Sorptive removal of dissolved organic matter in biologically-treated effluent by functionalized biochar and carbon nanotubes: importance of sorbent functionality

The sorptive removal of dissolved organic matter (DOM) in biologically-treated effluent was studied by using multi-walled carbon nanotube (MWCNT), carboxylic functionalised MWCNT (MWCNT-COOH), hydroxyl functionalized MWCNT (MWCNT-OH) and functionalized biochar (fBC). DOM was dominated by hydrophilic fraction (79.6%) with a significantly lower hydrophobic fraction (20.4%). The sorption of hydrophobic DOM was not significantly affected by the sorbent functionality (~10.4% variation) and sorption capacity followed the order of MWCNT > MWCNT-COOH > MWCNT-OH > fBC. In comparison, the sorption of hydrophilic fraction of DOM changed significantly (~37.35% variation) with the change of sorbent functionality with adsorption capacity decreasing as MWCNT-OH > MWCNT-COOH > MWCNT > fBC. Furthermore, the affinity of adsorbents toward a hydrophilic compound (dinitrobenzene), a hydrophobic compound (pyrene) and humic acid was also evaluated to validate the proposed mechanisms. The results provided important insights on the type of sorbents which are most effective to remove different DOM fractions.


Introduction
Dissolved organic matter (DOM) with a wide variety of chemical compositions and molecular sizes is ubiquitous in different types of water (Fu et al., 2017;Shimabuku et al., 2017). DOM is originated from plant litter, soil, humus, microbial biomass (with aquagenic and pedogenic sources) degradation, root exudates, and living or decayed vegetation (Conte et al., 2011;Fu et al., 2017). DOM has an average concentration of 0.5 to 10.0 mg L -1 in natural water (Genz et al., 2008). DOM anthropogenically impacted water and wastewater and plays an essential role in environmental and engineered aquatic systems (Fu et al., 2017;Shimabuku et al., 2017). For example, DOM can serve as an energy source for bacteria, attenuate light, and influence the fate and transport of contaminants. Although DOM is regarded as non-toxic, however, many problems may arise with the naturally coloured groundwater. These leads direct concern to taste and odour problems of water (Cornelissen et al., 2008;Genz et al., 2008). DOM is also responsible for membrane fouling; poor oxidation of iron and manganese; and biological instability of drinking water in distribution systems (re-growth) ( Golea et al., 2017). Furthermore, residual natural organic matter can promote bacterial regrowth and pipe corrosion in the drinking water distribution system (Song et al., 2009). The content of NOM (especially for the hydrophobic fraction) in surface waters is increasing rapidly due to change in natural environment (Levchuk et al., 2017). The water-derived DOM contains a low amount of phenolic and aromatic compounds, whereas soil-derived DOM contains higher lignin content and aromatic fraction (Fu et al., 2017). Based on hydrophobicity, the components in DOM can be divided into hydrophobic and hydrophilic fractions.
To remove DOM, most commonly used methods are coagulation and flocculation followed by sedimentation/flotation and granular media filtration. However, coagulation mainly removes hydrophobic fraction of DOM rather than the hydrophilic fraction (Sharp et al., 2006). Unfortunately, the residual DOM after coagulation generally have significant DBP formation potential and need additional treatments to remove the residual DOM (Sharp et al., 2006). It is recommended by the US EPA for the control of DBP precursors (Kim and Kang, 2008). Furthermore, Biological system can promote the microbial growth and biofilm formation in the biofilters, subsequently enhancing the removal of NOM. Biopolymers can be well removed by biofiltration system, whereas low removal was found for humic acid fractions such as humic substances, building blocks, low MW neutrals and low MW acids (Chen et al., 2016). Granular activated carbon (GAC) has been employed to remove DOM as well as taste and odour through direct competition column bed and mentioned that pore blockage occurred due to fouling by DOM (Summers et al., 2013). Carbon nanotubes (CNTs), with their high surface area, hydrophobicity, porosity, and rapid sorption kinetics, have been Therefore, materials characteristics, DOM molecular structure and composition, and the solution chemistry (e.g. pH, water temperature, and ionic strength) are prime factors affecting sorption of DOM (Hyung and Kim, 2008;Li et al., 2014;Wang et al., 2013). However, it is more difficult to explain how DOM characteristics influence its adsorption behaviour as DOM is composed of a mixture of heterogeneous compounds possessing highly different physical and chemical properties. Several studies have shown that DOM molecular size can determine its adsorption behaviour (Newcombe et al., 1997a;Newcombe et al., 1997b;Velten et al., 2011). In addition, DOM aromaticity and polarity can be important characteristics (Zietzschmann et al., 2015), yet it remains unclear which characteristics govern DOM adsorption.
It is generally expected that sorption of hydrophobic and hydrophilic organic compounds is affected by the increase or decrease of the sorbent hydrophobicity or hydrophilicity (Peng et al., 2017b;Teixidó et al., 2011). However, the comprehensive studies in the literature examining the adsorption of DOM fractions, based on sorbent functionality, by carbonaceous materials (CMs) are rare. Therefore, the main objective is to apply several CMs having unique functionalities (such as functionalized biochar (fBC), MWCNT, MWCNT-COOH and MWCNT-OH) to investigate the effect of functional groups for the adsorption of hydrophilic and hydrophobic fractions of DOM from membrane bioreactor (MBR) effluent. This will provide detailed knowledge on how the sorbents' surface functional groups influence the removal of different DOM fractions and what type of adsorbent is most effective for DOM removal.
Biochar was first prepared from Eucalyptus globulus wood via pyrolysis at 380 0 C under continuous nitrogen supply at 1 psi. Then produced biochar was washed with milli-Q water for several times and adjusted the solution pH to 7.0 and finally dried at furnace at 105 o C. fBC was prepared from biochar (produced at 380 o C) using phosphoric acid as activating and functionalized agent at 600 0 C. A detailed preparation procedure of fBC is reported in our previous studies (Ahmed et al., 2017a;2017c;).

