Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
Black tea liquor ultrafiltration: effect of ethanol pre-treatment upon
fouling and cleaning characteristics
Iain S. Argyle1, Arto Pihlajamӓki2, Michael R. Bird1*
1Membrane Applications Laboratory, Department of Chemical Engineering, University of
Bath, Bath, BA2 7AY, UK 2Laboratory of Separation Technology, Group of Membrane Technology, Lappeenranta
University of Technology, Lappeenranta, Finland
*corresponding author - [email protected]
Keywords: ultrafiltration; pre-treatment; black tea; fouling; cleaning
Abstract
This paper reports on the use of polymeric ultrafiltration (UF) membrane ethanol pre-
treatment as a strategy to improve filtration performance in terms of both flux increase, and
membrane water flux recovery following cleaning. A 4-fold pure water flux (PWF) increase
was observed for a 100 kDa polysulfone membrane. Marked increases in permeate flux were
recorded for ethanol treated UF membranes over a range of molecular weight cut-off values.
Ethanol treatment also aided fluxes over multiple foul-clean cycles, and enabled the enhanced
transmission of polyphenols during UF for the clarification of black tea liquor. Following tea
fouling and NaOH cleaning repetitions, PWF values of treated membranes were returned to
values of >150% of the untreated virgin membrane PWF over 4/5 consecutive cycles,
indicating that the ethanol pre-treatment strategy adopted had a prolonged effect upon
subsequent performance.
1
1. Introduction
Fouling of membranes leads to reductions in performance and thus process efficiency. For
membrane systems this is manifested in flux losses and often inefficient membrane
regeneration through expensive and time-consuming cleaning operations. Returning
membranes to their pristine condition and achieving a performance similar to that of a new
membrane is the over-arching aim of the cleaning process, with recovery typically being
measured as percentage of original membrane permeability [1].
There is a general requirement for technological improvements in tea product manufacturing
processes due to the growing worldwide popularity of iced tea, or more generally, RTD tea
products, evident in the fact that they occupied 8.7% of the ca $500 billion soft drinks market
in 2012-13 [2, 3]. Application of UF in black tea processing has received specific focus for
its potential to remove the natural tea cream which forms upon cooling of tea liquors [4].
Membrane processes offer distinct advantages over other haze removal methods such as
alkali solubilisation, enzymatic treatments, centrifugation or fining by eliminating or reducing
chemicals addition, energy consumption and processing time [5, 6], over the aforementioned
methods. Membrane processing can additionally offer a continuous processing option and
integrated pasteurisation process [7].
Tea haze is noticeable by the mottled appearance which develops on the surface of a tea brew
as cooling commences. This haze forms an unpleasant haze characteristic in tea; the haze
agglomerates having been shown to include tea constituents which contribute to the taste and
colour of black tea and thus loss through hazing (through the consequent phase separation/
precipitation) and result in a perceived lowering of quality [8]. The formation of haze has
been attributed predominantly to the interaction of polyphenols with themselves or other
components e.g. proteins, caffeine and metal cations [8-10] and is a function of pH,
concentration and temperature-time history [11] through interactions by homo-association of
polyphenolic galloyl groups (π – π stacking interactions) or hetero-association of species
through H-bond, hydrophobic or [10, 12, 13]. 85% of haze particulates in the 0.1 – 100 µm
have been shown to be below 1.03 µm suggesting removal of these particulates through
physical size exclusion is a feasible strategy using microfiltration (MF) or UF [5]. Solubility
experiments on tea cream moieties have shown that it is stable below 40 °C and remains
partially formed up to 90 °C [11], indicating that formation begins from the instant an
2
infusion begins to cool (at ca 90 °C) and that removal of haze initiators (i.e. early stage haze
agglomerates) can be carried out using membranes prior to haze maturation.
A drawback of membrane filtration of tea is the preferential rejection of polyphenols over
other constituents as indicated by Chandini et al. (2013) [6]. The phenolic species have also
been identified as being prolific foulants of membranes [14]. Evans et al. (2008) showed that
negatively charged fluoropolymer (FP) membranes fouled by tea inherited additional
negative charges associated with tea constituents, before cleaning returned them to a charge
between that of the virgin membranes and fouled membranes. In further work, repetitive
fouling and cleaning of an FP 30 kDa cut-off membrane showed progressive membrane
surface hydrophilisation upon cleaning, the contact angle being reduced from 65° to 52° after
23 fouling and cleaning cycles [15]. The study highlighted the synergy of tea fouling and
NaOH cleaning and the resulting modification in surface chemistry. Also apparent from
previous work is the importance of membrane material with particular relevance to surface
hydrophilicity and through pore charge with respect to fouling propensity and species
transmission [16].
