Introduction of an ordered porous polymer network into a ceramic alumina membrane via non-hydrolytic sol-gel methodology for targeted dynamic separation

Article abstract of DOI:10.1039/c4ra06166g

A highly selective composite imprinted alumina membrane (CIAM) for gentisic acid (GA) was successfully synthesized by the nonhydrolytic sol-gel (NHSG) method with room temperature ionic liquid (RTIL) as the pore template. The carboxylic acid was used as both the functional monomer and the catalyst to form the alumina membrane-based porous imprinted polymer layer. The adsorption capacities, fluxes and permeation selectivities of various CIAMs suggested that cinnamic acid (CA) was the promising functional monomer for preparing a CIAM to separate GA from salicylic acid (SA) and the incorporation of RTIL improved the selectivity of GA over CIAM. The amount of porous imprinted polymer layer on the CIAM significantly affected the separation of GA from SA over CIAM. A three-level Box-Behnken experimental design with three factors combining the response surface modeling was used to optimize the dynamic separation process. The experimental data were well fitted to a second-order polynomial equation using multiple regression analysis. The optimal conditions for the separation of GA from SA were as follows: GA concentration of 0.0325 mmol L^-1, temperature of 15.0 °C and flow rate of 1.0 mL min^-1. Under these conditions, the experimental separation factor was 13.26 ± 0.87%, which was close to the predicted separation factor. the Partner Organisations 2014.

Full text of DOI:10.1039/c4ra06166g

RSC Advances
View Article Online
View Journal | View Issue
PAPER
Introduction of an ordered porous polymer
network into a ceramic alumina membrane via non-
hydrolytic sol–gel methodology for targeted
dynamic separation
Cite this: RSC Adv., 2014, 4, 38630
a
a
a
b
a
Minjia Meng, Yan Liu, Min Zhang, Yonghai Feng and Yongsheng Yan*
A highly selective composite imprinted alumina membrane (CIAM) for gentisic acid (GA) was successfully
synthesized by the nonhydrolytic sol–gel (NHSG) method with room temperature ionic liquid (RTIL) as
the pore template. The carboxylic acid was used as both the functional monomer and the catalyst to
form the alumina membrane-based porous imprinted polymer layer. The adsorption capacities, fluxes
and permeation selectivities of various CIAMs suggested that cinnamic acid (CA) was the promising
functional monomer for preparing a CIAM to separate GA from salicylic acid (SA) and the incorporation
of RTIL improved the selectivity of GA over CIAM. The amount of porous imprinted polymer layer on the
CIAM significantly affected the separation of GA from SA over CIAM. A three-level Box–Behnken
experimental design with three factors combining the response surface modeling was used to optimize
the dynamic separation process. The experimental data were well fitted to a second-order polynomial
equation using multiple regression analysis. The optimal conditions for the separation of GA from SA
Received 24th June 2014
Accepted 14th August 2014
ꢀ1
ꢁ
were as follows: GA concentration of 0.0325 mmol L , temperature of 15.0 C and flow rate of 1.0 mL
minꢀ1. Under these conditions, the experimental separation factor was 13.26 ꢂ 0.87%, which was close
to the predicted separation factor.
DOI: 10.1039/c4ra06166g
www.rsc.org/advances
commercial membranes. Therefore, methods of preparing
composite membranes with controlled specicity for selective
1
Introduction
Salicylic acid (SA) has been used worldwide in many industrial separation of analogues have attracted considerable attentions.
applications, including for ointments, liquids, creams or plas-
Molecular imprinting is a good method for preparing poly-
6
ters, for the treatment of acne, psoriasis, warts, ichthyosis and mers with specic recognition for template molecule, which
1
other hyperkeratotic disorders. It is also used as a raw material are called the molecularly imprinted polymers (MIPs) and
2
7
8
for the production of aspirin. Commercial SA is usually widely used in chromatography, solid phase extraction, chiral
13,14
9,10
11,12
produced along with a mixture of isomers and analogues, which separations,
synthesis and catalysis,
enzyme mimics
are usually toxic and harmful to life. From among these, gentisic and so on. Using the molecular imprinting technique to prepare
acid (GA) is usually generated as an SA analogue in the produc- molecularly imprinted membrane has attracted great attention
tion of SA. The separation of GA from SA is hardly achieved by from researchers because it suggested an innovative and facile
conventional separation methods due to their similar structures. way to synthesize composite membrane for selective separation
15
Therefore, it is necessary to develop efficient methods to separate of analogues. Yohichi Tagawa et al. have investigated poly-
GA from SA for ensuring the high purity of SA.
meric membranes in detail for chiral separation of pharma-
Membrane separation is an emerging technology with ceuticals and chemicals. The pioneer work demonstrated that
advantages of low energy costs, low environmental impact, and the use of imprinting membranes is of great importance for
possible integrated processes with selective removal of certain chiral separation. However, the recognition sites and adsorp-
3,4
components. However, usual commercial membranes do not tion capacity of the polymeric membranes were relative low due
5
allow the selective separation of analogues, implying that the to the shortcomings of the preparation method. In these cases,
separation of analogues is hard to achieve by using the improving the recognition sites and adsorption capacity of the
imprinting membranes for selective separation of analogues are
still a challenge.
a
School of Chemistry and Chemical Engineering, Jiangsu University, Zhenjiang
Recently, the sol–gel imprinting method has been thought to
an alternative method for preparing MIPs, through which
organic/inorganic materials with unique structure and highly
2
b
12013, China. E-mail: mmjybyq@aliyun.com
School of Material Science and Engineering, Jiangsu University, Zhenjiang 212013,
China
38630 | RSC Adv., 2014, 4, 38630–38642
This journal is © The Royal Society of Chemistry 2014

