0960–3085/02/$10.00+0.

00
# Institution of Chemical Engineers
Trans IChemE, Vol 80, Part C, June 2002

TREATMENT OF UF MEMBRANES WITH SIMPLE

AND FORMULATED CLEANING AGENTS
P. VÄISÄNEN1 , M. R. BIRD2 and M. NYSTRÖM1
1
Laboratory of Membrane Technology and Technical Polymer Chemistry, Lappeenranta University of Technology, Finland
2
Department of Chemical Engineering, University of Bath, UK

B
oth hydrophilic (C 30F) and hydrophobic (PA 50H and PES 50H) ultraŽ ltration
membranes were fouled with (i) a 3.5 wt% whey protein solution or (ii) ground wood
mill circulation water from an integrated pulp and paper mill. The membranes were
subsequently treated with different cleaning agents (NaOH, HNO3, Ultrasil 11 and Libranone
960). For whey protein fouled membranes, the effectiveness of the cleaning protocol was a
strong function of the sodium hydroxide concentration used. After treatment with Libranone
960, membranes fouled with ground wood mill water displayed a substantial increase in water
permeability. The less hydrophilic the membrane surface, the greater the observed  ux increase
following cleaning. After a short period of cleaning with Libranone 960, the pure water
permeability decreased as the surfactant desorbed from the membrane surface. No such trend
was seen when cleaning with Ultrasil 11. FTIR, SEM and AFM techniques were used to
investigate the nature of the membrane surface before and after fouling and cleaning. These
techniques conŽ rmed the efŽ cacy of the Libranone 960 cleaning protocol.

Keywords: fouling; cleaning; ultraŽltration; lignosulphates.

INTRODUCTION of non-ionic surfactants in the range C9–11E6, and was

Industrial membrane Ž ltration processes are widely used as
wastewater and circulation water treatment methods. The
effectiveness of membrane Ž ltration depends on the
membrane type; for example, ultraŽ ltration can be used to
remove dissolved or suspended macromolecules; nanoŽ ltra-
tion can be used to remove salts. Under practical operating
conditions, any Ž ltration process invariably leads to fouling,
and this necessitates frequent cleaning. This results in addi-
tional operating costs due to cleaning chemicals, energy and
labour, and can also generate large quantities of often highly
alkaline ef uent1.
UltraŽ ltration membranes are used widely in the pulp and
paper industry for the separation of lignosulphonates from
salts, and for ef uent treatment2–6. The pulp and paper
industry is moving towards partially or totally closed water
circuits6. Membrane Ž ltration can help achieve this aim, by
providing effective water puriŽ cation. Closed water circuits
result in a reduction of water consumed, and chemical and
Ž bre losses are also reduced6.
UltraŽ ltration membranes are also widely used in the
processing of whey, for protein recovery and as a partial
sterilization technique7–9.
There are a wide range of commercially available
compounds for membrane cleaning, from acidic to alkaline
solutions with or without enzymes and surfactants. Cleaning
agents can be formulated according to the fouling
substances present, the process conditions, and the nature
of the membrane material. In this study, four cleaning agents
were selected: (i) Libranone 960 (Shell Company) consists
chosen because non-ionic cleaning agents have a good
record in removing foulants such as fatty acids, sterols
and lignans which are likely to be present; (ii) Ultrasil 11
(Henkel Ecolab) and (iii) sodium hydroxide were selected as
examples of formulated and simple alkali cleaning agents.
Sodium hydroxide is generally effective at removing whey
protein based deposition10. Nitric acid (iv) is also commonly
used both in industry and the laboratory to remove mineral
based deposition1.
In addition to the cleaning agent formulation, cleaning
efŽ ciency depends upon the detergent concentration,
temperature, cross- ow velocity, and trans-membrane
pressure11–18.
The effect of the adsorption or removal of fouling
substances to or from the membrane can be measured in
many ways. The most common parameter measured has been
permeate  ux. However, the permeate  ux does not uniquely
determine the state of a surface, as different deposits result
in different membrane permeation characteristics. Several
surface characterization methods have recently been
exploited. Contaminants can be investigated using infrared
spectra, provided a virgin membrane is also scanned for
comparison purposes18–21.
Microscopic techniques are mainly used to provide topo-
graphical images of unused surfaces, but can also be used to
provide information on contaminated surfaces22–27.
The two aims of this study were, Ž rstly, to test the effect
of the cleaning agent composition (including non-ionic
surfactants, which do not Ž nd widespread use as membrane
cleaners) upon UF membranes fouled with industrially

