Separation and Purification Technology 217 (2019) 147–153

Contents lists available at ScienceDirect

Separation and Purification Technology

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

Tuning the porosity of asymmetric membranes via simple post-synthesis T

solvent-treatment for non-aqueous applications
Yun Lia, Ines Nulensa, Rhea Verbekea, Hanne Mariëna, Tönjes Koschinea, Marcel Dickmannb,
⁎
Werner Eggerc, Ivo F.J. Vankelecoma,
a
Centre for Surface Chemistry and Catalysis, KU Leuven, Celestijnenlaan 200F, Box 2461, 3001 Heverlee, Belgium
b
Heinz Maier-Leibnitz Zentrum (MLZ) and Physik Department E21, Technische Universität München, Lichtenbergstraße 1, 85748 Garching, Germany
c
Institut für Angewandte Physik und Messtechnik, Universität der Bundeswehr München, 85577 Neubiberg, Germany

A R T I C LE I N FO A B S T R A C T

Keywords: Tuning the cohesive and swelling forces in porous membranes by post-synthesis solvent-treatment was already
Porous membrane proven to be a successful way to increase the membrane selectivity in aqueous redox flow battery application
Solvent treatment without sacrificing too much permeability of the membranes. This method was now extended to non-aqueous
Solvent evaporation applications and the mechanism behind the tuning of the pore sizes upon solvent-treatment was studied in more
Non-aqueous
detail. Porous polyvinylidene fluoride (PVDF) and polysulfone (PSF) membranes were thus prepared via phase
SRNF
inversion and subsequently treated with different solvents followed by the evaporation of this solvent to adjust
the structural porosities of these asymmetric membranes and better understand the underlying mechanism. For
this purpose, water and acetonitrile permeances were determined before and after the solvent-treatment and
these permeances were linked to the (solvent)-(membrane polymer) Hansen solubility parameters. The mem-
brane density changes were characterized by positron annihilation lifetime spectroscopy (PALS). It was proven
that the pore adjustment that was created by the solvent-treatment was reversible when the membrane was later
on applied in a feed solution with high affinity for the membrane polymer.

1. Introduction after the phase inversion process [9]. Since the phase inversion process
is a very fast process, the solidification fixes the chains in a certain non-
There is a long-standing interest in developing membranes with equilibrium state. Subsequent membrane swelling in these organic
highly interconnected, well-defined narrow pores which show both a solvents causes reorganization of the polymer chains and consequently
high permeability and selectivity to meet the needs of challenging se- alters membrane porosity and performance [10,11]. It was hypothe-
parations of small molecules and ions, like in nanofiltration and sec- sized that upon immersion into a solvent, the membrane pore walls are
ondary batteries [1,2]. Phase inversion is a commonly used process for subjected to a combination of cohesive and swelling forces. The cohe-
porous membrane preparation [3–6]. It is based on the controlled sive force is an intrinsic material property, while the swelling force is a
transition of a polymer solution from a liquid state into a porous solid result of the interaction between solvent and polymer. After re-
state. To induce this transformation, immersion precipitation, vapor arrangement of the polymer chains, the solvent is evaporated from the
phase precipitation or thermal-induced phase separation can be ap- membrane. The decrease in solvent content in the membrane pores
plied. By synthesizing membranes using the phase inversion method, causes the swelling force to decrease and the cohesive force to take
membranes with interconnected, porous top-layers can be obtained over. As a result, the pores shrink, but do not collapse. Due to the
[7,8]. However, as the final pores become smaller, the pore inter- solvent-dependent initial rearrangement of the polymer chains upon
connectivity in the top layer often decreases. swelling, the shrinkage can go beyond the pristine pore size. The con-
Recently, a breakthrough has been obtained in porous membrane served pore interconnectivity results in a high conductivity of ions
preparation for battery applications by tuning the cohesive and swelling when applying these modified membranes in aqueous flow battery
forces of swollen porous membranes, by the so-called ‘solvent-treat- applications, whereas the reduced pore sizes lead to high ion selectivity.
ment’ method which allows membranes to become equilibrated in a This method is interesting because of the low cost, the efficiency and
solvent, such as isopropanol, n-hexane, ethanol, water or methanol, simplicity compared to other post-synthesis treatments. Recently, it was

