Electrostatically Directed Self-Assembly of Ultrathin Supramolecular Polymer Microcapsules

Supramolecular self-assembly offers routes to challenging architectures on the molecular and macroscopic scale. Coupled with microfluidics it has been used to make microcapsules—where a 2D sheet is shaped in 3D, encapsulating the volume within. In this paper, a versatile methodology to direct the accumulation of capsule-forming components to the droplet interface using electrostatic interactions is described. In this approach, charged copolymers are selectively partitioned to the microdroplet interface by a complementary charged surfactant for subsequent supramolecular cross-linking via cucurbit[8]uril. This dynamic assembly process is employed to selectively form both hollow, ultrathin microcapsules and solid microparticles from a single solution. The ability to dictate the distribution of a mixture of charged copolymers within the microdroplet, as demonstrated by the single-step fabrication of distinct core–shell microcapsules, gives access to a new generation of innovative self-assembled constructs.

ii S1: Microfluidic device design and microdroplet generation Microfluidic devices were manufactured from polydimethylsiloxane (PDMS) via soft lithography, whereby: (i) the microchannel network was designed in silico (AutoCAD), (ii) printed as a negative photo-mask and (iii) transferred onto a silicon wafer spin-coated with SU-8 photoresist via UV-photolithography to form a mould. PDMS and the cross linker (Sylgard 184 elastomer kit, Dow Corning) in a 10:1 ratio were poured onto this mould and allowed to stand overnight at 70 °C. The PDMS layer, imprinted with the microfluidic channel design, was removed and using a biopsy punch (1.0 mm) inlets and an outlet were formed. The imprinted PDMS and a glass substrate were exposed to oxygen plasma for 8 s and then pressed together to seal the microfluidic channels.
Figure S1 | Schematic of microfluidic devices. (A) Water-in-oil microdroplets were generated using a hydrophobic microfluidic device comprising a 60 µm flow-focusing nozzle. (B) Analysis of static microdroplets was conducted in a microfluidic reservoir. (C) A droplettrapping reservoir was employed when a continuous oil flow was required; the droplet-trapping pillars are shown in the inset. All microfluidic channels were 50 µm in depth.
Monodisperse water-in-oil microdroplets were generated with a hydrophobic flow-focusing microfluidic channel ( Figure  S1A). The diameter of the junction was 60 µm with a channel depth of 50 µm. To generate microdroplets, the continuous oil phase and the discrete aqueous phase were injected into the microfluidic device via two syringe pumps (PHD 2000, Harvard Apparatus) with controlled flow rates of 150 and 75 µLh -1 respectively. At the intersection, the shear forces caused the formation of aqueous droplets in oil (Ø = 79.0 ± 0.7 µm, Figure S2). The continuous phase comprised of the perfluorinated oil, Fluorinert FC-40 (3M), with 2 wt% surfactant (XL-01-171, Sphere Fluidics). To this was added up to 1.0 wt% of either the carboxylic acid (K (-) ) or amine-terminated (K (+) ) poly(hexafluoropropylene oxide) dopant. The dispersed phase consisted of an aqueous solution of charged, functionalized copolymer(s), and where appropriate a molar equivalent of CB[8]:guests. For a typical experiment, a concentration of 60 µM was used for each polymer-bound guest and CB[8] to allow formation of the ternary 1:1:1 complex.
Once formed, microdroplets were collected either into a microfluidic reservoir ( Figure S1B,C) or on the surface of a glass slide for further study and evaporative formation of microcapsules. Residual surfactant was removed by washing with solvent (FC-40), upon which it can be recycled. iii