98
 TREATMENT OF UF MEMBRANES 99

important liquids. These foulants were (i) whey protein Ž xed solids. The total organic carbon (TOC) was analysed
concentrates, and (ii) ground wood mill circulation water with a Shimadzu TOC-5050A analyser, the temperature was
from an integrated pulp and paper mill. The second aim was 680¯ C, and the carrier gas was high purity air with a  ow
to investigate the linkage between  ux recovery measure- rate of 150 ml=minute.
ments and surface characterization techniques in determin-
ing the cleaning performance of UF membranes. Whey protein concentrate powder analysis
The reason for selecting wood pulp liquor and whey
An analysis of the whey protein concentrate powder is
protein concentrate was that they are two of the most shown in Table 2. The powder consisted of 35 wt% protein,
industrially relevant liquids with which to evaluate UF
and b-lactoglobulin was the main component, comprising
membrane performance. Three important membranes
50 wt% of the total.
(regenerated cellulose, polyethersulphoneand aromatic poly-
amide) and a suite of analytical techniques have been used to
link surface characteristics to permeate  ux recovery. UF Membranes and Membrane Analyses
Virgin membranes were characterized using contact angle
and streaming potential measurements. Fouled and cleaned The characteristics of the  at-sheet membranes used in
membranes were examined using Fourier Transform Infra- this study are shown in Table 3. The ultraŽ ltration
red (FTIR) spectroscopy with the Attenuated Total Re ec- membranes selected are made from a range of materials.
tance (ATR) technique, Scanning Electron Microscopy A narrow range of molecular weight cut-offs was selected.
(SEM) and Atomic Force Microscopy (AFM). The permeate  ux was measured during experiments and
IR-spectra were obtained for both clean, unused membranes,
and membranes which had been subjected to fouling and
MATERIALS AND METHODS cleaning cycles. The surface properties of the virgin
membranes were analysed by measuring the zeta potentials
Water and Whey Protein Concentrate Powder
and contact angles. The membrane surfaces were analysed
Analysis
with a JEOL JSM T330 SEM, a NanoScope IIIa AFM and a
Water analysis
An analysis of the ground wood mill water is presented in
Table 2. Analysis of whey protein concentrate (WPC) powder.
Table 1, and is the same procedure as presented in Huuhilo
et al.6. The TOC, conductivity, pH, turbidity, absorbance, Component Weight %
total carbon, total solids and Ž xed solids were determined.
The turbidity was measured using a Hach DR=2000 direct Protein 35
Lactose 50
reading spectrophotometer at 450 nm. The lignous material Fat 7
was measured using a Varian Cary 1C UV-visible spectro- Ash 8
photometer. Water was driven off the samples at 100¯ C to Total 100
obtain the total solids, and heated to 550¯ C to obtain the The protein composition was as follows:
Protein
b-lactoglobulin 50
Table 1. Analysis of ground wood mill circulation water. a-lactalbumin 15
Bovine serum albumin 5
pH 4.9§ 0.1 Immuno-globulin 8
Conductivity, mS cm¡ 1 1.0§ 0.3 Proteose peptones 22
TOC, mg l¡1 500§ 160 Total 100
Turbidity, FTU 360§ 280
Absorbance, 280 nm 6.0§ 1.0 Whey protein concentrate (WPC) is a spray dried powder. The
Total solids, g l¡1 2.0§ 0.3 WPC used in this study was supplied by Carberry Limited,
Fixed solids, g l¡ 1 0.7§ 0.3 Michelstown, Co. Cork, Eire. The powder contained 35 wt%
protein.

Table 3. Characteristics of the membranes tested.

C 30F PES 50H PA 50H

Manufacturer Celgard= Hoechst Celgard= Hoechst Celgard= Hoechst

Material
active=support Regenerated Polyethersulfone=PO Aromatic
cellulose=PET polyamide=PO
Cut-off, kDa 30 50 50
Thickness, mm
backing= 190 § 30= 220 § 30= 220 § 30=
whole membrane 200–280 270–330 250–310
pH resistance, 20¯ C 1–13 1–14 1–12
T, max, ¯ C 70 90 90
Flux, l m2 h¡1, 300–600 250–500 600–2500
Measured at 3 bar, and 20¯ C
Retention, % PVP K30 80–88 75–83 70–85
Chemical resistance Good Good Excellent
PO ˆ PolyoleŽ n; PET ˆ Polyethyleneterefthalate; PVP K30 ˆ Polyvinylpyrrolidone.

