Microporous and Mesoporous Materials 158 (2012) 292–299

Contents lists available at SciVerse ScienceDirect

Microporous and Mesoporous Materials

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

Sonication-induced Ostwald ripening of ZIF-8 nanoparticles and formationof ZIF-8/polymer composite membranes

Joshua A. Thompson a, Karena W. Chapman b, William J. Koros a, Christopher W. Jones a,⇑, Sankar Nair a,⇑a School of Chemical & Biomolecular Engineering, Georgia Institute of Technology, 311 Ferst Dr. NW, Atlanta, GA 30332-0100, USAb X-Ray Science Division, Advanced Photon Source, Argonne National Laboratory, 9700 South Cass Ave., Bldg. 433, Argonne, IL 60439-4858, USA

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

Article history:Received 15 February 2012Received in revised form 29 March 2012Accepted 30 March 2012Available online 9 April 2012

Keywords:Zeolitic imidazolate frameworkMixed-matrix membranesOstwald ripeningSonicationMembrane formation

1387-1811/$ - see front matter � 2012 Elsevier Inc. Ahttp://dx.doi.org/10.1016/j.micromeso.2012.03.052

⇑ Corresponding authors.E-mail addresses: cjones@chbe.gatech.edu (C.W. Jo

ch.edu (S. Nair).

The effect of typical membrane processing conditions on the structure, interfacial morphology, and gasseparation performance of MOF/polymer nanocomposite membranes is investigated. In particular, theZIF-8/Matrimid� nanocomposite membrane system is examined, and it is shown that ultrasonication –a commonly employed particle dispersion method – induces significant changes in the shape, size distri-bution, and structure of ZIF-8 particles suspended in an organic solvent during membrane processing.Dynamic light scattering and electron microscopy reveal that ZIF-8 nanoparticles undergo substantialOstwald ripening when subjected to high intensity ultrasonication as often required in the formationof MOF/polymer nanocomposite membranes. Other characterization techniques reveal that the ripenedparticles exhibit lower pore volumes and lower surface areas compared to the as-made material. ZIF-8/Matrimid� composite membranes fabricated using two sonication methods show significant differ-ences in microstructure. Permeation measurements show significant enhancement in permeability ofCO2 and increased CO2/CH4 selectivity in membranes fabricated with high-intensity sonication. In con-trast, composite membranes prepared with low-intensity sonication are found to be defective. A carefulevaluation of MOF membrane processing conditions, as well as knowledge of the properties of the MOFmaterial after these membrane processing steps, are necessary to develop reliable processing–structure–property relations for MOF-containing membranes.

� 2012 Elsevier Inc. All rights reserved.

1. Introduction

Zeolitic imidazolate frameworks [1] (ZIFs) are a promising classof nanoporous materials for molecular sieving applications (e.g.,separations or catalysis [2]), due to high internal surface area andtunable crystalline structure and porosity. In particular, the mate-rial ZIF-8 has recently attracted considerable attention for theseapplications due to its facile synthesis coupled with its good chem-ical and thermal stability in comparison to other classes of metal–organic frameworks (MOFs) [3]. For example, Yaghi et al. showedthat refluxing ZIF-8 particles in benzene, methanol, water, andeven concentrated alkaline solutions had no effect on the long-range crystal structure, as determined by X-ray diffraction (XRD)[4]. In contrast, other well-known MOFs such as MOF-5 may havevery poor stability, even when exposed to water at low activity [5].

There have been a number of reports on the adsorption and sep-aration properties of ZIF-8 [6–8]. Caro et al. explored the separa-tion performance of ZIF-8 membranes for gas pairs such as CO2/CH4, C2H4/C2H6, and H2/C3H8 [9,10]. Although ZIF-8 has a crystallo-

ll rights reserved.

nes), sankar.nair@chbe.gate-

graphically-determined pore aperture of 0.34 nm and should hencebe a good molecular sieve for the above gas pairs, it exhibited apoor separation performance with the exception of gases that areconsiderably different in size (H2/C3H8) [11]. While the low selec-tivity can be explained by the potential presence of defects in themembrane, it has been shown that ZIF-8 adsorbs gas molecules lar-ger than its nominal pore size [12]. This has led to several ques-tions regarding the molecular transport and adsorptionmechanisms in ZIF-8 and MOFs in general. It has been shown thatZIF-8 is a flexible or ‘‘gate-opening’’ framework whose pores swingopen by rotation of imidazolate linkers and expand when probedwith N2 [13]. This phenomenon could have wide implications forevaluating the molecular sieving capabilities in any MOF or relatedmaterial that exhibits flexibility.

An alternative to fabricating pure ZIF-8 membranes for gas sep-arations is to combine crystals of ZIF-8 with a polymer to form acomposite or ‘‘mixed-matrix’’ membrane (MMM). It has provendifficult to form defect-free composite films with inorganic mate-rials (zeolites) as fillers and glassy polymers as the surroundingmatrix without the introduction of additional processing steps[14–18]. The selection of matrix and filler materials that are com-patible both for interfacial adhesion and for desired separationproperties can be challenging. Fortunately, the incorporation of

J.A. Thompson et al. / Microporous and Mesoporous Materials 158 (2012) 292–299 293

MOFs in polymer membranes has proven to be much less likely tocreate the types of defects that are commonly encountered whenusing zeolites as fillers (e.g., voids and low-density regions of poly-mer around the zeolite surface or polymer rigidification around thezeolite particles) [19–21]. Ordoñez et al. studied ZIF-8/Matrimid�

mixed-matrix membranes at different weight loadings; however,the results did not show consistent trends with regard to ZIF par-ticle loading, as would be expected for a defect-free membrane[19]. Bae et al. have reported the preparation of ZIF-90/polymercomposite membranes with good separation properties; however,the effect of particle loading was not explored due to difficulties inpreparing defect-free membranes with higher ZIF-90 loadings [20].To date, only one MOF mixed-matrix membrane publication hasshown data phenomenologically consistent with established theo-retical models (e.g., the Maxwell model) [22].

