Article pubs.acs.org/Langmuir
Simulation-Aided Design and Synthesis of Hierarchically Porous Membranes Fan Li,† Molly B. Wilker, and Andreas Stein* Department of Chemistry, University of Minnesota, 207 Pleasant St. S.E., Minneapolis, Minnesota 55455, United States S Supporting Information *
ABSTRACT: Free-standing silica membranes with hierarchical porosity (ca. 300 nm macropores surrounded by 6−8 nm mesopores) and controllable mesopore architecture were prepared by a dual-templating method, with the structural design aided by mesoscale simulation. To create a twodimensional, hexagonal macropore array, polymeric colloidal hemisphere arrays were synthesized by a two-step annealing process starting with non-close-packed polystyrene sphere arrays on silicon coated with a sacrificial alumina layer. A silica precursor containing a poly(ethylene) oxide−poly(propylene oxide)−poly(ethylene) oxide (PEO−PPO−PEO) triblockcopolymer surfactant as template for mesopore creation was spin-coated onto the support and aged and then converted into the free-standing membranes by dissolving both templates and the alumina layer. To test the hypothesis that the mesopore architecture may be influenced by confinement of the surfactant-containing precursor solution in the colloidal array and by its interactions with the polymeric colloids, the system was studied theoretically by dissipative particle dynamics (DPD) simulations and experimentally by examining the pore structures of silica membranes via electron microscopy. The DPD simulations demonstrated that, while only tilted columnar structure can be formed through tuning the interaction with the substrate, perfect alignment of 2D hexagonal micelles perpendicular to the plane of the membrane is achievable by confinement between parallel walls that interact preferentially with the hydrophilic components (PEO blocks, silicate, and solvent). The simulations predicted that this alignment could be maintained across a span of up to 10 columns of micelles, the same length scale defined by the colloidal array. In the actual membranes, we manipulated the mesopore alignment by tuning the solvent polarity relative to the polar surface characteristics of the colloidal hemispheres. With methanol as a solvent, columnar mesopores parallel to the substrate were observed; with a methanol−water mixed solvent, individual spherical mesopores were present; and with water as the only solvent, twisted columnar structures were seen.
■
INTRODUCTION Ultrathin films or membranes with equally sized, well-defined through-hole pores have applications in various separation, filtration, and purification processes.1 In this regard, precise control of parameters such as pore size, pore spacing, and film thickness is highly desirable.2 A narrow pore size distribution is critical for size-exclusive separations with a definite “cutoff” size, whereas high pore density and thin film thickness are necessary to allow high flux and improve the separation efficiency.3 Commercially available products, including fiber glass membranes, cellulose, particle-track-etched films, etc., are disadvantageous due to their irregular porosity, large thickness, or low pore density. Technologies such as lithography and anodic corrosion are capable of generating membranes with densely packed, uniform, cylindrical pores. However, these methods suffer from high cost or limited availability of materials. Alternatively, the fabrication of porous membranes at the submicrometer length scale may be realized following selfassembly approaches.2 In contrast to top-down techniques, selfassembly relies on the intrinsic ability of the materials to achieve spontaneous structural organization. Porous membranes have been synthesized from surfactants4 or block copolymers.5 Block copolymers are capable of realizing phase © 2012 American Chemical Society
separation from several nanometers to micrometers, and a porous membrane could thus be produced by selective etching of one block/component. The phase behavior of block copolymers is influenced by affinity, confinement, shearing, external field, temperature, and other parameters, providing a rich toolbox for tailoring porous structures.5 Another self-assembly approach for porous thin films is based on colloidal templating.6 Colloidal assembly leads to many interesting structures in one, two, and three dimensions, and the templating strategy opens up opportunities for the replication of the structures with a broad range of materials.7 Monodisperse colloidal spheres can self-assemble into a monolayer of an ordered, hexagonally close-packed array, namely a 2D colloidal crystal, on a substrate. By filling the interstitial space with another material followed by removal of the spheres, a porous membrane with uniform circular pores is formed. As a hard templating approach, colloidal templating is readily available for a wide range of materials.6 Received: January 3, 2012 Revised: March 9, 2012 Published: April 21, 2012 7484
dx.doi.org/10.1021/la300037q | Langmuir 2012, 28, 7484−7491
Langmuir
Article
Figure 1. Schematic illustration of the fabrication of a 2D non-close-packed colloidal hemisphere array and its use as a template for a 2D submicrometer porous membrane.
Figure 2. Schematic illustrations describing the choice of the simulation cell. (a) Blue: unit cell of 2D hexagonal lattice; red: reduced cell for simulation. (b) Cross-sectional view of reduced cell in (a). (c) Construction of the actual simulation cell.
