Influence of the hydrothermal synthetic parameters on the pervaporative separation performances of CHA-type zeolite membranes

Microporous and Mesoporous Materials 143 (2011) 270–276

Contents lists available at ScienceDirect

Microporous and Mesoporous Materials

journal homepage: www.elsevier .com/locate /micromeso

Influence of the hydrothermal synthetic parameters on the pervaporativeseparation performances of CHA-type zeolite membranes

Xiansen Li a,⇑, Hidetoshi Kita a, Hua Zhu a, Zhenjia Zhang b, Kazuhiro Tanaka a, Ken-ichi Okamoto a

a Environmental Science and Engineering, Graduate School of Science and Engineering, Yamaguchi University, Tokiwadai 2-16-1, Ube, Yamaguchi 755-8611, Japanb School of Environmental Science and Engineering, Shanghai Jiao Tong University, Shanghai City, China

a r t i c l e i n f o a b s t r a c t

Article history:Received 18 December 2010Received in revised form 3 March 2011Accepted 4 March 2011Available online 9 March 2011

Keywords:Chabazite-structured zeoliteZeolite membraneDehydrationPervaporation process

1387-1811/$ - see front matter � 2011 Elsevier Inc. Adoi:10.1016/j.micromeso.2011.03.006

⇑ Corresponding author. Present address: UniversApplied Energy Research (CAER), 2540 Research Park8410, United States. Tel.: +1 859 257 0271; fax: +1 8

E-mail address: [email protected] (X. Li).

The hydrothermal synthesis of intermediate-silica CHA-type zeolite membranes from the precursor mix-tures containing single K+ alkali cations was studied in detail under different conditions, including theinfluencing factors of silica source, the crystallization temperature, water concentration and gel alkalinityof the initial gels. Additionally, CHA-type zeolite membranes were also prepared tentatively by the inter-zeolitic transformation method. The pervaporation (PV) measurements on these resulting CHA-typemembranes for the dehydration of alcohol aqueous solutions were performed to investigate the qualityof these membranes. As a result of these pervaporative studies, the optimal synthetic recipe was formu-lated with an optimized crystallization temperature spanning from 130 to 150 �C. The PV data showedthat the separation performances of CHA membranes obtained were strongly dependent on the type ofsilica source employed with Ludox HS-30 being the optimal silica source. Furthermore, synthetic temper-ature had a strong influence on the morphology of the resultant CHA membranes. Under the optimizedsynthetic condition, the total permeation flux and separation factor of a representative membrane were2.20 kg/m2 h and 3900, respectively, for the separation of 10 wt.% H2O/EtOH mixtures at 75 �C. The CHAmembranes were characterized by X-ray diffraction, field emission-scanning electron microscopy andenergy dispersive X-ray spectroscopy.

� 2011 Elsevier Inc. All rights reserved.

1. Introduction

Chabazite zeolites are built from double 6-membered rings(D6Rs) linked by tilted 4-membered rings (MRs) to give the overallstructure [1]. This results in a tri-dimensional pore channel systemthat contains 8-MR pore windows. These 8-MR windows have anestimated dimension of 3.7 � 4.2 Å2 (hydrated) or 3.1 � 4.4 Å2

(dehydrated). The framework Si/Al atomic ratio of CHA-type zeo-lites can be finely tuned from ca. 2 to 1 by dosing optionally anappropriate structure-directing agent (SDA) (e.g., N,N,N-tri-methyl-1-adamantammonium hydroxide) into a synthetic gelprecursor.

Owing to the environmental effect of CO2 emission on globalwarming, there has been a substantial focus on improving tech-niques for CO2 capture and sequestration. It has been previouslyreported that among many kinds of naturally occurring and syn-thetic zeolites, chabazite (CHA) and 13X are most suitable forCO2 separation by pressure swing adsorption techniques [2]. In an-

ll rights reserved.

ity of Kentucky, Center forDrive, Lexington, KY 40511-

59 257 0302.

other study covering adsorption-based CO2 capture [3], it wasfound that NaCHA and CaCHA held comparative advantages forhigh temperature CO2 separation while NaX zeolites showed supe-rior performance at relatively low temperatures. Therefore, theCHA zeolite membrane was herein chosen for intensive researchwith an attempt to evaluate its CO2 separation behavior later.

Naturally occurring chabazite has been used successfully atsome nuclear facilities to remove radioactive cesium-137 andstrontium-90 from processing waters [4,5]. It is also used commer-cially as a desiccant. It was reported that water vapor sorption dataon chabazite zeolites demonstrated a value of 0.35 g/g at P/Ps = 0.95 and 20 �C [6], whereas the corresponding value obtainedon NaA zeolites was 0.25 g/g at P/Ps = 0.9 and 25 �C [7]. Moreimportantly, they are stable at a pH of 2.5, which renders themsuitable for removing water from hydrogen chloride gas streams,as well as trace gas removal, such as nitrogen removal from argon,or oxygen separation from argon [8–10]. Quite recently, Hasegawaet al. disclosed that chabazite zeolite membrane could separateselectively water from an ethanol solution of around pH 2 for10 h with the feed solution acidity adjusted by adding 1 ml ofHCl solution (10 M) [11]. On the other hand, with recourse to suchcharacteristics as strong hygroscopic properties, proton conductorsand moderate acid tolerance, H-form chabazite zeolites were

