ww.sciencedirect.com

wat e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 0 6 5e2 0 7 4

Available online at w

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

Composite hollow fiber nanofiltration membranes forrecovery of glyphosate from saline wastewater

Jianfeng Song a, Xue-Mei Li a,*, Albeto Figoli b, Hua Huang c, Cheng Pan c, Tao He a,*,Biao Jiang a

a Laboratory for Membrane Materials and Separation Technology, Shanghai Advanced Research Institute, Chinese Academy of Sciences,

Shanghai 201210, Chinab Institute on Membrane Technology, ITM-CNR, c/o University of Calabria, Via P. Bucci, Cubo 17/C, 87030 Rende (CS), ItalycNanjing University of Technology, Nanjing 210009, Jiangsu, China

a r t i c l e i n f o

Article history:

Received 17 June 2012

Received in revised form

6 November 2012

Accepted 20 January 2013

Available online 30 January 2013

Keywords:

Nanofiltration

Glyphosate

Hollow fiber membranes

Composite membranes

Wastewater

* Corresponding authors. Tel.: þ86 21 203251E-mail address: het@sari.ac.cn (X.-M. Li).

0043-1354/$ e see front matter ª 2013 Elsevhttp://dx.doi.org/10.1016/j.watres.2013.01.032

a b s t r a c t

A high performance versatile composite hollow fiber nanofiltration (NF) membrane is

reported for the separation of glyphosate from saline waste streams. Preparation of SPEEK

based on an amorphous poly (ether ether ketone, PEEK) was investigated. The membrane

was prepared by coating sulfonated polyether ether ketone (SPEEK) onto a polyethersulfone

(PES) ultrafiltration (UF) hollow fiber membrane. The composite membrane was charac-

terized by water permeability, scanning electron microscopy, and rejection toward sodium

sulfate (Na2SO4), sodium chloride (NaCl), and calcium chloride (CaCl2). About 90% rejection

toward sulfate anions and only 10% rejection for calcium cations were obtained. A water

permeability around 10e13 LMHBar and 90% rejection for polyethylene glycol (PEG) with

a molecular weight of 4000e6000 Da were observed. In the separation of glyphosate from

saline wastewater, the membrane rejected less than 20% of NaCl and higher than 90% of

glyphosate at an operating pressure of 5 bars and pH ¼ 11.0. An economic analysis indi-

cated that the cost for recovery of glyphosate was comparably low to the value gained by

an increase in the productivity. The results may lead to a new promising low energy so-

lution for the environmental problem faced by the herbicide industry.

ª 2013 Elsevier Ltd. All rights reserved.

1. Introduction glyphosate for drinking water (GB5749-2006 China) is 0.7 mg/L

Glyphosate (N-(phosphonomethyl) glycine) is a broad-

spectrum systemic herbicide for killing weeds, especially

annual broadleaf weeds and grasses. Glyphosate is harmful to

humans, plants and animals in terms of enzyme activity

dysfunction (Freuze et al., 2007), cytotoxicity and DNA dam-

ages for human cells (Gasnier et al., 2009). Thus, discharge of

glyphosate into the environment is prohibited. The limit for

62; fax: þ86 21 20325034.

ier Ltd. All rights reserved

(CNSC, 2006). In the production process, to obtain 1 ton of

glyphosate, 5e6 tons of wastewater is produced, containing

1e2 wt. % glyphosate sodium salt, 12e16 wt. % NaCl, 1 wt. %

Na2HPO3 and a small amount of triethylamine (Xie et al., 2010).

A highly effective and energy-efficient method for the recov-

ery of glyphosate from thewaste streams is critical to improve

the glyphosate production efficiency and to reduce its envi-

ronment impact.

.

wat e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 0 6 5e2 0 7 42066

Nanofiltration is a membrane separation process based on

both sieving effect and Donnan exclusion effect (Bowen and

Mohammad, 1998a,b). Sieving effect is based on size exclu-

sion while Donnan effect is basically a result of charge

repulsion. The particular characteristics of both glyphosate

and arsenic are that they are dissociable at various pH con-

ditions and may exist as multivalent anions. These charac-

teristics make nanofiltration process a suitable technology for

separation and removal of ions from the contaminated

streams.

Xie et al. (2010) reported the use of a commercial spiral-

wound nanofiltration membrane for glyphosate wastewater

treatment. A flux of 7.32 L/m2 h and a removal rate of 95.5%

were obtained at 20 bars. However, the use of a narrow spacer

and membrane fouling make the spiral wound membrane

module an unlikely economic configuration. Spiral-wound

membranes have narrow channels and the backwashing is

difficult. Consequently, the membrane fouling is problematic.

Therefore, a pre-filtration is essential. The clarification of the

feed streams using various technologies adds up extra costs to

the treatment process, resulting in a low economical com-

petence. Hollow fiber membranes are well-defined in feed

channels and are back-washable (Frank et al., 2001), thus may

be more promising than spiral-wound membranes. We have

reported the preparation of composite hollow fiber nano-

filtration membrane by coating an ultrafiltration support

membrane with sulfonated polyether ether ketone (SPEEK)

(He et al., 2008). The membranes showed rejection toward

negatively charged dyes including sunset yellow and procion

blue. The principle of separation was based on the Donnan-

Steric effects (Bowen and Mohammad, 1998a). However, the

glyphosate rejection properties of this membrane were not

known. Besides, the SPEEK coating layer was made by sulfo-

nation of a semi-crystalline PEEK. The sulfonation was rather

inhomogeneous (Bishop et al., 1985), which may cause in-

homogeneity in the final SPEEK and thus in the coating layer.

