Photopatterned Nanoporosity in Polyelectrolyte Multilayer Films

Mar 5, 2008 - Cuauhtémoc Pozos Vázquez , Thomas Boudou , Virginie Dulong , Claire Nicolas , Catherine Picart and Karine Glinel. Langmuir 2009 25 (6), ...
0 downloads 0 Views 499KB Size
5172

Langmuir 2008, 24, 5172-5178

Photopatterned Nanoporosity in Polyelectrolyte Multilayer Films Solar C. Olugebefola, William A. Kuhlman, Michael F. Rubner, and Anne M. Mayes* Department of Materials Science and Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139 ReceiVed December 16, 2007. In Final Form: January 16, 2008 We report on spatial control of nanoporosity in polyelectrolyte multilayer (PEM) films using photopatterning and its effects on film optical and adsorption properties. Multilayers assembled from poly(acrylic acid-ran-vinylbenzyl acrylate) (PAArVBA), a photo-cross-linking polymer, and poly(allylamine hydrochloric acid) (PAH) were patterned using ultraviolet light followed by immersion in low pH and then neutral pH solutions to induce nanoporosity in unexposed regions. Model charged small molecules rhodamine B, fluorescein, and propidium iodide and the model protein albumin exhibit increased adsorption to nanoporous regions of patterned PEM films as shown by fluorescence microscopy and radiolabeling experiments. Films assembled with alternating stacks of PAH/poly(sodium-4-styrene sulfonate) (SPS), which do not become nanoporous, and stacks of PAH/PAArVBA were patterned to create nanoporous capillary channels. Interdigitated channels demonstrated simultaneous, separate wicking of dimethyl sulfoxide-solvated fluorescein and rhodamine B. In addition, these heterostack structures exhibited patternable Bragg reflectivity of greater than 25% due to refractive index differences between the nanoporous and nonporous stacks. Finally, the PEM assembly process coupled with photo-cross-linking was used to create films with two separate stacked reflective patterns with a doubling in reflectivity where patterns overlapped. The combined adsorptive and reflective properties of these films hold promise for applications in diagnostic arrays and therapeutics delivery.

Introduction The field of layer-by-layer (LbL) film assembly is well established and constantly expanding to incorporate new materials to achieve novel functionalities. The patterning of multilayer films to spatially control their chemical and physical properties has been one of the important areas of exploration within this field in the search for new applications. Film patterning has been demonstrated through a variety of techniques. Polymer-onpolymer stamping1,2 has been used to deposit patterned single polymer layers atop preassembled films to control the deposition of colloids and magnetic particles3-5 and direct cell adhesion.6-8 This versatile approach has also been used to deposit or transfer fully assembled films to new substrates.9,10 Other techniques to control the deposition of LbL films have been demonstrated, such as allowing assembly solutions to flow through microfluidic channels,11,12 using dip-pen nanolithography to “write” base layers for further assembly,13 and the use of patterned sacrificial * Corresponding author. E-mail: [email protected]. (1) Jiang, X.; Zheng, H.; Gourdin, S.; Hammond, P. T. Langmuir 2002, 18, 2607-2615. (2) Berg, M. C.; Choi, J.; Hammond, P. T.; Rubner, M. F. Langmuir 2003, 19, 2231-2237. (3) Zheng, H.; Lee, I.; Rubner, M. F.; Hammond, P. T. AdV. Mater. 2002, 14, 569-572. (4) Zheng, H.; Rubner, M. F.; Hammond, P. T. Langmuir 2002, 18, 45054510. (5) Lyles, B. F.; Terrot, M. S.; Hammond, P. T.; Gast, A. P. Langmuir 2004, 20, 3028-3031. (6) Kim, H.; Doh, J.; Irvine, D. J.; Cohen, R. E.; Hammond, P. T. Biomacromolecules 2004, 5, 822-827. (7) Berg, M. C.; Yang, S. Y.; Hammond, P. T.; Rubner, M. F. Langmuir 2004, 20, 1362-1368. (8) Salloum, D. S.; Olenych, S. G.; Keller, T. C. S.; Schlenoff, J. B. Biomacromolecules 2005, 6, 161-167. (9) Park, J.; Hammond, P. T. AdV. Mater. 2004, 16, 520-525. (10) Lee, N. Y.; Lim, J. R.; Lee, M. J.; Park, S.; Kim, Y. S. Langmuir 2006, 22, 7689-7694. (11) Shchukin, D. G.; Kommireddy, D. S.; Zhao, Y.; Cui, T.; Sukhorukov, G. B.; Lvov, Y. M. AdV. Mater. 2004, 16, 389-393. (12) Reyes, D. R.; Perruccio, E. M.; Becerra, S. P.; Locascio, L. E.; Gaitan, M. Langmuir 2004, 20, 8805-8811. (13) Lee, S. W.; Sanedrin, R. G.; Oh, B.-K.; Mirkin, C. A. AdV. Mater. 2005, 17, 2749-2753.