μm.13 Here, the target structure is an ultrathin membrane based on a single colloidal layer, which in turn provides us an opportunity to precisely manipulate and observe the mesoporous structures. Additionally, it is known that 2D confinement at the submicrometer scale can have a profound impact on the resulting mesostructures.14 To aid the design of the membrane structure, we used a simulation based on dissipative particle dynamics (DPD) to elucidate the phase behavior of the block copolymer within 2D colloidal confinement and to explore possible routes for structural control based on the mechanistic understanding. Several different methods have been described to replicate the colloidal array, including vapor phase deposition15 and electrochemical plating.6,8 A simpler approach is to directly cast a precursor onto the colloidal film in liquid form and allow it to solidify. In this regard, spin-coating is an effective and reliable approach to ensure a uniform coating on a substrate over a large scale. Jiang et al. developed a method to prepare porous membranes through codeposition of mixtures of a sol precursor and colloids that were spin-coated onto a substrate, the colloids being removed after precursor gelation.16,17 However, films derived in this manner were rather thick. An alternative way would be to spin-cast a precursor on a preformed colloidal film template.9 However, direct application of a colloidal monolayer by spin-coating was proven to be unsuccessful due to the low affinity of the colloids with the substrate.18 DPD as a mesoscopic simulation method is a suitable tool to examine the phase behavior of polymer systems.19 Mesoscale modeling was developed to address the incompatibility between fast molecular kinetics and macroscale properties by treating polymer chains at the coarse-grained level.20 Although previous work mainly focused on the phase study of block copolymer melts, recently additional factors were incorporated, such as solvents, nanoparticles, confinement, and even dynamic systems.21−24 In a previous study, mesoscopic simulations (MesoDyn) were utilized to study the process of mesoporous materials synthesis. However, no synthesis work was conducted based on the simulation results.25 Herein we employ the DPD
In this work, we investigated the synthesis of a membrane with ordered circular pores and smaller textural mesopores by using both colloidal spheres and block copolymer surfactants as cotemplates. Because of their different sizes, an integration of colloidal crystals and surfactants led to simultaneously templating at two different levels and produced porous structures with two or more distinct pore sizes. While 3D hierarchically porous structures based on this strategy have been extensively studied, few reports were devoted to creating similar 2D porous structure,8−10 especially ultrathin membranes based on the templating of a single layer of colloids. Herein, we fabricated a non-close-packed hemispherical array based on etching and annealing of a polystyrene (PS) colloidal monolayer as a robust substrate for templating the porous membrane through spin-coating (see Figure 1). In addition, a triblock copolymer (the nonionic surfactant Pluronic P123, PEO20PPO70PEO20) was introduced into the precursor as a template for textural porosity. In a membrane with dual porosity, the bigger pores may serve for size selection, and the smaller mesopores can provide additional pathways for fluid transport. Therefore, we expected that the hierarchically porous membrane can offer higher throughput with a lower pressure buildup.11 For this purpose, it is necessary to ensure that the mesopores are open and easily accessible on both sides of the membrane. A desirable structure would consist of a 2D hexagonal (P6mm) mesopore array perpendicular to the plane of the membrane; cubic mesopore structures with 3D connectivity are also preferred. Previously, we explored the alignment of 2D hexagonal mesopores within 3D colloidal crystals and suggested that the mesoporous structures were associated with the block copolymer structure and composition, the surface properties of the polymer colloid, and spatial confinement.12 However, the exact relationship between these multiple factors and the underlying mechanism was still unclear. Through the dual templating technique, we also demonstrated the fabrication of a composite porous membrane with functionalized mesoporous spheres inside a hierarchically porous matrix, but the membrane thickness was still above 10 7485
dx.doi.org/10.1021/la300037q | Langmuir 2012, 28, 7484−7491
Langmuir
Article
hexagonally close-packed, single-layer PS sphere array was deposited on top of the Al2O3 layer by spin-coating a 1:1 ethylene glycol (EG)/ H2O suspension containing ca. 15 vol % PS spheres (diameter 468 ± 12 nm) at 1000 rpm for 16 min. Parameters were adjusted to maximize coverage of the wafer by a colloid monolayer without the formation of any multiple layers. The colloidal film was O2 plasma etched in a STS 320 etcher running at low power (50 W), where the colloids were etched isotropically resulting in non-close-packed sphere array. Afterward, the colloidal film was annealed to convert spheres into domelike structures. The annealing should be performed immediately after the reactive ion etching process, as non-close-packed spheres are unstable before annealing and spontaneously aggregate. To finely adjust the profile of the hemisphere, a two-step annealing process was developed. The freshly etched template was first subjected to 1 min “pulse annealing” at 140 °C for 10 cycles, with 1 min cooling at room temperature between cycles. During the second stage, the sample was annealed at 115 °C, and after 75 min, the colloids had formed uniform hemispherical shapes on the substrate. The wafer was then cut into ca. 2 × 2 cm2 chips. To compare the stability of the pristine colloidal monolayer and the annealed hemisphere layer, the samples were placed in ethanol and subjected to gentle sonication for certain periods of time, and changes in sample structure were tracked using reflectance UV−vis spectroscopy at a 90° angle of incidence (Spectral Instruments 400 spectrophotometer) (Figure S5). Dual Templating of Hierarchically Porous Membrane. A silica sol−gel precursor was prepared by mixing 1 g of tetramethyl orthosilicate (TMOS, Aldrich), 1 g of Pluronic P123 (donated by BASF), 0.5 g of HCl (0.1 M) (Mallinckrodt), and 6 g of H2O or methanol (MeOH, Aldrich) and stirring for 1 h (molar ratio = 1 TMOS:0.27 P123:0.008 HCl:4.3−56.1 H2O:0−29.1 MeOH). A drop of this precursor solution (ca. 50 μL) was applied to a silicon chip with a colloidal hemisphere array, and the chip was then spun with a homemade spin-coater at 300 rpm in ambient humidity until the solvent evaporation was complete. After being aged in a closed container at 50 °C for 24 h, the chip was immersed in toluene (Mallinckrodt) to dissolve the templates and dried at 50 °C. To obtain a free-standing membrane, the sample was soaked in 5 M H3PO4 for several hours to dissolve the underlying Al2O3 layer. Characterization. SEM micrographs were recorded on a JEOL JEM-6700 field emission microscope using an accelerating voltage of 5 kV and an emission current of 20 mA. Chips were directly loaded into the microscope without any conductive coating layer.