Trans IChemE, Vol 80, Part C, June 2002

100 VÄISÄNEN et al.
Perkin-Elmer System 2000 FTIR spectrometer with an ATR
accessory.
Simultaneous  ux and streaming potential measurements
were carried out using a  at-sheet cell with a Ž ltration area
of 0.00216 m2. Ag=AgCl electrodes were placed on both
sides of the membrane in order to measure the streaming
potential through the pores of the membrane. The measure-
ments were made with 0.001 M KCl solution, the tempera-
ture was 25¯ C and the cross- ow velocity 0.5 m s¡1 in all
streaming potential measurements26. The zeta potentials for
the membranes were calculated from the measured stream-
ing potentials (Helmholtz-Smoluchowski equation28), and
the contact angle measurements were carried out using the
sessile-drop method. Before streaming potential and contact
angle measurements were taken, the virgin membranes were
cleaned in an ultrasonic bath for Ž ve minutes. The standard
deviations in contact angle and streaming potential measure-
ments were one degree and 0.01 mV respectively.
For the SEM an accelerating voltage of 15 kV was
utilized, and the distance between the sample and the
electron beam was 25 mm. The membranes were coated
with gold before visualization (the Ž lm thickness was
approximately 30 nm) as the samples were not conducting.
The AFM was used in tapping mode in the air, and the drive
frequency was 297–392 kHz. The sensor type was NCH,
and the cantilever length was 123 mm. For the FTIR studies,
KRS-5 (thallium bromo-iodide) was used as an ATR crystal,
the angle of incidence was 45¯ and the resolution 2 cm¡ 1.
All the samples were scanned 50 times 19.
Cleaning Agents
The following cleaning solutions were used (i) 0.1 wt%
Libranone 960 (Shell Company), (ii) 0.5 wt% Ultrasil 11
(Henkel Ecolab) and (iii) 0.05 to 1.0 wt% sodium hydroxide
followed by 0.3 wt% nitric acid (both supplied by Fisher
ScientiŽc).
Libranone 960 is an alcohol ethoxylate (C9–11E6) non-
ionic surfactant with a critical micelle concentration (cmc)
of 0.018 wt%. P3 Ultrasil 11 is a caustic based cleaning
agent containing EDTA and surfactants.
Membrane Fouling Protocol
Membranes were subjected to two different fouling
protocols:
(i) A 3.5 wt% solution of reconstituted whey protein
concentrate (WPC) was made using reverse osmosis
water and Carbelac 35, a free  owing WPC powder,
supplied by Carberry Milk Products Ltd, (Republic of
Ireland). The procedure started with a pure water  ux
measurement at 1 bar, followed by a fouling with whey
protein solution at 1 bar for 30 minutes. The membrane
was subsequently rinsed for 10 minutes, and a pure
water  ux measurement was taken at 1 bar. A tempera-
ture of 40¯ C and a cross- ow velocity of 2 m s¡ 1 were
maintained for all stages of the process. The whey
protein fouling=cleaning process and simultaneous
 ux measurement experiments were carried out at the
University of Bath, UK, using a  at-sheet cell with a
Ž ltration area of 0.00837 m2. The membrane fouling=
cleaning apparatus consisted of two tanks, one for the
rinsing water and another for the whey protein=cleaning
solution. The whey protein solution (or rinsing water)
was pumped through a mass  ow control unit into the
membrane cell. Permeate  ux was measured using an
electronic load cell, and subsequently recycled into the
feed tank. The transmembrane pressure, the permeate
 ux, the cross- ow velocity and temperature data were
collected using a computer. A schematic diagram of
the fouling=cleaning equipment is shown in Figure 1.
The retentate and permeate were recycled back to the
fouling=cleaning tank.
(ii) A second set of membranes were fouled at an integrated
Finnish pulp and paper mill using the ground wood mill
circulation water6. Each of the three types of membranes
tested were treated with ground wood mill circulation
water for as close to ten minutes as possible, using a
transmembrane pressure of 1.0 bar. The ground wood
mill water temperature during the fouling procedure
varied from 46–78¯ C at the paper mill, and could not
be closely controlled in this industrial environment, as
these were the actual process stream temperatures. At
the paper mill, all the membranes were fouled simulta-
neously. The pH of the circulation water was 5,
the conductivity 0.8–1.3 ms cm¡1 and the COD app.
2000 mg l¡1. The turbidity of the water varied between
200–1300 FTU. The absorbance at 280 nm was between
5 and 8.
Membrane Cleaning Protocol
The cleaning procedure started with the pure water  ux
measurement at 1 bar followed by a chemical clean at 0.5
bar for 30 minutes. This was followed by a subsequent pure
water  ux measurement taken at 1 bar. Lowering the
transmembrane pressure from 1.0 bar during fouling to 0.5
bar during cleaning gives an improved cleaning perfor-
mance, as foulants are not forced further into the pores of
the membrane12,29. A cross- ow velocity of 2 m s¡ 1 and a
water temperature of 40¯ C was used in all membrane  ux
measurements.
The percentage  ux recovery due to cleaning is deŽ ned as
100£ the ratio of the pure water  ux of a membrane
Figure 1. Schematic diagram of the fouling=cleaning equipment.
