Long term pervaporation desalination of tubular MFI zeolite membranes

Journal of Membrane Science 415–416 (2012) 816–823

Contents lists available at SciVerse ScienceDirect

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

Long term pervaporation desalination of tubular MFI zeolite membranes Martin Drobek a, Christelle Yacou b, Julius Motuzas a, Anne Julbe a, Liping Ding b, Joa~ o C. Diniz da Costa b,n a b

Institut Europe´en des Membranes (UMR 5635 CNRS), Universite´ Montpellier 2, CC47, Place Euge ne Bataillon, 34095 Montpellier Cedex 5, France The University of Queensland, FIMLab—Films and Inorganic Membrane Laboratory, School of Chemical Engineering, Brisbane, Qld 4072, Australia

a r t i c l e i n f o

a b s t r a c t

Article history: Received 13 March 2012 Received in revised form 22 May 2012 Accepted 30 May 2012 Available online 5 June 2012

Silicalite-1 (S-1) and ZSM-5 membranes prepared by secondary growth on tubular ceramic supports were tested using a pervaporation set up for the desalination of aqueous solutions containing NaCl in concentrations corresponding to brackish (0.3–1 wt%), sea (3.5 wt%) and brine (7.5–15 wt%) water. ZSM-5 membranes delivered higher water fluxes when compared with S-1 membranes due to enhanced hydrophilicity of the Al-rich zeolite structure leading to fluxes as high as 11.5 kg m  2 h  1 for 0.3 wt% NaCl feed solutions at 75 1C. At higher salt concentration the water flux decreased alongside with the salt rejection rate, however the hydrophilic ZMS-5 membrane became more susceptible to performance loss, particularly at high temperatures. Detailed stability experiments carried out for up to 560 h showed the dissolution of both S-1 and ZSM-5 top layers. This was attributed to the combined effects of ion exchange and water dissolution mechanisms. Interesting though, the MFI structure zeolite was still observed on the XRD patterns whilst EDX depth profile showed the presence of silica up to 30 mm depth into the support. These results suggest the formation/infiltration of a compact amorphous phase in the substrate, derived from the aggregation of the dissolved silicate species and possibly together with the arrangement of MFI nano-slabs during the desalination tests. Despite generating lower water fluxes, the S-1 membrane exhibited relatively high robustness under the long term testing conditions which included temperature cycling, delivering salt rejections from initial 4 99% to o 80% at the end of the testing period (560 h). & 2012 Elsevier B.V. All rights reserved.

Keywords: Desalination Zeolite membranes MFI Silicalite-1 ZSM-5

1. Introduction Zeolite membranes have attracted the concerted effort of the research community for gas [1,2] and liquid phase separations [3,4] and membrane reactor applications [5,6]. The main advantages of zeolite membranes are directly related to their high separation performance, in tandem with catalytic activity, good thermal and chemical stability, and low fouling tendencies [7,8]. These properties make them ideal to operate in processes which cannot be met by conventional polymeric membranes. In addition, the precise tuning of the pore sizes of zeolite membranes confers them great process capabilities for separation processes based on molecular size and shape [9]. In the case of liquid separation, zeolite membranes with relatively large pores have been considered as enabling technologies in reverse osmosis and nano- or ultrafiltration [10,11]. Recent studies have shown the zeolite membranes as unique materials in the field of desalination taking an advantage of their ability to remove ions from aqueous saline solutions, and leading to high ionic (i.e., salt) rejections [12]. The separation efficiency was found to increase with

n

Corresponding author. Tel.: þ61 7 3365 6960; fax: þ61 7 3365 4199. E-mail address: [email protected] (J.C. Diniz da Costa).

0376-7388/$ - see front matter & 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.memsci.2012.05.074

the ion valence suggesting that the separation mechanism is based on rejection of the hydrated ions by size exclusion and ions interactions, which in turn are influenced by the charged double layer formed in inter-crystalline gaps of zeolite membranes [10]. If the separation mechanism based on pore size exclusion cannot be achieved, a functionalization of the zeolite cage might be an option to enhance the membrane separation efficiency as observed in the case of Al-rich zeolite membranes [13], where the membranes properties like surface hydrophobicity and surface charge are significantly changed [14]. The number of zeolite membrane types reported in the literature for desalination is however quite limited. The major constraint of these membranes in desalination application is their pore size allowing the diffusion of the small water molecules ˚ and the larger molecules represented by salt ions e.g., (dk ¼2.6 A) ˚ respecNa þ and Cl  , whose hydrated sizes are 7.2 A˚ and 6.6 A, tively [10,15]. Ideally, if the zeolite pore size and intercrystalline ˚ then zeolite membrane interconnections are well below 6.6 A, would selectively separate water from hydrate salt ions whilst delivering high ion or salt rejections. These criteria can be met theoretically by NaA and MFI zeolite membranes exhibiting 4 A˚ [16] and  5.5 A˚ pore sizes [17], respectively. Table 1 lists the testing parameters and performance of zeolite membranes reported for desalination application. Membranes

M. Drobek et al. / Journal of Membrane Science 415–416 (2012) 816–823

Table 1 LTA and MFI zeolite membranes in desalination applications. Membrane

Salt concentration (NaCl wt%) Transmembrane pressure (MPa) Support geometry Temperature(1C) Flux (kg m  2 h  1) Rejection (%) Reference n

LTA-NaA

MFI-S-1 (Si/Al ¼N)

MFI-ZSM-5 (Si/Al¼ 50–65)

MFI-ZSM-5 (Si/Al ¼100)

3.1

0.5

0.5

3.8

n

nn

nn

0.1

2.07

2.8

0.1n

Tubular 79 4.392 99.9 [18]

Flat 25 0.121 75 [19]

Flat 25 1.129 93 [20]

Flat 80 0.7 99 [12]

Membrane distillation. Reverse osmosis.

