual of 1 ± 0.4 mg L-1 after the 24 h holding time. After 24 h, the trihalomethanes and haloacetic acids formed during the UFC tests were extracted according to Standard Method 6232 B and Standard Method 6251 B (APHA, 2005) and analysed on an Agilent 7890B GC-ECD [Standard Method 6232 B and Standard Method 6251 B (APHA, 2005)]. Blanks and 20 mg L-1 check standards were analysed after every 10 samples. All statistical analyses, including Tukey's method for multiple comparisons to establish a point of practical equilibrium and linear regression of the adsorption data to determine KF and KSR, were conducted at the 95% confidence level. Reported error values represent half of the calculated confidence interval unless otherwise specified. 2.3. Sample preparation NOM degradation studies were conducted in a high-intensity UV reactor equipped with UV LED lamps emitting UVA light at 365 cm with an average irradiance of 6.25 mW/cm2. Unchlorinated raw water from the Peterborough Water Treatment Plant (Otonabee River water) was dispensed into three 50-mL batch reactors, dosed with 0.25 g L-1 of P25 nanoparticles, allowed to mix in the dark for 1 min, and then exposed to the LED light for times ranging from 0 to 60 min. The treated samples were chlorinated according to the UFC method and analysed for THMfp and HAAfp as described in Section 2.2. The time required to reach a stable adsorption equilibrium between NOM and TiO2 nanoparticles was determined by adding 75 mL of raw unchlorinated water to duplicate 125 mL amber bottles and, when necessary, adjusting the pH to 4, 6, or 8 with 1 N HCl or 1 N NaOH. The bottles were dosed with 0.5 g L-1 of Evonik P25 TiO2 nanoparticles and mixed end-over-end in a box mixer for
366
S.L. Gora, S.A. Andrews / Chemosphere 174 (2017) 363e370
times ranging from 1 to 8 h. This time range was chosen because previous experiments (results not shown) had indicated that most NOM adsorption occurred within 5 min and that equilibrium likely occurred between 1 and 4 h. For the bottle point isotherm tests, eight 250 mL amber bottles were filled with 150 mL of raw water; adjusted to pH 4, 6, or 8; dosed with 0, 0.01, 0.025, 0.05, 0.1, 0.25, 0.5, or 1 g L-1 of Evonik P25 TiO2 nanoparticles; and then mixed continuously in the dark for 4 h in an end-over-end box mixer. All experiments were run in triplicate. All replicate samples were analysed for DOC, UV254 and SUVA. The samples from one replicate experiment were used to establish chlorine demand, and the remaining samples were analysed for THMfp and HAAfp. The DOC, THMfp and HAAfp results of the bottle point isotherm tests were evaluated for fit against the linearised Freundlich and SR models. All raw and treated samples were filtered through a 0.45-mm polyethersulphone (Pall) laboratory filter in a standard vacuum filtration apparatus to remove particulate matter and TiO2 ahead of DOC, UV254, DBPfp and LC-OCD analysis. 3. Results and discussion 3.1. DBP formation during photocatalysis During the irradiation tests, the THMfp and HAAfp of the raw Otonabee River water were modestly reduced by adsorption alone (15% and 10%) but both increased upon irradiation (Fig. 1). The impact on THMfp was particularly dramatic. After only 5 min of irradiation, THMfp increased by 61% relative to the control. After 30 min, THMfp began to decrease and after 60 min the THMfp of the treated water matched that of the control. The increase in HAAfp upon irradiation was smaller than that of THMfp and was reversed after 30 min of irradiation. After 60 min of irradiation, HAAfp was reduced by 35% relative to the control. These results, which are similar to those obtained by Liu et al. (2008) and Philippe et al. (2010), illustrate the risk associated with the use of photocatalysis for NOM and DBPfp reduction. Although both THMfp and HAAfp were eventually reduced by the treatment, both increased at treatment times between 0 and 15 min. Shorter treatment times are desirable as they minimise the amount of space and energy required at full scale. The modest reductions of THMfp and HAAfp through adsorption, however, suggest an alternative treatment optiondcould higher concentrations of TiO2 adsorb a sufficient amount of DBP precursors to provide a viable reduction in overall DBPfp and do certain water quality conditions (e.g. pH) favour the adsorption of DBP precursors to TiO2?
