Supramolecular perspectives in colloid science

Let us first return to our classical example: the soap micelle. Such a particle forms by spontaneous aggregation of amphiphilic objects, e.g., molecules with one soluble and one insoluble part, see Fig. 1. The primary driving force for the aggregation is provided by the insoluble part: for each free molecule transferred to the aggregate’s interior the free energy of solvation drops by a certain amount (‘volume term’). However, a small aggregate does not have so much interior: a substantial fraction of the molecules is still in contact with the solvent rather than with molecules of its own kind. Hence, a positive correction to the free energy must be introduced, the importance of which obviously scales with the surface area of the aggregate (‘surface tension term’). The sum of these two free energies varies non-monotonically with the aggregates’ size: first it goes up because the surface term dominates; at larger sizes, it decreases because the volume term becomes more important.

If this were the whole picture, particles, once formed, would tend to grow to very large sizes (because the free energy would continue to decrease with size). In fact the two terms just mentioned are the ingredients of classical nucleation theory [3]. However, free growth is prohibited by the lyophilic moieties (head groups) of the molecules: the free energy increases as soon as these loose contact with the solvent by being ‘buried’ in the domain of insoluble parts. Hence, aggregate growth in all three spatial dimensions has to stop. This leaves two possibilities: growth has to stop altogether (we then get spherical micelles), or it must be limited to one dimension (cylindrical micelles) or two dimensions (lamellae) [4]. Due to the balance between lyophobic attraction and lyophilic repulsion, the free energy as a function of aggregation number (or of aggregation number per unit length or area) has a distinct minimum, which limits the extent to which size fluctuations can occur.

The case of cylindrical micelles is interesting. Long cylindrical micelles share many properties with polymers because of their ability to undergo thermal shape (bending) fluctuations. The typical difference is, of course, that the covalent bonds in an ordinary polymer are extremely long-lived, whereas these in a cylindrical micelle break and form rapidly at experimental time scales, which has consequences for their rheological behavior. Hence, these micelles can be seen as part of a special class of ‘reversible supramolecular polymers’ (RSP’s), and quite some research has been devoted to them to understand their static and dynamic properties [5, 6]. Cylindrical surfactant micelles are not the only possible kind of RSP; there are many more. For example, in almost all living creatures, the protein actin occurs; it forms filamentous assemblies that are part of the cytoskeleton, the structure responsible for the cell’s mechanical performance [7]. In a biomimetic approach, we have recently prepared a new protein polymer which can be triggered to form very similar fibrils; this polymer will be discussed later in this review.

Another RSP subclass consists of bifunctional molecules that are capable to form reversible bonds with exactly two (complementary) partners. One recent example that attracted quite some attention is based upon the DNA-inspired ureido-pyrimidinone (UPy) moiety, which dimerizes in non-aqueous systems due to hydrogen bonding [8]. A spacer carrying UPy moieties at both ends forms reversible chains, and new materials based upon this idea are now being developed. In our own group, we have recently explored the use of coordination chemistry for making reversible polymers with coordinated metal ions in the backbone. A pyridine ring carrying two carboxylic acid groups turned out to be a good ligand; a wide range of metal ions could form a bidentate complex with two such ligands, so that two such ligands connected back-to-back by a suitable spacer (denoted as ‘L₂EO _n ’) should be able to form water-soluble coordination polymers (Fig. 2) [9].

This turned out to be the case: at the proper (1:2) metal/ligand ratio (e.g., with Zn as the coordinated metal), chains were found in the solution. However, small rings consisting of two bisligands and two metal ions were also detected in cases where the spacer had just the right length. Typically, low concentrations would then favor rings, whereas high concentrations would induce ring-opening polymerization of the rings to chains [10]. With large trivalent ions like neodymium (Nd³⁺), triple coordination can occur, leading to branched polymers and reversible gels [11]. Clearly, RSP’s are an intriguing class of spontaneously assembling materials that can be considered as colloids in their own right (Fig. 3).