⁎
Corresponding author.
E-mail address: [email protected] (I.F.J. Vankelecom).

https://doi.org/10.1016/j.seppur.2019.02.012
Received 26 October 2018; Received in revised form 6 February 2019; Accepted 6 February 2019
Available online 07 February 2019
1383-5866/ © 2019 Elsevier B.V. All rights reserved.
Y. Li, et al. Separation and Purification Technology 217 (2019) 147–153

Fig. 1. Schematic diagram of porous, selective membranes with highly interconnected pores prepared via post-synthesis solvent-treatment.

found that different solvents from which to dry membranes have dif- room temperature. The dry coupons were stored dry, in water or in
ferent effects on the membrane performances. The drying of mem- acetonitrile, depending on the subsequent characterization technique
branes would result in a collapse of pores, or enhanced chemical in- used.
teractions between polymer chains [12,13]. The mechanism behind this
method of solvent treatment and its effect on membrane morphology 2.2. Membrane characterization
still need to be better understood.
This work intends to create better understanding of the exact me- 2.2.1. Scanning electron microscopy
chanism behind this effect of solvent evaporation on membrane mor- The cross-sections and the surfaces of the prepared membranes were
phology and to extend this solvent-treatment approach from aqueous to visualized using a scanning electron microscope (SEM, JEOL-JSM-
non-aqueous applications (Fig. 1). Membrane performance will be 6010LV). The cross-sections were obtained by breaking the membrane
tested in non-aqueous pressure-driven membrane separation systems samples in liquid nitrogen. Before analysis, all samples were coated
after solvent-treatments and the relationship between solvent-mem- with a thin (1.5–2 nm) Ag/Pd layer by using a JEOL JFC-1300 auto fine
brane affinity and membrane performance will be estimated. As one of coater to ensure conductance of the samples.
the most important solvents in solvent-resistant nanofiltration and non-
aqueous batteries [14–16], acetonitrile (ACN) was chosen to evaluate 2.2.2. Membrane porosity
the membranes. Polyvinylidene fluoride (PVDF) and polysulfone (PSF) The thicknesses the membrane coupons were measured with a spiral
were used to prepare porous membranes. micrometer. After that, the surfaces of the wet coupons were wiped to
remove excess water and the samples were weighed. Subsequently, the
2. Experimental coupons were dried at 100 °C for 24 h to remove all water. The dry
coupons were then weighed again to calculate the loss of water. Finally,
2.1. Membrane preparation the porosity (ε) of the membrane was calculated according to the fol-
lowing equation:
The porous membranes were prepared by phase inversion from a m w − md
solution containing polymer dissolved in N, N-dimethylacetamide ε (%) = × 100
ρ×S×l (1)
(DMAc). The casting solution compositions are listed in Table 1. Poly-
vinylpyrrolidone (PVP) with an average molecular weight of 10 kDa with mw the mass of the wet membrane coupon (g), md the mass of the
was used as additive for the porous PSF membrane preparation. dried membrane coupon (g), ρ the water density (g cm−3), S the surface
Polymer solutions were cast on a glass plate using an automatic film area (cm2) of the dried membrane and l the thickness (cm) of the dried
applicator (Braive Instruments, Luik, Belgium). Hereafter, the cast films membrane.
were immersed in a coagulation bath containing distilled water. The
casting knife had an opening thickness of 200 µm and the casting speed 2.2.3. Water contact angle measurements
was 0.021 m s−1. In this way, PVDF and PSF membranes with a Water contact angle measurements, with distilled water as probing
thickness of around 100 µm were obtained. liquid, were executed on dense PVDF and PSF membranes to obtain an
The membranes, stored in distilled water after synthesis, were cut in indication of their hydrophilicity. The drop shape analysis (Krüss, DSA
coupons. The coupons destined for a certain solvent evaporation were 10Mk2) was used.
immersed in that solvent for 24 h. The solvents were acetone, acet-
onitrile, distilled water, ethanol, ethyl acetate, isopropanol, methanol, 2.2.4. Membrane swelling
n-hexane and n-heptane. The coupons destined for n-hexane, n-heptane The swelling was obtained by first measuring the length of the dried
or ethyl acetate evaporation were first immersed in ethanol for 24 h membranes (dried at 100 °C for 24 h). The membranes were then im-
prior to their final solvent immersion to facilitate the displacement of mersed in corresponding solvents for 24 h. The membranes were taken
water. Afterwards, the coupons were placed in plastic weighing boats out and wiped to remove excess solvent, and the length was measured
and put in a hood for 24 h to allow full evaporation of the solvent at again. The swelling was calculated by the following formula:

Table 1
Membrane properties: polymer type, polymer concentration and PVP concentration in the DMAc-based casting solution, membrane porosity and membrane contact
angle with water.
Polymer Polymer concentration (wt.%) PVP concentration (wt.%) Porosity (%) Water contact angle (°)

PVDF, Solef® 1015 25 – 57 90

PSF, Udel® P-1835 25 9.7 78 60

148
Y. Li, et al. Separation and Purification Technology 217 (2019) 147–153

Fig. 2. SEM images of membranes prepared by phase inversion: (a) and (b) cross-sections of the porous PVDF membrane at different magnifications (scale bar: 20 μm
and 1 μm); (c) and (d) cross-sections of the porous PSF membrane at different magnifications (scale bar: 20 μm and 1 μm).

l w − ld million counts were collected with a count rate of about 10,000 cts/s
SR (%) = × 100
ld (2) and a time resolution of about 250 ps. The resolution function was
determined by measuring p-doped SiC, a sample with known lifetimes.
with lw the length (cm) of the wet membrane sample and ld the length
All spectra were evaluated with POSWIN and deconvoluted into four
(cm) of the dried membrane sample.
lifetime components (p-Ps, free e+ and two o-Ps) leading to good
overall fit variances from 1.0 to 1.15 [26]. To prepare the samples, the
2.2.5. Positron annihilation lifetime spectroscopy (PALS) solvents were evaporated from the membranes and the resulting dry
PALS is a method used to determine the free volume element size in coupons were cut in pieces of 1.5 × 1.5 cm2 and measured.
polymer membranes [17–20]. After implantation, the positron ther-
malizes rapidly and can form a hydrogen-like bound state with an
2.2.6. Filtration experiments
electron, the so-called positronium (Ps). It occurs either as para-posi-
Permeances were determined with water or ACN using a dead-end
tronium (p-Ps) in a spin-singlet state or as ortho-positronium (o-Ps) in a
high-throughput filtration system. Three samples of the same mem-
spin-triplet state. The positron of the o-Ps interacts with an electron of
branes were tested to determine the average permeance. The per-
its surroundings and annihilates into two gamma-quanta with an
meance (J, L m−2 h−1 bar−1) was calculated as follows:
average lifetime, the pick-off lifetime, which can be measured and di-
rectly correlated to the free volume hole size according to the Tao- V
J=
Eldrup model [21,22]. This model, assuming spherical holes with in- S × Δp × Δt (4)
finite potential walls, gives a quantum mechanical description of the
correlation of free volume radius and o-Ps lifetime: where V is the solvent volume (L) that permeated over a time period (Δt
in h), S is the effective area of measured membrane (0.18 × 10−4 m2),
τ3 = 1/2 ⎡1 −
R
+
1
sin ⎛
2πR ⎞ ⎤
ns Δp the pressure applied in the dead-end high-throughput filtration cells
⎢
⎣ R + ΔR 2π ⎝ R + ΔR ⎠ ⎥
⎦ (3) (bar). Permeances were normalized for viscosity. The viscosity of water
at 20 °C is equal to 1.00 mPa s or 2.78 × 10−12 bar h. The viscosity of
In this equation τ3 is the o-Ps kick-off lifetime, R the radius of the
ACN at 20 °C is equal to 0.39 mPa s or 1.08 × 10−12 bar h.
free volume hole and ΔR the thickness of the electron layer surrounding
the free volume hole, which is assumed to be 0.166 nm. This model is
valid for lifetimes up to 10 ns, but it can be easily extended to longer 2.3. Calculation of membrane and solvent
lifetimes and thus to larger free-volume element sizes [23]. The in-
tensity of the o-Ps is a product of Ps formation probability and the The affinity of two species can be quantified using the Hansen so-
concentration of free volume holes. Therefore, it can be used in some lubility parameters (HSPs) [6]. The HSPs exist of three parameters
cases to estimate relative changes in the hole concentration. identifying the strength of dispersion interactions (δD ) , polar interac-
All membranes were measured with the pulsed low energy positron tions (δP ) and hydrogen bonding interactions (δH ) . These three para-
beam system (PLEPS), which is operated at the neutron induced posi- meters are summarized in a single solubility parameter (δt ) .
tron source Munich (NEPOMUC) [24,25]. An implantation energy of δt = δD 2 + δP 2 + δH 2 (5)
1 keV was used, coinciding with a penetration depth of ca. 40 nm (when
a polymer density of about 1 g cm−3 is assumed). For each spectra, 4 The more the interactions of two species are alike, the smaller the

