Membrane performance and application of ultrafiltration.pdf

8/19/2019 Membrane performance and application of ultrafiltration.pdf

1/10

Membrane performance and application of ultrafiltrationand nanofiltration to ethanol/water extract of Eucalyptus bark

Paula C.R. Pinto ⇑, Inês F. Mota, José M. Loureiro, Alírio E. Rodrigues

LSRE – Laboratory of Separation and Reaction Engineering – Associate Laboratory LSRE/LCM, Faculdade de Engenharia, Universidade do Porto, Rua Dr. Roberto Frias, 4200-465

Porto, Portugal

a r t i c l e i n f o

Article history:

Received 7 January 2014

Received in revised form 26 April 2014

Accepted 28 April 2014

Available online 9 May 2014

Keywords:

Ultrafiltration

Nanofiltration

Ethanolic solutions

Bark extract

Polyphenols

Carbohydrates

a b s t r a c t

The aim of this work is to promote the enrichment of an ethanolic extract of Eucalyptus globulus bark in

polyphenolic compounds relatively to other compounds such as carbohydrates. Several flat sheet mem-

branes were tested with water and ethanol solutions (52% v/v and 80% v/v) assessing to permeability.

Rejections to gallic and tannic acids and maltose were evaluated for nanofiltration membranes and for

the ultrafiltration membrane of lower cut-off. The dependence of permeability and rejection relative to

ethanol percentage is discussed giving new insights about the membrane performance towards

ethanol/water solutions. Among the tested membranes, two ultrafiltration (JW 30,000 Da and PLEAIDE

5000 Da) and one nanofiltration (SolSep 90801) membranes were selected to the concentration process

of an ethanolic extract of Eucalyptus globulus bark produced at previously optimized conditions. The per-

formance of the three membranes was evaluated concerning polyphenolic compounds and carbohydrate

composition. The volume reduction factor was 1.76. JW membrane revealed the lowest total decrease on

permeability (53%) relative to the initial. All the three membranes showed selective retention of polyphe-

nolic compounds, however JW promoted the highest enrichment of formaldehyde-condensable tannins

(fcT) and proanthocyanidins (Pac) (17% and 28%, respectively). The final composition of the retentate

(in % weight/dry weight) was: TPC 39%, fcT 46%, Pac 38%, GalT 3.2% and TC 15%. The detailed sugar anal-

ysis revealed that some arabinose- and rhamnose-containing oligo/polysaccharides are preferentiallyretained, while those with glucose and galacturonic acid moieties are transported through the membrane

to permeate stream. Finally, cleaning performance of membranes was evaluated and 80–100% flux

recoveries were attained.

2014 Elsevier B.V. All rights reserved.

1. Introduction

The development of biorefinery platforms is currently undergo-

ing rapid expansion. Pulp and paper industries have a privileged

position due to the availability of side-streams lignocellulosicmate-

rials usually classified as by-products, such as bark which is a dis-

posal in mill site where the logs are debarked. This is the case of pulp plants in Portugal which produces about 124,000 tons of Euca-

lyptus globulus bark per medium size industrial unit. Bark is further

integrated in themill operation as energy source. The basic chemical

composition of bark and wood is similar concerning the major

macromolecular components: lignin, cellulose and hemicelluloses

[1,2]. However, the extractive and inorganic content is usually

higher in bark than in wood. This is one of the reasons why bark

has not been used for pulp production. Among the undesired

extractive fractions is the polyphenolic fraction. This is composed

by simple phenolics such as gallic and ellagic acids, flavonoids,

complex glycosides of phenolic compounds [3,4], hydrolysable tan-

nins, and proanthocyanidins [5,6], often called condensed tannins.

The awareness on these compounds is growing up due to theirproperties and biological activities with emerging applications on

cosmetics, nutricosmetic and fortified foods or supplements indus-

tries turning it on high added-value additives or active principles

[7,8].

In this perspective, E. globulus bark is a potential raw material to

produce polyphenolic enriched extracts. In our previous work, the

optimum conditions (time, temperature and ethanol %) for the

extraction of polyphenolic compounds from E. globulus bark were

reported. The extract produced at optimum conditions (OC extract)

was obtained in ethanol/water solution (52/48, v/v) and it

demonstrated important biological activity. The yield was 50 g of

material per kg of bark with 1/3 of the extracted material being

http://dx.doi.org/10.1016/j.seppur.2014.04.042

1383-5866/ 2014 Elsevier B.V. All rights reserved.

⇑ Corresponding author. Address: LSRE – Laboratory of Separation and Reaction

Engineering, Faculty of Engineering, Chemical Engineering Department, University

of Porto, Rua Dr. Roberto Frias s/n, 4200-465 Porto, Portugal. Tel.: +351 22 041

3606; fax: +351 22 508 1449.

E-mail address: [email protected] (P.C.R. Pinto).

Separation and Purification Technology 132 (2014) 234–243

Contents lists available at ScienceDirect

Separation and Purification Technology

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / s e p p u r

http://dx.doi.org/10.1016/j.seppur.2014.04.042mailto:[email protected]://dx.doi.org/10.1016/j.seppur.2014.04.042http://www.sciencedirect.com/science/journal/13835866http://www.elsevier.com/locate/seppurhttp://www.elsevier.com/locate/seppurhttp://www.sciencedirect.com/science/journal/13835866http://dx.doi.org/10.1016/j.seppur.2014.04.042mailto:[email protected]://dx.doi.org/10.1016/j.seppur.2014.04.042http://crossmark.crossref.org/dialog/?doi=10.1016/j.seppur.2014.04.042&domain=pdf


2/10

of phenolic nature (assessed by Folin–Ciocalteu method for

quantification of total phenolic compounds) [1]. Envisaging the

fractionation and the increase of the polyphenolic fraction as the

next step in the valorization process, membrane processing of OC

extract was carried out. Based on its principle, membranes

processing should lead to a fractionation of the polyphenolics

and other components in the extract according to their molecular

weight, hydrodynamic volume (size and shape of the hydrated/

solvated molecule) and solvent–solute–membrane interactions.

However, the adsorption and the build-up of a gel layer may act

as a secondary membrane, changing both solute retention and

permeate flux rate [9]. Moreover, the performance of a systemstrongly depends on the feed characteristics, operating conditions,

membrane, and system configuration.

Membrane separations have been applied to fractionate and

purify polyphenolic rich streams from several biomass resources

as recently reviewed [10]. Olive mill wastewaters [11–14], extracts

of grape seeds [15] and grape pomace [16,17] are the main exam-

ples of liquid streams derived from industrial activity processed by

ultrafiltration (UF) and/or nanofiltration (NF) for polyphenols

recovery. Concerning woody bark extracts, only one study for tan-

nins recovery by UF was found [18]. Moreover, most of the studies

in literature deal with aqueous solutions/extracts and just a few

report real streams of organic solvent or binary mixture, namely

ethanol/water [19–21].

In this work, seven commercial membranes were characterizedand the impact of solvent composition on membrane performance

was evaluated. Gallic acid (170 g mol1), tannic acid (1701 g

mol1) were used as models for phenolic compounds, and maltose

(342 g mol1) as model for carbohydrates, to test the NF mem-

branes and the UF membrane of lower cut-off. The OC extract

was submitted to UF and NF in concentration mode. The goal

was to evaluate the performance of membrane processing in the

polyphenol enrichment of the E. globulus extract. For this, the flux

declines were evaluated and the compositions of retentates and

permeates were assessed considering total non-volatile solids

(TS), total phenolic compounds (TPC), formaldehyde-condensable

tannins (fcT) quantified as Stiasny number (SN), proanthocyanidins

(Pac), gallotannins (GalT), and sugar composition allowing the

quantification of total carbohydrates (TC).

2. Experimental

2.1. Equipment, membranes and conditioning

Benchtop studies were conducted using a membrane cell sys-

tem Sepa CF II Med/High Foulant System (GE Osmonics, USA) with

an effective area of 0.014 m2 plus a flow meter, a diaphragm pump

Hydra-Cell, model M-3/G-13, (Wanner Engineering, Inc.) with a

frequency inverter (MC07, MovitracB, SEW Eurodrive), and a

manual hydraulic pump (P19, SPX Corporation, USA). The NF/UF

unit withstands a maximum operating pressure of 69 bar, and a

maximum operating temperature of 177 C. The temperature of the feed was assured by a Lauda thermostatic bath (Ecoline

Staredition Re 206) and a coil immersed on the feed tank. The feed

temperature was checked by an electronic contact thermometer

(VT-5 S40, VWR).

The UF and NF flat sheet membranes studied are listed in

Table 1. Aqueous solutions of ethanol (Panreac) were prepared

on a volume/volume basis using deionized water. All membranes

were preconditioned according to the protocol recommended in

the literature [22]. Prior to use, the membranes were first rinsed

with water and soaked overnight. Afterwards, the membranes

were soaked with ethanol solutions starting with 10% (v/v) ethanol

and then with increments of 10–20% ethanol until 52% or 80% (v/v)

ethanol, depending of the programed assays. For the experiences

with water, membranes were simply soaked with water for threetimes and left overnight. The SolSep membranes were directly

washed and conditioned in the working solvent as recommended

by the fabricant. Before operation, each membrane was prepared

by compressing it into the module by means of system hydraulic

pressure (about 10–15 bar more than the operating pressure in

the experiments), using water or ethanol/water solutions at a

transmembrane pressure (TMP) of 1 bar for about 30 min to

remove material from the pores. Then, using fresh solution, the

membranes were submitted to compaction with a TMP 1–2 bar

higher than the operating pressure in the experiments. The perme-

ate flux was measured and usually the time to ensure the steady

state was 1 h.

Ultra-pure water and analytical grade reagents were used for

membrane characterization.