817

In the present work, we report the performance for desalination of MFI membranes prepared by easily up-scalable secondary growth from MW-derived nano-seeds on mono-channel tubular ceramic supports and with a rapid heating rate for de-templating. Of particular interest, we compared the water fluxes, salt rejection and long-term stability of silicalite-1 and ZSM-5 zeolite membranes with Si/Al¼N and 100, respectively, in order to investigate the impact of surface hydrophobicity, surface charges and membrane steadiness on its performance in desalination process. All experiments were carried at a transmembrane pressure close to 1 atm, temperatures ranging from 21 1C to 75 1C, and for salt aqueous solutions containing NaCl concentrations corresponding to brackish (1 wt%), sea (3.5 wt%) and brine (7.5– 15 wt%) waters.

nn

were tested in either reverse osmosis (RO) with pressures in excess of 2 MPa or membrane distillation (MD) set up at 0.1 MPa. Results show that the MFI membranes prepared on flat supports with the optimized Si/Al ratio of 50 delivered the best water fluxes and maximum salt rejection. Duke et al. [12] demonstrated that ZSM-5 membranes with a higher Si/Al ratio and prepared by secondary growth gave similar high salt rejections. Generally speaking, the MFI (ZSM-5 and S-1) membranes delivered low water fluxes, associated with poor hydrophilicity as compared to LTA (NaA) membranes, in addition to small channel pore sizes which tend to increase the resistance to the permeation of water. Some level of intercrystalline gaps was also evident as salt rejections varied between 75 and 93% [19,20]. MFI membranes obtained by secondary growth clearly reduced the non-selective transport delivering higher salt rejections of 99%. Nevertheless, ˚ membranes yield water flux six-fold LTA zeolite (  4 A) higher than ZSM-5 membranes operating in MD set up at similar conditions of salt concentration, pressure and temperature. Secondary growth is a highly recognized method for preparing reproducible MFI membranes, although key steps such as seed formation, seed deposition and secondary growth conditions have to be carefully controlled in order to optimize both membrane quality and performance. Microwave-assisted heating (MW) is an attractive method for producing uniform nano-seeds yielding reproducible seed layers on macroporous supports [21]. The seed layer is non-infiltrated and sufficiently thin to avoid any remaining seed excess after the secondary growth step. The non-infiltration of the seed layer into the membrane support is advantageous as it reduces the overall membrane thickness and also the resistance of the preferentially permeated species at the interface of the zeolite film and substrate. Hence, high water flux can be expected as the flux in porous membranes is generally inversely proportional to the thickness of the rate limiting layer, in this case the zeolite film. In parallel, high salt rejection should be achieved with uniform zeolite membranes having a significantly low number of intercrystalline defects. Uniform zeolite layer with defined composition and thickness can be grown in a reproducible way by either MW assisted or classical heating method. By employing these synthesis methods, it has been shown that MFI membranes can be also prepared on pre-assembled capillary supports with high surface to volume ratio ( 41000 m2 m  3) [22], thus providing higher industrial potential when compared to low surface area flat membranes. Another important obstacle to industrial development of zeolite membranes is related to the long calcination step, due to the classically recommended extremely slow heating rates. This drawback can be easily overcome by ozonation [23]. It was also recently reported that a rapid thermal processing improves the quality and separation performance of thick columnar films of silicalite-1 by eliminating the defects, possibly by strengthening grain bonding at the grain boundaries [24].

2. Experimental 2.1. Membrane synthesis and characterization MFI zeolite membranes were synthesized by secondary growth from a layer of silicalite-1 (S-1) nano-seeds deposited on Pall-Exekia asymmetric a-Al2O3 tubular supports (OD/ID¼ 10/7 mm, length¼50 mm, with 5 mm enamel on both ends) and with an internal 200 nm pore size layer. S-1 nanoseeds (50–60 nm size) were prepared by a two-step microwave-assisted synthesis as described elsewhere [25,26]. The molar composition of the sol used for seeds preparation was SiO2: 0.4TPAOH: 19.5H2O: 4C2H5OH and the reaction conditions were the followings: T1 ¼80 1C, t1 ¼90 min, P1 ¼250 W for the first step and T2 ¼125 1C, t2 ¼60 min, P2 ¼400 W for the second step. A suspension of seeds was used to cast a uniform S-1 seed layer inside the a-Al2O3 supports by dip-coating. The seeded supports were placed vertically in an autoclave containing a sol whose composition was adjusted for the membrane growth step: 3TPAOH: 25SiO2: 100C2H5OH: 1500H2O for S-1 membranes and 3TPAOH: 0.125Al2O3: 25SiO2: 100C2H5OH: 1500H2O for ZSM-5 (Si/Al¼100) membranes. Membranes were grown at 160 1C for 9 h by conventional heating. After the synthesis the membranes were washed with distilled water, dried for 3 h at 155 1C and finally calcined in air, applying a heating rate of 17.7 1C min  1 up to 550 1C and a 4 h dwell time followed by cooling down to room temperature (20 1C min  1). The morphology, thickness and homogeneity of both seed layers and membranes were studied by field emission scanning electron microscopy (FESEM, Hitachi S-4500). The chemical composition of both surface and cross section of the membranes was analyzed by energy-dispersive X-ray spectroscopy (EDX, Quanta 200 FEG Electron Microscopy). The membrane crystalline structure was examined by X-ray diffraction (XRD, PANanalytical X-Pert Pro). 2.2. Membrane testing All the prepared membranes were initially tested by single gas permeation measurements to ensure their good quality in terms of affirming the absence of intercrystalline macro-defects (viscous flow contribution) and other irregularities in the zeolitic network. The permeation experiments were carried out in a stainless steel module with a dead-end set up equipped with silicon o-rings to fix the membrane. Single gas permeation was conducted (after outgassing the membranes) for N2 (dk ¼0.364 nm) and SF6 (dk ¼0.55 nm) at 21 1C, applying a transmembrane pressure DP¼100 kPa. Gas permeance was measured by a bubble flow meter connected to the atmosphere. The desalination testing was carried out using a classical pervaporation set-up schematically depicted in Fig. 1. The desalination

