w a t e r r e s e a r c h 5 7 ( 2 0 1 4 ) 1 7 2 e1 8 2

Available online at www.sciencedirect.com

ScienceDirect journal homepage: www.elsevier.com/locate/watres

Membrane scaling and flux decline during fertiliser-drawn forward osmosis desalination of brackish groundwater Sherub Phuntsho a, Fezeh Lotfi a, Seungkwan Hong b, Devin L. Shaffer c, Menachem Elimelech c, Ho Kyong Shon b,* a

Centre for Technology in Water and Wastewater, School of Civil and Environmental Engineering, University of Technology, Sydney (UTS), Broadway, NSW 2007, Australia b School of Civil, Environmental & Architectural Engineering, Korea University, 1, 5-ka, Anam-Dong, Sungbuk-Gu, Seoul 136-713, Republic of Korea c Department of Chemical and Environmental Engineering, Yale University, P.O. Box 208286, New Haven, CT 06520-8286, USA

article info

abstract

Article history:

Fertiliser-drawn forward osmosis (FDFO) desalination has been recently studied as one

Received 11 January 2014

feasible application of forward osmosis (FO) for irrigation. In this study, the potential of

Received in revised form

membrane scaling in the FDFO process has been investigated during the desalination of

27 February 2014

brackish groundwater (BGW). While most fertilisers containing monovalent ions did not

Accepted 16 March 2014

result in any scaling when used as an FO draw solution (DS), diammonium phosphate (DAP

Available online 27 March 2014

or (NH4)2HPO4) resulted in significant scaling, which contributed to severe flux decline. Membrane autopsy using scanning electron microscopy (SEM), energy-dispersive x-ray

Keywords:

spectroscopy (EDS), and x-ray diffraction (XRD) analysis indicated that the reverse diffu-

Forward osmosis

sion of DAP from the DS to the feed solution was primarily responsible for scale formation

Membrane scaling

during the FDFO process. Physical cleaning of the membrane with deionised water at

Inorganic fouling

varying crossflow velocities was employed to evaluate the reversibility of membrane

Desalination

scaling and the extent of flux recovery. For the membrane scaled using DAP as DS, 80e90%

Brackish groundwater

of the original flux was recovered when the crossflow velocity for physical cleaning was the same as the crossflow velocity during FDFO desalination. However, when a higher crossflow velocity or Reynolds number was used, the flux was recovered almost completely, irrespective of the DS concentration used. This study underscores the importance of selecting a suitable fertiliser for FDFO desalination of brackish groundwater to avoid membrane scaling and severe flux decline. ª 2014 Elsevier Ltd. All rights reserved.

* Corresponding author. E-mail addresses: [email protected], [email protected] (H.K. Shon). http://dx.doi.org/10.1016/j.watres.2014.03.034 0043-1354/ª 2014 Elsevier Ltd. All rights reserved.

w a t e r r e s e a r c h 5 7 ( 2 0 1 4 ) 1 7 2 e1 8 2

1.

Introduction

Growing water scarcity issues have been driving interest in the use of membrane processes, such as reverse osmosis (RO) and nanofiltration (NF), for seawater and brackish groundwater desalination and for wastewater reclamation to produce alternative water sources (Baker, 2010; Geise et al., 2010; Mi and Elimelech, 2010a; Elimelech and Phillip, 2011). However, inorganic scaling and organic fouling pose a significant challenge for the efficient operation of these pressure-based membrane processes. Scaling not only increases energy consumption, but it also increases the operation and maintenance costs and significantly shortens membrane life. Scaling occurs when the concentrations of some of the sparingly soluble salts in the feed solution reach supersaturation due to high product water recovery, and the salts crystallise directly on the membrane surface or crystallise in the bulk solution and deposit on the membrane (Mi and Elimelech, 2010a; Antony et al., 2011). Scaling in membrane processes occurs by surface crystallisation, bulk crystallisation, or both mechanisms, depending on the operating conditions (Gilron and Hasson, 1987; Hasson et al., 2001; Liu and Mi, 2012). The most common scaling salts during desalination of seawater or brackish water by RO/NF are calcium sulphate, barium sulphate, and calcium carbonate. However, phosphate scaling is a major issue when RO is applied for wastewater treatment (Zach-Maor et al., 2008; Chesters, 2009; Antony et al., 2011). The presence of phosphate ions can cause serious scaling problems for RO membranes, and effective anti-scaling agents have not been identified to prevent all types of phosphate scaling (Chesters, 2009; Antony et al., 2011). Forward osmosis (FO) is an emerging osmotic membrane process that has been studied for a wide variety of applications, including desalination and wastewater treatment, food processing, and power generation from salinity gradients (Cath et al., 2006; McGinnis and Elimelech, 2007; Charcosset, 2009). FO utilises the osmotic pressure gradient between a concentrated draw solution (DS) and a feed solution (FS) as a driving force to pull water molecules from the FS through a semi-permeable membrane (Cath et al., 2006). For desalination applications, the diluted DS is then further processed to separate potable water and reconcentrate the draw solution. Many of the promising applications of FO focus on the low fouling propensity of the process and the associated benefits for treating feed waters with high fouling potential (Zhao et al., 2012). Although the FO process can experience membrane scaling and fouling, the absence of hydraulic pressure in FO operation is advantageous in terms of fouling rate and cleaning efficiency. In fact, several studies have demonstrated that inorganic scaling and organic fouling are almost fully reversible by adopting simple physical cleaning/rinsing without the need for chemical cleaning reagents (Lee et al., 2010; Mi and Elimelech, 2010b,a). In some cases, the FO process has been observed to have less fouling potential or a reduced fouling rate compared to the RO process (Holloway et al., 2007; Cornelissen et al., 2008; Achilli et al., 2009; Zhang et al., 2012).