Characteristics of MBR effluent
Municipal effluent was collected from a membrane bioreactor at Central Park, Sydney, Australia. After collection, the MBR effluent was filtered through 1.2 m glass fiber filter and physicochemical properties such as pH, turbidity, UV254, total organic carbon (TOC), chemical oxygen demand (COD), conductivity, alkalinity, dissolved oxygen, inorganic ions, and metal ions were measured as listed in Table 1.

Sorption experiments
The sorption of DOM and its fractions in MBR effluent was conducted in 250 mL conical fluxes at 25 o C in duplicate on an orbital shaker over 48 h at 120 rpm using different CMs.
Dosages of fBC, MWCNT, MWCNT-COOH and MWCNT-OH were selected based on removal capacity of all DOM fractions. Pristine biochar was also used for DOM removal but found with low removal efficacy of 7.2%. Therefore, pristine biochar was not utilized for further experiments. Initial and diluted MBR effluent samples were used for DOM isotherm study. Dilution of MBR effluent was carried out with MQ water with the same pH as MBR effluent. The control experiments without sorbent were also conducted. Initial and final pH and conductivity were measured.
The sorption of DNB, PYE and humic acid was performed individually. The initial concentrations of DNB, PYE and humic acid were 5.0 mg L -1 , 1.0 mg L -1 and 13.67 mg L -1 , respectively. These concentration were chosen based on sorbent types and removal efficacy.
The adsorbent dosage used was 50-70 mg L -1 . Control experiments were also performed under the same condition. After the sorption experiments, the supernatants were filtered through 0.45-μm (for DNB and PYE) and 1.2-μm (for DOM and humic acid) syringe filters before analysis.