818

M. Drobek et al. / Journal of Membrane Science 415–416 (2012) 816–823

process was operated at a dead end mode, where the outer shell of the membrane was exposed to a feed salt solution and the permeate stream was connected to a vacuum pump through a cold trap. The salt solution was circulated/recycled from the feed tank with the help of a peristaltic pump (flow rate of 40 ml min  1), enabling a constant stirring to prevent concentration polarization on the feed side of the membrane. The feed solution contained NaCl (98%-Sigma Aldrich) in concentrations from 0.3 wt% to 15 wt% in deionized water in order to simulate the typical salt level of brackish water (0.3–1 wt%), sea water (3.5 wt%) and brine stream (7.5–15 wt%). For

Permeate stream

Recycling stream

each salt concentration, the temperature of the feed solutions was set at 21 1C, 50 1C and 75 1C using a temperature controlled hotplate and the permeate was collected at specific time intervals from the cold traps. The conductivities of both feed and permeate solution were measured by a labCHEM TPS conductivity meter. The permeate flux F (kg m  2 h  1) was determined according to the equation F¼m/(At), where ‘‘m’’ is the mass (kg) of the permeate retained in the cold trap, ‘‘A’’ (m2) is the membrane active area and ‘‘t’’ is the experiment duration (h). After measuring the permeate flux, the same water was run through the permeate stream of the membrane tube to flush any residual salt that may have crossed the zeolite film and crystallized in the substrate. The membrane salt rejection R (%) was calculated as R¼(Cf  Cp)/Cf  100%, where ‘‘Cf’’ and ‘‘Cp’’ are the feed and flushed permeate concentrations (wt%), respectively.

3. Results and discussions

Cold trap

3.1. Characteristics of the selected MFI membranes (silicalite-1 and ZSM-5) Membranes

Hot plate

Liquid N2

Peristaltic pump

Vacuum pump

Fig. 1. Pervaporation set up for desalination applications.

10

15

20 2θ,°

25

30

Prior to the desalination tests, the silicalite-1 (S-1) and ZSM-5 membranes were carefully characterized by scanning electron microscopy, XRD measurements and subjected to single gas permeation tests. FE-SEM micrographs of the cross-section and top-surface of the membranes are shown in Fig. 2. In both cases,

10

15

20 2θ,°

25

30

Fig. 2. FESEM micrographs of MFI membrane cross Section (a1, b1) and surfaces (a2, b2) (a) S-1 and (b) ZSM-5 (Si/Al ¼ 100)), and (a3, b3) corresponding XRD patterns.

M. Drobek et al. / Journal of Membrane Science 415–416 (2012) 816–823

homogeneous layers were obtained with a thickness of about 6 mm and 3.3 mm for S-1 and ZSM-5 membranes (Fig. 2a1 and b1), respectively. In a similar fashion, Fig. 2a2 and b2 revealed homogeneous top surfaces, made with uniform and interconnected zeolite crystals and showing no visible macro-defects. The XRD patterns in Fig. 2a3 and b3 are characteristic for the MFI structures with relevant 2y peaks at regions: 7–91, 12–151 and 21–251. Single gas permeances for N2 and SF6 alongside with the corresponding ideal selectivities measured at 21 1C are listed in Table 2. As expected, the smaller molecule (N2) permeated much faster than the larger molecule (SF6), resulting in reasonably high permselectivities of  40 for both membranes. These results clearly indicate that the membranes were free of any macro defects, otherwise the permeance of both gases would be much closer, leading to lower selectivity values. Although N2/SF6 permselectivity is a method commonly used in literature to assess MFI membrane integrity/quality, one must bear in mind that permeation measurements in the presence of molecules that swell the crystals, such as SF6, cannot clearly discriminate between membranes with few intercrystalline defects and those with many defects [24]. However, according to our previous studies [22] this test was found to give valuable indication of the membrane performance for ethanol/water separation. Therefore, in the present work, we chose membranes with a N2/SF6 permselectivity higher than 30 for the desalination tests Therefore, in the present work, we chose membranes with a N2/SF6 permselectivity higher than 30 for the desalination tests which are at least one order of magnitude higher than the ideal Knudsen selectivity of 2.28 for these two gases. Hence, a cut off