Fig. 1. THMfp and HAAfp of Otonabee River water treated with 0.25 g L-1 and irradiated by high intensity UVA-LED light.
3.2. NOM removal through adsorption e time series experiments The practical adsorption equilibrium, defined as the point at which the 95% confidence of neighbouring means began to overlap and the slope of the line of mean concentration vs. time could no longer be distinguished from zero, was determined on the basis of the results of the DOC and UV254 time series experiments conducted in each water source. Irrespective of the water type used or the parameter observed, the results indicated that the majority of NOM adsorption to the P25 TiO2 particles occurred within minutes and that a practical adsorption equilibrium was reached within 1 h (see Figs. S1eS3 in the supplemental file). This is similar to results obtained by some researchers working with TiO2 materials, including P25 nanoparticles (Kim and Shon, 2007; Ng et al., 2014), though others have suggested that a longer period of time might be required to reach full equilibrium (Mwaanga et al., 2014; Erhayem, 2013). On the basis of the results of the time series experiments conducted in this study, all subsequent equilibrium experiments were conducted with a 4-h adsorption period to ensure that all data were gathered at a point well beyond the practical point of equilibrium. The results of the time series experiments also indicated that adsorption was most effective on pH 4, and that UV254 and SUVA were more strongly affected by the treatment than DOC, hinting that aromatic NOM may have been preferentially adsorbed by the nanoparticles. The latter effect was more apparent in the Otonabee River water because it had an initial raw water SUVA that was two times that of the Lake Ontario water. 3.3. Effects of pH and TiO2 dose on adsorption Bottle point isotherm tests were conducted to further characterise the adsorption behaviour of the nanoparticles at the three pH conditions as the dose of TiO2 was varied from 0.01 to 1 g L-1. As shown in Fig. 2 and Fig. S4 in the supplemental file, at equilibrium, the DOC and UV254 removals observed in the Otonabee River and Lake Ontario samples were found to be pH dependent and in all cases, more NOM was removed at pH 4 than at pH 6 and pH 8. Irrespective of pH, increasing the dose of TiO2 added to the water resulted in a decrease in the amount of DOC remaining in the treated water. In both the Otonabee River (Fig. 2A) and Lake Ontario water trials (Fig. 2B), DOC removal increased quickly as the TiO2 dose was increased from 0.01 to 0.25 g L-1, but slowed thereafter, though no definitive plateau was reached at any pH, suggesting that further increases in TiO2 dose beyond the maximum applied in this study (1 g L-1) may have improved DOC removal even further. At pH 4 and 1 g L-1 of TiO2, the DOC of the Otonabee River water was reduced from 4.69 ± 0.12 to 1.10 ± 0.12 mg L-1 whereas that of the Lake Ontario water was reduced from 1.64 ± 0.05 to 0.69 ± 0.04 mg L-1. DOC removal from both water sources was statistically significantly lower at pH 6 and pH 8 than at pH 4. The UV254 results followed similar trends as DOC and are shown in Fig. 4. UV254 was removed more effectively than DOC, particularly from the Otonabee River water, where UV254 was reduced from 0.112 ± 0.002 to 0.013 ± 0.002 cm1, approximately 88%, by a 1 g L-1 dose of P25 nanoparticles at pH 4. UV254 is a measure of the aromaticity of the NOM present in water; hence, the results presented here suggest that aromatic NOM was preferentially removed over non-aromatic NOM during the adsorption process. As has been suggested by other researchers (Mwaanga et al., 2014), this pH dependence may be partially explained through charge interactions. Zeta potential measurements conducted on the P25 nanoparticles indicated that they had an isoelectric point (IEP) between pH 6 and pH 6.5, consistent with the literature (Kosmulski,
S.L. Gora, S.A. Andrews / Chemosphere 174 (2017) 363e370
367
Fig. 3. Size distribution of agglomerates of P25 nanoparticles in Otonabee River water adjusted to pH 4, pH 6 and pH 8.
