Colloids with high-definition surface structures

Abstract

Compared with the well equipped arsenal of surface modification methods for flat surfaces, techniques that are applicable to curved, colloidal surfaces are still in their infancy. This technological gap exists because spin-coating techniques used in traditional photolithographic processes are not applicable to the curved surfaces of spherical objects. By replacing spin-coated photoresist with a vapor-deposited, photodefinable polymer coating, we have now fabricated microstructured colloids with a wide range of surface patterns, including asymmetric and chiral surface structures, that so far were typically reserved for flat substrates. This high-throughput method can yield surface-structured colloidal particles at a rate of ≈10⁷ to 10⁸ particles per operator per day. Equipped with spatially defined binding pockets, microstructured colloids can engage in programmable interactions, which can lead to directed self-assembly. The ability to create a wide range of colloids with both simple and complex surface patterns may contribute to the genesis of previously unknown colloidal structures and may have important technological implications in a range of different applications, including photonic and phononic materials or chemical sensors.

Complex colloidal structures have been a major focus of fundamental and applied research because of their potential applications in photonic (1–6) or phononic (7) band gap materials, chemical sensors (8), or data storage devices (9). Many of these applications will require colloids to be organized in nontrivial colloidal crystal structures with exquisite lattice periodicity, such as the diamond lattice (10). In principle, cocrystallization of binary mixtures of oppositely charged particles can form a range of unusual colloidal crystals (11–16). An alternate approach relies on the use of colloids that have spatially defined binding patches to encode and ultimately direct lattice organization (10). Although the potential merits of such an approach, e.g., homogenous colloidal assemblies and potentially superior optical properties, have been widely recognized (10, 17), successful experimental implementation has been hampered by the limited availability of suitable surface modification protocols for colloidal particles (18–21). This limited availability of microstructuring techniques stands in clear contrast to the sizable number of patterning processes that have been developed for flat surfaces (22, 23), contributing to major technological breakthroughs in electronics and biotechnology (24).

Recently, chemical vapor deposition (CVD) polymerization, a solventless coating process, which relies on the deposition of reactive coatings made of functionalized poly-p-xylylenes, has been developed as a flexible surface modification approach (25). Reactive coatings have been used to prepare polymer films with a wide range of functional groups, such as photoreactive groups, on various substrate materials and geometries (25–29). Most importantly, the CVD technology is not intrinsically limited to flat substrates, but can, at least in principle, be equally applied to curved surfaces.

Results and Discussion

To generate microstructured colloidal particles, we pursued a route that uses solventless deposition of a photoreactive polymer coating onto colloidal particles followed by light-directed surface modification. As shown in Fig. 1, microstructured colloidal particles were fabricated by a two-step procedure: (i) coating of the colloids with a photodefinable polymer, poly[4-benzoyl-p-xylylene-co-p-xylylene], via CVD polymerization (26, 28) and (ii) spatially controlled surface reaction of the photoreactive coatings using a highly parallel projection lithographic patterning step. The rationale for selecting poly[4-benzoyl-p-xylylene-co-p-xylylene] as a base coating was that this CVD-based reactive polymer coating forms homogeneous interlayers on curved surfaces while providing photoreactive chemical groups for further surface modification (25, 26).

Fig. 1.

Schematic description of the 3D microstructuring technique. The method comprises two process steps: deposition of the photodefinable CVD coating (step 1) and subsequent projection lithographic rendering of the polymer-coated colloids (step 2).