X. Li et al. / Microporous and Mesoporous Materials 143 (2011) 270–276 271

incorporated into proton exchange membranes (e.g., Nafion) to en-hance high-temperature proton conductivity of these membranesby increasing water retention in the composite membranes at ele-vated operating temperatures (e.g., >100 �C) for the application inhigh-temperature direct methanol fuel cells [12]. Furthermore,chabazite membranes may be potentially applied for pervapora-tion-aided etherification or esterification to break up these equilib-rium-limited reactions, just as the roles of T-type and ZSM-5zeolite membranes [13,14].

Al-rich chabazite zeolite membranes have not been so far re-ported in the open literature except for Refs. [11,15,16], where itwas claimed that the polycrystalline CHA membranes could notbe grown successfully onto the porous a-Al2O3 tubular supports(2 mm OD, 1.6 mm ID and 30 mm length) from the syntheticmixtures without the addition of expensive Sr2+ cations. As aconsequence of this study, it is demonstrated that the addition ofSr2+ cations in a gel precursor is not indispensible for the prepara-tion of CHA zeolite membranes with superior dehydrativeperformances.

High-silica chabazite membranes designated as SSZ-13 mem-branes were tentatively synthesized on the inner surface of porousstainless steel tubes in the presence of N,N,N-trimethyl-1-adaman-tammonium hydroxide as an SDA for the single-/mixed-gas sepa-rations and dewatering of nitric acid aqueous mixtures by PV[17]. However, the reported separation performances need to beimproved further.

Two zeotypes including AlPO4-34 and SAPO-34, a microporousaluminophosphate and silicoaluminophosphate, respectively, areisostructural with CHA-type zeolites. The latter was made in theform of a membrane with a high selectivity towards CO2 compo-nent over light molecules such as H2, N2 and CH4 [18,19].

Herein, we reported a new recipe for the fabrication of alumina-rich CHA membranes in the synthetic gels containing single K+ ionsbut without Na+ or Sr2+ ions. The effect of a variety of syntheticparameters on the separation performances of the resulting CHAmembranes was investigated in detail. The primary objective ofthis work was to prepare economically and reproducibly CHAmembranes in the absence of any SDA with enhanced dehydrationperformances in terms of membrane permeation flux and selectiv-ity while separating several typical light alcohol aqueous solutions.

2. Experimental

2.1. Synthesis of CHA zeolites

Initially, powdery CHA zeolites were prepared as crystalseeds according to the reported solid-phase conversion route,as described in the patent [20]. This synthetic approach involvedtransforming directly Y zeolites into CHA zeolites in alkalinemedia. Herein, the protonated form of Y zeolites was selectedas feedstock. They were prepared by heating NH4

+-exchangedY zeolites (Sigma–Aldrich) to 550 �C, followed by calcination at550 �C for 2 h. Synthetic CHA zeolites were grown from a reac-tion mixture of batch composition: SiO2:0.19Al2O3:0.39K2O:0.03-Na2O:40H2O. In a typical synthesis, the reaction suspension wasprepared by dissolving a given amount of KOH pellets in water,followed by addition of the heat-treated HY zeolites to theabove alkaline solution. The resulting mixture was stirred atroom temperature for 10–20 min, and then charged into a poly-tetrafluoroethylene (PTFE) container. The hydrothermal synthesiswas conducted at 95 �C for 4 days. After hydrothermal treat-ment, the sample was quenched with cold tap water, recoveredby suction filtration and washed with copious amounts ofdistilled water. It was finally air dried at 100 �C overnight forfuture use.

2.2. Synthesis of CHA zeolite membranes

CHA membranes were hydrothermally grown onto 10-cm-long tubular mullite supports (12 mm OD, 1.5 mm wallthickness and 1.3 lm average pore size, Nikkaido Corporationin Japan). Prior to preparation, the outer surface of mullite tubewas simply coated with water slurry of CHA crystal seeds as pre-pared above (ca. 5.5 lm crystal size and 95.7% crystallinity usingthe most crystalline CHA zeolites as a reference). The seededsupport was then air dried at 80 �C for several hours.