Fig. 2. Adsorption of DOC from raw unchlorinated water from Otonabee River water (A) and Lake Ontario water (B) adjusted to pH 4, pH 6 and pH 8 and mixed with 0.5 g L1 of P25 TiO2 nanoparticles for 4 h.
2009). Thus, at pH 4, they were positively charged; at pH 8, they were negatively charged; and at pH 6, they were approximately neutral. At pH 4, most NOM compounds would have been neutral or slightly negatively charged and both charge and hydrophobic interactions likely contributed to adsorption. At pH 6, hydrophobic interactions between NOM compounds were likely the main contributors to adsorption. At pH 8, both the nanoparticles and the NOM were negatively charged and thus charge interactions would result in repulsion, rather than attraction, and any adsorption that occurred would have been attributable to hydrophobic interactions. Nanoparticle agglomeration and its effects on surface area may also have contributed to the changes in adsorption efficiency observed at different pHs. Agglomeration is most likely to occur when the pH is near the isoelectric point/point of zero charge of the material in question because at this pH repulsive forces between individual particles are at a minimum (Liu et al., 2013). P25 has an IEP of approximately 6.5, which means that the nanoparticles were more likely to agglomerate at pH 6 than at pH 4 or pH 8. Indeed, the particle size distributions presented in Fig. 3 indicate that the agglomerates formed by the nanoparticles were larger in Otonabee River water adjusted to pH 6 than in Otonabee River water adjusted to pH 4 or pH 8. In this study, better adsorption was observed at pH 4 than at pH 6. Agglomeration and the subsequent reduction in available surface area at pH 6 versus at pH 4 may have contributed to the poorer adsorption observed at pH 6, whereas charge repulsion between NOM and the TiO2 nanoparticles was more of a driver at pH 8. 3.4. Modelling of adsorption isotherms The DOC results of the bottle point isotherm tests were
Fig. 4. DOC data from Otonabee River water tests (A) and Lake Ontario water tests (B) fitted to the Summers and Roberts (SR) model.
modelled using the linearised forms of the Freundlich and SR models, as shown in Fig. 4 (SR model) and Fig. S5 of the supplement (Freundlich model). The isotherm parameters are summarised in Table 1 and Table S2 in the supplemental file. At all pHs and in both water sources, the SR model was a better fit to the data, defined as a higher R2 value, than the Freundlich model. The SR model is generally thought to provide a more accurate fit for data from highly heterodisperse systems and when the isotherms are developed using variable doses of adsorbent; hence, this result was not surprising. Other researchers have observed that aromatic NOM and humic acids are preferentially adsorbed to P25 nanoparticles over other types of NOM (Erhayem and Sohn, 2014). Given that the two water
368
S.L. Gora, S.A. Andrews / Chemosphere 174 (2017) 363e370
Table 1 Summary of isotherm parameters for the adsorption of NOM from Otonabee River water onto P25 TiO2 nanoparticles at pH 4, pH 6 and pH 8. Parameter
DOC 1/nSR KSR (mg DOC/g TiO2)11/n R2 THMfp 1/nSR KSR (mg THMfp/g TiO2)11/n R2 HAAfp 1/nSR KSR (mg HAAfp/g TiO2)11/n R2
Otonabee River
Lake Ontario
pH 4
pH 6
pH 8
pH 4
pH 6
pH 8
0.37 ± 0.03a 3.7 ± 1.1 0.98
0.49 ± 0.04 1.3 ± 1.2 0.97
0.47 ± 0.06 1.0 ± 1.2 0.94
0.58 ± 0.05 1.5 ± 1.1 0.98
0.57 ± 0.07 0.8 ± 1.2 0.95
0.59 ± 0.10 0.5 ± 1.3 0.90
0.39 ± 0.09 27 ± 2 0.91
0.42 ± 0.10 13 ± 2 0.90
0.42 ± 0.20 8±4 0.71
eb e e
e e e
e e e
0.39 ± 0.15 8±2 0.81
0.48 ± 0.22 4±3 0.76
0.69 ± 0.29 1±5 0.76
e e e
e e e
e e e
a
