Universal surface modification by aldehydes on polymeric membranes for isopropanol dehydration via pervaporation

Universal surface modification by aldehydes on polymeric membranes for isopropanol dehydration via pervaporation

Journal of Membrane Science 492 (2015) 197–208 Contents lists available at ScienceDirect Journal of Membrane Science journal homepage: www.elsevier...

6MB Sizes 2 Downloads 75 Views

Journal of Membrane Science 492 (2015) 197–208

Contents lists available at ScienceDirect

Journal of Membrane Science journal homepage: www.elsevier.com/locate/memsci

Universal surface modification by aldehydes on polymeric membranes for isopropanol dehydration via pervaporation Dan Hua, Tai-Shung Chung n NUS Graduate School for Integrative Sciences and Engineering, Department of Chemical and Biomolecular Engineering, Faculty of Engineering, National University of Singapore, 10 Kent Ridge Crescent, Singapore 119260, Singapore

art ic l e i nf o

a b s t r a c t

Article history: Received 27 February 2015 Received in revised form 2 May 2015 Accepted 6 May 2015 Available online 5 June 2015

This work reports a novel and facile aldehyde modification method to transform polymeric hollow fiber membranes for pervaporation dehydration of alcohols and biofuels with much enhanced separation performance and chemical stability. A universal approach consisting of hyper-branched polyethyleneimine (HPEI) pre-treatment and aldehyde modification was proposed and successfully applied to various polymeric membranes such as Torlons, Ultems, and PES hollow fibers with greatly improved separation performance. The surface modified hollow fibers were characterized by positron annihilation spectroscopy, Fourier transform infrared spectroscopy, X-ray photoelectron spectroscopy, and water contact angle measurements, etc. Results show the HPEI pre-treatment not only mitigates surface defects but also augments membrane hydrophilicity favoring water transport, while the cross-linking reaction between HPEI and aldehydes tightens polymeric chains and enhances selectivity. Besides, the newly molecularly designed membranes display satisfactory separation performance at various feed temperatures and compositions as well as good long-term stability. This study may open up new perspectives to design advanced pervaporation membranes by synergistic combination of the super reactivity and unique chemical interactions from HPEI and aldehydes. & 2015 Elsevier B.V. All rights reserved.

Keywords: Aldehydes Surface modification Hollow fiber (HF) Isopropanol Pervaporation

1. Introduction Pervaporation is a promising membrane technique for alcohol dehydration and biofuel production due to its high separation efficiency, ability to break the azeotrope and low energy consumption [1–5]. As the heart of pervaporation processes, the ideal membrane must possess high productivity, good selectivity and long-term stability. Extensive researches have been conducted toward it via various molecular engineering of membrane materials and fabrication [2–7]. Polymers are of particular interest because they own many advantages over inorganic materials such as reasonable costs, easy fabrication and scale up [5]. However, most commercially available polymeric materials have some drawbacks such as limited separation performance and/or poor antiswelling properties. To overcome these material weaknesses, two strategies are available; namely, syntheses of novel polymers and modifications of the existing materials. Compared with novel material syntheses, modifications of the existing polymers provide benefits of simplicity and low risk [8]. Among available modification methods, chemical grafting and crosslinking have shown effectiveness to augment pervaporation

n

Corresponding author. Tel.: þ 65 6516 6645; fax: þ 65 6779 1936. E-mail address: [email protected] (T.-S. Chung).

http://dx.doi.org/10.1016/j.memsci.2015.05.056 0376-7388/& 2015 Elsevier B.V. All rights reserved.

membranes with enhanced separation factor and durability [2–5,9,10]. For instance, aldehyde-induced crosslinking modifications have successfully improved the physicochemical properties of early-generation pervaporation membranes made of poly (acrylic acid) (PAA), poly(vinyl alcohol) (PVA), sodium alginate and chitosan for dehydration. These hydrophilic materials originally have weak mechanical properties and swell severely in aqueous solutions [2–4,11]. Recently, polyimides and polyamide-imides have emerged as potential pervaporation materials for alcohol dehydration due to their excellent thermal, chemical and mechanical stabilities and high water selectivity [5,8,12,13]. Though these two materials have lower degrees of swelling than the aforementioned polymers, swelling reduces their performance too [5,14]. Therefore, various crosslinking modification methods have been proposed to control their swelling and stability, especially in harsh environments [2,5,15]. However, using aldehydes as cross-linkers to modify polyamide-imides (as well as polyimides) is rarely reported. Aldehydes are organic compounds containing aldehyde groups, which are highly reactive in various chemical reactions. For example, formaldehyde and glutaraldehyde have worked as cross-linkers to modify polymers containing hydroxide groups such as PVA, chitosan, alginate, and cellulose [2–4,11,16,17]. Aldehydes also react with both primary amines (RNH2) and secondary amines (R2NH) to form carbinolamines, which then

198

D. Hua, T.-S. Chung / Journal of Membrane Science 492 (2015) 197–208

Fig. 1. The molecular structures of aldehydes used to modify polymeric hollow fibers.

dehydrate to form substituted imines and enamines, respectively [18–20]. Several aldehydes were reported to crosslink proteins and chemicals containing amino groups [21–25]. In addition, formaldehyde was found to crosslink polyurethane with its free-NHCO groups [26]. With such high reactivity, aldehydes inspire us to explore the use of them to modify polyamide-imides and other polymeric membranes for dehydration pervaporation of alcohols. Therefore, the objectives of this manuscript are to investigate (1) if we could use aldehydes to directly modify polyamide-imide membranes for pervaporation and (2) if we could develop a universal approach to modify polyamide-imides, polyimides as well as other polymeric membranes using aldehydes. To meet the 1st objective, a Torlons 4000T-MV polyamideimide was chosen because (i) Torlons contains abundant secondary amino groups (–NH–), (ii) Torlons owns high thermal stability, good mechanical properties and solvent resistance which are critical for industrial applications. Several types of aldehydes, including glyoxylic acid (GA), benzaldehyde (BA), glucose (GL) and imidazolate-2-carboxyaldehyde (ICA) as shown in Fig. 1, were chosen to modify the surface of Torlons hollow fiber (referred to as HF thereafter) membranes. To meet the 2nd objective, a hyper-branched polyethyleneimine (HPEI) gutter layer would be firstly formed on the membrane surface before the aldehyde-induced crosslinking reaction. The HPEI pre-treatment would not only introduce abundant primary amino groups on the membrane surface for the subsequent aldehyde modification but also improve the overall hydrophilicity as well as effectively seal the membrane defects [27–30]. A covalent assembly of HPEI–glutaraldehyde multilayer on hydrolyzed polyacrylonitrile HF membranes had been fabricated by the layer-by-layer method for alcohol dehydration [31], but this method needed vacuum assistance during assembly and involved multiple steps including washing and drying after each assembly, which might incur additional costs and prolong production duration. Since the HPEI mediated aldehyde modification method (including HPEI pretreatment and then aldehyde modification) can be easily extended to other polymeric membranes such as Ultems, PES and others; this study may provide useful insights of using aldehydes to design next generation membranes for pervaporation applications.