The desired gel composition was SiO2:0.19Al2O3:0.39K2O:40-H2O on a molar basis. To examine the effect of silica source,three colloidal silicas (30 wt.% SiO2, 0.34 wt.% Na2O, Ludox HS-30, Sigma–Aldrich; 30 wt.% SiO2, Ludox LS-30, Sigma–Aldrich;30 wt.% SiO2, 0.43 wt.% Na2O, Snowtex S, Nissan Chemical Ind.,Ltd., in Japan) and precipitated hydrated silica (88 wt.% SiO2,Nipsil VN3, Nihon Silica Industry in Japan) were employed. Here-after, they were, respectively abbreviated as HS-30, LS-30, NS-30and PS. In a typical synthesis, water, potassium hydroxide pellets(Wako Pure Chemicals Industry in Japan), aluminum hydroxidefine powders (Wako Pure Chemicals Industry in Japan) and oneof silica sources were sequentially combined with intense stir-ring at room temperature. The stirring was maintained over-night. The resultant mixture was then charged into anautoclave, into which the seeded support was subsequently in-serted. After the autoclave was tightly sealed, the hydrothermalsynthesis was carried out at 100–150 �C for a given period oftime. At the completion of hydrothermal treatment, the mem-brane was recovered and washed with plenty of distilled water.It was ultimately air dried at 80–100 �C overnight for furthermeasurements.

As a complementary strategy to the conventional sol–gel oneas described above, CHA membranes were also prepared byusing an inter-zeolitic transformation method, resembling thepreparative protocol for CHA crystal seeds, as depicted in Section2.1. Briefly, after homogenizing and then autoclaving the reactivesuspension containing the calculated amounts of KOH solutionand HY zeolites, the seeded mullite tube was transferred intothe autoclave, followed by tight sealing. The hydrothermal syn-thesis was conducted at ca. 100 �C for roughly 5 d with or with-out a horizontal rotation of the autoclave. The remainingprocedure was the same as the traditional sol–gel syntheticroute, as described above.

2.3. Methods of characterization

XRD (X-ray diffraction) patterns of the membranes and pow-ders were recorded on Rigaku RINT 2200 X-ray diffractometerusing Cu-Ka radiation in the 2h range of 5–45� with a scanning rateof 0.05�/s.

Membrane microstructural morphology, average crystal sizeand apparent membrane thickness were determined by FE-SEM(field emission-scanning electron microscopy). Platinum was sput-ter-coated onto the samples under high vacuum before observa-tion on an FE-SEM microscope (JEOL JSM-6335F) with anacceleration voltage of 15 kV. Energy dispersive X-ray (EDX) spec-troscopy was used to determine the bulk Si/Al ratio and elementalcomposition of CHA zeolite membranes. To avoid any degree ofpeak overlapping between Al and Si elements, prior to elementalanalysis, the membrane samples were sputter-coated with a thinlayer of carbon film by rapidly burning off a piece of thread usinga high vacuum sputter (CED 020, Balzers Union). The EDX spectrawere recorded on an EDX spectroscope (X-max, Oxfordinstruments).

Fig. 1. XRD patterns of CHA membranes prepared at the synthetic temperatures of100 �C (4 d), 110 �C (2.5 d), 120 �C (2 d) and 150 �C (1 d). (SiO2:0.19Al2O3:0.39-K2O:40H2O with HS-30 as silica source;*: mullite crystalline phase).

272 X. Li et al. / Microporous and Mesoporous Materials 143 (2011) 270–276

2.4. Procedure for PV runs

PV experiments for dewatering alcohol aqueous mixtures werecarried out at 60 or 75 �C by use of a PV experimental setup de-scribed in Refs. [21,22]. The effective permeation membrane areain direct contact with the liquid feed was approximately 19 cm2.During on-stream PV operation, the downstream side of the mem-brane was always kept under high vacuum ranging from 1 to 13 Pa.The permeate vapor was collected by a cold trap condensed withliquid nitrogen. The compositions of the feed and permeate wereanalyzed by gas chromatograph (Shimadzu GC-8A). The total massflux (J in kg/m2 h) was defined as the mass of permeate producedper unit membrane area per unit time. The separation factor (a)was defined as aA/B = (YA/YB)/(XA/XB), in which XA, XB, YA and YB rep-resented the mass fractions of components A and B at the feed andpermeate sides, respectively.

3. Results and discussion

3.1. CHA membranes prepared by the inter-zeolitic transformationmethod

Initially, we briefly explored the impact of non-conventionalpreparative methodology on the separation properties of CHAmembranes to find out the suitable membrane growth conditions.The permeation performances of the CHA membranes prepared bythe inter-zeolitic transformation technique instead of the routinesol–gel synthetic approach were given in Table 1. After roughly5 d of hydrothermal treatment at ca. 100 �C, unfortunately, neitherof the HY-derived CHA membranes was completely free of defects,irrespectively of static or dynamical crystallization fashion. Thismight be a result of the inhomogeneous reactive suspension in-volved in these syntheses, which would give rise to a large concen-tration gradient along the axial direction of the tubular membranesover the course of hydrothermal reaction.

3.2. CHA membranes prepared by the conventional sol–gel method

3.2.1. Effect of the crystallization temperature3.2.1.1. XRD and FE-SEM characterizations. Fig. 1 showed the typicalXRD patterns of CHA membranes prepared at different crystalliza-tion temperatures for varying periods of time. The peaks markedwith asterisks referred to the reflection peaks from mullite supportphase. As revealed in Fig. 1, in addition to the reflection peaks cor-responding to mullite crystalline phase, these as-synthesized fourmembranes all show the presence of pure-phase CHA zeolitesand the diffraction patterns are in agreement with the reporteddata [23]. Well-defined XRD profiles assigned to the CHA topologysuggest good crystallinities of the constituent crystals in the mem-brane layers. Prolonged durations of hydrothermal reaction timefor low-temperature synthetic cases allow one to achieve similarcrystallinity to the cases prepared at higher crystallization temper-atures. It is worth noting that only 1 d of hydrothermal reaction at150 �C is sufficient for preparing highly crystalline CHA mem-branes. On the other hand, by comparing the relative intensity of

Table 1Leak detection results for CHA membranes prepared by inter-zeolite transformationapproacha.