149
Y. Li, et al. Separation and Purification Technology 217 (2019) 147–153

difference between solubility parameters will be. This can be quanti-

tatively summarized into one parameter, the solubility parameter
“distance” Ra, calculated as:

Ra = 4(δD1 − δD2)2 + (δP1 − δP 2)2 + (δH 1 − δH 2)2 (6)

1/2
where Ra, δD , δP and δH are expressed in MPa .

3. Results and discussion

3.1. Membrane morphology

Fig. 2a and b show the cross-sections of porous PVDF membranes.

Three regions can be observed: a thin top-layer, a region with macro-
voids and a spongier sub-layer. The thickness of the top-layer cross-
section was estimated to be 0.25 μm. When analyzing the SEM image of
the cross-section of the porous PSF membrane (Fig. 2c and d), a thick
dense top-layer of about 2.0 μm can be seen on top of macrovoids in the
sublayer, which now reach the membrane bottom side. These ob-
servations coincide with the membranes prepared from PSF/PVP/
Fig. 3. The swelling of the PVDF membrane in different solvents, plotted as a
DMAc systems reported earlier [7–9].
function of the solvent-PVDF Ra-value.

3.2. Solvent-treated PVDF membranes

The Ra of PVDF was calculated in different solvents based on HSPs

(Table 2) and subsequently compared to experimental values of swel-
ling (Fig. 3). Many solvents were used to make sure swelling occurred in
a wide range (from 2% to 10%). Additionally, small Ra values resulted
in the highest degree of swelling, as anticipated from theory [27].
It was clear that the solvent-treated membranes generally showed a
smaller water permeance than the pristine ones (Fig. 4a). The nor-
malized water permeance for the pristine PVDF membrane was
1.21 ± 0.28 × 10−12 m. The solvent evaporation step thus indeed
caused the pores to shrink significantly, as the permeance decreased
minimally with 77%. Fig. 4b shows the normalized acetonitrile (ACN)
permeances of pristine and solvent-treated porous PVDF membranes.
For the untreated membrane, the ACN permeance (5.85 × 10−11 m)
was almost 50 times higher than the water permeance
(1.21 × 10−12 m). The higher ACN permeance as compared to the
water permeance is consistent with previous research which relates
solvent flux to solvent properties, like viscosity and surface tension.
Indeed, both viscosity and surface tension are lower for ACN than for
water. In addition, the higher ACN permeance is also partly attributed
to the interaction between solvent and polymer [31–33].
Surprisingly, no clear difference was seen here between the un-
treated and the treated membranes with respect to their ACN pear-
meances, which is in clear contrast with the results from the water
permeance. The average water permeance of the different solvent-
treated membranes was 0.17 × 10−12 m. In contrast, the average value
of the ACN permeance of the solvent-treated membranes was
5.38 × 10−11 m, i.e. 300 times the average water permeance. This

Table 2 Fig. 4. Normalized (a) water and (b) acetonitrile permeances of the pristine
The HSPs of solvents, and their Ra-values with PVDF and PSF (MPa1/2) [27–30]. porous PVDF membrane and of that same membrane after treatment in dif-
ferent solvents (the solvents ranked as a function of decreasing Ra PVDF).
δd δp δh δt Ra PVDF Ra PSF