A critical step in fabrication of these composite membranes isthe dispersion of the filler particles in the polymer solution. Theuse of surfactants and salts can provide colloidal stability and pre-vent particle aggregation, but ultrasonication is widely used to dis-perse particles for composite membrane fabrication. While someMOFs such as ZIF-8 have shown chemical and thermal stabilityin various solvents, the stability of MOFs upon exposure to high-powered ultrasonication is not currently known. Indeed, it is notuncommon for sonication to generate chemical reactions due tolocalized high temperatures, and some MOFs such as MOF-177have been synthesized by a sonochemical route [23].

In the present work, we first explore the effects of sonication onZIF-8 nanoparticles during the fabrication of ZIF-8/Matrimid�

MMMs, using two distinct sonication methods. When high-inten-sity sonication is used for particle dispersion, there is a significantchange in particle size, polydispersity, and morphology of the crys-tals. We present evidence that the phenomenon is an Ostwald rip-ening effect induced by sonication and characterize in furtherdetail the structure of the ripened ZIF-8 particles. Dynamic lightscattering measurements and electron microscopy show a dis-tinctly bimodal particle size distribution that appears when sonica-tion is applied to ZIF-8 dispersions at different concentrations. Wethen show that ZIF-8/Matrimid� MMMs have different microstruc-tures and gas permeation behavior, depending on the sonicationconditions used to prepare the polymer-nanoparticle dispersions.The observed membrane morphology-permeation property rela-tionships are shown to be wholly consistent with expected MMMbehavior, as described by Maxwell models for the different mem-brane microstructures. A significant conclusion of the present workis that a systematic understanding of MOF-MMM behavior is in-deed possible through detailed structure–property correlations,and careful evaluation of the effects of commonly used processingconditions is necessary in the fabrication of such membranes. Dueto the systematic and consistent trends observed in the perfor-mance of our MOF/polymer MMMs, we are able to reliablyestimate the permeation characteristics of the ZIF-8 crystals asexisting in the MMMs and show that these separation proper-ties are indeed different from those obtained by permeation mea-surements with pure ZIF-8 membranes or predicted by adsorptionand diffusion studies with ZIF-8 crystals or by computationalpredictions.

2. Experimental methods

2.1. ZIF-8 Nanoparticle synthesis

ZIF-8 nanoparticles were synthesized in a manner similar to apreviously published procedure [2]. Two reactant solutions wereprepared: 1.50 g Zn(NO3)2�6H2O (Sigma–Aldrich, 99%) in 50 mLmethanol; and 1.67 g 2-methylimidazole (2-MeIM, Sigma–Aldrich,

99%) in 50 mL methanol. The former was poured into the latter andstirred at room temperature for 1 h. The resulting milky solutionwas centrifuged at 7000 rpm for 5 min, and the supernatant wasremoved. After washing the precipitate with methanol, this pro-cess was repeated three more times. The resulting powder wasdried in an oven at 358 K.

2.2. Ultrasonication studies

The sonication horn used for direct sonication was a FisherSciUltrasonic Model 500 Dismembrator with an average power out-put of 200 W (400 W at 50% amplitude) and 20 kHz frequency.The sonication bath used for indirect sonication was a VWR Ultra-sonication water bath, operating at 120 W and 40 kHz. To investi-gate ZIF-8 stability using direct sonication, dispersions wereprepared in 20 mL borosilicate vials with constant ZIF-8 concentra-tions of 1.0 g/L (0.025 g ZIF-8 in 5 mL tetrahydrofuran (THF)) and20 g/L (0.1 g ZIF-8 in 5 mL THF). These solutions were sonicatedin 30-s bursts and allowed to cool before sonicating again to min-imize solvent evaporation. The dispersions were then filtered andwashed with DI H2O to obtain the resulting powder, which wasdried in an oven at 358 K.

2.3. Composite membrane fabrication

ZIF-8/Matrimid� mixed-matrix membranes were prepared by asolution-casting technique. Matrimid� is a glassy polyimide with ahydrophobic backbone; its structure is shown in Figure S1 in theSupporting Information. Dried ZIF-8 particles (0.1–0.25 g) werefirst dispersed in 5–10 mL THF (Sigma–Aldrich, 99%) by either di-rect or indirect sonication. Direct sonication was done by insertinga sonication horn into the colloidal solution, and indirect sonica-tion was done by submerging the vial containing the colloidal solu-tion in a sonication water bath. The mixture was sonicated for5 min. Then, 1.0 g of 10 wt% Matrimid�/THF priming solution wasadded. Because the ZIF and polymer are assumed to have affinityfor each other, the priming step is suggested to help the ZIF disper-sion by allowing the polymer to adhere to the ZIF surface, therebyproviding steric stability and preventing aggregation of the nano-particles. This primed dispersion was sonicated for another2 min. A balance of Matrimid� powder dried in an oven at 358 Kwas added to obtain the desired composite membrane composi-tion, and the dispersion was tumbled overnight. The resultingZIF-polymer dope was placed in a glove bag, flushed with N2,and saturated with THF. The dope was poured across a glass plate,and a film was cast manually using a 200 lm doctor’s blade. Aftersolvent evaporated and the membrane vitrified, it was annealed at523 K under vacuum for 12 h.