Error values represent half of the 95% confidence interval for each parameter. Since the concentrations of THMs and HAAs detected in the raw Lake Ontario water after chlorination at UFC conditions were near the analytical detection limit, THMfp and HAAfp isotherm experiments were not conducted with this water source. b
sources differ mainly in terms of their NOM concentration and aromaticity, it is not surprising that the KSR values obtained from the Lake Ontario tests were lower than those obtained from the Otonabee River tests. The higher 1/nSR values in the Lake Ontario tests also indicate that adsorption was less favourable in this water than in the Otonabee River water. Within each water source, 1/nSR was nearly constant irrespective of pH and KSR was larger at pH 4 than at pH 6 and pH 8, which once again indicates that the adsorption of NOM to P25 TiO2 nanoparticles was more effective at pH 4 than at pH 6 or pH 8 in both water sources. The pH 6 and pH 8 confidence intervals for KSR overlapped in both water sources, perhaps suggesting that pH became less of a driver of adsorption capacity when the pH of the water was equal to or higher than the IEP. These findings are in agreement with those of other researchers (Mwaanga et al., 2014; Erhayem and Sohn, 2014; Sun and Lee, 2012), which have demonstrated that NOM adsorption to P25 TiO2 nanomaterials occurs more readily at low pH than at high pH. They are also in agreement with studies that have demonstrated that NOM adsorption to TiO2 can be modelled using the Freundlich (Wiszniowski et al., 2002) and SR models (Erhayem and Sohn, 2014), though it should be noted that all the aforementioned studies made use of standardised NOM or humic acid isolates (e.g. IHSS) in synthetic water matrices rather than natural water sources. The KSR and 1/nSR results of this study and those of other TiO2 researchers are lower than those achieved by other groups working with activated carbon and nanoscale carbon adsorbents, but not dramatically so. In their original study, which was conducted with four NOM isolates and GAC doses similar to the TiO2 doses used in this study, Summers and Roberts (1988) observed KSR values ranging from 4.22 to 11.4 (mg C/g GAC)11/n and 1/nSR values ranging from 0.254 to 0.347. Karanfil et al. (1999) evaluated the adsorption of commercially available NOM isolates and NOM from natural water onto a series of commercially available and modified activated carbons. They observed KSR values ranging from 1.754 to 10.695 (mg C/g GAC)11/n in the natural water matrices. Hyung and Kim calculated KSR values ranging from 5.471 to 13.088 (mg C/g MWNT)11/n and 1/nSR values ranging from 0.212 to 0.384 when they evaluated the adsorption of commercially available NOM isolates onto multi-walled carbon nanotubes.
method; hence, the adsorption of DBP precursors to the P25 nanoparticles was not explored for this water source. The raw water and treated samples prepared during the Otonabee River adsorption tests were analysed for THMfp and HAAfp, and the results are shown in Fig. 5. As observed with DOC and UV254 removal, THMfp reduction through adsorption was pH dependent. Maximum THMfp reduction, 147 ± 16 mg L-1 to 38 ± 16 mg L-1 (74% reduction), was achieved at pH 4 and a TiO2 dose of 1 g L-1. Less removal was achieved at this dose at pH 6 (153 ± 11 mg L-1 to 75 ± 11 mg L-1, 51% reduction) and pH 8 (154 ± 15 mg L-1 to 104 ± 15 mg L-1, 34% reduction). HAA precursors were also removed through adsorption and this removal was pH dependent, though less so than for THM precursors. Fig. 5B shows that at the highest concentration of TiO2
3.5. Adsorption of DBP precursors Negligible amounts of THMs and HAAs were formed in the raw Lake Ontario water when it was chlorinated according to the UFC
Fig. 5. THMfp (A) and HAAfp (B) of Otonabee River water treated with increasing concentrations of P25 TiO2 nanoparticles at pH 4, pH 6 and pH 8.