photoresist sublayers.14,15 Photolithography methods for patterning polyelectrolyte multilayer films have also been explored and have been primarily focused on cross-linking for selective etching, such as the extensive work done using diazo resins.16-20 Other multilayer photo-cross-linking systems have been demonstrated, such as through the addition of hydroxybenzoyl functional groups to PAA for hydrogen-bonded multilayer films21 and a benzophenone-modified poly(acrylic acid) (PAA), poly(allylamine hydrochloride) (PAH) pair used to stabilize multilayer colloidal shells against dissolution.22,23 Recently, we devised a similar approach of coupling a photoactive group to PAA to take advantage of the pH-based structural control offered by a weak polyelectrolyte to make polyelectrolyte multilayer films with patterned micro- and nanoporosity.24 This was accomplished through the synthesis of poly(acrylic acid-ran-vinylbenzyl acrylate) (PAArVBA) and its incorporation in place of PAA in multilayer assemblies.24 The photoactive vinylbenzyl group acted to cross-link the film in areas exposed to ultraviolet light, preventing the normally observed porosity transition.25 Nanoporous multilayer films have been shown to exhibit significantly different properties from those of dense films, including a greater retention of guest molecules and the ability (14) Cui, T.; Hua, F.; Lvov, Y. Sens. Actuators, A 2004, 114, 501-504. (15) Cho, J.; Jang, H.; Yeom, B.; Kim, H.; Kim, R.; Kim, S.; Char, K.; Caruso, F. Langmuir 2006, 22, 1356-1364. (16) Cao, S.; Zhao, C.; Cao, W. Polym. Int. 1998, 45, 142-146. (17) Sun, J.; Cheng, L.; Liu, F.; Dong, S.; Wang, Z.; Zhang, X.; Shen, J. Colloids Surf. A: Physicochemical and Engineering Aspects 2000, 169, 209217. (18) Shi, F.; Dong, B.; Qiu, D.; Sun, J.; Wu, T.; Zhang, X. AdV. Mater. 2002, 14, 805-809. (19) Shi, F.; Wang, Z.; Zhao, N.; Zhang, X. Langmuir 2005, 21, 1599-1602. (20) Sun, J.; Wu, T.; Liu, F.; Wang, Z.; Zhang, X.; Shen, J. Langmuir 2000, 16, 4620-4624. (21) Yang, S. Y.; Rubner, M. F. J. Am. Chem. Soc. 2002, 124, 2100-2101. (22) Park, M.-K.; Deng, S.; Advincula, R. C. J. Am. Chem. Soc. 2004, 126, 13723-13731. (23) Park, M.-K.; Deng, S.; Advincula, R. C. Langmuir 2005, 21, 5272-5277. (24) Olugebefola, S. C.; Ryu, S.-W.; Nolte, A. J.; Rubner, M. F.; Mayes, A. M. Langmuir 2006, 22, 5958-5962. (25) Mendelsohn, J. D.; Barrett, C. J.; Chan, V. V.; Pal, A. J.; Mayes, A. M.; Rubner, M. F. Langmuir 2000, 16, 5017-5023.