Nomenclature

List of symbols A effective membrane area (m2)Ara arabinoseC p concentration in the permeate (g L

1)C r concentration in the retentate (g L

1)

Gal galactoseGalA galacturonic acidGalT gallotannins (% w/w)Glc glucosefcT formaldehyde-condensable tannins (% w/w) J p volumetric flux through membrane (L m

2 h1)L p membrane permeability coefficient (L m

2 h1 bar1)NF nanofiltrationpHPZC point of zero chargeOC optimum conditionsPac proanthocyanidins (% w/w)Q p permeate flow rate (L h

1)Rha rhamnose

Man mannoseRm membrane resistance coefficient (m

1)R j apparent solute rejection coefficientSN Stiasny number (% w/w)TMP transmembrane pressure (bar)

TPC total phenolic compounds (% w/w)TS total non-volatile solids (g L 1)TC total carbohydrates (% w/w)UF ultrafiltrationV f feed volume (L)V r retentate volume (L)VRF volume reduction factorXyl xylose

Greek lettersl dynamic viscosity of water/solvent (kg m1 s1)

P.C.R. Pinto et al. / Separation and Purification Technology 132 (2014) 234–243 235

http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


3/10

2.2. Operating procedure

2.2.1. Membrane permeability

After membrane conditioning and compaction, the flux was

measured for water and ethanol solutions 52% and 80% at different

TMP for membrane permeability assessment. Besides the intrinsic

resistance of each membrane, these assays allow evaluating the

irreversible fouling or membrane damage after each experiment

with the extract by comparison with measurements carried out

at the same conditions. For UF membranes the applied TMP were

in the range 1–9 bar and for NF the range was 1–30 bar.

Membrane permeability was evaluated in total recirculation

mode (recirculation of permeate and retentate to the feed tank).

The different fluxes were monitored over time after a period of sta-

bilization at a feed flow rate of 4.5 ± 0.25 L h1 and 35 C.

2.2.2. Standard rejections

Membrane selectivity is a measure of the membrane ability to

reject a particular solute in detriment of another due to differences

in interactions between solute, nature of membrane and type of

solvent. It is affected by several operating variables such as TMP,

turbulence near membrane surface triggered by tangential flow,temperature, solute type and concentration, pH, ionic strength, sol-

vent or other factors that can modify the shape and the molecule

conformation.

Feed solutions of 0.6 g L 1 of gallic acid (170 g mol1, Acros

organics, 98%), tannic acid (1701 g mol1, Sigma Aldrich, P98%)

and maltose (342 g mol1, Sigma Aldrich, P98%) were prepared

in water and ethanol/water solutions (52% and 80% ethanol, v/v).

The experiments were conducted at constant feed flow rate and

temperature, 4.5 L h1 and 35 C, respectively. TMP was also

constant at 4 bar for GE membrane, 30 bar for NF270 and 14 bar

for the SolSep, 090801 and 080105 membranes.

Membrane selectivity was evaluated in total recirculation

mode. The feed solution was circulated for about 1 h until the

steady state. Afterwards, at preset time intervals, permeate was

collected and further analyzed. Gallic acid and tannic acid concen-

tration was assessed by the Folin–Ciocalteu assay and the disac-

charide was quantified by HPLC, as described in Section 2.4.

Experiments were carried out from the lowest to the highest

solute molecular weight. Each membrane was then washed with

the solvent or cleaning solution until at least 75% of the initial

solvent flux.

2.2.3. Concentration of the E. globulus bark extract

Several UF and NF membranes were initially tested for perfor-

mance and stability with the ethanolic solutions and, among these,

three membranes were selected to proceed for E. globulus extracts

processing: NF SolSep 090801, UF Pleiade 5 kDa and UF JW 30 kDa.

The selection was based on some important attributes duringmembrane characterization such as solvent flux, applied TMP,

solute rejection coefficients, robustness and stability, and cleaning

cycles to retrieve the initial permeability. The point of zero charge

(pHPZC) of the membranes was determined by the pH drift test

described in the literature [23].

E. globulus bark extractions were performed according to the

optimum conditions found in our previous work [1] but using N2as inert atmosphere. This extract is denoted in this work as

optimum conditions extract (OC extract). Prior to the membrane

separation experiments, the extract was filtered through a polycar-

bonate membrane (Nucleopore, Whatman) with 10 lm of pore size

to remove small particles of biomass, accounting for 0.08 g L 1. The

turbidity and total suspended solids are 411 NTU and 0.50 g L 1,

respectively. The TS content of this extract is 7.9 g/L with the

following composition on % weight/dry weight: TPC 37.9%; fcT

36.0%; Pac 32.1%; GalT 3.1%; and TC 16.0%.

The membrane separation operations were carried out in con-

centration mode, at 35 C and a fixed TMP (14 bar for NFmembrane

and 4 bar for UF membranes) was applied. In concentration mode

the retentate stream is recycled into the feed tank whereas the

permeate stream is separately collected, resulting in a continuous

volume decline in the feed tank. Volume permeation fluxes were

measured up to a volume reduction factor (VRF) of about 1.76.

VRF is the ratio between the initial feed volume and the remainingvolume of retentate at a given operating time. The feed flow rate

was adjusted to 4.5 L h1.

Permeate and retentate samples were collected during each run

for composition analysis regarding the same parameters used for

OC extract characterization: TS, TPC, fcT quantified by SN, Pac, GalT

and TC.

2.3. Cleaning and storage

In this study, immediately after the experiments, the system was

initially flushed with fresh solvent (water, ethanol 52% or ethanol

80%, depending on the experiment) with no TMP. The solution

was discharged and subsequent cleaning cycles (as described belowfor standards and for extract experiments)in fullrecirculation mode

and applying a TMP of 1 bar for 60 min were performed as many

times as necessary to attain at least 75% of the initial feed flowrate.

The feed flowrate was 4.5 L h1 and the maximum temperature

40 C. Alkaline solutions for cleaning were prepared in water, etha-

nol 52% or 80% according to each case. Supplier recommendations

were followed: SolSep membranes were cleaned using NaHCO30.1 M andfor Osmotic andOrelis membranes NaOH 0.1 M was used.

For the experiments using standard compounds (gallic acid,

maltose or tannic acid), after flushing, the first two cleaning cycles

were performed with the corresponding solvent (water, ethanol

52% or ethanol 80%) used in the experiment. After this, the

permeability was evaluated and, if necessary, subsequent cleaning

cycles with alkaline solution (water, ethanol 52% or 80%, as thecase may be) were performed for 60 min and 1 bar of TMP.

Table 1

Characteristics of the NF/UF membranes.

Membrane Rejection to typical solutes/MWCO Producer Composition Operational pH T max (C) P max (bar)

80105 >99% colorant (500 Da) in ethanol SolSep Polyamide derivative 1–12 120 20

90801 50% NaCl 90% colorant (350 Da) in ethanol SolSep Polyamide derivative 1–14 80 20

NF270 97% MgSO4 D ow–Filmte c Polya mide t hin film c omposite 3–10a

1–12b45 41

GE 1000 GE Osmonics Polyamide composite 2–11 – –

PLEAIDE 5000 Orelis Environnement Polyethersulfone (PES) 3–14 50 – JW 30,000 GE Osmonics Polyvinylidene fluoride (PVDF) 1–11 75 –

EW 60,000 GE Osmonics Polysulfone (PS) 0.5–13 – –

a Continuous operations.b Short-term cleaning (30 min).

236 P.C.R. Pinto et al. / Separation and Purification Technology 132 (2014) 234–243


4/10

In the case of OC extract experiments, after the first flush, the

system was additionally flushed with the alkaline solution (in

for 10 min without TMP). The cleaning cycles were performed

with alkaline solution of 52% ethanol (0.1 M NaOH or NaHCO3depending on the membrane). After the alkaline washing, the

system was first flushed with 52% ethanol followed by a double

washing at TMP of 1 bar for 30 min to remove the residual alkali

of the membrane. Finally, the permeability was evaluated usingfresh solution of 52% ethanol.

After cleaning, each membrane was stored in 0.5% Na2S2O5aqueous or ethanolic (52% or 80%) solution.

2.4. Analytical methods

For rejection evaluation, gallic acid and tannic acid were quan-

tified by Folin–Ciocalteu method as described previously [1] using

gallic acid or tannic acid as standards. The disaccharide maltose

was quantified in a HPLC Knauer unit equipped with a Smartline

5000 online degasser, a Smartline 1000 quaternary pump, and a

Smartline 2300 refractive index detector. The analytical column

was SHODEX SC-1011 (300 8.0 mm, 6 lm) using ultra-pure

water as eluent at 0.7 mL min1. Chromatograms were run at

80 C and the volume of injection was 20 lL.

Quantification of TS, TPC, fcT, Pac and TC were performed as

described in our previous work [1]. Briefly, TS were quantified by

weighting the dried extract/fractions; TPC quantification was

based on Folin–Ciocalteu method using gallic acid as standard;

fcT quantified by SN is a comparative parameter which includes

all the monoflavonoids, biflavonoids and oligomers susceptible to

form methylene linkages and polymerizing through the reaction

with formaldehyde; the value is obtained by the weight of the

produced precipitate [24,25]; Pac was quantified by Bate-Smith

reaction: in this method, proanthocyanidins (also referred as

condensed tannins) are cleaved to yield anthocyanidins [26,27]

and further quantified by absorbance using a mimosa extract as

standard. TC analysis was performed by acid methanolysis

followed by identification and quantification of sugar moieties byGC–MS and GC–FID, respectively.

GalT estimation was based on the quantification of gallic acid

methyl ester liberated in the methanolysis of the extract [28]:

dried samples (10 mg) were treated with anhydrous methanolic

HCl (2 mL) 2 M for 17 h at 100 C, the time needed to obtain a

maximum of gallic acid methyl ester from the extract and

fractions. After completion, the solutions were cooled to room tem-

perature, evaporated at reduced pressure, redissolved in methanol

and diluted in methanol/water 90/10 containing 0.1% HCOOH

before injection. Gallic acid methyl ester was quantified using an

analytical column YMC-Park ODS-A in the same HPLC described

above equipped with Smartline UV/DAD Detector 2600 operating

at 280 nm. The separation was performed at room temperature

and eluent flowrate 0.4 mL min1

using a gradient composed bytwo solutions: A – water/methanol: 95/5 (v/v) containing 0.1%

(v/v) formic acid and B – 5:95 (v/v) with 0.1% (v/v) formic acid.