View Article Online
Paper
RSC Advances
16–18
selective recognition could be formed.
This methodology is acidic aqueous layer was decanted and the lower ionic liquid
based on the non-hydrolytic sol–gel (NHSG) process and the portion was washed with water until the washings were no
molecular imprinting technique. The NHSG technique does not longer acidic. The ionic liquid ([BMIM]PF ) was then heated
6
ꢁ
require aging and drying steps at high temperatures. Since little under vacuum at 70 C to remove any excess water.
or no water is evolved in the synthesis, the cracking and
2.2.2 Preparation of CIAM. In this research, varied
shrinking of the gelatin by the hydrolytic sol–gel method are composite imprinted alumina membranes (CIAM) were
avoided. Thus, NHSG process can subtly retain binding sites synthesized. The compositions are presented in Table 1.
19
and increase the selectivity of the imprinted material.
2 3
At rst, the surfaces of Al O microporous ceramic
Room temperature ionic liquid (RTIL) as the interesting membranes were boiled in 30% hydrogen peroxide for 30 min
solvents has been widely used in organic synthesis, catalysis, to introduce –OH groups on the surface for further modica-
20–22
electrochemical purposes, etc.
template in the sol–gel reaction,
synthesis and improved the selectivity of imprinted polymers.
RTIL is also a promising pore tion. They were then boiled in deionized water for 15 min to
23–25
which accelerated the clean the surface and then dried under nitrogen condition.
Secondly, GA (1.0 mmol), functional monomer (MAA, AA or
In the present work, a novel way was developed for selective CA) (4.0 mmol) and acetonitrile (12.0 mL) were mixed in a
separation of GA from SA. Three kinds of composite imprinted conical ask for 4 h in order to obtain the self-assembled
alumina membranes (CIAM) for GA were prepared by NHSG composites with GA. Then MPS, [BMIM]PF and AIBN were
6
imprinted method with RTIL used as the pore template. The as- added to the solution forming the prepolymerization mixture
prepared CIAM were characterized by Fourier transform (see Table 1). Then several pretreated Al O microporous
2
3
infrared spectrometer (FT-IR), transmission electron micro- ceramic membranes were immersed into the prepolymerization
scope (SEM) and atomic force microscope (AFM) techniques. solution and the prepolymerization mixture was submerged
ꢁ
The as-prepared CIAM showed highly selective recognition of into a 54 C water bath for 12 h.
GA. A three-level Box–Behnken experimental design with three Finally, the imprinted membranes were moved out of the
factors combining the response surface modeling was used for water bath and immediately ushed with several volumes of
Optimization of dynamic separation of GA from SA.
mixed solvents of methanol and acetic acid (9 : 1, v/v) by ultra-
sound to remove the unreacted reagents, the RTIL and the
imprinting molecule GA. Then the composite imprinted
alumina membrane (CIAM) for GA was obtained by rinsing with
methanol and drying and the specic recognition sites for GA
were le aer removing the template (Scheme 1).
2
Experimental
2.1. Materials
Al
2
O
3
microporous membranes with a nominal pore size d
p
¼ 4
The composite non-imprinted alumina membrane (CNAM)
was prepared in parallel without addition of GA.
mm and a thickness of 2 mm, were purchased from Hefei Great
Wall Xinyuan lm Technology Co., Ltd. Salicylic acid (SA),
gentisic acid (GA), methacrylic acid (MAA), acrylic acid (AA),
cinnamic acid (CA), acetonitrile, methacryloxypropyltrimethox-
2.3. Characterization
0
ysilane (MPS), 2,2 -azobis(2-isobutyronitrile) (AIBN), 1-methyl-
The prepared RTIL was characterized by FT-IR spectra (Nicolet
NEXUS-470, USA). The morphologies of surface and cross-
sectional structures of the samples were obtained by scanning
electron microscopy (SEM, S-4800). Topographical atomic force
microscope (AFM) images of unmodied and imprinted
alumina membranes were made in tapping mode (Veeco
Instruments Inc.) with an atomic force microscope (MicroNano
AFM-III).
imidazole, chlorobutane and hexauorophosphoric acid were
all purchased from Sinopharm Chemical Reagent (Shanghai,
China). All other chemicals used were of analytical grade and
obtained commercially. Ultra pure water used throughout the
experiments was obtained from laboratory purication system.
2.2. Synthesis of composite imprinted alumina membrane
(
CIAM)
2.2.1. Preparation of ionic liquid (RTIL). 1-Butyl-3-methyl-
imidazolium chloride ([BMIM]Cl) was prepared according to 2.4. Batch selective rebinding assay
26
the report by Rogers et al. In the typical experiment, 61.58 g of
-methylimidazole (0.75 mol) and 69.43 g of chlorobutane
A piece of composite imprinted membrane (or a piece of blank
membrane) was added into conical ask, each of which con-
tained 20 mL solution of GA and SA, respectively. Then it was
1
(

0.75 mol) were added in a 250 mL of round-bottomed ask
tted with a reux condenser by heating and stirring at 70 C for
ꢁ
ꢁ
shocked at 25 C for 4 h in a water bath oscillator. Then,
72 h. The resulting viscous liquid ([BMIM]Cl) was allowed to
substrate concentration (GA and SA) in the solution was
analyzed using UV spectrophotometry and the binding amount
of each substrate can be calculated by the following equation:
cool to room temperature and then was washed three times with
2
00 mL portions of ethyl acetate. Aer the last washing, the
ꢁ
remaining ethyl acetate was removed by heating to 70 C under
ðC
0
ꢀ C ÞV
e
vacuum. To prepare the ionic liquid, hexauorophosphoric acid
q
e
¼
(1)
W
(94.88 g, 0.65 mol) was added (slowly to prevent the temperature
ꢀ
1
ꢀ1
from rising signicantly) to a mixture of [BMIM]Cl (87.34 g, where C
0
(mmol L ) and C
e
(mmol L ) represent the
1
mol) in 500 mL of water. Aer stirring for 12 h, the upper concentration of substrates measured at an initial and
This journal is © The Royal Society of Chemistry 2014
RSC Adv., 2014, 4, 38630–38642 | 38631