Acetone 15.5 10.4 7.0 19.9 5.3 8.8 indicates that ACN either was able to restore the pore sizes more or less
Acetonitrile (ACN) 15.3 18.0 6.1 24.4 7.9 13.3 back to their original state, or that ACN was able to travel through the
Ethyl acetate 15.8 5.3 7.2 18.2 8.5 8.4 dense membrane matrix by extensively solvating the polymer chains
Isopropanol (IPA) 15.8 6.1 16.4 23.6 9.1 11.5
Ethanol (EtOH) 15.8 8.8 19.4 26.5 9.9 13.6
and creating additional pathways for permeation. To investigate whe-
Methanol 15.1 12.3 22.3 29.6 12.4 17.2 ther ACN permeation through the dense PVDF matrix was possible,
n-Heptane 15.3 0.0 0.0 15.3 16.9 14.7 fully-dense PVDF membranes with a thickness of 40 µm were prepared
n-Hexane 14.1 0.0 0.0 14.1 17.5 16.2 by mere solvent evaporation of a cast PVDF-solution. Four samples
Water 15.6 16.0 42.8 48.3 32.5 36.3
were tested and the ACN permeances were below 0.6 × 10−13 m. This
PVDF 17.1 12.6 10.6 23.7 0.0 –
PSF 19.7 8.3 8.3 22.9 – 0.0 would correspond to a maximum permeance of 0.96 × 10−11 m
through the matrix of PVDF membrane, assuming a 250 nm thick

150
Y. Li, et al. Separation and Purification Technology 217 (2019) 147–153

Fig. 5. Free-volume element radius and intensity of short-living o-Ps (a and b) and long-living o-Ps (c and d) in the PVDF membrane.

selective layer. This is 5 times less than what was experimentally ob- component is significantly lower. Since the chemical environment of
tained, suggesting that ACN cannot travel though the dense membrane. the free-volume is expected to be similar between these samples, the Ps-
In order to check the hypothesis of the restoration of the original formation probability will most probably not differ much. Therefore,
pore size when contacted with ACN, and thus to investigate whether differences in intensities may carefully be related to differences in
immersion of the solvent-treated membrane in ACN eliminates the ef- porosities. It can hence be concluded that the IPA-treated membrane
fect of prior solvent-treatment, PALS measurements were performed on has more small pores, which are more or less the same in size as the
membranes in two sets of experiments. In a first set, membranes treated IPA- and EtOH-treated membranes, while its bigger pores are much
with water, ethanol or isopropanol and subsequently evaporated, were bigger in size but less in concentration. Additionally, the smallest Ra
characterized. In a second set of experiments, the membranes that were was also observed for the PVDF/IPA pair, suggesting that, indeed, PVDF
solvent-treated with water, hexane or acetonitrile, were subsequently will interact more strongly with IPA than with the other solvents. This
immersed in acetonitrile. The samples of the latter set of experiments difference in pore size, porosity and Ra-values between the 3 samples is
are named H2O_ACN, HEX_ACN and ACN_ACN, according to the sol- however not translated into the macroscopic transport properties of the
vents used for the solvent-treatment prior to the ACN immersion. membranes, as their water permeances are not significantly different
Two o-Ps species with lifetimes of ca. 2.3 ns and 4 ns were observed (Fig. 4a). For the membranes that had undergone ACN immersion after
in each sample and further referred to as short and long-living o-Ps, the solvent-treatment, both free-volume element radii and intensities
respectively. The free-volume element radii and intensities for both the are much more similar to each other (Fig. 5), in agreement with their
short-living o-Ps and the long-living o-Ps are plotted in Fig. 5. The comparable ACN permeances (Fig. 4b). The differences observed be-
short-living o-Ps corresponds to pores of around 0.7 nm and the long- tween H2O_ACN, n-HEX_ACN and ACN_ACN are relatively small, in-
living o-Ps to pores of around 3–4 nm in diameter. It can thus be con- dicating that the final immersion of the solvent-treated membranes in
cluded that the membranes have a bimodal pore size distribution. In ACN overrules the influence on pore size and porosity of the former
contrast to the short-living component, the long-living component solvent-treatment. The changes in membrane pore size are thus re-
showed a pronounced difference in free-volume element radii between versible. This can possibly be explained by the Hansen solubility
the membranes, when no ACN immersion was performed. Especially parameters (Table 2). The low Ra value of 7.9 suggests a high affinity
the IPA-treated membrane is significantly different, with its big pores between PVDF and ACN. Therefore, during the contact with ACN, there
having twice the size compared to those of the water and ethanol- is a much higher swelling than cohesive force on the pore walls. Hence,
treated membranes. When investigating the o-Ps intensity, the IPA- the membrane swells again significantly to a level that is only de-
treated membrane again largely differs from the others. Its intensity of termined by the final ACN treatment, i.e. without any link to the in-
the short-living component is much bigger, while that of the long-living termediate solvent-treatment. Hence, all membranes show similar ACN