method to investigate the mesopore formation by a mixture of surfactant and silica precursor within colloidal hemispherical confinement. Besides helping verify our hypothesis that colloids can be used to control mesopore alignment, we also expect the simulation system to provide valuable insight about tuning the interactions between colloids and surfactants for the experimental design. Hierarchically structured porous silica membranes were then fabricated by dual templating with surfactants and two-dimensional hemispherical polymer arrays, and the effects of solvent polarity on mesopore architecture were investigated.
■
EXPERIMENTAL SECTION
DPD Simulation. The DPD simulations were performed using the Materials Studio v4.3 package (Accelrys). A rectangular unit cell of the non-close-packed 2D hemispherical array is shown in Figure 2a (blue box). Assuming 400 nm colloids are used as the template, the volume of the unit cell would be over 50 000 000 nm3, several orders of magnitude higher than typically involved in mesoscopic simulations. As a simplification, we chose the region between two adjacent spheres for the simulations (Figure 2a, red box). Because the sphere radius is much larger than the surfactant molecules, this structure can be further simplified into a confined space between two parallel walls, together with a substrate and a free upper surface (Figure 2b), and the actual simulation cell using these assumptions is shown in Figure 2c. The parallel walls and substrate were composed of frozen hard spheres, whereas the free surface was simulated by using a “bad solvent”.26 The “bad solvent” was bounded at the top of the simulation cell and was immiscible with the block copolymer/silica phase. The presence of the deformable air phase (“bad solvent”) eliminated the case of structural frustration potentially caused by hard confinement. The dimensions of the simulation cell were 16 × 8−32 × 32−64 of DPD grids, with each grid unit corresponding to a length of ∼2 nm in reality. In the mesoscopic simulation, the block copolymers were represented by strings of beads connected by springs, following dynamics governed by Newton’s law.19 The density of the system was represented by the number of beads per cubic grid unit and was set to 3 by default. Each bead represented a minimal polymer coil whose internal structure was small compared to the macroscopic scale and was considered to have little influence on the phase behaviors.27 The interactions between different blocks in the polymer, as defined by the Flory−Huggins parameter χ, were replaced by the bead-to-bead repulsion parameter εij = 25 + 3.5χij. Although using more beads could have improved the details of the simulated structures and allowed more precise mapping of a specific real polymer, it would also have demanded more computation time. Since this study dealt with a very large system whose dimensions were determined by the size of the colloidal confinement and our focus was to acquire qualitative insight into direct the experimental design, smaller numbers of beads were more practical choices. Here we used one bead for each block and built a triblock copolymer as A1B1A1, with an additional bead (M) representing a certain volume of the fluidic silica precursor. The ratio of A1B1A1 to M was 4:1, which resulted in a stable 2D hexagonal mesophase in the absence of external confinement. Previous studies on polymer dispersions showed that such a simplification could still present reasonable phase information.19,28 The parametrization of the simulation is discussed in the Results and Discussion section. The initial state of the system was randomly mixed except for the substrate, ceiling, and walls, which consisted of beads with fixed positions. The simulation time was set to be 200 000 steps, and the evolution of the phase changes was monitored by the diffusivity (Figure S1 in the Supporting Information). Snapshots were taken every 20 000 steps (Figures S2−S4). Fabrication of a 2D Hemispherical Array. PS colloidal spheres were prepared via emulsifier-free emulsion polymerizations of styrene as described previously.29 A pristine silicon wafer [p-doped, (100) surface orientation] was first coated with 50 nm Al2O3 by atomic layer deposition (ALD) using a Cambridge NanoTech ALD system. A
■
RESULTS AND DISCUSSION Controlling Mesopore Alignment via Surface Interactions and Confinement. The interaction parameters of different components in the DPD simulation system are summarized in Table 1. In the parametrization process, the first Table 1. Definitions of Interaction Parameters Used in the DPD Simulation System
block (A) block (B) silica precursor (M)
block (A)
block (B)
silica precursor (M)
substrate (S)
parallel walls (W)
εAA εAB εAM
εAB εBB εBM
εAM εBM εMM
εAS εBS εMS
εAW εBW εMW
step is to determine the interactions between two blocks (εAB) of the P123 surfactant and between blocks and the silica precursor (εAM and εBM). Herein, the parameters were adopted from previous reports of successful simulations of similar Pluronic surfactants and rescaled based on the number of beads used in the current simulations.28 The interactions between different chains and water blocks are represented by calculating repulsion parameters, which are linearly correlated to polymer 7486
dx.doi.org/10.1021/la300037q | Langmuir 2012, 28, 7484−7491
Langmuir
Article
Figure 3. Different alignments of columnar mesopores induced by different affinities with the substrate. (a) εBS − εAS = 0, (b) εBS − εAS = 100, and (c) εBS − εAS = 300. The air is represented by the green dotted layer above the mesophases. Block A is colored olive green, block B blue, and the silica precursor red. The solid green surface is the isosurface defining the boundary between hydrophobic and hydrophilic blocks.