10.1021/la703936p CCC: $40.75 © 2008 American Chemical Society Published on Web 03/05/2008

Photopatterned Nanoporosity in Multilayer Films

to wick solvent from a standing reservoir.26 In addition, porous multilayer films demonstrate significantly lower refractive indices from dense films,27 and films containing alternating dense and porous stacks have shown optical characteristics of Bragg reflectors.28 Here we further explore the potential offered by photopatterning of PEM films for making distinct nanoporous regions within the film at controlled depths. This property is examined with respect to guest molecule adsorption, including model protein adsorption and the formation of capillary structures for the simultaneous loading of two model dyes. In addition, we exploit the refractive index differences between nanoporous and nonporous regions to pattern overall film optical properties. Experimental Section Materials. Vinylbenzyl chloride, sodium iodide, poly(allylamine hydrochloride) (PAH, 70 000 g/mol), triethylamine (TEA), potassium iodate, poly(sodium-4-styrene sulfonate) (SPS) 1,8-diazobicyclo[5.4.0]undec-7-ene (DBU), dimethyl formamide (DMF), anhydrous dimethyl sulfoxide (DMSO), tetrahydrofuran (THF), fluorescein isothiocyanate-labeled bovine serum albumin (FITC-BSA, SigmaAldrich), potassium iodide (Mallinckrodt), poly(acrylic acid) (PAA, 90 000 g/mol in 25 wt % aqueous solution (Polysciences), sodium tetraborate decahydrate, fluorescein sodium salt, rhodamine B, propidium iodide (Fluka), deuterated dimethyl sulfoxide (DMSOd6) (Cambridge Isotope Labs), iodo beads, D-salt polyacrylamide columns (Pierce), and iodine-125 radionuclide (Perkin-Elmer) were used as received. All aqueous solutions were prepared using Milli-Q Academic water (Millipore) with a total organic content (TOC) of 515 nm) or red (>590 nm) light. Images were processed using IGOR (Wavemetrics, Inc.) software.

Results and Discussion Controlling Nanoporosity in PAH 8.5/PAArVBA 3.5 Films. As previously observed, subjecting PAH 8.5/PAArVBA 3.5 films to low pH/neutral pH treatment results in the formation of nanopores, which leads to an increase in the film thickness and a corresponding decrease in the refractive index.27 In addition, the pores, once formed, can be closed and opened by alternately immersing the films in pH 2.3 (closed) and pH 5.5-7 (open) solutions. This pseudoreversible morphological transformation may be explained by the swelling and collapse of the nanoporous network as the environmental pH is cycled above and below the pKa of the carboxylic acid groups (pH 2.5) within the film. This is observed in Figure 2, which shows thickness and refractive (30) Markwell, M. A. K. Anal. Biochem. 1982, 125, 427-432.

Olugebefola et al.

Figure 2. Thickness and refractive index measurements for asbuilt (empty symbols) and cross-linked (filled symbols) films after the nanopore opening (pH 5.5) and closing (pH 2.3) immersion steps.

Figure 3. (a) Microscopy image and (b) height profile of the UVpatterned nanoporous (PAH 8.5/PAArVBA 3.5)10.5 film. A deliberate scratch was made to determine the overall film thickness.

index values for a control (as-built) film cycled in this manner (empty symbols). The as-built films increase in thickness/porosity after each immersion at higher pH but return to their initial thickness after each immersion at pH 2.3. By contrast, films exposed to 254 nm light for 15 min show negligible changes in their thickness and refractive index with pH cycling (filled symbols). Exposure of these films to UV radiation results in the formation of covalent cross-links that are resistant to morphological rearrangement. This difference in behavior can be used to create patterned porous films through selective UV exposure. Shown in Figure 3 are an optical micrograph and an associated height profile of a (PAH 8.5/PAArVBA 3.5)10.5 film built on glass, patterned using a mask with a series of parallel lines. The film thickness (∼150 nm) determined from the scratch is consistent with that of PAH 8.5/PAA 3.5 films, showing similar building behavior of PAArVBA to that of PAA. The unexposed lines are clearly taller as a result of the increase in thickness from pore formation. The decrease in refractive index coupled with the change in

Photopatterned Nanoporosity in Multilayer Films

Langmuir, Vol. 24, No. 9, 2008 5175

Figure 5. 125I-labeled bovine serum albumin adsorption to (PAH 8.5/PAArVBA 3.5)10.5 films after various treatments (as assembled, nanoporous and then thermally cross-linked at 180 °C, thermally cross-linked only at 180 °C, and UV cross-linked) and a fluorescence image of a nanoporous, thermally cross-linked film after the adsorption of FITC-BSA (inset). Figure 4. Fluorescence images of UV-patterned nonporous (lefthand) and porous (right-hand) films after the adsorption of (top) fluorescein (negatively charged), (middle) rhodamine B (zwitterionic), and (bottom) propidium iodide (positively charged). The scale bar is 200 µm.