151
Y. Li, et al. Separation and Purification Technology 217 (2019) 147–153

for ACN and water, compared to PVDF. The Ra values of PSF with the
different solvents now vary in a relatively narrow range (Table 2), so it
is reasonable to find all experimentally determined swells of PSF
membranes in the range of 3–6% (Fig. 6). Ethyl acetate and acetone
showed an extremely high affinity for PSF (Ra below 9) and the mem-
branes got visually damaged, not allowing exact determination of the
swelling.
The water and ACN permeances of treated PSF membranes are
shown in Fig. 7, with the solvents ranked as a function of decreasing Ra.
It can be seen that all solvent-treated membranes show a significantly
lower water permeance, compared to the pristine PSF membrane
(1.68 ± 0.13 × 10−10 m). Between the different solvent-treated
membranes, distinct permeance differences are also observed. The
trend, however, did not coincide with the expectations based on HSP. It
was expected that a solvent with a lower Ra, hence with more beneficial
interactions between the solvent and the polymer, would induce a
higher impact on the pore size of the membrane upon immersion and
evaporation, thus resulting in smaller pores and a lower permeance.
For the PSF system, the trend for the ACN permeance was more or
less the same as the trend seen for water permeance. This indicates that
Fig. 6. Swelling (%) of a porous PSF membrane in different solvents, plotted
according to their Ra-values (see Table 2). the re-swelling in ACN during filtration still preserves the modified pore
morphology that was created during the preceding solvent-treatment,
as opposed to PVDF. This can be explained by the lower affinity of PSF
for ACN (Ra of 13.3), compared to PVDF (Ra of 7.9). Furthermore, the
ACN permeance through the pristine PSF membranes was more than 4
times higher than the water permeance. This is agreement with the
differences in Ra between PSF/water and PSF/ACN and consistent with
the reported PVDF filtration results.

4. Conclusions

The pore sizes of membranes can be tuned by solvent-treatment, i.e.

immersion in a solvent followed by the evaporation of this solvent, but
the final performance of the solvent-treated membranes highly depends
on the solvents in which membranes are finally used. Water permeances
of porous PVDF and PSF membrane varied after solvent-treatment as a
function of the solvent type, indicating that there are pore size changes
and polymer chain rearrangements. In non-aqueous systems, the acet-
onitrile flux of porous PVDF did not change at all after solvent-treat-
ment. PALS measurements showed that pore sizes changed after the
solvent-treatment, but that afterwards, when filtering in acetonitrile,
the initial influence of different solvents on the membranes pore size
was offset by re-swelling in ACN. The changes in membrane pore size
were thus found to be reversible. In contrast, porous PSF membranes
did show different ACN permeances after the treatments in different
solvents. This was attributed to the lower ACN-polymer affinity i.e.
(Ra = 7.9) for PVDF and (Ra = 13.3) for PSF.
The solvent-treatment method can thus be used in non-aqueous
solvents to modify pore structures, but the affinity between the mem-
brane polymers and the solvent in which the membrane will be used
can erase the treatment when solvents and polymers interact strongly
during filtration.
Fig. 7. Normalized water and ACN permeances of pristine and solvent-treated
porous PSF membranes. The solvents on the x-axis are ranked as a function of Acknowledgements
decreasing Ra PSF.
The authors gratefully acknowledge Belgian government (IAP-PAI
permeances and similar pore sizes. In water, on the contrary, the networking) and KU Leuven (OT/11/061 project). Yun Li acknowledges
swelling force is not strong enough to change the pores due to the low the China Scholarship Council for financial support. R.V. thanks
affinity of water for PVDF (Ra equals 32.5). Research Foundation-Flanders (FWO) for a SB-PhD fellowship
(1S00917N). This work is based upon experiments performed at the
NEPOMUC PLEPS instrument operated by FRM-II at the Heinz Maier-
3.3. PSF membranes Leibnitz Zentrum (MLZ), Garching, Germany.