Improving the homogeneity of sulfonation would improve the

performance of the final composite nanofiltration mem-

branes. Moreover, the support hollow fiber membranes were

tailor-made, instead of a commercial product, making the

scale up of this process unrealistic.

Hereby, we explore systematically the possibility of pre-

paring a composite hollow fiber nanofiltration membrane

using a SPEEK from amorphous PEEK and a commercial hol-

low fiber support membrane for the separation of glyphosate

fromwaste streams. The sulfonation degree of the amorphous

PEEK was controlled by variation of the sulfonation time. The

performance of the membranes was optimized and charac-

terized by water permeability, salt rejection, field emission

scanning electron microscopy (FESEM) and molecular weight

cutoff. In the NF treatment of glyphosate wastewater streams,

operation conditions were investigated including trans-

membrane pressure, feed concentration, pH, and ionic

strength with respect to glyphosate rejection. The long time

membrane performance was tested in order to evaluate the

performance stability of the membrane. Furthermore, an

economic analysis of the recovery of glyphosate using NF

membrane is given, which may assist in evaluating the

application of such a membrane process for herbicide

industry.

2. Materials and methods

2.1. Membranes and materials

Commercial PES hollow fibers UF membrane was provided by

Nanjing Altrateck Co. (MWCO 70,000 Da, inner/outer

diameter ¼ 0.8/1.3 mm). Amorphous polyether ether ketone

(PEEK), VESTAKEEP 4000P, was kindly provided by DEGUSSA

(Germany). Semi-crystalline PEEK (Victrex 450PF) was pur-

chased from ICI. Glyphosate (�95% purity, MW: 196.6 Da) was

provided by Nantong Jiangshan Agrochemical & Chemicals

Co. Analytical grade Na2SO4, NaCl, CaCl2, BaCl2, MgCl2, NaOH,

sulfuric acid (H2SO4), iodine (I2), methanol and other reagents

were supplied by Sinopharm Chemical Reagent Co. Ltd and

used as received. Polyethylene glycol (PEG) of different mo-

lecular weight was obtained from Alfa Aesar. Deionized water

was used in solution preparation.

2.2. Sulfonation of PEEK

To a three necked 1 L round-bottom flask equipped with

amechanical stirrer was added 500mL of concentrated H2SO4.

PEEK powder (50 g) was then introduced while stirring. The

mixture was stirred at 25 �C for 24e96 h, which yielded a light

yellow to dark brown solution. The solution was suspended

drop-wise into ice cold deionized water (5 L). Then the gran-

ular SPEEK precipitates were washed with deionized water

repeatedly to remove residue sulfuric acid and filtered. The

SPEEK polymer was then dried in the air for 48 h and com-

pletely dried up in vacuum oven for at least 7 days at 30 �C.Sulfonated PEEK was denoted as SPEEK-XX, where the XX

represents the sulfonation time, h. The degree of sulfonation

(DS) was determined by the acid-base titration (Li et al., 2003).

2.3. Preparation of the composite nanofiltrationmembrane

SPEEK-48 was dissolved inmethanol at a certain concentration

at ambient temperature. The solutionwas filteredwith a 40 mm

stainless steel filter, and then coated onto the inner surface of

commercial hollow fiber PES UF membranes. The coating so-

lution was raised from the bottom to the top of the fiber and

allowed to contact the membrane inner surface for 3 s. The

solution was then drained from the bottom of the fiber by

gravity. After introducing the coating layer, the bore side was

dried under a nitrogen stream and subsequently dried at 65 �Cfor a certain time or specified otherwise. The composite

membraneswerestoredatambient temperaturebefore testing.

2.4. Molecular Weight Cut-Off (MWCO)

Six composite hollow fiber membranes with effective lengths

of 16 cm were potted into a nylon tube with epoxy resin. The

MWCOof the compositemembranewasdetermined by a cross

flow nanofiltration setup as shown in Fig. 1. Feed solution was

circulated and partially fed to the inlet of the test module.

Pressure and flow rate through the lumen of the hollow fiber

nanofiltration membranes were controlled manually. The

pressure drop between the upstream and downstream of the

Fig. 1 e Nanofiltration cross-flow filtration setup. 1. feed

tank; 2. magnetic stirrer; 3. filter; 4. control valve; 5. gear

pump; 6. pressure gauge; 7. hollow fiber membrane

module; 8. permeate collector; 9. rota flowmeter.

wat e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 0 6 5e2 0 7 4 2067

module was controlled within 0.5e1.0 bar. The solution tem-

perature was thermostatically maintained at 20 � 2 �C.PEG of different molecular weight was used for testing the

molecular weight cutoff. The concentration was fixed at 0.5 wt

% and the operation trans-membrane pressurewas 3 bars. The

concentrations of PEG in the feedandpermeateweremeasured

as follows (Liu et al., 2004): certain amount of PEG standard

solution was added 1mL iodine standard solution (0.05mol/L),

followed by addition of 1 mL BaCl2 standard solution (5 wt %).

After mixing thoroughly, the total volume was fixed with

deionizedwater to25mL.After 5min, theabsorbanceat 661nm

wasmeasuredbyUVeVis spectrophotometer (UV-2802,UNICO

(Shanghai) Instruments Co., Ltd.). The concentration of PEG

was determined according to a standard calibration curve. The

MWCOvaluewasdefinedas thePEGmolecularweight atwhich

the membrane shows a rejection of 90%.