thickness also results in a change in film color in the porous regions. Guest Molecule Adsorption into Patterned Nanoporous PAH/PAArVBA Films. A previous study of hydrophobic drug adsorption and release in PAH 8.5/PAA 3.5 films found a significant increase in the degree of uptake of ketoprofen (negatively charged) and cyctochalasin D (uncharged) into porous versus nonporous films when adsorbed from DMSO.26 In that study, the increased adsorption was thought to be due to increased surface area and possibly hydrophobic surface association. Here, to elucidate the effect of charge on adsorption to porous PEM structures, variously charged dyes were adsorbed on UV-patterned films before and after the porosity and thermal cross-linking treatments were carried out in PBS solution (Figure 4). Fluorescence images of UV-patterned, nonporous control films are shown on the left side of Figure 4, and patterned, porous films are shown on the right. For each dye, the same light exposure conditions were used for images of porous and control films. As expected, the porous patterned films show greater adsorption of all dyes into the porous regions of the film irrespective of dye charge, attributable to the increased effective surface area in these film regions. For nonporous control films, masked film regions also show either similar (fluorescein) or stronger (rhodamine B, propidium iodide) dye uptake than do UV-exposed areas. A previous examination of UV-patterned PAH/PAArVBA films showed higher swelling in such unexposed film regions,24 and swelling has been clearly correlated with the degree of smallmolecule adsorption in weak PEM films.31 Interestingly, both fluorescein (negatively charged) and rhodamine B (zwitterionic) show overall higher adsorption by control films whereas propidium iodide (positively charged) is adsorbed more strongly by porous films. The increased adsorption of the positively charged dye molecule suggests that after the porosity treatment and thermal cross-linking, film surfaces are negatively charged. The ability of these porous films to pattern adsorption regardless of charge suggests their broad utility in applications involving molecule capture or storage. Porous regions of patterned films also show a greater adsorption of the model protein bovine serum albumin (BSA), which is a significantly larger molecule. Shown in the inset of Figure 5 is

a fluorescence micrograph of a (PAH 8.5/PAArVBA 3.5)10.5 film on glass after UV patterning, exposure to five low pH/ neutral pH cycles, and thermal cross-linking, followed by 24 h of immersion in a solution of 1.0 mg/mL FITC-labeled BSA (FITC-BSA) in PBS. The difference between the two regions is apparent. A quantitative comparison of adsorption using radiolabeled BSA under the same conditions shows a 3- to 5-fold enhancement of adsorption for thermally cross-linked porous samples over that of as-built and UV-exposed control samples, as seen in Figure 5. (Note that porous, non-cross-linked films can structurally evolve with extended immersion in aqueous solution,25 making it difficult to obtain reliable values for their protein adsorption, and thus were not included as a control.) For thermally cross-linked porous samples, the surface coverage value, Γ, of 25 mg/m2 is well above that predicted theoretically for an albumin monolayer on a flat surface (4.5 mg/m2),32 consistent with the increase in effective surface area due to the creation of nanopores. An estimate of the pore size can be made using a purely geometric argument. Assuming that the pores are roughly tubular, the change in film volume per unit area is equivalent to the change in film thickness and can be written as

∆V ) ∆T ) Fπr2l A

(1)

where V is the volume, A is the area, T is the thickness, r is the pore radius, l is the tube length, and F is the number of pores per unit area. The surface area of each pore is 2πrl, leading to an expression for the change in surface area per unit area of

∆SA ) 2

∆T r

(2)

The pore radius can be determined using the value of 25 nm for ∆T taken from profilometry and ∆SA extrapolated from the ratio of 125I-BSA adsorption between porous and nonporous thermally cross-linked samples. This leads to an internal surface area per unit area of 5.2 and a rough estimate of the pore radius of ∼9.6 nm. Previous studies of albumin adsorption to strongly charged multilayer films such as SPS/PAH have suggested that although the surface charge plays a role in the degree of adsorption a minimum level of adsorption is always observed.33,34 Although (31) Chung, A. J.; Rubner, M. F. Langmuir 2002, 18, 1176-1183. (32) Fainerman, V. B.; Miller, R. Langmuir 1999, 15, 1812-1816. (33) Salloum, D. S.; Schlenoff, J. B. Biomacromolecules 2004, 5, 1089-1096.