View Article Online
RSC Advances
Paper
a
Table 1 Protocol of prepolymerization solutions
b
Membranes
SA (mmol)
Carboxylic acid
MPS (mmol)
RTIL (mL)
AIBN (mmol)
AA–CIAM
1
0
1
0
1
1
0
1
1
AA
AA
MAA
MAA
CA
CA
CA
CA
CA
8
8
8
8
4
8
8
16
8
2
2
2
2
2
2
2
2
0
0.12
0.12
0.12
0.12
0.09
0.12
0.12
0.2
AA–CNAM
MAA–CIAM
MAA–CNAM
CA–CIAM1
CA–CIAM2
CA–CNAM2
CA–CIAM3
CA–CIAM0
0.12
a
b
ACN was added to ensure a total volume of 12 mL. The amount of three carboxylic acid (AA, MAA and CA) are the same 4 mmol according to the
results in Fig. 1.
saturated binding time. V (L) and W (g) are the volume of the (100 mL) in methanol was placed in the le-hand side chamber,
solution and the weight of the imprinted membrane, while 100 mL methanol was placed in the right-hand side
ꢁ
respectively.
chamber. The permeation experiments were done at 25 C.
The distribution coefficients (K
d
), selectivity coefficients (a) Finally the concentration of both GA and SA in the permeate
of GA with respect to SA can be obtained according to the solution was determined by a UV system. The amounts of the
following equation: substrate that permeated the membranes were determined by a
ꢀ
2
ꢀ1
UV system. The ux J (mmol cm
s ), permeability coefficient
q
e
K
d
¼
(2)
2
ꢀ1
P (cm s ) and separation factor (perm-selectivity) aASA/SA can
C
e
be calculated by the following equations:
ꢀ
1
ꢀ1
Kd (mL g ) represents the distribution coefficient, q (mmol g
)
e
ꢀ
1
DC V
i
and C (mmol L ) are the equilibrium binding amount and the
e
Ji ¼
i ¼ SA; GA
i ¼ SA; GA
(5)
(6)
DtA
equilibrium concentration of the adsorbate, respectively. The
selectivity coefficient a can be obtained according to the
following equation:
i
J d
P ¼
ðCFi ꢀ CRi
Þ
K
dðGAÞ
a ¼
(3)
K
dj
P
SA
a
SA=GA ¼ P
(7)
GA
where, Kdj represents the distribution coefficient of competition
i
where DC /Dt is the concentration change rate of the receiving
solution which can be obtained by calculating the slope value of
species.
the diffused analyte concentration–time curve, V is the volume
2
.5. Membrane ux experiment
ꢀ
2
of solution, A is the effective membrane area (cm ), d is the
membrane thickness and (CFi ꢀ CRi) is the concentration
difference between feeding and receiving chambers.
ꢀ1
The aqueous solution containing 0.65 mmol L
GA was
prepared as feed solution, and the composite imprinted
membrane and blank membrane were tted on the ultraltra-
tion cell with 25 mm diameter (UF-8010, Amicon). The feed
solution was permeated through the membrane (operation
2.7. Optimization of dynamic separation of GA from SA
pressure: 0.15 MPa), and the ux of the feed solution passing In order to be able to predict and control the dynamic separa-
different membranes were calculated by the following equation: tion of GA from ASA, optimization studies should be conducted.
The response surface methodology (RSM), introduced by Box
V
F ¼
(4)
and Wilson, which is a collection of mathematical and statis-
tical techniques whose purpose is to analyze, by an empirical
st
Where V is volume of permeate solution (mL), t and s represent
operation time (min) and effective area of membrane (cm ),
27
model, problems as the one posed. RSM includes much more
2
model tting and analysis of them. The relation between vari-
ables and response is theoretically described by a function that
is the underlying physical mechanism to the problem under
study. Second-order model are usually studied for tting
respectively.
2
.6. Permeation experiments
The permeate experiments were carried out using the mixture response surfaces.
solutions with varied concentrations of GA and SA as the feed
In the study, Box–Behnken design (BBD) model of RSM was
solution. The permeate experiments were performed in a two- used for the determination of optimum levels of the processing
2
28
compartment cell (effective membrane area ¼ 1.5 cm ; capacity variables for the parameters studied. A 3-factor-3-level BBD
150 mL ꢃ 2). The membrane was xed between the was applied to design the experimental conditions using
compartments. Then, the mixture solution of GA and SA Design-Expert soware involving three factors, the
¼
38632 | RSC Adv., 2014, 4, 38630–38642
This journal is © The Royal Society of Chemistry 2014

View Article Online
Paper
RSC Advances
Table 2 BBD experiment design and result for the optimization of the
separation factors between SA and GA
C
ꢀ1
ꢁ
ꢀ1
Run
A (C: mmol L
)
B (T: C)
(V: mL min
)
b
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
0.0325
0.325
0.0325
0.325
0.179
0.179
0.179
0.325
0.179
0.179
0.179
0.179
0.179
0.179
0.0325
0.325
0.0325
45.0
30.0
30.0
15.0
15.0
15.0
30.0
45.0
30.0
30.0
45.0
30.0
45.0
30.0
15.0
30.0
30.0
3.0
1.0
1.0
3.0
5.0
1.0
3.0
3.0
3.0
3.0
5.0
3.0
1.0
3.0
3.0
5.0
5.0
8.890
2.123
13.01
1.712
1.179
3.495
2.018
1.023
1.512
2.101
1.312
2.301
2.971
2.716
10.78
0.837
7.081
0
1
2
3
4
5
6
7
À
Á
Á
C
eSA
C
fGA ꢀ CeGA
b ¼
À
(8)
C
eGA
C
fSA ꢀ CeSA
where b is the selective separation factor during the dynamic
separation process, CfGA and CfSA are the initial concentration of
GA and SA in the feeding solutions (mmol L ), respectively,
ꢀ1
CeGA and CeSA are the effluent concentration of GA and SA from
ꢀ
1
the column (mmol L ), respectively.
The mathematical relationship between responses (Y) and
the three signicant independent variables (A, B, and C) were
usually described by estimating coefficients of the following
28
second-order polynomial model based on experimental data.
2
2
2
Y ¼ b + b A + b B + b C + b A + b B + b C + b AB
0
1
2
3
11
22
33
12
+
b AC + b BC
13 23
where, Y is the response function, b
0
is an constant, b
1
, b
2
and
3
b are the coefficients of the linear, b11, b22, and b33 are
quadratic coefficients, b12, b13 and b23 are the interaction
coefficients between the three factors.
A multiple regression analysis was done to obtain the coef-

cients and the equation could be used to estimate the
response. A total of 17 experiments were needed to estimate the
full model. All the experimental data were statistically analyzed
by the soware package Design-Expert 8.0.5b. p-Values of less
than 0.05 were considered to be statistically signicant.
Scheme 1 Scheme and possible mechanism of the RTIL mediated
NHSG route to prepare the composite imprinted alumina membrane.
3
Results and discussion
concentration of GA (A), temperature (B) and ow rate (C) of 3.1. UV study
separation process, which were chosen as independent vari-
The interaction between the functional monomers (MAA, AA or
ables. The selected responses for analysis were separation factor
b) based on eqn (8). The levels of the variable factors in the
CA) and GA was studied by UV spectroscopic analysis due to that
the functional monomers strongly interact with the template
(
experiment and the experimental design with the parameters
are given in Table 2.
29
and form stable host–guest complexes prior to polymerization.
The results were shown in Fig. 1.
This journal is © The Royal Society of Chemistry 2014
RSC Adv., 2014, 4, 38630–38642 | 38633

