EFFECT OF DENSE CO2 ON POLYMERIC REVERSE OSMOSIS AND NANOFILTRATION MEMBRANES AND PERMEATION OF MIXTURES OF MACAUBA OIL (Acrocomia aculeata) AND CO2

Journal of Membrane Science 481 (2015) 195–206

Contents lists available at ScienceDirect

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

Effect of dense CO2 on polymeric reverse osmosis and nanofiltration membranes and permeation of mixtures of macauba oil (Acrocomia aculeata) and CO2 Katia Rezzadori a, Josamaique Gilson Veneral a, Júlia Carolina Medeiros Silveira a, Frederico Marques Penha a, José Carlos Cunha Petrus a, Pedro Prádanos b, Laura Palacio b, Antonio Hernández b, Marco Di Luccio a,n a b

EQA-CTC/UFSC, Chemical and Food Engineering Department, Federal University of Santa Catarina, C.P. 476, CEP 88040-900 Florianopolis, SC, Brazil SMAP, Surfaces and Porous Materials Group, Faculty of Science, University of Valladolid, P. Belén 7, 47011 Valladolid, Spain

art ic l e i nf o

a b s t r a c t

Article history: Received 28 May 2014 Received in revised form 18 November 2014 Accepted 11 December 2014 Available online 5 January 2015

The performance of four commercial membranes upon subcritical and supercritical CO2 treatments have been characterised by different methods. The membranes were treated up to 8 h in static process, using two different conditions (18 MPa/313.15 K – supercritical and 8 MPa/293.15 K – subcritical). The possible changes in membrane characteristics were investigated by atomic force microscopy (AFM), contact angle and surface free energy, thermogravimetric analysis (TGA), scanning electronic microscopy (SEM). Membrane performance was also evaluated by measuring CO2 flux and macauba oil retention factors. Changes in membrane roughness and in contact angles were observed for all membranes after sub and supercritical CO2 treatments. Moreover, the surface free energy and the polar component showed a decrease after CO2 exposure, confirming an increase in surface hydrophobicity of membranes detected from contact angle results. This empowerment of hydrophobicity is associated mainly with the CO2– polymer interaction. Other intra and interchain effects should not affect the thermal stability leaving TGA results unchanged after CO2 exposure. ORAK and NP030 membranes showed high macauba oil retentions; 95% and 85%, respectively in supercritical condition. This latter revealed to cause important changes in the membranes due to the higher solubility of CO2 in polymeric matrix at this condition. However, the selectivity was not changed and it is possible to use commercial polymeric membranes in supercritical systems for CO2 regeneration and partial fatty acid fractioning. & 2015 Elsevier B.V. All rights reserved.

Keywords: Polymeric membranes Supercritical CO2 Membrane stability

1. Introduction Membrane technology has been recognised as a useful tool in many industrial applications such as water purification, gas and vapour separation, and fuel cells [1]. This technology is known due to its simple scale up, low energy demand compared to the more energy intensive conventional processes and because it can be carried out under moderate pressure and temperature conditions [2]. Recently, the use of dense (i.e. supercritical) CO2 has become a great alternative to replace organic solvents in the extraction of solute molecules, due to the tuneable dissolving power of CO2 [3,4]. One of the main issues in high-pressure applications is the recovery of the supercritical fluid, which represents the major part of the operational costs [5].

n

Correspondent author. Tel.: þ 55 48 3721 2529. E-mail address: [email protected] (M. Di Luccio).

http://dx.doi.org/10.1016/j.memsci.2014.12.010 0376-7388/& 2015 Elsevier B.V. All rights reserved.

Coupling membrane separation with supercritical CO2 extraction is a relatively new technology, and attempts to separate the extracted components from CO2, with benefits in terms of low energy consumption and compact process design compared with other technologies. Moreover, the use of membrane separation can lead to a reduction in pressure losses and recompression costs [1]. Furthermore, membrane technology offers a suitable option for continuous regeneration of supercritical fluids. Detailed studies about the use of commercial polymeric membranes in supercritical CO2 systems are still limited, although there are some instances of applications of this approach using commercial thin film membranes with supercritical CO2 [4,6,7]. Studies report that different membranes made of polyvinyl alcohol (PVA) and a polyamide copolymer (IPC) were available with a good CO2 permeability and offer possibilities for efficient regeneration of carbon dioxide while maintaining supercritical conditions [7]. Sarmento et al. [3] verified that the polymeric membranes used (DL, HL – Osmonics and NF – Dowfilmtec) showed a good mechanical resistance and good permeability values for pure

196

K. Rezzadori et al. / Journal of Membrane Science 481 (2015) 195–206

supercritical CO2 and for its mixture with cocoa extract. The polyphenol retention index for these membranes were all greater than 90% when subject to a differential pressure of 1 MPa. Also, for long-chain polyphenols, the retention was high. Nevertheless, thin film polyamide membranes were not considered stable under supercritical conditions by Akin and Temelli [8], especially upon the application of depressurisation, due to the stability of the membranes. The membrane structure was affected due to compaction, resulting in a lower CO2 flux and higher oleic acid retention at higher transmembrane pressures. The main motivation behind the research on membrane properties and stability in the field of supercritical extraction has been the strong interaction of CO2 with polymer materials at high pressures and better understanding of the way how such interactions would improve or worsen membrane performance. These interactions can lead to phenomena as the swelling of the polymeric matrix, plasticisation or even dissolution of membrane material and subsequent loss of morphological structure, causing changes in separation properties, mainly in the membrane selectivity, and/or mechanical resistance under pressure [7,9]. Therefore studies in this field can promote benefits to different industries and stimulate the development of new membranes to be used with supercritical CO2. Macauba is a Brazilian endemic palm tree species whose fruits are eatable and present high oil yields in its pulp and kernel. An important amount of oil is obtained from the coconuts of this palm, with productivity yields between 1500 and 5000 kg of oil per hectare per year, which is the second largest productivity after palm oil (Elaeis guineensis) [10]. Researches involving macauba and its by-products are increasing due to its high productivity and multiple possibilities such as producing cooking oil, substitute for hydrogenated fat and component in several candies, besides being an alternative on energy scenario. However, macauba productive chain is still emerging. The oil extraction is often performed mechanically and in a rudimentary way. One potential solution is the use of pressurised liquid carbon dioxide as a processing solvent, which is a non-flammable and non-toxic solvent. Oil solubility in subcritical and supercritical carbon dioxide varies as a function of the density and temperature. Both sub- and supercritical CO2 have a significantly lower viscosity than the oil and organic solvents [10,11] and they are less expensive than organic solvents. In this context, the aim of this study was to investigate the effect of supercritical and subcritical CO2 on those morphological and physicochemical changes in four commercial membranes (NF270, NP030 – nanofiltration; ORAK and BW30 – reverse osmosis) and investigate the membranes’ performance in the permeation and retention of macauba (Acrocomia aculeata) palm oil using subcritical and supercritical conditions.

2. Materials and methods 2.1. Material specifications Four commercial membranes were studied, 2 nanofiltration (NF) and 2 reverse osmosis (RO). The main characteristics of each membrane, according to manufacturer’s description, are shown in Table 1. The solvent consisted of commercial CO2 (99.95% – White Martins, Brazil). Raw macauba oil (Acrocomia aculeata), obtained through kernel mechanical extraction, was purchased from Cocal Special Oils Ltd. (Abaeté, Minas Gerais, Brazil). Oil was stored in 1 L amber glass flasks, in inert atmosphere (N2) at approximately 255.15 K, to minimise degradation.