Polymers, while Ultems 1010 polyetherimide was purchased from SABIC Innovative Plastics. N-methyl-2-pyrrolidone (NMP, analytical grade, Merck) was acquired as the solvent, while polyethylene glycol 400 (PEG, analytical grade, Merck), ethanol (analytical grade, Fisher), and deionized water were used as non-solvents for the preparation of spinning solutions. The bore fluids were prepared from mixtures of NMP, n-butanol (analytical grade, Fisher) and deionized water. Several aldehydes, including glyoxylic acid monohydrate (GA), benzaldehyde (BA), and glucose (GL) purchased from Sigma-Aldrich, as well as imidazolate-2carboxyaldehyde (ICA) provided by Alfa Aesar were used to modify polymeric HF membranes. A hyper-branched polyethyleneimine (HPEI, 50 wt% in aqueous solutions, Sigma Aldrich) with a molecular weight of 60 kg/mol was acquired to prepare solutions for the pretreatment of membrane surface. Moreover, methanol (reagent grade, Merck) and n-hexane (analytical grade, Fisher) were employed as solvents in this study. Isopropanol (IPA, reagent grade, Fisher) and deionized water were used to prepare the feed mixture for pervaporation experiments. The Torlons, Ultems and PES polymers were vacuum dried at 120 1C overnight before use, while all other chemicals were used as received.

2. Experimental section

2.3. The modifications of HF membranes by aldehydes and HPEI mediated aldehydes

2.2. The fabrication of HF substrates The HF substrates were obtained via a dry-jet wet spinning process as described elsewhere [32]. The dope solution was poured into an ISCO syringe pump and degassed overnight. The bore fluid was stored in another syringe pump shortly before the spinning started. During spinning, the dope solution and bore fluid were co-extruded via the spinneret with controlled flow rates. The nascent fiber traveled a certain air-gap distance and then entered a water coagulation tank for phase inversion. The HF substrate was then collected by a rolling drum with a controlled take up speed. The details of spinning parameters are listed in Table 1. The obtained HF substrates were cut and rinsed in tap water for at least 3 days to remove residual solvents. Afterwards, they were solvent exchanged with methanol followed by hexane for three consecutive times to further remove residual solvents within the membrane. Finally, they were air dried for further characterizations, post-modifications and pervaporation tests.

2.1. Materials Commercial Torlons 4000T-MV polyamide-imide and Radels A polyethersulfone (PES) were supplied by Solvay Advanced

Firstly, the aldehyde solutions were prepared by dissolving specific aldehydes in methanol with a concentration of 0.05 mmol/L. Fig. 2(a) illustrates the aldehyde modification scheme on Torlons HF

D. Hua, T.-S. Chung / Journal of Membrane Science 492 (2015) 197–208

substrates using an aldehyde solution at 65 1C for 4 h because aldehyde molecules would react with the secondary amino groups (–NH–) of Torlons. Then the modified HF was taken out, rinsed by fresh methanol, and air dried. Fig. 2(b) elucidates the HPEI mediated aldehyde scheme so that the aldehyde modification scheme could be applicable to HF membranes made from other materials. Firstly, the primary amino groups (–NH2) were grafted to the membrane surface by immersing the HF substrate in a 1 wt% HPEI aqueous solution at 60 1C for 1 h. Then, the HF was wiped with tissue papers and subsequently immersed into an aldehyde solution via a similar aldehyde modification scheme. As a result, aldehyde molecules would react easily with the primary amino groups of HPEI, and modify the membrane surface. 2.4. The fabrication of HF modules Each module consists of a 3/8 in. perfluoroalkoxy tube and two Swagelok stainless steel male run tees. HFs with an effective

199

length of 14–15 cm were then loaded into the module and sealed with epoxy at both ends of the male run tees. All the HF modules were cured at room temperature for 48 h before tests. 2.5. Characterizations of HF substrates and aldehyde-modified HFs The morphologies of HFs were observed via a field-emission scanning electron microscope (FESEM, JEOL JSM-6700LV). The membrane surface topology was characterized by a Nanoscope V atomic force microscope (AFM) from Bruker Dimension ICON. To study the surface hydrophilicity of pristine and aldehyde modified HF substrates, the water contact angle was measured by a Sigma 701 Tensiometer from KSV Instruments Limited. Each sample was measured more than 3 times. The variations of free volume and depth profile of these membranes were studied by the slow beam positron annihilation spectroscopy–Doppler broadening energy spectroscopy (DBES). One of the characteristic parameters of DBES, R parameter, defined as the ratio of the total counts from

Table 1 Spinning parameters for the polymeric HF substrates used in this worka. Spinning conditions

Torlons HF substrate

Ultems HF substrate

PES HF substrate

Dope formulation Bore fluid Dope flow rate (ml/min) Bore fluid flow rate (ml/min) Air gap (cm) Take up speed (m/min)

Torlons /PEG/NMP (17/15/68 wt%) NMP/water (95/5 wt%) 4.0 3.0 5.0 15

Ultems/ethanol/NMP (23/5/72 wt%) NMP/butanol (95/5 wt%) 4.0 3.0 5.0 15

PES/PEG/water/NMP (20/25/5/50 wt%) NMP/water (90/10 wt%) 6.0 4.0 3.0 22

a

All HFs were spun at ambient temperature around 23 1C.

Fig. 2. The scheme of procedures for (a) aldehyde modification and (b) HPEI mediated aldehyde modification of HF substrate. Note that if there is no α-H in the aldehyde molecule, the intermediated carbinolamines cannot be dehydrated to give enamines, but form derivatives of amine, as described in literatures [19,33].