173

One of the practical applications of FO technology is for irrigation, in which the diluted fertiliser DS that contains desalinated water can be used directly for fertigation instead of requiring a separation process to reconcentrate the draw solution. Termed fertiliser-drawn forward osmosis (FDFO) desalination (Phuntsho et al., 2011b, 2012a; Phuntsho, 2012), this process can use commercially-available soluble fertilisers as draw solutions to generate very high osmotic pressures. The FDFO process has been recently investigated for desalination of both seawater (Phuntsho et al., 2011b; Phuntsho et al. 2012a) and brackish groundwater (Phuntsho et al., 2013a). Fertilisers are composed of compounds that contain major essential elements for plants, such N, P, K, Ca, Mg, and S. However, when such compounds are used as DS in FO, especially for desalination of BGW containing scaling precursor ions, ionic species of these essential fertiliser elements could adversely affect the process through membrane scaling. 2 Ca2þ, Mg2þ and PO3 4 /HPO4 ions are all scaling precursors that accelerate membrane scaling and fouling, thereby undermining the efficiency of RO desalination plants (Hatziantoniou and Howell, 2002; Li and Elimelech, 2004; Lee and Elimelech, 2006). In RO, these scaling precursors have been shown to interact with ions in the feed solution to promote membrane scaling, resulting in flux decline and ultimately undermining the process efficiency. Fouling in FO is influenced by the feed water quality, similar to pressure-based membrane processes such as RO and NF. However, unlike RO/NF processes, scaling and fouling in the FO process could be influenced by reverse draw solute flux from the DS to the FS. Reverse diffusion of draw solutes has been cited as one of the major challenges of the FO process because it not only results in economic losses from the cost to replenish the DS, but it also could complicate FS concentrate management and enhance membrane fouling potential. Reverse draw solute flux into the feed solution has been observed to influence colloidal and organic fouling (Lee et al., 2010; Boo et al., 2012). It also exacerbates cake-enhanced osmotic pressure on the membrane surface in contact with the feed solution, thereby reducing the effective driving force for permeation and causing flux decline (Lee et al., 2010). When the DS contains divalent ions, such as Mg2þ or Ca2þ, the reverse diffusion of these ions could result in interaction with dissolved organic matter present in the FS through a bridging effect, significantly affecting foulant cake formation and flux decline (Hatziantoniou and Howell, 2002; Lee and Elimelech, 2006). The effects of fertiliser draw solutions on membrane inorganic scaling during the FDFO process has not been studied. Higher feed recovery rates can be achieved in FDFO when high fertiliser DS concentrations are used, and hence, scaling may become a significant issue due to both feed solution super-saturation and reverse diffusion of fertiliser salts. The objective of this study was to investigate flux decline due to inorganic scaling of FO membranes during the desalination of brackish groundwater by the FDFO process. The major factors affecting membrane fouling during the FDFO process were evaluated, including FS and DS properties. The study also investigated the effects of physical cleaning on membrane flux recovery.

174

w a t e r r e s e a r c h 5 7 ( 2 0 1 4 ) 1 7 2 e1 8 2

Table 1 e Composition of the synthetic brackish groundwater (BGW) feed solutions with various total dissolved solids (TDS) concentrations. This composition simulates the BGW usually found at the Buronga salt interception scheme in the Murray-Darling Basin, Australia (Phuntsho, 2012; Phuntsho et al., 2013a,b). Osmotic pressure was calculated using OLI Stream Analyser 3.2. BGW10, BGW20 and BGW35 denote BGW with TDS concentrations of 7.824, 15.647 and 27.382 g/L, respectively. Concentration (g/L) /

CompoundsY

BGW10

BGW20

BGW35

3.713 1.794 0.134 0.317 3.947 0.094 7.824 5.35 7.72

7.426 3.588 0.268 0.634 7.895 0.189 15.647 10.56 7.63

13.000 6.280 0.470 1.110 13.820 0.330 27.382 18.56 7.33

NaCl Na2SO4 KCl CaCl2$2H2O MgCl2$6H2O NaHCO3 Total dissolved solids, TDS (g/L) Osmotic pressure, p (atm) pH

2.

Materials and methods

2.1.

Draw and feed solutions

Fig. 1 e A schematic diagram of the laboratory-scale forward osmosis (FO) system used in this study.

2.2.

Six different fertilisers were used as DS: potassium chloride (KCl), monoammonium phosphate or NH4H2PO4 (MAP), diammonium phosphate or ((NH4)2HPO4) (DAP), potassium nitrate (KNO3), ammonium sulphate or (NH4)2SO4 (SOA), and calcium nitrate or Ca(NO3)2.4H2O. All chemicals were reagent grade and were supplied by Sigma Aldrich, Australia. The DS were prepared by dissolving the fertiliser salts in deionised (DI) water with the aid of a magnetic stirrer. The feed solution consisted of a synthetic brackish groundwater (BGW), and its composition is shown in Table 1. This composition simulates the BGW usually found at the Buronga salt interception scheme (SIS) in the Murray-Darling Basin, Australia (MDB) (Phuntsho, 2012; Phuntsho et al., 2013a). All the compounds in Table 1 were reagent grade and supplied by Sigma Aldrich. Feed solutions with different total dissolved solids (TDS) concentrations were prepared. They were designated as BGW10, BGW20 and BGW35 and had TDS concentrations of 7.824, 15.647, and 27.382 g/L, respectively. More background information regarding the BGW composition can be found in our earlier publications (Phuntsho, 2012; Phuntsho et al., 2013a).

Forward osmosis membrane

A commercial flat-sheet cellulose triacetate (CTA) FO membrane (Hydration Technology Innovations or HTI, Albany, USA) was used in this study. The CTA membrane is made from cellulose acetate embedded in a polyester woven mesh, and the characteristics of this membrane are presented in Table 2. More information on the properties and characteristics of the CTA FO membrane can be found in other publications (McCutcheon et al., 2005; Cath et al., 2006; Tang et al., 2010). The pure water permeability coefficient of the FO membrane was determined in the RO mode (A ¼ 0.28194  0.008 mm/s/atm or 1.00  0.03 L m2 h1 bar1).