Table 2 Single gas permeance data (7 2%) for the studied silicalite-1 and ZSM5membranes. Gas permeation (mol s  1 m  2 Pa  1)

N2

SF6

N2/SF6 permselectivity

Silicalite-1 ZSM-5 (Si/Al¼ 100)

1.16  10  6 1.73  10  6

2.97  10  8 4.09  10  8

39.1 42.4

3.2. Desalination performance of the MFI (silicalite-1 and ZSM-5) membranes 3.2.1. Influence of salt concentration and temperature on membrane performance Fig. 3 shows the water fluxes and salt rejection of the membranes tested under varying salt feed concentrations and temperatures. At the initial 0.3 wt% salt feed concentration, the water flux increased with temperature for both membranes. For the S-1 membranes, the water flux increased by 15% to 57% as the feed temperature was raised from 21 1C to 75 1C, respectively, reaching a maximum value of 11.5 kg m  2 h  1 at 75 1C. Under these experimental conditions, ZSM-5 membranes provided higher water fluxes than S-1 membranes. This result is attributed first to the smaller ZSM-5 membrane thickness as observed from the FE-SEM micrographs (Fig. 2), but mainly to the higher hydrophilicity of the ZSM-5 structure enabling better wettability of the membrane and favouring water transport. The measurements carried out at higher feed concentrations revealed that water fluxes are reduced. Contrary to this trend, the water fluxes increased for the ZSM-5 membranes when the salt concentration was higher than 3.5 wt%. This improvement was probably caused by a partial alteration of the membrane structure operating in the harsh salty environment, as far as this flux increase occurred with a significant decrease of the membrane salt rejection efficiency. High salt feed concentration induced changes in the protective electrically charged double layer which form in the zeolite pores, thus affecting salt rejection in tandem with water fluxes. Further inspection of the salt rejection parameter reveals that the ZSM-5 membrane started losing its separation efficiency shortly after the initial testing period (salt feed concentration at 0.3%). The loss of performance was further exaggerated when the feed concentration exceeded 3.5%, as salt rejections decreased from  90% to less than 75%. Additional tests also revealed that the increase of temperature further altered the ZSM-5 membrane efficiency, as the salt rejection became as low as  62% at 75 1C. These results confirm the poor stability of the ZSM-5 membrane

25

25

20

20

15

15

10

10

5

5

100

90

90 T=20°C T=50°C T=75°C

80

80

70 60

70

2.5

5.0

7.5

10.0 12.5 15.0 5.0 7.5 2.5 Salt concentration (wt%) in the feed

10.0

12.5

15.0

Salt Rejection (%)

100

Flux (kg.m-2.h-1)

Flux (kg.m-2.h-1)

permselectivity value of 30 is an effective criterion for selecting high quality MFI zeolite membranes.

ZSM-5

S-1

Salt Rejection (%)

819

60

Fig. 3. Desalination performance of membranes S-1 (left) and ZSM-5 (right) in terms of permeate flux and salt rejection at different operation temperatures.

820

M. Drobek et al. / Journal of Membrane Science 415–416 (2012) 816–823

Time (hour) 12

200

300

400

500

600

Salt water

S-1 ZSM-5

10 Flux (kg m-2 h-1)

100

0

8 6 4 2 Salt concentration 7.5w%

Temperature (°C)

Salt concentration 3.5w%

80

75°C

60

50°C

Salt Rejection (%)

50°C

40 21°C

21°C

20

Pure water flux (kg m-2 h-1)

75°C

21°C

80 60 40 20 20

S-1 ZSM-5

15 10 5 0

0

100

200

300 Time (hour)

400

500

600

Fig. 4. Long term testing of membranes S-1 (circle) and ZSM-5 (triangle): (a) water flux under aqueous saline solution feed (b) during thermal cycling, (c) corresponding salt rejections, and (d) water flux under pure water feed.

in the desalination process, the membrane degradation being further enhanced at high salt concentration and temperatures. In comparison, the S-1 membrane was found to be much more stable during the whole testing period, yielding high salt rejections in excess of 96% even when the temperature was ramped up to 75 1C. It is of particular interest to note that the salt rejections for the S-1 membrane at 21 1C were generally in excess of 99% irrespective of the salt concentration in the feed.

3.2.2. Evolution of membrane performance upon cycling and long term operation In order to investigate the long-term performance of both membranes in MD operation, stability tests were conducted under fresh water and various salt concentrations and temperatures for 560 h cycling. The cycling conditions and the corresponding evolution of the membrane performance are depicted in Fig. 4. During the first 300 h, the salt concentration in the feed was set at sea water salt concentration (3.5 wt%) and the temperature of the feed streams was varied up and down between 21 1C and 75 1C. For testing times in excess of 300 h,