The gradient program was 0–3.30 min 90% A, 6.7–20.0 min 80%

A, 40.0–43.3 min 40% A, 43.3–46.7 min 0% A. The calibration was

performed with gallic acid (P98%, Sigma). Standard solutions

and samples were filtered before injection using a 0.2 lm

disposable filter (Millipore). The quantification data were reported

as gallic acid equivalent. The content on GalT was calculated by

subtracting the content of gallic acid in the extract/fractions to

the value resulting from the methanolysis.

2.5. Membrane parameters and calculations

The volumetric flux through membrane ( J p, L m2

h1

) isgiven by

J p ¼ Q p A

ð1Þ

where Q p is the permeate flow rate (L h1) and A is the effective

membrane area (m2).

Considering negligible the osmotic pressure at membrane sur-

face and in the permeate, J p is proportional to the differential pres-

sure across membrane or transmembrane pressure (TMP, bar) and

given by the following equation:

J p ¼ TMP

lRmð2Þ

The membrane permeability coefficient is a common parameter

to evaluate the performance of membranes, representing the liquid

crossing the membrane per time unit, per membrane area unit and

per TMP unit. Experimentally, it is calculated by the slope of J p vs

TMP for the system:

J p ¼ L pTMP ð3Þ

where l is the dynamic viscosity of water/solvent (kg m1 s1), Rmis the membrane resistance coefficient (m1) and L p the membrane

permeability coefficient (L m2 h1 bar1).

The volume reduction factor (VRF) is given by:

VRF ¼ V f V r

ð4Þ

with V f representing feed volume (L) and V r the retentate volume

(L).

The apparent solute rejection coefficient (R j) for gallic acid,

maltose and tannic acid is defined as,

R j ¼ 1 C pC r

ð5Þ

being, for each compound family, C p concentration in the permeate

(g L 1) and C r concentration in the retentate (g L 1).

3. Results and discussion

3.1. Membranes characterization

Membranes were characterized concerning their L p in water,

ethanol 52% and 80% at 35 C and constant feed flowrate of

4.5 L h1. The purpose of studying water and ethanol 80% was to

provide additional data to evaluate the effect of the ethanol

percentage on the membranes performance. Moreover, in the

previous published work [1], where the optimum of 52% ethanol

for TPC extraction was reported, it was also found that the extract

produced with 80% ethanol presented a high biological activity. If

the fractionation of this extract would be considered in the future,

the characterization of the membranes with the corresponding

0

50

100

150

200

250

80105 90801 NF270 GE PLEAIDE JW EW

L p ( L m - 2 h - 1 b a r - 1 )

Water

Ethanol 52%

Ethanol 80%

membrane

0

5

10

80105 90801 NF270 GE

Fig. 1. Permeability of polymeric membranes for water, ethanol 52% and ethanol80% at 35 C and feed flowrate 4.5 L h1.


http://-/?-http://-/?-


5/10

solvent would be also valuable. Fig. 1 shows the permeability for

each membrane in water, ethanol 52% and ethanol 80%.

UF membranes were chosen based on supplier reference for

molecular cut-off. The aim was to test a widespread cut-off range

(from 1000 Da to 60,000 Da) and different membrane composition,

as stated in Table 1. Concerning NF membranes, two polyamide-

based membranes, SolSep 80105 and 90801, and also the NF270,

a piperazine-based polyamide membrane [29] were tested.Pure water and aqueous solutions of ethanol 52% (the solvent

composition giving the best results on extraction experiments

[1]) and ethanol 80% were used for permeation studies to evaluate

their effect on membrane performance. Since SolSep 80105 and

90801 are typically membranes for organic solvents, these were

not tested with 100% water. Solvent permeation through mem-

branes is a diffusion process enhanced by the interaction between

the solvent and the hydrophilic and/or hydrophobic domain of the

membrane; the affinity of water to a membrane with hydrophilic

properties would promote the transport, while for ethanol/water

system, in the same membrane, flux would be considerably lower

due to the limited hydrogen bonding capability of the ethanol. In

general, for membrane with dominant hydrophilic properties, the

greater the solvent polarity the highest is the flux and the oppo-

site is observed for membranes with hydrophobic characteristics

[30]. From Fig. 1 it is clear that as the membrane cut-off increases,

the highest is the water permeability due to the decrease of the

resistance offered in fluid transport. For all the tested membranes,

a high decrease on L p was observed from pure water to 52% eth-

anol, which is in accordance with the increase of viscosity of

fluids: from water (0.7202 103 Pa s, 35 C) to ethanol 52%

(1.7896 103 Pa s, 35 C) [31]. However, a L p decrease was also

observed for ethanol 80%, in spite of the lower viscosity of this

solution (1.436 103 Pa s, 35 C) [31] compared to ethanol 52%.

This clearly shows that solvent viscosity is not the main feature

affecting flux when using ethanol solutions. Other phenomena,

such as the decrease of solvent polarity had impact on L p for all

above mentioned membranes, except for EW. In this case the L p

increases when changing 52% to 80% of ethanol. According toEq. (2) there is a linear correlation between flux and the inverse

of viscosity if the flux decrease is dominated by solution viscosity

[30,32]. Fig. 2 shows the plot of J p and 1u for the studied

membranes. EW was the only one showing a linear relationship

demonstrating that the viscosity of the solution is the dominant

parameter affecting solvent flow through membrane. For the

other membranes the effect of other solvent characteristics has

more impact than viscosity alone. The solvent/solvent mixture

characteristics with already reported impact on L p are surface

tension, molar volume and dielectric constant [33]. Surface ten-

sion and dielectric constant decrease in the order water, ethanol

52% and ethanol 80% [31,34,35], while molar volume increases

in this order. Molar volume could, at least, partially explain the

L p decrease for NF membrane in the order water to ethanol 80%,

while surface tension would have the opposite effect on L p. How-

ever, dielectric constant of the mixture and the related factor,

polarity, could have a practical effect on membrane due to theinteraction (or its absence) with the polymeric material. This

effect depends also on surface energy of the polymer as stated

in the literature [30].

Hence, for membranes with predominant hydrophilic character,

the flux would be considerably lower for ethanol/water mixtures

than for water due to the limited hydrogen bonding capability of

the ethanol delaying and impairing the transport through the

membrane. In accordance, other authors have concluded that the

hydrophilicity and porosity are the most important characteristics

of membranes affecting the fluid transport, while viscosity and

polarity are the solvent properties with highest influence on per-

meability [36].

From this preliminary study, it was possible to evaluate the

behavior of the different membranes concerning L p as the net

result of different factors. Among the UF membranes, those of

medium cut-off and simultaneously presenting L p values higher

than 35 L m2 h1 bar1 were selected for the concentration

assays: membranes JW and PLEAIDE. Among the NF membranes,

the 090801 is the membrane with highest L p for 52% ethanol

(5.8 L m2 h1 bar1) and, thus, the most promising for the concen-

tration process. In spite of this, a study of rejection coefficients for

three compounds was performed for NF membranes: two phenolic

compounds (gallic acid and tannic acid) and one disaccharide,

maltose, as general representatives of the families of compounds

present in the real extract. In this study, the membrane GE was

included due to its small cut-off. The overall results are presented

in Fig. 3.

Among all membranes, the NF270 presents the highest rejection

for the three compounds. NF 270 is a NF membrane stated to havehigh rejection indexes for simple sugars and salts in water (about

90% for glucose in water at 37 C and at 8 bar [37] and more than

97% to MgSO4 (manufacturer data – Table 1). However, Restolho

and co-workers [38] have reported rejections of 52% and 34% (in

water, 25 C, 18 bar and 2.0 L min1) for glucose and xylose,

respectively. The values found in this work for rejections are 40%

for gallic acid, 67% for maltose and 95% for tannic acid in 52%

ethanol. The membrane 80105 presented the lowest rejections in

52% ethanol (between 4% and 22%).

0

250

500

750

1000

0.0 0.5 1.0 1.5 2.0

J p ( L m - 2 h - 1 )

1/µ (m s kg-1)

EW

JW

Pleaide

GE

NF270

Ethanol

80%

Ethanol

52%

0

5

10

15

20

0.0 0.5 1.0 1.5

GE

Water

Fig. 2. Correlation between J p and 1l

for membranes EW, JW, PLEAIDE, GE and NF270 for water, ethanol 52% and ethanol 80% at 35 C and feed flowrate 4.5 L h1.



6/10

In the conditions of this work, the relative rejection in the mem-

branes 90801, NF270 and GE follows the relative molecular weight

order (gallic acid, maltose, tannic acid) in both solvents. However,

for the membrane 80105 a change was observed: the rejection to

gallic acid is higher than to the disaccharide. These observations

can only be explained by different membrane characteristics since

solvent–solute interactions are the same as for the other mem-

branes. The polarity differences between the molecules can leadto a modification of the expected relative rejection based on molec-

ular weight due to polarity itself (interaction with membrane) and

hydrodynamic volume (interaction with the solvent). The trans-

port of the gallic acid relative to the disaccharide (the phenolic

compound is less polar than the disaccharide) was improved by

an additional affinity with this membrane. Among all membranes,

the 80105 seems to be the most hydrophobic. This is in accordance

with the increase of the rejection of the three compounds (but

maintaining the relative rejections) with the increase of ethanol

percentage (52% to 80%) (and consequently, with decrease on sol-

vent polarity) observed for membrane 80105 – Fig. 3A and B.