2.3. FO performance tests for membrane scaling/fouling and physical cleaning The performance tests for the FDFO process were conducted using a lab-scale FO unit consisting of an acrylic FO cell with internal dimensions of 7.7 cm length, 2.6 cm width, and 0.3 cm depth (effective membrane area of 2.002  103 m2) on both sides of the membrane. The schematic layout of the FDFO process is presented in Fig. 1, similar to the unit used in our earlier studies (Phuntsho et al., 2011b, 2012b, 2013b). Both the FS and DS were supplied at crossflow velocities of 8.5 cm/s (400 mL/min or Reynolds number Re 455) all under co-current

Table 2 e Physical and chemical properties of cellulose triacetate (CTA) membranes as provided by the manufacturer and from various literature sources. Sample CTA

Active layer material Cellulose triacetate

Pure water permeability A (L m2 h1 bar1)

NaCl rejection R (%)

Salt permeability B (107 m/s)

Membrane thickness (mm)

1.00  0.03

w93% at 10 bar

9.8

93  3

Operating pH 3e8

w a t e r r e s e a r c h 5 7 ( 2 0 1 4 ) 1 7 2 e1 8 2

flow and in FO mode (active layer facing FS). The temperature of the FS and DS in all cases was maintained at 25  C with the use of an automated heater/chiller control system connected to a water bath. All FO experiments were conducted in the batch mode of operation. The DS and FS were recycled back to their respective tanks after passing through the FO membrane cell. The initial volumes of both DS and FS were fixed at 2.0 L each. As the volume of DS increased due to permeation, its concentration decreased. Correspondingly, the volume of the FS decreased with time, thereby becoming more concentrated Water flux was measured online by placing the DS tank on a digital mass scale and automatically recording the change in mass over time due to permeate flux. A new FO membrane was used for each experiment, and the membrane system was stabilized for 20 min with DI water on both the FS and DS sides of the membrane. The initial baseline flux of the virgin membrane was then obtained using 1.0 M NaCl as DS and DI as FS under the operating conditions described above. The decline in the baseline flux curve is due to the dilution of the DS and concentration of the FS over time because of batch mode operation. After the baseline experiment, all FDFO scaling experiments were then conducted using different fertiliser solutions as DS and synthetic BGW as FS under the operating conditions described earlier. The concentration of the fertiliser DS and the TDS concentration of the synthetic BGW FS were varied depending on the need of the experimental conditions (characteristics of the BGW are shown in Table 1). Each FDFO scaling experiment was conducted for at least 20 h before the membrane cleaning procedure was performed. After the FDFO scaling experiment, physical membrane cleaning was performed to evaluate the flux recovery after the FO membrane was subjected to inorganic scaling. The DS and FS were replaced with DI water, and the membrane was subjected to similar operating conditions as the FDFO process (crossflow velocity of 8.5 cm/s or Re 455) for 20 min. Following this physical cleaning, the flux recovery was assessed by measuring the flux under the same conditions as the baseline experiment (1.0 M NaCl as DS and DI as FS). The percentage ratio of the recovered flux after cleaning to initial virgin baseline flux (normalised) was assessed as a flux recovery rate.

3.

The permeate fluxes as a function of operation time for the six fertiliser DS tested at 1 M concentration are presented in Fig. 2 using BGW10 as FS. Amongst the six fertiliser DS considered in this study, Ca(NO3)2, KCl, and (NH4)2SO4 showed the highest initial water fluxes, while the lowest initial water flux was observed for DAP, consistent with our earlier studies (Phuntsho et al., 2011b, 2012a, 2013a). Most fertilisers showed a small gradual decrease in the permeate flux with time, except for DAP, which experienced rapid flux decline. The slight gradual decrease in the permeate flux is attributed to the decrease in the driving force (osmotic pressure difference) with time due to dilution of the DS and concentration of the FS in batch mode operation. However, the sharp decline in permeate flux with DAP as DS indicates that membrane scaling occurred. After about 4 h of operation, the flux rapidly declined to almost zero. These results illustrate that long-term observation of the performance of the fertiliser DS is important, as some of the DS could promote membrane scaling and fouling of the membrane. Such significant flux decline was not observed during the short-term lab-scale experiments using BGW in our earlier studies (Phuntsho et al., 2013a). An autopsy of the fouled membrane with DAP as DS was conducted to study the composition of the membrane scaling layer formed during the FO experiment. Membrane autopsy involved both destructive and non-destructive observations of the fouled membrane, which can provide an understanding of scaling and fouling at the microscopic level (Li et al., 2003; Phuntsho et al., 2011a). Fig. 3 presents SEM images showing the formation of significant scaling on the membrane surface when DAP was used as DS. Scalants were observed on both the active layer and support layer surfaces, as shown in Fig. 3(a) and (c). However, the scaling layer formed on the active layer surface (facing the FS) was more significant. Scaling was also observed on the walls of the acrylic membrane cell facing

Results and discussion

3.1. Influence of draw solution (DS) properties on inorganic scaling Six different fertilisers were selected from our earlier studies to investigate their impacts on membrane scaling propensity during long-term operation of the FDFO process. Although these fertilisers were previously studied as DS for the desalination of BGW, the earlier experiments were mostly shortterm and did not observe any significant flux decline due to scaling (Phuntsho et al., 2013a). Other studies of the FDFO process did not consider BGW but instead focused on evaluating the suitability of fertilisers as draw solutes. In these studies, the performance experiments were conducted using either pure water or NaCl solutions as the feed solution (Phuntsho et al., 2011b, 2012a; Phuntsho, 2012).