In step i, i.e., the deposition of a thin photoreactive coating on the surface of the colloids, the starting material, 4-benzoyl[2.2]paracyclophane, is polymerized without the need of solvent, initiator, or any additional additives (25). In a series of initial experiments, reaction conditions were adjusted to ensure that the polymerization shown in Fig. 1 proceeded with a growth rate of ≈0.5 Å/s, as estimated based on in situ quartz crystal microbalancing analysis. Ellipsometric analysis revealed that polymer films with thicknesses of 60–80 nm were deposited under these conditions. With respect to subsequent modification steps, it was important that the photoreactive coatings were chemically well defined and could be deposited without significant loss of functionality. The chemical structure of the deposited polymer films was examined with a combination of adsorption/reflection FTIR and x-ray photoelectron (XPS) spectroscopy and was found to be in accordance with the predicted polymer structure of poly[4-benzoyl-p-xylylene-co-p-xylylene] as shown in Fig. 1. Specifically, the FTIR spectra revealed the characteristic carbonyl stretches at 1,604 and 1,663 cm⁻¹, whereas XPS survey spectra [see supporting information (SI) Fig. 5] indicated an average chemical composition of 95.5 atom-% carbon and 4.5 atom-% oxygen, which is in close agreement with the theoretically obtained values of 95.8 atom-% for carbon and 4.2 atom-% for oxygen (26, 28). The high-resolution C_1s XPS spectrum of the polymer film provided further information regarding the chemical fine-structure of the polymer films. The C_1s-spectrum shows characteristic signals of aliphatic and aromatic carbon atoms (C C, CH, 285.0 eV), carbon atoms in α-position to the carbonyl group (CCO, 285.9 eV), carbonyl carbon (CO, 286.1 eV), as well as a π→π* satellite signal at 291.5 eV, which is characteristic for aromatic poly-p-xylylene polymers (29). It is known from similar systems that the photoactivated carbonyl groups of the polymer are metastable and rapidly react via insertion into CH or NH bonds (30). This insertion reaction can be exploited for covalent immobilization of adjunct molecules onto illuminated surface areas.

Once the deposition of the photoreactive coatings on colloidal particles has been demonstrated, spatially directed microstructuring becomes achievable. To obtain spatially controlled surface patches on colloids, we selectively illuminated certain areas of previously coated colloids with light at 365 nm by using a high-throughput projection technique that has been previously used for in situ synthesis of peptides and DNA on microarrays (31–33). Programmable patterns were created by using a 1,024- × 768-pixel digital micromirror device (DMD) (Fig. 1). Whereas the entire surface of the colloids was coated with the photoreactive coating during CVD polymerization, only the areas illuminated with the DMD underwent photochemical conversion of the carbon–oxygen double bond from the singlet ground state into the corresponding triplet state (34). To obtain the patterned colloids shown in Fig. 2, a biotin derivative was used as a model molecule because of its established relevance as a biomolecular linker for subsequent biological modification and its high binding affinity toward strepdavidin (35). To immobilize the biotin-based linker, photoreactive colloids were immersed into an aqueous biotin solution during the photopatterning step. The colloids were then selectively activated via spatially directed illumination with the DMD, resulting in exclusive immobilization onto light-exposed surface areas only. To visualize the micropatterns, surface-modified colloids were further incubated with TRITC-labeled strepdavidin, which has been shown to bind to surface areas that present biotin ligands with high affinity (36). Fig. 2 shows typical examples of colloids coated with photoreactive CVD coating and modified via projection lithography. Both, the “UM” patterns shown in Fig. 2 A and B as well as the stripe patterns (Fig. 2C) demonstrate high selectivity of strepdavidin toward active binding patches, i.e., the illuminated areas. The 3D confocal laser scanning microscopy (CLSM) images further reveal high contrast, as indicated by precise contour lines between pattern and unmodified areas, even for the smallest colloids with a radius of 50 μm.

Fig. 2.

High-precision microstructuring of colloids. Three-dimensional CLSM micrographs of designer colloids obtained by spatially controlled photoimmobilization of biotin and subsequent binding of TRITC-strepdavidin. (A and B) UM patterns on polystyrene microspheres with average radii of 200 μm at different magnifications. (C) Stripe patterns on polystyrene microspheres with average radii of 50 μm. (Scale bars, 100 μm.)

Based on these findings, several aspects of the microstructuring process are noteworthy. (i) High pattern fidelity and excellent contrast are observed, suggesting efficient conversion of the photoactive groups in the exposed areas only. (ii) A range of different patterns has been created, demonstrating that patterns can be selected with high flexibility. This procedure is in clear contrast to the currently existing methods for patterning of colloids, which are all subject to constrains imposed by the patterning technique itself. (iii) High-resolution patterns can be created on spheres with different radii. Particle radii of the examples shown in the Fig. 2 range from 50 to 200 μm. Because CVD coatings can be conformally deposited on submicrometer features without loss of topological information (37), the lower resolution of the 3D printing process is not limited by the CVD step but is defined by the photopatterning process and, more specifically, by the maximum resolution of the DMD. In this study, a DMD with a maximum resolution of 27.5- × 27.5-μm² pixels was used. It is important to recognize that this pixel size is not a fundamental limit but rather a limit imposed by the available optics in our laboratories. Commercially available projection lithography instrumentation can result in pixel sizes as small as a few micrometers squared (38, 39), suggesting that the proposed approach will have significant potential for covering a major size range of colloidal particle science ranging from hundreds of micrometers down to the single-digit micrometer regime.