View Article Online
RSC Advances
Paper
when increased the concentration of functional monomer. The
changes of the maximum absorbance for any functional
monomer became smaller when the molar ratio of the template
and functional monomer was greater than 1 : 4. It demon-
strated that 1 : 4 was the optimal molar ratio chosen for the
synthesis of IAM Since too large amount of functional mono-
mers may lead to their own association and increase nonse-
lective binding sites. Among the three functional monomers,
CA had more obvious changes, indicating stronger interaction
between CA and GA.
3.2. FTIR spectrum
To ascertain the successful preparation of the ionic liquid, FT-
IR spectra (Fig. 2) were obtained for [BMIM]Cl and [BMIM]PF .
6
ꢀ1
ꢀ
1
For [BMIM]Cl, the band around 3145 cm and 2800–3000 cm
are, respectively, aromatic and aliphatic C–H stretches. The
ꢀ
1
wide and strong adsorption band around 3421 cm could be
ascribed to stretching vibrations of O–H of water which absor-
bed by [BMIM]Cl for its hydrophily. While no peak was found
ꢀ
1
around 3421 cm , indicating that the ion liquid ([BMIM]PF
6
) is
ill-suited to absorb water. Compared with [BMIM]Cl, a charac-
ꢀ
1
teristic feature of [BMIM]PF was the band around 833 cm
6
ꢀ
belonging to the adsorption peak of [PF ] . It suggested that the
6
ionic liquid was successfully synthetized.
3.3. Characterization
The surface microstructure (two and three-dimension) of Al O3
2
ceramic membranes before and aer imprinting polymeriza-
tion was examined by AFM (Fig. 3). It can be observed that the
surface topography (Fig. 3a) of Al O ceramic membranes
2 3
before modication visually shows a smooth surface. Aer
imprinting polymerization, the surface topography of CA–CIAM
exhibits orderly strip structure from Fig. 3c. Moreover, the
topography of CA–CIAM greatly changes including the peak-to-
valley height and roughness by comparison of the three-
dimension images of Al O ceramic membranes (Fig. 3b) and
2
3
CA–CIAM (Fig. 3d). It is clear that the surface roughness of CA–
CIAM become bigger aer polymerization, while the topography
Fig. 1 Adsorption spectra of the GA in the presence of various
concentrations of functional monomers (MAA (a), AA (b), CA (c)) in
ꢀ1
acetonitrile. Concentration of GA: 0.1 mmol L , corresponding pure
functional monomer solutions as blanks.
As shown in Fig. 1a–c, it was clearly observed that the
maximum absorbance at 215 nm of the mixture solutions
decreased with the increase in the functional monomer
concentration. The maximum absorption wavelength showed
variable changes of redshi, showing that the interaction
6
between functional monomer and GA was getting stronger Fig. 2 FTIR spectrum of [BMIM]Cl and [BMIM]BF .
38634 | RSC Adv., 2014, 4, 38630–38642
This journal is © The Royal Society of Chemistry 2014

View Article Online
Paper
RSC Advances
Fig. 3 AFM topographical images revealing changes in the surface of ceramic membrane: Al
dimension), CA–CIAM (c: two dimension, d: three-dimension).
O
2 3
ceramic membrane (a: two dimension, b: three-
of Al
which revealing that the well-distributed imprinted layer
formed on the surface of Al ceramic membranes.
2 3
O ceramic membranes appeared more homogeneous,
2 3
O
SEM analysis was used to further investigate the membrane
surface and cross section microstructure. The microscopic
structures of various prepared CIAM were studied by using SEM
with gold-sputtered samples. As seen in Fig. 4a, Al O ceramic
2
3
membrane exhibited a more porous and smooth structure.
Fig. 4b–d show that the surface of the MAA–CIAM, AA–CIAM
and CA–CIAM were all mostly covered by a rough imprinted
layer aer the polymerization procedure, which can be clearly
seen in the inset of the gure. Compared to the MAA–CIAM and
AA–CIAM, it is evident that the formation of more regular pores
is observed from the morphology of the imprinted layer on
surface of CA–CIAM2.
Fig. 4d–f also presented the microphotographs of the surface
of CA–CIAMs, which were prepared with pre-polymerization
solutions with different amount of crosslinker (shown in
Table 1). Comparing with the differences in the morphology of
CA–CIAM1, CA–CIAM2 and CA–CIAM3, it is found that CA–
CIAM2 presents more regular pores than CA–CIAM1 and CA–
CIAM3, which indicated the amount of crosslinker affected the
structure of CIAM.
Fig. 5 presents the microphotographs of the cross section of
the original membrane, CA–CIAM0, CA–CIAM2, CA–CNAM2. As
can be seen from Fig. 5, there were signicant difference in
2 3
cross section morphology between nascent and imprinted Fig. 4 SEM images of (a) Al O ceramic membrane, (b) MAA–CIAM, (c)
AA–CIAM, (d) CA–CIAM1, (e) CA–CIAM2, (f) CA–CIAM3 and the inset
shows an enlargement of part of the four images, respectively.
2 3
membranes. The original Al O ceramic membrane exhibited a
smooth cross section structure. Fig. 5b–d show that the cross
section of the CA–CIAM0, CA–CIAM2, CA–CNAM2 are all mostly
covered by a thin imprinted layer on the base of Al O ceramic became much looser and formed more micropores, which
2
3
membrane aer the polymerization procedure. Compared with would make the imprinted molecules access the binding sites
CA–CIAM0, the imprinted layer of CA–CIAM2 and CA–CNAM2 exposed at the imprinted layer more easily. It was proved that
This journal is © The Royal Society of Chemistry 2014
RSC Adv., 2014, 4, 38630–38642 | 38635

View Article Online
RSC Advances
Paper
Fig. 5 SEM images of cross section of the studied membranes (a–d):
(
a) original membrane, (b) CA–CIAM0, (c) CA–CIAM2, (d) CA–CNAM2.
the low vapor pressure of RTIL could assist in reducing the
problem of gel shrinkage, and also act as pore templates in the
sol–gel reaction.
Compared with imprinted membrane (CA–CIAM2), there
were less differences between the morphology of imprinted and
non-imprinted membrane (CA–CNAM2). Therefore, the distinct
properties for imprinted and non-imprinted membrane could
not entirely be attributed to the morphology difference, but to
the imprinting effect.
Fig. 6 The flux of GA aqueous solution through different membranes.
All these changes of surface and cross section morphology of
the Al O ceramic membrane before and aer polymerization
2 3
indicated that the reaction between the original ceramic
membrane and pre-polymerization solution was indeed
reliable.
membranes (CA–CIAM1, CA–CIAM2 and CA–CIAM3) coated
with various amounts of imprinted polymer.
The ux of GA decreased with the increase in the amount of
coated polymer, especially for CA–CIAM3, the ux of which was
lower than the original alumina membrane. It might be
ascribed to that excessive polymer on the membrane will greatly
reduce the porosity of membrane.
3
.4. Membrane ux studies
The relationship between operation time and ux of GA
aqueous solution through different membranes was all recor-
3.5. Batch adsorption studies of membrane
ded in Fig. 6. With the increase of operation time, the ux of The adsorption capacity at equilibrium of varied CIAM and
membranes declined stablely. It was also observed from Fig. 6a CNAM are shown in Fig. 7. The adsorption capacity of GA at
ꢀ
ꢀ3
3
ꢀ1
ꢀ1
ꢀ3
that the water ux of GA aqueous solution was in an order of equilibrium was ca. 100.4 ꢃ 10 mmol g (AA), 125.7 ꢃ 10
ꢀ
1
alumina membrane <CA–CIAM0 < CA–CNAM2 < CA–CIAM2. mmol g (MAA), 143.6 ꢃ 10 mmol g (CA) for CIAM; and
ꢀ
3
ꢀ1
ꢀ3
ꢀ1
The higher water ux of CA–CIAM0, CA–CIAM2 and CA–CNAM2 62.56 ꢃ 10 mmol g (AA), 78.45 ꢃ 10 mmol g (MAA),
ꢀ
3
ꢀ1
than that of alumina membrane was probably due to lots of 82.65 ꢃ 10 mmol g (CA) for CNAM, respectively. It was found
binding sites on the outer surface and inner pores in the that the membrane (CIAM or CNAM) prepared with CA as the
alumina membrane, which was benecial for the accessibility functional monomer had a higher adsorption capacity than the
and mass transport for GA. Additionally, CA–CIAM2 has a other two corresponding membranes. The results indicated that
higher ux of GA compared with the non-imprinted membrane it was much easier for GA to access CA–CIAM surface than the
(
CA–CNAM2), which might be ascribed to more selective cavi- AA–CIAM and CA–CNAM. The nature of the carboxylic acid (AA,
ties or adsorption sites obtained on the surface of CA–CIAM2 MAA, CA) signicantly affected not only the morphology of the
which promoted the permeation of GA. prepared membrane (Fig. 4b–d), but also the adsorption
The content of polymerization layer on the imprinted performance. The data also showed that the imprinted
membrane was expected to strongly affect the ux of membrane signicantly adsorbed more GA than the non-
membranes. Fig. 6b shows the GA ux through the imprinted imprinted membrane for either functional monomer, which
38636 | RSC Adv., 2014, 4, 38630–38642
This journal is © The Royal Society of Chemistry 2014