Chemical analysis of MBR effluent
The concentration and the fractionation of DOM were carried out using liquid chromatography-organic carbon detector (LC-OCD). LC-OCD result also provides detail information on the hydrophilic and hydrophobic fractions of DOM together with quantitative and qualitative results regarding molecular size distribution of organics present in water and wastewater. As required, samples were pre-filtered using 0.45 μm cellulose nitrate membrane filters (Fisher Scientific, USA). UV254 absorbance analysis of treated and untreated MBR effluent at 254 nm was measured at room temperature using Shimadzu UV-visible spectroscopy instrument (UV-1700). Metal cation contents in raw and treated MBR effluent were analysed using inductive coupled plasma mass spectroscopy (ICP-MS 7900, Agilent Technologies, Japan). Inorganic anions were analysed for phosphate, chloride, nitrate, nitrite and sulphate using a Metrohm ion chromatography (IC) (model 790 Personal IC). The IC was equipped with an autosampler and conductivity cell detector. Na2CO3 (3.2 mmol L -1 ) and NaHCO3 (1.0 mmol L -1 ) were used as a mobile phase with 0.7 mL min -1 flow rate. The results provide a more robust understanding of the adsorption behaviour of DOM fractions on MWCNTs and fBC for MBR effluent treatment.
DNB concentration (as a representative hydrophilic compound) was analysed by a high performance liquid chromatography (HPLC) equipment with an auto-sampler and a UV detector. A reverse-phase Zorbax Bonus RP C18 column (5.0 μm, 2.1 1.50 mm) was used throughout for detection and quantification of HOCs. The volume of injection was 100 μL.
Mobile phase A was composed of acetonitrile and formic acid (99.9: 0.1) while mobile phase B was composed of Milli-Q water and formic acid (99.9: 0.1). The elution used 40% of A and 60% of B at a flow rate of 0.4 mL min -1 and maintained 6 min. The UV wavelength for the whole method was selected at 285 nm and switched to 240 nm from 4.00 min to 5.00 min for DNB detection. PYE concentration was measured using a UV-Vis spectroscopy with a specific wavelength of 240 nm. The DOM concentration of humic acid solution was measured by total organic carbon (TOC) analyser and using a UV-Vis spectroscopy with a specific wavelength of 254 nm. Any samples that could not be analysed immediately were stored at 4 0 C.

Physical characterizations of sorbents
The physicochemical characteristics of fBC and MWCNTs were extensively examined using Fourier transform infrared spectroscopy (FTIR, Miracle-10: Shimadzu), scanning electron microscopy (SEM, Zeiss Evo-SEM system), X-ray diffraction (XRD) analysis, and zeta potential measurement. SEM was used to determine the morphological properties of CMs.
XRD analysis of the samples was carried out using a Bruker D8 Discover diffractometer using Cu Kα radiation, in the scattering angle 2θ range 20°-60°. FTIR was used to determine surface functional groups. The iso-electric values (zeta potential) of fBC and MWCNTs (dosage, 40-70 mg L -1 ) were measured by suspending into 1 mM KCl solution in the pH range of 1.45-10.20 separately using a Zetasizer Nano instrument (Nano ZS Zen3600, Malvern, UK). Samples were pre-equilibrated for ~48 h. Measurements in triplicate (average 30 scans with settling time of 5 s) were made to minimise undesirable biases, and the average value with standard deviation was used for data analyses.

Modelling of sorption data
Adsorption capacity, i.e. solid phase sorption (qs, g g -1 ) of DOM was calculated using following equations.
where C0 and Ce are the initial and equilibrium concentration of DOM (mg L -1 ), m is the mas of sorbent used (g) and V is the volume of solution (L).
Isotherm models employed to fit the adsorption isotherms are as follows: where qmax is the maximum adsorption capacity (mg g -1 ) and KL is the Langmuir fitting parameter (L mg -1 ). Parameters were estimated by nonlinear regression weighted by the dependent variable.
The Freundlich model: = where qs is the solid-phase sorbed capacity (mg g -1 ) of DOM fractions, n is a dimensionless number related to surface heterogeneity, and Kf is the Freundlich affinity coefficient (mg 1-n L n g -1 ). Ce represents the aqueous-phase concentration of solute (mg L -1 ) at 25 0 C. All model equations were fitted by origin-pro, with model parameters being obtained with a standard coefficient of determination (r 2 ) and adjusted coefficient of determination (radj 2 ).

Characteristics of sorbents
All three MWCNTs have randomly distributed surface and grooves. MWCNTs appear to form bundles or aggregates driven by van der Waals forces due to their high hydrophobicity.
These surface, groove, interstitial and inner areas on MWCNTs offer many sites for adsorption of organic chemicals .
XRD spectra of fBC consists of amorphous structure, and broad peak appeared at 25-

Adsorption affinity of DOM and its fractions
Data on DOM adsorption have been fitted to two different isotherm models ( Table 3) and LMW acids (MW < 350) were 58. 50, 0.74, 1.14 and 0 mg L -1 , respectively.