S.L. Gora, S.A. Andrews / Chemosphere 174 (2017) 363e370
(1 g L-1) HAAfp was reduced from 39 ± 7 mg L-1 to 19 ± 5 mg L-1 (50% reduction) at pH 4 and from 36 ± 2 mg L-1 to 23 ± 3 mg L-1 (40% reduction) at pH 6. HAAfp reduction at pH 8 was not statistically significant at the 95% confidence level. Some of the variability in the HAA results can be explained by the fact that the UFC test, which was used to evaluate the THMfp and HAAfp of the raw and TiO2treated samples in this study, is conducted at pH 8, which does not favour the formation of HAAs. As a result, all the raw and TiO2treated samples had low HAAfp, making it difficult to isolate the effects of TiO2 adsorption on HAAfp removal, particularly at pH 8. The agreement between the THMfp and HAAfp datasets and the SR model are shown in Fig. S6, and the isotherm parameters are summarised in Table 1. Although the R2 values of the THMfp and HAAfp isotherms were lower than those of the DOC isotherms, the general trends indicate that, with the exception of HAAfp at pH 8, TiO2 could remove significant amounts of THM and HAA precursors from Otonabee River water through adsorption and that this adsorption could be modelled using the SR isotherm model. Nonetheless, as a whole, the isotherm parameters for the two classes of DBPs should be approached with caution because they were developed using a small dataset that contained substantial variation at low TiO2 doses. Additional experiments at higher TiO2 doses, using water sources with higher concentrations of DBP precursors, and/or employing chlorination regimes more likely to result in THM and HAA formation may help clarify the how well the SR model can predict the removal of DBP precursors from drinking water by TiO2 as well as the suitability of the model at different pHs.
3.6. Effects of pH on adsorption of LC-OCD fractions A selection of raw and TiO2-treated water samples from the Otonabee River experiment was analysed using LC-OCD to determine whether any specific fractions were being removed during adsorption and whether pH impacted the fractions adsorbed. As shown in Fig. 6, the biopolymers and humic substances fractions were targeted for adsorption at all pHs but most effectively removed at pH 4. The building blocks fraction was also removed to some degree at pH 4 but was essentially unaffected at pH 6 and pH 8. The LMWA and LMWN fractions were not adsorbed at any pH. These results indicate that, consistent with the findings of other researchers (Erhayem and Sohn, 2014), large and aromatic NOM compounds were preferentially adsorbed by TiO2 nanoparticles and help explain why U254, which is associated with the humic substances fraction, was sometimes removed more effectively than overall DOC.
369
4. Conclusions The results of this study show that during adsorption aromatic NOM (as measured by UV254) is preferentially removed over nonaromatic NOM and that the efficiency of NOM adsorption to TiO2 can vary by water source. They also demonstrate that TiO2 nanoparticles preferentially adsorb larger NOM molecules including the biopolymers and humic substances fractions. pH was shown to have a strong impact on the removal of NOM, including DBP precursors, from surface water by TiO2 nanoparticles. Specifically, more adsorption occurred at low pH than at higher pH. The poorer adsorption observed at pH 6 and pH 8 may be related to both agglomeration and charge repulsion at higher pH, with the former dominating at pH 6 and the latter at pH 8. A modified version of the Freundlich isotherm model (SR model) was found to provide an excellent fit to the DOC data gathered in this study. The resulting isotherm parameters were within but at the low end of the range usually observed during NOM adsorption to GAC and carbon nanomaterials, indicating that, particularly at neutral pH, the TiO2 nanoparticles were less effective than the adsorbents currently used in drinking water plants. Unlike TiO2, however, GAC is generally expensive and energy intensive to regenerate and the regeneration must usually be conducted offsite, whereas TiO2 is potentially regenerable and reusable in place. The THMfp and HAAfp datasets were also fitted to the SR model, with generally positive results. The results presented in this paper show that TiO2 adsorption is a viable way to remove NOM and DBP precursors from drinking water and that this removal can be modelled using simple isotherm models. The results also suggest that researchers hoping to design adsorption-based TiO2 processes should keep in mind that pH adjustment might be required to optimise performance. Acknowledgements The authors would like to acknowledge the training provided by Jim Wang, the laboratory assistance provided by Yijun (Jessie) Gai and Michelli Park, and the support of the Drinking Water Research Group at the University of Toronto. The authors are also grateful to Dr. Monica Tudorancea and Dr. Sigrid Peldzsus (University of Waterloo) for performing LC-OCD analyses. Funding was provided by the Natural Sciences and Engineering Research Council of Canada and the Ontario Ministry of Training, Colleges, and Universities. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.chemosphere.2017.01.125. References
Fig. 6. LC-OCD fractions present in raw unchlorinated Otonabee River water and water adjusted to pH 4, pH 6 and pH 8 and mixed with 0.5 g L-1 of P25 TiO2 nanoparticles for 4 h.