175

Fig. 2 e Variation of permeate flux with time for six different fertiliser draw solutions (DS) during FDFO desalination of synthetic brackish groundwater feed solution (FS). Operating conditions are DS: 1.0 M; FS: BGW10; fertilisers: calcium nitrate (CAN), potassium chloride (KCl), ammonium sulphate (SOA), potassium nitrate (KNO3), monoammonium phosphate (MAP), diammonium phosphate (DAP).

176

w a t e r r e s e a r c h 5 7 ( 2 0 1 4 ) 1 7 2 e1 8 2

Fig. 3 e SEM images of CTA membrane after experiment with diammonium phosphate (DAP) as draw solution (DS) and synthetic brackish groundwater (BGW) as feed solution (FS). a) active layer of the membrane surface, (b) closer SEM image of the scaling layer formed on the membrane active layer, (c) both active layer and support layer of the FO membrane and (d) picture of the scales on the wall of the acrylic membrane cell facing the FS (taken using a normal digital camera). Experimental conditions are DS: 3 M DAP, FS: BGW35.

the FS (Fig. 3(d)). A closer inspection of the scaling layer reveals a flaky crystal layer formed on the membrane surface (Fig. 3(b)). The sharp flux decline observed when DAP was used as DS is attributed to the growth of this scaling layer on the membrane surface. There are two possible reasons for the severe membrane scaling observed during FDFO experiments with DAP as DS. The first relates to the super-saturation of ions in the FS, and the second relates to the reverse diffusion of draw solute ions that may interact with the feed solution ions to form scalants. In pressure-based membrane processes, such as RO and NF, scaling occurs due to complex phenomena that involve both crystal nucleation and transport mechanisms (Antony et al., 2011). Nucleation and growth of inorganic scales on the membrane surface, termed surface crystallisation, is a common problem for RO systems (Antony et al., 2011). Surface crystallisation occurs due to the super-saturation of scalant ions in the feed solution as permeate is extracted. The highest concentration of scalants occurs at the membrane surface during permeation, and this super-saturation promotes crystal nucleation and growth of sparingly-soluble salts (Gilron and Hasson, 1987; Hasson et al., 2001; Liu and Mi, 2012). Scaling in RO/NF therefore increases with increasing system recovery rates. A similar scaling mechanism may occur during the FDFO process because the BGW feed contains several scaling

2 precursor ions, such as Mg2þ, Ca2þ, HCO 3 , and SO4 , which could precipitate as scalants such as CaCO3, MgCO3 CaSO4. However, the fact that flux decline was not observed for all fertiliser DS as the FS was concentrated over time indicates that the specific properties of the DS also play an important role in membrane scaling. The DS can influence scaling on the membrane active layer by reverse flux of draw solute ions across the membrane from the DS to the FS. Reverse diffusion of certain draw solute ions such as Ca2þ and Mg2þ has also been found responsible for accelerating organic fouling by interacting with the colloids and organic matter (Lee et al., 2010; Boo et al., 2012). In order to identify the composition of the scaling layer, EDS analysis was conducted on the active layer of the scaled membrane (Fig. 4). The results show that the scaling layer is composed primarily of Na, Ca, Mg, and P. Other than P, all other elements detected by EDS analysis are already present in the FS and are related to the feed properties. The presence of P in the scaling layer on the active layer side of the FO membrane indicates that reverse diffusion of ions from the DAP or (NH4)2HPO4 DS may have played a role in the membrane scaling. Our earlier study with 1 M DAP as DS and 5000 mg/L NaCl as FS has shown that for every litre of water extracted by FO process, 0.26 g/L of P diffuse towards the FS from the DS under these conditions (Phuntsho et al., 2012a). The reverse diffusion of ammonium and phosphate ions from

w a t e r r e s e a r c h 5 7 ( 2 0 1 4 ) 1 7 2 e1 8 2

Fig. 4 e The EDS spectrum for the active layer of the membrane. Experimental conditions are DS: 3 M DAP, FS: BGW35.

177

DAP across the membrane could lead to reaction with the ions present in the FS, such as Mg2þ and Ca2þ. Phosphorous may combine with Ca and Mg to form sparingly soluble calcium phosphate and magnesium phosphate salts, and membrane scaling by these salts could be the cause of the severe flux decline observed during the FDFO experiments. Calcium phosphate and magnesium phosphate scaling is often a significant issue when the RO process is applied for wastewater treatment because of the presence of phosphate nutrients in the FS (Zach-Maor et al., 2008; Chesters, 2009; Antony et al., 2011). The combined effects of increased phosphate levels with high calcium and bicarbonate concentrations increase the likelihood of phosphate scale formation on the membrane (Chesters et al., 2007). There are several types of phosphate salts of calcium because phosphate can form different polymeric ions, and the most thermodynamically stable forms are calcium

Fig. 5 e XRD analyses of the scaling layer formed on the membrane surface showing the presence of (a) magnesium phosphate and (b) struvite. Experimental conditions are DS: 3 M DAP, FS: BGW35.