Sample Feedstock tsynth. (d) Rotation rate (rpm) Tsynth. (�C) a

No. 1 HY crystal 5.5 0 95 LeakNo. 2 HY crystal 5 37.5 100 Leak

a SiO2:0.19Al2O3:0.39K2O:0.03Na2O:40H2O.

CHA peaks, it is concluded that CHA crystals in all membranesare aligned in a random orientation manner.

In this study, the influence of synthesis temperature upon thethickness and morphology of the resulting membranes was quali-tatively investigated by means of FE-SEM microscopy with an aimof minimizing the crystallization time necessary for the synthesisof defect-free CHA membranes. Examples of FE-SEM images ofCHA membranes grown at various crystallization temperatureson the seeded supports were presented in Fig. 2. It is well knownthat crystal growth rates vary exponentially with crystallizationtemperature [24] since the crystallization of zeolites is an activa-tion-controlled process. However, the growth of zeolite crystal isgenerally anisotropic; namely, the growth rates differ along eachindividual crystallographic axis. Therefore, the crystallization tem-perature will affect strongly the crystal shape and consequentlythe eventual microstructure of the membrane. This growth phe-nomenon has been unequivocally corroborated in Fig. 2. The find-ings indicate that the morphology of CHA membranes is closelyassociated with the synthesis temperature since crystal growthrates alter exponentially with crystallization temperature. By com-parison, these photographs clearly reveal that there exists a transi-tional synthesis temperature centered around 110 �C whichfeatures as an abrupt transformation in individual particle geome-try. Membrane growth at 100 �C that is located well before thetransition temperature indeed shows a distinct microstructure,whereas membrane allowed to prepare up to the transition tem-perature exhibits a mixed nature, where prismatic edges of colum-nar crystals coexist with chain-shaped microcrystals. Finally,membrane growths that are positioned well beyond the transitiontemperature exclusively display large facetted crystals. This mightbe a result of the strong dependence of crystal growth rate uponthe synthetic temperature. Presently, it is still unclear how thetransition in crystal shape from the leaflets to faceted crystals, asdemonstrated in Fig. 2, takes place by increasing the synthetic tem-perature. The exact reason for that issue will be elucidated in detailelsewhere in due course. A continuous CHA layer involving lots ofwell-intergrown crystallites with a characteristic geometry isformed, independently of the respective synthesis temperature.This is a notable benefit of pre-seeding. From the cross-sectionalSEM images revealed in Fig. 2, except for the membrane preparedat 100 �C possessing a thickness lower than 10 lm, there exists adense skin layer on any other support with an apparent thicknessof ca. 10 lm.

3.2.1.2. PV separation performances. Table 2 showed the PV data for10 wt.% H2O/EtOH and H2O/IPA mixtures at 75 �C through CHAmembranes prepared at different synthesis temperatures. Sincecrystal growth rates are correlated exponentially with crystalliza-

Fig. 2. FE-SEM surface and cross-sectional images of CHA membranes prepared at the synthetic temperatures of 100 �C (A and B), 110 �C (C and D), 120 �C (E and F) and150 �C (G and H): top views (A, C, E and G) and cross-sectional views (B, D, F and H). Same synthetic conditions as in Fig. 1.

X. Li et al. / Microporous and Mesoporous Materials 143 (2011) 270–276 273

tion temperature, the crystallization time is correspondingly short-ened at a higher synthetic temperature. Apparently, higher-tem-perature growth, e.g., at 130, 140 and 150 �C, allows one toreproducibly prepare higher-quality CHA membranes, as comparedto the lower-temperature synthetic cases. Although there was a

probability of acquiring perfect membranes either at 100 or120 �C, the synthetic reproducibility from batch to batch was farfrom satisfaction until now. Interestingly, at the transition temper-ature of 110 �C, as shown in Fig. 2, independently of syntheticduration ranging from 2.5 to 4 d, none of the membranes obtained

Table 2PV performances for 10 wt.% H2O/EtOH and H2O/IPA mixtures at 75 �C through CHAmembranes prepared at different synthesis temperaturesa.