2.4. Characterization methods

ZIF-8 particles were characterized with powder X-ray diffrac-tion (XRD), synchrotron X-ray pair distribution function (PDF)analysis, scanning electron microscopy (SEM), energy dispersiveX-ray spectroscopy (EDX), nitrogen physisorption, Fourier trans-form infrared spectroscopy (FTIR), Fourier transform Raman spec-troscopy (FT-Raman), and dynamic light scattering (DLS). PowderXRD measurements were done on an X’Pert Pro PANalytical X-ray Diffractometer. Experiments were carried out scanning from4� to 50� 2h, using an X’celerator detector. Total scattering datasuitable for PDF analysis were collected at beamline 11-ID-B atthe Advanced Photon Source at Argonne National Laboratory. Highenergy X-rays (58 keV, k = 0.2128 Å) were used, in combinationwith a large amorphous silicon-based area detector, to collect datato high values of momentum transfer, Qmax = 24 Å�1 [24,25].The two-dimensional images were reduced to one-dimensional

Fig. 1. SEM images of the ZIF-8 nanoparticles showing changes in particle size,polydispersity and morphology: (a) as-made; (b) 1 g/L dispersion sonicated for10 min; (c) 20 g/L dispersion sonicated for 10 min.

294 J.A. Thompson et al. / Microporous and Mesoporous Materials 158 (2012) 292–299

scattering data within fit2d [26]. The PDFs, G(r), were extractedwithin PDFgetX2 [27], subtracting contributions from thebackground and Compton scattering to the total scattering data.The PDFs were smoothed by r-averaging over the periodicity ofthe termination ripples (2p/Qmax) to minimize the influence ofthese artifacts on the subsequent analysis. The measured PDFscontain local structural information as a weighted histogram ofall interatomic distances within the material, regardless of crystal-linity. The intensity of features in the PDF was weighted by thescattering power of both atoms in a given correlation.

SEM imaging and EDX measurements were carried out withZeiss LEO 1530 and 1550 scanning electron microscopes. Sampleswere first coated with either gold or a gold–palladium mixtureby sputtering under vacuum. Images were taken at 10 kV acceler-ating voltage, and EDX analysis was done at 20 kV. Nitrogen phys-isorption measurements were done with a Micromeritics ASAP2020 surface area analyzer at 77 K. Samples were degassed for18 h at 308 K. The resulting isotherms were analyzed using theBET method, the Langmuir method, and the t-plot micropore vol-ume method. FTIR and FT-Raman measurements were done witha Bruker Vertex 80v FTIR/RAM II FT-Raman Analyzer. FTIR mea-surements were performed under vacuum with samples preparedin KBr pellets; FT-Raman measurements were done in open atmo-sphere with powders deposited in NMR tubes. Spectra were ana-lyzed from 400 to 4000 cm�1. DLS measurements wereperformed with a Protein Solutions DynaPro DLS. ZIF-8 (or soni-cated ZIF-8) powder was dispersed in filtered methanol with a son-ication bath. The colloidal solution was then inserted into plasticcuvettes using a 5 lm syringe filter and 3 mL syringe. Autocorrela-tion functions were analyzed by a regularization fit method solvedwith a non-negative least squares algorithm to obtain particle sizedistributions.

ZIF-8/Matrimid� films were characterized using SEM, FTIR anddifferential scanning calorimetry (DSC). Samples for SEM were pre-pared by fracturing small portions of the film under liquid N2 andcoating with gold or a gold–palladium mixture by sputtering. FTIRsamples were small portions of the film used as-is and measure-ments were done under vacuum. DSC measurements were doneon a Netzsch STA-409-PG thermogravimetric (TGA) and differentialscanning calorimeter (DSC). Samples were subjected to two heat-ing cycles from room temperature to 623 K at a 10 K/min ramprate. Glass transition temperatures of each film were determinedfrom the second heating cycle.

2.5. Permeation measurements

Permeation measurements were performed using a constantvolume permeation cell described in earlier work [28]. A small areaof the film was cut out, and using aluminum tape, a mask was pre-pared with approximately 1 cm in diameter. Film thickness wasmeasured with a micrometer in 8–10 different locations of themasked film and varied between 50 and 60 lm. At least two areasof a film and two separate films were tested for each membrane re-ported. After insertion into the permeation cell, the film was de-gassed at 308 K for at least 24 h before each permeation test.Leak tests were done before each permeation experiment, rangingfrom 10�7 to 10�6 Torr/s. Subsequent permeation tests were per-formed after degassing both sides of the film under vacuum for12–24 h and testing the leak rate. Permeation experiments wereperformed at 308 K with 3400 Torr of upstream pressure of eitherCO2 or CH4. Measurements started once upstream gas was intro-duced to the cell and the downstream was evacuated (<10�3 Torr),and permeability values were calculated after the pressure rise ratereached steady state, monitored by taking the derivative of thepressure rise. The time to reach steady state varied between mem-branes, decreasing as the permeability of the membrane increased

with loading of ZIF-8. It was typically in the range of 40 min-1.5 hfor CO2 and 8.5–20 h for CH4.

3. Results and discussion

3.1. Stability of ZIF-8 nanoparticles during ultrasonication

Fig. 1 shows the morphology of the as-synthesized ZIF-8 parti-cles and the particles obtained from 1 and 20 g/L dispersion after10 min of direct sonication. In comparison to as-made ZIF-8(Fig. 1a), there is a disparity in both particle size distribution andparticle morphology in Fig. 1b and c. In the 1 g/L dispersion, themorphology of the ZIF-8 particles was largely unchanged after son-ication, but a number of larger particles (>500 nm in diameter)were present. In the 20 g/L dispersion, there was no longer well-defined particle morphology; closer examination showed a num-ber of different structures with particle size reaching over 1 lm

J.A. Thompson et al. / Microporous and Mesoporous Materials 158 (2012) 292–299 295

(Figure S2 in Supporting Information). EDX spectra (Figures S3 andS4 in Supporting Information) revealed that these structures wereconsiderably more oxygen-rich than as-made ZIF-8, likely due tosurface defects containing hydroxyl groups that are formed uponwashing with water after sonication.