However, one can do more. The ingredients discussed above can be used as the starting point for assembling new colloidal species with increased complexity. As explained, poor solubility and a stop mechanism are necessary, but also sufficient ingredients of a self-assembly process. When we consider solubility, we usually think of a single molecular species in a solvent: poor solubility then means that the solute segregates from the solvent. However, solubility may also be affected by attraction between two soluble species. What options do we have here?

Around 1930, Bungenberg de Jong investigated mixtures of oppositely charged macromolecules like gelatin and gum arabic, which were found to form complexes which could phase separate (at appropriate compositions) into a dilute watery phase and a concentrated fluid polymer complex phase [12]. He introduced the term ‘complex coacervation’ for this phenomenon. The formation of a complex between macromolecules carrying opposite charges can be seen as an ion-exchange process: ions in the overlapping double layers (the counterion clouds surrounding any macroion) are released when the oppositely charged polymers meet in a close encounter. This leads to an entropy increase determined by the relative change in local concentration experienced by the small ions, and to an additional energy decrease, if the average distance between opposite charges becomes smaller. Hence, unless the ionic strength is too high, macroions of opposite charge attract and tend to aggregate in a dense phase. Of course, bringing together macromolecules of like charge confines rather than releases the counterions, which leads to repulsion, as is very commonly known in colloid science [13]. Hence, what happens in mixtures of two oppositely charged macroions is therefore determined to a large extent by the charge stoichiometry, that is by how many charges are positive and how many are negative. A convenient way to express this is by the charge fraction f _i , the ratio between the number of (polymer) charges of a given sign (i; plus or minus), and the total number of charges (plus and minus) on all polymers in the system. If f is equal to 0.5, the macroion charges are exactly matched and all counterions can be released; hence, there is no repulsion from small ions left and we expect phase separation. Of course, some left over small ions generate a repulsion that can suppress formation of a dense phase, but a certain number of them can be tolerated, so that phase separation can occur in a window of f values around 0.5. The width of this window decreases with ionic strength. Outside this window, at f values closer to 0 or 1, there are too many small ions that cannot be released and these prevent the system from phase separating. On the basis of these considerations one can draw a generic phase diagram, see Fig. 4.

The key issue is that we can have poor solubility with pairs of oppositely charged macromolecules. This might be employed to make films and capsules. The now very popular ‘layer-by-layer’ method [14] aims at doing so, and indeed works provided redissolution upon exposure to a polyelectrolyte solution is slow or even totally suppressed so that it cannot occur during a dipping cycle. Typically, polymers with strong ionic groups (e.g., sulfonates and quaternary ammonium groups) are good at this, and backbone hydrophobicity also helps. In other cases, however, redissolution does occur [15].

What about a stop mechanism? This is straightforward: an uncharged block or graft added to a charged polymer can very well fulfill this role, provided it has no tendency to enter the dense complex phase. Given the general tendency of polymers to avoid mixing unless there is a specific attraction, this is very unlikely to happen. Hence, we can make micelles with a pair of oppositely charged macromolecules, at least one of which carries a water-soluble, non-ionic block [16]. Such micelles are now known for about 12 years, in many variations. They are usually spherical, having radii varying between 10 and 30 nm, and typically consist of a relatively dense core which contains the insoluble polyelectrolyte complex, surrounded by a more dilute corona of swollen neutral and hydrophilic chains; we denote them accordingly as Complex Coacervate Core Micelles or C3M’s. Other terms used are ‘PIC micelles’ (polyion complex micelles) or ‘BIC’s’ (‘block ionomer complexes’). In several respects, they are very similar to the more familiar micelles formed by amphiphilic diblock copolymers such as, e.g., poly ethylene oxide–b-polystyrene. New features are, however, that they are sensitive to solution parameters that affect the ionic complex, such as salt concentration and pH, and that they often respond reversibly to changes in polymer composition: they appear near-stoichiometric conditions, but disappear upon leaving the near-stoichiometric range, just like the complex coacervates do. Micelles of this kind are highly interesting candidates for targeted drug delivery to tumors, since they have just the right size to benefit of the Enhanced Permeability and Retention (EPR) effect that occurs in tumor tissue due to the leaking of blood vessels, and they can be loaded with various anticancer drugs [17].