Sample Tsynth. (�C) tsynth. (d) Feed J (kg/m2 h) a

No. 3 100 4 H2O/EtOH 2.13 210No. 4 100 4 H2O/EtOH 2.37 1700No. 5 110 2.5 H2O/EtOH 2.67 59No. 6 110 2.5 H2O/EtOH 2.60 120No. 7 110 3 H2O/EtOH 2.42 63No. 8 110 3 H2O/EtOH 2.70 110No. 9 110 4 H2O/EtOH 2.39 140No. 10 120 2 H2O/EtOH 2.90 33No. 11 120 2 H2O/EtOH 2.27 110

H2O/IPA 2.69 140No. 12 120 2 H2O/EtOH 2.37 3700No. 13 130 2 H2O/EtOH 2.64 990No. 14 130 2 H2O/EtOH 2.54 1600No. 15 140 1 H2O/EtOH 2.36 1100No. 16 140 1 H2O/EtOH 2.54 1700No. 17 150 1 H2O/EtOH 1.93 1000No. 18 150 1 H2O/EtOH 1.87 1600No. 19 150 1 H2O/EtOH 2.20 3900

a SiO2:0.19Al2O3:0.39K2O:40H2O with HS-30 as silica source.

274 X. Li et al. / Microporous and Mesoporous Materials 143 (2011) 270–276

had an appreciable separation performance, presumably as a resultof a less compact top layer with a mixed microstructure. As illus-trated in Fig. 2, in contrast to prism-like crystal morphology inthe higher-temperature syntheses, the observed anisotropic shapein the lower-temperature synthesis is leaflet-like crystals (Fig. 2 A),in which case the membrane permeation flux is definitely crystalgrowth orientation dependent. The latter CHA membrane had avery thin layer with a thickness of roughly 5 lm, which makes itdifficult for the crystals to grow very densely on the superficiallyrough mullite tubes. The fact that the reproducibility for the low-temperature synthesis is relatively inferior can be well understood.Therefore, morphology control study (i.e., microstructural controlstudy) should not be ignored in the synthesis of zeolite mem-branes, especially the membranes built upon the anisotropic crys-tals. On the other hand, it is demonstrated from Table 2 thatanother pronounced merit from the higher-temperature syntheseslies in that the required duration of the crystallization time isgreatly shortened from 4 d at 100 �C to 1 d at 150 �C. Overall, thepreferred synthetic temperature window lies in the 130–150 �Crange.

3.2.2. Effect of silica sourceMintova et al. [25] reported that the size of the silicalite-1 nano-

crystals was strongly dependent on the silica source applied.Hence, the nature of the selected silica source is an essential factorinfluencing the ultimate membrane separation performances. Inthis work, the impact of four types of commercial available silicasources was in detail examined including NS-30, LS-30, HS-30and PS. The former three colloidal silicas all contain Na+ stabilizingcounterion. HS-30 (pH 9.8) differs slightly from both LS-30 (pH 8.2)and NS-30 (pH 10) by sol pH values. The particulate specific surfacearea of LS-30 (�215 m2/g) is similar to that of HS-30 (�220 m2/g).Based upon the sol pH level, it is deduced that the residual Na+ con-tent in the sols decreases in the following order: NS-30(0.43 wt.%) > HS-30 (0.34 wt.%) > LS-30. PS is synthetic amorphoussilica product, and distinguishes itself from colloidal silica on thebasis of agglomerate particle size. PS typically has an agglomeratesize of 1–40 lm, whereas colloidal silica generally has a particlesize altering from approximately 0.03–0.1 lm in diameter.

Table 3 presented the permeation properties of CHA mem-branes synthesized with different silica sources for the separationof the feed mixtures of 10 wt.% H2O/alcohol at 75 �C. While HS-30was used as silica source, Membrane No. 4 gave both high selectiv-

ity and high permeation flux of 1700 and 2.37 kg/m2 h, respec-tively, for 10 wt.% H2O/EtOH solution at 75 �C. Apart frommeasuring the separation ability for 10 wt.% H2O/EtOH feed mix-ture, this membrane was as well utilized for the dehydration of a90 wt.% iso-propanol (IPA) aqueous solution by PV at 75 �C, yield-ing a flux of 2.93 kg/m2�h and a water separation factor of 1800.It is shown that the separation capacity of CHA membranes is high-er in H2O/IPA separation compared with that in H2O/EtOH separa-tion, primarily arising from the bigger kinetic diameter of IPAmolecule (5.2 Å) than EtOH one (4.5 Å). As shown in Table 3, effi-cient separation of MeOH aqueous mixtures has so far posed agreat challenge by way of CHA membranes because of the smalldifference in molecular size between MeOH (3.6 Å) and H2O(2.6 Å), both of which are less than the pore dimension of hydratedCHA zeolites (3.7 � 4.2 Å2). On the other hand, the small separationfactor is quite likely responsible for the existence of a certain con-centration of non-selective interzeolitic pores with a pore diametergreater than the intrinsic CHA zeolitic pores. By comparison, it isfound from Table 3 that the separation performances of theseCHA membranes are substantially affected by the silica sourceused. For example, Membrane No. 23 derived from NS-30 hadthe separation performances of 2.67 kg/m2 h in total flux and of130 in selectivity. In contrast, when LS-30 was chosen as one ofthe starting materials, through the resulting Membrane No. 21the corresponding values were 2.24 kg/m2 h and 42, respectively.PS was also not an attractive candidate for high-quality CHA mem-brane acquisition under the present synthetic conditions. An impli-cation of such minor modification is that the physicochemicalnature of silica source used plays an important role in dictatingthe two characteristics of CHA membranes, selectivity and produc-tivity. Highly condensed and polymerized silica source such as PShas an adverse impact on CHA membrane separation perfor-mances. Additionally, a trace amount of Na+ cations contained inthe silica source may contribute to some degree to the membranequality, as evidenced by the poorest selectivity collected on the PS-based CHA Membrane No. 26. It was reported that CHA zeolitescould be successfully prepared in a tertiary Na+/K+/TMA+ (tetra-methylammonium cation) mixed ionic system by a traditionalsol–gel technique [26]. On the other hand, by comparing Mem-brane No. 4 with Membrane No. 22, with increasing duration ofthe synthetic period at 100 �C, the permeation flux decreased from2.37 to 0.95 kg/m2 h as a consequence of the expected increasedmembrane thickness. In conclusion, altering silica source resultsin notable difference in separation performance in the order: HS-30 > NS-30 � PS � LS-30. Therefore, HS-30 is selected as an opti-mum silica source for the synthesis of CHA membranes in the fol-lowing work.