There are three main strategies for enhancing the performance of commercial membranes:
i. Optimising the hydrodynamic conditions or changing physicochemical variables
(temperature, ionic strength, pH) to either prevent foulant deposition or to induce
changes to solution or solute properties.
ii. Improving foulant removal by optimisation of detergent concentration and cleaning
thermo-hydraulics to improve membrane regeneration.
iii. Pre-treating membranes to modify membrane performance by changing the chemical
or physical nature of the membrane surface and/or underlying porous matrix.
Pre-treatment of UF membranes with alcohols prior to aqueous filtration is an under-
investigated area of research, with little work being reported in the literature [17, 18].
Kochan et al. (2009) [18] reported near 3-fold increases when treating polysulfone flat sheet
membranes with 80 wt.% ethanol solutions, the treatment showing an improvement in
performance for up to 5 days when permeability was tested with pure water. The authors
showed a flux uplift during sludge supernatant filtration when the same membrane was pre-
treated with 80 wt.% ethanol for 2 hours. This shows that this apparently simple pre-
3
treatment method can have positive and lasting effects on membranes. Little attention has
been given to the effects of surface chemistry as a result of alcohol modification. Further, the
application of modified membranes to the filtration of an industrially relevant feed (tea),
whose filtration performance has been shown to be influenced by surface chemical
parameters such as charge and hydrophobicity is made here. Tea as a surface modifier also
shows interesting phenomena relating to membrane surface chemistry, when considering
fouling and cleaning synergy.
This paper reports on results of a standard fouling and cleaning protocol applied multiple
times to polymeric UF membranes which have been subjected to pre-treatment using ethanol.
An emphasis on filtration performance, PWF recovery and transport mechanisms is made, as
well as connections between these factors and the underlying surface chemical changes
associated with ethanol pre-treatment and subsequent tea fouling and cleaning. Flux uplifts
during sub-limiting flux operation are reported. A lasting effect of this treatment method is
seen over multiple fouling and cleaning cycles, and influences to species transmission are
apparent. Overall, improvements in the membrane filtration of tea is documented, which has
relevance to the broader food process engineering sector in the context of developing
adjusted processes, improving process efficiency and improving product quality.
4
2. Materials and methods
2.1. Membranes and filtration system
UF was performed using a DSS Lab-Unit M10 (DSS, Denmark). Three different flat sheet
polysulfone membranes (PS50, PS100, MFG1) with nMWCOs of 50 kDa and 100 kDa, and
one with 0.1 µm nominal pore size (Alfa Laval, Denmark) were used for experimentation. An
effective filtration area of 336 cm2 was available within the plate and frame module used.
After being subjected to hot water conditioning to remove anti-humectant, membrane PWFs
were measured at 22 °C and 1.0 bar transmembrane pressure (TMP).
2.2. Pre-treatment
Following rinsing and hot water conditioning, membranes were used in their untreated state
or subjected further to ethanol treatment. Ethanol (50 wt.% or 100 wt.%) (ACS grade,
obtained from Sigma Aldrich) was used to purge and fill the membrane module. Nitrogen gas
was used to pressurise an ethanol solution reservoir and subsequently force ethanol through
the membrane at 1.0 bar (in the manner of dead-end filtration). When saturated the
membranes remained immersed and under 1.0 bar of hydrostatic pressure for 24 hours before
purging with reverse osmosis (RO) water and re-measurement of PWF. PWF after treatment
was divided by PWF prior to treatment to give relative flux change (equation 1).
Jrel=J final
J initial (1)
Where Jrel is the relative flux change, J final is the post-treatment water flux and J initial is the
original PWF.
2.3. Feed and cleaning regimes
Soluble spray-dried black tea powder was supplied by Unilever, UK. Tea powder was
reconstituted with RO water at 90 °C to make 8 L of 0.5 wt. %. Temperature was adjusted to
50 °C before 60 minute filtration. For cleaning, 0.5 wt. % NaOH at 60 °C was used at 1.0 bar
TMP for 10 minutes following ambient rinsing for 15 minutes at 1.0 bar TMP. Cross flow
velocities (CFV) for PWF measurement, fouling, and cleaning were fixed at 1.0 ms-1, rinsing
was carried out at 1.5 ms-1. Between each phase of the foul-clean cycle, PWF was measured
at 1.0 bar. See Table 1 for full protocol.