In order to verify a material dependent effect of the solvent-treat- References

ment, porous PSF membranes were also prepared. As suggested by the
Hansen solubility parameters, PSF is supposed to have a lower affinity [1] H. Zhang, H. Zhang, X. Li, Z. Mai, J. Zhang, Nanofiltration (NF) membranes: the

152
Y. Li, et al. Separation and Purification Technology 217 (2019) 147–153

next generation separators for all vanadium redox flow batteries (VRBs), Energy I.S. Kim, Characterization of pore size distribution (PSD) in cellulose triacetate
Environ. Sci. 4 (2011) 1676–1679. (CTA) and polyamide (PA) thin active layers by positron annihilation lifetime
[2] W. Lu, Z. Yuan, Y. Zhao, H. Zhang, H. Zhang, X. Li, Porous membranes in secondary spectroscopy (PALS) and fractional rejection (FR) method, J. Membr. Sci. 527
battery technologies, Chem. Soc. Rev. 46 (2017) 2199–2236. (2017) 143–151.
[3] P. Marchetti, M.F. Jimenez Solomon, G. Szekely, A.G. Livingston, Molecular se- [19] T. Shintani, A. Shimazu, S. Yahagi, H. Matsuyama, Characterization of methyl-
paration with organic solvent nanofiltration: a critical review, Chem. Rev. 114 substituted polyamides used for reverse osmosis membranes by positron annihila-
(2014) 10735–10806. tion lifetime spectroscopy and MD simulation, J. Appl. Polym. Sci. 113 (2009)
[4] J.R. Werber, C.O. Osuji, M. Elimelech, Materials for next-generation desalination 1757–1762.
and water purification membranes, Nat. Rev. Mater. 1 (2016) 16018. [20] A. Cano-Odena, P. Vandezande, K. Hendrix, R. Zaman, K. Mostafa, W. Egger,
[5] J. De Jong, R.G.H. Lammertink, M. Wessling, Membranes and microfluidics: a re- P. Sperr, J.D. Baerdemaeker, I.F.J. Vankelecom, Probing the molecular level of
view, Lab Chip 6 (2006) 1125–1139. polyimide-based solvent resistant nanofiltration membranes with positron annihi-
[6] P. Vandezande, X. Li, L.E.M. Gevers, I.F.J. Vankelecom, High throughput study of lation spectroscopy, J. Phys. Chem. B 113 (2009) 10170–10176.
phase inversion parameters for polyimide-based SRNF membranes, J. Membr. Sci. [21] M. Eldrup, D. Lightbody, J.N. Sherwood, The temperature dependence of positron
330 (2009) 307–318. lifetimes in solid pivalic acid, Chem. Phys. 63 (1981) 51–58.
[7] A.K. Hołda, B. Aernouts, W. Saeys, I.F.J. Vankelecom, Study of polymer con- [22] S. Tao, Positronium annihilation in molecular substances, J. Chem. Phys. 56 (1972)
centration and evaporation time as phase inversion parameters for polysulfone- 5499–5510.
based SRNF membranes, J. Membr. Sci. 442 (2013) 196–205. [23] T.L. Dull, W.E. Frieze, D.W. Gidley, J.N. Sun, A.F. Yee, Determination of pore size in
[8] A.K. Hołda, M. De Roeck, K. Hendrix, I.F.J. Vankelecom, The influence of polymer mesoporous thin films from the annihilation lifetime of positronium, J. Phys. Chem.
purity and molecular weight on the synthesis of integrally skinned polysulfone B 105 (2001) 4657–4662.
membranes, J. Membr. Sci. 446 (2013) 113–120. [24] C. Hugenschmidt, C. Piochacz, M. Reiner, K. Schreckenbach, The NEPOMUC up-
[9] W. Lu, Z. Yuan, Y. Zhao, X. Li, H. Zhang, I.F.J. Vankelecom, High-performance grade and advanced positron beam experiments, New J. Phys. 14 (2012) 055027.
porous uncharged membranes for vanadium flow battery applications created by [25] C. Hugenschmidt, G. Dollinger, W. Egger, G. Kögel, B. Löwe, J. Mayer, P. Pikart,