To obtain quantitative insight into the above observations, DLSmeasurements were carried out over a range of sonication timesand the particle size distributions (shown in Fig. 2) obtained fromthe regularization fit were analyzed for both 1 and 20 g/L disper-sions. Plotting the PSD over time for the 1 g/L dispersion, there isfirst a broadening of the size distribution; however, after a shortamount of sonication time, the distribution becomes bimodal. Thiscorroborates well with SEM observations that there are two differ-ent particle populations after 10 min sonication. The PSD in Fig. 2bshows the evolution of ZIF-8 nanoparticles as a function of sonica-tion time at 20 g/L dispersion concentration. Due to an increase involume fraction of nanoparticles [29], the rate of particle growthincreases considerably, accounting for the much larger change inPSD from DLS. To verify these changes in PSD are directly relatedto the ultrasonication process, ZIF-8 nanoparticles were solvother-mally treated in 10 mL THF at 373 K for 24 h using a Parr digestion

Fig. 2. Evolution of ZIF-8 particle size distribution in 1 g/L (a) and 20 g/L (b)suspensions during sonication, showing the shifts as well as broadening of the sizedistribution: (i) 0 min sonication; (ii) 0.5 min; (iii) 1 min; (iv) 2 min; (v) 5 min; (vi)10 min. The appearance of a bimodal distribution is indicative of an Ostwaldripening phenomenon.

bomb. After treatment, no particle size changes were observed byDLS, nor were there any changes in the long-range crystallographicstructure as observed by powder XRD (see Figures S5 and S6 inSupporting Information).

Combining the observations of SEM and DLS, the above resultssupport the occurrence of a sonication-induced Ostwald ripeningmechanism involving preferential dissolution of smaller ZIF-8 par-ticles and recrystallization and growth of larger ZIF-8 particles[29–31]. During ultrasonication, cavitation from sonic waves cre-ates localized areas wherein the pressure and temperature are sig-nificantly higher than that of the surrounding medium, known as‘‘hot spots’’ [32]. We hypothesize that the cavitational effects leadto the dissolution of the ZIF-8 constituents at the particle surfaces.This is followed by diffusion of the dissolved species and rapidrecrystallization on the other particles. Since the smaller particleshave a lower thermodynamic stability and a higher surface-to-volume ratio [31], the ripening process leads to the slow disap-pearance of small particles and the growth of large particles tominimize the surface free energy, thereby leading to an increasein the average particle size and a shift from a narrow to a bimodalPSD (Fig. 2).

Because the PSD becomes bimodal, it is important to deconvo-lute the two distributions to analyze the growth kinetics [33]. Ifthe ripening process is limited by diffusion of the solute from thesmall particles to the large particles, then the radius (r) of the largeparticles should grow with time (t) at a rate given by r � Kt1/3

according to LSW theory [34,35]. In Fig. 3, r is plotted as a functionof t1/3. The initial lack of particle growth could be interpreted as an‘‘induction’’ period, which may be necessary to gain a critical dis-parity in particle sizes before there are any noticeable ripening ef-fects. After this induction period, there are large enough particlepopulations above and below the critical radius for Ostwald ripen-ing to occur [36]. However, it is important to note that under son-ication conditions two assumptions of LSW theory are possibly notsatisfied, resulting in the observed non-ideal coarsening behavior.The particles are likely colliding from Brownian motion and fromcavitation in the sonicated solution; therefore, the particles arenot fixed in space and are experiencing interparticle interactions,both of which are not accounted for by LSW theory.

The size of the (dissolving) smaller particles is also shown inFig. 3 for both dispersions, revealing an overall decrease in thesmaller particle size with sonication time. This confirms the slow

Fig. 3. Ostwald ripening of ZIF-8 suspensions: r as a function of sonication time, t1/3.Squares: 1 g/L; Circles: 20 g/L; Closed symbols: growing particles; Open symbols:shrinking particles.

Fig. 5. Synchrotron X-ray pair distribution functions of sonicated and as-made ZIF-8. Solid line: as-made ZIF-8; Dashed line: ZIF-8 sonicated for 5 min; Dotted line:sonicated for 10 min; Dashed-Dotted line: sonicated for 20 min.

296 J.A. Thompson et al. / Microporous and Mesoporous Materials 158 (2012) 292–299

disappearance of the smaller particle population, an effect of Ost-wald ripening. In both dispersions, the larger particles show pri-marily an increasing particle size with time; however, in the20 g/L dispersion the larger particle size actually shows an appar-ent decrease at long sonication times (10 min). This leads to thehypothesis that elastic stress effects may play a role in the particleripening (see Fig. 4 and discussion). It has been shown that a largeincrease in elastic stress in ‘‘soft’’ particles leads to their breakageinto smaller particles to minimize the energy of the system [29].This may also be accompanied by a morphological change; for in-stance, spherical nanoparticles can become cuboidal as the ratio ofthe interfacial energy to the elastic energy decreases [37]. As the20 g/L dispersion ripens at a faster rate than the 1 g/L suspension,this ratio would decrease more rapidly in the former case, leadingto particles breaking to reduce the elastic stress and formation ofdifferent morphologies seen in SEM.