5
Table 1. Pre-treatment and filtration protocol
Phase of operation # Fluid (concentration) Conditions (T, TMP, CFV, t)
PWF 1 RO water (-) 22 °C, 1.0 bar, 1.0 ms-1, 10 min
Pre-treatment 0/1 None/ethanol (50/100 wt. %) 22 °C, 1.0 bar, static, 24 hours
PWF 0/1 RO water (-) 22 °C, 1.0 bar, 1.0 ms-1, 10 min
Filtration 4/5 Black tea (0.5 wt. %) 50 °C, 1.0 bar, 1.0 ms-1, 60 min
PWF 4/5 RO water (-) 22 °C, 1.0 bar, 1.0 ms-1, 10 min
Rinse 4/5 RO water (-) 22 °C, 1.0 bar, 1.5 ms-1, 15 min
PWF 4/5 RO water (-) 22 °C, 1.0 bar, 1.0 ms-1, 10 min
Clean 4/5 NaOH (0.5 wt. %) 60 °C, 1.0 bar, 1.0 ms-1, 10 min
PWF 4/5 RO water (-) 22 °C, 1.0 bar, 1.0 ms-1, 10 min
2.4. Solute analysis
Permeate samples were collected at 15 minute intervals during each filtration with retentate
samples taken at filtration commencement. All collected permeate (except samples) was
returned to the feed tank to maintain constant conditions. Samples were analysed for total
solute concentration via dry weight measurement by dehydration at 60 °C and for total
polyphenols via the method as described by Singleton and Rossi (1965) [19]. For total
polyphenols, measurements were carried out in 96 well micro-titre plates at 765 nm
wavelength using a Synergy HT multi-mode plate reader (BioTek, Winooski, USA). Gallic
acid standards were used as a reference. Apparent rejection (Rapp) of total solids and
polyphenols was made by calculation using both sample concentration (C p) and
feed/retentate concentration (C r) measurement using equation 2.
Rapp=1−C p
C r (2)
2.5. Turbidity
Degree of haze was measured by way of turbidity in consistently diluted samples. Permeate
sample concentration was approximated by use of refractive index indicating a °Brix value.
Permeates were then diluted to 0.3 wt.% based on this measurement and turbidity recorded
using an HI 93703 turbidimeter (Hanna Instruments, Woonsocket, USA) against turbidity
standards (0, 10 and 100 NTU) at room temperature within 2 hours of filtration.
6
2.6. Tea colour
Samples were loaded into cuvettes before L* (lightness), a* (redness) and b* (yellowness)
parameters (CIELAB colour space) were measured and analysed using a Shimadzu UV-1601
spectrophotometer over a 380 – 770 nm wavelength range and interpreted using UVPC Color
Analysis software v.3.0 (also Shimadzu).
2.7. Contact angle
Water contact angles were measured via the sessile drop technique using an OCA 15 Pro
goniometer (Dataphysics GMBH, Germany). Samples dried in room temperature air and
stored in a desiccator prior to loading with a 1.0 µL drop of pure water. OCA 15 software
was used to calculate the contact angle.
2.8. Apparent zeta potential
Apparent zeta potential (ZP) was calculated by measurement of streaming potential through
membranes pores as per the method described by Nyström et al. (1994) [20]. 1.0 mM KCl
solution was made using Milli-Q water and was fluxed through 10.4 cm2 membrane samples
held in a custom module housing two Ag/AgCl electrodes; one positioned near the membrane
surface at the retentate side, and one near the support layer at the permeate. Streaming
potential was measured over a range of TMPs (typically 0.5 – 1.0 bar) before pH adjustment.
Three pH values were measured in the decreasing pH direction before purging and rinsing of
the electrolyte solution. Streaming potential for three further pH values in the increasing pH
direction were then conducted. All measurements were in the range of between pH 3 and pH
7. Zeta potential was calculated by assessing the rate of change streaming potential with
respect to TMP using the uncorrected form of the Helmholtz-Smoluchowski equation
(equation 3).
ζ = ΔETMP
μkε0 εr
(3)
Where ζ is the zeta potential, TMP is transmembrane pressure, μ is electrolyte solution
viscosity, k is conductivity, ∆ E is streaming potential change, ε 0 and ε r are permittivity of
free space and the dielectric constant of water respectively.
7
3. Results and discussion
3.1. Effect of alcohol pre-treatment
Figure 1 shows the relative flux increases observed following treatment of the membranes
used. For the PS50 membrane (untreated average virgin PWF of 110.1 ± 5.2 Lm-2hr-1),
treatment with 50 wt.% and 100 wt.% ethanol resulted in relative increases of 1.32 ± 0.01 and
2.32 ± 0.04. For the PS100 membrane (untreated average virgin PWF of 254.7 ± 7.2 Lm-2hr-
1), treatment gave increases of 3.75 ± 0.11 and 3.84 ± 0.06, this membrane giving the most
prolific flux uplifts. For MFG1, with PWF of 309.4 ± 1.5 Lm-2hr-1, the relative uplifts were
2.09 ± 0.04 and 2.13 ± 0.03.
PS50 PS100 MFG1
Relative flux (
J rel) (-)
0
1
2
3
4
5No treatment 50 wt.% ethanol 100 wt.% ethanol
Figure 1. Relative PWF for membranes following 24 hours of static pre-treatment under 1.0
bar pressure in various ethanol-water mixtures.