178

w a t e r r e s e a r c h 5 7 ( 2 0 1 4 ) 1 7 2 e1 8 2

hydroxyapatite (Ca5(PO4)3(OH) or HAP) and calcium fluoroapatite (Chesters et al., 2007). Hydroxyapatite has been found to cause serious inorganic scaling during the RO process (Antony et al., 2011). Calcium sulphate (CaSO4 or gypsum) is another potential scaling compound because the BGW FS contains both Ca2þ and SO2 4 ions. Gypsum is one of the most significant scaling compounds encountered in the RO process (Chesters, 2009; Greenlee et al., 2009; Tang et al., 2011) and also in the FO process (Mi and Elimelech, 2010a; Zhao and Zou, 2011). In order to identify the scaling compounds on the membrane, XRD analysis was conducted for the fouled membrane samples. Immediately after the FO experiment, the scaled membrane was removed from the FO cell and then rinsed thoroughly with the DI. The membrane was then dried overnight in a desiccator, and the thick and loose top layer of the scale was brushed off from the membrane active layer. This scalant layer was then ground into a uniform powder for XRD analysis. The XRD results (Fig. 5) indicate that the scalants on the membrane surface are primarily composed of magnesium phosphate (MgHPO4) (Fig. 5(a)) and magnesium ammonium phosphate (MgNH4PO4.6H2O or struvite) (Fig. 5(b)). These insoluble compounds are formed because of the reverse diffusion of ammonium and phosphate ions from the DAP ((NH4)2HPO4) DS towards the feed during FDF. The phosphate ions then react with the magnesium ions present in the FS to form sparingly-soluble magnesium phosphate salts. Struvite precipitation is one of the major operational problems during wastewater treatment (Williams, 1999), including for the RO treatment process (Chesters, 2009). A slight increase in FS pH (from 7.33 to 8.0) was observed after the FDFO process was started, which may have favoured the formation of struvite, as alkaline pH is a conducive environment for struvite for2þ mation besides other factors such as adequate NHþ 4 and Mg concentrations (Stratful et al., 2001). Calcium phosphate and calcium sulphate could also be responsible for scale formation. However, the XRD analysis could not clearly confirm the presence of calcium phosphate in the scaling layer, probably because of its very low composition in the layer of crystal scale. XRD was performed on the scalant layer after it was brushed from the membrane surface. Ca was clearly detected by EDS on the membrane surface after the loose scalant layer was removed, as shown in Fig. 4, which could indicate that formation of Ca scalants was more rapid than the formation of Mg scalants. Therefore, Ca scaling might have occurred more rapidly at the membrane surface while the Mg scales were formed later in time at the membrane surface or were deposited later on the membrane surface as a loose scaling layer. Neither the EDS analysis nor the XRD analysis could detect the presence of sulphur or sulphate, and hence, it does not appear that gypsum scaling was formed in this study. This is a reasonable conclusion given that the lab-scale FDFO experiments were conducted at feed recovery rates less than 5%, and, thus, super-saturation of CaSO4 is not expected. The results in Fig. 2 show that the severity of the flux decline increases with operation time, finally approaching zero permeate flux after w16 h of operation. This observation indicates that the scaling layer becomes more significant with time as result of continuous reverse diffusion of DAP from the

DS. The reverse draw solute flux is expected to affect the rate of scale formation and the severity of flux decline. The reverse draw solute flux increases when a higher DS concentration is used. Therefore, FDFO experiments were conducted to see how the DS concentration affects the scaling potential and the flux decline during the FDFO process using BGW as FS. Monoammonium phosphate (NH4H2PO4 or MAP) was used as a typical non-scaling fertiliser DS and compared with DAP as DS at various concentrations. It is interesting to note the difference between the performances of MAP and DAP, the two phosphate-containing fertiliser DS. Unlike FDFO desalination with DAP as DS, no severe flux decline was observed from scaling when MAP was used as DS. For the FDFO experiments with MAP as DS, the pH of the FS was observed to decrease from pH 7.3 to 5.4 (using 1 M MAP as DS with BGW35 as FS). However, for similar experiments with DAP as DS, the pH of the BGW35 FS increased from 7.3 to 8.9. This is understandable given that MAP has lower pKa values than DAP, and hence, MAP solution is slightly acidic (pH 4.06 at 1 M) while DAP is slightly basic (pH 8.04 at 1 M). The formation of phosphate precipitates of Ca and Mg are generally more conducive under alkaline pH condition (Harrison et al., 2011). Consequently, the absence of membrane scaling with MAP as DS is likely due to its lower pKa values that could lower the pH of the FS as a result of reverse draw solute flux. A separate set of experiments were conducted to study the reverse draw solute flux of MAP and DAP using both DI and NaCl solutions as FS. The results from this study (data not presented) confirmed the results from our earlier studies showing that MAP has significantly higher reverse flux than DAP (Phuntsho et al., 2011b, 2012a). To further study the absence of scaling by phosphate precipitates with MAP as DS, 1.08 g/L of MAP (concentration similar to the specific reverse solute flux of MAP) was mixed directly with the BGW35 FS. No precipitation was observed. However, the pH of the BGW FS decreased from 7.33 to 5.36 upon addition of MAP. When the pH of the BGW FS was adjusted to its original pH of 7.33 using NaOH, precipitation occurred in the solution. This study demonstrates that, despite its higher reverse solute flux, MAP has a lower scaling potential than DAP because of its lower pKa values and its tendency to lower the pH of the FS as a result of reverse solute flux. Fig. 6 shows the influence of DS concentration on the water flux during the FDFO process using BGW10 as FS. The increase in MAP concentration increases the water flux proportionately without any sudden or sharp flux decline as shown in Fig. 6(a). Similar observations were made with the other four fertiliser DS, and hence, only MAP is presented as a representative experimental result. However, DAP DS concentration showed a significant influence on the membrane scaling and flux decline as shown in Fig. 6(b). The induction time until membrane scaling is reduced as the DAP DS concentration is increased, and the extent of flux decline is generally greater with higher DAP DS concentrations. These results indicate that when higher DS concentrations are used, the reverse draw solute flux increases as it is a direct function of concentration gradient (DC) (Hancock and Cath, 2009; Phillip et al., 2010; Boo et al., 2012; She et al., 2012), and the rate of scale formation at the membrane surface also increases, resulting in more severe flux decline.

w a t e r r e s e a r c h 5 7 ( 2 0 1 4 ) 1 7 2 e1 8 2

Fig. 6 e Variation of permeate flux with time during FDFO desalination of synthetic brackish groundwater (BGW) at different draw solution (DS) concentrations using (a) MAP and (b) DAP as DS. Experimental conditions are DS: DAP/ MAP, FS: BGW10.