2.5. Rejection test

Repeated NF testes were carried out at 3e8 bars and the

average value within 10% deviation was reported. For the salt

solution, the feed concentration was 1000 ppm if not other-

wise stated. The glyphosate feed concentration was 500 mg/L,

pH was 11.0 for the NF process, unless stated otherwise. The

pH values of the solutions were measured using a pH meter

(Sartorius Scientific Instruments Co., Beijing).

The permeation flux J (L/m2 h), determined by collecting

the water permeation for a certain period of time was calcu-

lated as follows:

J ¼ VA� t

(1)

Where V, A, t represent the total volume (L) of permeation, the

membrane area (m2) and the operation time (h), respectively.

The observed rejection, Robs, was defined by the following

equation:

Robs ¼ 1� Cp

Cf(2)

where Cf and Cp are the bulk feed and permeate concentra-

tions, respectively.

Glyphosate concentration was measured according to the

Chinese Standard Regulation (GB12686-2004) by UVeVis spec-

trophotometry. Briefly, glyphosate was treated with sodium

nitrite under acidic conditions to generate a nitrosoglyphosate

derivative which has a UV absorption peak at 242 nm. A stan-

dard curvewas calibratedbeforehand. For amixture of sodium

chloride and glyphosate, the sodium concentration was

measured using an Inductively Coupled Plasma Atomic Emis-

sion Spectrometer (ICP-AES, Optima 2000DV).

2.6. Scanning electron microscopy (FESEM)

A Hitachi TM1000 Scanning Electron Microscopy (SEM) and

a Hitachi S-4800 Field Emission Scanning Electron Microscopy

(FESEM) were used for imaging the morphology of the mem-

branes. Samples were prepared by cryogenic breaking with

liquid nitrogen, allowed to dry under vacuum at 30 �C over-

night and coated with a thin gold layer.

3. Results and discussion

3.1. Membrane preparation and characterization

3.1.1. SPEEK coating polymerSulfonated poly (ether ether ketone) (SPEEK)was prepared and

characterized following the same procedure as reported in

literature (Bishop et al., 1985). An amorphous PEEK was used

instead of the semi-crystalline ones. It was found that amor-

phous PEEK dissolved faster in sulfuric acid than the crystal-

line PEEK. Moreover, solubility tests indicated that at the same

sulfonation time, semicrystalline SPEEK was less soluble than

the amorphous SPEEK (He, 2001). It requires only 48 h of reac-

tion time to obtain a methanol soluble SPEEK for amorphous

PEEK insteadof 120h for the semicrystallinePEEK (Table1). The

relatively faster sulfonationmaybepartially due to thequicker

dissolution of an amorphous PEEK polymer in sulfuric acid.

The solubility of amorphous SPEEK was determined in

order to select an appropriate solvent for coating. The test

results (Table 1) indicates that dimethyl formamide (DMF),

dimethylacetamide (DMAc) and methanol are good solvents

for SPEEK with a sulfonation degree (SD) greater than 0.69.

Because PESmaterial is soluble in polar solvent like N-methyl-

2-pyrrolidone (NMP), DMAc and DMF, methanol was used as

the solvent for preparing SPEEK coating solution. In order to

prevent aggressive swelling of coating in nanofiltration,

a SPEEK with sulfonation degree of 0.69 was selected.

3.1.2. Characteristics of the support membraneA commercial PES hollow fiber ultrafiltration membrane was

used as the support. The membrane has an inner/outer

diameter of 0.8/1.3 mm, respectively, typical for commercial

UF membranes with a water permeability of 600 � 50 L/

(m2 h bar) and amolecular weight cutoff (MCWO) of 70,000 Da.

SEM images of the cross-section and the inner-skin and mid-

structure of the hollow fiber membrane are shown in Fig. 2.

Very open outer surface, sponger-like interior and rather thin

skin layer are observed. The PES hollow fiber membranes

showed an elongation at break of 96% with a force of 3.70 N at

breakage, which was rather high values even comparable to

Table 1 e Solubility of SPEEK prepared at different sulfonation time.

Polymers S-24 S-48 S-72 S-48-Victrexa S-120 Victrex

Sulfonation time(hr.) 24 48 72 48 120

IEC(meq g-1) 1.6 1.9 2.1 1.2 2.3

Sulfonation degree, SD (%) 0.47 0.69 0.75 e e

NMP þ þ e þ e

DMF þ e e NA NA

DMAc þ e e þ e

Methanol þ e e þ e

Ethanol o þ e O þþH2O o þ þþ O þ

O: no change; þ: swollen; þþ: highly swollen; e: dissolved.

a Results adopted from reference (He, 2001).

wat e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 0 6 5e2 0 7 42068

an elastic PVDF membrane (Sukitpaneenit and Chung, 2009),

given that PES is a glassy polymer. This comparison indicates

that the PES membrane is rather robust, which may be toler-

ant to mechanical scrubbing during backwashing and clean-

ing processes.