Powder XRD patterns of the sonicated samples (20 g/L concen-tration) are shown in Fig. 4a. Although the low-angle region of theXRD patterns indicates that the long-range topology of the ZIF-8framework is preserved, the XRD peaks of the sonicated samplesare both shifted as well as broadened. Because XRD peaks at higher2h (lower d-spacing) disappeared with longer sonication time, thisindicates the framework became more locally disordered. These

Fig. 4. Powder XRD patterns (a) and Williamson-Hall plots (b) of as-made ZIF-8 (i,black squares), ZIF-8 sonicated for 5 min (ii, red circles), and ZIF-8 sonicated for10 min (iii, blue triangles). (For interpretation of the references to color in thisfigure legend, the reader is referred to the web version of this article.)

observations suggest that the ZIF-8 nanoparticles maintain long-range crystallinity, but may have a greater degree of static disorder(e.g., rotational orientation of the imidazolate linkers). The crystal-lite domain size and crystal strain as a result of sonication can beestimated respectively from the intercept and the slope of a Wil-liamson-Hall plot, as shown in Fig. 4b and Eq. (1) [38]:

b cos h ¼ g sin hþ KkL

ð1Þ

where b is the full-width half-maximum (FWHM) of the diffractionpeak, h is the diffraction angle, K is a constant equal to 0.9, k is theX-ray wavelength, g is the crystalline strain, and L is the crystallitesize. The FWHM of a set of low-angle peaks was determined by fit-ting the data to a Cauchy–Lorentz peak shape. The as-made ZIF-8and ZIF-8 sonicated for 5 minutes both show a negligible slope,indicating that the crystal lattice is not significantly strained evenafter 5 minutes of intense sonication. However at 10 minutes ofsonication, there is a clear development of lattice strain (�4%), aswell as an increase in crystallite domain size consistent with signif-icant Ostwald ripening and formation of larger crystallites (see insetTable in Fig. 4b). The lattice strain developed at 10 min of sonicationcorrelates well with the behavior observed in Fig. 3. As a large elas-tic stress develops in the particles, they begin to break apart and/orform different morphologies (both shown in Fig. 1b and c) to reducethe overall energy of the system. The synchrotron X-ray PDF pat-terns (Fig. 5) also corroborate the observations from powder XRDpatterns of maintenance of the crystal structure. The distributionsof short-range atomic distances (<1 nm) in sonicated and as-madeZIF-8 are essentially identical, thereby indicating that sonicationdoes not cause localized defects (such as missing metal centers orligands) in the bulk structure. Although there are slight changesin the PDFs at 3.26 and 8.33 Å, this is not indicative of significantdifferences between the materials, since the changes do not appearto be a function of sonication time.

Nitrogen physisorption isotherms of as-made and sonicatedZIF-8 are shown in Fig. 6. The calculated BET surface area,Langmuir surface area, and t-plot micropore volume are shownin Table 1. Because ZIF-8 has an adsorption inflection at P/P0 � 0.005, selection of the range for BET and Langmuir surfacearea calculations were done by consistency criteria establishedpreviously [39]. There was a noticeable decrease (�10%) in themicropore volume and Langmuir surface area after sonication,although the BET surface area did no change significantly.

Fig. 6. Nitrogen physisorption isotherms at 77 K. Squares: as-made ZIF-8; Circles:ZIF-8 sonicated for 5 min; Triangles: ZIF-8 sonicated for 10 min.

Table 1Surface area and micropore volume of ZIF-8 based on N2 physisorption isotherms at77 K, showing decreases in both Langmuir surface area and t-plot micropore volume.

Sonication time(min)

BET SA(m2/g)

Langmuir SA(m2/g)

t-plot micropore volume(cm3/g)

0 1700 ± 60 1870 ± 50 0.64 ± 0.035 1650 ± 60 1720 ± 50 0.56 ± 0.03

10 1710 ± 60 1740 ± 50 0.58 ± 0.03

J.A. Thompson et al. / Microporous and Mesoporous Materials 158 (2012) 292–299 297

FTIR and FT-Raman spectra (Figure S7 in Supporting Informa-tion) were used to further examine the chemical bonding changesin the material. After sonication for 5–10 min, an additional vibra-tional band appeared in the FTIR spectra at 480 cm�1 and thevibration at 3400 cm�1 increased in intensity relative to othervibrations in the framework. These vibrations correspond tov(Zn–O) and v(–OH), respectively [40]. EDX analysis of sonicatedZIF-8 nanoparticles (Figure S3 and S4) indicates that these particleshave significantly higher oxygen content than as-made ZIF-8.Interestingly, there were no changes in the FT-Raman spectra aftersonicating for 10 min. Since the appearance of Raman-active pho-nons in ZnO is strongly related to the inherent symmetry of thecrystalline wurtzite structure and the wurtzite structure generallyhas different morphology than ZIF-8, these oxygen-rich domainscan be attributed to surface hydroxyl (Zn–OH) groups [41,42].

Overall, the characterization results are consistent with the con-clusion that the bulk structure of ZIF-8 remains essentially intactafter sonication and even after Ostwald ripening, albeit with asomewhat higher degree of static disorder shown by XRD. Whilethe bulk of the resulting ZIF-8 particles may retain a structure closeto that of the as-made ZIF-8 crystals, our observations indicate thatthe outer regions and surfaces of the sonicated particles likely con-tain more localized defects. The reduction in Langmuir surface areaand micropore volume are also consistent with an increased fre-quency of defects (e.g., Zn–OH bonds occurring at the particle sur-face) and pore blockages created during rapid recrystallization ofZIF-8 particles upon sonication. This could alter the rate of molec-ular diffusion into the crystal if the micropores on the surface arepartially blocked, resulting in ZIF-8 composite membrane perme-ation behavior that cannot be predicted from the structure of theas-made ZIF-8 material or from permeation data collected fromas-made ZIF-8 membranes [43].