View Article Online
Paper
RSC Advances
Fig. 7 Batch adsorption of GA by CIAM and CNAM prepared with AA,
MAA and CA as functional monomers, respectively.
might be ascribed to more cavities or adsorption sites obtained
on the surface of imprinted membrane.
The effect of RTIL and the amount of polymer layer on the
imprinted membranes were expected to strongly affect the
adsorption of GA. Fig. 8 shows the GA adsorption on the
membranes (CA–CIAM0, CA–CIAM1, CA–CIAM2, and CA–
CIAM3) coated with various amounts of imprinted polymer. The
adsorption capacity of GA at equilibrium for CA–CIAM1
ꢀ
3
ꢀ1
(
143.6 ꢃ 10 mmol g ) was much higher than that of CA–
ꢀ3 ꢀ1
CIAM0 (86.77 ꢃ 10 mmol g ). It could be ascribed to that
when no RTIL was involved, the imprinted layer on the
membrane was less porous, which was responsible for the poor
recognition ability, since less porosity led to less chances for the
analyte to access the recognition site.
Comparing with CA–CIAM1, CA–CAM2 and CA–CIAM3, the
amount of GA adsorption increased with the increase in the
amount of coated polymer due to more adsorption sites on the
surface of CA–CIAM. However, the increasing trend was not very
obvious, which might be ascribed to that excessive polymer on
the surface limited the effect on the specic adsorption of GA.
With the increase in the amount of polymer, less porous on the
membrane reduced efficient recognition sites.
Fig. 9 Time-permeation curves of SA and GA through the various
membranes (feed concentration ¼ 0.360 mmol L ).
ꢀ1
on the permeability performances, Fig. 9a shows the time-
dependent permeation curves of SA and GA on the blank Al O3
2
ceramic membrane and the CA–CIAM (1, 2, 3), respectively. As
shown in Fig. 9a, compared with the Al O ceramic membrane,
2 3
3
.6. Permeability
3.6.1 Permselectivity of composite imprinted membranes.
the concentration of SA in the receiving chamber was much
higher than that of GA for the three CA–CIAMs, indicating that
the three CA–CIAMs had a special selectivity of GA. Comparing
the three CA–CIAMs, it was also found that the diffusion rate of
both SA and GA through CA–CIAM3 was much lower than other
two imprinted membranes (CA–CIAM1 and CA–CIAM2). More-
over, it can be clearly seen that the CA–CIAM2 shows the
greatest differences in the concentration between SA and GA in
the receiving chamber among the three CA–CIAMs, indicating
that the CA–CIAM2 has better separation effect for SA and GA. It
In the permeation experiment, the permeation concentrations
of SA and GA through different membranes at the feed mixture
ꢀ
1
concentration of 0.35 mmol L in 3 h were shown in Fig. 9. In
order to study the effect of the content of polymerization layer
2 3
may suggest that CA–CIAM1 with little polymer on the Al O
ceramic membrane has limit recognition sites, while excessive
polymer will increase non specic adsorption and decrease the
permeation rate of competing molecule, which affected the
separation effect.
Fig. 9b also shows the time-dependent permeation curves of
the CA–CIAM0, CA–CIAM2, CNAM, respectively. Compared to
Fig. 8 Batch adsorption of GA by varied imprinted membrane.
This journal is © The Royal Society of Chemistry 2014
RSC Adv., 2014, 4, 38630–38642 | 38637