the salt concentration was increased to 7.5 wt% in tandem with a thermal cycling between 21 1C and 75 1C. As shown in Fig. 4a, both membranes displayed similar performance during the first cycling step. Upon completing the tests at 75 1C, temperature was reduced to 21 1C, and the water fluxes and salt rejections at 300 h were only recovered for the S-1 membrane as compared to the initial performance below l00 h. The ZSM-5 membrane was slightly affected by these test conditions resulting in minor increase of permeate flux although this came at the loss of salt rejection from o99% to 95%. In the following cycling run (from 300 to 560 h and 7.5 wt% salt feed concentration), the ZSM-5 membrane underwent severe structural degradation, resulting in the simultaneous increase in water flux with significant decrease in salt rejection to below 30%. At these testing conditions, the S-1 membrane also showed performance decay as salt rejection decreased from 96% to 82%. These results strongly suggest that high salt concentration and temperature detrimentally affected both membranes, but this effect was more severe for the ZSM-5 membranes. After 530 h of testing, the water flux of the S-1 membrane did not change as the temperature was reduced from 75 to 21 1C, and did not revert to the initial values measured at 300 h. This clearly indicates the negative impact of high salt concentration on the membrane integrity. Nevertheless, compared to the unstable ZSM-5 membranes, the S-1 membranes exhibited superior long term stability at all temperature and concentration ranges. Both ZSM-5 and S-1 membranes were also tested using salt free water (e.g., deionised water) under the same temperature and cycling conditions as previously described. The results are depicted in Fig. 4d. The S-1 membrane showed almost no changes in water flux for each temperature from the first to the second cycle, thus demonstrating good long term hydrostability properties. The ZSM-5 membrane showed small variations, particularly at 75 1C in the second cycle, though upon cooling to 25 1C after 530 h the water flux increased by 40%. These results therefore suggest that ZSM-5 membranes underwent structural alteration to prolonged high temperature exposure concurrently with temperature cycling, though not as severe as salt solution exposure as observed in Fig. 4a.

3.3. Membrane evolution during long term operation and mechanism postulations In order to understand the observed loss of performance, postexperimental characterization of both membranes submitted to long term desalination testing (560 h, between 21 and 75 1C and up to 7.5 wt% NaCl) was carried out by SEM and XRD. The micrographs presented in Fig. 5a–c clearly display significant alteration of the zeolite layer for both types of membranes. The uniform zeolite intergrowth top layer (6 and 3.3 mm thick, for S-1 and ZSM-5, respectively) which was evidenced prior to testing (Fig. 2a and b) had completely disappeared. Instead, a very thin (  0.15 mm) indented structure was now detected on the surface (Fig. 5a2 and b2). These post testing observations clearly indicate that the spherical holes in the indented structure fits the size of the S-1 seeds (50–100 nm) used for growing the two membranes and thus correspond to their remaining footprint. If this is true, the new material which has formed between the seeds during the long term desalination test is certainly very compact and much less sensitive to salt dissolution in comparison with the very porous MFI zeolite membrane and seeds. A progressive dissolution of both the zeolite membrane and remaining seed layer can then be suggested from these observations, revealing the competition between the dissolution and the formation of a new material compacted within the underlying support porosity.

M. Drobek et al. / Journal of Membrane Science 415–416 (2012) 816–823

3.00 µm

3.00 µm

3.00 µm

3.00 µm

[011]/ [101]

[011]/ [101] [020]/ [200]

Al O [020]/ [200] [002] [012]

10

15

[511] [051] [033] [501]

20 2θ,°

25

821

Al O

[002] [012]

30

10

15

[511] [051] [033] [501]

20 2θ,°

25

30

Fig. 5. FESEM micrographs of MFI membrane for (a) S-1 and (b) ZSM-5 (Si/Al¼ 100) membranes. (a1, b1) Cross sections, (a2, b2) surfaces and (a3, b3) XRD patterns.

0.25

Ratio Si/Al

0.20 0.15 ZSM-5 S-1

0.10 0.05

0

5

10 15 20 Depth in α-Al2O3 support (μm)

25

30

Fig. 6. EDX depth profile for S-1 and ZSM-5 membranes tested for 560 h under desalination experimental conditions.

Interestingly, the XRD patterns in Fig. 5a3 and b3 still confirm the presence of the MFI diffraction lines in the 2y range of 7.5–91 and 13–141 for the membranes that underwent stability testing,

although their relative intensity is slightly reduced in comparison with the intense diffraction line of the a-Al2O3 substrate at 2y ¼25.51. This point is certainly surprising as the zeolite top layers were basically destroyed. This result demonstrates that the MFI structure still remained (at least partially) in the top layer and/or in the new composite membrane material derived from the dissolution of the zeolite top layer and the forced transportation of the species through the support. This could suggest the formation of zeolite nano-slabs during the dissolution process, which might partially rearrange when forced through the support porosity. The mechanisms leading to the formation of zeolites can be quite complex but this type of rearrangement cannot be ignored, considering the temperature (75 1C) and high pressure gradient applied in the pores of the support under operation as reported elsewhere [27]. These results confirm that the MFI zeolite membranes layers underwent dissolution when exposed to long term desalination experimental conditions, leading to an increase in the size and number of intercrystalline pores. This process occurs through the dissolution of synthesis residues and amorphous silica [28]. Indeed, silica tends to be hydrophilic, and silanol groups become mobile from the silica matrix, shifting to the low energy areas at the entrance of smaller pores, thus forming large pores [29]. Unless silica can be anchored by surfactants [30–33] or metal oxides [34] in water/desalination processes, the continuous reaction of silica and water favours pore size enlargement. In agreement with the above comments, post-analysis gas permeation revealed an increase in the SF6 permeance up to