The polar nature of water influences its organisation at the solid-fluid interface and is
dependent on the charge interactions between the two phases. Considering that the negative
charge is reduced to a more neutral value relative to the untreated state (as later shown in
figure 8), the adherence of water molecules to the wall would be reduced. The effect of
reduced charge would thus have the effect of increasing the slip at the wall. Normally it
would be assumed that fluid flowing through pores would observe a no slip boundary
8
condition where the outermost molecules are stationary due to this adherence. The results
indicate that if a charge modification to the pore brings about such evident permeability
increases, generally assuming a no slip condition for membranes with similar properties
(moderate surface charge and hydrophobicity) would thus be insufficient and that
incorporation of slip due to surface charge (such as for cases involving the modelling of fluid
flow in porous structures) should be a general consideration.
The exact mode of action of the alcohol remains unclear at this stage though a number of
hypotheses can be suggested. Either ethanol molecules are adhering in a film with their
orientation such that their polar end is attached to the wall and the remaining non-polar
hydrophobic tail protrudes into the pore space reducing the adhesion of water. The charge
modification could thus be a result of charge masking given its lower conductivity. Another
theory is that alcohol is partially dissolving the polymer matrix (e.g. swelling) which, as well
as potentially modifying the membrane morphology, could bring about changes in its
chemical nature. The action of swelling, and any morphological changes which occur as a
result, could increase the effective pore diameter of some membranes. Swelling in three-
dimensional porous structures can lead to two opposing effects: 1) macroporous swelling
would lead to a dilation of the entire structure causing expansion of pores and 2) microporous
swelling would result in the polymer swelling though the effective total space occupied by
the membrane would change negligibly. Since both effects will occur simultaneously, the
relative proportions of these two phenomena may differ between membranes offering
differing degrees of uplift in pure water flux following treatment.
The PS50 membrane shows a greater rise in flux between 50 wt.% and 100 wt.% ethanol
treatment, whereas the PS100 shows its most significant rise in flux between being untreated,
and treated with 50 wt.% ethanol. If adsorption of ethanol molecules to the pore walls is
considered, the initial magnitude of charge would influence the amount of ethanol adsorbed
since ethanol must displace water at the solid-fluid interface in order to adsorb to the wall.
For a more strongly charged virgin membrane (-15 mV for the PS50 membrane) this
saturation effect (or equilibrium) would thus be favoured towards the molecule with greater
relative polarity (water) and would thus require a higher presence of ethanol molecules to
effect a change over the more neutrally charged surfaces (PS100).
3.2. Tea filtration
9
The benefits of the treatment protocol can be seen when performing filtration with 0.5 wt.%
black tea. As for the post-treatment PWFs, it is the difference between 50 wt.% and 100 wt.%
treatment that made the greatest difference in filtration fluxes for PS50 (Figure 2a). For the
PS50 without treatment, the relative terminal filtration flux was 0.255 ± 0.010. On the fourth
filtration, this rose to 0.304 ± 0.011. Contrastingly to the untreated operation, repetitive
fouling of conditioned membranes showed a decrease in terminal filtration fluxes; falling
from 0.455 ± 0.016 to 0.380 ± 0.013. Quantifying these values in terms of percentage change,
the PS50 with 50 wt.% ethanol treatment shows a 21% improvement falling to a 6%
improvement after 4 consecutive cycles. For 100 wt.% ethanol treatment, a 79% increase
during the first filtration is observed falling to a 25% enhancement after 4 cycles. The PS50
membrane generally showed large and prolonged increases when treated with ethanol, with
ethanol concentration increases correlating positively with maintained flux improvements
over multiple cycles.
Figure 2b and 2c show similar results for PS100 and MFG1 respectively. The increases are
less significant than for the PS50 membrane though the similar trends emerge in terms of
depleted performance after multiple cycles for treated membranes, and an improved
performance for untreated membranes. Another observation is that the pure water
permeability values mirror the changes during filtration, with 50 wt.% and 100 wt.% ethanol
treatment giving more closely allied terminal flux values compared to the untreated case,
whereas the PS50 treated with 100 wt.% ethanol gave by far and away the best performance
over untreated and 50 wt.% ethanol treatment.