The most relevant difference on rejection between ethanol 52%

and ethanol 80% concerns to tannic acid: for GE and for 90801

membranes, the rejection decreases with the change of the per-centage of ethanol from 52% to 80%; this means that the presence

of ethanol promotes the transport of this compound through the

membrane pores. This could be explained based on interaction of

solvent–membrane and the solvation of the pores: for membranes

with some hydrophilic character, the hydration of the pore wall

occurs, leading to a decrease on their effective pore size [36]; con-

sidering that these membranes are not so effectively solvated by

ethanol as by water, the increase of ethanol percentage led to an

higher effective pore diameter and thus, to a lower hindrance to

the transport of the tannic acid reflected in a higher rejection in

52% ethanol than in 80% ethanol. This effect is noticeable for tannic

acid probably due to its higher molecular weight. For membrane

NF270, rejection in 52% and 80% ethanol is similar for all com-

pounds. Although for this membrane no effect was perceptiblefrom 52% to 80%, the lower rejections observed for this membrane

(when compared with literature data for water as solvent, as

referred above) is probably due to the introduction of ethanol in

the feed, increasing the effective pore size relative to that obtained

in water.

On the opposite, for membrane 80105, besides the lowest

rejections of all the membranes in 52% ethanol, it was noticed a

high increase on rejection when the solvent is changed to 80% eth-

anol for all compounds. This could be due to the effect of the higheraffinity for ethanol, occurring an effective solvation of the

pores (reducing the permeability to the compounds) since the

solute–solvent is the same as for the other membranes. Thus,

the character/nature of this membrane should be much less hydro-

philic than for the other tested membranes, as was stated before.

These considerations allow a better knowledge about the

behavior of the different NF membranes concerning fluxes and

permeability to different solutes in ethanol 52% and ethanol 80%.

Among the NF membranes and GE, the one selected for processing

the OC extract was the SolSep 090801: the flux on 52% ethanol was

the highest with a high rejection to tannic acid (70%) over the

disaccharide (27%) which is favorable in the perspective of a

selective concentration, considering these two standards and

comparing with the other membranes.

3.2. Concentration of OC extract from E. globulus bark

The next step was to apply the OC extract in concentration

mode using the selected membranes: JW, PLEAIDE and SolSep

090801. Fig. 4 shows the flux along the concentration experiment

regarding operating time (A) and VRF (B) for each membrane.

The solute build up in the membrane boundary layer

established in laminar flow and in the first minutes of operation

is known as concentration polarization. It is the main cause for

the flux deviation from the solvent flux. After that, the fouling

A

0

20

40

60

80

100

80105 90801 NF270 GE

R f ( % )

Membrane

Gallic acid

Maltose

Tannic acid

0

20

40

60

80

100

80105 90801 NF270 GE

R f ( % )

Membrane

Gallic acid

Maltose

Tannic acid

Ethanol 80%

Ethanol 52%

B

Fig. 3. Apparent rejection coefficients to gallic acid, tannic acid and maltose in

ethanol 52% (A) and in ethanol 80% (B) for the NF membranes and for the UF

membrane GE at 35 C and feed flowrate 4.5 L h1.

A

B

0

10

20

30

40

50

60

J p ( L h - 1 m

- 2 )

VCR

JW PLEAIDE 90801

0.0

0.2

0.4

0.6

0.8

1.0

1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8

0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0

J t / J 0

Operation time (h)

JW PLEAIDE 90801

Fig. 4. (A) Normalized permeate flux (ratio of instantaneous permeate flux at time

t , J t , and at initial time, J 0) along the operating time. (B) Instantaneous permeate flux

along the VRF for the extract OC. Conditions: concentration mode, 35 C, feed

flowrate 4.5 L h

1, TMP 4 bar for UF JW and PLEAIDE and 14 bar for NF 90801;operating time for VRF 1.76: 4.3 h – JW; 7.4 h – PLEAIDE; 5.6 h – 90801.


http://-/?-


7/10

due to solute adsorption onto the membrane surface or mem-

brane pore plugging becomes noticeable and contributes to the

flux drop [9]. In the OC extract processing, the decrease pattern

of permeate fluxes is different for the three membranes

(Fig. 4A). 90801 is the membrane with the highest absolute flux

but also with the higher initial decline, notorious up to the VRF

1.37 – Fig. 4B – which corresponds to about 2.5 h of operation.

From this point onward, the membrane 90801 presents lower J p value than the UF membrane JW (Fig. 4B). These differences

suggest that the 90801 membrane is the most susceptible one

to concentration polarization and fouling in the conditions of this

work, thus creating an additional resistance to permeate flux. At

the pH of the OC extract (4.3) the compounds are mainly neutral

and protonated in solution. In the conditions of the processing,

the membrane 090801 (pHPZC 3.3) is negatively charged and

electrostatic attractions should be expected between the surface

and some of the components of the extract. For JW (pH PZC 6.1)

and PLEAIDE (pHPZC 6.4) membranes, electrostatic repulsions are

expected due to the positively charged surface. Thus, considering

this data, fouling should be more severe in the case of PLEAIDE.

However, besides this effect, hydrophobicity of the membranes

would play also an important role in this process. The balance

between electrostatic repulsion and van der Waals interactions

determines the outcomes of membrane fouling, as well as the effi-

ciency of the membrane cleaning after processing.

The initial decline of normalized flux J t / J 0 (Fig. 4A) for UF mem-

branes is the same, suggesting that the concentration polarization

effect is similar for both. However, after about 1 h of operation, the

membrane JW holds up higher normalized and absolute flux. At

the end of the concentration stage, for the same concentration

factor in volume, the flux reductions relative to the initial were

53%, 68% and 85% for JW, PLEAIDE and 90801 membranes, respec-

tively. Among these, JW presents simultaneously the lowest

decrease of flux by fouling, reaching the steady state earlier than

the others, and the highest final flux.

3.3. Retention of polyphenolic compounds: characterization of

retentates

The extract OC contains, as major components, polyphenolic

compounds (polyhydroxy aromatics with an amphiphilic charac-

ter which is controlled by phenolic- and by carboxyl-groups) as

well as carbohydrates, including pectins and fragments of hemi-

celluloses and glucans from bark structure [1]. The approach of

this work was to quantify families of compounds for assessing

the composition of the retentates produced in the concentration

process of the E. globulus bark extract. For TS and TPC, as well

as TC, the permeates were also analyzed, allowing to calculate

the respective apparent rejection coefficient R j depicted in Table 2.

The rejection values for TPC increases in the order JW, PLEAIDE,9080, following the trend of the membranes cut-off; however,

for TS no significant difference was found between PLEAIDE and

90801. Lower rejections were found for TC indicating that it

would be possible the elimination of sugar moieties in the perme-

ate during UF and NF.

Besides TS, TPC and TC, final retentates were also analyzed for

Pac, fcT and TC. The global results are depicted in Fig. 5.

For the same VRF, the membranes showed different perfor-

mances as denoted by the composition of the retentate. The

common feature is an accentuated enrichment effect on fcT

(monoflavonoids, biflavonoids and some oligomers). However,

membranes with lower cut-off (PLEAIDE and 900801) retain more

material carrying phenolic hydroxyl groups, as indicated by TPC in

the respective retentates, probably those compounds of lower

molecular weight. This is not reflected on Pac content, assessed

by butanol-acid assay, since this family is mainly composed by fla-

vonoid oligomers. In accordance, JW membrane promoted higher

enrichment of this parameter than did for TPC and fcT, probably

due to a noteworthy contribution of compounds with high

molecular weight that responds to the butanol-acid method. Onother side, the Pac content of PLEAIDE and 90801 retentate is sim-

ilar to that of OC extract (which was not observed for TPC and fcT).

The reason for that could be a preferential entrapment or adsorp-

tion of Pac at the membranes contributing for fouling, as assessed

by the mass balance using the content on permeate for PLEAIDE

and 90801 (data not presented). This also occurs for JW, but in a

lower proportion. In accordance, JW showed the lowest flux

decline during the concentration process (Fig. 4).

Hydrolysable tannins (GalT) were quantified by the conventional

methodology: gallic acid analysis before and after methanolysis.

For this, gallic acid present in the extract is previously quantified

and them this value is deducted to the value quantified after acid

methanolysis. Acid methanolysis is the process of cleaving ester

linkages between monosaccharide and gallic acid units composingGalT, one of the typical structures of hydrolysable tannins. The

results for OC extract, retentates and permeates from each

membrane are displayed in Fig. 6.

The contribution of GalT (estimated as gallic acid) as a parcel of

polyphenolic compounds in the OC extract, permeates and reten-

tates is low (about 3% of the dried extract weigh). This could be

due to a natural low content on hydrolysable tannins in the extract

or due to ester bonds hydrolysis already in the extraction process.

Nevertheless, GalT quantification is a practical indication of the

membrane performance for this type of compounds. The mem-

brane JW presents higher permeability to GalT than 90801 and

PLEAIDE as revealed by the lower content on JW retentate. This

result is in accordance with the differences already referred for

TPC, suggesting that GalT have significant impact on the TPCquantification.

Table 2

R j for TS, TPC and TC during the concentration of OC extract with the membranes JW,

PLEAIDE and 90801. Conditions: 35 C, feed flowrate 4.5 L h1, TMP 4 bar for UF (JW

and PLEAIDE) and 14 bar for NF (90801); operating time for VRF 1.76: 4.3 h – JW;

7.4 h – PLEAIDE; 5.6 h – 90801.

Membrane TS TPC TC

JW 0.58 0.75 0.12

PLEAIDE 0.77 0.85 0.17

90801 0.78 0.92 0.15

0

10

20

30

40

50

TPC fcT Pac TC

% w

/ w

Parameter

Feed (OC extract)

JW

PLEAIDE

90801

Fig. 5. Composition of feed (OC extract) and composition of retentates produced

with membranes JW, PLEAIDE and 90801. The values are represented as % weight/

dry weight of the OC extract (the feed) or dry weight of the retentate obtained for

each membrane. Conditions: concentration mode, 35 C, feed flowrate 4.5 L h1,

TMP 4 bar for UF JW and PLEAIDE membranes and 14 bar for NF 90801 membrane;

VRF 1.76.



8/10


9/10

composition with the original oligo- and/or polysaccharides. The

aim of this analysis was to evaluate if there exists preferential

rejection for any carbohydrate type, as will be discussed below.

The molar fraction for each monosaccharide is presented in Fig. 7.