2.2. Membrane exposure to pure subcritical and supercritical CO2 Assays to determine the effect of the contact of polymeric membranes with pure CO2 were performed in static mode, i.e., the operating system was filled with CO2 and let stand for 8 h at different pressure and temperature conditions – subcritical: 8 MPa/293.15 K; supercritical: 18 MPa/313.15 K. The schematic diagram of CO2 and membrane system is presented in Fig. 1. The system was pressurised using a syringe pump (Model D260, Teledyne ISCO Inc., Lincoln, NE). A dead-end type stainless steel membrane module (Ce) was used to accommodate flat sheet membranes. The effective membrane filtration area was 2.624  10  3 m2. CO2 flux was controlled by the pump controller to maintain the system pressure constant. The latter was monitored by a pressure gauge (P) (Novus, Model HUBA-511, Switzerland). The module bottom had two micrometric valves (V3 and V4) (Hoke, model 1315G4Y, USA) used to depressurise the system at the end of process. A flowmeter (F) (Quantim, model QMBM4L1A3A1B3A1KB1B1A1CA, USA) was used to monitor the CO2 flow. For the static mode assays, initially the internal pump chamber was filled with solvent (CO2). The valve V1 was slowly opened (0.1 MPa s  1) until the system reached the working pressure. The valve V2 remained opened during system pressurisation to promote equal pressure on both sides of the membrane; therefore avoiding mechanical damages to it. Valve V2 was closed after the work pressure was stable. Cell (Ce1) temperature was adjusted by a thermostatic bath (BT2) connected to the cell’s jacket and controlled by a temperature sensor (thermocouple). Then, the CO2 permeate flux was measured closing the valve V2 to build up the transmembrane pressure by adjusting the micrometric valve (V4). The differential pressure between the upstream and downstream was monitored by a pressure gauge (P) and Δp ¼1 MPa was used. The permeate flux was monitored for 60 min. Thereafter, the valve V4 was closed and the system remained for 8 h in contact with CO2 in static mode. At the end of the 8 h contact period, the permeate flux was measured again, using the same conditions established at the beginning of the assay. Finally, the valve V1 was closed to isolate the system whilst the valve V4 was slowly opened to depressurise the system (0.1 MPa s  1). 2.3. Membrane characterisation 2.3.1. Atomic force microscopy – AFM The membranes submitted to the supercritical and subcritical CO2 static exposition for 8 h were analysed by AFM. Images of the membranes were obtained by atomic force microscopy (AFM) using a Nanoscope IIIA Multimode (Veeco Metrology Inc., Santa Barbara, CA). The Tapping Mode technique was used according to Carvalho et al. [12] methodology. In this measurement mode, the cantilever where the tip is located oscillates with its natural frequency and the sample topography is obtained from the subsequent changes in the oscillation amplitude. Quantitative roughness analysis was performed using the Nanoscope Software, also according to the same authors. The surface roughness was studied by image statistical analysis with areas between (1  1) μm2 and (5  5) μm2. Each measurement was carried out three times to obtain a mean value of the Rq roughness (Eq. 1). Several scan areas were used in order to improve comparison of surface roughness for the different samples. vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u n u1 X ð1Þ Rq ¼ t ðZ  Z m Þ2 ni¼1 i where Rq is the root-mean-square roughness, Zm is the mean value of the tip-to-surface distance, Zi over a reference baseline (Z).

K. Rezzadori et al. / Journal of Membrane Science 481 (2015) 195–206

197

Table 1 Membrane characteristics according to manufacturers' data. Membrane

NF270

NP030

BW30

ORAK

Manufacturer Material MWCOe (Da) Pmax (MPa) Tmax (K) pH range Rejection (%)

Dow Filmtech Polyamide TFCd 200–300 4.1 318.15 2–11 497.0a

Microdyn Nadir Polyethersulfone 400 4.0 368.15 0–14 80–97b

Dow Filmtech Polyamide TFCd – 4.1 318.15 2–11 99.5c

Osmonics Polyamide TFC – 2.7 323.15 4–11 99c

a

Rejection in MgSO4 (298.15 K; 0.48 MPa). Rejection in Na2SO4 (293.15 K; 4 MPa). Rejection in NaCl (298.15; 1.55 MPa). d TFC: thin film composite. e MWCO: molar weight cut-off. b c

Fig. 1. Experimental unit used for CO2-membrane separation coupled system experiments. (SR) cylinder; (B1) syringe pump for CO2; (B2) pump for oil; BT1 and BT2 temperature-controlled bath; (V1–V4) micrometric valves; (CV) back pressure valves; (P) pressure gauge; (Ce) stainless steel membrane module; (F) flow metre; (T) thermocouple.

2.3.2. Contact angle and surface free energy The contact angle measurements were carried out by a contact angle and surface tension instrument (FTA 200, Virginia, USA). The contact angle was measured from digital pictures by using “adhoc” software of analysis. Drops of three different standard liquids, deionised water, formamide and diiodomethane, with volumes 5 μL, 4 μL and 0.8 μL, respectively, were added by a motor-driven syringe at room temperature. The average value of the angles from both sides of each drop is counted as one measurement. Four measurements were carried out for each sample. The presented data correspond to the final average value. For more direct quantitative information on the surfaces of the new and treated membranes, surface energies and surface energy components were calculated from contact angles. The Good–van Oss theory [13] was used to calculate the acid-base components of solid surface free energies [14]. In this approach, the work of adhesion of a liquid phase (W) onto a solid substrate can be expressed according to Eq. (2): qffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffi W ¼ γ tot γ dli þ γ liþ γ s þ γ li γ sþ ð2Þ li ð1 þ cos θ i Þ ¼ 2 where γ dli ; γ li ; γ liþ ; γ ds ; γ s ; γ sþ are dispersive, basic and acidic components of the surface free energy of the liquid and the solid, respectively. The total liquid surface free energy was γ tot qffiffiffiffiffiffiffi li ¼ þ  d γ li þ 2 γ li γ li . The Equation can be solved to evaluate the three

variables corresponding to the three components of the free energy surface for the solid. It is actually difficult to measure the contact angle accurately, since the process of wetting when the liquid spreads on a surface is affected by some factors as the viscosity of the fluid, the roughness and heterogeneity of the surface, the temperature of both the fluid and the substrate, the volume of the drop deposited and the specific interactions of the fluid and the surface [15]. Actually, on a rough surface, the apparent contact angle is related to the ideal contact angle by the Wenzel’s equation [15,16]. cos θapp ¼ ϕr cos θ

ð3Þ

where θapp is the apparent contact angle, θ is the actual Young’s contact angle and ϕr is defined as the ratio between the area Ar of the topography determined by any measurement technique, in this case AFM, and the nominal or geometrically projected area Ag of the topography, and it is called roughness Wenzel’s coefficient (Eq. 4): ϕr ¼

Ar Ag

ð4Þ

Since Ag is obtained from the projection of the microscopic area, Ar, onto a reference surface, it is also known as apparent area. As such, Ag will be equal to the scan area (product of the transversal and longitudinal scan lengths), in this case (1  1) μm2 was used [17].