10
Total filtration time (s)0 3600 7200 10800 14400
Relative filtration flux (-)
0.0
0.2
0.4
0.6
0.8
1.0No treatment50% ethanol100% ethanol
Total filtration time (s)0 3600 7200 10800 14400 18000
Relative filtration flux (-)
0.0
0.1
0.2
0.3
0.4
0.5No treatment50% ethanol100% ethanol
Total filtration time (s)0 3600 7200 10800 14400
Relative filtration flux (-)
0.0
0.1
0.2
0.3
0.4
0.5No treatment50% ethanol100% ethanol
a
b
c
Figure 2. Relative tea filtration fluxes for 0.5 wt.% feed (a) PS50, (b) PS100, and (c) MFG1
over successive foul-clean cycles
11
3.3. Cleaning and flux recovery
The increase in flux is also observable over 4 foul and 4 cleaning (4F4C) cycles by
determination of the percentage flux recovery compared to the virgin untreated PWF (Figure
3). For PS50, the performance after cleaning multiple times mirrored the initial pre-treatment
flux uplifts, though degradation of the benefits of treatment was apparent. For 100 wt.%
ethanol treatment, membrane regeneration was 188% ± 12% after F1C1, falling to 166% ±
9% after 4F4C and for 50 wt.% ethanol, 116% ± 5% to 110% ± 2% for the same membrane
ages respectively. Considering the PS100 (Figure 3b); the untreated membrane is less well
regenerated than the PS50 membrane, and the MFG1 membrane (Figure 3c) is even less well
returned to its original condition. The PS100 offered ca 80% return to the virgin PWF, and
the MFG1 when cleaned using the experimental protocol described herein, only offered a 77
– 79% regeneration to the virgin water flux performance. When observing the regeneration of
the membrane after initial ethanol treatment, it is seen that the effect of 50 wt.% ethanol
treatment is diminished after F1C1, even when the substantial PWF gains after treatment and
prior to fouling are in evidence, although the effect can still be seen.
For the PS100 membrane with 100 wt.% ethanol treatment, the largest regeneration of the
PWF is seen, this showing a return of 203% ± 24% and representing a two-fold gain over the
untreated PWF following F1C1 and maintaining a 133% ± 2% figure after 5F5C. The MFG1
membrane also showed improvements over the untreated membrane for 4F4C. For the 50%
ethanol treatment case, this membrane offered a 118% ± 12% regeneration improvement after
1F1C, and 98% ± 8% recovery after 4 cycles, in comparison to an untreated membrane flux
recovery of 78% ± 2% after the same number of cycles. This pattern is also mirrored for the
100 wt.% ethanol treated membranes of similar specification.
When comparing Figures 2a, 2b, and 2c it is apparent that while untreated membrane
performance is improving with successive foul-clean cycles, the performance of treated
membranes stays constant, or is diminished slightly. The introduction of foulants (in the form
of tea species) adds complexity, when explaining the flux uplifts, but it is likely to be caused
by the same effects as those seen for the PWF uplifts. The greater convective mass transport
of dissolved material is a direct result of increased membrane permeability, though factors
such as surface fouling (both cake layer or by surface adsorption) and in-pore fouling present
limits to the ideal case (filtering at the PWF). Given that cleaning is not achieving PWF
regeneration to values comparable to that of the treated membrane, and the magnitude of
12
regeneration falls with successive cycles, it suggests that the treatment is not effective in
permanently modifying the membrane. Further work will be needed to determine the point
where there is no additional benefit in treating the membrane, or whether such effects are
noticeable over the life time of the membrane.
Foul 1 Foul 2 Foul 3 Foul 4
Flux recovery (% of virgin P
WF)
0
50
100
150
200
250 No treatment 50 wt.% ethanol100 wt.% ethanol
Foul 1 Foul 2 Foul 3 Foul 4 Foul 5Flux recovery (%
of virgin PW
F)
0
50
100
150
200
250 No treatment 50 wt.% ethanol100 wt.% ethanol
Foul 1 Foul 2 Foul 3 Foul 4
Flux recovery (% of virgin P
WF)
0
30
60
90
120
150 No treatment 50 wt.% ethanol100 wt.% ethanol
a
b
c
Figure 3. PWF recovery measured following each foul-clean cycle for (a) PS50, (b) PS100,
and (c) MFG1.
13
3.4. Solute transmission
Ethanol treatment had no effect on solids rejection for the PS50 membrane (see Figure 4a).
This was also the case for polyphenols rejection (Figure 5a) with the exception of the first
filtration cycle, with untreated membranes showing a more substantial polyphenol rejection
(0.59 ± 0.02 for F1 and 0.48 ± 0.01 for F2). This outcome was also observed for PS100
shown in Figure 4b (F1: 0.56 ± 0.02 and F2: 0.49 ± 0.01), and MFG1 (F1: 0.48 ± 0.02 and
F2: 0.41 ± 0.02) (Figure 5c). The solids transmission of PS100 and MFG1 also showed
deviation for the first filtration cycle, in comparison to subsequent cycles. An explanation of
this could relate to fouling effects during the UF of black tea, in which membrane charge
alteration was observed after multiple fouling and cleaning cycles [16].