200

D. Hua, T.-S. Chung / Journal of Membrane Science 492 (2015) 197–208

ortho-positronium 3γ (o-3γ) annihilation to the total counts from the 511 keV peak region (due to 2γ annihilation) [34], was employed as a qualitative measure of the free volume and the existence of pores in polymeric membranes. Moreover, to quantitatively determine the selective layer thickness, the resultant R parameter data were fitted against positron incident energy using the well-established VEPFIT computer program with a three-layer model [35]. The chemical structures of the HFs before and after the aldehyde modification were analyzed by using Fourier transform infrared spectroscopy (FTIR, Perkin-Elmer Spectrum 2000 FTIR spectrometer) under the attenuated total reflectance mode. Besides, X-ray photoelectron spectroscopy (XPS, Kratos AXIS His spectrometer) was employed to characterize the variation of N element-containing functional groups at the surface of HFs before and after chemical modifications. The binding energy of C 1s at 284.5 eV was taken as the reference.

The membrane performance was evaluated by a lab-scale pervaporation apparatus described elsewhere [36]. The IPA/water feed solution of 2 L was circulated around the shell side of the HF module at a flow rate of 30 L/h. The permeable components transport across the HF membranes under a vacuum pressure of lower than 1 mbar at the lumen side of the membrane. After the system was stabilized for 2 h, the permeate samples were collected by a cold trap immersed in liquid nitrogen, weighted and then analyzed by gas chromatography (Hewlett-Packard GC 7890, HP-INNOWAX column, thermal conductivity detector). At least three permeate samples were collected to ensure the reliability of the results and at least two identical HF membranes were tested for each pervaporation condition. Afterwards, the average value was reported as the pervaporation performance. The flux J was defined as Q At

ð1Þ

Wherein, Q is the weight of the collected permeate sample, A stands for the effective membrane surface area, and t is the time interval during the sample collection. The separation factor is calculated by the following equation:

β¼

α ¼ P~ i =P~ j

ð4Þ

where the subscripts i and j refer to water and IPA, respectively, in this work.

3. Results and discussion

2.6. Pervaporation experiments



where P~ i and P i are the mole-based permeance and permeability of component i, respectively. l is the membrane thickness, J i , M i , γ i , and psat are the flux, molecular weight, activity coefficient, and the i saturated vapor pressure of component i, respectively. xi and yi refer to the mole fractions of component i in the feed and permeate, respectively. pp, the total pressure at the permeate side, was assumed to be zero due to the vacuum condition. Both γ i and psat were calculated using the AspenTech Process Modelling softi ware (version 7.2) based on the Wilson equation and the Antoine equation, respectively. Thereafter, the mole-based selectivity of the membrane (α) is obtained from

ywi =ywj xwi =xwj

ð2Þ

where ywi and ywj are the weight fractions of water and IPA in the permeate, respectively, while xwi and xwj refer to the weight fractions of water and IPA in the feed, respectively. Since permeance and selectivity can manifest the intrinsic properties of asymmetric membranes better than flux and separation factor [37–39], the mole-based permeance is calculated by Eq. (3) according to the solution–diffusion model [39,40]: P Ji  P~ i ¼ i ¼ p l M i xi γ i psat i  yi p

ð3Þ

Table 2 The separation performance of Torlons HFs before and after modification for aqueous IPA (85/15 wt% IPA/water) dehydration at 50 1C. Membrane

Water in permeate (wt%)

Flux Separation factor (gm  2 h  1) (water/IPA)

Torlons HF substrate GA modified Torlons HF HPEI coated Torlons HF HPEI mediated GA modified Torlons HF

89.8 97.5 94.9 99.3

3947 1744 2085 1521

50 225 108 791

3.1. Torlons HF membranes modified by GA and HPEI mediated GA methods The pervaporation performance of the pristine Torlons HF substrate was studied for the dehydration of an aqueous IPA solution (i.e., 85/15 wt% IPA/water) at 50 1C. As shown in Table 2, its separation factor is only 50 because the unmodified HF substrate is not defect-free, as observed by FESEM on its outer surface in Fig. 3(a). After individual GA and HPEI modifications, the separation factors are greatly enhanced but the fluxes of the resultant HFs drop. This is normal because both modifications reduce the surface defects on Torlons HF substrates, which is clearly observed from the outer surface and cross-section morphologies in Fig. 3(a)–(c). Interestingly, the HPEI modified HF has a higher flux but a lower separation factor than the GA modified one. This may be attributed to the fact that the former has higher hydrophilicity than the latter (i.e., contact angle: 67.4 73.81 vs. 84.8 73.21) because HPEI possesses abundant hydrophilic amino groups. Encouragingly, the separation factor of the HPEI pretreated and then GA modified HF (referred to as the HPEI mediated GA modified HF thereafter) can be further raised to 791 with a minor flux drop as compared with HFs modified by GA or HPEI alone. As a result, the permeate contains almost pure water of  99.3 wt%. In addition, the contact angle of the resultant HF is 72.1 74.41 (Fig. 3 (d)) which is between those HFs modified by GA and HPEI, separately. Clearly, an interpenetration layer between HPEI and GA molecules has been formed on the Torlons outer surface with synergic effects for water transport across the membrane. Not only does it have a less defective selective layer to enhance the separation factor but also possess greater hydrophilicity to augment the solubility selectivity of water/IPA during the penetrant transports. The differences in dense selective layer among these HFs were further verified by DBES. Fig. 4 shows R parameter curves of HFs as a function of positron incident energy, which determines the penetration depth of positrons into the membrane surface [34,41]. For all the HF membranes, the R parameter near the surface is observed to decrease rapidly with increasing positron incident energy, which is typical due to the back diffusion and scattering of positronium. With an increase in positron incident energy to around 1.2–3.0 keV, the R parameter quickly reaches the minimum value, which can be interpreted as the dense selective layer of the HF membranes [41]. Thereafter, the R parameter starts to increase as the positron incident energy increases, indicating the evolution of large pores in the membrane structure and the transition from the dense layer structure to the porous substrate structure. From

D. Hua, T.-S. Chung / Journal of Membrane Science 492 (2015) 197–208

201

Fig. 3. Water contact angles and FESEM images of the outer surfaces and cross-sections of (a) Torlons HF substrate, and (b–d) HF membranes modified by GA, HPEI and HPEI mediated GA, respectively. Table 3 R parameters and selective layer thicknesses of Torlons HF substrate and HFs modified by HPEI, GA, and HPEI mediated GAa. Sample

R1

L1 (nm)

Torlons HF substrate HPEI treatment GA modification HPEI mediated GA modification

0.4232 7 0.0003 0.42117 0.0004 0.42077 0.0003 0.41827 0.0004

997 8 3077 37 3727 50 3587 55

a R1 and L1 are the R parameter and thickness of the first layer, respectively (i.e. the selective layer), estimated by fitting the resultant R parameter data against positron incident energy using the VEPFIT computer program with a threelayer model.