of the relative affinity εBS − εAS where the columnar structure could adopt an ordered nonparallel 2D hexagonal arrangement. In fact, the tilted alignment was consistent with the prevailing morphology previously observed experimentally for mesoporous silica films prepared with a P123 surfactant template on a polymer-coated substrate as well as with the consensus that the substrate affinity itself is less effective in dictating the direction of columnar micelles.33 Therefore, the focus of the current study was to use additional walls, namely the colloid matrix, to further manipulate the mesoporous structure. We found that the addition of parallel walls could significantly alter the mesopore alignment. Figure 4 shows two examples with the same
Flory−Huggins parameters, and the correlation factor rescales inversely with the number of beads in the model.30 Because a simplified three-beads (A1B1A1) model was used to represent the P123 triblock surfactant in order to accommodate the large unit cell (up to 64 000 nm3) and the multiple components in the current study, the repulsion parameters were accordingly chosen as εAA = εBB = εAM = 25, εAB = 150, εBM = 200 (see Table S1 for additional parameters). Although such numbers are considerably larger than those used in homogeneous block copolymer simulations,25 a previous study used similar parameters to obtain a reasonable surfactant phase diagram.19 With these parameters, a highly ordered hexagonal mesophase could be obtained in a simulation box without external confinement in a reasonable amount of iterations (200 000 steps). When a mixture of surfactant (block copolymer) and inorganic precursor assembles on a substrate or within confinement, the alignment of the columnar structure formed by the lyotropic liquid crystal is a function of surfactant composition, surfactant−substrate interaction, and structural confinement.31 We adjusted the interactions of the P123 blocks/silica precursor with the substrate (S), namely εAS (εMS = εAS) and εBS, and monitored the impact on the mesoscopic phases (Figure 3). Here, the repulsion εAS was set at 300 and εBS was adjusted between 300 and 600, and in effect, only the net repulsion (εBS − εAS) dictated the final structures.26 When εBS − εAS = 0 (εBS = 300), columnar mesopores parallel to the substrate were obtained (Figure 3a), whereas when εBS − εAS was increased to 100 (εBS = 400), the columns became diagonal within the simulation cell (Figure 3b). Further increasing εBS = 600 again led to parallel mesoporous near the substrate (Figure 3c). Although micelles parallel to the surface could be seen at both neutral and highly hydrophilic surfaces, the exact structures at the interfaces were different as shown in Figure S6 of the Supporting Information. The neutral surface was covered by semicylindrical micelles with both hydrophilic and hydrophobic domains in contact with the substrate, whereas the highly hydrophilic surface was covered by complete cylindrical micelles with the hydrophilic components in touch with the substrate. Such structural similarity on completely different substrates has been experimentally confirmed by the observation of parallel cylindrical mesopores formed on both mica and graphite substrates, with a proposed interfacial geometry consistent with our simulation results as shown in Figure S6.32 While a quantitative interpretation was difficult due to the lack of detail in our simplified surfactant model, it was clear that the mesophases in the current system were sensitive to surfactant−substrate interactions, and there existed a window
Figure 4. Parallel walls with different interactions with the PPO (B) block. (a) εBW − εAW = 200 and (b) εBW − εAW = 300. Side views, 45° views; top and bottom views are shown. Colors are defined as in Figure 3.
simulation cells as above but having two parallel walls. If the walls (W) do not show sufficient affinity for the PEO block/ silicate phase (εAW = εMW = 200, net repulsion εBW − εAW ≤ 200), parts of the columnar mesopores may orient themselves perpendicular to the wall, similar to the situation in Figure 3b, and therefore create a hybrid mesostructure. On the other hand, when the PEO phase is more preferred (εAW = εMW = 200, εBW − εAW = 300), the preference for columns to be parallel to the walls can effectively align them vertically rather 7487
dx.doi.org/10.1021/la300037q | Langmuir 2012, 28, 7484−7491
Langmuir
Article
Figure 5. Mesopore alignment between parallel walls separated by a longer distance. (a) Air phase exclusion occurred. (b) Air phase remained stable. Colors are defined as in Figure 3.