5176 Langmuir, Vol. 24, No. 9, 2008

Olugebefola et al.

Figure 6. Creating nanoporous channels in a heterostack multilayer film. (a) Schematic of the patterning process for channel formation. (b) Experimental setup for the capillary adsorption of dye-loaded DMSO into a patterned film. (c) Fluorescence images of a film loaded with fluorescein and rhodamine B using a (top) green bandpass filter and (bottom) a red bandpass filter.

our system is different in composition, the quantities of albumin adsorbed by our porous films without attempts at optimization are comparable to the adsorption obtained by similar thickness SPS/PAH films33 or highly swollen heterostructured films composed of SPS/PAH and poly(glutamic acid)/poly(L-lysine) bilayers.35 The results are promising for therapeutics delivery, but studies of protein activity and release remain to be carried out. Previously, Berg et al. demonstrated that (PAH 4.0/SPS 4.0)(PAH 8.5/PAA 3.5)-(PAH 4.0/SPS 4.0) heterostack films exhibit capillarity when in contact with DMSO solutions, allowing the loading of small molecules through wicking into the porous middle stack.26 This property is potentially useful because DMSO is a popular solvent for hydrophobic drugs, making these structures interesting candidates for controlled drug delivery. In that study, the wicking was visualized by the change in refractive index of the heterostack film. What was not established was whether DMSO was solely traveling through the porous stack or whether it was also diffusing through the PAH/SPS stacks. Here we sought to determine if dyes could be loaded by capillary action into porous heterostack films incorporating PAArVBA, patterned to create channel structures. Single patterned porous heterostack films were assembled and processed as described previously with a pattern designed to create capillary channels beginning from opposite ends of each sample, terminating at the middle in close proximity, but not directly connected. After simultaneous wicking of fluorescein and rhodamine B dissolved in DMSO, fluorescence images of a sample film with red and green bandpass filters, shown in Figure 6c, reveal the location of both dyes. If the DMSO and dye were diffusing through the PAH/SPS stacks, then one would (34) Ladam, G.; Gergely, C.; Senger, B.; Decher, G.; Voegel, J.-C.; Schaaf, P.; Cuisinier, F. J. G. Biomacromolecules 2000, 1, 674-687. (35) Gergely, C.; Bahi, S.; Szalontai, B.; Flores, H.; Schaaf, P.; Voegel, J.-C.; Cuisinier, F. J. G. Langmuir 2004, 20, 5575-5582.

expect to see a uniform coverage of dye. If the solutions were diffusing throughout all of the PAH/PAArVBA stack, then one would expect to see dye adsorption akin to the immersion case shown previously, with all porous regions showing strong adsorption and intermixing of both dyes. Instead, the two dyes remain separated in their respective channels, indicating that the DMSO is confined within the porous film regions. The lack of either dye in the unconnected center diamonds of the pattern (compare with the adsorption pattern in Figure 4) further confirms this behavior. This capability could be used for the selective loading of multiple different therapeutic molecules within the same film through separate channels. In this way, incompatible molecules or molecules with a high reactivity to each other could be isolated on the same surface for later release and interaction. Patterned Nanoporous Bragg Reflectors. As noted previously, Zhai et al. studied the optical properties of films constructed from PAH 8.5/PAA 3.5 stacks sandwiched between PAH/SPS stacks.28 The outer stacks remain fully charged and dense when films are subjected to the low pH/high pH nanoporosity treatment. Porosity induced in the middle PAH 8.5/PAA 3.5 stack thus yields the requisite high/low/high refractive index profile normal to the film plane to achieve Bragg-type light reflection. Through lateral control over nanoporosity with UV crosslinking, the location of the Bragg reflection in the film plane can be controlled.24 To illustrate this, we assembled (PAH 4.0/SPS 4.0)50-(PAH 8.5/PAArVBA 3.5)5.5-(SPS 4.0/PAH 4.0)50.5 (one A stack) films on glass, UV patterned them, and subjected them to five low pH/neutral pH cycles to induce nanoporosity. Figure 7 shows an optical micrograph of the sample film after this treatment. The nanoporous regions reflect in the yellow region of the spectrum, which is consistent with the thickness of each stack as determined through profilometry. For this one A stack, the maximum reflectivity is more than 25% at 547 nm, whereas