Fig. 4. R parameter as a function of position incident energy of Torlons HF substrates, and HFs modified by HPEI, GA, and HPEI mediated GA.

the comparison in Fig. 4, it can be seen the HF modified by HPEI mediated GA has the smallest value of the minimum R parameter, suggesting it has the densest selective layer. Moreover, its increasing trend of R parameter starts the latest, implying it owns the thickest selective layer. Furthermore, to quantitatively analyze the results, the R parameters are fitted by the VEPFIT program with a three-layer model. The fitted R values are also plotted in Fig. 4, which achieve a good match with the experimental data. Table 3 summarizes the fitted R parameter and thickness of the selective layer (R1 and L1). As shown, the Torlons HF substrate has the largest R1 and smallest L1, indicating it possesses a loose and thin selective layer, and thus leads to its high

flux and low separation factor. On the contrary, the other three modified HFs show significant thickness increments, which cause their flux declines. Moreover, the lowest R1 value of the HPEImediated GA modified HF indicates that it has the densest selective layer and the highest separation factor. Fig. 5 compares the outer surface morphology of pristine and modified Torlons HFs measured by AFM. It is observed that the outer surface becomes smoother after either GA or HPEI modification. However, the surface roughness increases from 20.0 to 26.3 nm after the HPEI mediated GA modification because of forming sharp mountain-like nodules on its surface. Nevertheless, the concaves on the surface of the Torlons HF disappear and become smoother, which could be observed from the top view of AFM images. These nodules might be generated by the interpenetration reaction between the primary amino groups of HPEI and the aldehyde groups of GA. To confirm this, both unmodified and modified Torlons HF membranes were further characterized by FTIR and XPS.

202

D. Hua, T.-S. Chung / Journal of Membrane Science 492 (2015) 197–208

Fig. 5. 3-D AFM images (side and top views) of the outer surface morphologies of (a) Torlons HF substrate, and (b–d) HF membranes modified by GA, HPEI and HPEI mediated GA, respectively.

As shown in Fig. 6, two kinds of functional groups are present in the FTIR spectrum of the Torlons HF substrate, namely, imide groups and amide groups. The two peaks of the imide group appear at around 1770 cm  1 and 1315 cm  1 corresponding to the C ¼O stretching and C–N stretching, respectively, while the peaks of the amide group occur at around at around 1650 cm  1 and 1540 cm  1 representing the C ¼O stretching and N–H bending, respectively [42–44]. The HF modified by GA shows an enhanced

stretching of C–N at around 1315 cm  1. Since the HPEI coating offers free primary amino groups on the substrate, a broad peak around 3400–3500 cm  1 is observed on the HPEI modified HF due to the N–H stretching band [45]. In contrast, the HPEI mediated GA modification introduces a new –N ¼C peak at around 1615– 1710 cm  1 [46], but this peak is overlapped with the peak for C¼ O stretching. Therefore, XPS was adopted to further characterize the modified Torlons HFs.

D. Hua, T.-S. Chung / Journal of Membrane Science 492 (2015) 197–208

203

Fig. 6. The FTIR spectra of Torlons HF substrates and Torlons HF membranes modified by GA, HPEI and HPEI mediated GA.

High resolution N 1s core level XPS spectra of the Torlons HF substrate and chemical modified HF membranes were compared in Fig. 7. The Torlons substrate has a single peak at binding energy of 399.8 eV, which is ascribed to N atoms of N–C ¼O in amide and imide groups of Torlons. After the GA modification there is still a single peak representing N–C ¼O. Nevertheless, the successful grafting of GA on the Torlons substrate is evidenced by the changes of surface element concentrations. The O, N, C mass ratios change from the initial values of 19.5%, 6.6%, and 73.9% on the unmodified HF substrate to 22.5%, 5.4%, and 72.1%, respectively, after the GA modification. In the case of the HPEI modified HF, Fig. 7(c) shows a strong peak at binding energy of 398.2 eV due to the introduction of C–NH2 from HPEI [47]. The binding energy of N atoms in N–C ¼O is higher than that in C–NH2 because the former has a stronger electron-withdrawing environment induced by – C ¼O. Based on the ratio of N species in different states, the N atoms in the form of C–NH2 cover 82% of the total surface, indicating a successful HPEI coating on the HPEI modified HF. Interestingly, the surface compositions relative to the N species alter a lot after the HPEI mediated GA modification. As depicted in Fig. 7(d), a new peak emerges at a binding energy of 401.0 eV besides the N–C ¼ O peak (399.8 eV) and C–NH2. This new C ¼N–C peak is formed via the imine condensation reaction between primary amino groups of HPEI and aldehyde groups of GA. Quantitative XPS analyses on the HPEI mediated GA modified HF indicate that the atomic percentages of surface N in the C–NH2, N– C ¼O, C ¼ N–C are 38.4%, 35.7% and 25.9%, respectively. 3.2. Chemical modifications by other aldehydes on Torlons HF substrates To verify the general application of aldehyde modification methods, several other aldehydes such as BA, GL and ICA were used to modify the Torlons HF substrate. As shown in Table 4, all modified HFs

display much higher separation factors than the unmodified HF substrate, indicating that both aldehyde and HPEI mediated aldehyde modification methods are really effective. Especially, the HPEI mediated aldehyde modifications can further bring the separation factor higher than 300 and maintain the flux of around 2000 g  2 h  1 for the dehydration of aqueous IPA (85 wt%) solutions at 50 1C. Consistent with the surface change after the HPEI mediated GA modification shown in Fig. 3, the outer surfaces of Torlons HFs modified by these three new aldehydes also become much denser than the pristine Torlons HF substrate, as illustrated in Fig. 8. High resolution N 1s core level XPS spectra also confirm the hypotheses that the imine condensation reaction between primary amino groups of HPEI and aldehyde groups forms the new C ¼N–C bond in all three cases. As illustrated in Fig. 9, the newly formed N atoms have a strong peak at binding energy of around 400.8 eV, while the N atoms with binding energies of  399.4 eV and  398.5 eV belong to N–C ¼O in the Torlons HF substrate and C– NH2 in HPEI, respectively. Likewise, quantitative XPS analyses indicate that the atomic percentages of surface N in C ¼ N–C are 29.7%, 33.7% and 25.2% after the BA, GL and ICA modifications, respectively. 3.3. Universal applications of the HPEI mediated aldehyde modification method to other HF membranes Even though the aldehyde modification can only be applicable to polymeric membranes containing amino or hydroxide groups, the HPEI mediated aldehyde modification may break this limitation because the HPEI pre-treatment can inherently graft primary amino groups onto the membrane surface. To prove our hypotheses, Ultems polyetherimide and PES HF substrates without any amino or hydroxide groups were chosen and studied. Table 5 shows the separation performance of Ultems and PES HF substrates as well as their modified ones. The Ultems and PES HF