Figure 6. PS hemisphere array with different interparticle spaces controlled by the oxygen plasma time (t). (a) t = 90 s, (b) t = 105 s, and (c) t = 120 s.
Figure 7. SEM micrographs of the porous membrane. (a) Highly ordered hexagonal pore array. (b) Cross-sectional view confirming the uniformity of the membrane. (c) Porous membrane lifted off from the substrate.
than at a tilt angle (Figure 4b). Hence, this simulation proved that additional colloid walls with a high preference for the PEO/silicate components greatly improved our capability to adjust the mesophase alignment within a surfactant-templated mesoporous membrane. While the simulations showed that parallel walls with a preference for PEO could effectively assist the vertical alignment of mesopores, one remaining question was: how far from the wall could this effect still persist? We therefore expanded the simulation cell to increase the wall-to-wall distance four times and performed the simulation with the same parameters. However, an interesting observation was that the air phase (“bad solvent”) was pushed away from the top of the cell (Figure 5a). While it no longer corresponded to a physically meaningful system, it showed that the upper surface also had a certain impact on the self-assembled mesostructure. To avoid this phenomenon, we increased the thickness of the air phase, which effectively stabilized the system and prevented the occurrence of an exclusion. In this case, it was shown that the vertical orientation of the mesopores could be maintained between spaces of up to 10 columns (Figure 5b). Considering that each real mesopore was ca. 6−8 nm in diameter, the width of the current simulation cell (60−80 nm) was on the same length scale as that of the real colloidal hemisphere system where the distance was ca. 100 nm. Therefore, we may
conclude that using the colloidal hemispherical array may be an effective approach toward vertically aligned mesopores, provided that both the substrate and the colloid surface are tailored to dictate the micelle alignment at the interfaces. Morphology and Stability of Hemispherical Array. We used a two-step annealing process to convert a non-closepacked PS colloidal array into a hemispherical array. In this procedure, pulsed annealing at high temperature was believed to preferentially soften the external shell of the colloids, thereby helping to immobilize the spheres onto the substrate. The subsequent extended annealing step slightly above the PS glasstransition temperature then gradually changed spheres into hemispheres. If no pulse annealing was applied to enhance the connection with the substrate, a disordered colloidal array was produced with a high percentage of sphere agglomeration. By controlling the plasma etching time, the sphere size and the interparticle space could be readily tuned (Figure 6). The annealed hemisphere array was highly stable on the substrate compared to the original colloidal sphere array, and the mechanical stability was evaluated by sonication in ethanol and periodic monitoring with a reflectance UV−vis spectrometer. While the colloidal particles were quickly removed from the substrate in the case of an unannealed pristine film, a significant improvement in stability was observed for the pulse-annealed sample, and the two-step annealed sample showed virtually 7488
dx.doi.org/10.1021/la300037q | Langmuir 2012, 28, 7484−7491
Langmuir
Article
Figure 8. SEM images of textural mesoporosity. (a−d) Parallel mesopores formed with methanol as the solvent. (e−h) Transitional mesophases with methanol/water cosolvents. (i−l) Possible vertical mesophase and spherical mesopores formed in water.