With the selective modification of colloidal surfaces demonstrated, the focus of our study shifted toward fabrication of colloids with controlled binding sites, which can act as precisely designed binding pockets for programmable self-assembly. If colloids were indeed equipped with sticky binding pockets, this could manifest an important step toward the design of building blocks for programmable colloidal self-assembly. For proof-of-concept, we selected one of the simplest geometries for the binding patch: a 55- × 55-μm² square pattern consisting of either biotin or strepdavidin. Two visually distinct groups of equally sized colloids, i.e., polystyrene microspheres (dark) and gold-coated polystyrene microspheres (light), were surface-modified by using the colloidal patterning technique. Before photopatterning, all colloids were coated with a 60-nm-thick film of the photoreactive CVD coating. Subsequently, dark colloids were modified with 55- × 55-μm² square patterns of biotin, whereas light colloids were surface-modified to present strepdavidin (Fig. 3A). Before self-assembly, the presence of the binding patches was confirmed by incubating biotin-modified colloids with TRITC–strepdavidin. As shown in Fig. 3B, well defined, isolated squares can be observed on the surface of the colloids, confirming the successful introduction of the binding patches. Moreover, in accordance with the microstructured colloids shown in Fig. 2, excellent contrast between modified and unmodified areas was observed.

Fig. 3.

Self-organization of microspheres. (A) Micrographs showing self-assembly of gold-coated microspheres (light, streptavidin-modified) with polystyrene microspheres (dark, biotin-modified). (B) Patterns used for studying self-assembly. Green areas denote fluorescence signal detected by CLSM. Orange dotted lines show the perimeter of the colloids and are drawn to guide the eyes. (C) Distribution of self-assembled microspheres; the number on top denotes the P value obtained from a t test, showing that self-assembled AB pairs (first configuration) are significantly different (P = 0.000157) from the control group. (Scale bars, 100 μm.)

For the self-assembly experiments, an equivalent number of colloidal particles from both groups was simultaneously suspended in PBS (pH 7.4) solution for 90 min. During the course of each experiment, the colloids precipitated and self-oriented into a range of different shapes. Based on light microscopy, a total of 123 colloidal aggregates from three independent runs were statistically analyzed. A range of different configurations was identified and colloidal assemblies were categorized in nine subgroups as shown in Fig. 3C. As anticipated, the largest fraction of colloidal assemblies by far comprised the “retrozygous” AB couples consisting of one single biotin-modified polystyrene colloid (A) and one strepdavidin-modified gold colloid (B). The AB fraction was more than twice as large as the next largest fraction consisting of AA couples. This large abundance of the AB configuration suggests the dominance of specific interactions between the complementary biotin and strepdavidin binding patches. In addition to AA assemblies, a number of aggregated structures also were observed during the experiment, which may be attributed to nonspecific interactions. We further confirmed these results with a control experiment, in which colloids that were modified with the photoreactive polymer plus biotin, but lacking strepdavidin, were allowed to self-assemble in the same fashion. Specific biotin/streptavidin interactions were excluded in this experiment. In the control experiment, the AB configurations were not significantly different from AA configurations, indicating the absence of the specific biotin/strepdavidin interactions. Overall, the predominance of AB assemblies was highly significant for the encoded self-assembly as compared with the control group. Although simpler patterns, such as the line pattern shown in Fig. 2C, are, at least in principle, also obtainable by other techniques (18–21), we wanted to demonstrate the applicability of our 3D photopatterning technique to patterns of higher complexity.

To further extend the repertoire of surface structures that can be created on curved colloidal surfaces, structures consisting of two different patterns, each made of a different type of biomolecule, were designed first. As shown in Fig. 4A, we approached this challenge with a routine that involved two consecutive photopatterning cycles. After CVD of the photoreactive coating onto the colloidal surfaces, an initial photopatterning step resulted in a 110- × 137.5-μm green U pattern. These patterns were comprised of TRITC-labeled streptavidin, which was immobilized according to the previously described protocol. In a subsequent step, the red M pattern (110 × 137.5 μm) was formed by light-directed photoimmobilization of an amino-terminated star-polyethylene glycol. The resulting amino-functionalized surface areas supported covalent immobilization of a red-fluorescent atto-655-NHS ester. CLSM was used to examine the resulting colloids in details. As shown in Fig. 4B, two independent microstructures were clearly distinguishable on the colloidal surfaces. Based on their characteristic emission wavelengths, the structures could clearly be attributed to the strepdavidin (green color) and atto-ester (red color), respectively. Despite the additional complexity stemming from the presence of two different surface patches, excellent pattern fidelity, minimal cross-reactivity between different dyes, and excellent particle-to-particle reproduction were observed. Based on a relatively simple cascade of process steps, we were not only able to demonstrate the accessibility of precisely controlled, complex surface patches but also to establish a simple access route to multifunctional colloids. This method lends itself to the simple fabrication of colloids with unusual surface patterns, including asymmetric particles. For example, particle patterns shown in Fig. 4B are chiral.