204

D. Hua, T.-S. Chung / Journal of Membrane Science 492 (2015) 197–208

Fig. 7. N 1s XPS spectra of (a) Torlons HF substrates and (b–d) Torlons HF membranes modified by GA, HPEI, and HPEI mediated GA, respectively.

Table 4 The separation performance of Torlons HF substrates and HFs modified by various aldehydes or HPEI mediated aldehydes for aqueous IPA (85/15 wt% IPA/water) dehydration at 50 1C. Modification

Water in permeate (wt%)

Flux (gm  2 h  1)

Separation factor (water/IPA)

Torlons HF substrate BA modification HPEI mediated BA modification GL modification HPEI mediated GL modification ICA modification HPEI mediated ICA modification

89.8 97.4 98.2

3947 2078 2285

50 221 316

96.7 98.6

1989 2055

171 413

96.8 98.7

2282 1978

172 432

substrates are very porous and have larger surface pores than the Torlons HF (see Supplementary material, Figs. S1–S3). As a result, they have high fluxes and low separation factors. After the HPEI modification, both HFs show enhanced separation factors because HPEI not only seals parts of surface defects but also improves the hydrophilicity [27–30]. The HPEI modified PES HF still has a relatively low separation factor because the PES substrate is very porous and has almost no selectivity for IPA/water separation. After the HPEI mediated GA modification, the separation factors of both HF membranes were greatly enhanced. For the modified Ultems HF, it has a satisfactory pervaporation performance with a separation factor of 209 and a total flux of 2192 g  2 h  1. For the modified PES HF, it exhibits a greatly improved separation factor of 38 and a very high total flux of 5008 g  2 h  1. This separation factor can be further improved if better PES substrates with

smaller pores are used. The changes of outer-surface morphologies of Ultems and PES HFs, after the HPEI mediated GA modification, are shown in the Supplementary material (Fig. S4). The successes of these two examples further prove that the HPEI mediated aldehyde modification is universally applicable to modify polymeric HF membranes with enhanced pervaporation performance. 3.4. The effects of operation conditions on IPA dehydration performance To expand the potential of the HPEI mediated aldehyde modification for alcohol dehydration, the effects of feed temperature and composition as well as long-term stability on IPA dehydration were investigated and the HPEI mediated GA modified Torlons HF was chosen for the study. 3.4.1. Effects of feed temperature on pervaporation performance As illustrated in Fig. 10, permeate flux increases dramatically, while separation factor drops with an increase in feed temperature. However, the water concentration in the permeate remains as high as 98% at 80 1C. The flux increase at high feed temperatures is ascribed to two factors [3,10,48–50]: (1) increased driving forces for IPA and water transports, and (2) enhanced chain motion and expanded chain–chain distance. Since IPA has a larger molecular size than water, the expanded chain–chain distance at high temperatures will facilitate IPA transport more effectively than water transport. As a result, the separation factor drops. Table 6 shows the apparent activation energies of water and IPA transports across the membrane by plotting their fluxes against the reciprocal of temperature using the Van't Hoff–Arrhenius equation. The

D. Hua, T.-S. Chung / Journal of Membrane Science 492 (2015) 197–208

205

Fig. 8. FESEM images of outer surface of (a) Torlons HF substrate, and (b–d) Torlons HF membranes modified by HPEI mediated BA, HPEI mediated GL and HPEI mediated ICA, respectively.

Fig. 9. N1s XPS spectra of Torlons HF membranes modified by HPEI mediated BA, HPEI mediated GL, as well as HPEI mediated ICA.

Table 5 The separation performance of other polymeric HFs modified by HPEI mediated GA for aqueous IPA (85/15 wt% IPA/water) dehydration at 50 1C. Polymeric HF material used

Modification Water concentration in permeate (wt%)

Flux Separation (gm  2 h  1) factor

Ultem

Nil 66.5 HPEI 93.7 HPEI 97.2 mediated GA

4351 2634 2192

11 86 202

PES

Nil 16.9 HPEI 45.8 HPEI 86.8 mediated GA

41,430 11,989 5008

1 5 38

Table 6 The calculated activation energies of total flux, water flux, and IPA flux for the dehydration of IPA/water (85/15 wt%) mixtures. Flux

EJ (kJ/mol)

Total flux Water flux IPA flux

37.4 37.0 66.5

Fig. 11. The effects of feed composition on the separation performance of HPEI mediated GA modified Torlons HF for aqueous IPA dehydration at 50 1C.

Fig. 10. The effects of feed temperature on separation performance of HPEI mediated GA modified Torlons HF for aqueous IPA (85/15 wt% IPA/water) dehydration.

water transport has a much lower activation energy than that of IPA, indicating that this modified HF membrane is highly waterselective.

206

D. Hua, T.-S. Chung / Journal of Membrane Science 492 (2015) 197–208

3.4.2. Effects of feed composition on pervaporation performance Fig. 11 depicts the effects of feed composition on separation performance. Three trends can be observed: (1) total flux decreases with increasing IPA concentration, (2) separation factor also increases with IPA concentration but reaches the highest value at the feed IPA concentration of 85 wt% then drops, and (3) water content in the permeate is always higher than 95 wt% even at the feed IPA concentration of 95 wt%.