3.1.3. Effect of the coating solution concentrationSolutions with different SPEEK concentration were employed

as coating for composite hollow fiber nanofiltration mem-

brane. Their effects on the flux and salt rejection of the NF

membranes are shown in Fig. 3. With the increase of the

SPEEK concentration from 1.0 wt. % to 3.0 wt. %, the Na2SO4

rejection rate increased slightly from 92.8% to 96.8%. The flux

remained the same about 43.8 L/m2 h for 1.0 wt% and 1.5 wt%,

but dropped significantly at 2.0% and further down to 9.6 L/

m2 h at 3%. The high SO42� rejection is ascribed to electrostatic

repulsion because of the presence of the sulfonic groups in

SPEEK (Ismail and Lau, 2009). The slight increase in the

rejection with increase of the coating concentration is

Fig. 2 e SEM photos of polyethersulfone ultrafiltration hollow fi

morphology with very open interior structure and a rather thin

ascribed to increased coating layer thickness that leads to

higher mass transfer resistance and consequently the lower

permeation flux (Kim et al., 2000). For optimized permeability

and acceptable rejection, the SPEEK concentrationwas fixed at

1.5 wt. % for further investigation.

3.1.4. MWCO of the composite hollow fiber nanofiltrationmembranePEG molecules are often used to determine the molecular

weight cutoff (MWCO) of UF membranes due to their wide

range availability in molecular weight. Five PEGs with mo-

lecular weight in the range of 400 Dae8000 Da were tested and

the retention rates were shown in Fig. 4. No retention was

observed for PEG 400 and 600. For PEG 1000, the rejection rate

was about 60%. From Mw > 2000 Da on, the rejection rate

slowly increased and reached nearly 100% at Mw ¼ 8000 Da.

The definition of the MWCO is the Mw of PEG at which a 90%

rejection is obtained. Thus, the MWCO of the nanofiltration

membrane was estimated to be between 4000 and 6000 Da.

ber membrane. The membrane shows a sponge-like

skin layer.

0.5 1.0 1.5 2.0 2.5 3.0 3.50

20

40

60

80

100

FluxRejection

SPEEK Conc. (wt.%)

Flu

x (L

/m

2 h)

0

20

40

60

80

100

Re

je

ctio

n (%

)

Fig. 3 e Effect of SPEEK concentration on nanofiltration

membranes performance. Na2SO4 concentration [ 1000

mg/L, trans-membrane pressure [ 3 bars.

wat e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 0 6 5e2 0 7 4 2069

The pore radius of the SPEEK composite membrane was

then estimated based on Eq. (3) as follows

log10rs ¼ �1:3363þ 0:395log10Mw (3)

Where rs, represents the pore radius and is in nm, Mw. corre-

sponds to the MWCO of the membrane, and is in g/mol.

This equation was adopted from the research work for the

characterization of the nanofiltration membrane pore size for

neutral organic chemicals by Bowen et al. (Bowen and

Mohammad, 1998a). Using the MWCO, the molecular radius

of the solute, rs, was calculated by using the StokeseEinstein

equation. By fitting the experimental data between the mo-

lecular weight and the Stokes radius, an empirical relation-

ship is obtained as Eq. (3). Assuming the Stokes radius of the

solute to be very close to the pore radius, the SPEEK coated

nanofiltration membrane has a pore radius in the range of

1.22e1.43 nm, significantly smaller than the pore size of the

support membrane.

3.1.5. Morphology of the composite membraneField emission SEM images of the composite membranes are

shown in Fig. 5. The inner surface of the PES support ultrafil-

trationmembranes shows crevice-like pores (Fig. 5A), but that

of composite membrane appears quite smooth (Fig. 5B).

Compared to the support membrane (Fig. 5C), a dense layer is

0 2000 4000 6000 80000

25

50

75

100

Rejectio

n (%

)

Molecular Weight PEG

Fig. 4 e Rejection rate of SPEEK coated composite

membrane for PEG with molecular weight ranging from

400 to 8000 Da at a trans-membrane pressure of 3 bars.

clearly observed from the cross section view (Fig. 5D) in the

composite with a thickness of approximately 0.4 mm. It is

likely that there exists penetration of SPEEK polymer into the

support membrane pore as seen from Fig. 5D. This pore pen-

etration makes the quantification of the dense skin layer

based on the SEM photos difficult.

3.1.6. Salt rejectionTheNFmembraneswere tested for their salt rejections toward

Na2SO4, NaCl, and CaCl2 (Table 2). The salt rejection rates fol-

low an order of RNa2SO4 > RNaCl > RCaCl2 . The rejection of NF

process is based on sieving effect andDonnan exclusion effect.

Considering that thehydraulic radii (rh) of SO42� (0.300nm), Cl�

(0.178 nm) and Ca2þ (0.253 nm) aremuch smaller than the pore

radius (1.22e1.43 nm) of the composite nanofiltration mem-

branes as estimated by MWCO, sieving effect may play less

significant role in the salt rejection. On the other hand, the

SPEEK nanofiltration membranes are negatively charged. Ac-

cording to Donnan exclusive effect, it should have a high

rejection toward bivalent anions and a low rejection toward

multiple valent cations (Petersen, 1993), and thus the highest

rejection was found for Na2SO4, and the lowest for CaCl2. For

example, Ca2þ is bivalent, which is attracted by themembrane

more strongly than monovalent Naþ. Because the Ca2þ ions

diffuse through the membrane, Cl� anions will follow due to

electro-neutral requirements. Thus, the rejection of the cal-

cium chloride was lower than that of NaCl.

3.2. Separation properties of glyphosate

Theconditions fornanofiltrationof glyphosatewereoptimized

based on glyphosate solutions prepared in the laboratory. The

NF stability performance was carried out using production

glyphosate wastewater streams as feeds (after dilution).