An interesting observation was for polyphenol rejection from the PS100 membrane (and to a
lesser extent from the MFG1) (Figures 5b and 5c). The effect of ethanol has a profound effect
upon the transmission of polyphenols species, these species contribute to the pleasant
astringency and bitterness of black tea, and thus enhanced transmission of these species is of
paramount importance; removal from the feed will cause a perceivably lower quality
beverage. For the PS100 membrane the rejection of polyphenols is drastically lowered for the
first filtration, and each subsequent filtration remains at these lower values, though the
untreated membrane is incrementally encroaching on the treated membrane values with each
foul-clean cycle; this provides a further indication that the effectiveness of ethanol treatment
reduces cycle by cycle. This implies that ethanol treatment can give a double benefit;
improved fluxes and improved transmissions of essential polyphenolics during tea
clarification.
The drop in transmission following the first to the second fouling cycle could be a result of
membrane shrinkage after use from its pre-treated swelled state. The PS100 membrane
showed the greatest uplift in pure water flux, and the sharpest reduction in pure water flux,
after cleaning cycles were completed. Another possibility is that the removal of pore forming
agents from the membrane upon pre-treatment could lead to changes in fouling propensity
and thus the effectiveness of the cleaning regimemay reduce as these species are removed.
For untreated membranes, the effect of caustic (though at relatively mild concentration) could
also induce swelling. This will be the subject of a future investigation.
14
Foul 1 Foul 2 Foul 3 Foul 4
Rejection coefficient (-)
0.0
0.1
0.2
0.3
0.4
0.5
No treatment50 wt.% ethanol100 wt.% ethanol
Foul 1 Foul 2 Foul 3 Foul 4 Foul 5
Rejection coefficient (-)
0.0
0.1
0.2
0.3
0.4
0.5
No treatment50 wt.% ethanol100 wt.% ethanol
Foul 1 Foul 2 Foul 3 Foul 4
Rejection coefficient (-)
0.0
0.1
0.2
0.3
0.4
0.5
No treatment50 wt.% ethanol100 wt.% ethanol
a
b
c
Figure 4. Total tea solids rejection for (a) PS50, (b) PS100, (c) MFG1
15
Foul 1 Foul 2 Foul 3 Foul 4
Rejection coefficient (-)
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
No treatment50 wt.% ethanol100 wt.% ethanol
Foul 1 Foul 2 Foul 3 Foul 4 Foul 5
Rejection coefficient (-)
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
No treatment50 wt.% ethanol100 wt.% ethanol
Foul 1 Foul 2 Foul 3 Foul 4
Rejection coefficient (-)
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
No treatment50 wt.% ethanol100 wt.% ethanol
a
b
c
Figure 5. Total phenolics rejection for (a) PS50, (b) PS100, (c) MFG1
16
3.5. Colour
Measurement of tea colour by spectrophotometric means not only gives an idea as to the
consumer perception of a tea infusion but also an indication to the relative concentrations of
certain tea species. Redness is known to give an indication to the presence of thearubigins in
tea [21], a class of highly polymerised polyphenols offering complexity in taste and aroma.
Data collected shows the highly increased redness in the permeate stream when filtering with
a PS100 treated membrane (Figure 6). A high degree of the redness is stripped from the tea
although this loss decreases as the membrane ages. In accordance with the polyphenols
rejection data, the redness is vastly improved for the treated membrane permeate for PS100
(this was also the case with MFG1 although not shown in this paper) and the incremental
redness increase after each foul-clean cycle is again shown with treated membranes. This
shows that as the membranes age with subsequent foul-clean cycles, its ability to transmit
polyphenols improves regardless of membrane pre-treatment.
Filtration time (min)10 20 30 40 50 60
a* (-)
-2
0
2
4
6
8
Foul 1Foul 2Foul 3Foul 4Foul 5
Filtration time (min)10 20 30 40 50 60
a* (-)
-2
0
2
4
6
8
Foul 1Foul 2Foul 3Foul 4Foul 5
Filtration time (min)10 20 30 40 50 60
a* (-)
-2
0
2
4
6
8
Foul 1Foul 2Foul 3Foul 4Foul 5
No treatment 50% ethanol 100% ethanol
Figure 6. Redness of tea permeates over filtration cycles for PS100 membranes
17
3.6. Haze removal
Haze removal was highly efficient as shown in Figure 7, which shows the average of 4
permeate samples between 15 minutes and 60 minutes of filtration. Neither membrane type
nor treatment type affected the decreaming capability. The highest haze value recorded (ca
3.0 NTU for the untreated MFG1 membrane) is not statistically relevant as being any
different from all other values measured and still represents removal in the range of an order
of magnitude from the unfiltered tea (96 ± 5 NTU for 0.3 wt.% tea at room temperature).
