Desolventizing organic solvent-soybean oil miscella using ultrafiltration ceramic membranes

Desolventizing organic solvent-soybean oil miscella using ultrafiltration ceramic membranes

Journal of Membrane Science 475 (2015) 357–366 Contents lists available at ScienceDirect Journal of Membrane Science journal homepage: www.elsevier...

827KB Sizes 3 Downloads 54 Views

Journal of Membrane Science 475 (2015) 357–366

Contents lists available at ScienceDirect

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

Desolventizing organic solvent-soybean oil miscella using ultrafiltration ceramic membranes Jonas R.M. de Melo a, Marcus V. Tres a,b,n, Juliana Steffens a, J. Vladimir Oliveira c, Marco Di Luccio c a

Department of Food Engineering, URI Erechim, Erechim, RS, 99700-000, Brazil UFSM, Cachoeira do Sul, RS 96506-302, Brazil c Department of Chemical and Food Engineering, UFSC, Florianópolis, SC, 88040-900, Brazil b

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 13 October 2014 Accepted 15 October 2014 Available online 5 November 2014

This work reports the use of ceramic membranes with different cut-offs (5, 10 and 20 kDa) for the separation of synthetic mixtures simulating soybean oil industrial miscellas in organic solvents (n-hexane, ethanol and isopropanol). The mass ratios oil/solvent investigated in this work were 1:4, 1:3 and 1:1 (w/w) for the feed pressures of 0.5–4 bar depending on the miscella. It is shown that n-butanol was the best solvent for the proper conditioning of 20 kDa membrane, since it increased permeate flux of n-hexane, up to 314 L/m2 h at 1 bar of transmembrane pressure. The desolventizing of oil/solvent mixtures was strongly affected by solvent nature, and on the solute–solvent–membrane affinity. The highest retentions were observed for oil/ethanol mixtures, with values usually close to 100%, as a consequence of polarity as well as low solvation power. Crude oil mixture with n-hexane (industrial mixture) yielded greater retention and lower flux than those obtained with refined oil, due to the polarized layer formed by gums and phospholipids. Results reported in this work indicate the potential applicability of this technology in vegetable oil processing and biodiesel industries in the solvent recovery step. & 2014 Elsevier B.V. All rights reserved.

Keywords: Ceramic membrane Soybean oil Organic solvents Separation Desolventizing

1. Introduction The projections of world population growth in addition to the limited area for food production have put forward a maximization scenery and valorization of food chain market. Within such segment, soybean has become undoubtedly one of the major world commodities. Although the production process of soybean oil has reached desirable levels, the soybean oil extraction system has been based on an old process with technical developments mostly focused on production control systems [1]. The conventional extraction process is based on the oil extraction by organic solvent (n-hexane), and its separation through solvent evaporation/distillation involves enormous energy costs. According to Johnson and Lusas [2], the solvent separation from the mixture by distillation is so costly, that if the whole extraction industry is analyzed, only the separation step is responsible for 2/3 of the total energy demand by the industry. Other studies report

n Corresponding author at: UFSM, Cachoeira do Sul, RS, 96506-302, Brazil. Tel.: þ 55 51 3722 3247; fax: þ 55 51 3722 3057. E-mail address: [email protected] (M.V. Tres).

http://dx.doi.org/10.1016/j.memsci.2014.10.029 0376-7388/& 2014 Elsevier B.V. All rights reserved.

the substantial energy savings achieved by application of membrane separations in oil industry [3]. Then, it seems that a huge energy economy would be possible by making use of membrane technology in this kind of industry. Despite appearing revolutionary at first glance, membrane technology is commercially used for more than six decades, being born in 1950 for the desalinization of water from the sea. Many advantages of membrane technology can be envisaged, such as its robustness, easy installation with room saving, energy economy, the high selectivity reached, operation and scale-up. Furthermore, low temperature processing may be a great advantage for thermosensitive compounds affording products of higher quality, with lesser sensorial and nutritional changes, and of course significant reduction in energy costs [4]. The first report on membrane separation for oil/solvent mixtures was focused on using polymeric membranes, which present as main disadvantages inherent incrustation, plasticization and swelling when in contact with solutes and organic solvents, hence leading to the reduction of its industrial life time [5–12]. Ceramic membranes, in spite of the higher initial cost, present a great potential application in the separation of solutes from non-aqueous solutions, since the interactions of the solutes and solvents with the

358

J.R.M. de Melo et al. / Journal of Membrane Science 475 (2015) 357–366

membrane ceramic material are much smaller when compared with polymeric membranes [13–15]. Such characteristic increases membrane stability and, consequently, provides longer shelf life. However, recent studies have shown that solute–solvent–membrane interactions play an important role even when ceramic membranes are used in non-aqueous permeations [16–18]. The use of alternative solvents to the conventional n-hexane seeks a more economical and sustainable system, since n-hexane comes from fossil reserves and tends to increase prices with the unavoidable depletion of natural petroleum reserves. Besides the price uptrend in coming decades [19], n-hexane is a poisonous solvent, which generates great interest in its substitution for less harmful solvents and eco-friendly compounds, following the tendencies of the concept of Green Chemistry [20]. In this sense, several other solvents can be used for soybean extraction just varying some process parameters. Among some potential solvents, ethanol, isopropanol and their azeotropes have been proposed as alternatives to n-hexane in the vegetable oil extraction industry, in order to reduce the environmental risks while affording a healthier final product [21]. Darvishmanesh et al. [22] and Kwiatkowski and Cheryan [23] proposed the use of renewable solvents for the soybean and corn oil extraction, respectively, and the solvent recovery using commercial nanofiltration polymeric membranes. These authors presented interesting results proving that the membrane technology can be successfully applied in traditional industrial processes for solvent recovery. Thus, the aim of the present work was to select and explore the application of ceramic membranes in soybean oil separation from conventional solvent (n-hexane) and green solvents (ethanol and isopropanol).