The rise of feed IPA concentration results in total flux decline is a typical phenomenon in pervaporative dehydration because of less water-induced membrane swelling at higher IPA concentrations and the lower driving force for water permeation [17,25,49–51]. The water content in the permeate is constantly high (99.3 wt%) when the feed IPA concentration varies from 70 to 85 wt% but drops slightly at higher IPA concentrations. This is due to the fact that a higher IPA feed concentration results in a greater driving force for isopropanol transport across the membrane [25]. However, the water content in the permeate always stays higher than 95 wt% even at the feed IPA concentration of 95 wt%, indicating the modified HF is chemically stable and has a high selectivity for IPA/water separation. 3.4.3. Long-term stability for IPA dehydration Fig. 12 displays the long-term performance of the HPEI mediated GA modified Torlons HF membrane based on a labscale pervaporation experiment continuously run for more than 200 h at 50 1C. Both the flux and water concentration in the permeate stay almost constant, evidencing good chemical and mechanical stability for dehydration applications. 3.5. Benchmarking for the HF membranes modified by HPEI mediated aldehyde

Fig. 12. Long-term pervaporation performance of the HPEI mediated GA modified Torlons HF for the dehydration of aqueous IPA solutions (IPA/water 85/15 wt%) at 50 1C.

Table 7 compares the pervaporation performance of the HPEI mediated aldehyde modified HF membrane with other polymeric HF membranes for pervaporative IPA dehydration available in literatures. The newly molecularly designed HPEI mediated aldehyde modified Torlons HF possesses both high permeance and high selectivity

Table 7 Performance benchmarking of the HPEI mediated aldehyde modified HF membranes with other polymeric HF membranes for pervaporative IPA dehydration in literatures. Polymeric HF membrane

Temperature (1C)

Feed (IPA wt%)

Flux Separation factor (gm  2 h  1) (water/IPA)

Mole-based water permeance (mol m  2 h  1 kPa  1)

Mole-based selectivity (water/IPA)

Ref.

Heat treated polyacrylonitrile (PAN) HF Chitosan/polyacrylonitrile (PAN) composite HF Cellulose/polysulfone dual-layer HF Torlons single-layer HF Heat treated Torlons/Ultems dual-layer HF Torlons single-layer HF PXD crosslinked Torlons/P84 blend single-layer HF P84 co-polyimide HF Heat treated P84/PES dual layer HF (HPEI/GA)4/PAN composite HF MPD-TMC polyamide TFC/Torlons dual-layer HF HPEI-TMC polyamide TFC/Torlons dual-layer HF HPEI-HGOTMS polyamide TFC/ Ultems HF HPEI/MPD-TMC polyamide TFC/ Ultems tri-bore HF Nexar Block Copolymer/Ultems HPEI mediated GA modified Torlons HF HPEI mediated BA modified Torlons HF HPEI mediated GL modified Torlons HF HPEI mediated ICA modified Torlons HF HPEI mediated GA modified Ultems HF

25

90

186

1116

5.5

1022

[49]

25

90

145

2991

4.3

2740

[50]

25

95

4

94,981

0.2

66,731

[52]

60 60

85 85

801 765

7 1944

1.7 3.0

8 2199

[5] [5]

60 60

85 85

850 1000

15 185

2.3 3.8

15 222

[53] [53]

60 60 50 50

85 85 95 85

883 570 426 1374

10,585 125 18,981 53

3.4 2.1 4.7 7.8

13,227 145 13,626 62

[36] [32] [31] [43]

50

85

1280

624

8.0

729

[43]

50

85

3519

278

21.8

324

[54]

50

85

2647

261

15.8

294

[30]

50 50

85 85

2060 1521

509 791

12.9 9.5

595 899

50

85

2285

316

14.2

353

50

85

2055

413

12.8

461

50

85

1978

432

12.3

488

50

85

2192

202

13.4

230

[55] This work This work This work This work This work

D. Hua, T.-S. Chung / Journal of Membrane Science 492 (2015) 197–208

among the reported data, indicating their great potential for industrial applications.

4. Conclusion In summary, not only we have demonstrated that aldehydes can directly modify Torlons polyamide-imide hollow fibers for improved dehydration of IPA solutions, but more importantly, a novel universal modification method based on the synergistic combination of HPEI and aldehyde has been discovered and successfully applied to various polymeric membranes, such as Ultems and PES, with enhanced pervaporative dehydration. It was found that the HPEI pre-treatment can mitigate surface defects and enhance membrane hydrophilicity toward water transport. In addition, the reaction between HPEI and aldehydes at the outer surface of the HF membranes enhances selectivity and chemical resistance. The HPEI mediated aldehyde modified hollow fibers possess satisfactory separation performance in wide ranges of feed temperature and composition ranges, and also exhibit a good long-term stability, suggesting their potential for industrial IPA dehydration applications.

NomenclaturesAbbreviations AFM BA DBES FESEM FTIR GA GL HF HPEI ICA IPA NMP PEG PES XPS

atomic force microscope benzaldehyde Doppler broadening energy spectroscopy field-emission scanning electron microscope Fourier transform infrared spectroscopy glyoxylic acid glucose hollow fiber hyper-branched polyethyleneimine imidazolate-2-carboxyaldehyde isopropanol N-methyl-2-pyrrolidone polyethylene glycol polyethersulfone X-ray photoelectron spectroscopy

Symbols A EJ J Ji l Pi pp psat i P~ i ; P~ j Q t xi xwi ; xwj yi ywi ; ywj

α β γi

effective surface area of the membrane (m2) apparent activation energy of flux (kJ mol  1) flux of the membrane (g m  2 h  1) permeation flux of component i (g m  2 h  1) membrane thickness (m) mole-based permeability of component i (mol m  1 h  1 kPa  1) the total pressure at the permeate side (kPa) saturated vapor pressure of component i (kPa) mole-based permeance of component i or j (mol m  2 h  1 kPa  1) weight of the sample collected as permeate of the membrane (g) time interval during the sample collection (h) mole fraction of the component i in the permeate weight fraction of the component i or j in the feed mole fraction of the component i in the permeate weight fraction of the component i or j in the permeate mole-based selectivity (water/IPA) separation factor (water/IPA) activity coefficient of component i

207

Acknowledgments The authors are grateful to the National Research Foundation, Prime Minister's Office, Singapore for funding this research under its Competitive Research Program for the project entitled, “New Biotechnology for Processing Metropolitan Organic Wastes into Value-Added Products” (CRP Award no. NRF-CRP 5-2009-5 (NUS Grant no. R-279-000-311-281)). Special thanks are due to Dr. Yee Kang Ong, Dr. Kuo-Sung Liao, and Dr. Yu Pan Tang for their valuable suggestions.