822

M. Drobek et al. / Journal of Membrane Science 415–416 (2012) 816–823

3.0  10  7 (ZSM-5) and 4.2  10  7 (S-1) mol m  2 s  1 Pa  1, an increase of one order of magnitude as compared to the values obtained prior to testing as seen in Table 2. The N2/SF6 selectivity was reduced by almost four-fold to  11 for both membranes, thus also suggesting a significant increase in pore size as a result of the long term test which altered the membrane’s microporosity. This correlates well with the evolution of the desalination performance, as the dissolution of the zeolite top-layer leads to enhanced water flux at the expense of salt rejection. However, this does not explain the presence of zeolites in the material resulting from dissolution after long term desalination exposure. To shed further light on this important issue, we carried out EDX analysis of the long term tested membrane cross-sections at several depths. Fig. 6 shows both the depth profile evolution of the Si/Al ratio in the membrane, and the EDX patterns of representative points. These results clearly indicate that silica species infiltrated the substrate structure up to 30 mm depth. In tandem with the confirmation of XRD zeolite patterns in Fig. 5, we infer that the silica measured by EDX is related to a mixture of silica and MFI zeolite arranged nano-slabs in an amorphous but quite compact matrix. Further analysis of Fig. 6 reveals that the Si/ Al ratios were quite similar for both membranes at 0.18 for S-1 and 0.14 for ZSM-5, close to the membrane surface, though the trend diverged from thereon. The Si/Al ratio for the ZMS-5 membrane greatly reduced to 0.03 at 4 mm and it remained almost constant all the way down to 30 mm depth. Contrary to this trend, the S-1 membrane Si/Al ratio was kept almost constant (  0.18) down to 12 mm depth. These results suggest that the composite material (dissolved zeolite nano-slabs, silica and/aAl2O3 support) was in the region of o1 and 14–16 mm for the ZSM-5 and S-1 membranes, respectively. In addition, the Si/Al depth profile trends resemble the performance trends observed in Figs. 2 and 3, where the S-1 membrane delivered superior salt rejection than the ZSM-5 membrane. This implies that the combined ZSM-5/a-Al2O3 structure no longer provided pore size exclusion. On the other hand, the much thicker combined S-1/aAl2O3 structure maintained microporosity, as the rejection of the large hydrate ions was reasonably high and above 80% even at 560 h exposure to sea/brine waters. Although the membranes were fully washed with deionized water after the long term testing, we still observed by EDX the presence of Na þ at a few measured points of the ZSM-5 membrane surface, but not on the S-1 membrane. Duke and coworkers [35] reported that zeolite powders interact with sea salt solutions containing Na þ , K þ , Ca2 þ , and Mg2 þ , whereas the ion activity was observed at the microporous and mesoporous intercrystalline regions. The ion exchange mechanism is directly related to the charge attraction of Na þ with the negative charge of the zeolite cage. It is known that ion exchange interaction between MFI and Na þ strongly impacts on the diffusion properties, generating an increase of intercrystalline pore size and can result in the destruction of the zeolitic material [34]. This supports the case for the accelerated loss of performance of the ZMS-5 membrane, and possibly S-1 membranes at the highest salt concentrations. We therefore postulate that the mechanisms of water dissolution and ion exchange are responsible for the partial destruction/ restructuring of the MFI zeolite membranes. The ion exchange occurs within the zeolite cage by charge transfer, thus weakening and altering the zeolite crystalline structure. Concomitantly, water is constantly accessing and reacting with the altered zeolite structure, particularly the hydrophilic silica groups. These groups are mobile and tend to move towards the low energy areas, at the entrance of smaller pore sizes, and leaving behind larger pore sizes, thus explaining the increase of water flux and reduction of salt rejection, particularly for the thinner ZSM-5 membrane. Water per se also

affected the crystal structure of ZMS-5, though to a much lesser degree. As this process of combined mechanisms continues with time, the initial zeolite top layer is destroyed by alteration of the intercrystalline connections. The subsequent infiltration of the solubilized silicate species and MFI nano-slabs into the support forms a compact phase that is essentially amorphous, thus explaining the remaining X-ray diffraction lines. The present work clearly points out that MFI zeolite membranes are susceptible to structural changes and loss of performance when exposed to aqueous solutions containing salt. Hence, there is a requirement to build zeolite cages which can oppose ionic exchange. Potential pathways to address this problem could include substituting silica with titania [36], or phosphorous in ZMS-5 zeolites [37]. Another alternative could include the preparation of bi-layered zeolite structure where a more stable toplayer could serve as a preventative barrier protecting the lower and less stable zeolite membrane as suggested by O’Brien and coworkers [38]. These strategies may provide future directions to tackle the long term effect of salts on the stability of zeolite membranes for desalination processes.