2. Material and methods 2.1. Material and reagents The commercial refined soybean oil used in the assays was purchased at the local market. The industrial mixture was kindly provided by a soybean oil extraction industry (Erechim, RS, Brazil) and comprised of crude soybean oil in n-hexane (1:4). The solvents used in the experiments had a minimum purity of 95%. The ceramic membrane of 20 kDa and its stainless steel membrane module were purchased from the Jiangsu Jiuwu Hitech Co. (Jiangsu, China). The membrane has 19 internal channels with diameter of 0.004 m and each membrane has 1.016 m length. The ceramic membranes of 5 and 10 kDa were purchased from Pall Corporation (New York, USA). The membranes are single channel with 0.010 m with 0.25 m length. Table 1 shows the specifications of the membranes used in this work. The choice of the membranes investigated in this work was based on previous experience of our research group [5–9]. In this work, retentate and permeate solutions returned to the feed tank for keeping constant feed concentration. Fig. 1 shows the experimental apparatus used in the present work. For the temperature control a thermostatic bath was used (MA 083, Marconi, São Paulo, Brazil). The rotameter was manufactured by Conaut (Model 440, São Paulo, Brazil). The gear pump was purchased from Micropump-Idex Corporation (Vancouver, WA, USA).

For the evaporation of the solvents, a vacuum oven was employed (Quimis, Q819v2, São Paulo, Brazil). An analytical balance was used for samples weighing (Shimadzu, AY220), with a precision of 0.005 g.

2.2. Mixtures preparation The minimum volume of feeding solution was defined based on the needed volume for each membrane unit, taking also into account the necessary sampling. Thus, for the 20 kDa membrane, the feed volume stipulated was 1.5 L, and for both 5 and 10 kDa membranes, 0.25 L was adopted. Preparation of the oil/solvent mixtures was made in mass basis, following the mass ratios established in the study. The following mass ratios in the feed were investigated for ethanol, isopropanol and n-hexane: 1:4, 1:4, 1:3, 1:4, 1:3,1:1, respectively.

2.3. Membrane conditioning Due to the high polarity exhibited by ceramic membranes it is necessary a conditioning step before using, in order to increase the solvent flux and hence make those membranes suitable for industrial applications. A priori, the conditioning, or pretreatment, involves membrane washing in appropriate solvents to either increase or remove preservatives and humectants from membrane surface and its pores [24]. The 20 kDa membrane had its flux checked and it was the only one submitted to tests with n-hexane. In such case, 20 kDa membrane was immersed in several solvents, which were then exchanged to n-hexane. The pre-conditionings that afforded the highest permeate flux of n-hexane was chosen. Some more recent studies show that the conditioning or pretreatment step of hydrophilic membranes can change its hydrophilic character, allowing an increase in organic solvent flux. Basically, the idea is to expose the membrane to a variety of solvents with decreasing polarities [25]. In the case of the 20 kDa membrane, n-butanol was chosen as preconditioning solvent before n-hexane permeation. Briefly, the module was completely filled with n-butanol at room temperature and the membrane was kept soaked in the solvent for 24 h. After removing n-butanol, 3 rinses with n-hexane were carried out, and then the module was maintained filled with n-hexane for 24 h prior to the assays [7].

Fig. 1. Schematic diagram of experimental apparatus. A- feed tank, B- thermostatic bath, C- gear pump, D- membrane module, E- manometer, F- valve, G- rotameter.

Table 1 Membrane description and classification. Class UF UF UF

Membrane material zirconium zirconium α-aluminia oxide/zirconium

Manufacturer Pall Co. Pall Co. Jiangsu Jiuwu Hitech Co.

MWCO 5 kDa 10 kDa 20 kDa

Permeation area 2

0.0055 m 0.0055 m2 0.24 m2

Model

pH limit

S700-01446 S700-01447 CMF 19040-OD30

0–14 0–14 0–14

J.R.M. de Melo et al. / Journal of Membrane Science 475 (2015) 357–366

2.4. Determination of permeate flux After membrane conditioning, the permeate solvent flux was determined as a means of determining the degree of membrane cleaning. The separation assays were only started if the flux of pure solvent was at least 80% of the initial flux. The initial flux for these comparisons is the permeated flux of the solvent in the clean membrane. Sampling in the conditioning step was accomplished after stabilization of pressure and feed flux (1 L/min) for at least 30 min. The permeate flux was calculated by average of three samples by: J¼

V AU t

ð1Þ

where J is the permeate flux (L/m2 h), V is the permeate volume (L), A is the membrane area (m2) and t is the time (h) of sampling. The mass flux was straightforwardly calculated by multiplying the volumetric flux by the solvent specific mass. 2.5. Separation assays After determining the solvent flux and checking that membranes were appropriate for use, the solvent was removed from the system, and the freshly prepared mixture of oil and solvent was fed in the system. For all assays the temperature of the solvent was 20 1C, except for the experiments with ethanol, which were conducted at 40 1C due to the low solubility of ethanol/oil at room temperature. For instance, the solubility of soybean oil in ethanol at 40 1C is 20.4 wt%, in isopropanol at 20 1C is 45.1 wt%, while at that same temperature in n-hexane the oil is completely miscible [21]. Permeate samples were withdrawn every 10 min and stored in nitrogen-atmosphere glass flasks. Oil concentration was determined gravimetrically. Samples were quickly weighed and then submitted to vacuum oven evaporation to constant weigh at 2 1C below the solvent boiling point to prevent bubble formation and possible sample overflow. Then, the samples were placed in desiccators for cooling. The oil retention was calculated by:    Cp R ¼ 100 1  ð2Þ Cr where Cp is the oil concentration in the permeate and Cr denotes the oil concentration in the retentate. The assays carried out with each membrane, as well as the mass ratios investigated, are shown in Table 2. All experimental runs were performed at least in duplicates and the experimental errors were always smaller than 10%. 2.6. Cleaning procedure After the assays with the oil/solvent mixtures, a series of rinses were carried out, firstly in the solvent in which the experiment