American Public Health Association, 2005. Standard Methods for the Examination of Water and Wastewater, twenty-first ed. APHA, Washington D.C. Avisar, D., Horovitz, I., Lozzi, L., Ruggieri, F., Baker, M., Abel, M.-L., Mamane, H., 2013. Impact of water quality on removal of carbamazepine in natural waters by Ndoped TiO2 photo-catalytic thin film surfaces. J. Hazard. Mater. 244e245, 463e471. Erhayem, M., 2013. Effect of Naturally Occurring Organic Matter (NOOM) Type and Source on NOOM Adsorption onto Titanium Dioxide Nanoparticles under Varying Environmental Conditions. Florida Institute of Technology, USA. Thesis. Erhayem, M., Sohn, M., 2014. Stability studies for titanium dioxide nanoparticles upon adsorption of Suwannee River humic and fulvic acids and natural organic matter. Sci. Total Environ. 468e469, 249e257. Fotiou, T., Triantis, T.M., Kaloudis, T., Hiskia, A., 2015. Evaluation of the photocatalytic activity of TiO2 based catalysts for the degradation and mineralization of cyanobacterial toxins and water off-odor compounds under UV-A, solar, and visible light. Chem. Eng. J. 261, 17e26. Huang, X., Leal, M., Li, Q., 2008. Degradation of natural organic matter by TiO2 photocatalytic oxidation and its effect on fouling of low-pressure membranes.
370
S.L. Gora, S.A. Andrews / Chemosphere 174 (2017) 363e370
Water Res. 42, 1142e1150. Huber, S.A., Balz, A., Abert, M., Pronk, W., 2011. Characterisation of aquatic humic and non-humic matter with size-exclusion chromatography-organic carbon detection and organic nitrogen detection (LC-OCD-OND). Water Res. 45, 879e888. Hyung, H., Kim, J.-H., 2008. Natural organic matter (NOM) adsorption to multiwalled carbon nanotubes: effect of NOM characteristics and water quality parameters. Environ. Sci. Technol. 42, 4416e4421. International Organization for Standardization, 2010. Nanotechnologies e Methodology for the Classification and Categorization of Nanomaterials. ISO/TR 11360:2010. Kanakaraju, D., Glass, B.D., Oelgemoller, M., 2014. Titanium dioxide photocatalysis for pharmaceutical wastewater treatment. Environ. Chem. Lett. 12, 27e47. Karanfil, T., Kitis, M., Kilduff, J.E., Wigton, A., 1999. Role of granular activated carbon surface chemistry on the adsorption of organic compounds 2. Environ. Sci. Technol. 33, 3225e3233. Kim, S.-H., Shon, H.K., 2007. Adsorption characterization for multi-component organic matters by titanium oxide (TiO2) in wastewater. Sep. Sci. Technol. 42, 1775e1792. Kosmulski, M., 2009. Compilation of PZC and IEP of sparingly soluble metal oxides and hydroxides from literature. Adv. Colloid Interface Sci. 152, 14e25. Li, F., Yuasa, A., Ebie, K., Azuma, Y., Hagishita, T., Matsui, Y., 2002. Factors affecting the adsorption capacity of dissolved organic matter onto activated carbon: modified isotherm analysis. Water Res. 36, 4994-4604. Liu, S., Lim, M., Fabris, R., Chow, C., Drikas, M., Amal, R., 2008. TiO2 photocatalysis of natural organic matter in surface water: impact on trihalomethane and haloacetic acid formation potential. Environ. Sci. Technol. 