Glucose (Glc) and galactose (Gal) are the predominant moieties

in the OC extract accounting, together, to almost 50% of the TC. Glc

is certainly coming from an accessible and amorphous fraction of

cellulose and/or starch of the bark (starch was already found inE. globulus wood – [39]); other sources would be glucomanans

and the sugar moiety of some tannins. The linkage or association

of carbohydrates with polyphenols is also well known in both

hydrolysable [40] and condensed tannins [41] as well as in more

simple phenolics as those identified in E. globulus bark [3]. Gal

would arise from hemicelluloses branching [42] and/or from a

pectin fraction [43]. Arabinose (Ara) and galacturonic acid (GalA)

compose about 25% of the TC. As far as we know, there is no study

in the literature about pectins in E. globulus bark. However, the

composition of pectins in general [43] and in woody materials

[44] as well as data on pectins from bark of other species [45],

suggest that GalA, Gal, Ara and rhamnose (Rha) have arisen from

pectins, which were partially extracted with ethanol/water in the

conditions used in this work. Xylose (Xyl) and mannose (Man)

(8–10% of TC, each) are probably part of solubilized hemicellulose.

While Xyl is the main residue of xylans in wood [46], Man is usu-

ally associated with glucomanans, a minor hemicellulosic fraction.

Interestingly, the content of Man in the extract is higher than Xyl,

which is not in accordance with the relative percentages in the

wood [46] and bark [1].

Concerning the composition of the retentates and permeates:

the molar fractions of Xyl and Man are similar in OC, permeates

and retentates, indicative of a similar distribution within the three

carbohydrate fractions due to a non-selective permeation. How-

ever, the molar fractions of Ara, Rha, Man and Gal are lower in

the permeate, particularly for Ara and Rha with about 50% less than

in OC extract and respective retentates. On the opposite, the TC

fraction in permeate became enriched in Glc and particularly in

GalA. For permeate produced by membrane 90801, the GalAcontent in the TC increased about 2-fold. The same trend of reten-

tion/permeation was observed for the three membranes. However,

NF membrane (90801) stands out by the lowest and highest rela-

tive rejection for Rha and GalA, respectively (Fig. 7). Considering

these observations, it is possible to conclude that the membrane

processing of OC extract leads to a modification of the relative

composition of carbohydrate fraction.

3.5. Cleaning

One additional factor taken into account in the membrane

selection is the evaluation of the initial permeability recovery

through cleaning cycles. The colloidal nature of polyphenolic com-

pounds, as well the co-extracted oligomeric or polymeric polysac-charides contributes for the gel layer on membrane surface.

However, this phenomenon is usually reversible by cleaning with

the same solvent of the extract. More difficult to overcome is the

membrane fouling due to adsorption or internal pore plugging.

Usually, this phenomenon is the cause of internal fouling resis-

tance [47] and a chemical treatment is necessary to restore the

membrane characteristics. Alkaline washing was required for all

membranes applied for the OC extract concentration and it was

applied as described in the experimental part. Fig. 8 summarizes

the influence of cleaning cycles on permeability recovery.

The membrane JW recovered the initial permeability after the

first cleaning cycle while PLEAIDE with two cleaning cycles

restored 80% of the initial permeability. For the cleaning of NF

membrane, NaOH solutions are not recommended; therefore, aweaker base, NaHCO3, was applied. After the second cycle, this

membrane has restored 78% of permeability, achieving the initial

permeability with a third cycle (Fig. 8). Using buffered solutions

combined with a detergent would be a good alternative to improve

washing performance of 90801 membranes. The recover obtainedfor the three membranes is within the acceptance limits to con-

tinue using the membranes, meaning that irreversible fouling

was not significant. Nevertheless, new OC extract concentrations

must be performed to evaluate the rejection performance and

the productivity in successive experiments.

4. Conclusions

The aim of this work was to test different membranes for the

concentration process of polyphenolic compounds from an etha-

nolic extract of E. globulus bark. Characterization of the selected

membranes concerning permeability to water, ethanol 52% and

80% and rejection to standards was the first step. When changing

water to ethanol 52%, the permeability decreased between 47%(PLEAIDE) and 80% (EW); further increase of ethanol to 80% led

to an additional decrease of permeability for all membranes,

except for EW (UF) and 80105 (NF) membranes. These effects

are related to ethanol/water properties and solvent interaction

with the membranes. 90801 and GE membranes showed a favorable

rejection ratio tannic acid/disaccharide for ethanol 52%, indicative

that some selective enrichment in polyphenolic compounds

would be possible. Among the tested membranes, JW (30 kDa),

PLEAIDE (5 kDa) and 90801 were selected for concentration

process of the bark extract. The best flux performance during

the concentration was found for JW. The three membranes

promoted an enrichment of fcT (20–30%) for a VRF of 1.76. JW

promoted the highest concentration of Pac and lower rejection

for TC, what is advantageous considering the purpose of the mem-brane processing. GalT, the fraction of hydrolysable tannins

detected in the extract, were preferentially rejected by 90801

and PLEAIDE but not by JW membrane. The detailed carbohydrate

analysis showed some selective permeation to glucose and galact-

uronic acid-containing oligo-/polysaccharides. Higher rejection of

rhamnose and arabinose could be related to the association

of these moieties with polyphenolic compounds. Flux recoveries

of 80–100% were attained for all the membranes. However, JW

was the easiest membrane to clean.

Membrane process was successfully applied for concentration

of an ethanolic bark extract achieving, with a VRF of 1.76, an

enrichment of polyphenolic compounds of flavonoid nature.

This process could be the primary step in a separation process

envisaging the purification of the Pac fraction of this extract forhigh-added value applications.

0

20

40

60

80

100

0 1 2 3 4 5

F l u x r e c o v e r y ( % )

Cleaning cycles

JW

PLEAIDE

90801

Fig. 8. The influence of cleaning cycles on permeability recovery of membranes JW,

PLEAIDE and 90801.



10/10

Acknowledgements

This work was carried out under the Project BIIPP No.

11551 – Integrated Biorefinery in Pulp and Paper Industry – funded

by the European Regional Development Fund (ERDF) through the

Operational Programme for Competitiveness Factors (POFC) of

the National Strategic Reference Framework (NSRF). This work

was co-financed by FCT and FEDER under Programme COMPETE(Project PEst-C/EQB/LA0020/2013).

Eng. Maria Eduarda Baptista and Dr. Sergio Morales Torres (LA

LSRE/LCM) are acknowledged for support in some of the

permeability and pHPZC assays, respectively.

References

[1] I. Mota, P.C. Rodrigues Pinto, C. Novo, G. Sousa, O. Guerreiro, A.R. Guerra, M.F.Duarte, A.E. Rodrigues, Extraction of polyphenolic compounds from Eucalyptus

globulus bark: process optimization and screening for biological activity, Ind.Eng. Chem. Res. 51 (2012) 6991–7000.

[2] I. Miranda, J. Gominho, I. Mirra, H. Pereira, Fractioning and chemicalcharacterization of barks of Betula pendula and Eucalyptus globulus, Ind. Crop.Prod. 41 (2013) 299–305.

[3] S.A.O. Santos, C.S.R. Freire, M.R.M. Domingues, A.J.D. Silvestre, C. Pascoal Neto,Characterization of phenolic components in polar extracts of Eucalyptus

globulus Labill. bark by high-performance liquid chromatography–massspectrometry, J. Agric. Food Chem. 59 (2011) 9386–9393.

[4] E. Conde, E. Cadahía, M.C. García-Vallejo, F. Tomás-Barberán, Low molecularweight polyphenols in wood and bark of Eucalyptus globulus, Wood Fiber Sci.27 (1995) 379–383.

[5] P.C.R. Pinto, G. Sousa, F. Crispim, A.J.D. Silvestre, C. Pascoal Neto, Eucalyptus globulus bark as source of tannin extracts for application in leather industry,ACS Sust. Chem. Eng. 1 (2013) 950–955.

[6] E. Cadahía, E. Conde, B. Fernández de Simón, M.C. García-Vallejo, Tannincomposition of Eucalyptus camaldulensis, E. globulus and E. rudis. Part II. Bark,Holzforschung 51 (1997) 125–129.

[7] T. Stevanovic, P.N. Diouf, M.E. Garcia-Perez, Bioactive polyphenols fromhealthy diets and forest biomass, Curr. Nutr. Food Sci. 5 (2009) 264–295.

[8] M. Royer, M. Prado, M.E. García-Pérez, P.N. Diouf, T. Stevanovic, Study of nutraceutical, nutricosmetics and cosmeceutical potentials of polyphenolicbark extracts from Canadian forest species, PharmaNutrition 1 (2013) 158–167.

[9] M. Cheryan, Ultrafiltration and Microfiltration Handbook, TechnomicPublishing Co. Inc., Lancaster, USA, 1998.

[10] E. Conde, B. Díaz Reinoso, M.J. González-Muñoz, A. Moure, H. Domínguez, J.C.Parajó, Recovery and concentration of antioxidants from industrial effluentsand from processing streams of underutilized vegetal biomass, Food PublicHealth 3 (2013) 69–91.

[11] A. Cassano, C. Conidi, L. Giorno, E. Drioli, Fractionation of olive millwastewaters by membrane separation techniques, J. Hazard. Mater. 248–249(2013) 185–193.

[12] C. Russo, A new membrane process for the selective fractionation and totalrecovery of polyphenols, water and organic substances from vegetation waters(VW), J. Membr. Sci. 288 (2007) 239–246.

[13] E. Garcia-Castello, A. Cassano, A. Criscuoli, C. Conidi, E. Drioli, Recovery andconcentration of polyphenols from olive mill wastewaters by integratedmembrane system, Water Res. 44 (2010) 3883–3892.

[14] A. Cassano, C. Conidi, E. Drioli, Comparison of the performance of UFmembranes in olive mill wastewaters treatment, Water Res. 45 (2011)3197–3204.

[15] B. Santamaría, G. Salazar, S. Beltrán, J.L. Cabezas, Membrane sequences forfractionation of polyphenolic extracts from defatted milled grape seeds,Desalination 148 (2002) 103–109.