3.2. ZIF-8/Matrimid� composite membranes

Composite membranes were prepared by the solution-castingtechnique. Permeation measurements were performed on mem-branes containing 0, 10, and 25 wt% of ZIF-8 and subjected totwo different sonication methods (direct and indirect) for dispers-ing the ZIF-8 particles prior to membrane casting. Different sonica-tion conditions were used to illustrate the large differences inultrasound energy intensity between direct and indirect sonica-tion. The power output per unit area was 156 W/cm2 for directsonication and 3.78 W/cm2 for indirect sonication [44]. Cross-sec-tions of membranes prepared by both sonication methods areshown in Fig. 7. While the dispersion of nanoparticles in the poly-mer matrix varies, Fig. 7a and b show no large agglomerates. Over-all, the ZIF-8 particles showed good adhesion to the polymer.

FTIR measurements of as-made films indicated that there was asmall shift for the symmetric imide carbonyl group frequency ofthe polymer (Figure S8 in Supporting Information) in the compos-ite membranes when compared to the pure polymer membrane.This suggests that the carbonyl groups of the polymer interact withthe functional groups on the surface of the ZIF-8 particles. Uponannealing at 523 K, the imide carbonyl group vibrational frequency(�1775 cm�1) is similar for the composite films while the pureMatrimid� film still shows a higher carbonyl frequency(�1780 cm�1). Additionally, the broadening of the carbonyl bandin the composite membranes (FWHM �20 cm�1) is noticeably lar-ger than in pure Matrimid� (�10 cm�1). These observations indi-cate that the composite membrane carbonyl groups are indifferent, more widely distributed electronic environments thanthe pure polymer. The annealed films were analyzed by DSC todetermine shifts in the glass transition temperature (Tg). In Fig. 8,there is a clear, upward shift in Tg from 583 to 593 K. This shift doesnot appear to be a strong function of the loading of ZIF-8. Previ-ously, it was shown that only a single Tg was observable in ZIF-8/poly-(ethersulfone) composite membranes, which showed noapparent Tg shift [45]. Considering the more electronegative natureof the carbonyl versus the sulfonyl functional groups, we hypothe-size that the shifts seen in both FTIR and DSC are due to a change inthe conjugation of the carbonyl functional groups at the ZIF-8 sur-face, interacting with terminal imidazolate groups.

Membranes prepared by indirect sonication are shown in Fig. 7cand d. There was no change in the particle size, polydispersity, ormorphology in comparison to the as-made ZIF-8. Because of itsmuch smaller power density, the indirect sonication method doesnot provide enough energy to fully break apart nanoparticle aggre-gates or cause nanoparticle ripening. Although the ZIF-8 particlesstill showed good adhesion with the polymer, the nanoparticlesin the matrix were in large aggregates, ranging up to severalmicrometers in size (emphasized with red circles in Fig. 7c andd). Poor dispersion typically creates paths for unselective transportof gases between the particles during permeation, resulting in alarge increase in permeability and a decrease in selectivity [14].On the other hand, the use of the direct sonication method allowedsuccessful dispersion of the ZIF-8 nanoparticles in the polymersolution, but also lead to significant changes in the particle size dis-tribution (Ostwald ripening) and the quality of the ZIF-8 crystals atthe end of the sonication procedure.

The results of pure gas permeation measurements of the com-posite membranes are shown in Fig. 9. Table S1 (Supporting Infor-mation) lists the values plotted in Fig. 9. The error reportedrepresents the standard deviation of the average permeability ob-tained for each membrane reported. The CO2/CH4 gas pair separa-tion was considered due to previous studies showing pure ZIF-8membranes have poor separation performance. Membranes pre-pared with direct sonication show a large increase in the permeabil-ity of CO2 as well as a modest increase in ideal gas selectivity

Fig. 7. SEM images of cross sections of ZIF-8/Matrimid� composite membranes prepared by direct (a, b) and indirect (c, d) sonication: (a, c) 10 wt% loading; (b, d) 25 wt%loading.

Fig. 8. DSC curves of the second heating cycle for (a) Pure Matrimid�, (b) 10 wt%ZIF-8/Matrimid�, and (c) 25 wt% ZIF-8/Matrimid�. There is a clear shift in Tg from583 to 593 K; however, the Tg does not appear to be a strong function of the weightloading of ZIF-8.

Fig. 9. Permeation properties at 308 K of ZIF-8/Matrimid� composite membranes,prepared using either direct or indirect sonication. The wt% loadings of ZIF-8 in themembrane are indicated. The Maxwell model is used to predict the membraneperformance up to 50 wt% ZIF-8 loading. Squares: direct sonication; Circles: indirectsonication; Solid line: Upper Bound; Dotted line: Maxwell model predictions.

298 J.A. Thompson et al. / Microporous and Mesoporous Materials 158 (2012) 292–299

compared to the pure polymer film. On the other hand, the use ofindirect sonication results in membranes with poor performance,in particular a drop in the selectivity and very high CO2 permeabil-ity even at low ZIF-8 loadings. These characteristics are due to unse-lective transport through the void spaces in the regions occupied byaggregates of ZIF-8 nanoparticles. The permeation data for compos-ite membranes prepared with direct sonication exhibit characteris-tics of a generally defect-free microstructure. Furthermore, our datashow a clear and consistent trend in enhancement of membraneperformance as a function of ZIF-8 loading. Such a set of permeationdata can be analyzed in terms of the Maxwell model (see Eq. (2)),which can be used to estimate the permeation properties of theZIF-8 filler particles from composite membrane data [46–48]:

Peff ¼ PmPf þ 2Pm � 2Uf ðPm � Pf ÞPf þ 2Pm þUf ðPm � PfÞ

� �ð2Þ

where P is the permeability (Barrer), U is the volume fraction, andsubscripts eff, m, and f stand for effective (composite), matrix

(polymer), and filler (ZIF-8) phases. This analysis leads to a CO2

permeability of 300 Barrer and CO2/CH4 ideal selectivity of 85 forZIF-8, based on fitting the experimental data from Fig. 9.