359

was performed, for removal of the residual oil from the membrane surface. In the specific case of the 20 kDa membrane, after three rinses in n-hexane, three more rinses with n-butanol were carried out. If necessary, the chemical cleaning was performed. For such cleaning, the solvents were exchanged to water following the reverse order used for membrane pretreatment. After water rinsing, a solution of 1% of sodium hypochlorite (NaOCl, Indústrias Anhembi, 2.5% w/w) and 1% of sodium hydroxide (NaOH, Cromato Produtos Químicos, PM40) at 80 1C was recirculated in the system for 15 min. Then the system was washed with ultrapure water (Milli-Q), followed by recirculation of an aqueous solution of NaOH 2% for 30 min at 80 1C. Another cycle of ultrapure water washing was carried out, followed by recirculation of an aqueous solution of nitric acid 1.5% (HNO3, Quimex, 65%) and a final ultrapure water washing. 3. Results and discussion 3.1. Membrane conditioning The performance of the membrane is influenced by its chemical composition, temperature, pressure, flux and interactions between components of the feed stream to the membrane surface [26]. Ceramic membranes exhibit highly hydrophilic characteristics, since they are composed mostly by metal (aluminum, titanium, silicon or zirconium) oxides, nitrides and carbides [27]. Such point was confirmed by measuring the permeability of pure solvents for 20 kDa membrane, as shown in Table 3, where the permeability observed for pure solvents and for n-hexane after preconditioning in different solvents are presented. The permeate flux of n-hexane for the membrane before conditioning was very low ( o0.25 L/m2 h bar). It can be seen from Table 3 the highest permeability was obtained with water, the most polar solvent used. The permeate flux decreased when the solvent polarity is reduced. Such fact is in Table 3 Permeability of the pure solvents and n-hexane after pretreatment in each solvent for 20 kDa membrane. Permeability of the solvents (L/m2 h bar) Water Ethanol n-Propanol Isopropanol n-Butanol

192.5 120.5 76.1 70.4 60.6

Permeability of n-hexane after pretreatment in each solvent (L/m2 h bar) Ethanol 9.2 n-Propanol 4.8 Isopropanol 172.4 n-Butanol 313.8 Note: dielectric constants of solvents at 25 1C: water¼ 80.1, etanol ¼24.5, n-propanol¼ 20.3, isopropanol¼19.9, n-butanol ¼17.5, n-hexane ¼ 1.88 [42]

Table 2 Mixtures and solvents used in each assay. Solvent 20 kDa membrane n-hexane

Ethanol Isopropanol 5 and 10 kDa membranes Ethanol Isopropanol

Oil/solvent mass ratio

Pressure

Feed flow rate

Temperature

1:4 1:3 1:1 1:4 1:4 1:3

1 bar

1 L/min

20 1C

1:4 1:4 1:3

2, 3 and 4 bar

40 1C 20 1C

1 L/min

40 1C 20 1C

360

J.R.M. de Melo et al. / Journal of Membrane Science 475 (2015) 357–366

accordance with Machado et al. [28], which reported that the flux is closely related to the hydrophyilicity or hydrophobicity of the membrane simultaneously with the ability to form hydrogen bonds with solvent. Thus, for membranes with polar characteristic, the higher is the solvent chain length, the lower will be the permeated solvent flux. When n-butanol was used, higher n-hexane permeate flux was obtained, differently when only n-hexane was used in preconditioning step. Following the trend observed for the pure solvent flux, now using the same analogy for preconditioning and subsequent n-hexane permeation, it was observed that the solvent which had closest polarity to n-hexane generated higher permeate flux. The use of solvents with different polarities generated different results. So, it was observed a clear advantage for the permeate fluxes when using n-butanol rather than n-hexane. Shukla and Cheryan [25] studied this gradual solvent transition and concluded that less abrupt and thus more gradual changes in solvent polarity in the conditioning process provide a synergism effect that increases the flux. The possible reason for the better performance of the pretreatment with n-butanol could be related to the adsorption of nbutanol onto the pore surface. The polar head interacts with the polar pore surface and the nonpolar tail would then increase the surface hydrophobicity, leading to an increase in the interaction of pore surfaces with the nonpolar solvent n-hexane, improving its flux. The adsorption of polar solutes on ceramic membrane surfaces and its effects on membrane performance is already reported in the literature. Li et al. [29] showed that the fouling of zirconia ceramic membranes was severely affected by pH of aqueous feed solution, since membrane charge is altered by pH and adsorption of solutes on the membrane matrix is favored depending on pH. A systematic study on solvent flux behavior of ceramic and nanofiltration membranes was reported by Buekehoudt et al. [18]. The authors emphasized the importance of viscosity in solvent transport. However, they showed that polarity difference between the membrane surface and the solvent played an important role, depending on the membrane pore size. The authors cite the history effects (that we here call pretreatment or membrane conditioning) to be responsible for changes in solvent permeation behavior. Solvents like alcohols can adsorb on surface hydroxyl groups affecting subsequent permeation assays. Thus, we can confirm that the alcohol pretreatment is an effective way to improve solvent permeation through ceramic membranes.