Trans IChemE, Vol 80, Part C, June 2002
 TREATMENT OF UF MEMBRANES
subjected to fouling and cleaning protocols, divided by pure
water  ux of the unfouled (pristine) membrane.
%Jr ˆ 100 £ Jfc =Ju
Where %Jr is the percentage  ux recovery, Jfc is the pure water
 ux of a membrane subjected to fouling and cleaning proto-
cols, and Ju is the pure water  ux of a pristine membrane.
Whey protein fouled membranes were cleaned initially
with sodium hydroxide at a range of concentrations between
0.05 and 1.0 wt%. The results were subsequently compared
to those obtained when using Ultrasil 11 and Libranone 960
as cleaning agents. An alkaline=acidic sequence cleaning of
0.075 wt% sodium hydroxide followed by 0.3 wt% nitric
acid was applied to test the effect of nitric acid in removing
calcium phosphate mineral deposition from the membrane.
Following effective cleaning membranes were subsequently
refouled for further experimentation.
Membranes fouled industrially with the ground wood mill
circulation water were cleaned with both Ultrasil 11 and
Libranone 960.
RESULTS AND DISCUSSION
Measured Characteristics of Clean Membranes
The charge and the hydrophobicity of the virgin
membranes were analysed before experimentation, in order
to examine the relationship between hydrophobicity and
charge to fouling propensity and cleaning characteristics.
The zeta potentials of the three clean membranes differed
considerably (see Table 4). The PES 50H and C 30F
membranes had a weak negative charge (ranging from
0.73–1.55 mV) over the pH range used in the measurements.
The PA 50H membrane was a little more negatively charged
than the PES 50H and the C 30F membranes, with zeta
potentials in the range 1.09–3.55 mV. A median value of
zeta potential for the PES 50H membrane was 2.69 mV.
Of the three membranes used in this study, the C 30F
membrane had the most hydrophilic surface, with a contact
angle of 44¯ . The PES 50H and PA 50H membranes were less
hydrophilic, with contact angles of 54¯ and 58¯ respectively.
Cleaning of Membranes
Membranes fouled with whey protein deposits
The pure water  uxes before fouling for pristine C 30F,
PA 50H and PES 50H membranes were 162, 69 and
114 l m¡2 hr¡1 respectively. The pure water  ux values
Table 4. Calculated zeta potentials for the clean membranes used in the
experiments.
zeta potential, mV
pH PES 50H PA 50H C 30F
7 ¡ 0.98 ¡ 3.55 ¡ 1.02
6.3 ¡ 0.77 ¡ 3.38 ¡ 1.00
5.6 ¡ 0.83 ¡ 2.95 ¡ 0.96
5 ¡ 0.76 ¡ 2.69 ¡ 0.91
4.4 ¡ 0.73 ¡ 2.18 ¡ 0.84
3.5 ¡ 0.93 ¡ 1.33 ¡ 0.86
2.8 – ¡ 1.09 ¡ 1.55
Trans IChemE, Vol 80, Part C, June 2002
101
obtained after fouling with WPC solution were 10%, 19%
and 16% for the C 30F, PA 50H and PES 50H membranes
respectively. These Ž gures are the average of all fouling
…1† experiments under taken. The  ux recoveries reported for
cleaning are for single experiments.
The effect of different cleaning agents upon the PA 50H
membrane fouled with 3.5 wt% whey protein solution at
1 bar for 30 minutes is shown in Figure 2 (the results from
the experiments with C 30F and PES 50 H membranes are
not shown). Figure 2 shows that sodium hydroxide concen-
tration had a major effect upon the cleaning efŽ ciency. There
is not sufŽ cient data to claim that an optimum concentration
of sodium hydroxide existed that maximized cleaning
performance. However, for all the membranes cleaned with
NaOH, a concentration of 0.075 wt% had the best cleaning
efŽ ciency. A second experiment using 0.075 wt% NaOH is
reported on right hand side of Figure 2. This was carried out
as a cross check for the Ž rst experiment. The similar  ux
recovery values conŽ rm the validity of the Ž rst result.
For the C 30F, PA 50H and PES 50H membranes the  ux
recoveries after cleaning with 0.075 wt% NaOH were 53%,
88% and 40% respectively. At sodium hydroxide concentra-
tions above and below 0.075 wt%, cleaning was less efŽ -
cient. Cleaning the PA 50H membrane with 0.75 wt%
sodium hydroxide resulted in a  ux recovery of only 53%,
compared to a  ux recovery of 88% when 0.075 wt% was
used. The existence of concentration optima in the cleaning
of whey protein fouled microŽ ltration membranes is well
documented3,4,24. Although the data is limited, these results
imply that a similar pattern occurs for UF membranes.
Compared to the performance of sodium hydroxide, the
cleaning efŽ ciency of 0.5 wt% Ultrasil 11 solution was poor
under the conditions tested. Fouled PES 50H membranes
cleaned with 0.5 wt% Ultrasil 11 and 0.075 wt% sodium
hydroxide showed  ux recoveries of 32 and 40% respec-
tively. The equivalent Ž gures for the C 30 membrane were
24 and 53%.
Libranone 960 was the most effective cleaning agent
tested in this study. PA 50H and PES 50H membranes
showed  ux recoveries of 110 and 55% respectively when
cleaned with Libranone 960. No data was collected for the
effect of cleaning the C 30F membrane with Libranone 960.
The dairy industry standard for nitric acid concentration
in CIP is 1.0 wt%. It was felt that 1.0 wt% would be too high
Figure 2. Flux recoveries for a PA 50H-membrane cleaned with sodium
hydroxide, Libranone 960 and Ultrasil 11 at 0.5 bar.
102 VÄISÄNEN et al.