The permeability and selectivity values estimated here reflectthe properties of the ZIF-8 material actually constituting the fillerin the membranes. These cannot be considered to be the same asthose of the as-made ZIF-8 material. Indeed, it is not clear if thereis a single ‘‘definitive’’ set of permeation properties of ZIF-8 mate-rials in their practically applied form since the findings from thispaper indicate clearly that MOFs such as ZIF-8 can be significantlyaltered by commonly used membrane processing conditions. Thereis also increasing evidence that many MOFs possess flexible frame-works capable of adsorption and permeation of molecules largerthan the nominal crystallographic pore size [12,13,49]. It has alsobeen shown that the ZIF-8 framework is much less rigid in compar-ison with other ZIF materials [50]. With these considerations, it is

J.A. Thompson et al. / Microporous and Mesoporous Materials 158 (2012) 292–299 299

reasonable that composite membranes containing ZIF-8 may showquite different gas separation performance than pure ZIF-8 mem-branes due to alterations in the structure of the ZIF material duringdifferent processing conditions. A recent publication [51] showedthat ZIF-8 enhanced the performance of Ultem� hollow-fiber mem-branes for CO2/N2 separation, whereas other studies have shownZIF-8 to have poor separation performance. With the sonication-in-duced structural changes, adsorption of polymer to the surface ofthe ZIF-8 crystals, and subsequent annealing at high temperatures,the effective pore size and flexibility of ZIF-8 could be reduced toprovide the separation enhancement observed in ZIF-8 compositemembranes presented in this work. Finally, the Maxwell model isalso used (Fig. 9) to extrapolate composite membrane performanceup to a hypothetical 50 wt% ZIF-8/Matrimid� membrane. Althoughthe present membranes show an increase in separation perfor-mance, the present combination of ZIF-8 and Matrimid� is not pre-dicted to allow the fabrication of composite membranes that reachthe ‘‘upper bound’’ of pure polymer membrane separation perfor-mance [52]; therefore, other polymers can be selected as matrixmaterials for ZIF-8 if desired [53].

4. Conclusions

High-intensity ultrasonication with a sonication horn is shownto induce Ostwald ripening on ZIF-8 nanoparticles in THF.Although there are significant changes in the particle morphology,there are only minor losses in crystallinity and microporosity asconcluded from powder XRD, PDF analysis and nitrogen physisorp-tion. From FTIR and EDX data, an increase in Zn–OH groups is ob-served, most likely due to surface defects created during theripening process. Light scattering measurements show a cleartrend in ripening of particles, and sonication eventually creates abimodal particle size distribution. The subsequent behavior ofthe bimodal distribution is affected by the sonication time andthe development of elastic stresses in the growing particles. Com-posite films prepared with both direct and indirect sonicationshow apparently good adhesion between the polymer and ZIF-8phases; however, films fabricated using indirect sonication exhibitsevere agglomeration of nanoparticles while direct sonication pro-duced ripened nanoparticles with variable dispersion. Permeationmeasurements reveal that direct sonication produces an effectivecomposite membrane system whose properties are enhanced overthe pure polymer material, and show a full consistency with theMaxwell model. The latter fact also enables a reliable estimationof the permeation properties of the ZIF-8 particles existing in thecomposite membranes, and they are shown to be quite differentfrom those obtained using measurements of ZIF-8 crystals, pureZIF-8 membranes, or computational predictions.

Acknowledgments

This work was supported by King Abdullah University of Sci-ence and Technology under Award No. KUS-I1-011-21. Work doneat Argonne National Laboratory and use of the Advanced PhotonSource, an Office of Science User Facility operated for the U.S.DOE Office of Science by Argonne National Laboratory, were sup-ported by the U.S. DOE under Contract No. DE-AC02-06CH11357.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.micromeso.2012.03.052.

References

[1] X.-C. Huang, Y.-Y. Lin, J.-P. Zhang, X.-M. Chen, Angew. Chem. Inter. 45 (2006)1557–1559.

[2] U.P.N. Tran, K.K.A. Le, N.T.S. Phan, ACS Catal. 1 (2011) 120–127.[3] J. Cravillon, S. Münzer, S.J. Lohmeier, A. Feldhoff, K. Huber, M. Wiebcke, Chem.

Mater. 21 (2009) 1410–1412.[4] K.S. Park, Z. Ni, A.P. Côté, J.Y. Choi, R. Huang, F.J. Uribe-Romo, H.K. Chae, M.

O’Keeffe, O.M. Yaghi, Proc. Nat. Acad. Sci. 103 (2006) 10186–10191.[5] J.A. Greathouse, M.D. Allendorf, J. Am. Chem. Soc. 128 (2006) 10678–10679.[6] H. Bux, F. Liang, Y. Li, J. Cravillon, M. Wiebcke, J. Caro, J. Am. Chem. Soc. 131

(2009) 16000–16001.[7] S.R. Venna, M.A. Carreon, J. Am. Chem. Soc. 132 (2010) 76–78.[8] M.C. McCarthy, V. Varela-Guerrero, G.V. Barnett, H.-K. Jeong, Langmuir 26

(2010) 14636–14641.[9] C. Chemik, H. Bux, H. Vob, J. Caro, Chem. Ing. Tech. 83 (2011) 104–112.

[10] H. Bux, A. Feldhoff, J. Cravillon, M. Wiebcke, Y.S. Li, J. Caro, Chem. Mater. 23(2011) 2262–2269.

[11] E. Haldoupis, S. Nair, D.S. Sholl, J. Am. Chem. Soc. 132 (2010) 7528–7539.[12] K. Li, D.H. Olson, J. Seidel, T.J. Emge, H. Gong, H. Zeng, J. Li, J. Am. Chem. Soc. 131

(2009) 10368–10369.[13] D. Fairen-Jimenez, S.A. Moggach, M.T. Wharmby, P.A. Wright, S. Parsons, T.