Tres et al. [7] investigated the best solvent for the conditioning step in an attempt to increase n-hexane flux in a 50 kDa polymeric hollow fiber membrane. The authors tested a homologous series of alcohols (ethanol, n-propanol and n-butanol) and n-propanol was the one that provided the largest n-hexane flux. The authors attributed that effect to lower polarity of this solvent compared to ethanol, which might have caused agglomeration of hydrophilic and hydrophobic groups, hence reducing the hydrophilicity of the membrane. Despite n-butanol presents the lowest polarity in that homologous series, it did not lead to the highest n-hexane flux, hence demonstrating the complexity of the interfacial phenomena involved in polymeric membrane separation processes. Ribeiro et al. [14] obtained practically no n-hexane flux, when using a 0.01 mm pore size ceramic membrane (Atech Innovations GmbH) for soybean oil/n-hexane degumming. However, when ethanol was used as pre-conditioning solvent, measurable nhexane fluxes were recorded. The authors attributed this fact to the gradual solvents transition, as ethanol has an intermediate polarity and is soluble in both water and n-hexane. Other authors investigated different conditionings for UF polyamide polymeric membranes [30] and also polyamide and polysulfone covered with silicone, [31] and showed that the pretreatment with ethanol increases the n-hexane permeate flux. Vankelecom et al. [32] affirmed that the solvent transport by the membrane depends on several solvent characteristics such as polarity, viscosity and surface tension. The effect of solute–solvent–membrane interactions on nanofiltration ceramic membrane performance has been extensively studied by Marchetti et al. [16,17]. They showed that Hagen– Poiseuille equation cannot describe solvent permeation though nanofiltration ceramic membranes, and proposed correction factors based on capillary pressure, dipole and steric effects. The effect of solute charge was important in water, but was less significant when an organic solvent was present.

3.2. Soybean oil/n-hexane mixtures After determining the best preconditioning treatment for the 20 kDa membrane, the first mixture was used to simulate the separation of industrial soybean oil/n-hexane mixture, varying the proportion of oil/n-hexane in 1:4, 1:3 and 1:1 (w/w). Fig. 2 shows the experimental data of permeate flux (A) and retention (B).

100

10 9

80

8

70

7 Retention (%)

Permeate flux (kg/m 2 h)

90

60 50 40

1:1 (w/w) 1:3 (w/w)

30

1:4 (w/w)

6 5 4 3

1:1 (w/w)

20

2

1:3 (w/w)

10

1

1:4 (w/w)

0

0

10

20

30

40

Time (min)

50

60

70

0

0

10

20

30

40

50

60

70

Time (min)

Fig. 2. Total mass flux (A) and oil retention (B) for the soybean oil/n-hexane mixture in the mass ratios 1:4, 1:3 and 1:1 for 20 kDa membrane at 1 bar transmembrane pressure.

J.R.M. de Melo et al. / Journal of Membrane Science 475 (2015) 357–366

polymeric membranes, for which a fast decline of the permeate flux is usually noted. 3.3. Industrial mixture After the usage of synthetic mixtures of soybean oil and n-hexane, an industrial mixture was tested. The mixture composition was determined evaporating the solvent (n-hexane) and considering the remaining amount constituted only by the oil, on the basis that oil and solvent are the major components in the mixture. As a result, the percentage of soybean oil was determined to be 25.9 wt% in the mixture. The permeate flux (A) and oil retention (B) results are shown in Fig. 3. It can be noticed that the fluxes were from 35% to 75% smaller than those obtained with the prepared synthetic mixture refined soybean oil/n-hexane. At 1 bar, the flux falls from 53 kg/m2 h to 14 kg/m2 h, probably due to the presence of gums (phospholipids) in the mixture leaving the extraction equipment. These phospholipids might have deposited onto the membrane surface, causing a decrease in the permeate flux. This effect was also reported in the literature by Moura et al. [37], which have shown that flux and rejection were timedependent, and attributed this behavior to the deposition of gums and phospholipids onto the membrane surface. However, it is interesting to note that the fluxes remained almost invariant during the experimental time, hence suggesting that this membrane is little affected by the fouling, differently from polymeric membranes [5,38]. With the increase in the polarized layer, increases the membrane resistance making the flux less sensitive to the working pressure. The increment in the pressure usually promotes larger permeated fluxes. However, here it will now only represent a small increment in the flux. The polarized layer formed by the gums tends to increase with the applied pressure, increasing the resistance to the permeation. Such phenomenon was also observed by Souza et al. [10], using a ceramic membrane of alumina (MWCO 0.05 mm) in the separation of components from non-refined corn oil. Fig. 3 (B) shows the results of oil retention observed for 20 kDa membrane as applied to the industrial mixture. Corroborating with results obtained for the permeate flux, larger oil retentions (o7%) than those obtained compared to the synthetic mixtures of refined soybean oil were observed. These results confirm the fouling influence, increasing the retention at the operating pressures applied in this work, also reported elsewhere [37]. Wu and Lee [13] used a ceramic membrane of alumina (disk shape, average pore size 0.02 mm) for the separation of crude soybean oil (33% of the mixture) and n-hexane. Oil retentions obtained in their assays were around 20% and permeate fluxes were up to 240 L/m2 h, at 8 bar of transmembrane pressure. They obtained a higher permeate flux, possibly due to the higher pressure applied in that work. The oil retention obtained by Wu

16

14

15

12

14

Retention (%)

Permeate flux (kg/m²h)