Appendix A. Supporting information Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.memsci.2015.05. 056.

References [1] L.M. Vane, A review of pervaporation for product recovery from biomass fermentation processes, J. Chem. Technol. Biotechnol. 80 (2005) 603–629. [2] P.D. Chapman, T. Oliveira, A.G. Livingston, K. Li, Membranes for the dehydration of solvents by pervaporation, J. Membr. Sci. 318 (2008) 5–37. [3] X. Feng, R.Y.M. Huang, Liquid separation by membrane pervaporation: a review, Ind. Eng. Chem. Res. 36 (1997) 1048–1066. [4] P. Shao, R.Y.M. Huang, Polymeric membrane pervaporation, J. Membr. Sci. 287 (2007) 162–179. [5] L.Y. Jiang, Y. Wang, T.S. Chung, X.Y. Qiao, J.Y. Lai, Polyimides membranes for pervaporation and biofuels separation, Prog. Polym. Sci. 34 (2009) 1135–1160. [6] G. Liu, W. Wei, W. Jin, Pervaporation membranes for biobutanol production, ACS Sustain. Chem. Eng. 2 (2014) 546–560. [7] Y.C. Sue, J.W. Wu, S.E. Chung, C.H. Kang, K.L. Tung, K.C.W. Wu, F.K. Shieh, Synthesis of hierarchical micro/mesoporous structures via solid–aqueous interface growth: zeolitic imidazolate framework-8 on siliceous mesocellular foams for enhanced pervaporation of water/ethanol mixtures, ACS Appl. Mater. Interfaces 6 (2014) 5192–5198. [8] N.L. Le, T.S. Chung, High-performance sulfonated polyimide/polyimide/polyhedral oligosilsesquioxane hybrid membranes for ethanol dehydration applications, J. Membr. Sci. 454 (2014) 62–73. [9] K. Hunger, N. Schmeling, H.B.T. Jeazet, C. Janiak, C. Staudt, K. Kleinermanns, Investigation of cross-linked and additive containing polymer materials for membranes with improved performance in pervaporation and gas separation, Membranes 2 (2012) 727–763. [10] Y.M. Lee, S.Y. Nam, S.Y. Ha, Pervaporation of water/isopropanol mixtures through polyaniline membranes doped with poly(acrylic acid), J. Membr. Sci. 159 (1999) 41–46. [11] B. Bolto, T. Tran, M. Hoang, Z. Xie, Crosslinked poly(vinyl alcohol) membranes, Prog. Polym. Sci. 34 (2009) 969–981. [12] K. Vanherck, G. Koeckelberghs, I.F.J. Vankelecom, Crosslinking polyimides for membrane applications: a review, Prog. Polym. Sci. 38 (2013) 874–896. [13] Y. Xu, C. Chen, J. Li, Experimental study on physical properties and pervaporation performances of polyimide membranes, Chem. Eng. Sci. 62 (2007) 2466–2473. [14] X.Y. Qiao, T.S. Chung, Fundamental characteristics of sorption, swelling, and permeation of P84 co-polyimide membranes for pervaporation dehydration of alcohols, Ind. Eng. Chem. Res. (2005) 8938–8943. [15] D.W. Mangindaan, N.M. Woon, G.M. Shi, T.S. Chung, P84 polyimide membranes modified by a tripodal amine for enhanced pervaporation dehydration of acetone, Chem. Eng. Sci. 122 (2015) 14–23. [16] J.G. Frick, R.J. Harper, Crosslinking cotton cellulose with aldehydes, J. Appl. Polym. Sci. 27 (1982) 983–988. [17] J. Yu, C.H. Lee, W.H. Hong, Performances of crosslinked asymmetric poly(vinyl alcohol) membranes for isopropanol dehydration by pervaporation, Chem. Eng. Process. 41 (2002) 693–698. [18] J.E. McMurry, Organic Chemistry, Mary Finch, 2011. [19] M.A. Sprung, A summary of the reactions of aldehydes with amines, Chem. Rev. 26 (1940) 297–338. [20] A. Huang, J. Caro, Covalent post-functionalization of zeolitic imidazolate framework ZIF-90 membrane for enhanced hydrogen selectivity, Angew. Chem. Int. Ed. 50 (2011) 4979–4982. [21] P. Hernández-Muñoz, R. Villalobos, A. Chiralt, Effect of cross-linking using aldehydes on properties of glutenin-rich films, Food Hydrocolloids 18 (2004) 403–411. [22] N.A. Danilov, N.Y. Ignatieva, E.N. Iomdina, S.A. Semenova, G.N. Rudenskaya T.E. Grokhovskaya, V.V. Lunin, Stabilization of scleral collagen by glycerol aldehyde cross-linking, Biochim. Biophys. Acta, Gen. Subj. 1780 (2008) 764–772.