preferentially stay in the more polar PEO phase. Therefore, the effective polarity of the PEO blocks changes when solvents of different polarities are used, and correspondingly the interactions between PEO and the substrate/colloidal surface are altered as well. The textural mesoporosity of the porous membranes was examined by high definition SEM imaging, and three distinct mesophases were identified. Methanol as a solvent led to columnar mesopores that were oriented parallel to the membrane (Figure 8a). Previously in a 3D colloidal crystal system we observed that the mesopores usually chose a preferred alignment (parallel or perpendicular) with respect to the colloid surface.12 However, this effect was not prominent for the current sample prepared in methanol. It appeared that the mesopores adopt a roughly unidirectional alignment in rather large domains indifferent to the colloids. As a result, both mesopore alignments were observed, either parallel or perpendicular to the colloidal hemispheres (Figure 8b,c), although clearly all the mesopores were aligned parallel to the substrate (Figure 8d). This may be related to the kinetics of the assembly processrapid drying of the methanol solvent during spinning prevented the formation of more desired structures.34 Hence, water was added as a cosolvent to increase the polarity, and a different mesophase was observed which appeared to be at an intermediate stage between parallel and vertical phases (Figure 8d). Individual spherical mesopores instead of extended columnar structures were observed from
unchanged spectra after 2 min sonication (Figure S7). It is also worth mentioning that, although the colloidal spheres consisted of nonpolar polystyrene, the oxygen plasma treatment was known to significantly increase the surface oxygen content and enhance the polarity. Hierarchically Porous Membranes through Dual Templating. After spin-coating of the silica/block copolymer precursor and template removal, a highly ordered hexagonal pore array was observed by SEM (Figure 7a). The pore size of the membrane was slightly larger (ca. 360 nm) than the etched hemispheres (ca. 350 nm), presumably due to shrinkage during the drying process, whereas the d-spacing remained the same (460 nm). The membrane also displayed a uniform thickness, which is an advantage of the spin coating technique (Figure 7b). For applications targeting separations, it is necessary to have free-standing membranes and a liftoff process is required. Here, the porous membrane fabricated on Al2O3 could be readily detached through etching in H3PO4 to dissolve the Al2O3 sacrificial layer (Figure 7c). Following the acid treatment, an additional 30% shrinkage was also observed as a result of increased condensation of the silicate. As shown in the DPD simulation, the mesophase was determined by the relative affinities of PPO and PEO blocks with the substrate (εBS − εAS) and the colloidal surface (εBW − εAW). Herein, the affinity was adjusted by using solvents of different polarities in the syntheses. In the lyotropic liquid crystal phase of the P123 surfactant, the solvent will 7489
dx.doi.org/10.1021/la300037q | Langmuir 2012, 28, 7484−7491
Langmuir
■
the cross-sectional images (Figure 8f,g), consistent with previous reports describing the change of direction of the hexagonal mesophase (Figure 8h).35,36 At last, when only water was used as the solvent, a 2D hexagonal dot pattern was observed in a top view image (Figure 8i). The mesopores appeared to be quite uniform, although the compression effect of colloidal confinement was apparent as marked with arrows in Figure 8i. However, from the cross-sectional views, we were unable to identify vertically aligned mesostructures as shown in the simulation (Figures 4b and 5b). Nevertheless, Figure 8j,k shows areas that may be described as twisted columnar structures. It was possible that vertically aligned mesopores were formed initially; however, solvent evaporation prior to the maturation of the sol−gel network led to a decrease in film thickness, and therefore, mesopores became twisted (Figure 8l). Another possibility is that the phase transition was incomplete, as spherical mesopores were also observed. Therefore, tuning solvent polarity was shown to be effective in controlling the mesopore alignment. The DPD simulation suggested the combination of a weak preference of PEO over PPO by the substrate, and a strong preference by the colloidal surface could lead to the vertical columnar mesopore (Figure 4b). Considering that the colloids have a high polarity due to the oxygen plasma treatment, addition of water as the cosolvent can improve the affinity of the PEO phase with the colloids, whereas less polar methanol reduces the relative affinity. While we did not deliberately differentiate the surface properties of the substrate and the colloidal surface, it is apparent that the precursor with water as the sole solvent fulfilled the above requirements and accounted for the unusual mesostructures. We attempted to further evaluate the effects of different solvent polarity in the simulation, but direct tackling of solvent polarity seems to be problematic as it easily leads to solvent phase separation from the PEO block (Figure S8). Additional model refinement is needed to help improve our understanding of this effect.
■
Article
ASSOCIATED CONTENT
S Supporting Information *
Diffusivities of the system components as a function of simulation time, snapshots of the mesostructures as they evolve throughout the simulation, setup of the microscopecoupled UV−vis measurement system and corresponding spectra, zoomed-in images corresponding to Figure 3, simulation of a system with reduced solvent polarity, table of interaction parameters. This material is available free of charge via the Internet at http://pubs.acs.org.
■
AUTHOR INFORMATION
Corresponding Author
*Tel 612-624-1802; e-mail
[email protected]. Present Address †
DuPont Central Research & Development, Rt. 141 and Henry Clay, Wilmington, DE 19880. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS This work was financially supported by the National Science Foundation (DMR-0704312). Parts of this work were carried out in the University of Minnesota Characterization Facility, which receives partial support from the NSF through the MRSEC, ERC, MRI, and NNIN programs. This work benefited from resources at the Minnesota Supercomputing Institute.