198

K. Rezzadori et al. / Journal of Membrane Science 481 (2015) 195–206

2.3.3. Environmental scanning electron microscopy (ESEM) Membrane samples were dried in desiccators for 24 h, before performing the ESEM analysis, to remove any residual solvent. The samples were analysed in a FEI Quanta 200 FEG (FEI Company, Model Quanta 200 FEG, USA). Cross-section samples were obtained from the fracture in liquid nitrogen. 2.3.4. Thermogravimetric analysis (TGA) Thermogravimetric analysis (TGA) was performed in a thermal analysis instrument (Q500, New Castle, USA). Disc samples cut from films with weights between 5 and 15 mg were tested. When running dynamic scans, High Resolution (HiRes) mode was used, where the heating rate is automatically adjusted in response to changes in the rate of weight loss, which results in improved resolution, with an initial heating rate of 283.15 K/min under a nitrogen flux. 2.3.5. Determination of macauba oil retention For the macauba permeation assays, the same experimental unit (Fig. 1) was used, and an HPLC pump (AcuFlow, Model Series III HPLC, USA) was included to displace the macauba oil to the membrane module. The streams of CO2 and oil were mixed in a static mixer before reaching the thermostatic bath and the module. The feed flow rates of both streams were adjusted to reach the desired feed ratio of oil and CO2 (1:1 v/v). The temperature of the mixture of CO2 plus oil was controlled using a water bath (BT2) (T). Firstly the internal pump chamber was filled with solvent (CO2). The valve V1 was slowly opened (0.1 MPa s  1) until the system reached the working pressure. Valve V2 remained opened during system pressurisation to promote equal pressure on both sides of the membrane; therefore, avoiding mechanical damages. When the system reached the working pressure, the pump B2 was switched on for the oil transfer and the valve V2 was closed. The transmembrane pressure was set up by adjusting the micrometric valves (V3 and V4). The pressure difference between the upstream and downstream was monitored by a pressure gauge (P) and set at 5 MPa. The permeate flux was monitored for 60 min and permeate and retentate samples were collected using the valves V4 and V3, respectively. At the end of the process, the valve V1 was closed to isolate the system whilst the valve V3 was slowly opened to depressurise the system (0.1 MPa s  1). The same conditions listed in Section 2.2 were used to macauba permeation. In order to quantify the macauba oil feed concentration, samples of the retentate were periodically collected by the micrometre valve (V3). The macauba oil permeate flux was measured gravimetrically at time intervals by evaluation of the mass deposited in a collecting vial. The retention coefficient for the oil or fatty acids (%R) was calculated by Eq. (5), according to Ribeiro et al. [18]: %R ¼ ½ðcf  cp Þ100=cf

ð5Þ

where cf and cp are the concentrations of the solute in the feed and in the permeate, respectively. A selectivity factor (β) previously defined by Sarrade et al. [19] was used to compare membrane selectivity to each component of the oil. This factor is based on retentate and permeate compositions, according to Eq. (6). β ¼ %χ P =%χ R

ð6Þ

where % χP is the mass percentage of component in the permeate and % χR is the mass percentage of component in the retentate. If the selectivity factor was greater than one, then the component would permeate or pass through the membrane readily. If β were less than one, the component would be significantly retained or rejected by the membrane.

The fatty acid profile of the oil in the permeate and in the retentate was determined by gas chromatography coupled to a mass selective detector (GC–MS) (Shimadzu, model GCQT8030, USA), with GC–MS Solutions software (version 4.11), using the following conditions: column RTX-WAX 25 m  0.25 mm  0.25 μm, injection split 1:50, injector temperature 523 K, ionisation source temperature 523 K, interface temperature 523 K and carrier gas flow (helium/1.8 mL/min). Mixed pattern FAME C4-C24 (Supelco, Lot no. LB-80955) was used as reference material.

3. Results and discussion 3.1. Pure CO2 permeate flux Fig. 2 presents the curves for pure CO2 permeation with time, before and after CO2 permeation for 8 h. In general, independently of the membrane or its forming material, permeate flux was higher under supercritical conditions. High fluxes obtained after CO2 permeation in this supercritical condition suggest that the polymeric matrix is changing, which is probably due to swelling and/or plasticisation of the polymer by CO2. The mechanisms that cause the swelling and/or plasticisation were described by the weakening effect on the polymer structure due to swelling via reduction of interactions between polymer chains [4,20]. The solubility of CO2 in polymers depends on the specific interactions between CO2 and the polymer, and the nature of this interaction is mostly of a Lewis acid-base kind [21]. According to Akin and Temelli [4], CO2 has a quadrupole moment, accommodating both acid and base sites. These interactions can be considered the most important between several polymers and CO2 and different physicochemical properties such as polymer chain mobility, steric hindrance, aliphatic or aromaticity, molecular weight, side chain size, free volume and others can affect molecular interactions. Even for the case of well-known polymer structures, there are no methods available that entirely predict whether the polymer will be plasticised by CO2 or not. Furthermore, when CO2 is subjected to higher pressures and temperatures its capacity of interacting with the polymer gets stronger, which may lead to more severe changes in membrane structure, resulting in more pronounced swelling or plasticisation of the polymer, as a result for the higher mobility of the polymeric chain [9,22]. When the membranes were made of polyamide (NF270, ORAK and BW30), they present carbonyl and amino groups only partially cross-linked within its structure with a certain tendency to present hydrogen bonding. CO2 is a strongly sorbing penetrant, and the higher the feed pressure, the stronger the interactions with the polymer, especially if the polymer material has polar functional groups. The interaction of CO2 with polar groups where the oxygen of CO2 would interact with hydrogen has been reported previously [23]. High pressures, like those used in this study, can reduce the intermolecular hydrogen bonding in the polymer network and improve CO2 interaction with hydrogen of carboxyl and amide groups [4]. Hydrogen bonding can result in a relatively high mobility on the polymeric chain, when compared to crosslinking, especially in harsher conditions of pressure and temperature [8] increasing the permeate flux, as observed in these assays. In less severe (subcritical) conditions, further relaxation of the polymeric matrix would be necessary, thus longer time to observe dramatic increases in flux, justifying the lower fluxes obtained for this condition. Permeate fluxes for NP030 were lower than fluxes obtained for the other membranes in both tested conditions. For this membrane, in subcritical conditions, there were practically no differences before and after CO2 exposure, whereas in supercritical condition, a tendency to flux increasing during permeation time

K. Rezzadori et al. / Journal of Membrane Science 481 (2015) 195–206

199

Fig. 2. Pure CO2 permeate flux before and after CO2 exposure for the membranes: (a) NF270 (b) NP030 (c) ORAK (d) BW30.

can be observed. This increase in the permeate flux can be related to the solubilisation of dense CO2 in the polymeric matrix, causing swelling [24], which does not occur in subcritical conditions. Also, according to Kim and Lee [1], some polymers, like polyethersulfone, hold a relatively small free volume in the polymeric chain, hindering swelling and consequently do not yield higher fluxes, which agrees with the results found in the present work for NP030. After membrane exposure to CO2 for 8 h, an increase in the permeate flux for both conditions tested was also observed. The results are presented in Table 2. It can be verified that permeate flux increases around 140% and 165%, for subcritical and supercritical conditions, respectively. For NP030 in subcritical condition, however, the flux did not change after CO2 exposure for 8 h. As mentioned earlier, the use of high pressure also eases the diffusion and solubility of CO2 through the polymeric matrix, causing expansion and swelling of the polymeric material. Consequently, it promotes the plasticisation of the membrane and changes in physical and mechanical properties, leading to the occurrence of higher fluxes at the end of the 8 h period. Another reason for the increase in permeate flux after CO2 exposure for 8 h could be the higher surface hydrophobicity arising from dense CO2 polar group interactions. Higher hydrophobicity on the top surface could lead to higher dissolution rates of CO2, resulting in higher flux when compared with the new membrane. This fact is