Fig. 4.

Colloids with multifunctional surfaces. (A) Experimental approach used for immobilization of multiple (bio)molecules to create two specific patterns (green U and red M). (B) Three-dimensional assembly of CLSM micrographs showing close-to-identical colloids modified with two different proteins according to the process described for A. (C) Three-dimensional assembly of CLSM micrographs showing nine colloids with distinct surface modifications fabricated through parallel processing using two process cycles; dotted lines show the perimeter of the colloids and are drawn to guide the eyes. (Scale bars, 100 μm.)

In addition to the controlled fabrication of a series of identical colloids with multiple, biologically distinct binding patches, parallel fabrication of multiple types of distinct colloids is of specific interest to a range of different applications, such as biosensors (18). By using our 3D patterning approach, a large number of individually customized colloids could, in principle, be prepared simultaneously. However, individual addressability of each single colloid becomes imperative. Using a protocol similar to the one established for multifunctional colloids (Fig. 4A), we created an array of colloids with two distinct structures: a red square and a green cross (both 137.5 × 137.5 μm in size). Fig. 4C shows a 3 × 3 array of colloids, which resembles a miniaturized game board. In this array, four colloids were modified with a red square, whereas four other colloids were surface-structured with green crosses. The colloid located in the upper right corner was intentionally not modified to demonstrate independent addressability and designability. To generate the colloidal assembly shown in the confocal micrograph of Fig. 4C, only two consecutive photopatterning cycles were needed. Again, both pattern fidelity and reproduction were superb and were not compromised by the use of multiple photopatterning and washing steps.

Conclusions

Although significant attention has been given in the past to controlling sizes and shapes of colloids, much less is known about the precise engineering of colloidal surfaces (18–21). It was our intent to design colloids with microstructured surfaces that rival their flat counterparts with respect to pattern precision and quality. Because traditional spin-coating techniques used during photolithographic processing are intrinsically limited to flat substrates, we replaced the spin-coated photoresist with a vapor-deposited, photodefinable polymer coating. These coatings were applied as ultrathin, homogeneous films on curved surfaces and supported high-precision microstructuring of colloidal particles with radii between 50 and 200 μm. These findings mark an important progression toward microstructured colloidal particles, because well established lithographic processes, such as the projection lithography used in this study, can now be extended to the curved surfaces of spherical objects. With the concept of colloidal surface patterning demonstrated, future research will need to be directed toward fine-tuning binding affinities and patch geometries to yield optimal interactions during self-organization and toward extending the operative particle size ranges from tens and hundreds of micrometers down to the submicrometer regime.

Appearing on the horizon are novel microstructuring technologies that may establish a shift in paradigm with respect to ease of fabrication as well as variability and complexity of available surface motifs. The CVD-based technology described here relies on flexible surface chemistries for surface modification and is therefore compatible with a wide range of different surface markers. This feature may provide access to colloids with delicate balances between repulsive and attractive surface interactions, which are believed to play an essential role in programmed self-assembly of next-generation optoelectronic materials (11).

Materials and Methods

CVD Polymerization.