View Article Online
RSC Advances
Paper
imprinted membranes, the diffusion rate of SA through the
The following equation was used to compute the overall
CNAM was closed to that of GA. This means that the non- mass transfer coefficient (KOV) using concentration as the
32
imprinted membrane exhibited non specic adsorption to SA driving force:
ꢀ
ꢁ
ꢀ
ꢁ
and GA. Comparing CA–CIAM0 and CA–CIAM2, it was also
K
OV
A
1
C
F;0
clearly found that both the diffusion rate of SA and GA through
CA–CIAM0 were much lower than that of CA–CIAM2. Especially,
CA–CIAM2 exhibited excellent separation effect for SA and GA where f ¼
t ¼
ln
(9)
V
F
1 þ f
ð1 þ fÞC
F
ꢀ fCF;0
VF
VR
.
compared with CA–CIAM0. It can be attributed to the CA–
CIAM0 with no RTIL involved was less porous, which caused the
blocking permeation of SA and GA through the imprinted estimate KOV. This plot yields a linear curve with the slope
membranes. And there were less efficient recognition sites, (KOVA/V ) containing KOV. The values of KOV can be determined
which will affect the separation effect. So RTIL incorporated will from a plot of CF,0 against t using nonlinear regression analysis.
The right hand side of equation is plotted against time to
R
play an important role in the separation process for imprinted
membrane.
As shown in Table 3, the experimental data were well tted
by the model for varied membrane with high R values. Fig. 10
2
3
.6.2 Mass transfer performance. According to the shows the experimental points and nonlinear predicted points
mentioned above, the imprinted membrane itself is of high according to the parameters of the mass transfer model in Table
importance in success of the membrane separation. The inherent 3. Apparently, the nonlinear model provided ne tting of SA
nature of the polymer membranes is their specic recognition and GA for various membranes in Fig. 10a and b.
properties. From permeation experiments, the permeability
coefficients of SA and GA through different imprinted coefficients of SA was close to that of GA for Al
membranes were calculated by eqn (6) as shown in Table 3. The while the KOV values of SA were much higher than that of GA for
results show that the Al
membrane has a similar permeability CA–CIAM1, CA–CIAM2 and CA–CIAM3, respectively. However,
coefficients for both SA (3.736 ꢃ 10 cm s ) and GA (3.841 ꢃ both the mass transfer coefficients of SA and GA for CA–CIAM3
In Table 3, the experimentally determined mass transfer
2 3
O
membrane,
2 3
O
ꢀ
4
2
ꢀ1
ꢀ
4
2
ꢀ1
10
cm s ) compared to the imprinted and non-imprinted were lower than that of CA–CIAM1 and CA–CIAM2, These
membrane, indicating that Al
membrane shows almost the means that the imprinted layer can greatly affect the mass
same mass transfer ability during the permeation process. transfer ability of the Al ceramic membrane.
Moreover, as for the imprinted membranes (CA–CIAM0, CA–
Additionally, the mass transfer coefficients of SA for CA–
CIAM1, CA–CIAM2, CA–CIAM3), all the p-values of SA is much CIAM2, CA–CIAM0 and CNAM2 were 5.667 ꢃ 10 , 2.668 ꢃ
2 3
O
O
2 3
ꢀ
3
ꢀ
3
ꢀ3
ꢀ1
higher than GA. This suggests that GA has a much stronger 10 , 5.330 ꢃ 10 cm s , respectively, while the mass transfer
ꢀ3
interaction with imprinted membranes and diffuses faster than coefficients of GA for the three membranes were 1.120 ꢃ 10
,
ꢀ
3
ꢀ3
ꢀ1
SA. Because the SA molecules compete with GA molecules to go 1.009 ꢃ 10 , 3.001 ꢃ 10
cm s . Among the three
through the inner pores of the imprinted layer, GA as template membranes, CA–CIAM2 has a high mass transfer coefficients of
can bind with the group on the imprinted membrane more SA and low mass transfer coefficients of GA, showing a favorable
powerfully and then SA will get through the composite separation effect of SA and GA. The inverse of the mass transfer
membrane preferentially. Thus, the imprinted membrane has a coefficient is the mass transfer resistance. Compared with the
special selectivity to GA. Additionally, the CA–CIAM2 shows a CA–CNAM2, CA–CIAM2 has the smallest mass transfer coeffi-
higher selectivity than that of other imprinted membranes and cient for GA, suggesting the CA–CIAM2 has maximum resis-
CA–CNAM2 according to the results of the values separation tance for transferring GA.
factors (aSA/GA) summarized in Table 3, which was agreed with
the earlier report by J. Komiyama et al.
In the mass transfer process, analyte transport takes place
from the feed solution to the receiving solution. In our research,
30,31
Table 3 Experimental permeability results for SA and GA through the various imprinted membranes
Mass transfer model
7
ꢀ2 ꢀ1
P ꢃ 10ꢀ4 (cm s
2
ꢀ1
OV ꢃ 10ꢀ3 (cm sꢀ1
2
Membranes
Al membrane
Substrate
J ꢃ 10ꢀ (mmol cm
s
)
)
a
SA/GA
K
)
R
2
O
3
SA
GA
SA
GA
SA
GA
SA
GA
SA
GA
SA
GA
6.537
6.723
8.995
4.675
8.447
3.136
6.409
3.107
6.033
2.897
8.098
6.214
3.736
3.841
5.140
2.671
4.827
1.792
3.662
1.775
3.448
1.655
4.628
3.551
0.9723
1.924
2.693
2.062
2.083
1.303
8.681
7.001
7.329
1.953
5.667
1.120
3.012
1.099
2.668
1.009
5.330
3.001
0.9700
0.9733
0.9809
0.9758
0.9880
0.9868
0.9955
0.9907
0.9957
0.9931
0.9967
0.9942
CA–CIAM1
CA–CIAM2
CA–CIAM3
CA–CIAM0
CA–CNAM
38638 | RSC Adv., 2014, 4, 38630–38642
This journal is © The Royal Society of Chemistry 2014

View Article Online
Paper
RSC Advances

lms can be negligible compared to those of the membrane. It
may conclude that the mass transfer across the membrane
phase is the rate-controlling step. This general assumption is
33
consistent with the reported literature.
3.7. Selective permeation mechanism
The permeability experiments mentioned above showed that
the imprinted membranes showed higher diffusion rate for the
completing molecule (SA) than for the template (GA). So the
major mechanisms applied in analyzing the selective transport
can be regarded retarded permeation, which was due to affinity
binding-faster transport of SA, until a saturation of recognition
34
sites with GA is reached. The transport of the target analytes go
across the membrane by adsorbing on and desorbing from the
2 3
imprinted skin on the Al O membrane. In the permeation
experiment, the template molecule (GA) can preferentially bond
with the functional groups in the recognizing sites on the
imprinted membrane. Meanwhile, the SA molecules can't
match the recognition cavity well in size and shape. So, the
imprinted membrane had hardly recognition effect for non-
template and SA molecules can be easily desorbed from the
membrane. As a result, SA had more chance to transfer
continuously from one side to the other side of the composite
membrane. So, the imprinted membrane acted as an adsorptive
35
membrane. Therefore, such transport mechanism for perme-
ation of the SA and GA towards the imprinted Al O membrane
2
3
was accordance with the retarded mechanism mentioned
above.
3.8. Design of RSM for optimization of dynamic separation
A numerical optimization combined with a desirability function
was applied to calculate the optimum values of the concentra-
Fig. 10 Non-linear regression of mass transfer model for SA and GA
through different membranes.
ꢀ
1
ꢁ
tion of GA (mmol L ), the temperature ( C) and the ow rates
ꢀ
1
(
mL min ) to provide more theoretical basis in the dynamic
since the solvents in both reservoirs are the same, the liquid separation process. For industrial application, a dynamic
resistances are greatly reduced. Additionally, our aim was to separation study is essential for continuous ow systems. The
obtain the separation of SA and GA by forming different mass optimization step of proposed method was carried out using a
transfer rate via the interaction between the analytes and the Box–Behnken design (BBD). In this research, CA–CIAM2 was
imprinted layer on the Al
2
O
3
membrane. On the basis of Table selected to study the dynamic separation of GA and SA.
3.8.1 Statistical analysis and the model tting. There were
3, different membranes show different overall mass transfer
coefficients for both SA and GA. So the resistances of the liquid a total of 17 runs with different combinations of the physical
Table 4 Estimated regression model of relationship between response variables (b) and independent variables (A, B, C)
Factor
SS
b
MS
F
p
Model
A
B
219.7
145.1
1.100
15.65
0.360
5.390
0.110
51.51
2.776 ꢃ 10
0.077
4.63
9
1
1
1
1
1
1
1
1
1
3
24.42
145.05
1.10
15.65
0.36
5.39
0.11
51.51
2.776 ꢃ 10
0.077
1.54
63.67
378.29
2.88
40.83
0.94
14.06
0.28
134.33
7.239 ꢃ 10
0.20
<0.0001
<0.0001
0.1338
0.0004
0.3644
0.0072
0.6122
<0.0001
0.9346
0.6669
0.0530
Signicant
C
AB
AC
BC
2
A
2
ꢀ3
ꢀ3
ꢀ3
B
2
C
Lack of t
6.36
Not signicant
This journal is © The Royal Society of Chemistry 2014
RSC Adv., 2014, 4, 38630–38642 | 38639