3.2. Influence of feed water properties on the water flux decline Membrane scaling in the FDFO experiments can be attributed to sparingly soluble Ca, Mg, and P salts. While the P was contributed from the DS due to reverse draw solute flux, Ca and Mg were already present in the FS, and they have been long recognised as the scaling precursors in salt rejecting membrane processes such as RO and NF (Gilron and Hasson, 1987; Chesters, 2009; Antony et al., 2011; Tang et al., 2011). Therefore, it is clear that the feed properties play a significant role in the formation of membrane scaling or fouling and the associated flux decline. The influence of feed properties on the flux decline was investigated using 1 M DAP as DS and by varying the TDS

179

concentration of the BGW FS and excluding Ca or Mg ions from the feed solution. The observed declines in permeate flux under these different conditions are presented in Fig. 7. As shown Fig. 7(a), the initial permeate flux was reduced, the flux decline was more rapid, and the time to induction of flux decline decreased when the TDS concentration of the BGW feed increased. When a BGW with a higher TDS concentration is used, the greater concentration of scaling ions, such as Ca and Mg, promotes more rapid scaling when combined with the reverse flux of DAP from the DS. In order to assess whether Ca or Mg ions are more responsible for membrane scaling with BGW FS and DAP DS, experiments were conducted that removed either Ca or Mg or both from the FS and replaced them with NaCl to maintain a similar TDS concentration. The results presented in Fig. 7(b) show that when both Ca and Mg ions are absent in the FS, the water flux exhibits only gradual decline attributed to decrease in the driving force due to batch operation of the FDFO system. These results confirm that Ca and Mg are primarily responsible for the formation of the scales during the FDFO process using DAP as DS. When either Mg or Ca ions were removed from the FS, permeate flux decline was still observed, although the flux decline was more severe when Mg ions were removed and only Ca was present as a precursor. This result indicates that Ca is responsible for more severe fouling than Mg during the FDFO process with DAP. XRD analysis of the membrane scaling layer formed for the BGW35 FS with Mg ions removed from the solution is shown in Fig. 7(c). The results indicate that the scaling layer is composed of hydroxyapatite or Ca10(PO4)5(OH)3 when magnesium was removed from the FS. Hydroxyapatite has very low solubility under alkaline pH, which promotes rapid scale formation and severe flux decline under these conditions. Furthermore, hydroxyapatite solubility is decreased in the presence of spectator ions such as Naþ (used to replace Mg2þ), which thereby contributed to more severe flux decline (Rootare et al., 1962). XRD analysis for scales formed after removing Ca2þ ions from FS indicates the presence of both magnesium phosphate and struvite in the scales. Given that both hydroxyapatite and struvite have minimum solubility at moderately alkaline pH (Harrison et al., 2011), it is difficult to correctly ascertain the actual cause of more severe flux decline with Ca than Mg. It is likely because, under alkaline pH of the FS, calcium phosphate is less soluble than struvite, and hence, its scale formation is more rapid than magnesium phosphate or struvite scales.

3.3. Influence of physical cleaning on permeate flux recovery The effectiveness of physical cleaning on the recovery of permeate flux in the FDFO process after scaling was studied. Before the membrane was subjected to the FDFO experiment, the baseline flux was obtained by using 1 M NaCl as DS and DI water as FS. After the FDFO experiments, the membranes were subjected to physical cleaning using DI water on both the feed and draw sides of the membrane at the same crossflow velocity as the fouling experiment (400 mL/min or 8.5 cm/s or Re 455 for 20 min). The baseline flux was again measured using 1 M NaCl as DS and DI water as FS to study the flux recovery.

180

w a t e r r e s e a r c h 5 7 ( 2 0 1 4 ) 1 7 2 e1 8 2

The percentage of water flux restored is expressed as a percentage ratio of the restored flux to the initial baseline flux (normalised) of the virgin membrane. The results of physical cleaning experiments for different fertiliser DS and for different concentrations of DAP as DS are presented in Fig. 8. The three different fertilisers subjected to physical cleaning experiments were calcium nitrate (CAN), which demonstrated the highest water flux in previous testing; KNO3, which demonstrated the lowest initial permeate flux in previous testing; and DAP, which resulted in severe membrane scaling. The results in Fig. 8(a) show that the physical cleaning is able to completely restore the permeate flux for both CAN and KNO3. This demonstrates that the scaling layer was either absent, or even it was formed, the physical cleaning was adequate to fully remove it. However, for DAP as DS, the permeate flux could be restored only up to about 80% when the normal cleaning regime of 400 mL/min (Re 455) for 20 min is applied. This shows that a higher crossflow velocity and/or longer cleaning duration are required to clean the membrane more effectively. Therefore, a higher crossflow velocity

Fig. 7 e Influence of feed solution (FS) properties on the permeate flux during FDFO desalination of synthetic brackish groundwater (BGW). The decline in permeate flux with operation time is shown for (a) BGW FS of different total dissolved solids (TDS) concentrations using 3 M DAP as draw solution (DS) and (b) BGW35 FS containing different cation compositions and using 3 M DAP as DS and (c) XRD analysis of the scaling layer formed when Mg2D ions were removed from the BGW35 FS.

Fig. 8 e Membrane flux recovery after membrane scaling (a) using three different types of DS at 3 M and BGW35 as FS and (b) using different concentrations of DAP as DS and BGW10 as FS. The cleaning regime was conducted at a crossflow velocity of 400 mL/min (Re 455) for 20 min except for DAP (1600), which was conducted at 1600 mL/min (Re 1820) crossflow velocity.