tuning cohesive and swelling forces, Energy Environ. Sci. 9 (2016) 2319–2325. C. Piochacz, R. Repper, K. Schreckenbach, Surface and bulk investigations at the
[10] F.M. Penha, K. Rezzadori, M.C. Proner, V. Zanatta, G. Zin, D.W. Tondo, J. Vladimir high intensity positron beam facility NEPOMUC, Appl. Surf. Sci. 255 (2008) 29–32.
de Oliveira, J.C.C. Petrus, M. Di Luccio, Influence of different solvent and time of [26] D. Bochert, Optimierte Analyse von Lebenszeitspektren aus
pre-treatment on commercial polymeric ultrafiltration membranes applied to non- Positronenstrahlsystemen, Master’s thesis Universität der Bundeswehr München,
aqueous solvent permeation, Eur. Polym. J. 66 (2015) 492–501. 2004.
[11] H. Mariën, I.F.J. Vankelecom, Transformation of cross-linked polyimide UF mem- [27] C.M. Hansen, Hansen Solubility Parameters: A User's Handbook, CRC Press, Boca
branes into highly permeable SRNF membranes via solvent annealing, J. Membr. Raton, FL, 2007.
Sci. 541 (2017) 205–213. [28] M.R.M. Abed, S.C. Kumbharkar, A.M. Groth, K. Li, Ultrafiltration PVDF hollow fibre
[12] J. da Silva Burgal, L. Peeva, P. Marchetti, A. Livingston, Controlling molecular membranes with interconnected bicontinuous structures produced via a single-step
weight cut-off of PEEK nanofiltration membranes using a drying method, J. Membr. phase inversion technique, J. Membr. Sci. 407–408 (2012) 145–154.
Sci. 493 (2015) 524–538. [29] J. Ma, R.M. Larsen, Use of Hansen solubility parameters to predict dispersion and
[13] M. Mertens, C. Van Goethem, M. Thijs, G. Koeckelberghs, I.F.J. Vankelecom, strain transfer of functionalized single-walled carbon nanotubes in poly(vinylidene
Crosslinked PVDF-membranes for solvent resistant nanofiltration, J. Membr. Sci. fluoride) composites, J. Thermoplast. Compos. Mater. 27 (2014) 801–815.
566 (2018) 223–230. [30] T. Xiao, P. Wang, X. Yang, X. Cai, J. Lu, Fabrication and characterization of novel
[14] P. Vandezande, L.E.M. Gevers, I.F.J. Vankelecom, Solvent resistant nanofiltration: asymmetric polyvinylidene fluoride (PVDF) membranes by the nonsolvent ther-
separating on a molecular level, Chem. Soc. Rev. 37 (2008) 365–405. mally induced phase separation (NTIPS) method for membrane distillation appli-
[15] S.-H. Shin, S.-H. Yun, S.-H. Moon, A review of current developments in non-aqueous cations, J. Membr. Sci. 489 (2015) 160–174.
redox flow batteries: characterization of their membranes for design perspective, [31] M. Bastin, K. Hendrix, I. Vankelecom, Solvent resistant nanofiltration for acetoni-
RSC Adv. 3 (2013) 9095–9116. trile based feeds: a membrane screening, J. Membr. Sci. 536 (2017) 176–185.
[16] K. Gong, Q. Fang, S. Gu, S.F.Y. Li, Y. Yan, Nonaqueous redox-flow batteries: organic [32] D.O.R. Machado, D. Hasson, R. Semiat, Effect of solvent properties on permeate
solvents, supporting electrolytes, and redox pairs, Energy Environ. Sci. 8 (2015) flow through nanofiltration membranes: Part II. Transport model, J. Membr. Sci.
3515–3530. 166 (2000) 63–69.
[17] Y. Htwe Htwe, Z. Yin, W. Ma, D. Zhu, A review of studies of polymeric membranes [33] D.O.R. Machado, D. Hasson, R. Semiat, Effect of solvent properties on permeate
by positron annihilation lifetime spectroscopy, Plasma Sci. Technol. 7 (2005) 3062. flow through nanofiltration membranes. Part I: investigation of parameters af-
[18] S.-J. Kim, S. Kook, B.E. O'Rourke, J. Lee, M. Hwang, Y. Kobayashi, R. Suzuki, fecting solvent flux, J. Membr. Sci. 163 (1999) 93–102.

153