View Article Online
RSC Advances
Paper
parameters for optimizing the three individual parameters in
the BBD. The experimental conditions and the selective sepa-
ration factors of GA and SA according to the factorial design
were shown in Table 2. The goal of the optimization was to
maximize the separation factors. Results also showed that the
response values (b) ranged from 0.837 to 13.01. The results of
this limited number of experiments provided a statistical model
that was used to identify trends in high separation factor for the
dynamic separation process. Table 2 shows the matrix and the
separation factors (b). The quadratic models below illustrate the
relationship of the three variables and Y (b).
Y ¼ 2.13 ꢀ 4.26A ꢀ 0.37B ꢀ 1.40C + 0.30AB + 1.16AC + 0.16BC
2
2
2
+
3.50A ꢀ 0.026B + 0.14C
The analyses of variance (ANOVA) results for the response
surface quadratic model are shown in Table 4. The p-value was
used to check the signicance of each coefficient, and the
smaller the p-value was, the more signicant the corresponding
coefficient was. The ANOVA of the quadratic regression models
demonstrated the models to be signicant with low probability
(p < 0.0001). The predicted versus observed values of the sepa-
ration factors indicated good agreement between the poly-
nomial regression model and the experimental data, with the
2
coefficient of determination (R ) of 0.9879. The lack of t
measures the failure of the model to represent data in the
experimental domain at points which are not included in the
regression. The non-signicant value of lack of t (>0.05)
revealed that the quadratic model is statistically signicant for
the response (b).
The contribution of each parameter/factor (rst and second
order) besides the interaction between them on the separation
effect was shown in Table 4. All coefficients with p-values less
than 0.05 are signicant. As seen in Table 4, the analyses indi-
2
cated that A, C, AC, A are signicant model terms for separation
2
of GA from SA, while the other term coefficients (B, AB, BC, B
C ) were not signicant (p > 0.05).
2
3.8.2 Response surface plots and response optimization.
Response surface methodology was used to determine the
optimal response for the separation of GA from SA. The results
of separation factors affected by concentration of GA, ow rate
and temperature are presented in Fig. 11. Three-dimensional
(3D) surface plots showed visually the effects and interaction of
two independent variables on the responding variable as third
independent variable was kept at level zero.
Fig. 11a giving the separation factor between GA and SA as a
function of concentration and ow rate at xed temperature
Fig. 11 Response surfaces plots for selective factors as a function of
ꢁ
(30.0 C), indicated that the separation factor rapidly increased
(a): concentration of GA and flow rate, (b) concentration of GA and
with the decrease of the concentration of GA from 0.325 to
temperature, (c) flow rate and temperature.
ꢀ
1
0
.0325 mmol L and decreased with the increase of the ow
ꢀ
1
rate from 1.0 to 5.0 mL min . It could be explained a decrease
in the residence time, which restricted the contact of GA and
SA mixed solution to the CA–CIAM. At higher ow rate, the GA
did not have enough time to diffuse into the binding sites on
CA–CIAM and they passed the membrane fast before equilib-
rium occurred. Moreover, the increase of the separation factor
when the concentration of GA in the mixture of GA and SA
decreased could be related with the chemical environment, the
binding sites preservation of the CA–CIAM. The number of
effective binding sites was limited aer imprinting on the
membrane. So it suggested that the imprinted membrane
38640 | RSC Adv., 2014, 4, 38630–38642
This journal is © The Royal Society of Chemistry 2014