Table 2 CO2 permeate flux increase after sub and supercritical CO2 exposure for 8 h. CO2 permeate flux increase (%)

Subcritical Supercritical

NF270

NP030

ORAK

BW30

126 161

0 170

163 161

152 160

consistent with the increase in contact angles observed after the CO2 exposure, which will be discussed in section 3.2.2. 3.2. Membrane characterisation 3.2.1. Atomic force microscopy – AFM Atomic Force Microscopy (AFM) was performed to analyse the surface topology of the commercial membranes before and after CO2 exposure. The micrographs are presented in Fig. 3. The valleys are characterised by the darkest coloration and, depending on the type of membrane, they can be considered as the pores, or internodular domains for dense membranes. The brightest spots represent the elevations. It can be verified that before the CO2 exposure, the studied membranes presented a heterogeneous topology, with some organisation patterns for each membrane in

200

K. Rezzadori et al. / Journal of Membrane Science 481 (2015) 195–206

Fig. 3. AFM images of: NF270 (A1, a1) new; (A2, a2) subcritical condition and (A3, a3) supercritical condition; NP030 (B1, b1) new; (B2, b2) subcritical condition and (B3, b3) supercritical condition; ORAK (C1, c1) new; (C2, c2) subcritical condition and (C3, c3) supercritical condition. BW30 (D1, d1) new; (D2, d2) subcritical condition and (D3, d3) supercritical condition.

K. Rezzadori et al. / Journal of Membrane Science 481 (2015) 195–206

Table 3 Membrane roughness calculated from different surface areas and the Wenzel’s roughness coefficient (for the 1  1 mm2 scanned areas) for each studied condition. Membranes

ϕr

Rq (nm) (5  5) lm2

(1  1) lm2

NF270 New 8 MPa/293.15 K 18 MPa/313.15 K

4.5 8.7 45.0

2.5 4.5 7.7

1.040 1.047 1.048

NP030 New 8 MPa/293.15 K 18 MPa/313.15 K

15.5 17.7 63.7

3.1 4.7 5.2

1.021 1.022 1.051

ORAK New 8 MPa/293.15 K 18 MPa/313.15 K

73.1 92.6 103.5

37.0 48.5 56.0

1.053 1.398 1.318

BW30 New 8 MPa/293.15 K 18 MPa/313.15 K

45.3 51.7 62.3

21.0 29.8 38.5

1.382 1.752 1.925

a nanometric level. These distinct profiles are a direct result of membrane composition and polymeric arrangement. For example, membranes NF270 and NP030 exhibit a surface with an active layer more regular and smooth, whilst ORAK and BW30 surfaces are more irregular and rough. These results also agree with roughness measurements presented in Table 3. In general, the possible reason for such differences is that the reverse osmosis membranes used in this work are comprised of a polyamide thin film, obtained by different reactions and configurations of the aromatic ring substituents (ortho-, meta-, or para-) of monomers [25]. This way, the highest roughness occurs for these membranes due to the higher distance between metaoriented amine monomers (m-PDA), which results in a lack of hydrogen bonds and in the formation of irregular polymers when compared, for instance, with NF270, which consists in a semiaromatic piperazine based polyamide on a polysulphone microporous membrane with a nonwoven polyester support [26]. These results are consistent with studies performed by Tang et al. [27] and Akin and Temelli [28]. Still, according to Fig. 3 and Table 3, it was verified that the influence of CO2 is stronger in the reverse osmosis membranes (BW30 and ORAK). Also, for these membranes, the roughness changes are more intense after exposure to supercritical conditions (comparing the figures with the same letter – A1, A2 and A3: new membrane, subcritical and supercritical conditions, respectively). For nanofiltration membranes, few or no changes in roughness factor were observed. As commented earlier in this paper, in pressure and temperature conditions above the critical point (7.4 MPa and 304.15 K), the solubility of the CO2 increases, thus, the amount of CO2 dissolved in the polymer is higher, which can lead to the swelling of the polymeric matrix and consequently to the increase of free volume of the polymer chains. Patil et al. [7] reported that the swelling of the polymeric chain could occur due to interactions between CO2 and polar groups from the polymeric material of the membrane. Koros [29] indicated that the carbonyl groups had the strongest effect on plasticisation of the polymer network. CO2 also has the potential to induce hydrogen-bonding interactions. Reverchon and Cardea [30] reported that the increase in segmental mobility of the polymer, caused by the swelling or plasticisation, results in membrane’s structural changes. According to Ronova et al. [31], the swelling of the polymeric matrix leads to a reduction in glass transition temperature and a reduction in the energy barrier on the polymer, i.e., the polymer with larger free volume needs less

201

energy for segmental mobility, causing rearrangements in the matrix and increasing the roughness. According to Ronova et al. [31], structural changes due to swelling and plasticisation of the polymer can lead to leaching of polymeric material from the active layer due to detachment or degradation, which may lead to an increase in roughness. In fact, the results obtained here strongly suggest that swelling and/or plasticisation occurs. However, the confirmation of such phenomena could only be obtained by the determination of the glass transition temperature (Tg) of the membrane material. This is not easy to be accomplished, since at ambient conditions, no residual CO2 is left in membrane matrix, and as the top layer is very thin, an accurate measurement of the Tg is not possible. According to Akin and Temelli [4], even for the case of well-known polymer structures, there are no accurate methods available to entirely predict whether the polymer will be plasticised by CO2. In Table 3, it is also observed that the scanned area plays a significant role: the larger the scanned area, the higher the roughness. This behaviour is common and was reported by other authors [14,32,33]. According to Boussu et al. [33], the phenomenon of increasing roughness with increasing scan area can be related to the dependency of the roughness on the spatial wavelength of the scanned area or the frequency. For a small surface area, only the roughness of the “higher” frequencies is measured. When a larger surface area is scanned, the roughness caused by additional lower frequencies also has to be taken into account. This results in a larger roughness value when a larger surface area is scanned. Another explanation for the increase of roughness with increasing scan size may be the formation of a fractal structure on the membrane surface when polymers are assembled to nodules or aggregates of nodules. So, when the scan size is changed, it is possible to get a different surface topography, resulting in a different roughness. Therefore it is crucial that the same scan size range is used when comparing the surface roughness for different samples [32–34]. It is known that the roughness of the active layer affects the contact angle or wettability of the membrane, since there is a direct correlation between the surface roughness and the contact angle, as shown in Eq. 3. This means that membranes with lower roughness will present lower apparent contact angles (more hydrophilic) and rougher membranes, usually give higher apparent contact angles (more hydrophobic) [35]. Therefore, in the next section, the changes in the contact angles after the exposure to sub- and supercritical CO2 will be discussed, highlighting the modification in the active layer surface regarding its hydrophobicity. The ϕr Wenzel’s roughness factors shown in Table 3 were obtained by averaging AFM scanned areas of 1.0 mm2, since at this scale there is still enough resolution but macroscopic features, not attributable to the membrane, are not considered [36].