Hydrophobic DOM removal by sorbents
The physical properties of MWCNTs play an important role in the adsorption of DOM.  Figure 2. PYE is a π-electron donor by virtue of its π-electron donor ability. The previous study showed that PYE was able to form a π-π complex between the π-electrons of benzene rings of PYE and active graphene layers on activated carbon due to electron-donor-acceptor (EDA) mechanism (Xiao et al.,

2015). Our study shows that adsorption of PYE was maximum for MWCNT followed by
MWCNT-COOH, MWCNT-OH and fBC (Figure 2). This can be explained by the graphene layer of MWCNT which is hydrophobic moiety and PYE is also hydrophobic properties.

Hydrophilic DOM removal by sorbents
Hydrophilic DOM fraction (79.6%) usually has more aliphatic carbon and nitrogenous compounds, such as carbohydrates, sugars and amino acids (Fu et al., 2017). The hydrophilic fraction is generally humic substances (57.2%, MW ~1000), comprised of humic acids (HA), fulvic acids (FA) and humin, which make up more than 50% of the total organic carbon in water (Cornelissen et al., 2008).

Humic substance removal by sorption
The number-average molecular weight (Mw, 734) and number-average molecular number (Mn, 378) of humic substances changed with sorbent type. Among different sorbents, MWCNT caused the largest reduction of Mn from 378 to 309, more than other adsorbents such as fBC (from 378 to 377), MWCNT-COOH (from 378 to 331) and MWCNT-OH (from 378 to 330).
The ratio of Mw to Mn (i.e. Mw/Mn) also changed simultaneously.  Higher adsorption of humic acid by MWCNTs than fBC was due to the mainly higher surface area of MWCNTs. However, maximum sorption capacity of MWCNT-OH and MWCNT-COOH should be due to EDA interactions and hydrogen bonds formation by the oxygenated functional groups. The absence of oxygenated functional groups in MWCNT did not facilitate the formation of hydrogen bond, hence leading to lower adsorption capacity of humic acid by MWCNT. Figure 5b shows the removal of dissolved organic nitrogen (DON) present in humic acid.
It was found that functionalized MWCNTs had higher sorption of DON than MWCNT. The presence of (hetero)aromatic amine cations can act as π acceptors in forming π + -π electron donor-acceptor (EDA) interactions with the π electron-rich, the polyaromatic surface of CMs (Xiao and Pignatello, 2015). Therefore, due to EDA interactions and hydrogen bonding capability of functionalized MWCNTs they can adsorb higher amount of DON.

Removal of LMW neutral compound by sorbents
MWCNT showed highest sorption of LMW neutral (2.16 mg g -1 ) followed by MWCNT-OH (1.93mg g -1 ), MWCNT (1.77 mg g -1 ) and fBC (0.61 mg g -1 ) (Figure 4). Slightly higher sorption of LMW neutral by MWCNT also indicated that the adsorption was similar to the adsorption of the hydrophobic compound. Therefore, the hydrophobicity and the hydrophilicity of surface functional groups in both adsorbent and adsorbate play a vital role in their sorption behaviour.

Removal of biopolymer and building block by sorbents
The removal of biopolymer and building blockers also followed the same trend as observed for humic substances removal, i.e. MWCNT-OH > MWCNT-COOH > MWCNT. However, sorption of biopolymers was found minimum compared to humic substances, LMW neutrals and building blockers. This might be due to their large molecular structure and their adsorption onto CMs requires greater potential and molecular arrangement toward the specific direction of CMs and on its functional groups. Moreover, CMs and their functional groups were unable to interact with biopolymer molecules therefore showing lower adsorption.
Among fractionations of hydrophilic DOM showed the general tread of humic substances > building blockers > LMW neutral substances > biopolymers. Table 4 shows the removal of metal ions by different CMs. It can be observed that all the CMs could remove different metal ions to some extent. This is desirable as MBR effluent had an initial pH of 7.94 and the iso-electric values of fBC and all MWCNTs were below 5.14. It is common that oppositely charged ions (i.e. negative surface of CMs and the positive surface of heavy metal ions) can interact together due to electrostatic interaction. Hence, the electrostatic interaction was the main cause for competitive adsorption of metal ions.

Conclusions
The maximum DOM sorption capacity was observed for functionalized sorbent MWCNT-OH

Conflict of Interests
The authors declare no competing financial interest.            Sorbent functionality had great influence on sorption of DOM fractions.

List of Figures
 Hydrophobic fraction was better removed by MWCNT than functionalized sorbents.
 Hydrophilic fraction was better removed by functionalized MWCNTs than MWCNT.
 EDA, H-bond and hydrophobic interactions were the main sorption mechanism.