[16] B. Díaz-Reinoso, A. Moure, H. Domínguez, J.C. Parajó, Ultra- and nanofiltrationof aqueous extracts from distilled fermented grape pomace, J. Food Eng. 91(2009) 587–593.

[17] B. Díaz-Reinoso, N. González-López, A. Moure, H. Domínguez, J.C. Parajó,Recovery of antioxidants from industrial waste liquors using membranes andpolymeric resins, J. Food Eng. 96 (2010) 127–133.

[18] Y. Yazaki, Improved ultrafiltration of extracts from Pinus radiata bark,Holzforschung 39 (1985).

[19] E.M. Tsui, M. Cheryan, Membrane processing of xanthophylls in ethanolextracts of corn, J. Food Eng. 83 (2007) 590–595.

[20] B.C.B.S. Mello, J.C.C. Petrus, M.D. Hubinger, Concentration of flavonoids andphenolic compounds in aqueous and ethanolic propolis extracts throughnanofiltration, J. Food Eng. 96 (2010) 533–539.

[21] J. Li, H. Chase, Applications of membrane techniques for purification of naturalproducts, Biotechnol. Lett. 32 (2010) 601–608.

[22] E.M. Tsui, M. Cheryan, Characteristics of nanofiltration membranes in aqueousethanol, J. Membr. Sci. 237 (2004) 61–69.

[23] L.M. Pastrana-Martínez, S. Morales-Torres, S.K. Papageorgiou, F.K. Katsaros,G.E. Romanos, J.L. Figueiredo, J.L. Faria, P. Falaras, A.M.T. Silva, Photocatalyticbehaviour of nanocarbon–TiO2 composites and immobilization into hollow

fibres, Appl. Catal., B 142–143 (2013) 101–111.[24] S. Feng, S. Cheng, Z. Yuan, M. Leitch, C. Xu, Valorization of bark for chemicals

and materials: a review, Renew. Sust. Energy Rev. 26 (2013) 560–578.[25] P.M. Tahir, O.C. Musgrave, Z. Ashaari, Determination of polyphenolic content of

bark extracts for wood adhesives, Holzforschung 56 (2002) 267–272.[26] L.J. Porter, L.N. Hrstich, B.G. Chan, The conversion of procyanidins and

prodelphinidins to cyanidin and delphinidin, Phytochemistry 25 (1986) 223–230.

[27] A.E. Hagerman, L.G. Butler, Choosing appropriate methods and standards forassaying tannin, J. Chem. Ecol. 15 (1988) 1795–1810.

[28] Z. Lei, J. Jervis, R.F. Helm, Use of methanolysis for the determination of totalellagic and gallic acid contents of wood and food products, J. Agric. Food Chem.49 (2001) 1165–1168.

[29] V. Freger, J. Gilron, S. Belfer, TFC polyamide membranes modified by grafting of hydrophilic polymers: an FT-IR/AFM/TEM study, J. Membr. Sci. 209 (2002)283–292.

[30] D. Bhanushali, S. Kloos, C. Kurth, D. Bhattacharyya, Performance of solvent-resistant membranes for non-aqueous systems: solvent permeation resultsand modeling, J. Membr. Sci. 189 (2001) 1–21.

[31] I. Khattab, F. Bandarkar, M. Fakhree, A. Jouyban, Density, viscosity, and surfacetension of water + ethanol mixtures from 293 to 323 K, Korean J. Chem. Eng.29 (2012) 812–817.

[32] R. Shukla, M. Cheryan, Performance of ultrafiltration membranes in ethanol–water solutions: effect of membrane conditioning, J. Membr. Sci. 198 (2002)75–85.

[33] D.R. Machado, D. Hasson, R. Semiat, Effect of solvent properties on permeateflow through nanofiltration membranes. Part I: Investigation of parametersaffecting solvent flux, J. Membr. Sci. 163 (1999) 93–102.

[34] G. Akerlof, Dielectric constants of some organic solvent–water mixtures atvarious temperatures, J. Am. Chem. Soc. 54 (1932) 4125–4139.

[35] C. Wohlfarth, Dielectric constant of the mixture (1) water; (2) ethanol, in: M.D.Lechner (Ed.), Supplement to IV/6, Springer, Berlin Heidelberg, 2008, pp. 520–523.

[36] J. Geens, K. Peeters, B. Van der Bruggen, C. Vandecasteele, Polymericnanofiltration of binary water–alcohol mixtures: influence of feedcomposition and membrane properties on permeability and rejection, J.Membr. Sci. 255 (2005) 255–264.

[37] M. Mänttäri, A. Pihlajamäki, E. Kaipainen, M. Nyström, Effect of temperatureand membrane pre-treatment by pressure on the filtration properties of nanofiltration membranes, Desalination 145 (2002) 81–86.

[38] J .A. Restolho, A. Prates, M.N. de Pinho, M.D. Afonso, Sugars andlignosulphonates recovery from eucalyptus spent sulphite liquor bymembrane processes, Biomass Bioenergy 33 (2009) 1558–1566.

[39] S.A. Lisboa, D.V. Evtuguin, C. Pascoal Neto, Characterization of non-cellulosicglucans in Eucalyptus globulus Labill. wood and kraft pulp, Holzforschung 61(2007) 478–482.

[40] S. Quideau, K.S. Feldman, Ellagitannin chemistry, Chem. Rev. (Washington, DC,U.S.) 96 (1996) 475–503.

[41] A.A. Watrelot, C. Le Bourvellec, A. Imberty, C.M. Renard, Neutral sugar sidechains of pectins limit interactions with procyanidins, Carbohydr. Polym. 99(2014) 527–536.

[42] A.A. Shatalov, D.V. Evtuguin, C. Pascoal Neto, (2-O-a-D-Galactopyranosyl-4-O-methyl-a-D-glucurono)-D-xylan from Eucalyptus globulus Labill, Carbohydr.Res. 320 (1999) 93–99.

[43] M. Le Normand, U. Edlund, B. Holmbom, M. Ek, Hot-water extraction andcharacterization of spruce bark non-cellulosic polysaccharides, Nord. Pulp Pap.

Res. J. 27 (1) (2012) 18–23.[44] D. Mohnen, Pectin structure and biosynthesis, Curr. Opin. Plant Biol. 11 (2008)

266–277.[45] E. Sjöström, U. Westermark, Analytical methods in wood chemistry, pulping,

and papermaking, in: Chemical Composition of Wood and Pulps: BasicConstituents and Their Distribution, Springer Series in Wood Science, 1999,pp 1–19.

[46] P.C. Pinto, D.V. Evtuguin, C. Pascoal Neto, Structure of hardwoodglucuronoxylans: modifications and impact on pulp retention during woodkraft pulping, Carbohydr. Polym. 60 (2005) 489–497.

[47] S. Lee, C.H. Lee, Microfiltration and ultrafiltration as a pretreatment fornanofiltration of surface water, Sep. Sci. Technol. 41 (2006) 1–23.