3.2.2. Contact angle The contact angle measurements provide information on the hydrophobicity and hydrophilicity of membranes surface. The apparent contact angles, for each membrane and condition are given in Table 4. As mentioned before, these results can be improved by taking into account the roughness of the membrane surface that were measured by AFM to evaluate the Young’s angles shown in Table 4. Young equation defines the equilibrium contact angle on flat and homogenous surfaces. However, most surfaces are naturally rough and the fundamental theories for rough and heterogeneous surfaces proposed by Wenzel and the theoretical models are available. Among others, the Wenzel model (Eq. (3)) assumes that the liquid wets the whole rough substrate. The surface chemistry and structure directly affect contact angle, whereas the changes in the active layer result in a variable

202

K. Rezzadori et al. / Journal of Membrane Science 481 (2015) 195–206

Table 4 Young’s and apparent contact angles for new membranes and for membranes after dense (supercritical) CO2 treatment. Apparent angle (θapp)

Young’s angle (θYoung)

New

8 MPa/293.15 K

18 MPa/313.15 K

New

8 MPa/293.15 K

18 MPa/313.15 K

NF270 W F D

12.4 7 0.9 13.17 0.9 37.0 7 1.8

33.2 7 0.8 22.7 7 0.7 33.9 7 1.3

56.8 7 1.2 26.1 7 1.4 26.7 7 1.2

20.1a 70.9 20.5a 70.9 39.8ab 71.8

37.0b 70.8 28.2b 70.7 37.6bc 71.3

58.5c 71.2 31.0b 71.4 31.5c 71.2

NP030 W F D

52.8 7 1.3 53.6 7 1.5 38.0 7 3.1

69.17 1.3 42.17 1.4 18.2 7 2.4

85.4 7 2.4 59.1 7 0.7 24.07 1.9

53.7a 71.3 53.6a 71.5 54.5a 73.1

69.6b 71.3 43.5b 71.4 21.6b 72.4

85.6c 72.4 60.7c 70.7 29.6b 71.9

ORAK W F D

48.07 0.6 21.9 7 0.3 34.3 7 1.8

49.27 0.8 29.5 7 1.6 24.5 7 3.1

53.2 7 1.9 37.2 7 1.2 23.1 7 1.3

50.5a 70.6 28.2a 70.3 51.2a 71.8

62.1b 70.8 51.5b 71.6 49.4a 73.1

63.0b 71.9 52.8b 71.2 29.1b 70.3

BW30 W F D

34.9 7 3.3 21.9 7 3.4 39.4 7 3.5

39.4 7 1.9 38.9 7 1.3 33.9 7 2.3

67.6 7 2.6 45.6 7 1.3 18.9 7 2.4

53.6a 73.3 47.8a 73.4 66.3a 73.4

63.8b 71.9 63.6b 71.3 61.7a 72.3

78.6c 72.6 68.7b 71.3 46.8b 72.4

W¼ water; F ¼ formamide; D¼ diiodomethane. Values followed by different superscript letters, indicate significant differences (significance level of 5 %) between the samples in different conditions for the same membrane and wetting liquid

contribution to the contact angle, i.e. the lower the roughness, the closer the apparent contact angle is to Young’s contact angle. This is because the increase of surface area due to roughness can influence the measurement and interpretation of contact angles due to the irregular interface formed [37]. The Young contact angles increase after the roughness factors are considereded because the roughness factor, ϕr, is always greater than 1. The Wenzel´s model predicts that the apparent contact angle when a liquid wets a surface (θo901) will decrease if the surface becomes rough (θapp oθYoung). Once this taken into account, the water Young’s angles increase for all the membranes studied, especially for the reverse osmosis membranes and after contact with dense CO2. In general, higher contact angles with water indicate that the surface is more hydrophobic. Equivalently for liquids with low polarity, smaller angle values correspond to higher wettability [38]. In this context, after the exposure to CO2 a reduction in contact angles was observed for the less polar liquid (diiodomethane), while the angle increases for formamide and significantly for water. This indicates an increase in hydrophobicity of the surface after CO2 exposure. The Young’s contact angle with water increased after exposure to CO2 in supercritical condition (18 MPa/313.15 K) for NF270, NP030 and BW30, by 3-fold, 1.6-fold and 1.5-fold, respectively. For subcritical CO2 (8 MPa/293.15 K), the increase in the contact angle with water was around 1.2-fold. ORAK membrane presented the same increase (1.2-fold) both for sub- and supercritical conditions. In general, for the supercritical condition, the increases in water contact angle were higher. This behaviour highlights the higher solvation capacity of CO2 in these conditions. Also, when diiodomethane was used for the contact angle determination, the angle decreased in the same sequence, confirming that the exposure to CO2 caused an increase in surface hydrophobicity. However, when using formamide, which is a liquid with intermediary polarity, a clear tendency was not observed. In some cases, the contact angle decreases with sub critical CO2 conditions and increases with supercritical conditions. According to Akin et al. [4], the polar groups present in the polymeric material of the membranes easily interacts with CO2 and these changes in contact angle can be a good indicator of the interaction between these groups. Table 5 presents the surface free energy data and their components for the studied membranes. A reduction in the

Table 5 Work of adhesion (W), surface free energy data (γtotal ) and its components: s dispersive (γds ), acid and basic (γsþ and γs ) and polar (γps ) for each membrane. γtotal s

γd

γp

γþ

γ

Wwater

NF270 New 8 MPa/293.15 K 18 MPa/313.15 K

53.28 51.23 51.09

39.71 40.80 43.59

13.57 10.43 7.50

0.89 0.70 1.22

51.98 38.70 14.19

145.13 134.43 115.47

NP030 New 8 MPa/293.15 K 18 MPa/313.15 K

35.62 43.71 49.77

31.73 42.61 47.30

3.89 1.10 2.47

0.11 0.11 0.18

35.28 2.61 8.65

117.33 92.01 91.99

ORAK New 8 MPa/293.15 K 18 MPa/313.15 K

53.18 39.52 38.58

33.60 34.61 35.60

19.58 4.91 2.98

0.84 0.28 0.30

22.10 21.24 20.95

114.70 109.04 109.77

BW30 New 8 MPa/293.15 K 18 MPa/313.15 K

40.69 29.24 28.32

29.04 18.50 20.96

11.65 10.74 7.36

0.18 0.02 0.23

30.40 28.95 12.52

112.07 96.53 85.34

surface free energy with the CO2 exposition for all polyamide membranes can be observed. For the polyethersulfone NP030 membrane, surface free energy increased probably due to the strong increment of the dispersive contribution, although the work of adhesion also decreased. The reverse osmosis membranes showed the greatest drop in surface energy after CO2 exposure (1.4-fold). Also, for these membranes, only slight differences between the membranes exposed to subcritical and supercritical CO2 could be detected. All membranes presented γd value higher than γp, which means that these membranes presented nonpolar properties [39]. The changes on the surface energy after CO2 exposition occur mainly in the polar component, showing a decrease, which is more evident in supercritical conditions. This decrease in polar component agrees with the total surface free energy data, also suggesting a decrease in the hydrophilicity of the membranes after their exposition to CO2. Moreover, all the membranes exhibited higher electron donor components (γ  ) and relatively lower electron acceptor components (γ þ ), suggesting that the basic component is what controls the polar contribution of the surface free energy [40].