a concentration for use in cleaning such thin Ž lms, particu- acidic cleaning was therefore carried out at concentrations
larly as effective cleaning was found at such low NaOH of 0.075 wt% sodium hydroxide and 0.3 wt% nitric acid.
concentrations. A lower, but still relevant concentration of Following the acidic clean, the pure water  ux recovery
0.3 wt% nitric acid was therefore selected. Sequential alkali= decreased in all cases. The PA 50H membrane showed a  ux

Figure 3. (a) The cleaning of C 30F, PA 50H and PES 50H membranes with Libranone 960 and Ultrasil 11; (b) Pure water  ux recovery after cleaning with
Ultrasil 11 and Libranone 960.

Trans IChemE, Vol 80, Part C, June 2002

 TREATMENT OF UF MEMBRANES
recovery of 85% after the Ž rst stage cleaning with sodium
hydroxide, and a  ux recovery of only 72% following the
second stage clean with nitric acid. The PES 50H membrane
showed  ux recoveries of 32 and 28% following alkali and
acid cleaning respectively. The equivalent recovery Ž gures
for the C 30F membrane were 53 and 32%.
This decline in  ux following the acid step is unexpected.
The acid step should remove any calcium phosphate deposits
from the membrane surface and weaken the deposit struc-
ture11. However, any partially hydrolysed proteinaceous
deposits remaining on the surface following alkaline cleaning
step may adopt a more shrunken morphology with a smaller
voidage on contact with nitric acid, as the deposit is likely
to gel at very low pH values. This hydrogel could resist
removal, and thereby reduce the subsequent pure water  ux
reading.
Membranes fouled with ground wood mill circulation
water
Figures 3(a) and (b) show  ux data obtained from the pulp
and paper factory. Each of the three types of membranes
tested were treated with ground wood mill circulation water
Figure 4. FTIR=ATR spectra of the virgin, fouled and cleaned C 30F membranes and possible functional groups present.
Trans IChemE, Vol 80, Part C, June 2002
103
for as close to ten minutes as possible. Cleaning cycles lasted
between 20 and 40 minutes, using either Libranone or
Ultrasil as cleaning agents. Pure water  ux was then
measured again during the subsequent pure water rinsing
cycle. The differences in starting points in the  uxes for
C30F and PA 50H membranes occur due to differences in
process conditions in the factory, and the inherent variability
in this system. The trends in the  ux recovery curves
following cleaning are the most important consideration.
Following treatment with Libranone 960, all membranes
that had been used in the pulp factory displayed a substantial
increase in pure water  ux, when rinsing started (see
Figure 3(a)). However, after a short period of time the
pure water permeability decreased. The less hydrophilic
the membrane surface, the greater the increase in  ux
following treatment with Libranone 960.
For the C 30F membrane (the most hydrophilic of the
membranes tested) the pure water permeability before
cleaning was 150 l m¡2 hr¡1. After cleaning with Libranone
the permeability had increased to 199 l m¡ 2 hr¡1. However,
after 7 minutes of rinsing this Ž gure had decreased to
170 l m¡2 hr¡1. For the PES 50H (least hydrophilic
104 VÄISÄNEN et al.
Figure 5. FTIR=ATR spectra of the virgin, fouled and cleaned PA 50H membranes and possible functional groups present.
membrane surface) the pure water permeability before
cleaning was 26 l m¡2 hr¡1 and after cleaning this Ž gure
had increased to 81 l m¡2 hr¡1. After Ž ve minutes of
rinsing, the permeability had decreased to 41 l m¡2 hr¡1.
All the measurements were taken at 1 bar pressure. When
using Ultrasil 11 cleaning agent, this pattern of substantial
 ux increase followed by a gradual decrease was not seen.
There can be no doubt that Libranone 960 is an effective
cleaning agent for ground wood mill circulation water
fouled membranes. The initial substantial increase and
subsequent fall in the  ux following cleaning can be
explained by the adsorption and subsequent desorption of
the non-ionic surfactants comprising Libranone 960. This
hypothesis is supported by the observation that the  ux
recovery rises most rapidly and falls most quickly for the
PES 50H membrane, which is the least hydrophilic of the
three examined. The fact that the terminal  ux decline value
on all Libranone cleaned membranes is still substantially
higher than the value prior to cleaning gives good conŽ -
dence of the effectiveness of this cleaning procedure.
It is clear from these results that when Libranone 960 is
used, the subsequent rinsing step results in a lower  ux.