View Article Online
Paper
RSC Advances
favored the separation of the mixture containing lower GA 2012CBB21500), Ph.D. Programs Foundation of Ministry of
initial concentrations. Education of China (No. 20123227120015) and Natural Science
At the meanwhile, Fig. 11b shows the 3-D response surface Foundation of Jiangsu Province (BK2011461, SBK2011459,
plot at varying temperature and concentration at xed ow rates BK2011514, BK20140580), China Postdoctoral Science Foun-
ꢀ1
of 3.0 mL min . The maximum separation factor can be ach- dation funded project (Nos. 2012M511220, 2013M530240) and
ꢁ
ieved when temperature and concentration of GA at 15.0 C and Special Financial Grant from the China Postdoctoral Science
ꢀ
1
0
.0325 mmol L . At high temperatures, less GA molecules were Foundation (2014T70488).
required to satisfy the maximum adsorption capacity of the CA–
CAIM. Therefore, the low temperature favors the separation
effect of GA from SA, indicating an exothermic process.
Fig. 11c shows the 3-D response surface plot with varying
temperature and ow rates at xed concentration of GA of
References
1 N. H. Ullsten, M. G ¨a llstedt, E. Johansson, A. Gr ¨a slund and
M. S. Hedenqvist, Biomacromolecules, 2006, 7, 771.
2 M. Hussain, A. Javeed, M. Ashraf, Y. Zhao, M. M. Mukhtar
and M. Ur Rehman, Int. Immunopharmacol., 2012, 12, 10.
3 A. L. Ahmad, M. R. Othman and H. Mukhtar, Int. J. Hydrogen
Energy, 2004, 29, 817.
ꢀ
1
0.179 mmol L . The maximum separation factor can be ach-
ieved in the case of temperature and ow rate at the value of
ꢁ
ꢀ1
1
5.0 C and 1.0 mL min , respectively. The results are con-
sitent with the discussion mentioned above.
.8.3 Optimization and verication of the model. Accord-
3
ing to Fig. 10a–c, it can be concluded that the optimal condi-
tions giving the maximum response for selectivity coefficient
4 H. M. Zhang, X. Quan, S. Chen and H. M. Zhao, Environ. Sci.
Technol., 2006, 40, 6104.
5 N. Hilal, V. Kochkodan, G. Busca, O. Kochkodan and
B. P. Atkin, Sep. Purif. Technol., 2003, 31, 281.
6 T. A. Sergeyeva, O. O. Brovko, E. V. Piletska, S. A. Piletsky,
L. A. Goncharova, L. V. Karabanova, L. M. Sergeyeva and
A. V. El'skaya, Anal. Chim. Acta, 2007, 582, 311.
7 M. S. da Silva, R. Viveiros, A. Aguiar-Ricardo,
V. D. B. Bonif ´a cio and T. Casimiro, RSC Adv., 2012, 2, 5075.
8 C. Y. He, Y. Y. Long, J. L. Pan, K. Li and F. Liu, J. Biochem.
Biophys. Methods, 2007, 70, 133.
ꢀ
1
(
1
13.39) was found in conditions of A ¼ 0.0325 mmol L , B ¼
ꢁ
ꢀ1
5 C, C ¼ 1.0 mL min
.
For their validation of the optimum conditions, triplicate
conrmatory experiments were carried out under the optimized
conditions and the average separation factor (b) was 13.26 ꢂ
0.87%. The results are closely related with the data obtained
from optimization analysis, indicating Box–Behnken design
could be effectively used to optimize the separation GA from SA.
9
Z. Zhang, M. Zhang, Y. Liu, X. Yang, L. Luo and S. Yao, Sep.
Purif. Technol., 2012, 87, 142.
4
. Conclusions
A novel composite imprinted alumina membrane (CIAM) was 10 R. J. Ansell, J. K. Kuah, D. Wang, C. E. Jackson, K. D. Bartle
successfully prepared via the RTIL-mediated NHSG method- and A. A. Clifford, J. Chromatogr. A, 2012, 1264, 117.
ology for selective separation of GA from SA. The adsorption 11 G. Zhu, J. Fan, Y. Gao, X. Gao and J. Wang, Talanta, 2011, 84,
capacity, ux and permeation selectivity of the imprinted 1124.
membranes depended on the functional monomers and the 12 H.
crosslinking degree of polymer layer on the alumina
membrane. CA–CIAM2 was found to be the promising imprin-
Henschel,
N.
Kirsch,
J.
Hedin-Dahlstr ¨o m,
M. J. Whitcombe, S. Wikman and I. A. Nicholls, J. Mol.
Catal. B: Enzym., 2011, 72, 199.
ted membrane to increase the effect on separating GA from SA. 13 C. Baggiani, F. Biagioli, L. Anfossi, C. Giovannoli, C. Passini
Additionally, incorporation of RTIL can greatly increase the and G. Giraudi, React. Funct. Polym., 2013, 73, 833.
porosity, ux and recognition ability, and further improve the 14 J. L. Urraca, M. D. Marazuela, E. R. Merino, G. Orellana and
selectivity of the CIAM to GA. M. C. Moreno-Bondi, J. Chromatogr. A, 2006, 1116, 127.
RSM was used to estimate and optimize the experimental 15 A. Higuchi, M. Tamai, Y. Ko, Y. Tagawa, Y. Wu, B. Freeman,
variables (the concentration of GA, temperature and ow rate) J. Bing, Y. Chang and Q. Ling, Polym. Rev., 2010, 50, 113.
in the dynamic separation process. Under the optimal condi- 16 M. E. D ´ı az-Garc ´ı a and R. B. La ´ı n˜ o, Microchim. Acta, 2005, 149,
ꢀ
1
tions of GA concentration at 0.0325 mmol L , temperature at
19.
5.0 C, ow rate at 1.0 mL min , the experimental selectivity 17 S. Fireman-Shoresh, I. Popov, D. Avnir and S. Marx, J. Am.
Chem. Soc., 2005, 127, 2650.
ꢁ
ꢀ1
1
factor was 13.26 ꢂ 0.87%, which was close to the predicted
selectivity factor value. We expect that the RTIL-mediated NHSG 18 Z. Jiang, Y. Yu and H. Wu, J. Membr. Sci., 2006, 280, 876.
protocol is promising as a general strategy for the fabrication of 19 H. F. Wang, Y. Z. Zhu, X. P. Yan, R. Y. Gao and J. Y. Zheng,
high selective imprinted membrane for purication of trace
Adv. Mater., 2006, 18, 3266.
analytes in the complex matrix.
20 S. Panja, P. K. Mohapatra, S. C. Tripathi, P. M. Gandhi and
P. Janardan, Sep. Purif. Technol., 2012, 96, 289.
2
2
1 T. Welton, Coord. Chem. Rev., 2004, 248, 2459.
2 P. Hapiot and C. Lagrost, Chem. Rev., 2008, 108, 2238.
Acknowledgements
This work was nancially supported by the National Natural 23 Y. Zhou, J. H. Schattka and M. Antonietti, Nano Lett., 2004, 4,
Science Foundation of China (No. 21207051), National key 477.
basic research development program (973 Program, No. 24 Y. Zhou and M. Antonietti, Chem. Mater., 2004, 16, 544.
This journal is © The Royal Society of Chemistry 2014
RSC Adv., 2014, 4, 38630–38642 | 38641

View Article Online
RSC Advances
Paper
2
5 Y. Liu, M. Wang, Z. Li, H. Liu, P. He and J. Li, Langmuir, 2005, 31 K. Taki, I. Arita, M. Satoh and J. Komiyama, J. Polym. Sci.,
1, 1618. Part B: Polym. Phys., 1999, 37, 1035.
6 J. G. Huddleston, H. D. Willauer, R. P. Swatloski, A. E. Visser 32 R. Valadez-Blanco and A. G. Livingston, J. Membr. Sci., 2009,
and R. D. Rogers, Chem. Commun., 1998, 1765. 326, 332.
7 G. E. P. Box and K. B. Wilson, J. Roy. Stat. Soc. B, 1951, 13, 1. 33 M. Dingemans, J. Dewulf, L. Braeckman, H. V. Langenhove,
2
2
2
2
8 J. Han, Y. Wang, Y. Liu, Y. F. Li, Y. Lu, Y. S. Yan and L. Ni,
K. Friess, V. Hynek and M. Sipek, J. Membr. Sci., 2008, 322,
Anal. Bioanal. Chem., 2013, 405, 1245.
234.
2
9 J. F. He, Q. H. Zhu and Q. Y. Deng, Spectrochim. Acta, Part A, 34 M. Ulbricht, J. Chromatogr. B: Anal. Technol. Biomed. Life Sci.,
007, 67, 1297. 2004, 804, 113.
0 Z. Y. Jiang, Y. X. Yu and H. Wu, J. Membr. Sci., 2006, 280, 876– 35 D. Keith Roper and E. N. Lightfoot, J. Chromatogr. A, 1995,
82. 702, 3.
2
3
8
38642 | RSC Adv., 2014, 4, 38630–38642
This journal is © The Royal Society of Chemistry 2014