From a colloid science viewpoint, one may now ask the question: to what extent do these systems enrich our science, by introducing new types of colloids and posing new questions? As our first example, we discuss particles with mixed coronas. The very fact that it takes two species rather than one to make these micelles allows us to choose a pair of dissimilar water-soluble blocks for the corona. Since, generally, dissimilar polymers tend to avoid mixing, this generates micelles which force chains which would otherwise stay apart to ‘live together’ due to the attraction between the charged blocks they are connected to. Hence, the corona tries to phase separate without destroying the micelle, and the simplest way it can do this is to become a bipolar particle, with the different corona chains sticking out on either side. Such micelles would be the first instance of reversible (self-assembled) bipolar particles; we denote them as ‘Janus micelles’ after the double-faced Roman deity Janus. As the reader will appreciate, proving the ‘Janus’ character is not a trivial task since typical methods for obtaining the relevant structural information (e.g., scattering) tend to be insufficiently sensitive at the nanolevel. However, we succeeded; in Figs. 5–7, we present the evidence which convinced us [18].

The polymers used were polyacrylate-b-polyacryl amide (PAA₄₂-b-PAAm₄₁₇) and poly ethylene oxide-b-poly-N-methyl-2-vinylpyridinium iodide (P2MVP₄₂-b-PEO₄₄₆), mixed so that f ₌  0.5, in dilute aqueous solution. First, cryo-TEM images (Fig. 5) show a mixture of circular and elliptical objects, as if one is looking at a collection of randomly oriented discs. Usually, the corona is invisible in such pictures so that, probably, we deal with disc-shaped cores.

This is confirmed by SANS data (Fig. 6). A fit with a shape factor for ellipsoids with main axes of 18 and 2.5 nm, respectively, is the only one giving satisfactory agreement (see residuals), without having to assume very exotic polydispersities which are not seen in cryo-TEM. Moreover, the sizes found closely match those of the images.

So far, this only tells us we have a non-spherical core shape, which would be very unlikely without some sort of symmetry breaking tendency. A more direct piece of evidence that the corona is not mixed comes from NMR-NOESY experiments. In such experiments, a particular nuclear resonance can be observed by irradiation of a nucleus which is chemically different but spatially close: magnetization is then transferred by coupling between the spins. As a result, the 2D NOESY spectrum of true mixtures of two polymers shows off-diagonal peaks, whereas such peaks would be absent for a demixed corona. Both cases do indeed occur (Fig. 7): polyacryl amide and poly glyceryl methacrylate mix (Fig. 7, left diagram), but poly acrylamide and poly (ethylene oxide) do not (Fig. 7, right diagram), which strongly supports the Janus hypothesis.

Finally, the alleged Janus system shows depolarization of scattered polarized light. This not only betrays a certain amount of optical anisotropy, but also allows to estimate the rotational diffusion coefficient. The data suggest an ellipsoidal shape, but much larger than that is seen by SANS. The way this can be accounted for is that each ‘face’ of the disc-like core carries a ‘lobe’ of swollen corona chains, one side PEO, the other side PAAm. Hence, our Janus particles can be represented by the sketch in Fig. 8.

Theoretical modeling also shows that a double-lobe structure is quite likely to appear (Fig. 9) provided the repulsion (represented by a Flory–Huggins interaction parameter) is large enough [19].

As is clear from the example just discussed, the very fact that C3M’s consist of two oppositely charged species opens new options. A further step is that one can choose one of the ‘lobes’ of a Janus micelle to be a temperature-sensitive polymer such as, e.g., PNIPAM (poly N-isopropyl acrylamide) or PDEGMA (poly diethylene glycol methacrylate). Upon increasing the temperature above the lower critical solubility temperature (LCST), the Janus micelle will have one side in a poor solvent condition, leading to collapse and attraction. As a result, several micelles will cluster to form a large particle, but, again, the presence of the solvated sides will prevent macroscopic phase separation. Hence, some sort of ‘supermicelle’ will appear as long as the temperature exceeds the LCST. This is precisely what we observe (Fig. 10) with micelles consisting of PAA-b-PAAm and PDMAEMA-b-PDEGMA (this latter block has an LCST of about 35 °C), but also with PAA-b-PNIPAM and PEO-b-PVPy (PNIPAM has an LCST of about 32 °C).