Consequently, it is recommended that the rinsing step be
omitted in this system.
Under the conditions tested, Libranone 960 was found to
be superior to Ultrasil 11 (see Figure 3(b)). For the PES 50H
membrane the difference in the efŽ ciency of the two cleaning
agents was small; the pure water  ux after cleaning increased
to 58% when using Libranone 960, and 42% when using
Ultrasil 11. The pure water  ux decreased after cleaning the
PA 50H and C 30F membranes with Ultrasil 11.
Membrane Analysis
FTIR measurement
FTIR=ATR measurements were made on (i) virgin
membranes; (ii) membranes fouled with ground wood mill
circulation water; and (iii) membranes following cleaning.
C 30F membranes
The spectra of the virgin and fouled C 30F membranes
(see Figure 4) are quite similar; the main difference is an
extra peak in the spectrum of the virgin membrane around
wavenumber 1743 cm¡1 (CˆO). The intensity of the
Trans IChemE, Vol 80, Part C, June 2002
 TREATMENT OF UF MEMBRANES
Figure 6. FTIR=ATR spectra of the virgin, fouled and cleaned PES 50H membranes and possible functional groups present.
responses at 1236 cm¡1 (C¡O) and 1315 cm¡1 (C¡H) are
also greater in the spectrum of the virgin membrane, when
compared to a fouled membrane.
In the spectrum of the cleaned C 30F membranes (treated
with Ultrasil or Libranone) there is a greater response at
wavenumber 1542 cm¡1 (C¡N) than seen in either fouled
or virgin membranes. For the membrane cleaned with
Ultrasil 11 there are also greater responses at wave numbers
556 cm¡1, 874 cm¡ 1 and 925 cm¡1 (CˆC, C¡H).
The spectrum of the membrane treated with Libranone is
close to the virgin membrane spectrum. This cannot be said
of the spectrum for the equivalent membrane cleaned with
Ultrasil 11. It can, therefore, be concluded that the spectra
support the  ux recovery data in indicating that membranes
treated with Libranone 960 are cleaner than those treated
with Ultrasil 11.
PA 50H membranes
The spectra of the virgin, fouled and cleaned PA 50H
membranes show similar trends to those discussed for the
C 30F membranes (see Figure 5).
Trans IChemE, Vol 80, Part C, June 2002
105
Peaks in the virgin membrane are masked by the wood mill
foulants (the response at wavenumber 1167 cm¡1 (C¡O,
C¡H) is a particular example). At wavenumber 1017 cm¡1
(C¡H) a singlet is seen in the spectrum of the virgin
membrane, whereas in the spectrum of the fouled membrane
it is a doublet. In addition, one peak has changed its position
and widened from 1117 cm¡1 (C¡H) in the virgin membrane,
to 1126 cm¡1 in the spectrum of the fouled membrane.
The intensity of some peaks also differs between the virgin
and fouled membranes. Peaks at 3313 cm¡ 1 (OH, NH) and
668 cm¡1 are smaller in the spectrum of the virgin membrane,
and the peaks at 720 cm¡1 (C¡H) and 1288 cm¡ 1 (C¡O) are
larger in the spectrum of the virgin membrane.
In the spectrum of the Ultrasil 11 cleaned PA 50H
membrane, there are additional peaks at wavenumbers
582 cm¡1 and 557 cm¡ 1 (OˆC¡N) not seen for the virgin
membrane, indicating that the membrane has not been fully
cleaned. The small peak at wavenumber 1407 cm¡1 (C¡H)
is similar to that of a virgin membrane.
The spectrum for the Libranone 960 cleaned membrane is
closer to that of the virgin membrane than the spectrum
106
Figure 7. SEM-pictures showing the support layer of the virgin, fouled and
cleaned PA 50H-membranes.
shown by the Ultrasil 11 cleaned membrane. This is seen
particularly at wavenumbers 2400 and 1350. Ultrasil 11
seems to give a better Ž t to the virgin membrane spectrum at
582 (OˆC¡N).
PES 50H
Figure 6 shows the spectra of the virgin, fouled and
cleaned PES 50H membranes. From the closeness of the
spectra, it is not possible to make deŽ nitive statements on
the effectiveness of the cleaning regimes.
VÄISÄNEN et al.
Figure 8. SEM-pictures of the virgin and fouled C 30F-membranes.
SEM pictures
Figure 7 shows the PA 50H membrane in virgin, fouled
and cleaned states. The membrane consists of an active and
a support layer (In Figure 7 this is best seen in the
micrograph of the fouled membrane). Deposition is seen
in the form of nodules and aggregates less than 1 mm across.
In the cleaned membrane (see Figure 7(c)), the majority of
the larger aggregates have been removed from the
membrane surface, but a number of smaller deposits still
remain within the structure.
Figure 8 shows a cross section of the C 30F membrane in
both virgin and fouled states. The complex morphology of
the membrane is seen, with an active layer (on the right of
Figure 8(a) and 8(b)) and two distinctly different structures