S2: '1A (-) ⊂CB[8]⊂1B (-) ' Microcapsules
Microdroplets comprising the negatively-charged copolymer, 1A (-) were prepared at 60 µM concentration of azobenzene guest. During droplet generation the concentration of the charged-dopant K (+) within the continuous phase was increased from 0.0 to 1.0 wt% and for comparison in the presence of 1.0 wt% K (-) . The microdroplets were collected into a storage reservoir ( Figure S1B) and the distribution of the fluorescent copolymer within the microdroplet immediately analyzed by laser-scanning confocal microscopy (LSCM). As shown in Figure S3A, when a neutral surfactant is employed 1A (-) remains uniformly distributed throughout the microdroplet. However upon increasing the concentration of the complementarilycharged K (+) the proportion of copolymer accumulated at the interface increases, with near-quantitative assembly at 0.6 wt% K (+) . A similar trend is observed for 1B (-) , with significant accumulation at the interface observed with just 0.4 wt% K (+) ( Figure  S3B  iv Any significant accumulation of copolymer at the interface on evaporation will typically lead to the formation of hollow microcapsules ( Figure S4A), while microdroplets containing dispersed copolymer will form solid microparticles ( Figure S4B). The hollow microcapsule forms creases and folds as it collapses, contrasting strongly with the smooth, dense microparticle.
The ability to trigger capsule formation through evaporative concentration allows for the micro-architecture to be tailored post-droplet formation ( Figure S4C). Methods for dynamically switching the location of charged copolymers within the droplet include: changes in dopant concentration (e.g. evaporation or dilution of solvent, addition of dopant) or the addition of a conflicting charged-dopant. It should be noted that once supramolecular cross-linking has progressed sufficiently to form a microcapsule skin or hydrogel, the micro-architecture is fixed. v The accumulation of 1A (-) at the droplet interface was studied by real-time LSCM using expanded microfluidic exit and oil channels (200 x 50 µm) to slow the droplet flow velocity. Figure S5 plots the ratio of the fluorescence intensity at the droplet interface against the bulk volume of the droplet, as the microdroplet flows along the microfluidic channel. In the presence of K (+) , 1A (-) will begin to move to the oil-water boundary upon droplet formation at the flow-focusing junction, with this process continuing until fluorescence is only observed at the interface. The rate at which 1A (-) is partitioned slows on lowering the concentration of K (+) within the carrier oil , with partitioning not observed for low or neutral concentrations (≤ 0.4 wt% K (+) ).

Figure S5 | Temporal study of the assembly of 1A
(-) at the droplet interface. The ratio of the fluorescence intensity at the droplet interface against that within the bulk volume of the microdroplet is plotted as a function of time lapsed from droplet generation at the flow focus. The rate of movement to the droplet interface increases with increasing the concentration of complementary-charged surfactant, K (+) , in contrast no accumulation is observed under neutral conditions (red). Lines are added to guide the eye.
vi vii In the absence of CB[8], although partitioning within the microdroplet is unperturbed, only smooth, polymer microparticles are formed upon evaporation. This control experiment confirms that supramolecular cross-linking between copolymer chains is crucial to undergo the microcapsule formation ( Figure S7). Hydrophilic cargo can be simultaneously loaded within the microdroplet, allowing one-step encapsulation with high efficiency. Fluorescein-labelled dextran (250 kDa) was used as a model cargo, allowing its location to be tracked by LSCM. Osmotic rehydration of the microcapsule, as shown in Figure S8B, confirms the cargo (green) is retained within the thin polymer shell (red). In contrast, microparticles are smaller, with the cargo homogeneously mixed throughout the polymer network ( Figure S8C). ix The accumulation of positively-charged 2B (+) at the droplet interface was studied by real-time LSCM using expanded microfluidic exit and oil channels (200x50 µm) to slow the droplet flow velocity (droplet diameter = 194 µm, 5 Hz). In the presence of K (-) , rapid partitioning to the interface was observed ( Figure S9). In the absence of K (-) the driving force is removed, with 2B (+) remaining dispersed throughout the droplet. x To investigate the kinetics of electrostatic-driven assembly of charged copolymers, microdroplets containing positivelycharged 2B (+) were prepared at 60 µM concentration of viologen guest and confined within microfluidic 'traps' (Figure S1C) under positive pressure by a continuous flow of oil (250 µLh -1 ). On introduction of 1 wt% of K (-) to the continuous oil flow, 2B (+) was immediately drawn to the interface of the upstream droplet. This accumulation of 2B (+) can be reversed over the same timescale by reverting to a flow of 1 wt% K (+) . Repeated cycling between K (-) and K (+) leads to alternating between diffuse and interfacial assembly, with no loss of efficacy or material observed ( Figure S10).