It can be noticed from Fig. 2 that an increase in the oil mass ratio led to a reduction in the permeate flux, which might be explained in terms of two main factors: the first related to a raise in oil concentration that can cause an increase of solution viscosity, and as permeability is affected by viscosity, higher viscosities lead to smaller fluxes [32]. The second point refers to the polarized layer generated, as higher is the solute concentration in the mixture, the larger will be the polarized layer and this will produce a greater resistance to solvent permeation [28]. Permeate fluxes obtained in this work can be considered significant, due partly to the good membrane conditioning technique employed and the large average pore size of the membranes. For example, Ribeiro et al. [14] used a commercial alumina ceramic membrane (MWCO 0.01 mm and 0.2 m2 of permeation area) for the mixtures degumming, containing 32 wt% of soybean oil in n-hexane, and obtained at the best experimental condition (2 bar and 3.4 m/s) a permeate flux of 40.5 kg/m2h, which is lower than the one shown in Fig. 2 at similar conditions. Marenchino et al. [33] working with a zirconium oxide membrane (MWCO 15 kDa) obtained a permeate flux of 30.0 L/m2 h for a soybean oil/n-hexane mixture with 25 wt% of oil. The oil retentions for the 20 kDa membrane are shown in the Fig. 2(B) where it can be noticed an opposite behavior for the permeate flux. Higher oil/solvent mass ratios in the mixture tend to increase the oil retention for this membrane. This behavior may also be related to the concentration polarization, since higher oil concentrations increase the resistance to permeation, for both solvent and oil. The oil retention obtained by the ceramic membrane of 20 kDa investigated in this work reached a maximum of 7.2%, a very low value, showing that the average pore size of this membrane is not adequate to promote satisfactory oil retention. Gupta [34] reported the average size of micelles formed by soybean oil/n-hexane is around 20 kDa, which means that smaller cut-offs are required to allow better separation. Ribeiro et al. [35] using a polymeric membrane Sepa GH (MWCO 1 kDa), obtained permeate fluxes for mixtures of soybean oil/n-hexane mixtures (1:3 w/w) up to 30 L/m2 h at 25 bar and 45 1C. The retention levels varied between 36.6% and 67.1% and an increment in the feed pressure also led to reductions in retention levels and increase the permeate flux. The low average membrane pore size used by Ribeiro et al. [35] required the use of a relatively high working pressure, on the other hand, due to the smaller pore size, oil retention values obtained were comparatively much better. It can also be observed that there was a good stability in the permeate flux within the process time for all the assays. Bottino et al. [36] showed that lower transmembrane pressure in the tangential filtration helps maintaining a near constant flux. For ceramic membranes, the reduction of the permeate flux is gradual, differently from

13 12

11

0.5 Bar

10

1.0 Bar

9 8

2.0 Bar 0

10

20

30

40

Time (min)

50

60

70

361

10 8 6 4

0.5 Bar

2

1.0 Bar

0

2.0 Bar

0

10

20

30

40

50

60

70

Time (min)

Fig. 3. Total mass flux (A) and oil retention (B) for the industrial mixture for 20 kDa membrane at 0.5, 1.0 and 2.0 bar transmembrane pressure.

362

J.R.M. de Melo et al. / Journal of Membrane Science 475 (2015) 357–366

and Lee [13] was more than twice higher than that observed in the present study. The main reason for this fact can be due to the nature of the feed solutions that those authors used. As the oil consists in a crude oil that contains gums, these gums might have deposited onto the membrane surface, creating a dynamic membrane that took to the larger retention. Ribeiro et al. [14] using a ceramic membrane (MWCO 0.01 mm) in the permeation of crude soybean oil with 32% of n-hexane obtained fluxes up to 40.5 kg/m2 h at 2 bar and tangential speed of 3.4 m/s. An increase in oil content in the feed led to a decrease in the permeate flux. Alicieo et al. [39] employed a ceramic membrane for the filtration of crude soybean oil, without n-hexane addition, with the same diameter of pore diameter of Ribeiro et al. [14], (MWCO 0.01 mm), and obtained a maximum permeated flux of 4.16 kg/m2 h at 6 bar. Thus, it can be inferred that higher solution viscosity may lead to lower permeate fluxes. Souza et al. [10] investigated the performance of a ceramic alumina membrane (MWCO  5 mm with 19 channels, 1 m length and filtration area of 0.2 m2) using corn oil and n-hexane, with pressures of 0.5 and 1.5 bar and tangential speeds of 1.4 and 2.4 m/s for the mass ratios of crude corn oil/n-hexane of 1:3 and 1:1.85 (w/w). The authors showed that an increase in the transmembrane pressure, without changes in the tangential speed and oil concentration, led to an increment in phospholipids retention for all oil concentrations in the mixtures. An increase in oil concentration in the mixture, and accordingly an increase in mixture viscosity, led to a decrease in the permeate flux. Garcia et al. [40] carried out sunflower oil filtration, in tubular membranes of polyethersulfone (4 and 9 kDa). The authors noticed

that an increase in pressure besides causing an increase in the permeate flux increased the oil retention. Although the authors did not present any specific justification, possibly the presence of gums in the polarized layer tends to increase the retention of the membrane, as well as larger working pressures increase the driving force of the system. 3.4. Soybean oil/ethanol mixture Fig. 4 shows the permeate fluxes in the mixture of soybean oil/ ethanol for the 20 kDa (A), 10 kDa (B) and 5 kDa (C) membranes. Comparison of the fluxes of the three membranes shows that a reduction in membrane pore size markedly reduced the permeate flux. In some cases, when combining high oil concentrations and low operating pressures no permeate flux was measurable. When compared to the fluxes obtained using n-hexane as solvent, greater fluxes are recorded even when the membrane was pretreated and conditioned with n-hexane before mixture permeation. Naturally, ethanol has a higher dipole moment and dielectric constant than n-hexane (see Table 3), hence justifying the higher permeate fluxes observed. Again, permeate flux is strictly related to oil concentration in the mixture and an increase in oil mass ratio causes a raise in the mixture viscosity, leading to a decrease of convective flux through the porous medium and then the decrease of permeate fluxes. It is remarkable how the increase in the oil mass ratio led to a drop in permeate flux, namely, from15 kg/m2 h to 3 kg/m2 h, when the mass ratio was increased from 1:3 to 1:1.