Figure 7. Haze of tea permeates over consecutive filtration cycles for PS100 membranes
3.7. Contact angle
Table 2 displays values for the contact angles measured at various points during the filtration
cycles. The resulting contact angle after ethanol treatment lead to increases, which generally
implies a stronger fouling propensity. As was mentioned in the introduction, tea fouling has
been shown to hydrophilise membranes following cleaning. This result is again shown here
for untreated membranes. For 100 wt.% ethanol treated PS50 membranes, fouling and
cleaning returns the membrane to a hydrophilicity almost identical to the untreated state. The
returning of the membrane to their untreated state (following treatment and foul-clean cycle
repetition) did not occur for the PS100 membrane though rehydrophilisation was occurring at
a slower rate. This indicates for both membranes that incomplete foulant removal is present
with species of a hydrophilic nature remaining on the surface. This can be looked at in a
positive manner as hydrophilisation is a common strategy for providing anti-foul membranes
18
though this would classically be achieved through surface grafting of hydrophilic molecules
or plasma treatment to yield more hydrophilic functional surface groups.
Table 2. Contact angle for membranes at various points during the foul/clean cycles (all units
in degrees)
Membrane Treatment Virgin F1C0 F1C1 F4C3F4/5* C4/5*
PS50
None 74.2 ±1.9 35.5 ± 2.5 66.5 ± 0.7 38.1 ± 4.1 65.3 ± 2.1
100 wt.% ethanol
85.0 ± 1.0 28.8 ± 2.4 73.7 ± 3.7 34.4 ± 4.8 72.4 ± 1.0
PS100
None 66.1 ± 1.7 21.5 ± 4.4 70.0 ± 2.8 - *69.4 ± 2.8
100 wt.% ethanol
73.9 ± 0.4 16.1 ± 5.5 72.3 ± 2.9 - *70.4 ± 0.6
3.8. Zeta potential
Figure 8 shows that for virgin membranes (both PS50 and PS100) the negative charge is
reduced upon ethanol treatment, which would normally suggest a greater fouling tendency
towards negatively charged phenolate ions (dissociated polyphenol species). The PS50
membrane shows the largest charge alteration, and after fouling and cleaning, this reduction
is still apparent. The fouled ZP for the untreated membrane is less negatively charged than for
the virgin, F1C1 and F4C4 membranes. The fouled ZP for the treated PS50 measures
similarly to that of the untreated equivalent suggesting the charge of the fouling layer is
consistent and not influenced heavily by the membrane. This is also true for PS100. Upon
cleaning, the reduced charge of the treated membranes is likely to reduce the adhesion of
negatively charged cations to the membrane and is a possible explanation for the enhanced
flux recovery seen, especially for the PS100 membrane which displayed the lowest
magnitude of charge following cleaning. The untreated PS100 membrane (Figure 8c) shows
19
little change after F1C0, F1C1 and F4C4 whereas the charge on the treated membrane
(Figure 8d) is not returned to the lower magnitude of negative charge of the virgin treated
membrane. This suggests that foulants that are less easily removed from the treated
membrane following cleaning are potentially aiding the transmission of phenolic species.
This would seem logical, as lower negative charge repulsion would occur with a lower
magnitude of negative ZP.
pH (-)3 4 5 6 7 8
Apparent ZP
(mV
)
-20
-15
-10
-5
0virginfouled 1fouled 1 cleaned 1fouled 4 cleaned 4
pH (-)3 4 5 6 7 8
Apparent ZP
(mV
)
-20
-15
-10
-5
0
virgin membranefouled 1fouled 1 cleaned 1fouled 4 cleaned 4
a b
pH (-)3 4 5 6 7 8
Apparent ZP
(mV
)
-20
-15
-10
-5
0
virginfouled 1fouled 1 cleaned 1fouled 4 cleaned 4
pH (-)3 4 5 6 7 8
Apparent ZP
(mV
)
-20
-15
-10
-5
0
virgin membranefouled 1fouled 1 cleaned 1fouled 4 cleaned 4
c d
Figure 8. Apparent zeta potentials calculated from streaming current measurements through
membrane pores for PS50. (a) untreated, and (b) 100 wt.% ethanol treated. PS100 (c)
untreated, and (d) 100 wt.% ethanol treated
20
4. Conclusions
Ethanol treatment of polymeric membranes has been shown to be effective in both improving
filtration fluxes and prolonging the performance of membranes over multiple foul-clean
cycles. Ethanol treatment has also shown that performance modification facilitates enhanced
transmission of polyphenolic species and the normally disproportionate rejection of
polyphenolic species is mediated somewhat by modifications in surface charge, this also
being apparent over multiple foul-clean cycles. Whilst negative charge reduction of the
membrane following treatment would normally be seen as detrimental to performance, the
propensity for fouling with species which act to beneficially modify the membrane is
improved. Tea filtered through modified membranes has been shown to be of superior quality
to the unmodified membrane permeates (and improves cycle on cycle), shown through
measurement of total phenolics and colour. Additional experiments to examine the ageing
over longer time scales would be of interest to determine whether ethanol could affect the
membrane through chemical degradation, and whether re-treatment after a given number of
cycles can recharge the optimal performance observed after the first fouling cycle in this
study. The likelihood is that re-treatment with solvents could also enhance the cleaning of the
membranes and help to return them to a near pristine state. Another requirement when
assessing the commercial benefit of such an operation would be to match the choice of
solvent to the subsequent filtration performance of the particular feed in question, as such an
optimal solvent choice may be feed specific.