http://refhub.elsevier.com/S1383-5866(14)00253-6/h0005http://refhub.elsevier.com/S1383-5866(14)00253-6/h0005http://refhub.elsevier.com/S1383-5866(14)00253-6/h0005http://refhub.elsevier.com/S1383-5866(14)00253-6/h0005http://refhub.elsevier.com/S1383-5866(14)00253-6/h0005http://refhub.elsevier.com/S1383-5866(14)00253-6/h0005http://refhub.elsevier.com/S1383-5866(14)00253-6/h0010http://refhub.elsevier.com/S1383-5866(14)00253-6/h0010http://refhub.elsevier.com/S1383-5866(14)00253-6/h0010http://refhub.elsevier.com/S1383-5866(14)00253-6/h0010http://refhub.elsevier.com/S1383-5866(14)00253-6/h0010http://refhub.elsevier.com/S1383-5866(14)00253-6/h0010http://refhub.elsevier.com/S1383-5866(14)00253-6/h0010http://refhub.elsevier.com/S1383-5866(14)00253-6/h0015http://refhub.elsevier.com/S1383-5866(14)00253-6/h0015http://refhub.elsevier.com/S1383-5866(14)00253-6/h0015http://refhub.elsevier.com/S1383-5866(14)00253-6/h0015http://refhub.elsevier.com/S1383-5866(14)00253-6/h0015http://refhub.elsevier.com/S1383-5866(14)00253-6/h0015http://refhub.elsevier.com/S1383-5866(14)00253-6/h0020http://refhub.elsevier.com/S1383-5866(14)00253-6/h0020http://refhub.elsevier.com/S1383-5866(14)00253-6/h0020http://refhub.elsevier.com/S1383-5866(14)00253-6/h0020http://refhub.elsevier.com/S1383-5866(14)00253-6/h0020http://refhub.elsevier.com/S1383-5866(14)00253-6/h0025http://refhub.elsevier.com/S1383-5866(14)00253-6/h0025http://refhub.elsevier.com/S1383-5866(14)00253-6/h0025http://refhub.elsevier.com/S1383-5866(14)00253-6/h0025http://refhub.elsevier.com/S1383-5866(14)00253-6/h0025http://refhub.elsevier.com/S1383-5866(14)00253-6/h0245http://refhub.elsevier.com/S1383-5866(14)00253-6/h0245http://refhub.elsevier.com/S1383-5866(14)00253-6/h0245http://refhub.elsevier.com/S1383-5866(14)00253-6/h0245http://refhub.elsevier.com/S1383-5866(14)00253-6/h0245http://refhub.elsevier.com/S1383-5866(14)00253-6/h0245http://refhub.elsevier.com/S1383-5866(14)00253-6/h0245http://refhub.elsevier.com/S1383-5866(14)00253-6/h0245http://refhub.elsevier.com/S1383-5866(14)00253-6/h0245http://refhub.elsevier.com/S1383-5866(14)00253-6/h0245http://refhub.elsevier.com/S1383-5866(14)00253-6/h0035http://refhub.elsevier.com/S1383-5866(14)00253-6/h0035http://refhub.elsevier.com/S1383-5866(14)00253-6/h0040http://refhub.elsevier.com/S1383-5866(14)00253-6/h0040http://refhub.elsevier.com/S1383-5866(14)00253-6/h0040http://refhub.elsevier.com/S1383-5866(14)00253-6/h0040http://refhub.elsevier.com/S1383-5866(14)00253-6/h0045http://refhub.elsevier.com/S1383-5866(14)00253-6/h0045http://refhub.elsevier.com/S1383-5866(14)00253-6/h0050http://refhub.elsevier.com/S1383-5866(14)00253-6/h0050http://refhub.elsevier.com/S1383-5866(14)00253-6/h0050http://refhub.elsevier.com/S1383-5866(14)00253-6/h0050http://refhub.elsevier.com/S1383-5866(14)00253-6/h0055http://refhub.elsevier.com/S1383-5866(14)00253-6/h0055http://refhub.elsevier.com/S1383-5866(14)00253-6/h0055http://refhub.elsevier.com/S1383-5866(14)00253-6/h0060http://refhub.elsevier.com/S1383-5866(14)00253-6/h0060http://refhub.elsevier.com/S1383-5866(14)00253-6/h0060http://refhub.elsevier.com/S1383-5866(14)00253-6/h0065http://refhub.elsevier.com/S1383-5866(14)00253-6/h0065http://refhub.elsevier.com/S1383-5866(14)00253-6/h0065http://refhub.elsevier.com/S1383-5866(14)00253-6/h0065http://refhub.elsevier.com/S1383-5866(14)00253-6/h0070http://refhub.elsevier.com/S1383-5866(14)00253-6/h0070http://refhub.elsevier.com/S1383-5866(14)00253-6/h0070http://refhub.elsevier.com/S1383-5866(14)00253-6/h0075http://refhub.elsevier.com/S1383-5866(14)00253-6/h0075http://refhub.elsevier.com/S1383-5866(14)00253-6/h0075http://refhub.elsevier.com/S1383-5866(14)00253-6/h0080http://refhub.elsevier.com/S1383-5866(14)00253-6/h0080http://refhub.elsevier.com/S1383-5866(14)00253-6/h0080http://refhub.elsevier.com/S1383-5866(14)00253-6/h0085http://refhub.elsevier.com/S1383-5866(14)00253-6/h0085http://refhub.elsevier.com/S1383-5866(14)00253-6/h0085http://refhub.elsevier.com/S1383-5866(14)00253-6/h0085http://refhub.elsevier.com/S1383-5866(14)00253-6/h0090http://refhub.elsevier.com/S1383-5866(14)00253-6/h0090http://refhub.elsevier.com/S1383-5866(14)00253-6/h0095http://refhub.elsevier.com/S1383-5866(14)00253-6/h0095http://refhub.elsevier.com/S1383-5866(14)00253-6/h0095http://refhub.elsevier.com/S1383-5866(14)00253-6/h0100http://refhub.elsevier.com/S1383-5866(14)00253-6/h0100http://refhub.elsevier.com/S1383-5866(14)00253-6/h0100http://refhub.elsevier.com/S1383-5866(14)00253-6/h0105http://refhub.elsevier.com/S1383-5866(14)00253-6/h0105http://refhub.elsevier.com/S1383-5866(14)00253-6/h0110http://refhub.elsevier.com/S1383-5866(14)00253-6/h0110http://refhub.elsevier.com/S1383-5866(14)00253-6/h0110http://refhub.elsevier.com/S1383-5866(14)00253-6/h0115http://refhub.elsevier.com/S1383-5866(14)00253-6/h0115http://refhub.elsevier.com/S1383-5866(14)00253-6/h0115http://refhub.elsevier.com/S1383-5866(14)00253-6/h0115http://refhub.elsevier.com/S1383-5866(14)00253-6/h0115http://refhub.elsevier.com/S1383-5866(14)00253-6/h0115http://refhub.elsevier.com/S1383-5866(14)00253-6/h0120http://refhub.elsevier.com/S1383-5866(14)00253-6/h0120http://refhub.elsevier.com/S1383-5866(14)00253-6/h0125http://refhub.elsevier.com/S1383-5866(14)00253-6/h0125http://refhub.elsevier.com/S1383-5866(14)00253-6/h0125http://refhub.elsevier.com/S1383-5866(14)00253-6/h0130http://refhub.elsevier.com/S1383-5866(14)00253-6/h0130http://refhub.elsevier.com/S1383-5866(14)00253-6/h0130http://refhub.elsevier.com/S1383-5866(14)00253-6/h0135http://refhub.elsevier.com/S1383-5866(14)00253-6/h0135http://refhub.elsevier.com/S1383-5866(14)00253-6/h0140http://refhub.elsevier.com/S1383-5866(14)00253-6/h0140http://refhub.elsevier.com/S1383-5866(14)00253-6/h0140http://refhub.elsevier.com/S1383-5866(14)00253-6/h0145http://refhub.elsevier.com/S1383-5866(14)00253-6/h0145http://refhub.elsevier.com/S1383-5866(14)00253-6/h0145http://refhub.elsevier.com/S1383-5866(14)00253-6/h0150http://refhub.elsevier.com/S1383-5866(14)00253-6/h0150http://refhub.elsevier.com/S1383-5866(14)00253-6/h0150http://refhub.elsevier.com/S1383-5866(14)00253-6/h0150http://refhub.elsevier.com/S1383-5866(14)00253-6/h0155http://refhub.elsevier.com/S1383-5866(14)00253-6/h0155http://refhub.elsevier.com/S1383-5866(14)00253-6/h0155http://refhub.elsevier.com/S1383-5866(14)00253-6/h0160http://refhub.elsevier.com/S1383-5866(14)00253-6/h0160http://refhub.elsevier.com/S1383-5866(14)00253-6/h0160http://refhub.elsevier.com/S1383-5866(14)00253-6/h0160http://refhub.elsevier.com/S1383-5866(14)00253-6/h0165http://refhub.elsevier.com/S1383-5866(14)00253-6/h0165http://refhub.elsevier.com/S1383-5866(14)00253-6/h0165http://refhub.elsevier.com/S1383-5866(14)00253-6/h0165http://refhub.elsevier.com/S1383-5866(14)00253-6/h0170http://refhub.elsevier.com/S1383-5866(14)00253-6/h0170http://refhub.elsevier.com/S1383-5866(14)00253-6/h0175http://refhub.elsevier.com/S1383-5866(14)00253-6/h0175http://refhub.elsevier.com/S1383-5866(14)00253-6/h0175http://refhub.elsevier.com/S1383-5866(14)00253-6/h0180http://refhub.elsevier.com/S1383-5866(14)00253-6/h0180http://refhub.elsevier.com/S1383-5866(14)00253-6/h0180http://refhub.elsevier.com/S1383-5866(14)00253-6/h0180http://refhub.elsevier.com/S1383-5866(14)00253-6/h0185http://refhub.elsevier.com/S1383-5866(14)00253-6/h0185http://refhub.elsevier.com/S1383-5866(14)00253-6/h0185http://refhub.elsevier.com/S1383-5866(14)00253-6/h0190http://refhub.elsevier.com/S1383-5866(14)00253-6/h0190http://refhub.elsevier.com/S1383-5866(14)00253-6/h0190http://refhub.elsevier.com/S1383-5866(14)00253-6/h0195http://refhub.elsevier.com/S1383-5866(14)00253-6/h0195http://refhub.elsevier.com/S1383-5866(14)00253-6/h0195http://refhub.elsevier.com/S1383-5866(14)00253-6/h0195http://refhub.elsevier.com/S1383-5866(14)00253-6/h0195http://refhub.elsevier.com/S1383-5866(14)00253-6/h0250http://refhub.elsevier.com/S1383-5866(14)00253-6/h0250http://refhub.elsevier.com/S1383-5866(14)00253-6/h0205http://refhub.elsevier.com/S1383-5866(14)00253-6/h0205http://refhub.elsevier.com/S1383-5866(14)00253-6/h0205http://refhub.elsevier.com/S1383-5866(14)00253-6/h0215http://refhub.elsevier.com/S1383-5866(14)00253-6/h0215http://refhub.elsevier.com/S1383-5866(14)00253-6/h0215http://refhub.elsevier.com/S1383-5866(14)00253-6/h0215http://refhub.elsevier.com/S1383-5866(14)00253-6/h0215http://refhub.elsevier.com/S1383-5866(14)00253-6/h0215http://refhub.elsevier.com/S1383-5866(14)00253-6