Fig. 4. Permeate fluxes for the mixture soybean oil/ethanol for 20 kDa (A), 10 kDa (B) and 5 kDa (C) membranes in the mass ratio 1:4.

J.R.M. de Melo et al. / Journal of Membrane Science 475 (2015) 357–366

For the 10 kDa membrane (Fig. 4B) the effect of the applied pressure is visible. Higher applied pressure promotes higher permeate fluxes at the same oil mass ratio. With the 5 kDa membrane (Fig. 4C), there was a remarkable drop in the flux, which do not exceed, under any working condition, the value of 2.5 kg/m2 h. Comparing the results with those obtained with the membranes of 20 kDa and 10 kDa, there was a large decrease in the permeate flux. The oil retention in the soybean oil/ethanol mixture, in the same manner as the permeate flux, presented a positive trend: the greater the oil content, the higher the retention is obtained. Fig. 5(A) shows the observed data for the retention of oil in a mixture of soybean oil in ethanol for the 20 kDa membrane. Retention levels between 88% and 96% were higher than those observed in the oil/n-hexane mixture (Fig. 3B). This high level of retention may be due to polarity difference of ethanol over the oil itself, thus facilitating the selectivity of the membrane having polar characteristics. Fig. 5(B) shows the oil retention data for the soybean oil/ ethanol mixtures through the 10 kDa membrane. It is observed that the oil retention remained high, always above 80%. It can also be noticed that there is a reduction in oil retention with increase of applied pressure, since for greater driving force, the greater the effect of polarization and fouling. Fig. 5(C) shows the oil retention for the 5 kDa membrane and soybean oil/ethanol mixture as the feed. It is observed that the retention was very high, similar to the 10 kDa membrane. The permeation of feed mixtures in ethanol through the 10 kDa membrane yielded the highest levels of retention among all experiments, which was already explained due to its higher polarity and lower solubility in oil.

363

Higher pressure causes greater compaction of the material retained on the membrane surface, increasing the retention, as well as higher oil concentration increases the solution viscosity, decreasing the permeate flux and increasing the membrane retention [10]. 3.5. Soybean oil/isopropanol mixture Fig. 6 shows the fluxes through the membranes investigated in this work, for the two soybean oil/isopropanol feed ratios. For 5 kDa and 10 kDa membranes, flux followed the expected trend of increasing with the feed pressure, and decreasing with oil concentration in feed, the latter due to the polarization effect. However, for 20 kDa membrane, flux was not affected by oil concentration. Experimental data obtained for 20 kDa membrane concerning oil retention are shown in Fig. 7. As verified for the permeate flux, oil retention was not influenced by oil concentration in the mixture. It should be emphasized that these assays were conducted twice, yielding the same results. Dafinov et al. [41] studied the adsorption and chemisorption of alcohols by alpha alumina, which is a component of the 20 kDa membrane. The strength of the interaction was measured by the temperature required to evaporate the solvent from alpha alumina. Ethanol and methanol required 130 1C, well above its boiling point temperature, indicating a physical interaction with the material. The propanol and butanol were evaporated from the membrane material in temperatures above 200 1C, indicating a strong interaction, possibly a chemical sorption with the alpha alumina. The authors found that the conditioning effects using alcohols as a result of the strong interactions taking place with the membrane can cause decrease in water permeation showing that

Fig. 5. Oil retention for the mixture of soybean oil/ethanol in the mass ratio 1:4 for the 20 kDa (a), 10 kDa (b) and 5 kDa (c) membranes.

364

J.R.M. de Melo et al. / Journal of Membrane Science 475 (2015) 357–366

Fig. 6. Permeate fluxes for the mixture soybean oil/isopropanol for 20 kDa (A), 10 kDa (B) and 5 kDa (C) membranes.

100 90

1:3 (w/w)

80

1:4 (w/w)

Retention (%)

70 60 50 40

30 20 10

0

0

10

20

30

40

50

60

70

Time (min)

Fig. 7. Oil retention in the mixture of soybean oil/isopropanol in mass ratios of 1:4 and 1:3 for 20 kDa membrane.

Figs. 8 and 9 show the retention of oil from soybean oil/ isopropanol mixtures for the 10 kDa and 5 kDa membranes, respectively. Regarding oil retention at different pressures and concentrations of oil in the mixture soybean oil/isopropanol, a distinct behavior was observed from the oil mixtures in ethanol. An unexpected behavior was observed, since oil retention declined significantly with time, while permeate flux was kept constant. This unusual behavior might be directly related to changes in solute and solvent interactions with membrane over permeation time. Marchetti et al. [16] described the role of competition between solvent-membrane and solute-membrane affinities on solute retention. When solute-membrane affinity is larger than the solvent-membrane affinity, retention tends to decrease. In the present case, the isopropanol may be strongly adsorbing to the pore walls, as suggested by Darvinov et al. [41], improving interaction of the membrane with the oil, which could be enhancing solute permeation causing the retention to decrease.

4. Conclusions interference permeation is not only affected by membrane pore size, but also may be dependent on the characteristics of membrane surface. The interaction with alcohol may have modified the membrane to a more hydrophobic nature. Membrane surface modification may completely alter the affinity between the filter medium and the species dissolved there, altering the membrane performance. Thus, the interaction of isopropanol with the membrane may have influenced the results, differentiating them from those obtained with the other solvents.