3.2.1. Effect of the trans-membrane pressureNanofiltration is a pressure drivenmembrane process and the

trans-membrane pressure (TMP) plays an important role. To

investigate the influence of the TMP on permeation and

rejection, a solution with glyphosate concentration of 500mg/

L was used as the feed at pH 11.0.

Fig.6 shows the permeation flux and the rejection toward

glyphosate of the composite membrane at TMP in the range of

3e8 bars. With the increase of the TMP, the permeation flux

increased linearly and the retention remainednearly constant.

At a high TMP, the transport of water molecules accelerates

(Al-Zoubi and Omar 2009), but the TMP did not accelerate the

diffusion of solute molecules (Hilal et al., 2005), thus the per-

meation concentration decreased, resulting in a higher

glyphosate retention.At 8bars, the retentionof glyphosatewas

about 100% and the permeation flux was 97.5 L/m2 h.

3.2.2. Effect of the feed concentration and pHTo examine the effect of the feed concentration on the

membrane flux and retention, glyphosate solutions with

concentrations varied from 100 to 1000 mg/L at pH 11.0 (TMP

3 bars) were tested. With the increase of feed concentration,

glyphosate retention declined from 100% to 92.3% and the

permeation flux dropped from 41.1 to 36.5 L/m2 h as shown in

Fig. 7a. The flux decrease is expected due to increased osmotic

Fig. 5 e SEM photos of the inner surfaces and inner skins of the PES ultrafiltration membrane and SPEEK coated

nanofiltration membranes. A: Inner surface of PES membrane; B: Inner surface of nanofiltration membrane; C: inner skin of

the PES membrane; D: inner skin of the nanofiltration membrane. The arrow indicates the coating layer.

0 2 4 6 8 100

30

60

90

Flu

x (L

/m

2

h)

TMP (bar)

98

99

100

n (%

)

wat e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 0 6 5e2 0 7 42070

pressure with increased solute concentration and thus lower

effective driving pressure. The decrease in the rejection agrees

to Donnan equilibrium theory in that: as the feed concentra-

tion increases, the increased positive charges dampen out the

electrostatic repulsion between the membrane and glypho-

sate, causing more glyphosate transfer across the membrane,

thus lower rejection rates. It appeared that the permeation

fluxwas nearly independent of the feed pH, indicating that the

membranes are rather stable toward acid and alkali environ-

ment (Richards et al., 2010).

To examine the effects of the feed pH on the membrane

flux and retention, 500 mg/L glyphosate solutions with pH

varied from 3.0 to 11.0 were tested. It was observed that the

retention rate changed significantly against the pH of the feed

solution (Fig. 7b). At pH ¼ 3.0, the nanofiltration membrane

showed a glyphosate retention rate of 32%; at pH ¼ 4.5, the

retention rate increased to 48%; when the pH was 6.0, the

retention rate increased significantly to 92%. At pH ¼ 11,

a nearly 100% rejection was observed.

The pH dependence of the rejection rate may be under-

stood by the dissociation status of glyphosate at different pH

Table 2 e Rejection properties of the compositemembrane against different saltsa.

Salt types Rejection (%) Permeation flux (L/m2 h)

Na2SO4 96.2 66.9

NaCl 42.8 68.9

CaCl2 34.0 69.1

a Salt concentration ¼1000 mg/L, trans-membrane pressure: 5 bar.

3 4 5 6 7 894

95

96

97

Rejectio

TMP (bar)

Fig. 6 e Influence of trans-membrane pressure on the flux

and glyphosate rejection. Glyphosate concentration [

500mg/L, pH[ 11.0.

0 200 400 600 800 100020

25

30

35

40

45

50

FluxRejection

Glyphosate Conc. (mg/L)

Flu

x (L

/m

2h

)

0

20

40

60

80

100

Re

je

ctio

n (%

)

3.0 4.5 6.0 7.5 9.0 10.5 12.00

20

40

60

80

100

pH

Re

je

ctio

n (%

)

Fig. 7 e Effect of feed concentration (a) and pH (b,

concentration [ 500 mg/L) on glyphosate rejection. Trans-

membrane pressure [ 3 bars, pH [ 11.0.

100

wat e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 0 6 5e2 0 7 4 2071

conditions. Glyphosate has carboxyl, amino and phosphoric

acid groups, and three dissociation states, as shown in

Scheme 1. At pH¼ 2.6e5.9, the glyphosate is presentmainly as

a mono-valent anion. At pH ¼ 5.9e10.9, it dissociates further

into a divalent anion; and at pH ¼ 10.9 and above, the glyph-

osate completely dissociates into a trivalent anion (Spethf,

1993). Although the molecular structure of glyphosate is

larger than Cl� and SO42�, the rejection at pH 3.0was only 32%.

This indicates that at mono-valent state, the glyphosate may

permeate through the membrane pore in a way similar to Cl�.At a pH > 6.0, the glyphosate is mainly present as a bi-valent

anion; the rejection was above 90%, which is close to the bi-

valent anion SO42�. At pH ¼ 11.0, glyphosate is present

mainly as a tri-valent anion. High repulsion leads to nearly

100% rejection. Based on above results, it can be seen that the

rejection of the nanofiltrationmembrane toward glyphosate is

mostly controlled by Donnan effect.

Scheme 1 e Dissociation of glyphosate at different pH.