Poly[4-benzoyl-p-xylylene-co-p-xylylene] was synthesized via CVD polymerization (25–27). The starting material, 4-benzoyl[2.2]paracyclophane, was sublimed under vacuum and converted by pyrolysis into the corresponding quinodimethanes, which spontaneously polymerized upon condensation to the substrate surface. A HPR-30 mass spectrometer (Hiden Analytical, Warrington, U.K.) was connected to the deposition chamber for in situ analysis. Mass spectra were recorded at an emission of 1,000 μÅ and an electron energy of 70.0 V by using a faraday detector scanning from 100 g/mol to 500 g/mol. A constant argon flow of 20 standard cubic centimeters per minute was used as the carrier. Sublimation temperatures were kept at 120°C followed by pyrolysis at 800°C. Subsequently, polymerization occurred on a rotating, cooled sample holder placed inside a stainless steel chamber with a wall temperature of 120°C. The coating pressure was 0.2 mbar (1 bar = 100 kPa) or below. The exit of the chamber was connected via a cooling trap to a mechanical pump. Both polystyrene and glass microspheres (Polysciences, Warrington, PA) were used in this study as spherical substrates, which were surface-modified via CVD polymerization to deposit the photodefinable CVD polymer (poly[4-benzoyl-p-xylylene-co-p-xylylene]). Sphere radii ranged from 50 to 200 μm. To generate contrast during microscopy, some microspheres were coated with a 10-nm-thick Ti adhesion layer and a 80-nm-thick gold layer.

Projection Lithography.

Microstructure blueprints were generated by a custom-made projection system (Fig. 1) consisting of a 1,024- × 768-pixel DMD (Texas Instruments, Dallas, TX), a high-pressure mercury light source (350–450 nm), and a downstream focus lens system (Brilliant Technology, Brentwood, TN). The projected pixel size is 27.5 × 27.5 μm. Microspheres were modified with photodefinable CVD polymer and then were aligned onto a glass substrate by using a microfabricated substrate with shallow holes. Subsequently, the resulting sample was immersed into a ligand solution followed by exposure to a predesigned pattern (UM, 247.5 × 137.5 μm). Deionized water was used after projection to wash away unreacted ligands.

Self-Organization Experiments.

Two pairs of polystyrene microspheres with radii between 50 and 60 μm were used to study self-organization. One pair was coated with gold (80 nm thick, yellow in color) to distinguish from a second pair (bare polystyrene, gray in color). After CVD modification, both groups were subsequently immersed into 10 mM biotin-hydrazide/PBS (pH 7.4) solution followed by exposure to a predesigned pattern (square, 55 × 55 μm). Subsequent incubation with 10 μg/ml streptavidin (Pierce, Rockford, IL) for 90 min was performed on gold-coated microspheres only. Finally, both groups were mixed/incubated in PBS (pH 7.4) solution for 90 min. Self-organized colloidal structures were observed by optical microscopy (BX 60; Olympus, Tokyo, Japan). In a parallel control experiment, self-assembly of particles without streptavidin was examined.

Multifunctional Modifications.

After CVD coating, colloids with radii between ≈180 and 210 μm were modified with predesigned illumination patterns and placed in a 20 mM biotin-PEO dimer solution (Pierce). Deionized water was carefully used to wash away unreacted biotin–PEO dimer after projection. The resulting samples were then exposed to a second predesigned pattern within 2% (wt/vol) 4-arm poly(ethylene oxide) solution (amine-terminated; Sigma–Aldrich, St. Louis, MO). Deionized water was used to wash away unreacted poly(ethylene oxide) after projection. For bioconjugation, samples were first incubated with 50 μg/ml atto-655-NHS ester (Sigma–Aldrich) in PBS containing 0.1% (wt/vol) bovine albumin and 0.02% (vol/vol) Tween 20 for 120 min and subsequently rinsed several times with PBS containing 0.1% (wt/vol) bovine albumin and 0.02% (vol/vol) Tween 20. Secondly, the rinsed samples were incubated with incubated with rhodamine-conjugated (TRITC) streptavidin at 50 μg/ml (Pierce) in PBS containing 0.1% (wt/vol) bovine albumin and 0.02% (vol/vol) Tween 20 for 90 min. Finally, samples were thoroughly rinsed with PBS containing 0.1% (wt/vol) bovine albumin and 0.02% (vol/vol) Tween 20.

Confocal Microscopy.

Samples were analyzed by CLSM (TCS SP2; Leica Microsystems, Bannockburn, IL) on an inverted microscope (DMIRE2; Leica Microsystems). An GreNe laser (wavelength, 543 nm) and a HeNe laser (wavelength, 633 nm) were used to excite TRITC-labeled streptavidin and atto-655-NHS ester respectively. The emission was confined to 560–595 nm for TRITC-labeled streptavidin and 650–700 nm for atto-655-NHS ester upon observation. Confocal z-stack imaging scanning was performed by using a 2.5-μm pixel resolution in the z direction.

Abbreviations

CVD

chemical vapor deposition

DMD

digital micromirror device

CLSM

confocal laser scanning microscopy.