K. Rezzadori et al. / Journal of Membrane Science 481 (2015) 195–206

The adhesion corresponds to the attraction of a certain material for another one; thus, the work of adhesion of water (Wwater) reflects the attraction between water and the membrane. This can indicate that the lower the value for Wwater, the less stable is the system. Therefore, for lower values of Wwater, it is easier to break the adhesion between the surfaces (membrane surface layer and water). In the present study, it can be verified that for all tested membranes, Wwater decreases after CO2 exposition, mostly in supercritical conditions, confirming once again the decrease of the membrane hydrophilic character. 3.2.3. Environmental scanning electron microscopy (ESEM) Fig. 4 shows the micrographs of the membranes NF270, NP030, ORAK and BW30, new and after exposition to sub- and supercritical CO2 for 8 h. After CO2 exposition for 8 h, some visual alterations were observed in the structure of the studied membranes. Magnifications in the cross-section of the membranes by 100,000 times allowed the observation of changes from an interconnected to a globular structure, which resembled microcapsules, after CO2 exposition for 8 h for all the polyamide based membranes (NF270, ORAK and BW30). These alterations were more

visible in supercritical condition (18 MPa/313.15 K). For NP030, the formation of bead-like structures does not occur in subcritical condition. Nevertheless, they are more evident in supercritical condition. The distension of the polymeric chain and development of bead-like structure can occur due to solvent-polymer interactions, leading to swelling and plasticisation phenomena, as stated earlier in this work, referring to the active skin layer as seen by AFM. 3.2.4. Thermogravimetric analysis (TGA) Thermogravimetric analysis was performed to evaluate the thermal stability of the membranes after CO2 exposure. Dynamic runs in Hi-Res mode, in a nitrogen atmosphere, showed an initial small weight loss (below a 3% in weight, from ambient temperature to approximately 773.15 K) in the membranes NF270, ORAK and BW30. NP030 showed the same behaviour; however the initial weight loss went only until 433.15 K, according to Fig. 5. This small initial weight loss for all membranes can be attributed to absorbed or adsorbed water and residual solvent in the sample [41,42]. NF270 and BW30 presented similar patterns with substantial weight losses and for these membranes, polymer degrada-

Subcritical CO2

Supercritical CO2

(8 MPa/293.15 K)

(18 MPa/313.15 K)

BW30

ORAK

NP030

NF270

New

203

Fig. 4. Photomicrographs of membrane cross-section, new and after CO2 exposure in sub and supercritical conditions for 8 h.

204

K. Rezzadori et al. / Journal of Membrane Science 481 (2015) 195–206

Fig. 5. Thermogravimetric analysis for the membranes NF270 (a), NP030 (b), ORAK (c) and BW30 (d).

3.2.5. Macauba oil retention Fig. 6 presents the permeate flux of macauba oil through the membranes NP030 and ORAK in supercritical condition (18 MPa/ 313.15 K) and Δp ¼5 MPa. For both membranes, during the first

30 ORAK NP030

25

-1

-2

Macauba oil permeate flux (kg.h .m )

tion occurs at 790.15 K. For ORAK, which consists in aromatic polyamide, two degradation stages were detected. The first stage occurred at 753.15 K with about 50% loss of material. The second stage was detected at 813.15 K, resulting in 20% weight loss. The final stage of weight loss is due to the thermal decomposition of the remaining aromatic polyimide segments, leaving a relatively high amount of carbonaceous residue [43]. NP030 showed a different behaviour from the other membranes, which may be related to the type of polymer (polyethersulfone). This membrane also presents two degradation stages, one at around 433.15 K, and other at 673.15 K, with losses of 20 and 40% of its original weight, respectively. The membrane material reported by the manufacturer for NP030 is only polyethersulfone, while the other three membranes contain a polyamide toplayer. This is possibly the main reason for such behaviour. None or slight alterations were observed in thermogravimetric analysis after the membrane exposure to sub- and supercritical CO2, i.e., the thermophysical properties of the polymers and consequently, of the studied membranes, are affected neither by subcritical nor supercritical CO2.

20

15

10

5

0 0

10

20

30

40

50

60

70

Time (min) Fig. 6. macauba oil permeate flux for the membranes NP030 and ORAK in supercritical condition (18 MPa/313.15 K) and Δp¼ 5 MPa.

25 min of experiment, the macauba oil flux suffers a reduction to around one half and reaches a stationary value. This reduction is probably due to compaction, concentration polarisation and fouling phenomena, which imply that the membrane resistance changes during the permeation process. At the end of the process a thin film of oil covering the membrane surface was observed. Furthermore, Akin and Temelli [4] reported that the decrease in

K. Rezzadori et al. / Journal of Membrane Science 481 (2015) 195–206

flux at a high Δp (5 MPa) could be also related to the suppression of swelling behaviour by compaction of the polymer network. Nanofiltration membrane NP030 showed higher initial fluxes (around 25 kg h  1 m  2) than the other membranes, while the initial flux for the ORAK (reverse osmosis membrane) was the smallest (approximately 15 kg h  1 m  2). This was expected, since the NP030 membrane has greater molar weight cut-off (MWCO¼400 Da). However, CO2 fluxes were always higher through ORAK. The permeate flux was almost stable after 40 min, with permeate flux around 5 kg h  1 m  2 for both membranes. Regarding the NF270 and BW30 membranes, neither oil nor CO2 permeation was observed. This may have occurred due to the polarity difference between the membrane surface and the mixture of oil and CO2. In this case, both membranes were initially quite hydrophilic, whilst the filtered mixture is non-polar. This difference can cause a total pore blockage in the membranes. Actually these membranes presented the lowest increase of permeability after supercritical CO2 treatment, as shown in Table 2 and Fig. 2, probably due to a milder induced change in their porous structure. Wall and Braun [44] also observed this behaviour in organic compounds in supercritical CO2 permeation through a hydrophilic microporous TiO2-membrane. In the same way, Moura et al. [45] have found permeate fluxes for modified triacylglycerols nearly null for the BW30 membrane. The authors used a transmembrane pressure of 1.5 MPa and temperature of 313.15 K and a pretreatment of the membranes in ethanol (4 h) followed by n-hexane (4 h). In subcritical conditions, also neither oil nor CO2 permeation was observed for all the membranes studied. In this condition the viscosities of the compressed liquids are higher than for the supercritical fluid, which is probably the reason for this behaviour. Furthermore, in supercritical conditions, CO2 has a lower viscosity and higher diffusivity than in subcritical state. Macauba oil retention factors showed that there was no apparent damage in the structures of membranes NP030 and ORAK, since retentions were high. The retention factor for NP030 and ORAK was 85% and 95%, respectively. Other authors report a decrease in retention in the permeation of supercritical CO2 and oil mixtures. For example, Spricigo et al. [46] reported that a reverse osmosis membrane, in the presence of supercritical CO2, may swell affecting its selective properties. The same authors reported that the retention of the nutmeg (genus Myristica) essential oil showed some fluctuations throughout the experiment period and when they used transmembrane pressure of 3.0 MPa, the nutmeg oil retention decreased. The average nutmeg essential oil retention was 96.4%. Akin and Temelli [4] reported that oleic acid retention of ORAK, using a feed pressure of 12 MPa, a temperature of 313.15 K and Δp¼ 4.0 MPa was 77%. At the end of the process (after 24 h), the authors observed an increase of 1.2-fold in the oleic acid retention. This behaviour was not observed in the present study. Besides the oil retention, the performance of the membranes was also evaluated based on fatty acids separation capacity, since the macauba oil usually has a high degree of hydrolysis, and thus high content of free fatty acids. Table 6 presents the fatty acids retention and the selectivity factor β for the membranes NP030 and ORAK. In general, the membranes NP030 and ORAK offered low separation capacity for the fatty acids that compose the macauba kernel oil. Nevertheless, a total retention for the palmitic acid was observed using NP030. Although NP030 has a MWCO greater than the palmitic molar mass, a total retention was observed, showing the influence of the solvent and membrane polymer interactions. Furthermore, for both membranes, negative retentions were observed for lauric acid (lower molar mass), i.e., this component

205

Table 6 Retention of fatty acids and selectivity factor β values for the membranes NP030 and ORAK in supercritical condition (18 MPa/313.15 K) and Δp ¼ 5 MPa.