3.2.3. Effect of the ionic strengthWater streams from the herbicide plant are complicated

mixtures containing various chemicals, among which NaCl is

highly concentrated. The existence of NaCl may influence the

separation characteristic of the membrane for glyphosate. In

order to investigate the effect of NaCl concentration, the

glyphosate concentration was fixed at 500 mg/L, and the NaCl

concentrationwas varied from100 to 5000mg/L at pH¼ 11.0 at

a TMP of 5 bars.

As seen in Fig. 8, a 100% glyphosate retention was observed

without NaCl. At NaCl concentration of 500mg/L, the rejection

value decreased slightly by 3.6%. With further increase in the

NaCl concentration, the rejection value declined slowly and

reached 90% at NaCl concentration of 5000 mg/L. As men-

tioned above, the retention of glyphosate is based on Donnan

effect.When the concentration of NaCl increases, the negative

charges of at the membrane surface are partially dampened

out and consequently reduced rejection rates are observed

(Ballet et al., 2007). Nevertheless, at 5000 mg/L of NaCl in the

feed, a 90% glyphosate retention was still achieved indicating

that nanofiltration membranes were capable of recovering

glyphosate from saline environment. In the following part,

a wastewater from glyphosate production facility was used as

feed for the NF process in order to evaluate the capability of

the NF membranes for glyphosate recovery from the produc-

tion wastewater.

3.2.4. Glyphosate recovery from production wastewaterThe glyphosate production wastewater contains a high con-

centration of NaCl in the range between 120 and 160 g/L and

a COD of 45e65 g/L. The other components are glyphosate and

Na2HPO3 with a concentration of 12e15 and 19e21 g/L,

respectively. The pH of the solution is in the range of 11e13.

Without dilution, the osmotic pressure of such a saline

water is much higher than that of the seawater. The operating

pressurewill be very high. Another critical issue is that at very

high salt concentration, the rejection rate of the glyphosate

would be low. Our solution to this problem is diafiltration

(Bowen and Mohammad, 1998a). Diafiltration is a unit oper-

ation that incorporates ultrafiltration/nanofiltration mem-

branes to obtain a complete separation of two solutes of

different molecular weight/properties. In diafiltration, the

feed is streamed continuously along amembrane unit and the

volume in the feed remains constant by adding water at a rate

0 1500 3000 4500 600080

84

88

92

96

Re

je

ctio

n (%

)

Conc. (mg/L)

Fig. 8 e Effect of NaCl concentration on glyphosate

rejection. Glyphosate concentration [ 500 mg/L, trans-

membrane pressure [ 5 bars, pH [ 11.0.

wat e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 0 6 5e2 0 7 42072

equal to the permeation rate. By diluting the feed solution,

both osmotic pressure and rejection issues may be solved to

a large extent and the membrane fouling will be decreased

(Ballet et al., 2007). Moreover, adoption of diafiltration may

achieve complete separation of glyphosate and NaCl.

Therefore, as a proof of concept, the diluted (10 times)

mother liquor was used as the feed to investigate the perfor-

mance of the membranes in separating salt from glyphosate.

The performance of the composite membrane was tested

continuously in a cross flow configuration for 8 h Fig. 9 shows

the rejection rates for glyphosate andNaCl in the test period. It

appeared that during thefirst 0.5 h, the rejections to glyphosate

andNaClwere91%and13%respectivelywithapermeationflux

of 31 L/m2 h at an operating pressure of 5 bars. As the operation

time elongated, the flux and the retention of glyphosate and

NaCl declined gradually and at the end of the test, the rejection

rates to glyphosate and NaCl reached 84% and 10%, respec-

tively. The significantly low flux was most probably due to the

instantaneous concentration polarization and progress in

membrane fouling. The decrease in rejection toward glypho-

sate and NaCl might also be ascribed to the concentration po-

larization in the NF processes as a result of water passage

through themembrane. It is expected that there is a local build-

upof theglyphosate andNaCl at andclose theactive separation

layer. Due to the localized higher concentration, more glypho-

sate and NaCl would transfer across the membrane leading to

decreased glyphosate and NaCl rejection.

Nevertheless, the long time test results showed the

chemicals within the waste streams did not attack or degrade

the SPEEK layer. This result gives a solid support to feasibility

of application of the NF membrane in pilot-scale. Larger

membrane modules will be developed and tested in an

extended time range and the possible cleaning and parameter

optimization will be investigated in future work.

3.3. Economic analysis

Based on the preliminary performance, cost estimation for the

recovery of glyphosate from the wastewater streams is pur-

sued. The assumptions and key cost parameters for such

a treatment system are listed in Table 3. The calculation is

based on the assumption of a glyphosatemanufacturing plant

with annual production of 100k tons. The membrane

0.0 1.5 3.0 4.5 6.0 7.5 9.00

20

40

60

80

100

Rejectio

n (%

)

Time (h)

Glyphosate RejectionNaCl Rejection

Fig. 9 e Effect of time on glyphosate and NaCl rejection.

Diluted glyphosate liquor, trans-membrane

pressure [ 5 bar.

permeability was assumed to be 10 L/(m2 h bar) and the

operating pressure was 5 bars. Based on an 8 inch membrane

hollow fiber membrane module, the number of membrane

modules is estimated to be 800 pieces (assume a 50% recovery

rate in comparison to 30e35% for seawater desalination). The

cost for the membranemodule was set to be 50 $/m2, which is

quite an optimistic cost, assuming that the SPEEK nano-

filtrationmembrane is in a large scale production. By adopting

the rule-of-the-thumb in reverse osmosis process, the life

time of the membrane system is set to be 30 years, and the

cost for a membrane system is 3 times as the cost of the

membranes (Baker and Lokhandwala, 2008). Therefore, by

estimating the cost for membrane, the cost for the nano-

filtration system is obtained as 4.2 million US dollars.