w a t e r r e s e a r c h 5 7 ( 2 0 1 4 ) 1 7 2 e1 8 2

(1600 mL/min or Re 1820 under laminar flow condition) was adopted for physical cleaning of the membrane scaled during the FDFO process using DAP as DS (maintaining the same cleaning duration of 20 min). Higher flow rate was not adopted because of the limitations with the pump used for this particular experiment. Fig. 8(a) shows that the permeate flux recovered almost completely when a higher crossflow velocity was adopted for physical cleaning. In the next cleaning experiment, the influence of cleaning on the permeate flux recovery was evaluated for FO membranes fouled to different degrees by operating at different DAP DS concentrations. Fig. 8(b) illustrates that when a higher DAP DS concentration was used, the permeate flux recovery was reduced compared to when a lower DAP DS concentration was used. These results are consistent with the observation of more severe membrane scaling and associated flux decline when using higher DS concentrations, as illustrated in Fig. 6(b). Reduced flux recovery after conditions of more severe membrane scaling indicate that the scalant layer formed on the membrane could be more extensive and more difficult to remove by physical cleaning. However, irrespective of the DAP DS concentrations, and the associated severity of flux decline due to membrane scaling, almost complete flux recovery was achieved when a higher crossflow rate or Re was adopted (data not shown). All the above results indicate that the reverse flux of draw solutes causes scaling on the feed side of the membrane, which can result in severe flux decline. However, permeate flux may be fully recovered by adopting appropriate physical cleaning regime.

4.

Conclusions

In this study, inorganic scaling and flux decline in the FDFO process during desalination of brackish groundwater were investigated. The study also evaluated the effectiveness of physical cleaning on the membrane flux recovery. The following conclusions are drawn from this study:  When FDFO is applied for the desalination of brackish groundwater containing multivalent ions, it is important to consider the influence of the reverse draw solute flux on membrane scaling and associated permeate flux decline.  Reverse diffusion of draw solutes plays a significant role in membrane scaling and flux decline during the FDFO desalination of brackish groundwater.  Among the six fertilisers tested as DS, (NH4)2HPO4 or DAP was observed to have the most severe scaling during the FDFO process using brackish groundwater.  Membrane autopsy using SEM, EDS and XRD indicates that magnesium phosphate, magnesium ammonium phosphate (struvite), and calcium phosphate (hydroxyapatite) were primarily responsible for membrane scaling when DAP is used as DS  Physical cleaning was observed to be effective in recovering the permeate flux of the membrane after FDFO process.  Although a rapid flux decline was observed using DAP as DS, the water flux recovered almost fully after physical cleaning.

181

Acknowledgements The authors acknowledge the financial support of the National Centre of Excellence in Desalination Australia, which is funded by the Australian Government through the National Urban Water and Desalination Plan.

references

Achilli, A., Cath, T.Y., Marchand, E.A., Childress, A.E., 2009. The forward osmosis membrane bioreactor: a low fouling alternative to MBR processes. Desalination 239 (1e3), 10e21. Antony, A., Low, J.H., Gray, S., Childress, A.E., Le-Clech, P., Leslie, G., 2011. Scale formation and control in high pressure membrane water treatment systems: a review. J. Membr. Sci. 383 (1e2), 1e16. Baker, R.W., 2010. Research needs in the membrane separation industry: looking back, looking forward. J. Membr. Sci. 362 (1e2), 134e136. Boo, C., Lee, S., Elimelech, M., Meng, Z., Hong, S., 2012. Colloidal fouling in forward osmosis: role of reverse salt diffusion. J. Membr. Sci. 390e391 (0), 277e284. Cath, T.Y., Childress, A.E., Elimelech, M., 2006. Forward osmosis: principles, applications, and recent developments: review. J. Membr. Sci. 281, 70e87. Charcosset, C., 2009. A review of membrane processes and renewable energies for desalination. Desalination 245 (1e3), 214e231. Chesters, S.P., 2009. Innovations in the inhibition and cleaning of reverse osmosis membrane scaling and fouling. Desalination 238 (1e3), 22e29. Chesters, S.P., del Vigo, F., Darton, E.G., 2007. Theoretical and Practical Experience of Calcium Phosphate Inhibition in RO Waters. IDA World Congress. Maspalomas, Gran CanariaSpain REF: IDAWC/MP07_182. Cornelissen, E.R., Harmsen, D., de Korte, K.F., Ruiken, C.J., Qin, J.J., Oo, H., Wessels, L.P., 2008. Membrane fouling and process performance of forward osmosis membranes on activated sludge. J. Membr. Sci. 319, 158e168. Elimelech, M., Phillip, W.A., 2011. The future of seawater desalination: energy, technology, and the environment. Science 333 (6043), 712e717. Geise, G.M., Lee, H.-S., Miller, D.J., Freeman, B.D., McGrath, J.E., Paul, D.R., 2010. Water purification by membranes: the role of polymer science. J. Polym. Sci. Part B: Polym. Phys. 48 (15), 1685e1718. Gilron, J., Hasson, D., 1987. Calcium sulphate fouling of reverse osmosis membranes: flux decline mechanism. Chem. Eng. Sci. 42, 2351e2360. Greenlee, L.F., Lawler, D.F., Freeman, B.D., Marrot, B., Moulin, P., 2009. Reverse osmosis desalination: water sources, technology, and today’s challenges. Water Res. 43 (9), 2317e2348. Hancock, N.T., Cath, T.Y., 2009. Solute coupled diffusion in osmotically driven membrane processes. Environ. Sci. Technol. 43 (17), 6769e6775. Harrison, M.L., Johns, M.R., White, E.T., Mehta, C.M., 2011. Growth rate kinetics for struvite crystallisation. Chem. Eng. Trans. 25 (2011), 309e314. Hasson, D., Drak, A., Semiat, R., 2001. Inception of CaSO4 scaling on RO membranes at various water recovery levels. Desalination 139, 73e81. Hatziantoniou, D., Howell, J.A., 2002. Influence of the properties and characteristics of sugar-beet pulp extract on its fouling