in the support layers. Fouling is seen inside the support layer
of the C 30F membrane. It is clear that in addition to fouling
of the active layer, fouling also occurs on the support layers
of the membranes investigated.
SEM pictures of the cross-section of the PES 50H
membranes show that it consist of three layers. Fouling
was found in the support and porous middle layers (pictures
not shown).
AFM-images
Examination of the AFM-image shows the difference in
the surface topography of the different membrane materials
and the depressions in the top layer of the C 30F membrane.
Trans IChemE, Vol 80, Part C, June 2002
 TREATMENT OF UF MEMBRANES
Figure 9. Tapping mode AFM-pictures of the virgin and fouled PA 50H-membranes.
AFM pictures indicate that the C 30F membrane has the
roughest surface (caused by the depressions in the
membrane) and the PA 50H has the smoothest surface.
The effect of fouling on the top of the PA 50H is shown
in Figure 9. The AFM pictures show that the effect of
fouling is to increase the surface roughness of the
membrane.
AFM pictures of the surfaces of the C 30F and the PES 50H
membranes were captured, but are not reproduced here as the
effect of cleaning is seen less clearly on these AFM images.
CONCLUSIONS
In this study three different UF membranes were fouled
with whey protein concentrate and ground wood mill
circulation water, and cleaned using two commercial surfac-
tant formulations, sodium hydroxide and nitric acid.
Results are presented which show for both classes of
deposition that the most effective cleaning agent was Libra-
none 960, a non-ionic surfactant comprised of surfactants in
the range C9–11E6. For WPC fouled membranes, sodium
hydroxide concentration had a major effect on the cleaning
efŽ ciency, with excessive concentrations being detrimental
to the cleaning process. These results are similar to the trend
reported for microŽ ltration membranes11,12,29.
When Libranone 960 was used as a cleaning agent for
ground wood mill circulation water, the permeability imme-
diately increased considerably when water rinsing was
introduced. The less hydrophilic the membrane surface,
the bigger the observed  ux increase on rinsing. The  ux
reduced from a maximum as the surfactant desorbed from
the membrane surface. The rate of  ux decline during
rinsing was greatest for the least hydrophilic membrane.
The terminal  ux decline value was still considerably higher
than the fouled membrane  ux, indicating that the
membrane had been partially cleaned. It is clear from
these results that when Libranone 960 is used, the subse-
quent rinsing step results in a lower  ux. Consequently, it is
recommended that the rinsing step be omitted in this system.
Examination of the FTIR-spectra revealed the foulants that
were present and the changes in the foulant composition after
cleaning (i.e. new peaks in the spectrum) for two of the three
UF membranes studied. The membranes fouled with ground
wood mill circulation water and cleaned with Libranone
Trans IChemE, Vol 80, Part C, June 2002
107
displayed spectra that were close to those of virgin
membranes. This supports the  ux recovery data, which
indicated that membranes fouled in this way had been cleaned
more effectively with Libranone than with Ultrasil 11.
SEM pictures conŽ rmed that the location of the fouled
deposits were primarily inside the membrane structure rather
than on top of the surface, explaining why attempts to
visualize the foulants using AFM pictures showed only
limited success.
FTIR spectra indicated the foulant groups present, and
SEM and AFM pictures showed the location and extent of
the deposition. Flux recovery data alone is insufŽ cient in
indicating the cleanliness of a membrane. A combination of
complementary surface characterization methods can give
important additional information on the effectiveness of a
membrane cleaning protocol.
REFERENCES
1. Bird, M. R. and Espig, S., 1994, Cost optimization of dairy cleaning in
place (CIP) cycles, Trans IChemE, Part C, Food and Bioprod Proc,
72(C1): 17–20.
2. Bar-Sinai, Y. L. and Wayman, M., 1976, Separation of sugars and lignin
in spent sulphite liquor by hydrolysis and ultraŽ ltration, Tappi, 59(3):
112–114.
3. Bansal, I. K. and Wiley, A. J., 1975, Membrane processes for fractiona-
tion and concentration of spent sulphite liquors, Tappi, 58: 125.
4. Domingo, G. S., Jacobs, E. P. and Swart, P., 2001, Characterization of
foulants present in ef uents emanating from the Piet retief Mondi Kraft
paper mill, Poster presentation 4th WISA-MTD Symp (Stellenbosch,