4. Conclusions MFI zeolite membranes grown on tubular ceramic supports pre-seeded with microwave derived nano-seeds were found to exhibit attractive desalination performance, with relatively high water fluxes and salt retention values. The hydrophilic Al-rich ZSM-5 (Si/Al ¼100) membranes delivered high fluxes than the organophilic S-1 (Si/Al¼N) membranes. At high salt concentrations the water flux decreased for all membranes. This was attributed to the reduced vapor pressure due to increased salt concentration. Concurrently, salt rejection was reduced mainly due to the dissolution of the protective electrically charged double layer in the zeolite pores. ZSM-5 membranes were severely affected by the testing conditions of salt feed concentration and temperature cycling, thus displaying poor stability for desalination processing. Long term exposure (560 h) confirmed a significant decrease in performance of the ZSM-5 membrane performance, as both salt concentration and temperature severely affected the zeolite structure. Despite delivering lower water fluxes, the long term testing confirmed that the S-1 membranes were robust enough for processing sea water (o3.5 wt%), though high brine concentrations led to loss of salt rejection from 499% to  80%. Extensive post-testing characterization (after 560 h, between 21 and 75 1C and up to 7.5 wt% NaCl) showed that both S-1 and ZSM-5 top layers were destroyed although it was still possible to discern relatively intense MFI diffraction lines for both membranes. Dissolution of the zeolite membranes explained the loss of performance. The experimental conditions during the long term desalination testing therefore created a driving force for the dissolution of the zeolite top films by attacking the connections between crystals. The formation/infiltration of a compact amorphous phase deriving from the aggregation of the silicate species and arrangement of MFI nano-slabs during the tests is a possible hypothesis supporting the characterization results and membrane performance. The penetration of the dissolved species into the a-Al2O3 substrate formed a partially amorphous zeolitic phase which penetrated up to 14–16 mm in depth for S-1, and only o1 mm in depth for ZSM-5, whilst silica was observed up to 30 mm depth for both membranes. This trend corresponded quite well with the long term performance testing of the membranes, as the thicker recombined S-1/a-Al2O3 membrane was able to maintain relatively high salt rejections, contrary to the severely affected and thinner ZSM-5/a-Al2O3 membrane.

M. Drobek et al. / Journal of Membrane Science 415–416 (2012) 816–823

References [1] F. Bonhomme, M.E. Welk, T.M. Neno, CO2 selectivity and lifetimes of high silica ZSM-5 membranes, Microporous Mesoporous Mater. 66 (2003) 181–188. [2] G.R. Gavalas, Materials Science of Membranes for Gas, Vapor Separation, in: B. Freeman (Ed.), Ed.,John Wiley and Sons Ltd., 2005. [3] L. Zhou, T. Wang, Q.T. Nguyen, J. Li, Y. Long, Z. Ping, Cordierite-supported ZSM-5 membrane: preparation and pervaporation properties in the dehydration of water–alcohol mixture, Sep. Purif. Technol. 44 (3) (2005) 266–270. [4] S. Sommer, T. Melin, Performance evaluation of microporous inorganic membranes in the dehydration of industrial solvents, Chem. Eng. Process. 44 (2005) 1138–1156. [5] A. Julbe, D. Farrusseng, J.C. Jalibert, C. Mirodatos, C. Guizard, Characteristics and performance in the oxidative dehydrogenation of propane of MFI and V-MFI zeolite membranes, Catal. Today 56 (2000) 199–209. [6] Z.A.E.P Vroon, A. Holmen, J.T.F. Keurentjes, Preparation of zeolite-coated pervaporation membranes for the integration of reaction and separation, Catal. Today 104 (2005) 288–295. [7] E.E. McLeary, J.C. Jansen, F. Kapteijn, Zeolite based films, membranes and membrane reactors: progress and prospects, Microporous Mesoporous Mater. 90 (2006) 198–220. [8] A. Julbe, Zeolite membranes—synthesis, characterization and applications, in Introduction to Zeolite Science and Practice, in: H. van Bekkum, J. Cejka, A. Corma, F. Schuth (Eds.), Studies in Surface Science and Catalysis, second ed.,Elsevier, Amsterdam, 2007, pp. 181–219, Chapter 6. [9] T.C Bowen, R.D. Noble, J.L. Falconer, Fundamentals and applications of pervaporation through zeolite membranes, J. Membr. Sci. 245 (2004) 1–33. [10] L. Li, J. Dong, T.M. Nenoff, R. Lee, Desalination by reverse osmosis of ionic aqueous solution on MFI zeolite membranes, Desalination 170 (2004) 309–316. [11] I. Galleo, R. Mallada, E.P. Urriolabeitia, R. Navarro, M. Memendez, J. Santamaria, Selective separation of homogeneous catalysts using silicalite-1 membrane, Inorg. Chim. Acta 357 (2004) 4577–4581. [12] M.C. Duke, J. O’Brien-Abraham, N. Milne, B. Zhu, J.Y.S. Lin, J.C. Diniz da Costa, Seawater desalination performance of MFI type membranes made by secondary growth, Sep. Purif. Technol. 68 (2009) 343–350. [13] V.A. Tuan, S. Li, J.L. Falconer, R.D. Noble, Separating organics from water by pervaporation with isomorphously-substituted MFI zeolite membranes, J. Membr. Sci. 196 (2002) 111–123. ¨ ¨ [14] M. Noack, P. Kolsch, A. Dittmar, M. Stohr, G. Georgi, R. Eckelt, J. Caro, Effect of crystal intergrowth supporting substances (ISS) on the permeation properties of MFI membranes with enhanced Al-content, Microporous Mesoporous Mater. 97 (2006) 88–96. [15] J. Lin, S Murad, A computer simulation study of the separation of aqueous solutions using thin zeolite membranes, Mol. Phys. 99 (2001) 1175–1181. [16] D. Shah, K. Kissick, A. Ghorpade, R. Hannah, D. Bhattacharyya, Pervaporation of alcohol-water and dimethylformamide-water mixtures using hydrophilic zeolite NaA membranes: mechanisms and experimental results, J. Membr. Sci. 179 (2000) 185–205. [17] Z. Lai, G. Bonilla, I. Diaz, J.G. Nery, K. Sujaoti, M.A. Amat, E. Kokkoli, O. Terasaki, R.W. Thompson, M. Tsapatsis, D.G. Vlachos, Microstructural optimization of a zeolite membrane for organic vapor separation, Science 300 (2003) 456–460. [18] C.H. Cho, K.Y. Oh, S.K. Kim, J.G. Yeo, P. Sharma, Pervaporative seawater desalination using NaA zeolite membrane: mechanisms of high water flux and high salt rejection, J. Membr. Sci. 371 (2011) 226–238. [19] L. Li, J. Dong, T.M. Nenoff, R. Lee, Desalination by reverse osmosis using MFI zeolite membranes, J. Membr. Sci. 243 (2004) 401–404.