/h0215http://refhub.elsevier.com/S1383-5866(14)00253-6/h0215http://refhub.elsevier.com/S1383-5866(14)00253-6/h0215http://refhub.elsevier.com/S1383-5866(14)00253-6/h0215http://refhub.elsevier.com/S1383-5866(14)00253-6/h0220http://refhub.elsevier.com/S1383-5866(14)00253-6/h0220http://refhub.elsevier.com/S1383-5866(14)00253-6/h0220http://refhub.elsevier.com/S1383-5866(14)00253-6/h0225http://refhub.elsevier.com/S1383-5866(14)00253-6/h0225http://refhub.elsevier.com/S1383-5866(14)00253-6/h0235http://refhub.elsevier.com/S1383-5866(14)00253-6/h0235http://refhub.elsevier.com/S1383-5866(14)00253-6/h0235http://refhub.elsevier.com/S1383-5866(14)00253-6/h0240http://refhub.elsevier.com/S1383-5866(14)00253-6/h0240http://refhub.elsevier.com/S1383-5866(14)00253-6/h0240http://refhub.elsevier.com/S1383-5866(14)00253-6/h0240http://refhub.elsevier.com/S1383-5866(14)00253-6/h0240http://refhub.elsevier.com/S1383-5866(14)00253-6/h0235http://refhub.elsevier.com/S1383-5866(14)00253-6/h0235http://refhub.elsevier.com/S1383-5866(14)00253-6/h0235http://refhub.elsevier.com/S1383-5866(14)00253-6/h0225http://refhub.elsevier.com/S1383-5866(14)00253-6/h0225http://refhub.elsevier.com/S1383-5866(14)00253-6/h0220http://refhub.elsevier.com/S1383-5866(14)00253-6/h0220http://refhub.elsevier.com/S1383-5866(14)00253-6/h0220http://refhub.elsevier.com/S1383-5866(14)00253-6/h0215http://refhub.elsevier.com/S1383-5866(14)00253-6/h0215http://refhub.elsevier.com/S1383-5866(14)00253-6/h0215http://refhub.elsevier.com/S1383-5866(14)00253-6/h0205http://refhub.elsevier.com/S1383-5866(14)00253-6/h0205http://refhub.elsevier.com/S1383-5866(14)00253-6/h0205http://refhub.elsevier.com/S1383-5866(14)00253-6/h0250http://refhub.elsevier.com/S1383-5866(14)00253-6/h0250http://refhub.elsevier.com/S1383-5866(14)00253-6/h0195http://refhub.elsevier.com/S1383-5866(14)00253-6/h0195http://refhub.elsevier.com/S1383-5866(14)00253-6/h0195http://refhub.elsevier.com/S1383-5866(14)00253-6/h0190http://refhub.elsevier.com/S1383-5866(14)00253-6/h0190http://refhub.elsevier.com/S1383-5866(14)00253-6/h0190http://refhub.elsevier.com/S1383-5866(14)00253-6/h0185http://refhub.elsevier.com/S1383-5866(14)00253-6/h0185http://refhub.elsevier.com/S1383-5866(14)00253-6/h0185http://refhub.elsevier.com/S1383-5866(14)00253-6/h0180http://refhub.elsevier.com/S1383-5866(14)00253-6/h0180http://refhub.elsevier.com/S1383-5866(14)00253-6/h0180http://refhub.elsevier.com/S1383-5866(14)00253-6/h0180http://refhub.elsevier.com/S1383-5866(14)00253-6/h0175http://refhub.elsevier.com/S1383-5866(14)00253-6/h0175http://refhub.elsevier.com/S1383-5866(14)00253-6/h0175http://refhub.elsevier.com/S1383-5866(14)00253-6/h0175http://refhub.elsevier.com/S1383-5866(14)00253-6/h0175http://refhub.elsevier.com/S1383-5866(14)00253-6/h0170http://refhub.elsevier.com/S1383-5866(14)00253-6/h0170http://refhub.elsevier.com/S1383-5866(14)00253-6/h0165http://refhub.elsevier.com/S1383-5866(14)00253-6/h0165http://refhub.elsevier.com/S1383-5866(14)00253-6/h0165http://refhub.elsevier.com/S1383-5866(14)00253-6/h0160http://refhub.elsevier.com/S1383-5866(14)00253-6/h0160http://refhub.elsevier.com/S1383-5866(14)00253-6/h0160http://refhub.elsevier.com/S1383-5866(14)00253-6/h0155http://refhub.elsevier.com/S1383-5866(14)00253-6/h0155http://refhub.elsevier.com/S1383-5866(14)00253-6/h0155http://refhub.elsevier.com/S1383-5866(14)00253-6/h0155http://refhub.elsevier.com/S1383-5866(14)00253-6/h0155http://refhub.elsevier.com/S1383-5866(14)00253-6/h0155http://refhub.elsevier.com/S1383-5866(14)00253-6/h0150http://refhub.elsevier.com/S1383-5866(14)00253-6/h0150http://refhub.elsevier.com/S1383-5866(14)00253-6/h0150http://refhub.elsevier.com/S1383-5866(14)00253-6/h0145http://refhub.elsevier.com/S1383-5866(14)00253-6/h0145http://refhub.elsevier.com/S1383-5866(14)00253-6/h0145http://refhub.elsevier.com/S1383-5866(14)00253-6/h0140http://refhub.elsevier.com/S1383-5866(14)00253-6/h0140http://refhub.elsevier.com/S1383-5866(14)00253-6/h0140http://refhub.elsevier.com/S1383-5866(14)00253-6/h0135http://refhub.elsevier.com/S1383-5866(14)00253-6/h0135http://refhub.elsevier.com/S1383-5866(14)00253-6/h0130http://refhub.elsevier.com/S1383-5866(14)00253-6/h0130http://refhub.elsevier.com/S1383-5866(14)00253-6/h0130http://refhub.elsevier.com/S1383-5866(14)00253-6/h0125http://refhub.elsevier.com/S1383-5866(14)00253-6/h0125http://refhub.elsevier.com/S1383-5866(14)00253-6/h0120http://refhub.elsevier.com/S1383-5866(14)00253-6/h0120http://refhub.elsevier.com/S1383-5866(14)00253-6/h0115http://refhub.elsevier.com/S1383-5866(14)00253-6/h0115http://refhub.elsevier.com/S1383-5866(14)00253-6/h0115http://refhub.elsevier.com/S1383-5866(14)00253-6/h0115http://refhub.elsevier.com/S1383-5866(14)00253-6/h0115http://refhub.elsevier.com/S1383-5866(14)00253-6/h0110http://refhub.elsevier.com/S1383-5866(14)00253-6/h0110http://refhub.elsevier.com/S1383-5866(14)00253-6/h0105http://refhub.elsevier.com/S1383-5866(14)00253-6/h0105http://refhub.elsevier.com/S1383-5866(14)00253-6/h0100http://refhub.elsevier.com/S1383-5866(14)00253-6/h0100http://refhub.elsevier.com/S1383-5866(14)00253-6/h0100http://refhub.elsevier.com/S1383-5866(14)00253-6/h0095http://refhub.elsevier.com/S1383-5866(14)00253-6/h0095http://-/?-http://refhub.elsevier.com/S1383-5866(14)00253-6/h0090http://refhub.elsevier.com/S1383-5866(14)00253-6/h0090http://-/?-http://refhub.elsevier.com/S1383-5866(14)00253-6/h0085http://refhub.elsevier.com/S1383-5866(14)00253-6/h0085http://refhub.elsevier.com/S1383-5866(14)00253-6/h0085http://-/?-http://refhub.elsevier.com/S1383-5866(14)00253-6/h0080http://refhub.elsevier.com/S1383-5866(14)00253-6/h0080http://refhub.elsevier.com/S1383-5866(14)00253-6/h0080http://-/?-http://refhub.elsevier.com/S1383-5866(14)00253-6/h0075http://refhub.elsevier.com/S1383-5866(14)00253-6/h0075http://refhub.elsevier.com/S1383-5866(14)00253-6/h0075http://-/?-http://refhub.elsevier.com/S1383-5866(14)00253-6/h0070http://refhub.elsevier.com/S1383-5866(14)00253-6/h0070http://refhub.elsevier.com/S1383-5866(14)00253-6/h0070http://-/?-http://refhub.elsevier.com/S1383-5866(14)00253-6/h0065http://refhub.elsevier.com/S1383-5866(14)00253-6/h0065http://refhub.elsevier.com/S1383-5866(14)00253-6/h0065http://-/?-http://refhub.elsevier.com/S1383-5866(14)00253-6/h0060http://refhub.elsevier.com/S1383-5866(14)00253-6/h0060http://refhub.elsevier.com/S1383-5866(14)00253-6/h0060http://-/?-http://refhub.elsevier.com/S1383-5866(14)00253-6/h0055http://refhub.elsevier.com/S1383-5866(14)00253-6/h0055http://refhub.elsevier.com/S1383-5866(14)00253-6/h0055http://-/?-http://refhub.elsevier.com/S1383-5866(14)00253-6/h0050http://refhub.elsevier.com/S1383-5866(14)00253-6/h0050http://refhub.elsevier.com/S1383-5866(14)00253-6/h0050http://refhub.elsevier.com/S1383-5866(14)00253-6/h0050http://-/?-http://refhub.elsevier.com/S1383-5866(14)00253-6/h0045http://refhub.elsevier.com/S1383-5866(14)00253-6/h0045http://refhub.elsevier.com/S1383-5866(14)00253-6/h0045http://refhub.elsevier.com/S1383-5866(14)00253-6/h0040http://refhub.elsevier.com/S1383-5866(14)00253-6/h0040http://refhub.elsevier.com/S1383-5866(14)00253-6/h0040http://refhub.elsevier.com/S1383-5866(14)00253-6/h0040http://refhub.elsevier.com/S1383-5866(14)00253-6/h0035http://refhub.elsevier.com/S1383-5866(14)00253-6/h0035http://refhub.elsevier.com/S1383-5866(14)00253-6/h0245http://refhub.elsevier.com/S1383-5866(14)00253-6/h0245http://refhub.elsevier.com/S1383-5866(14)00253-6/h0245http://refhub.elsevier.com/S1383-5866(14)00253-6/h0025http://refhub.elsevier.com/S1383-5866(14)00253-6/h0025http://refhub.elsevier.com/S1383-5866(14)00253-6/h0025http://refhub.elsevier.com/S1383-5866(14)00253-6/h0020http://refhub.elsevier.com/S1383-5866(14)00253-6/h0020http://refhub.elsevier.com/S1383-5866(14)00253-6/h0020http://refhub.elsevier.com/S1383-5866(14)00253-6/h0015http://refhub.elsevier.com/S1383-5866(14)00253-6/h0015http://refhub.elsevier.com/S1383-5866(14)00253-6/h0015http://refhub.elsevier.com/S1383-5866(14)00253-6/h0015http://refhub.elsevier.com/S1383-5866(14)00253-6/h0010http://refhub.elsevier.com/S1383-5866(14)00253-6/h0010http://refhub.elsevier.com/S1383-5866(14)00253-6/h0010http://refhub.elsevier.com/S1383-5866(14)00253-6/h0005http://refhub.elsevier.com/S1383-5866(14)00253-6/h0005http://refhub.elsevier.com/S1383-5866(14)00253-6/h0005http://refhub.elsevier.com/S1383-5866(14)00253-6/h0005http://-/?-http://-/?-

Documents

Membrane performance and application of ultrafiltration.pdf