C12:0 C14:0 C16:0 C18:0 C18:1 C18:2

(lauric acid) (myristic acid) (palmitic acid) (stearic acid) (oleic acid) (linoleic acid)

Fatty acids retention (%)

Selectivity factor (β)

NP030

ORAK

NP030

ORAK

 8.9 1.3 100.0 2.6 1.4 2.5

 18.3 6.5 21.0 26.0 7.8 4.3

1.2 0.6 0.01 0.1 1.0 1.0

4.2 0.9 0.8 0.8 0.9 0.9

permeates preferentially through the membranes, evidenced by the selectivity factor β, greater than one. Thus, lauric acid was found at low concentration in the retentate; whilst palmitic acid was found at low concentration in permeate samples. Negative retentions indicate that the solute is enriched in the permeate stream as compared to the bulk feed concentration [6]. In cases where there is stronger interaction of the solutes with the membrane material, negative rejections have been observed. This behaviour can occur when the solute-membrane affinity is larger than the solvent-membrane affinity-solute molecules accumulate at the pore wall and the solute flux is finally larger than the solvent flux [7]

4. Conclusions The effect of dense CO2 in sub- and supercritical conditions on physicochemical and morphological properties of four commercial membranes was presented and the stability of different nanofiltration and reverse osmosis membranes was assessed. CO2 permeate flux increased after 8 h of exposure in both sub- and supercritical conditions. Transport of CO2 under these conditions has certain interactions with the polymeric materials, possibly weakening the polymeric matrix structure and inducing a structural reorganisation. Higher permeate flux was observed for the membranes subjected to supercritical CO2, due to an induced higher surface hydrophobicity arising from CO2 polar group interactions. Higher hydrophobicity on the top surface could lead to higher dissolution rates of CO2 resulting in higher flux compared to the new membrane. The corresponding increase of water contact angle was accompanied by a decrease in the work of adhesion with an increase of hydrophobicity. Furthermore, the polar component of surface energy decreased for all the membranes after CO2 treatments. Membrane roughness increases after CO2 exposure, as shown by AFM analyses. Again, the effect of supercritical CO2 on the surface morphology of membranes was more pronounced than the effect caused by subcritical CO2. But, the thermal stability of the membranes did not change after exposure to CO2, although an actual relevant change could be masked by the influence of the support layers, which are probably quite similar in all cases. In general, the exposure to dense CO2 cause changes in chemical and morphological properties, which could improve the filtration of non-aqueous feed streams. Membranes NP030 and ORAK presented high oil retentions, 85.5% and 95.5%, respectively; while NF270 and BW30 were totally clogged without any permeation neither of macauba oil nor CO2. It is then possible to use NF and RO membranes to regenerate CO2 in supercritical conditions, since they presented good resistance to supercritical CO2. Regarding the fatty acids fractionation; both membranes presented permeation of the component with the lower molecular mass

206

K. Rezzadori et al. / Journal of Membrane Science 481 (2015) 195–206

(lauric acid); and NP030 showed almost total retention for palmitic acid. Thus, it is technically feasible to use these commercial polymeric membranes in supercritical systems for CO2 regeneration and partial fatty acid fractioning. However, it is essential to select the correct membrane for each specific use. Acknowledgements The authors wish to thank the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) for financial support (Grant nos. 304137/2012-0 and 480066/2011-6). Authors also thank Mrs. Bibiana C. Gándara and ICTP/CSIC for the TGA analyses. Dr. Maria Tereza Friedrich and Universidade de Passo Fundo are acknowledged for carrying out the chromatographic analyses. References [1] S. Kim, Y.M. Lee, High performance polymer membranes for CO2 separation, Curr. Opin. Chem. Eng. 2 (2013) 238–244. [2] R.W. Baker, Membrane gas-separation: applications, in: E. Drioli, L. Giorno (Eds.), Membrane Operations, Wiley-VCH Verlag GmbH and Co., 2009, pp. 167–194. [3] L.A.V. Sarmento, R.A.F. Machado, J.C.C. Petrus, T.R. Tamanini, A. Bolzan, Extraction of polyphenols from cocoa seeds and concentration through polymeric membranes, J. Supercrit. Fluids 45 (2008) 64–69. [4] O. Akin, F. Temelli, Effect of supercritical CO2 pressure on polymer membranes, J. Membr. Sci. 399–400 (2012) 1–10. [5] M.A. Shannon, P.W. Bohn, M. Elimelech, J.G. Georgiadis, B.J. Marinas, A. M. Mayes, Science and technology for water purification in the coming decades, Nature 452 (2008) 301–310. [6] L.H.C. Carlson, A. Bolzan, R.A.F. Machado, Separation of D-limonene from supercritical CO2 by means of membranes, J. Supercrit. Fluids 34 (2005) 143–147. [7] V.E. Patil, L.J.P. Van Den Broeke, F.F. Vercauteren, J.T.F. Keurentjes, Permeation of supercritical carbon dioxide through polymeric hollow fiber membranes, J. Membr. Sci. 271 (2006) 77–85. [8] O. Akin, F. Temelli, Performance characterization of polyamide reverse osmosismembranes upon supercritical CO2 processing, J. Supercrit. Fluids 66 (2012) 150–156. [9] J. Zhou, M. Tran, A.T. Haldeman, J. Jin, E.H. Wagener, S.M. Husson, Perfluorocyclobutyl polymer thin-film composite membranes for CO2 separations, J. Membrane Sci. 450 (2014) 478–486. [10] H.J. Navarro-Díaz, S.L. Gonzalez, B. Irigaray, I. Vieitez, I. Jachmanián, H. Hense, J. V. Oliveira, Macauba oil as an alternative feedstock for biodiesel: characterization and ester conversion by the supercritical method, Journal of Supercritical Fluids 93 (2014) 130–137. [11] G.D. Bothun, B.L. Knutson, H.J. Strobel, S.E. Nokes, E.A. Brignole, S. Díaz, Compressed solvents for the extraction of fermentation products within a hollow fiber membrane contactor, J Supercrit Fluids 25 (2003) 119–134. [12] A.L. Carvalho, F. Maugeri, V. Silva, A. Hernández, L. Palacio, P. Prádanos, AFM analysis of the surface of nanoporous membranes: application to the nanofiltration of potassium clavulanate, J. Membr. Sci. 46 (2011) 3356–3369. [13] C.J. Van Oss, R.J. Good, N.K. Chaudhury, Interfacial Lifschitz–van der Waals and polar interactions in macroscopic systems, Chem. Rev. 88 (1988) 927–941. [14] J.H. Clint, A.C. Wicks, Adhesion under water: surface energy considerations, Int. J. Adhes. Adhes. 21 (2001) 267–273. [15] M. Palencia, B. Rivas, E. Pereira, A. Hernández, P. Prádanos, Study of polymer– metal ion-membrane interactions in liquid-phase polymer-based retention (LPR) by continuous diafiltration, J. Membr. Sci. 336 (2009) 128–139. [16] K. Boussu, B. Van Der Bruggen, A. Volodin, J. Snauwaert, C. Van Haesendonck, C. Vandecasteele, Roughness and hydrophobicity studies of nanofiltration membranesusing different modes of AFM, J. Colloid Interface Sci. 286 (2005) 632–638. [17] P.J. Ramon-Torregrosa, M.A. Rodríguez-Valverde, A. Amirfazli, M.A. CabrerizoVílchez, Factors affecting the measurement of roughness factor of surfaces and its implications for wetting studies, Colloid Surf. A: Physicochem. Eng. Asp. 323 (2008) 83–93. [18] A.P.B. Ribeiro, J.M.L.N. Moura, L.A.G. Gonçalves, J.C.C. Petrus, L.A. Viotto, Solvent recovery from soybean oil/hexane miscella by polymeric membranes, J. Membr. Sci. 282 (2006) 328–336. [19] S. Sarrade, G.M. Rios, M. Carlés, Supercritical CO2 extraction coupled with nanofiltration separation. Applications to natural products, Sep. Purif. Technol. 14 (1998) 19–25.