To justify the assumptions, the estimated cost in the power

consumption, labor and membrane replacement is compared

with that in the seawater reverse osmosis desalination

(SWRO) plant (Atikol andAybar, 2005). For a SWRO, the cost for

power was reported to be 0.04 $/m3, which is very close to the

estimated cost for electricity of 0.035 $/m3. This indicates that

the assumption of an energy efficiency of 0.8 and estimation

of the energy cost is in a reasonable range. The cost for pre-

treatment and the maintenance is adopted from the SWRO

plant. The replacement for SPEEK membrane is estimated to

be 0.093 $/m3, which is significantly higher than the SWRO, in

which the membrane replacement cost is about 0.018e0.05 $/

m3. But this cost is compensated by themanpower, where the

cost for nanofiltration system is lower than SWRO. The cost

for the SPEEKmembranemay decrease if themarket becomes

large enough to cover the investment in the manufacturing of

such membranes. Thus, the cost for membrane may be lower

in the future. Less labor is assumed since the nanofiltration is

operated at a much lower pressure than RO and the hollow

fiber membrane may perform rather stably, assuming that

a good pre-treatment as SWRO is provided. The final gross

estimation for the cost of the nanofiltration process in treating

1 cubic wastewater was 0.224 $/m3, as listed in Table 3.

Within the nanofiltration membrane system, the mem-

brane cost takes nearly 42% of the total. Therefore, a further

research in decrease the cost for membrane is of utmost

importance for a large scale application of such membranes

for herbicide industry. It is expected that a 20e30% drop in the

membrane cost is realistic if the market is big enough. It

should be realized that the cost for interest rate, cost for land

and other peripheral facilities was not taken into calculation,

which may increase the cost slightly.

Currently, the market price of glyphosate is around

21,000e29,000 RMB/ton (Glyphosate Price, 2012). If taking the

average of 25,000 RMB/ton, or US dollar $3970/ton (exchange

ratio of 6.3 RMB to 1 dollar), the value of glyphosate in the

wastewater (1.2e1.5 g/L) is estimated to be 2.4e3.0 $/m3, noting

that the estimation assumes that a recovery of only 50% as

listed in Table 3. The cost to recover the glyphosate using

nanofiltration takes less than 10% of the value gained from the

recovered glyphosate. Although the further concentration of

the glyphosate and crystallization of the chemical may add up

to the cost, it is still a surprising return for the herbicide man-

ufacturer if the recovered glyphosate increases the productiv-

ity. Therefore, the big margin may act as a strong impetus for

the industry to invest in such a field. Our ongoing research is

Table 3 e Cost estimation of the nanofiltration membranes in recovery of the glyphosate from the wastewater.

Items Values Remarks

Assumptions System capacity (m3/h) 700 100k ton glyphosate per year; 5/1 mother liquor/glyphosate;

Diluted 10 times 700 m3/h

Working days (days) 300

membrane life time (years) 3 Suitable pre-treatment

membrane permeability (LMHBar) 10 Based on this workaRecovery rate (%) 50 30e35% for seawater reverse osmosis desalination

Membrane cost ($/m2) 50 Include support membrane and coating

Module (m2) 35 8 inch, 1.5 m module

Module No (pieces) 800

Operation Pressure (bars) 5 Based on this work

Membrane cost ($) 1,400,000

System cost ($) 4,200,000 As a rule-of-the-thumb, triple the membrane cost

(Baker and Lokhandwala, 2008)aSystem life time (years) 30 Assumption based on seawater reverse osmosis

desalination plant

Labor cost ($/year) 80,000 4 full time operator

Cost of electricity ($/KWhr) 0.1

Estimated cost Electricity ($/m3) 0.035 Power efficiency ¼ 0.8

Membrane replacement ($/m3) 0.093aPre-treatment ($/m3) 0.035 Assume the same as seawater desalination

Manpower ($/m3) 0.016 Based on labor costaMaintenance ($/m3) 0.017 2% of system cost every year

Depreciation in system ($/m3) 0.028 System cost divided by the amount of the annual

treated water

Total Cost ($/m3) 0.224

a Reference (Atikol and Aybar, 2005).

wat e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 0 6 5e2 0 7 4 2073

focusing on the preparation of a membrane module for a pilot

test in order to obtain realistic data to convince the herbicide

industry to adopt this new type of membrane process.

4. Conclusions

A composite hollow fiber nanofiltration membrane was pre-

pared by coating commercial ultrafiltration membrane with

sulfonated amorphous PEEK for the separation of glyphosate

from highly saline wastewater. Optimization of the mem-

brane preparation parameters was carried out. Membranes

with satisfactory properties were obtained by coating SPEEK of

1.5 wt. % solution onto a commercial UF membrane. The

membranes showed rejection toward salt in the order of

RNa2SO4 > RNaCl > RCaCl2 , indicating negatively charged surface

coating was realized.

The treatment of glyphosate effluent solution with the

composite NF membrane showed that the negatively charged

NF membrane separated glyphosate from NaCl efficiently,

which showed a new direction for the recovery of glyphosate

with a low energy process. Moreover, the use of commercial

hollow fibermembrane support holds promise for the scale up

application of this process. The coarse cost analysis based on

the present bench test indicates that the total cost for treating

glyphosate wastewater is about 0.224 $/m3, which is much

lower than the gain in recovery of the glyphosate to increase

the productivity. The future research will focus on the prep-

aration of a membrane module for a pilot test in order to

obtain realistic data to convince the herbicide industry to

adopt this new type of membrane process.