3.2.3. Effect of the dilution degree in the starting gelThe water concentration in the reaction mixture is also an

essential factor influencing the crystallization kinetics of zeolites.In the course of hydrothermal reaction, mass transport propertieswithin the mixture and the viscosity of the mixture are sensitiveto the variation in water content. Dilution of a reaction mixturecorresponds to a decrease in concentration of the various reactivecomponents in the liquid phase. This in turn slows down the crys-tallization rate of zeolite membranes. For example, for fluoride-mediated zeolite synthesis, water level in a starting gel plays amore important role in relation to hydroxide-based zeolite synthe-sis. Typically, the water content in the fluoride-containing media isdecreased down to an amount close to that of the structure-build-ing reactants [27,28]. In these cases, water concentration is so crit-ical that it can govern the phase selectivity of the product obtained.

The impact of water content on the growth of CHA membraneswas investigated using a gel with a general molar composition ofSiO2:0.19Al2O3:0.39K2O:xH2O, where x changed from 40 to 60.

Table 3PV performances for 10 wt.% H2O/alcohol mixtures at 75 �C through CHA membranes prepared with differing silica sourcesa.

Sample Si source Tsynth. (�C) tsynth. (d) Feed J (kg/m2 h) a

No. 20 LS-30 100 5 H2O/EtOH 2.79 27No. 21 LS-30 100 5 H2O/EtOH 2.24 42No. 22 HS-30 100 5 H2O/EtOH 0.95 320No. 4 HS-30 100 4 H2O/EtOH 2.37 1700

H2O/IPA 2.93 1800H2O/MeOHb 0.92 32

No. 3 HS-30 100 4 H2O/EtOH 2.13 210No. 23 NS-30 110 3 H2O/EtOH 2.67 130No. 24 NS-30 110 3 H2O/EtOH 2.75 31No. 25 PS 140 1 H2O/EtOH 3.20 120No. 26 PS 140 1 H2O/EtOH 4.79 17

a SiO2:0.19Al2O3:0.39K2O:40H2O.b PV test at 60 �C.

Table 4PV performances for 10 wt.% H2O/EtOH mixtures at 75 �C through CHA membranesprepared with different water contents in the initial gel precursorsa.

Sample H2O/SiO2 ratio tsynth. (d) J (kg/m2 h) a

No. 17 40 1 1.93 1000No. 18 40 1 1.87 1600No. 19 40 1 2.20 3900No. 27 50 1 2.45 320No. 28 50 1 2.33 330No. 29 55 1 2.44 130No. 30 60 1 1.94 420

a SiO2:0.19Al2O3:0.39K2O:xH2O at 150 �C with HS-30 as silica source.

X. Li et al. / Microporous and Mesoporous Materials 143 (2011) 270–276 275

Table 4 showed the PV results for 10 wt.% H2O/EtOH mixtures at75 �C through CHA membranes prepared with different water con-tents in the initial gel precursors. From Table 4, it is clear that thecurrently desired H2O/SiO2 molar ratio is located around 40 fromthe viewpoint of the combination of high productivity and highselectivity. In this dilution degree window, the constituent crystalsin the membrane layer are supposed to possess a higher crystallin-ity and more perfect intergrowth behavior. Beyond the optimumdilution degree, the separation performances are less susceptibleto the change in water concentration. It is worth noting that thecomparatively lower flux of Membrane No. 30 prepared at anH2O/SiO2 molar ratio of 60 for 1 d is probably attributed to thepoorly crystalline membrane barrier layer involved relative tothe comparably selective Membrane Nos. 27, 28 and 29 all pre-pared in a slightly more concentrated gel system.