208

D. Hua, T.-S. Chung / Journal of Membrane Science 492 (2015) 197–208

[23] J. Kawahara, K. Ishikawa, T. Uchimaru, H. Takaya, Chemical cross-linking by glutaraldehyde between amino groups: its mechanism and effects, Polym. Modif. (1997) 119–131. [24] Y. Cui, H. Wang, H. Wang, T.S. Chung, Micro-morphology and formation of layer-by-layer membranes and their performance in osmotically driven processes, Chem. Eng. Sci. 101 (2013) 13–26. [25] A. Svang-Ariyaskul, R.Y.M. Huang, P.L. Douglas, R. Pal, X. Feng, P. Chen, L. Liu, Blended chitosan and polyvinyl alcohol membranes for the pervaporation dehydration of isopropanol, J. Membr. Sci. 280 (2006) 815–823. [26] D. Dieter, M. Erwin, Crosslinking Aqueous Polyurethanes with Formaldehyde, US 3384606 A, 1968. [27] S.P. Sun, T.A. Hatton, T.S. Chung, Hyperbranched polyethyleneimine induced cross-linking of polyamide-imide nanofiltration hollow fiber membranes for effective removal of ciprofloxacin, Environ. Sci. Technol. 45 (2011) 4003–4009. [28] J. Gao, S.P. Sun, W.P. Zhu, T.S. Chung, Polyethyleneimine (PEI) cross-linked P84 nanofiltration (NF) hollow fiber membranes for Pb2 þ removal, J. Membr. Sci. 452 (2014) 300–310. [29] G.M. Shi, T.S. Chung, Thin film composite membranes on ceramic for pervaporation dehydration of isopropanol, J. Membr. Sci. 448 (2013) 34–43. [30] D. Hua, Y.K. Ong, P. Wang, T.S. Chung, Thin-film composite tri-bore hollow fiber (TFC TbHF) membranes for isopropanol dehydration by pervaporation, J. Membr. Sci. 471 (2014) 155–167. [31] G. Zhang, L. Dai, S. Ji, Dynamic pressure-driven covalent assembly of inner skin hollow fiber multilayer membrane, AIChE J. 57 (2011) 2746–2754. [32] R.X. Liu, X.Y. Qiao, T.S. Chung, Dual-layer P84/polyethersulfone hollow fibers for pervaporation dehydration of isopropanol, J. Membr. Sci. 294 (2007) 103–114. [33] J.E. Fernandez, G.B. Butler, The reaction of secondary amines with formaldehyde, J. Org. Chem. 28 (1963) 3258–3259. [34] H. Chen, W.S. Hung, C.H. Lo, S.H. Huang, M.L. Cheng, G. Liu, K.R. Lee, J.Y. Lai, Y. M. Sun, C.C. Hu, R. Suzuki, T. Ohdaira, N. Oshima, Y.C. Jean, Free-volume depth profile of polymeric membranes studied by positron annihilation spectroscopy: layer structure from interfacial polymerization, Macromolecules 40 (2007) 7542–7557. [35] A. van Veen, H. Schut, J. de Vries, R.A. Hakvoort, M.R. Ijpma, Analysis of positron profiling data by means of “VEPFIT”, AIP Conf Proc 218 (1991) 171–198. [36] R. Liu, X. Qiao, T.S. Chung, The development of high performance P84 copolyimide hollow fibers for pervaporation dehydration of isopropanol, Chem. Eng. Sci. 60 (2005) 6674–6686. [37] J.G. Wijmans, Process performance ¼ membrane properties þoperating conditions, J. Membr. Sci. 220 (2003) 1–3. [38] W.F. Guo, T.S. Chung, T. Matsuura, Pervaporation study on the dehydration of aqueous butanol solutions: a comparison of flux vs. permeance, separation factor vs. selectivity, J. Membr. Sci. 245 (2004) 199–210. [39] R.W. Baker, J.G. Wijmans, Y. Huang, Permeability, permeance and selectivity: a preferred way of reporting pervaporation performance data, J. Membr. Sci. 348 (2010) 346–352. [40] J.G. Wijmans, R.W. Baker, The solution–diffusion model: a review, J. Membr. Sci. 107 (1995) 1–21.

[41] J. Zuo, T.S. Chung, Design and synthesis of a fluoro-silane amine monomer for novel thin film composite membranes to dehydrate ethanol via pervaporation, J. Mater. Chem. A 1 (2013) 9814. [42] Y. Wang, T.S. Chung, H. Wang, Polyamide-imide membranes with surface immobilized cyclodextrin for butanol isomer separation via pervaporation, AIChE J. 57 (2011) 1470–1484. [43] J. Zuo, Y. Wang, S.P. Sun, T.S. Chung, Molecular design of thin film composite (TFC) hollow fiber membranes for isopropanol dehydration via pervaporation, J. Membr. Sci. 405–406 (2012) 123–133. [44] F.J. Holler, D.A. Skoog, S.R. Crouch, Principles of Instrumental Analysis, Belmont, 2007. [45] X. Cai, M. Lin, S. Tan, W. Mai, Y. Zhang, Z. Liang, Z. Lin, X. Zhang, The use of polyethyleneimine-modified reduced graphene oxide as a substrate for silver nanoparticles to produce a material with lower cytotoxicity and long-term antibacterial activity, Carbon 50 (2012) 3407–3415. [46] N.B. Colthup, Spectra–structure correlations in the infra-red region, J. Opt. Soc. Am. 40 (1950) 397–400. [47] A.P. Dementjev, A. de Graaf, M.C.M. van de Sanden, K.I. Maslakov A.V. Naumkin, A.A. Serov, X-Ray photoelectron spectroscopy reference data for identification of the C3N4 phase in carbon–nitrogen films, Diam. Relat. Mater. 9 (2000) 1904–1907. [48] C.K. Yeom, K.H. Lee, Characterization of sodium alginate membrane crosslinked with glutaraldehyde in pervaporation separation, J. Appl. Polym. Sci. 67 (1998) 209–219. [49] H.A. Tsai, Y.S. Ciou, C.C. Hu, K.R. Lee, D.G. Yu, J.Y. Lai, Heat-treatment effect on the morphology and pervaporation performances of asymmetric PAN hollow fiber membranes, J. Membr. Sci. 255 (2005) 33–47. [50] H.A. Tsai, W.H. Chen, C.Y. Kuo, K.R. Lee, J.Y. Lai, Study on the pervaporation performance and long-term stability of aqueous iso-propanol solution through chitosan/polyacrylonitrile hollow fiber membrane, J. Membr. Sci. 309 (2008) 146–155. [51] K. Koch, A. Górak, Pervaporation of binary and ternary mixtures of acetone, isopropyl alcohol and water using polymeric membranes: experimental characterisation and modelling, Chem. Eng. Sci. 115 (2014) 95–114. [52] Z. Mao, X. Jie, Y. Cao, L. Wang, M. Li, Q. Yuan, Preparation of dual-layer cellulose/polysulfone hollow fiber membrane and its performance for isopropanol dehydration and CO2 separation, Sep. Purif. Technol. 77 (2011) 179–184. [53] M.M. Teoh, T.S. Chung, K.Y. Wang, M.D. Guiver, Exploring Torlon/P84 copolyamide-imide blended hollow fibers and their chemical cross-linking modifications for pervaporation dehydration of isopropanol, Sep. Purif. Technol. 61 (2008) 404–413. [54] J. Zuo, Y. Wang, T.S. Chung, Novel organic–inorganic thin film composite membranes with separation performance surpassing ceramic membranes for isopropanol dehydration, J. Membr. Sci. 433 (2013) 60–71. [55] J. Zuo, G.M. Shi, S. Wei, T.S. Chung, The development of novel Nexar block copolymer/Ultem composite membranes for C2–C4 alcohols dehydration via pervaporation, ACS Appl. Mater. Interfaces 6 (2014) 13874–13883.