■
REFERENCES
(1) Striemer, C. C.; Gaborski, T. R.; McGrath, J. L.; Fauchet, P. M. Charge- and Size-Based Separation of Macromolecules Using Ultrathin Silicon Membranes. Nature 2007, 445, 749−753. (2) Xu, H.; Goedel, W. A. From Particle-Assisted Wetting to Thin Free-Standing Porous Membranes. Angew. Chem., Int. Ed. 2003, 42, 4694−4696. (3) Jackson, E. A.; Hillmyer, M. A. Nanoporous Membranes Derived From Block Copolymers: From Drug Delivery to Water Filtration. ACS Nano 2010, 4, 3548−3553. (4) Walcarius, A.; Sibottier, E.; Etienne, M.; Ghanbaja, J. Electrochemically Assisted Self-Assembly of Mesoporous Silica Thin Films. Nat. Mater. 2007, 6, 602−608. (5) Olson, D. A.; Chen, L.; Hillmyer, M. A. Templating Nanoporous Polymers With Ordered Block Copolymers. Chem. Mater. 2008, 20, 869−890. (6) Li, Y.; Cai, W.; Duan, G. Ordered Micro/Nanostructured Arrays Based on the Monolayer Colloidal Crystals. Chem. Mater. 2008, 20, 615−624. (7) Li, F.; Josephson, D. P.; Stein, A. Colloidal Assembly: The Road From Particles to Colloidal Molecules and Crystals. Angew. Chem., Int. Ed. 2011, 50, 360−388. (8) Etienne, M.; Sallard, S.; Schröder, M.; Guillemin, Y.; Mascotto, S.; Smarsly, B. M.; Walcarius, A. Electrochemical Generation of Thin Silica Films With Hierarchical Porosity. Chem. Mater. 2010, 22, 3426− 3432. (9) Villaescusa, L. A.; Mihi, A.; Rodriguez, I.; Garcia-Bennett, A. E.; Miguez, H. Growth of Mesoporous Materials Within Colloidal Crystal Films by Spin-Coating. J. Phys. Chem. B 2005, 109, 19643−19649. (10) Sel, O.; Sallard, S.; Brezesinski, T.; Rathousky, J.; Dunphy, D. R.; Collord, A.; Smarsly, B. M. Periodically Ordered Meso- and Macroporous SiO2 Thin Films and Their Induced Electrochemical Activity As a Function of Pore Hierarchy. Adv. Funct. Mater. 2007, 17, 3241−3250. (11) Innocenzi, P.; Malfatti, L.; Soler-Illia, G. J. A. A. Hierarchical Mesoporous Films: From Self-Assembly to Porosity with Different Length Scales. Chem. Mater. 2011, 23, 2501−2509.
CONCLUSIONS
In summary, we explored a dual templating approach toward ultrathin hierarchically porous membranes, in which a uniform honeycomb through-hole pattern was cast from a non-closepacked PS hemispherical array, and additional textural porosity was created by a self-assembled block copolymer surfactant. As an endeavor of computation-aided materials design,37 we employed DPD simulations to examine our hypothesis of using colloidal confinement to control the mesopore orientation. Key conclusions were that the relative affinity of the components in the synthesis mixture with the substrate and the colloidal surface dictates the mesopore structure and alignment and that the resulting configuration could extend from the interfaces to the whole structures. Experimentally, hierarchically porous membranes were successfully prepared and different mesopore structures were demonstrated by adjusting the solvent polarity. We expect that the mesoporous structures can be further refined in the future through (1) improved understanding of the mesoscale assembly behavior via DPD simulation and (2) more precise control of the template surface properties and interactions. Meanwhile, we also believe that such hierarchically porous membranes will find unique applications in fields such as high-throughput filtration, protein isolation, high-temperature separation, etc. 7490
dx.doi.org/10.1021/la300037q | Langmuir 2012, 28, 7484−7491
Langmuir
Article
(12) Li, F.; Wang, Z.; Ergang, N. S.; Fyfe, C. A.; Stein, A. Controlling the Shape and Alignment of Mesopores by Confinement in Colloidal Crystals: Designer Pathways to Silica Monoliths With Hierarchical Porosity. Langmuir 2007, 23, 3996−4003. (13) Li, F.; Stein, A. Functional Composite Membranes Based on Mesoporous Silica Spheres in a Hierarchically Porous Matrix. Chem. Mater. 2010, 22, 3790−3797. (14) Platschek, B.; Keilbach, A.; Bein, T. Mesoporous Structures Confined in Anodic Alumina Membranes. Adv. Mater. 2011, 23, 2395−2412. (15) Zhang, J.; Li, Y.; Zhang, X.; Yang, B. Colloidal Self-Assembly Meets Nanofabrication: From Two-Dimensional Colloidal Crystals to Nanostructure Arrays. Adv. Mater. 