As to the dynamics of formation, there is a fundamental difference between two-component and one-component micelles. As is often the case in a colloid preparation method, one starts out with a soluble precursor that is driven into a colloidal state by a change in solubility due to, e.g., chemical reaction, temperature change or solvent change. Polymer micelles are usually prepared by a ‘solvent quench’, i.e., a rapid change in solvent composition. When this is done with a conventional amphiphilic copolymer, the one block which becomes insoluble will collapse (which it can do without any other polymer molecule around) after which it aggregates into a multimolecular particle. Only the rate of the latter process is concentration dependent; in dilute solutions, the unimolecular collapse process will always be finished before any aggregation takes place. As a result, the formation of micelles proceeds rather smoothly, even though quite some shape variation is possible [20]. With C3M’s this is completely different. Here, the first step is always intermolecular, no unimolecular collapse is possible. What then happens depends strongly on composition. At compositions around f = 0.5, the random encounters between molecules of opposite charge initially leads to a dense, insoluble network that can subsequently rearrange into micelles. The rearrangement process requires the multiple breaking and forming of ions pairs which can be very slow due to the associated energy barriers [21], or may be even blocked altogether. At compositions far from f = 0.5, the build up of multimolecular species is much more gradual. Hence, the rate at which components are mixed has great consequences for the kinetic pathway taken by the system and, depending on relaxation rates, may sometimes lead to equilibrated structures, sometimes to frozen particles [22]. One can therefore say that these systems take an intermediate position between the ‘lyophobic’ and ‘lyophilic’ classes traditionally distinguished in colloid science.

What other options do we have? A diblock copolymer is very effective in providing a stop mechanism. Hence, this ingredient cannot be missed. However, for the other ingredient(s) one has quite some freedom. Simple polyelectrolytes of varying length have been used, but also proteins (enzymes) [23], ionic surfactants [24], and inorganic particles have been successfully incorporated. Systems composed of an ionic surfactant and diblock copolymers deserve some attention, since they can be considered as ‘micelles within micelles’ that is, an example of hierarchical structures. Low MW surfactant micelles are very dynamic, and therefore, the compound micelles based on them are equilibrium systems; their structure is entirely independent of sample history.

Another example of a hierarchical nanostructure is the C3M formed between a diblock copolymer and a charged water-soluble RSP. We have recently introduced such C3M’s by using a negatively charged coordination polymer formed by the L₂EO _n bisligands and metal ions, with diblocks consisting of PEO and positively charged P2MVP [25, 26]. In this case, it is the coordination compound which polymerizes (self-assembles) upon forming the complex with P2MVP, even though it is present in the form of small rings in the starting solution. The coordination polymer introduces metal ions into the micellar core (Zn²⁺, Ni²⁺, Fe²⁺, Fe³⁺, Nd³⁺,…) that endow it with high electron densities, rendering them nicely visible in EM pictures (Fig. 11) as well as with magnetic properties for appropriate metals (Fe, Co).

The fact that the coordination compound polymerizes reversibly manifests itself in other ways. One is by the fact that off-stoichiometric compositions behave in an asymmetric fashion. At f+ > 0.5, soluble complexes exist because excess PVPy provides the repulsive interaction which prevents mutual aggregation of the block copolymers. At f+ < 0.5, however, the excess coordination compound usually cannot polymerize beyond the dimer state (which is insufficient to stabilize small species) unless the concentration is high enough. Coordination complexes with large trivalent ions like, e.g., Nd³⁺, are more potent stabilizers, as they can form trivalent (L3Me) complexes with a single ion, leading to polymers with rings and branches which are much less likely to break up under thermal motion and therefore persist down to significantly lower concentrations [27]. This can be clearly seen in Fig. 12 which compares C3M’s with Zn and Nd. The implication is that in a solution of ring shaped L₂EO _n /Zn, any added PEO-PVPy will very soon produce colloidal C3M’s, whereas with L₂EO _n /Nd complexes one needs to approach f+ = 0.5 much closer before this happens.