823

[20] L. Li, N. Liu, B. McPherson, R. Lee, Enhanced water permeation of reverse osmosis through MFI-Type zeolite membranes with high aluminum contents, Ind. Eng. Chem. Res. 46 (2007) 1584–1589. [21] J. Motuzas, A. Julbe, R.D. Noble, A. van der Lee, Z.J. Beresnevicius, Rapid syntheoriented silicalite-1 membranes by microwave-assisted hydrothermal treatment, Microporous Mesoporous Mater. 92 (2006) 259–269. [22] M. Drobek, J. Motuzas, M. van Loon, R.W.J. Dirrix, R.A. Terpstra, A. Julbe, Coupling microwave-assisted and classical heating methods for scaling-up MFI zeolite membrane synthesis, J. Membr. Sci. DOI: 10.1016/j.memsci. 2012.01.045. [23] J. Motuzas, S. Heng, P.P. Sze Lau, K.L. Yeung, Z.J. Beresnevicius, A. Julbe, Ultrarapid production of MFI membranes by coupling microwave-assisted synthesis with either ozone or calcination treatment, Microporous Mesoporous Mater. 99 (2007) 197–205. [24] J. Choi, H.-K. Jeong, M.A. Snyder, J.A. Stoeger, R.I. Masel, M. Tsapatsis, Grain boundary defect elimination in a zeolite membrane by rapid thermal processing, Science 325 (2009) 590–593. [25] J. Motuzas, A. Julbe, R.D. Noble, C. Guizard, Z.J. Beresnevicius, D. Cot, Rapid synthesis of silicalite-1 seeds by microwave assisted hydrothermal treatment, Microporous Mesoporous Mater. 80 (2005) 73–83. [26] V. Sebastian, J. Motuzas, R.W.J Dirrix, R.A. Terpstra, R. Mallada, A. Julbe, Synthesis of capillary titanosilicalite TS-1 ceramic membranes by MWassisted hydrothermal heating for pervaporation application, Sep. Purif. Technol. 75 (2010) 249–256. [27] R. Singh, P.K. Dutta, MFI: a case study of zeolite synthesis, in: Scott M. Auerbach, Kathleen A. Carrado, Prabir K. Dutta (Eds.), Handbook of Zeolite Science & Technology, Ed.,Marcel Dekker, Inc., 2003, chapter 2. [28] C. Andersson, J. Hedlund, Effects of exposure to water and ethanol on silicalite-1 membranes, J. Membr. Sci. 313 (2008) 120–126. [29] M.C. Duke, J.C. Diniz da Costa, D.D. Do, G.Q. (Max), Lu,P.G. Gray, D. Thompsett, Hydrophilic hydrostable functionalised silica surfaces, Adv. Funct. Mater. 16 (2006) 1215–1220. [30] S. Wijaya, M. Duke, J.C. Diniz da Costa, Carbonised template silica membranes for Desalination, Desalination 236 (2009) 291–298. [31] J. Campaniello, C.W.R. Engelen, W.G. Haije, P.P.A.C. Pex, J.F. Vente, Long-term pervaporation performance of microporous methylated silica membranes, Chem. Commun. 7 (2004) 834–835. [32] M.C. Duke, S. Mee, J.C. Diniz da Costa, Performance of porous inorganic membrane in non-osmotic desalination, Water Res. 41 (17) (2007) 3998–4004. [33] B.P. Ladewig, Y.H. Tan, C.X.C. Lin, K. Ladewig, J.C. Diniz da Costa, S. Smart, Preparation, characterisation and performance of templated silica membranes in non-osmotic desalination, Materials 4 (2011) 845–856. [34] C.X.C. Lin, L. Ding, S. Smart, J.C. Diniz da Costa, Cobalt oxide silica membranes for desalination, J. Colloid Interface Sci. 368 (2012) 70–76. [35] B. Zhu, L. Zou, C.M. Doherty, A.J. Hill, Y.S. Lin, X. Hu, H. Wang, M. Duke, Investigation of the effects of ion and water interaction on structure and chemistry of silicalite MFI type zeolite for its potential use as a seawater desalination membrane, J. Mater. Chem. 20 (2010) 4675–4683. [36] J. Motuzas, R. Mikutaviciute, E. Gerardin, A. Julbe, Controlled growth of thin and uniform TS-1 membranes by MW-assisted heating, Microporous Mesoporous Mater. 128 (2010) 136–143. [37] G. Caeiro, P. Magnoux, J.M. Lopes, F.R. Ribeiro, S.M.C. Menezes, A.F. Costa, H.S. Cerqueira, Stabilization effect of phosphorus on steamed H-MFI zeolites, Appl. Catal. A: General 314 (2006) 160–171. [38] J. O’Brien-Abraham, M. Duke, Y.S. Lin, Xylene separation performance of composition-gradient MFI Zeolite membranes, Membr. Sci. Technol. 14 (2011) 195–212.