Photopatterned Nanoporosity in Multilayer Films

Langmuir, Vol. 24, No. 9, 2008 5177

Figure 7. Reflective heterostack multilayer film. (a) Optical micrograph of a patterned one-A-stack film with nanoporous regions appearing in yellow. (b) Profile of a one-A-stack film with a deliberate scratch to determine the film thickness.

for the cross-linked regions of the film, the maximum reflectivity of 11% is the same as that for the as-built film, indicating that there is no detectable porosity generated in those regions. One of the advantages of PEM assembly is the ability to build additional layers atop previously built or processed films. With the controlled porosity of PAArVBA-containing Bragg films, additional stacks can be assembled and patterned above an already patterned porous film, yielding regions with different degrees of reflectivity depending on the local refractive index profile. One A-stack films were constructed as (PAH 4.0/SPS 4.0)50(PAH 8.5/PAArVBA 3.5)5.5-(SPS 4.0/PAH 4.0)50.5 on glass and then UV patterned, subjected to the porosity treatment, and thermally cross-linked at 180 °C for 4 to 5 h to prevent further evolution of the porous structure during subsequent processing. After this, additional layers of (PAH 8.5/PAArVBA 3.5)5.5-(SPS 4.0/PAH 4.0)50.5 were assembled onto the patterned film, followed by a second UV patterning step and pore-inducing treatment to create a pseudo-two-A-stack film. Figure 8 shows an example of a two-pattern film with the upper and lower porous regions overlapping. The reflectivity of the overlapping regions is clearly higher than that for the two separate porous regions. UV-vis reflectance measurements of different regions of the patterned film (Figure 9) confirm the overlap region to have almost twice the maximum reflectivity of film regions containing only the upper or lower porous stack. Profilometry of the film surface reveals the height elevation for the singly porous regions and for the overlap region (Figure 8c). The lower porous stack regions exhibit slightly greater thickness increases compared with the upper porous regions (69 ( 3 nm vs 57 ( 5 nm), though the cause is not known. (It should be noted that long scan lengths such as that shown in Figure 8c, taken in order to display all regions of the pattern, showed some slope distortion near the end of the scan. Shorter scans showed both sides of the patterned film to be relatively flat.) The ability to control the thickness and depth placement of porous regions using the multilayer assembly technique, coupled with the lateral control over porosity through photopatterning, gives a unique, facile, inexpensive method of creating reflective

Figure 8. (a) Photograph of a dual-patterned Bragg reflector film. (b) Microscopy image of the central pattern. (c) Surface profile showing elevated single patterns and central overlap regions.

Figure 9. Reflectivity of distinct regions of a dual-patterned Bragg reflector film with the same structure as that of the sample shown in Figure 8.

patterns. By controlling the number of layers of each stack deposited,24,28 each patterned stack could potentially reflect at a different wavelength, allowing for pattern-specific optical filtering.

5178 Langmuir, Vol. 24, No. 9, 2008

Conclusions Photopatterning of polyelectrolyte multilayer films incorporating the weak PAArVBA polyelectrolyte allows control over the location of induced nanoporous regions within the film plane. In this study, the significant difference in adsorption properties between porous and nonporous regions was used to pattern the molecular uptake of small-molecule dyes as well as the model protein albumin. The adsorption of other molecules (e.g., therapeutics) should give similar results because increased adsorption in the nanoporous regions appears to be largely due to the increased interfacial area. The sandwiching of PAH/PAArVBA stacks between strongly charged PEM stacks allows the creation of nanoporous channels that enable the loading of multiple different molecule types into separate reservoirs by capillary action. In addition, differences in refractive index between the nanoporous and dense stacks

Olugebefola et al.

enable the creation of additive, optically reflective patterns within assembled films and control over the maximum reflective magnitude and wavelength of those patterns by specifying the number and thickness of the stacks. Such a capability could find application in signal amplification or transduction in bioassays and diagnostic arrays. Acknowledgment. We thank Professor Linda Griffith for the use of her laboratory in conducting radiolabeled albumin experiments. This work made use of shared experimental facilities under the MRSEC Program of the National Science Foundation under award DMR-0213282. This work was supported by the MRSEC Program of the National Science Foundation under award DMR-0213282. S.C.O. acknowledges support from the Lemelson-MIT Foundation. LA703936P