Another proof of reversibility is the fact that a CMC can be measured for these micelles. As the cohesive strength of the cores is highly salt-sensitive, we should expect a salt-dependent CMC, and we do find that, as can be seen in Fig. 13.

A central concern of colloid science is with forces between lyophobic colloidal particles. These can be electrodynamic in origin and arise from the body of the particles (van der Waals forces, dispersion forces), but quite often they must be assigned to properties of the particle/medium interface (true surface forces). As such, they are the result of all the interactions between the particle and the surrounding medium, including various solutes. The well-known double-layer interaction is an example that has been studied at length [28]; and also steric and depletion forces induced by soluble polymers have received much attention [29]. However, there are many more cases where subtle processes in solution show up as colloidal interactions or surface forces. We consider some cases where self-assembling objects are involved.

One example which has been discussed from a general theoretical viewpoint is that of RSP’s between surfaces [30]. RSP’s come in two classes: ‘non-directional’ and ‘directional’. The first class forms structures which possess inversion symmetry: the chains do not have a ‘head’ and a ‘tail’ or a ‘+’ and ‘−’ end. The second class does, because all the units have head and tail. The RSP’s based on coordination complexes clearly belong to the former class; an example of the former would be a single-stranded DNA chain the ends of which (when read in the same direction) are complementary. As can be shown, the non-directional RSP’s, when end-attached to identical surfaces, always generate attraction, whereas the non-directional ones can generate repulsion. Compatible mixtures of the two can even generate non-monotonic interactions. One example of attraction in the case of a non-directional RSP has been recently reported [31], but the ‘directional’ case has not been realized yet.

A second example is that of solutions of self-assembling species interacting with a solid/liquid interface [32]. A third example is a solution of self-assembling species close to a phase separation boundary. Here, the presence of a surface may induce a ‘precursor’ of the incipient phase when it prefers solute over solvent. If this occurs for liquid/liquid phase separation, one could refer to it as capillary condensation, and the ensuing force is a capillary force. A beautiful example was recently reported [33] and it seems that many more examples lie awaiting us in multicomponent complex fluids. They definitely deserve investigation, particularly from a dynamic point of view. Innovations from the side of synthetic chemistry will bring in new compounds that might induce such surface forces.

Let us discuss some ideas. Copolymers with a diblock structure tend to primarily form micelles, but copolymers with a triblock sequence (sometimes called telechelics) can do more. In particular, they can make networks of micelles. In the latter category we find, e.g., the associative thickeners based on polyethers (PEO) with insoluble (alkyl) end groups widely applied in water-based paints. Following up on this, one would expect that telechelics with charged end blocks would also be able to form similar networks in the presence of a stoichiometric amount of oppositely charged macroions. Remarkably, no literature reports on such networks can be found in the literature so that we have started an investigation in this direction.

A second example is the combination of two oppositely charged RSP’s. We already discussed an example of a negatively charged coordination polymer. Recently, we studied a positively charged protein polymer with a triblock structure (labeled CS^HS^HC) that self-assembled into linear structures (fibrils) whenever its charge was screened. This is a designed protein with a block-like sequence; its end blocks are hydrophilic and swell in water, whereas its middle block can form a so-called ‘beta roll’ which exposes hydrophobic faces and then polymerizes by hydrophobicity driven stacking. In a solution of neutral pH, the protein simply forms a molecular solution, but upon adding a dilute solution of the L₂EO _n /Zn compound, both the protein and the coordination RSP polymerized together, forming a tape-like structure that can be very well seen in cryo-TEM pictures (Fig. 14). Clearly, combining two (or more) ingredients that co-assemble rather than just self-assembly leads to very interesting new nano-objects and materials, and is a challenging direction for colloid science.

Colloid Science, with its roots in Physical Chemistry, and Supramolecular Chemistry, with its firm basis of synthetic organic and organometallic chemistry can mutually benefit a lot. The appearance of new structures and ways to tune and manipulate structures enriches both fields. Promising developments are to be found also in the realm of biopolymers, where the repertoire of possible conformations and internal interactions is so much larger than that for most synthetic polymers, and rules of thumb developed for simple polymers do not apply. Hence, nanoassemblies of biopolymers (and hybrids of these with synthetic compounds) are a great target for future work.