2010, 22, 4249−4269. (16) Jiang, P.; McFarland, M. J. Large-Scale Fabrication of Wafer-Size Colloidal Crystals, Macroporous Polymers and Nanocomposites by Spin-Coating. J. Am. Chem. Soc. 2004, 126, 13778−13786. (17) Jiang, P.; McFarland, M. J. Wafer-Scale Periodic Nanohole Arrays Templated From Two-Dimensional Nonclose-Packed Colloidal Crystals. J. Am. Chem. Soc. 2005, 127, 3710−3711. (18) Sun, F.; Cai, W.; Li, Y.; Cao, B.; Lei, Y.; Zhang, L. MorphologyControlled Growth of Large-Area Two-Dimensional Ordered Pore Arrays. Adv. Funct. Mater. 2004, 14, 283−288. (19) Ryjkina, E.; Kuhn, H.; Rehage, H.; Mueller, F.; Peggau, J. Molecular Dynamic Computer Simulations of Phase Behavior of NonIonic Surfactants. Angew. Chem., Int. Ed. 2002, 41, 983−986. (20) Nicolaides, D. Mesoscale Modelling. Mol. Simul. 2001, 26, 51− 72. (21) Feng, J.; Liu, H.; Hu, Y. Mesophase Separation of Diblock Copolymer Confined in a Cylindrical Tube Studied by Dissipative Particle Dynamics. Macromol. Theory Simul. 2006, 15, 674−685. (22) Huang, J.; Wang, Y.; Laradji, M. Flow Control by Smart Nanofluidic Channels: A Dissipative Particle Dynamics Simulation. Macromolecules 2006, 39, 5546−5554. (23) Chen, S.; Guo, C.; Hu, G. H.; Liu, H. Z.; Liang, X. F.; Wang, J.; Ma, J. H.; Zheng, L. Dissipative Particle Dynamics Simulation of Gold Nanoparticles Stabilization by PEO-PPO-PEO Block Copolymer Micelles. Colloid Polym. Sci. 2007, 285, 1543−1552. (24) Zhao, Y.; Chen, X.; Yang, C.; Zhang, G. Mesoscopic Simulation on Phase Behavior of Pluronic P123 Aqueous Solution. J. Phys. Chem. B 2007, 111, 13937−13942. (25) Yuan, S. L.; Zhang, X. Q.; Chan, K. Y. Effects of Shear and Charge on the Microphase Formation of P123 Polymer in the SBA-15 Synthesis Investigated by Mesoscale Simulations. Langmuir 2009, 25, 2034−2045. (26) Lyakhova, K. S.; Horvat, A.; Zvelindovsky, A. V.; Sevink, G. J. A. Dynamics of Terrace Formation in a Nanostructured Thin Block Copolymer Film. Langmuir 2006, 22, 5848−5855. (27) Groot, R. D.; Madden, T. J. Dynamic Simulation of Diblock Copolymer Microphase Separation. J. Chem. Phys. 1998, 108, 8713− 8724. (28) Cao, X.; Xu, G.; Li, Y.; Zhang, Z. Aggregation of Poly(Ethylene Oxide)-Poly(Propylene Oxide) Block Copolymers in Aqueous Solution: DPD Simulation Study. J. Phys. Chem. A 2005, 109, 10418−10423. (29) Holland, B. T.; Blanford, C. F.; Do, T.; Stein, A. Synthesis of Highly Ordered, Three-Dimensional, Macroporous Structures of Amorphous or Crystalline Inorganic Oxides, Phosphates, and Hybrid Composites. Chem. Mater. 1999, 11, 795−805. (30) Groot, R. D.; Warren, P. B. Dissipative Particle Dynamics: Bridging the Gap Between Atomistic and Mesoscopic Simulation. J. Chem. Phys. 1997, 107, 4423−4435. (31) Sevink, G. J. A.; Zvelindovsky, A. V. Block Copolymers Confined in a Nanopore: Pathfinding in a Curving and Frustrating Flatland. J. Chem. Phys. 2008, 128, 084901−1−084901/16. (32) Aksay, I. A.; Trau, M.; Manne, S.; Honma, I.; Yao, N.; Zhou, L.; Fenter, P.; Eisenberger, P. M.; Gruner, S. M. Biomimetic Pathways for Assembling Inorganic Thin Films. Science 1996, 273, 892−898. (33) Koganti, V. R.; Dunphy, D.; Gowrishankar, V.; McGehee, M. D.; Li, X.; Wang, J.; Rankin, S. E. Generalized Coating Route to Silica and
Titania Films With Orthogonally Tilted Cylindrical Nanopore Arrays. Nano Lett. 2006, 6, 2567−2570. (34) Cagnol, F.; Grosso, D.; Soler-Illia, G. J.; Crepaldi, E. L.; Babonneau, F.; Amenitsch, H.; Sanchez, C. Humidity-controlled Mesostructuration in CTAB-templated Silica Thin Film Processing. The Existance of a Modulable Steady State. J. Mater. Chem. 2003, 13, 61−66. (35) Huinink, H. P.; Brokken-Zijp, J. C. M.; van Dijk, M. A.; Sevink, G. J. A. Asymmetric Block Copolymers Confined in a Thin Film. J. Chem. Phys. 2000, 112, 2452−2462. (36) Yu, B.; Sun, P.; Chen, T.; Jin, Q.; Ding, D.; Li, B.; Shi, A. C. Confinement-Induced Novel Morphologies of Block Copolymers. Phys. Rev. Lett. 2006, 96, 138306−1−138306/4. (37) The White House, Materials Genome Initiative for Global Competitiveness, 2011.
7491
dx.doi.org/10.1021/la300037q | Langmuir 2012, 28, 7484−7491