The results of the preconditioning the 20 kDa membrane comprised of alpha alumina and zirconia have indicated that n-butanol is the best solvent for the proper conditioning. This solvent was the conditioner that increased permeate flux of n-hexane, reaching values of up to 314 L/m2 h at 1 bar of transmembrane pressure. The desolventizing of oil/solvent mixtures was strongly affected by solvent nature, and on the solute–solvent–membrane affinities. The highest retention was observed for oil/ethanol mixtures, with

J.R.M. de Melo et al. / Journal of Membrane Science 475 (2015) 357–366

100

90

Retention (%)

80 70

4 bar

90

3 bar

80

2 bar

70

Retention (%)

100

60 50 40 30

30 10

20

30

40

50

60

0

70

2 bar

40 20

10

3 bar

50

10 0

4 bar

60

20 0

365

0

10

20

30

Time (min)

40

50

60

70

Time (min)

Fig. 8. Oil retention in mixture of soybean oil/isopropanol at different pressures for the mass ratios of 1:4 (A) and 1:3 (B) for 10 kDa membrane.

100 90 70

4 bar

60

3 bar

50

2 bar

Retention (%)

Retention (%)

80

40 30 20 10 0

0

10

20

30

40

50

60

Time (min)

70

100 90 80 70 60 50 40 30 20 10 0

4 bar 3 bar 2 bar

0

10

20

30

40

50

60

70

Time (min)

Fig. 9. Oil retention for mixture of soybean oil/isopropanol at different pressures for mass ratios of 1:4 (A) and 1:3 (B) for 5 kDa membrane.

values usually close to 100%, possibly due to its low oil solvation power, causing strong oil exclusion by the membrane, since ethanol can interact stronger with the pore walls than the oil. When crude oil mixture with n-hexane (industrial mixture) was tested greater retention and lower flux than those obtained with refined oil, due to the polarized layer, which can be constituted of all parts of gums and phospholipids. Permeation of oil/isopropanol mixtures through 10 kDa and 5 kDa membranes presented a decrease in rejection over process time, while flux was constant. This behavior was attributed to the changes in oil-membrane affinity over time, due to adsorption of solvent onto pore walls. The results presented in this work indicate the potential applicability of this technology in vegetable oil processing and biodiesel industries in the solvent recovery step.

Acknowledgments The authors thank CNPq, FAPERGS, Transfertech Gestão de Inovações LTDA and URI Erechim for financial support and scholarships. References [1] A. Martinho, H.A. Matos, R. Gani, B. Sarup, W. Youngreen, Modeling and simulation of vegetable oil processes, Food Bioprod. Process. 86 (2008) 87–95. [2] L.A. Johnson, E.W. Lusas, Comparison of alternative solvents for oils extraction, J. Am. Oil Chem. Soc. 60 (1983) 229–242. [3] A. Buekenhoudt, H. Beckers, D. Ormerod, M. Bulut, P. Vandezande, R. Vleeschouwers, Solvent based membrane nanofiltration for process intensification, Chem. Ing. Tech 85 (2013) 1243–1247. [4] J.A. Howell, A. John, The Membrane Alternative – Energy Implications for Industry. The Watt Committee on Energy, CRC Press, London, UK, 1990. [5] M.V. Tres, S. Mohr, M.L. Corazza, M. Di Luccio, J.V. Oliveira, Separation of nbutane from soybean oil mixtures using membrane processes, J. Membr. Sci. 333 (2009) 141–146.

[6] M.V. Tres, J.C. Racoski, Z. Novello, J.R.M. Melo, H.C. Ferraz, M. Di Luccio, J.V. Oliveira, Separation of soybean oil/n-hexane miscellas using polymeric membranes, J. Food Sci. Eng. 2 (2012) 616–624. [7] M.V. Tres, R. Nobrega, R.B. Carvalho, J.V. Oliveira, M. Di Luccio, Solvent recovery from soybean oil/n-hexane mixtures using hollow fiber membrane, Braz. J. Chem. Eng. 29 (2012) 577–584. [8] M.V. Tres, J.C. Racoski, R. Nobrega, R.B. Carvalho, J.V. Oliveira, M. Di Luccio, Solvent recovery from soybean oil/n-butane mixtures using a hollow fiber ultrafiltration membrane, Braz. J. Chem. Eng. 31 (2014) 243–249. [9] M.V. Tres, J.C. Racoski, M. Di Luccio, J.V. Oliveira, H. Treichel, D. Oliveira, M.A. Mazutti, Separation of soybean oil/n-hexane and soybean oil/n-butane mixtures using ceramic membranes, Food Res. Int. 63 (2014) 33–41. [10] M.P. Souza, J.C.C. Petrus, L.A.G. Gonçalves, L.A. Viotto, Degumiming of corn oil/ hexane miscella using a ceramic membrane, J. Food Eng. 86 (2008) 557–564. [11] J.M.L.N. Moura, L.A.G. Gonçalves, L.A.V. Sarnento, J.C.C. Petrus, Purification of structured lipids using SCCO2 and membrane process, J. Membr. Sci. 299 (2007) 138–145. [12] D. Pioch, C. Larguèze, J. Graille, H. Ajana, J. Rouviere, Towards an efficient membrane based vegetable oils refining, Ind. Crop Prod. 7 (1998) 83–89. [13] J.C. Wu, E. Lee, Ultrafiltration of soybean oil/hexane extract by porous ceramic membranes, J. Membr. Sci. 154 (1999) 251–259. [14] A.P.B. Ribeiro, N. Bei, L.A.G. Gonçalves, J.C.C. Petrus, L.A. Viotto, The optimization of soybean oil degumming on a pilot plant scale using ceramic membrane, J. Food Eng. 87 (2008) 514–521. [15] C.C. Carvalho, M.P. Souza, T.D. Silva, L.A.G. Gonçalves, L.A. Viotto, Soybeans crude oil miscella degumming utilizing ceramic membranes: transmembrane pressure and velocity effects, Desalination 200 (2006) 543–545. [16] P. Marchetti, A. Butté, A.G. Livingston, NF in organic solvent/water mixtures: role of preferential solvation, J. Membr. Sci. 444 (2013) 101–115. [17] P. Marchetti, A. Butté, A.G. Livingston, An improved phenomenological model for prediction of solvent permeation through ceramic NF and UF membranes, J. Membr. Sci. 415–416 (2012) 444–458. [18] A. Buekenhoudt, F. Bisignano, G. De Luca, P. Vandezande, M. Wouters, K. Verhulst, Unravelling the solvent flux behaviour of ceramic nanofiltration and ultrafiltration membranes, J. Membr. Sci. 439 (2013) 36–47. [19] ASPO - Association for the Study of Peak Oil & Gas. International Energy Agency accepts Peak Oil: an analysis of chapter 3 of the world energy outlook, 2004. [20] J.V. Oliveira, M. Di Luccio, M.V. Tres, Sistema e processo para extração de óleos vegetaise essenciais utilizando fluidos pressurizados, BR Patent PI 1020120187523, Deposited in July 27, 2012a. [21] A.P. Gandhi, K.C. Joshi, K. Jha, V.S. Parihar, D.C. Srivastav, P. Raghunadh, J. Kawalkar, S.K. Jain, R.N. Tripathi, Studies on alternative solvents for the extraction of oil-I soybean, Int. J. Food Sci. Technol. 38 (2003) 369–375.