5. Acknowledgements
The authors would like to thank Dr Frank Lipnizki of Alfa Laval for kindly supplying the
membranes used in this study. Spray dried tea powder was kindly supplied by Unilever R&D
Colworth, UK. We thank the EPSRC for supporting this project through a Doctorial Training
Account.
21
5. REFERENCES
1. Bartlett, M., M. Bird, and J. Howell, 1995. An experimental study for the
development of a qualitative membrane cleaning model. Journal of Membrane
Science, 105(1-2): p. 147-157.
2. Euromonitor. Soft Drinks: Euromonitor from trade sources/national statistics. 2014
(accessed on 29/06/14; Available from: http://www.portal.euromonitor.com
3. Euromonitor, 2009. Global soft drinks: has the world finally acquired a taste for RTD
tea? Euromonitor International
4. Todisco, S., P. Tallarico, and B. Gupta, 2002. Mass transfer and polyphenols retention
in the clarification of black tea with ceramic membranes. Innovative Food Science &
Emerging Technologies, 3(3): p. 255-262.
5. Liang, Y. and Y. Xu, 2001. Effect of pH on cream particle formation and solids
extraction yield of black tea. Food Chemistry, 74(2): p. 155-160.
6. Chandini, S.K., L.J. Rao, and R. Subramanian, 2013. Membrane Clarification of
Black Tea Extracts. Food and Bioprocess Technology, 6(8): p. 1926-1943.
7. Head, L.E. and M.R. Bird, 2013. The removal of psychrotropic spores from Milk
Protein Isolate feeds using tubular ceramic microfilters. Journal of Food Process
Engineering, 36(1): p. 113-124.
8. Jobstl, E., et al., 2005. Creaming in black tea. J. Agric. Food Chem, 53(20): p. 7997-
8002.
9. Charlton, A., et al., 2000. The self-association of the black tea polyphenol theaflavin
and its complexation with caffeine. Journal of the Chemical Society, Perkin
Transactions 2, 2000(2): p. 317-322.
10. Jöbstl, E., et al., 2004. Molecular model for astringency produced by
polyphenol/protein interactions. Biomacromolecules, 5(3): p. 942-949.
11. Tolstoguzov, V., 2002. Thermodynamic aspects of biopolymer functionality in
biological systems, foods, and beverages. Critical reviews in biotechnology, 22(2): p.
89-174.
12. Siebert, K.J., 1999. Effects of protein-polyphenol interactions on beverage haze,
stabilization, and analysis. Journal of agricultural and food chemistry, 47(2): p. 353-
362.
22
13. Liang, Y., J. Lu, and L. Zhang, 2002. Comparative study of cream in infusions of
black tea and green tea. International Journal of Food Science & Technology, 37(6):
p. 627-634.
14. Wu, D. and M. Bird, 2007. The fouling and cleaning of ultrafiltration membranes
during the filtration of model tea component solutions. Journal of Food Process
Engineering, 30(3): p. 293-323.
15. Evans, P.J. and M.R. Bird, 2006. Solute-Membrane Fouling Interactions During the
Ultrafiltration of Black Tea Liquor. Food and Bioproducts Processing, 84(4): p. 292-
301.
16. Evans, P.J., et al., 2008. The influence of hydrophobicity, roughness and charge upon
ultrafiltration membranes for black tea liquor clarification. Journal of Membrane
Science, 313(1-2): p. 250-262.
17. Shukla, R. and M. Cheryan, 2002. Performance of ultrafiltration membranes in
ethanol-water solutions: effect of membrane conditioning. Journal of Membrane
Science, 198(1): p. 75-85.
18. Kochan, J., et al., 2009. Impact of wetting agents on the filtration performance of
polymeric ultrafiltration membranes. Desalination, 241(1-3): p. 34-42.
19. Singleton, V. and J.A. Rossi, 1965. Colorimetry of total phenolics with
phosphomolybdic-phosphotungstic acid reagents. American journal of Enology and
Viticulture, 16(3): p. 144-158.
20. Nyström, M., A. Pihlajamäki, and N. Ehsani, 1994. Characterization of ultrafiltration
membranes by simultaneous streaming potential and flux measurements. Journal of
Membrane Science, 87(3): p. 245-256.
21. Scharbert, S., N. Holzmann, and T. Hofmann, 2004. Identification of the astringent
taste compounds in black tea infusions by combining instrumental analysis and
human bioresponse. Journal of agricultural and food chemistry, 52(11): p. 3498-3508.
23