[20] A.F. Ismail, W. Lorna, Penetrant-induced plasticization phenomenon in glassy polymers for gas separation membrane, Sep. Purif. Technol. 27 (2002) 173–194. [21] S.P. Nalawade, F. Picchioni, J.H. Marsman, L.P.B.M. Janssen, The FT-IR studies of the interactions of CO2 and polymers having different chain groups, J. Supercrit. Fluids 36 (2006) 236–244. [22] J. Zhou, A.T. Haldeman, E.H. Wagener, S.M. Husson, CO2 plasticization and physical aging of perfluorocyclobutyl polymer selective layers, J. Membr. Sci. 454 (2014) 398–406. [23] P. Raveendran, Y. Ikushima, S.L. Wallen, Polar attributes of supercritical carbon dioxide, Acc. Chem. Res 38 (2005) 478–485. [24] M. Sauceau, J. Fages, A. Common, C. Nikitine, E. Rodier, New challenges in polymer foaming: a review of extrusion processes assisted by supercritical carbon dioxide, Prog. Polym. Sci. 36 (2011) 749–766. [25] Y.N. Kwon, C.Y. Tang, J.O. Leckie, Change of chemical composition and hydrogen bonding behaviour due to chlorination of crosslinked polyamide membranes, J. Appl. Polym. Sci. 108 (2008) 2061–2066. [26] S.P. Nunes, K.P. Peinemann, Membrane Technology in the Chemical Industry, Wiley-VCH, 2001 (299 pp.). [27] C.Y. Tang, Q.S. Fu, A.P. Robertson, C.S. Criddle, J.O. Leckie, Use of reverse osmosis membranes to remove perfluorooctane sulfonate (PFOS) from semiconductor wastewater, Environ. Sci. Technol. 40 (2006) 7343–7349. [28] O. Akin, F. Temelli, Effect of supercritical CO2 flux, temperature and processing time on physicochemical and morphological properties of commercial reverse osmosis membranes, J. Supercrit. Fluids 60 (2011) 81–88. [29] W.J. Koros, Simplified analysis of gas/polymer selective solubility behavior, J. Polym. Sci., Part B: Polym. Phys. 23 (1985) 1611–1628. [30] E. Reverchon, S. Cardea, Production of controlled polymeric foams by supercritical CO2, J. Supercrit. Fluids 40 (2007) 144–152. [31] I.A. Ronova, O.V. Sinitsyna, S.S. Abramchuk, A.N. Yu, S. Chisca, I. Sava, M. Bruma, Study of porous structure of polyimide films resulting by using various methods, J. Supercrit. Fluids 70 (2012) 146–155. [32] A.W. Mohammad, N. Hilal, L.Y. Pei, I.N.H.M. Amin, R. Raslan, Atomic force microscopy as a tool for asymmetric polymeric membrane characterization, Sains Malays. 40 (2011) 237–244. [33] K. Boussu, B. Van der Bruggen, A. Volodin, C. Van Haesendonck, J.A. Delcour, P. Van der Meeren, C. Vandecasteele, Characterization of commercial nanofiltration membranes and comparison with self-made polyethersulfone membranes, Desalination 191 (2006) 245–253. [34] A.L. Carvalho, F. Maugeri, V. Silva, A. Hernández, L. Palacio, P. Pradanos, AFM analysis of the surface of nanoporous membranes: application to the nanofiltration of potassium clavulanate, J. Mater. Sci. 46 (2011) 3356–3369. [35] A. Rahimpour, M. Jahanshahi, N. Mortazavian, S.S. Madaeni, Preparation and characterization of asymmetric polyethersulfone and thin-film composite polyamide nanofiltration membranes for water softening, Appl. Surf. Sci. 256 (2010) 1657–1663. [36] P. Ramón, M. Rodríguez, A. Amirfazli, M. Cabrerizo, Aspects affecting the measurement of roughness factor of surfaces and its implications for wettin studies, Colloids Surf. A: Physicochem. Eng. Asp. 323 (2008) 83. [37] A. Subramani, X.F. Huang, E.M.V. Hoek, Direct observation of bacterial deposition onto clean and organic-fouled polyamide membranes, J. Colloid Interface Sci. 336 (2009) 13–20. [38] R. Konradi, Weak Polyacid Brushes: Synthesis, Swelling Behaviour, Complex Formation, Germany, 2005. [39] Q. Wang, Z. Wang, C. Zhu, X. Mei, Z. Whu, Assessment of SMP fouling by foulant-membrane interaction energy analysis, J. Membr. Sci. 446 (2013) 154–163. [40] S. Bargir, S. Dunn, B. Jefferson, J. Macadam, S. Parsons, The use of contact angle measurements to estimate the adhesion propensity of calcium carbonate to solid substrates in water, Appl. Surf. Sci. 255 (2009) 4873–4879. [41] M.A.B. Meador, V.A. Cubon, D.A. Scheiman, W.R. Bennett, Effect of branching on rod-coil block polyimides as membrane materials for lithium polymer batteries, Chem. Mater. 15 (2003) 3018–3025. [42] A. Tena, A. Marcos-Fernandez, A.E. Lozano, J.G. Campa, J. de Abajo, L. Palacio, P. Pradanos, A. Hernandez, Thermally treated copoly(ether-imide)s made from bpda and aliphatic plus aromatic diamines. GAS separation properties with different aromatic diamimes, J. Membr. Sci. 387–388 (2012) 54–65. [43] K. Fukatsu, Thermal degradation behaviour of aromatic polyamide fiber blended with cotton fibre, Polym. Degrad. Stabil. 75 (2002) 479–484. [44] Y. Wall, G. Braun, Sorption effects of co-solvents and dissolved organic compounds on supercritical CO2 permeation through a micro-porous TiO2membrane, J. Supercrit. Fluids 66 (2012) 144–149. [45] J.M.L.N. Moura, L.A.G. Gonçalves, L.A.V. Sarmento, J.C.C. Petrus, Purification of structured lipids using SCCO2 and membrane process, J. Membr. Sci. 299 (2007) 138–145. [46] C.B. Sprícigo, A. Bolzan, R.A.F. Machado, L.H.C. Carlson, J.C.C. PETRUS, Separation of nutmeg essential oil and dense CO2 with a cellulose acetate reverse osmosis membrane, J. Membr. Sci. 188 (2001) 173–178.