Acknowledgments

The authors would like to thank the partial financial support

from National Science Fund China (Project No 20976083,

21176119), the National Key Basic Research Program of China

(973 Program) (Project No 2012CB932800), and TMSR from

Chinese Academy of Sciences (Project No. XDA02020100),

China-Israel Joint Research Program from Ministry of Science

and Technology (MOST).

r e f e r e n c e s

Al-Zoubi, H., Omar, W., 2009. Rejection of salt mixtures from highsaline by nanofiltration membranes. Korean Journal ofChemical Engineering 26 (3), 799e805.

Atikol, U., Aybar, H.S., 2005. Estimation of the water productioncost in feasibility analysis of RO systems. Desalination 184,253e258.

Baker, R.W., Lokhandwala, K., 2008. Natural gas processing withmembranes: an overview. Industrial & Engineering ChemistryResearch 47, 2109e2121.

Ballet, G.T., Hafiane, A., Dhahbi, M., 2007. Influence of operatingconditions on the retention of phosphate in water bynanofiltration. Journal of Membrane Science 290 (1e2),164e172.

Bishop, M.T., Karasz, F.E., Russo, P.S., 1985. Solubility andproperties of a poly(aryl ether ether ketone) in strong acids.Macromolecules 18, 86e93.

Bowen, W.R., Mohammad, A.W., 1998a. Diafiltration bynanofiltration: prediction and optimization. AIChE Journal 44(8), 1799e1812.

wat e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 0 6 5e2 0 7 42074

Bowen, W.R., Mohammad, A.W., 1998b. A theoretical basis forspecifying nanofiltration membranes-dye/salt/water streams.Desalination 117, 257e264.

CNSC, 2006. Standards for Drinking Water Quality (GB 5749-2006).China National Standardizing Committee.

Frank, M., Bargeman, G., Zwijnenburg, A., Wessling, M., 2001.Capillary hollow fiber nanofiltration membranes. Separationand Purification Technology 22-23, 499e506.

Freuze, I., Jadas-Hecart, A., Royer, A., Communal, P.-Y., 2007.Influenceof complexationphenomenawithmultivalent cationson the analysis of glyphosate and aminomethyl phosphonicacid in water. Journal of Chromatography A 1175 (2), 197e206.

Gasnier, C., Dumont, C., Benachour, N., Clair, E., Chagnon, M.C.,Seralini, G.E., 2009. Glyphosate-based herbicides are toxic andendocrine disruptors in human cell lines. Toxicology 262 (3),184e191.

Glyphosate Price, Aug. 7, 2012. Glyphosate Price Increase Boostingup Chinese Companies. http://news.agropages.com/News/NewsDetail%96-7540.htm.

He, T., 2001. Composite hollow fiber membranes for ionseparation and removal. Ph. D Thesis. University of Twente,Enschede.

He, T., Frank, M., Mulder, M.H.V., Wessling, M., 2008. Preparationand characterization of nanofiltration membranes by coatingpolyethersulfonehollowfiberswithSulfonatedpoly (ether etherketone) (SPEEK). Journal of Membrane Science 307, 62e72.

Hilal, N., Al-Zoubi, H., Mohammad, A.W., Darwish, N.A., 2005.Nanofiltration of highly concentrated salt solutions up toseawater salinity. Desalination 184 (1e3), 315e326.

Ismail, A.F., Lau, W.J., 2009. Influence of feed conditions on therejection of salt and dye in aqueous solution by different

characteristics of hollow fiber nanofiltration membranes.Desalination and Water Treatment 6, 281e288.

Kim, K.-J., Chowdhury, G., Matsuura, T., 2000. Low pressurereverse osmosis performances of sulfonated poly(2,6-dimethyl-1,4-phenylene oxide) thin film compositemembranes: effect of coating conditions and molecularweight of polymer. Journal of Membrane Science 179 (1e2),43e52.

Li, L., Zhang, J., Wang, Y., 2003. Sulfonated poly(ether etherketone) membranes for direct methanol fuel cell. Journal ofMembrane Science 226 (1e2), 159e167.

Liu, H., Zhang, H., Liu, J., Luo, Q., 2004. Determination theconcentration of different molecular weight of polyethyleneglycol by spectrophotometry. China Academic Journal ofHenan Chemical Industry 5, 36e37.

Petersen, R.J., 1993. Composite reverse osmosis and nanofiltrationmembranes. Journal of Membrane Science 83 (1), 81e150.

Richards, L.A., Vuachere, M., Schafer, A.I., 2010. Impact of pH onthe removal of fluoride, nitrate and boron by nanofiltration/reverse osmosis. Desalination 261 (3), 331e337.

Spethf, T.F., 1993. Glyphosate removal from drinking water.Journal of Environmental Engineering 119, 1139e1157.

Sukitpaneenit, P., Chung, T.-S., 2009. Molecular elucidation ofmorphology and mechanical properties of PVDF hollow fibermembranes from aspects of phase inversion, crystallizationand rheology. Journal of Membrane Science 340 (1e2),192e205.

Xie, M., Liu, Z., Xu, Y., 2010. Removal of glyphosate inneutralization liquor from the glycine-dimethylphosphitprocess by nanofiltration. Journal of Hazardous Materials 181(1e3), 975e980.