3.2.4. Effect of the alkalinity of the starting gelFig. 3 showed the EDX spectrum of Membrane No. 18. It was

found that this typical membrane had bulk SiO2/Al2O3, K/Al andK2O/SiO2 molar ratios of 5.72, 0.89 and 0.16, respectively. The mea-

Fig. 3. EDX spectrum of CH

sured bulk SiO2/Al2O3 ratio was similar to that reported in Ref. [15],which is obviously greater than that of LTA-type zeolite mem-branes. As a result, the hydrothermal and acid stabilities of CHAmembranes are superior to that of LTA membranes. That is, undercertain harsher operating conditions, CHA membranes are ex-pected to outperform LTA membranes in combination with a per-meation flux comparable to the latter [29]. K/Al molar ratiosmaller than unity is indicative that during synthetic period a smallamount of Na+ cations arising from the used raw materials (e.g.,colloidal silica) serves as extra-framework charge-counterbalanc-ing ions.

Table 5 showed the PV performances of CHA membranes pre-pared on the seeded mullite tubes as a function of gel KOH content.As a matter of fact, the variation in KOH concentration at least hasdual effects on the membrane formation. As a sole mineralizerdosed, KOH itself contributes to the alkalinity of the synthetic gelbesides the structure-directing and charge-counterbalancing rolesof K+ cations. In the current synthesis, the impact of K+ concentra-tion may be minor on the membrane separation performancessince the chosen gel K2O/SiO2 ratio in the 0.29–0.39 region isexcessive in regard to the constituent zeolites in the membranelayer with a corresponding value of 0.16 (vide supra). Namely,the alkalinity of the synthetic gel is quite likely to contribute to agreater extent than the structure-directing action of K+ cations.For H2O/EtOH system, it was demonstrated that Membrane No.19 prepared at a K2O/SiO2 ratio of 0.39 integrated a high selectivitywith a high flux. Meanwhile, CHA membranes obtained under thesame synthetic conditions as Membrane No. 19 yielded reasonablyreproducible separation performances for H2O/EtOH feed mixtures.As the alkalinity in terms of K2O/SiO2 ratio dropped from 0.39 to0.29, a comparatively poor intergrowth behavior was expectedfor Membrane Nos. 31 and 32 both grown from a synthesis gel thatcontained a lower amount of KOH (i.e., K2O/SiO2 = 0.29), as evi-denced by the deteriorated separation factor and raised flux. By

A Membrane No. 18.

Table 5PV performances for 10 wt.% H2O/EtOH mixtures at 75 �C through CHA membranesprepared at different alkalinities of the starting gela.

Sample K2O/SiO2 ratio tsynth. (d) J (kg/m2 h) a

No. 31 0.29 1 2.88 370No. 32 0.29 1 – LeakNo. 17 0.39 1 1.93 1000No. 18 0.39 1 1.87 1600No. 19 0.39 1 2.20 3900

a SiO2:0.19Al2O3:xK2O:40H2O at 150 �C with HS-30 as silica source.

Table 6Comparison of dehydration performances of CHA membranes prepared with andwithout Sr2+ ions for 10 wt.% H2O/EtOH mixtures at 75 �C.

Sample Cationtype

Tsynth.(�C)

tsynth.(h)

J (kg/m2 h)

a Ref.

CHA-1 K+/Sr2+ 140 18 4.14 >10,000 [11]No. 16 K+ 140 24 2.54 1700 This workCHA-2 K+/Sr2+ 150 5 14.00 >10,000 [16]No. 19 K+ 150 24 2.20 3900 This work

276 X. Li et al. / Microporous and Mesoporous Materials 143 (2011) 270–276

comparing Membrane Nos. 31 and 32 with Nos. 17, 18 and 19, it isconclusive that the preparations of defect-free CHA membraneswith abundant Al incorporation are favored by use of a CHA gelprecursor of slightly higher alkalinity.

3.3. Comparison of dehydration performances of CHA membranesprepared with and without Sr2+ ions

Dehydration performances of CHA membranes prepared withand without Sr2+ ions for the separation of 10 wt.% H2O/EtOH mix-tures at 75 �C were tabulated in Table 6. At the same synthetictemperature, the membranes prepared with mixed K+/Sr2+ systemsoutperformed significantly those prepared with K+-only gel precur-sors, as reflected with shorter synthetic times as well as larger totalpermeation fluxes and separation factors for the former mem-branes. The addition of Sr2+ ions is supposed to have a beneficialcontribution to the dewatering performances of CHA membranes.Nevertheless, the PV testing means by which the combined evacu-ation and He as a purging gas was applied on the permeation sidein Refs. [11,16], as compared to the individual evacuation wayadopted in the present work, may be another influencing factor,causing the distinct dehydration performances. Although the per-formances presented herein are not favorably comparable to thoseshown in Refs. [11,16], this study undoubtedly presents a new syn-thesis route for the hitherto poorly-documented CHA zeolitemembranes.