Düren, J. Am. Chem. Soc. 133 (2011) 8900–8902.[14] C. Liu, S. Kulprathipanja, A.M.W. Hillock, S. Husain, W.J. Koros, in: N.N. Li, A.G.

Fane, W.S.W. Ho, T. Matsuura (Eds.), Advanced Membrane Technology andApplications, John Wiley & Sons, Inc., Hoboken, N.J., 2008, pp. 789–819.

[15] S. Shu, S. Husain, W.J. Koros, J. Phys. Chem. C 111 (2007) 652–657.[16] T.H. Bae, J. Liu, J.K. Lee, W.J. Koros, C.W. Jones, S. Nair, J. Am. Chem. Soc. 131

(2009) 14662–14663.[17] T.H. Bae, J. Liu, J.A. Thompson, W.J. Koros, C.W. Jones, S. Nair, Micropor.

Mesopor. Mater. 139 (2011) 120–129.[18] S. Shu, S. Husain, W.J. Koros, Chem. Mater. 19 (2007) 4000–4006.[19] M.J.C. Ordoñez, K.J. Balkus, J.P. Ferraris, I.H. Musselman, J. Membr. Sci. 361

(2010) 28–37.[20] T.H. Bae, J.K. Lee, W. Qiu, W.J. Koros, C.W. Jones, S. Nair, Angew. Chem. Inter.

Ed. 49 (2010) 9863–9866.[21] R. Adams, C. Carson, J. Ward, R. Tannenbaum, W.J. Koros, Micropor. Mesopor.

Mater. 131 (2010) 13–20.[22] C. Zhang, Y. Dai, J.R. Johnson, O. Karvan, W.J. Koros, J. Membr. Sci. 389 (2012)

34–42.[23] D.W. Jung, D.A. Yang, Ju. Kim, Ja. Kim, W.S. Ahn, Dalton Trans. 39 (2010) 2883–

2887.[24] P.J. Chupas, X. Qiu, J.C. Hanson, P.L. Lee, C.P. Grey, S.J.L. Billinge, J. Appl.

Crystallogr. 36 (2003) 1342–1347.[25] P.J. Chupas, K.W. Chapman, P.L. Lee, J. Appl. Crystallogr. 40 (2007)

463–470.[26] A.P. Hammersley, S.O. Svensson, M. Hanfland, A.N. Fitch, D. Häusermann, High

Pressure Res. 14 (1996) 235–248.[27] X. Qiu, J.W. Thompson, S.J.L. Billinge, J. Appl. Crystallogr. 37 (2004) 678.[28] K.C. O’Brien, W.J. Koros, T.A. Barbari, J. Membr. Sci. 29 (1986) 229–238.[29] P.W. Voorhees, Annu. Rev. Mater. Sci. 22 (1992) 197–215.[30] E.M. Wong, J.E. Bonevich, P.C. Season, J. Phys. Chem. B 102 (1998) 7770–7775.[31] P.W. Voorhees, J. Stat. Phys. 38 (1985) 231–252.[32] K.S. Suslick, S.J. Doktycz, E.B. Flint, Ultrasonics 28 (1990) 280–290.[33] B.J. Schoeman, Zeolites 18 (1997) 97–105.[34] I.M. Lifshitz, V.V. Slyozov, J. Phys. Chem. Solids 19 (1961) 35–50.[35] C. Wagner, Z. Elektrochem. 65 (1961) 581–591.[36] R. Finsy, Langmuir 20 (2004) 2975–2976.[37] X. Li, K. Thornton, Q. Nie, P.W. Voorhees, J.S. Lowengrub, Acta Mater. 52 (2004)

5829–5843.[38] G.K. Williamson, W.H. Hall, Acta Metall. 1 (1953) 22–31.[39] K. Walton, R. Snurr, J Am. Chem. Soc. 129 (2007) 8552–8556.[40] G. Zhang, X. Shen, Y. Yang, J. Phys. Chem. C 115 (2011) 7145–7152.[41] M. Zhu, S.R. Venna, J.B. Jasinski, M.A. Carreon, Chem. Mater. 23 (2011) 3590–

3592.[42] X.L. Xu, S.P. Lau, B.K. Tay, Thin Solid Films 398–399 (2001) 244–249.[43] D.S. Sholl, Nature Chem. 3 (2011) 429–430.[44] T.J. Mason, J.P. Lorimer, D.M. Bates, Ultrasonics 30 (1992) 40–42.[45] K. Díaz, L. Garrido, M. López-González, L.F. del Castillo, E. Riande,

Macromolecules 43 (2010) 316–325.[46] R. Mahajan, W.J. Koros, Polym. Eng. Sci. 42 (2002) 1420–1431.[47] D.Q. Vu, W.J. Koros, S.J. Miller, J. Membr. Sci. 211 (2003) 335–348.[48] S.A. Hashemifard, A.F. Ismail, T. Matsuura, J. Membr. Sci. 347 (2010)

53–61.[49] E. Stavitski, E.A. Pidko, S. Couck, T. Remy, E.J.M. Hensen, B.M. Weckhuysen, J.

Denayer, J. Gascon, F. Kapteijn, Langmuir 27 (2011) 3970–3976.[50] C.J. Tan, T.D. Bennett, A.K. Cheetham, Proc. Nat. Acad. Sci. 107 (2010) 9938–

9943.[51] Y. Dai, J.R. Johnson, O. Karvan, D.S. Sholl, W.J. Koros, J. Membr. Sci. 401–402

(2012) 76–82.[52] L.M. Robeson, J. Membr. Sci. 62 (1991) 165–185.[53] S. Sridhar, B. Smitha, T.M. Aminabhavi, Sep. Purif. Rev. 36 (2007) 113–174.