42, 6218e6223. Liu, W., Sun, W., Borthwick, A., Ni, J., 2013. Comparison on aggregation and sedimentation of titanium dioxide titanate nanotubes and titanate nanotubes-TiO2: influence of pH, ionic strength, and natural organic matter. Colloids Surfaces A Physicochem. Eng. Aspects 434, 319e328. Liu, S., Lim, M., Amal, R., 2014. TiO2-coated natural zeolite: rapid humic acid adsorption and effective photocatalytic regeneration. Chem. Eng. Sci. 105, 46e52. Malato, S., Fernandez-Ibanez, P., Maldonado, M.I., Blanco, J., Gernjak, W., 2009. Decontamination and disinfection of water by solar photocatalysis: recent overview and trends. Catal. Today 147, 1e59. Mwaanga, P., Carraway, E.R., Schlautman, M.A., 2014. Preferential sorption of some natural organic matter fractions to titanium dioxide nanoparticles: influence of
pH and ionic strength. Environ. Monit. Assess. 186, 8833e8844. Nanotechnology Characterization Laboratory, 2009. NCL Method PCC-2: Measuring Zeta Potential of Nanomaterials. National Cancer Institute, U.S. National Institutes of Health. Ng, M., Kho, E.T., Liu, S., Lim, M., Amal, R., 2014. Highly adsorptive and regenerative magnetic TiO2 for natural organic matter (NOM) removal in water. Chem. Eng. J. 246, 196e203. Philippe, K.K., Hans, C., MacAdam, J., Jefferson, B., Hart, J., Parsons, S.A., 2010. Photocatalytic oxidation, GAC, and biotreatment combinations: an alternative for the coagulation of hydrophilic rich waters? Environ. Technol. 31, 1423e1434. Qi, S., Schideman, L.C., 2008. An overall isotherm for activated carbon adsorption of dissolved organic matter in water. Water Res. 42, 3353e3360. Shahbeig, H., Bagheri, N., Ghorbanian, S., Hallajisani, A., Poorkarimi, S., 2013. A new adsorption isotherm model of aqueous solutions on granular activated carbon. World J. Model. Simul. 9, 243e254. Shi, H., Magaye, R., Castranova, V., Zhao, J., 2013. Titanium dioxide nanoparticles: a review of current toxicological data. Part. Fibre Toxicol. 10, 15. Summers, R., Roberts, P., 1988. Activated carbon adsorption of humic substances: heterodisperse mixtures and desorption. J. Colloid Interface Sci. 122, 367e381. Summers, R.S., Hooper, S.M., Shukairy, H.M., Solarik, G., Owen, D., 1996. Assessing DBP yield: uniform formation conditions. J. Am. Water Works Assoc. 88, 80e93. Sun, D.D., Lee, P.F., 2012. TiO2 microsphere for the removal of humic acid from water: complex adsorption mechanisms. Sep. Purif. Technol. 91, 30e37. Wassink, J.D., Andrews, R.C., Peiris, R.H., Legge, R.L., 2011. Evaluation of fluorescence excitation-emission and LC-OCD as methods of detecting removal of NOM and DBP precursors by enhanced coagulation. Water Sci. Technol. Water Supply 11, 621. Wiszniowski, J., Robert, D., Surmacz-Gorska, J., Miksch, K., Weber, J.-V., 2002. Photocatalytic decomposition of humic acids on TiO2, Part I: discussion of adsorption and mechanism. J. Photochem. Photobiol. A Chem. 153, 267e273. rube , P.R., 2013. Surface shear stress and membrane Wray, H.E., Andrews, R.C., Be fouling when considering natural water matrices. Desalination 330, 22e27. Yang, Y., Westerhoff, P., 2014. Presence in, and release of, nanomaterials from consumer products, nanomaterials. In: Capco, D.G., Chen, Y. (Eds.), Advances in Experimental Medicine and Biology, 811. Springer Science þ Business Media, Berlin. Zhang, Y., Chen, Y., Westerhoff, P., Crittenden, J., 2009. Impact of natural organic matter and divalent cations on the stability of aqueous nanoparticles. Water Res. 43, 4249e4257.