Figure S10 | Dynamic manipulation of 2B
(+) within a microdroplet. LSCM micrographs of aqueous microdroplets of rhodamine-labelled 2B (+) [MV = 60 µM], held within the trap by a continuous oil flow (250 µL/h). On alternating between 1.0 wt% of orthogonally charged surfactants K (-) and K (+) within the carrier oil, 2B (+) can be repeatedly and controllably assembled at the interface or dispersed throughout the microdroplet, respectively. At 250 µLh -1 diffusion to/from the interface was complete within 25 s for all four droplets.

Figure S11 | The two step mechanism for the formation of the three component supramolecular complex between CB[8] and viologen and azobenzene guests.
xi

S5: Core-shell microcapsules
Core-shell microcapsules were prepared containing a dextran cargo (non-fluorescent, 70 kDa, 160 µM), that allowed for osmotic re-inflation of the collapsed, dry structures post-formation. On wetting, the microcapsule immediately began to inflate, with the loss of creases and folds ( Figure S12A); further expansion allows the three-dimensional nature of the capsule to be observed ( Figure S12B). After 40 minutes in water the core-shell microcapsule had returned to its initial spherical shape, with diameter (Ø = 55 µm) comparable to that seen prior to collapse during evaporative formation. In all cases the '2A

S6: '2A (+) ⊂CB[8]⊂2B (+) ' microcapsules formed in hexadecane oil
The electrostatically-directed self-assembly of supramolecular microcapsules is not dependent upon the use of perfluorinated oils in the continuous phase, but simply upon the presence of an appropriately charged surfactant at the interface. This is exemplified in Figure S13, where 2A (+) ⊂CB[8]⊂2B (+) microcapsules are analogously formed from microdroplets generated with hexadecane carrier oil, with controlled flow rates of 75 and 150 µLh -1 respectively. Here Span-80 (2.5 wt%) is used as a surfactant to stabilise the droplets, while palmitic acid (0.5 wt%) is used as a substitute for the negatively-charged perfluorinated dopant, K (-) . Upon evaporation of the aqueous phase, crumples became apparent on the microcapsule surface, resulting in eventual collapse on loss of the aqueous core. As before, evidence of a polymeric skin around the microdroplet is observed at circa 70 % of the initial droplet diameter. xii

S7: Synthetic methods
Instrumentation. 1 H NMR spectra (400 MHz) were collected on a Bruker Avance QNP 400 MHz Ultrashield spectrometer, equipped with a 5-mm BBO ATM probe with a z-gradient. Weight average molecular weight (M W ), number average molecular weight (M n ) and polydispersity (M W /M n ) were obtained by aqueous GPC. The aqueous GPC setup consisted of a Shodex OHpak SB column, connected in series with a Shimadzu SPD-M20A prominence diode array detector, a Wyatt DAWN HELEOS multi-angle light scattering detector and a Wyatt Optilab rEX refractive index detector.
Materials. All reagents were purchased from Sigma Aldrich, except fluorescein o-methacrylate, which was purchased from Alfa Aesar. propyl] trimethylammonium chloride solution, sodium 4-vinylbenzenesulfonate, Nhydroxyethyl acrylamide, N,N-dimethylaminoethyl methacrylate and acrylic acid monomers were passed through a column of silica gel and purged with high purity nitrogen for one hour prior to use. Solvents and reagents were used without further purification unless otherwise stated. All aqueous solutions were prepared in deionized water (Millipore Milli-Q Gradient A10) ensuring a resistivity of >15 MΩcm -1 .
To purify polyvinylalcohol (Mowiol 6-98, Mw-47kDa, 10.0 g), it was dissolved in water (150 mL) and heated (60 °C) for 3 h, before allowing it to settle overnight, decanted and freeze dried. The purified PVA (100 mg) was dissolved in anhydrous Nmethyl-2-pyrrolidone (10 mL) and to this were added simultaneously: rhodamine B isothiocyanate (1 mg), MV-Hex-NCO dihexafluorophosphate (127 mg, 10% mol ratio of PVA) and a drop of dibutyltin dilaurate. The mixture was stirred for 24 h at room temperature, dialyzed in water through a MWCO 6,000-8,000 membrane and freeze dried. The formed polymer was red in color and characterized as M W = 51 kDa, M n = 28 kDa and PDI = 1.3 from GPC. A monomer ratio of PVA: MV = 18.2: 1 was calculated from