South-Africa).
5. Carlsson, D. J., Dal-Cin, M. M., Black, P. and Lick, C. N., 1998, A
surface spectroscopic study of membranes fouled by pulp mill ef uent,
J Mem Sci, 142: 1–11.
6. Huuhilo, T., Väisänen, P., Nuortila-Jokinen, J. and Nyström, M., 2001,
In uence of shear on  ux in membrane Ž ltration of integrated pulp and
paper mill circulation water, Desalination, 141: 245–258.
7. Tragardh, G., 1991, Membrane applications in the food industry,
Poly J, 23: 521–529.
8. Merin, U. and DauŽ n, G., 1990, Cross  ow microŽ ltration in the diary
industry, Le Lait, 70: 281–291.
9. Mehra, R. K. and Donnelly, W. J., 1993, Fractionation of whey protein
components through large pore, hydrophilic, cellulose membranes,
Dairy Res, 60: 89–97.
10. Bird M. R. and Fryer P. J., 1991, An experimental study of the cleaning
of surfaces fouled by whey proteins, Trans IChemE, Part C, Food and
Bioprod Proc, 69(C1): 13–21.
11. Bartlett, M., Bird, M. R. and Howell, J. A., 1995, An experimental study
for the development of a qualitative membrane cleaning model, J Mem
Sci, 105: 147–158.
108 VÄISÄNEN et al.
12. Bird, M. R. and Bartlett, M., 1995, CIP optimization for the food
industry: Relationships between detergent concentration, temperature
and cleaning time, Trans IChemE, Part C, Food and Bioprod Proc,
73(C1): 63–70.
13. Gan, Q., Field, R. W., Bird, M. R., England, R., Howell, J. A.,
McKechinie, M. T. and O’Shaughnessy, C. L., 1997, Beer clariŽ cation
by cross ow microŽ ltration: Fouling mechanisms and  ux enhance-
ment, Trans IChemE, Part A, Chem Eng Res Des, 75(A1): 3–8.
14. Gan, Q., Howell, J. A., Field, R. W., England, R., Bird, M. R. and
McKechinie, M., 1999, Synergetic cleaning procedure for a ceramic
membrane fouled by beer microŽ ltration, J Mem Sci, 155: 277–289.
15. Geise, J. H., 1991, Sanitation: The key to food safety and public health,
Food Tech, 45: 74–80.
16. Nyström, M. and Zhu, H., 1997, Characterization of cleaning result
using combined  ux and streaming potential methods, Colloid Surfaces,
131: 195–205.
17. Parkin, M. F. and Marshall, K. R., 1976, The cleaning of tubular
cellulose acetate ultraŽ ltration membranes, NZ J Dairy Sci Tech, 11:
107–113.
18. Zhu, H. and Nyström, M., 1997, Cleaning results characterized by  ux,
streaming potential and FTIR measurements, Colloid Surfaces, 138:
309–321.
19. Fontyn, M., van’t Riet, K. and Bijsterboch, B. H., 1991, Surface
spectroscopic studies of pristane and fouled membranes. Part I.
Method development and pristine membrane characterization, Colloid
Surfaces, 54: 331–347.
20. Oldani, M. and Schock, G., 1989, Characterization of ultraŽ ltration
membranes by infrared spectroscopy, ESCA and contact angle measure-
ments, J Mem Sci, 43: 243–258.
21. Pihlajamäki, A., Väisänen, P. and Nyström, M., 1998, Characterization
of clean and fouled polymeric ultraŽ ltration membranes by Fourier
transform IR spectroscopy-attenuated total re ection, Colloid Surfaces,
138(2–3): 323–333.
22. Güell, C. and Davis, R. H., 1996, Membrane fouling during microŽ l-
tration of protein, J Mem Sci, 119: 269–284.
23. Hernández, A., Calvo, J. I., Prádanos, P. and Tejerina, F., 1998, Pore size
distribution of track-etched membranes; comparison of surface and bulk
porosities, Colloid Surfaces, 138: 391–401.
24. Noda, K., Masuda, H. and Kodama, T., 1999, AFM study on the rusting
process on iron under atmospheric corrosion environment, Inter Corro-
sion Council -CD-ROM.
25. Singh, S., Khulbe, K. C., Matsuura, T. and Ramamurthy, P., 1998,
Membrane characterization by solute transport and atomic force micro-
scopy, J Mem Sci, 142: 111–127.
26. Xinglong, X. and Coleman, M. R., 1997, Atomic force microscopy
images of ion-implanted 6FDA-pMDA polyimide Ž lms, J App Poly Sci,
66: 459–469.
27. Nyström, M., Pihlajamäki, A. and Ehsani, N., 1994, Characterization of
ultraŽ ltration membranes by simultaneous streaming potential and  ux
measurements, J Mem Sci, 87: 245–256.
28. Smoluchowski, M., 1905, Zur theorie der elektrischen kataphorese und
der ober ächenleitung, Phys Z, 6: 529–536.
29. Shorrock, C. and Bird, M. R., 1998, Membrane cleaning: The removal
of irreversibly fouled yeast deposits, Trans IChemE, Part C, Food and
Bioproducts Processing, 76(C1): 30–38.
ADDRESS
Correspondence concerning this paper should be addressed to Dr M. R.
Bird, Department of Chemical Engineering, University of Bath, Claverton
Down, Bath, BA2 7AY, UK. E-mail: m.r.bird@bath.ac.uk
The manusript was received 27 June 2001 and accepted for publication
after revision 13 February 2002.
Trans IChemE, Vol 80, Part C, June 2002