4. Conclusions

The variables affecting the quality of the resulting CHA-typezeolite membranes were optimized in detail in this work, includingsilica source, the crystallization temperature, water concentrationin the initial gels and gel alkalinity. In addition, an attempt wascarried out to fabricate CHA-type zeolite membranes by the in-ter-zeolitic transformation method, but finally failed to achieve de-

fect-free CHA membranes. Depending upon the silica sourceutilized, the PV performances of the resultant CHA membranes de-creased in the following order: HS-30 > NS-30 � PS � LS-30. Thesynthetic temperature involved affected severely the membranemorphology and separation performances. Higher crystallizationtemperature led to shorter synthetic duration, better syntheticreproducibility, higher selectivity and nearly unchanged total per-meation flux. After determining experimentally the optimal waterand potassium hydroxide concentrations in the initial precursormixtures, an optimal gel composition was thus formulated in thiswork. In comparison with other works reported until now, it wasshown that without recourse to any organic SDA or Sr2+ cations afeasibility study was performed herein that high-quality CHAmembranes could be synthesized from an individual K+-containingsynthetic medium.

Acknowledgement

The present work was partly supported by a Grant-in-Aid forDevelopment Scientific Research from the Ministry of Education,Culture, Sports, Science and Technology of Japan.

References

[1] M. Calligaris, G. Nardin, L. Randaccio, P.C. Chiaramonti, Acta Crystallogr. B 38(1982) 602.

[2] T. Inui, Y. Okugawa, M. Yasuda, Ind. Eng. Chem. Res. 27 (1988) 1103–1109.[3] J. Zhang, R. Singh, P.A. Webley, Microporous Mesoporous Mater. 111 (2008)

478–487.[4] D.T. Bostick, W.D. Arnold, B. Guo, M.W. Burgess, Sep. Sci. Technol. 32 (1997)

793.[5] J.J. Perona, AIChE J. 39 (1993) 1716.[6] D. Shah, K. Kissick, A. Ghorpade, R. Hannah, D. Bhattacharyya, J. Membr. Sci.

179 (2000) 185–205.[7] D.W. Breck, Zeolite Molecular Sieves: Structure Chemistry and Use, Wiley &

Sons, New York, 1974. pp. 83–90.[8] C.G. Coe, T.R. Gaffney, US Patent 4943304, (1990) 8.[9] M.W. Ackley, H. Saxena, G.W. Henzler, J.J. Nowobilski, WO03053546, 2003, p.

49.[10] S.U. Rege, R.T. Yang, K. Qian, M.A. Buzanowsi, Chem. Eng. Sci. 56 (2001) 2745.[11] Y. Hasegawa, H. Hotta, K. Sato, T. Nagase, F. Mizukami, J. Membr. Sci. 347

(2010) 193–196.[12] V. Baglio, A.S. Arico, A.D. Blasi, P.L. Antonucci, F. Nannetti, V. Tricoli, V.

Antonucci, J. Appl. Electrochem. 35 (2005) 207–212.[13] K. Tanaka, R. Yoshikawa, Y. Cui, H. Kita, K. Okamoto, Catal. Today 67 (1–3)

(2001) 121–125.[14] X. Li, H. Kita, H. Zhu, Z. Zhang, K. Tanaka, J. Membr. Sci. 339 (2009) 224–232.[15] Y. Hasegawa, C. Abe, M. Nishioka, K. Sato, T. Nagase, T. Hanaoka, J. Membr. Sci.

363 (2010) 256–264.[16] Y. Hasegawa, C. Abe, M. Nishioka, K. Sato, T. Nagase, T. Hanaoka, J. Membr. Sci.

364 (2010) 318–324.[17] H. Kalipcilar, T.C. Bowen, J.L. Falconer, R.D. Noble, Chem. Mater. 14 (2002)

3458–3464.[18] J.C. Poshusta, R.D. Noble, J.L. Falconer, J. Membr. Sci. 186 (1) (2001) 25–40.[19] S. Li, J.L. Falconer, R.D. Noble, Microporous Mesoporous Mater. 110 (2–3)

(2008) 310–317.[20] M. Bourgogne, J.-L. Guth, R. Wey, US Patent 4503024 (1985).[21] X. Lin, H. Kita, K. Okamoto, Ind. Eng. Chem. Res. 40 (19) (2001) 4069–4078.[22] Y. Cui, H. Kita, K. Okamoto, J. Membr. Sci. 236 (1–2) (2004) 17–27.[23] T.R. Gaffney, C.G. Coe, US Patent, 5026532 (1991).[24] W.C. Wong, L.T.Y. Au, C.T. Ariso, K.L. Yeung, J. Membr. Sci. 191 (2001) 143–163.[25] S. Mintova, V. Valtchev, Microporous Mesoporous Mater. 55 (2002) 171–179.[26] R.K. Singh, P. Webley, Adsorption 11 (2005) 173–177.[27] S.I. Zones, A.W. Burton, G.S. Lee, M.M. Olmstead, J. Am. Chem. Soc. 129 (29)

(2007) 9066–9079.[28] A.W. Burton, S.I. Zones, S. Elomari, Curr. Opin. Colloid Interface Sci. 10 (5–6)

(2005) 211–219.[29] Y. Morigami, M. Kondo, J. Abe, H. Kita, K. Okamoto, Sep. Purif. Technol. 25

(2001) 251–260.