182

w a t e r r e s e a r c h 5 7 ( 2 0 1 4 ) 1 7 2 e1 8 2

and rejection behaviour during membrane filtration. Desalination 148 (1e3), 67e72. Holloway, R.W., Childress, A.E., Dennett, K.E., Cath, T.Y., 2007. Forward osmosis for concentration of anaerobic digester centrate. Water Res. 41, 4005e4014. Lee, S., Boo, C., Elimelech, M., Hong, S., 2010. Comparison of fouling behavior in forward osmosis (FO) and reverse osmosis (RO). J. Membr. Sci. 365 (1e2), 34e39. Lee, S., Elimelech, M., 2006. Relating organic fouling of reverse osmosis membranes to intermolecular adhesion forces. Environ. Sci. Technol. 40 (3), 980e987. Li, J., Hallbauer, D.K., Sanderson, R.D., 2003. Direct monitoring of membrane fouling and cleaning during ultrafiltration using a non-invasive ultrasonic technique. J. Membr. Sci. 215 (1e2), 33e52. Li, Q., Elimelech, M., 2004. Organic fouling and chemical cleaning of nanofiltration membranes: measurements and mechanisms. Environ. Sci. Technol. 38 (17), 4683e4693. Liu, Y., Mi, B., 2012. Combined fouling of forward osmosis membranes: synergistic foulant interaction and direct observation of fouling layer formation. J. Membr. Sci. 407e408 (0), 136e144. McCutcheon, J.R., McGinnis, R.L., Elimelech, M., 2005. A novel ammoniaecarbon dioxide forward (direct) osmosis desalination process. Desalination 174, 1e11. McGinnis, R.L., Elimelech, M., 2007. Energy requirements of ammonia-carbon dioxide forward osmosis desalination. Desalination 207 (1e3), 370e382. Mi, B., Elimelech, M., 2010a. Gypsum scaling and cleaning in forward osmosis: measurements and mechanisms. Environ. Sci. Technol. 44 (6), 2022e2028. Mi, B., Elimelech, M., 2010b. Organic fouling of forward osmosis membranes: fouling reversibility and cleaning without chemical reagents. J. Membr. Sci. 348 (1e2), 337e345. Phillip, W.A., Yong, J.S., Elimelech, M., 2010. Reverse draw solute permeation in forward osmosis: modeling and experiments. Environ. Sci. Technol. 44 (13), 5170e5176. Phuntsho, S., 2012. A Novel Fertiliser Drawn Forward Osmosis Desalination for Fertigation. PhD Thesis. University of Technology Sydney. PhD. Phuntsho, S., Hong, S., Elimelech, M., Shon, H.K., 2013a. Forward osmosis desalination of brackish groundwater: meeting water quality requirements for fertigation by integrating nanofiltration. J. Membr. Sci. 436 (0), 1e15. Phuntsho, S., Listowski, A., Shon, H.K., Le-Clech, P., Vigneswaran, S., 2011a. Membrane autopsy of a 10 year old hollow fibre membrane from Sydney Olympic Park water reclamation plant. Desalination 271 (1e3), 241e247. Phuntsho, S., Mamisahebi, S., Majeed, T., Lotfi, F., Kim, J.E., Shon, H.K., 2013b. Assessing the major factors affecting the

performances of forward osmosis and its implications on the desalination process. Chem. Eng. J. 231, 484e496. Phuntsho, S., Shon, H.K., Hong, S.K., Lee, S.Y., Vigneswaran, S., 2011b. A novel low energy fertilizer driven forward osmosis desalination for direct fertigation: evaluating the performance of fertilizer draw solutions. J. Membr. Sci. 375, 172e181. Phuntsho, S., Shon, H.K., Majeed, T., El Salibya, I., Vigneswarana, S., Kandasamy, J., Hong, S., Leeb, S., 2012a. Blended fertilisers as draw solutions for fertiliser drawn forward osmosis desalination. Environ. Sci. Technol. 46 (8), 4567e4575. Phuntsho, S., Shon, H.K., Vigneswaran, S., Kandasamy, J., Hong, S.K., Lee, S.Y., 2012b. Influence of temperature and temperature difference in the performance of forward osmosis desalination process. J. Membr. Sci. 415e416, 734e744. Rootare, H.M., Deitz, V.R., Carpenter, F.G., 1962. Solubility product phenomena in hydroxyapatite-water systems. J. Colloid Sci. 17 (3), 179e206. She, Q., Jin, X., Li, Q., Tang, C.Y., 2012. Relating reverse and forward solute diffusion to membrane fouling in osmotically driven membrane processes. Water Res. 46 (7), 2478e2486. Stratful, I., Scrimshaw, M.D., Lester, J.N., 2001. Conditions influencing the precipitation of magnesium ammonium phosphate. Water Res. 35 (17), 4191e4199. Tang, C.Y., Chong, T.H., Fane, A.G., 2011. Colloidal interactions and fouling of NF and RO membranes: a review. Adv. Colloid Interface Sci. 164 (1e2), 126e143. Tang, C.Y., She, Q., Lay, W.C.L., Wang, R., Fane, A.G., 2010. Coupled effects of internal concentration polarization and fouling on flux behavior of forward osmosis membranes during humic acid filtration. J. Membr. Sci. 354 (1e2), 123e133. Williams, S., 1999. Struvite precipitation in the sludge stream at slough wastewater treatment plant and opportunities for phosphorus recovery. Environ. Technol. 20 (7), 743e747. Zach-Maor, A., Semiat, R., Rahardianto, A., Cohen, Y., Wilson, S., Gray, S.R., 2008. Diagnostic analysis of RO desalting treated wastewater. Desalination 230 (1e3), 239e247. Zhang, J., Loong, W.L.C., Chou, S., Tang, C.Y., Wang, R., Fane, A.G., 2012. Membrane biofouling and scaling in forward osmosis membrane bioreactor. J. Membr. Sci. 403e404, 8e14. Zhao, S., Zou, L., 2011. Effects of working temperature on separation performance, membrane scaling and cleaning in forward osmosis desalination. Desalination 278 (1e3), 157e164. Zhao, S., Zou, L., Tang, C.Y., Mulcahy, D., 2012. Recent developments in forward osmosis: opportunities and challenges. J. Membr. Sci. 396 (0), 1e21.