366

J.R.M. de Melo et al. / Journal of Membrane Science 475 (2015) 357–366

[22] S. Darvishmanesh, T. Robberecht, P. Luis, J. Degrève, B. Van der Bruggen, Performance of nanofiltration membranes for solvent purification in the oil industry, J. Am. Oil Chem. Soc. 88 (2011) 1255–1261. [23] J.R. Kwiatkowski, M. Cheryan, Recovery of corn oil from ethanol extracts of ground corn using membrane technology, J. Am. Oil Chem. Soc. 82 (2005) 221–227. [24] R. Kesting, Synthetic Polymeric Membranes, 2nd edition, Wiley, Hoboken, USA, 1985. [25] R. Shukla, M. Cheryan, Performance of ultrafiltration membranes in ethanol– water solutions: effect of membrane conditioning, J. Membr. Sci. 198 (2002) 75–85. [26] L. Lin, K.C. Rhee, S.S. Köseoğlu, Bench-scale membrane degumming of crude vegetable oil: process optimization, J. Membr. Sci. 134 (1997) 101–108. [27] L.J. Zeman, A.L. Zydney, Microfiltration and ultrafiltration: principles and application, Marcel Dekker, Inc., London, UK, 1996. [28] D.R. Machado, D. Hasson, R. Semiat, Effect of solvent properties on permeate flow through nanofiltration membranes: Part II. Transport model, J. Membr. Sci. 166 (2000) 63–69. [29] W. Li, W. Xing, W. Jin, N. Xu, Effect of pH on microfiltration of Chinese herb aqueous extract by zirconia membrane, Sep. Purif. Technol. 50 (2006) 92–96. [30] I. Kim, J. Kim, K. Lee, T. Tak, Phospholipids separations (degumming) from crude vegetable oil by polyimide ultrafiltration membrane, J. Membr. Sci. 205 (2002) 113–123. [31] R. Subramanian, M. Nakajima, Membrane degumming of crude soybean and rapeseed oils, J. Am. Oil Chem. Soc. 74 (1997) 971–975. [32] I.F.J. Vankelecom, K. De Smet, L.E.M. Gevers, A. Livingston, D. Nair, S. Aerts, S. Kuypers, Pierre A Jacobs, Physico-chemical interpretation of the SRNF

[33] [34] [35]

[36]

[37] [38]

[39] [40]

[41] [42]

transport mechanism for solvent through dense silicone membranes, J. Membr. Sci. 231 (2004) 99–108. R. Marenchino, C. Pagliero, M. Mattea, Vegetable oil degumming using inorganic membranes, Desalination 200 (2006) 562–564. A.K.S. Gupta, Purification process, U.S. Patent 44093540, 1978. 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. A. Bottino, G. Capannelli, A. Comite, F. Ferrari, F. Marotta, A. Mattei, A. Turchini, Application of membrane processes for the filtration of extra virgin olive oil, J. Food Eng. 65 (2004) 303–309. J.M.L.N. Moura, L.A.G. Gonçalves, J.C.C. Petrus, L.A. Viotto, Degumming of vegetable oil by microporous membrane, J. Food Eng. 70 (2005) 473–478. M.V. Tres, H.C. Ferraz, R.M. Dallago, M. Di Luccio, J.V. Oliveira, Characterization of polymeric membranes used in vegetable oil/organic solvents separation, J. Membr. Sci. 362 (2010) 495–500. T.V.R. Alicieo, E.S. Mendes, N.C. Pereira, O.C.M. Lima, Membrane ultrafiltration of crude soybean oil, Desalination 148 (2002) 99–102. A. García, S. Álvarez, F. Riera, R. Álvarez, J. Coca, Sunflower oil miscella degumming with polyethersulfone membranes, effect of process conditions and MWCO on fluxes and rejections, J. Food Eng. 74 (2006) 516–522. A. Dafinov, R. Garcia-Valls, J. Font, Modification of ceramic membranes by alcohol adsorption, J. Membr. Sci. 196 (2002) 69–77. Sigma-Aldrich, Dielectric constant. 〈http://www.sigmaaldrich.com/chemistry/ solvents.html〉, Accessed on September 2014.