3.1 Basic polymerization scheme

Table 1 gives the various recipes carried out in this study as well as the final parameters of the dispersions, i.e. rate of polymerization, particle size, polydispersity index, number average molar mass $\overline{M_n}$, molar mass distribution (MMD) $\overline{M_w}/\overline{M_n}$ and cycle contents at final conversion. Reactions were mainly conducted at room temperature (except run 5, conducted at 5 °C), at rather high contents in monomer (17 wt%, 12 wt% after hydrolysis/condensation) and surfactant (10 wt% compared to monomer) to generate small particles. The DCMS monomer was poured gradually to avoid temperature rising through acid/water mixing [2,26]. An electrosteric surfactant, namely polyoxyethylene(8) lauryl sulphate sodium (POELSNa), was chosen and its structure characterized by electrospray mass spectrometry (Fig. 1 ). It exhibits alkyl chains mainly composed of x = 12 methylene units with chains of 14 and 16 units that also show up. The ethoxylated moiety ranges between y = 0 to 10 ethoxy units, with a maximum set at 8 units. Even though this surfactant shows a fairly broad compositional distribution, it has exhibited remarkable qualities in stabilizing emulsions, even in high salt concentrations.

A typical SEC trace obtained in the course of a reaction is given in Fig. 2 . Blank injection of DCMS dissolved in toluene revealed that the monomer elutes at the same location as the toluene peak and also in negative polarity. Injection of samples of exact concentration did not reveal a variation of the negative peak with time after SEC trace normalization. This result indicates that, as expected [2], DCMS instantaneously hydrolyses once added in the reactor. In addition, only small cycles of typically 4 to 6 D units that elute at fixed retention time (Fig. 2) are present, since short disilanol terminated chains (typically, 3–4 D units) would elute after the toluene peak [8], where the baseline remains flat. High molar mass polymer is also observed right from the beginning.

The experiments presented in Table 1 were carried out using different lots of monomers, glassware and addition apparatus, but still produced similar data trends. The robustness of the process is established by compiling all particle size, polymer conversion, average molar mass $\overline{M_n}$ and polymolecularity results (including data from kinetics sampling not reported in Table 1) and plotting a master curve of these parameters (Fig. 3 ). Particle size decreases constantly, while the average molar mass increases with polymer conversion. Particle size reduction with conversion is an indication that the polymerization process obeys an emulsion (initial big monomer droplets that vanished to form smaller particles) rather than a microemulsion (initial monomer swollen micelles that collapse under polymerization to generate bigger particles) mechanism. Molar masses still evolve after the polymer conversion levels off at about 90%, whereas particle sizes reach a constant value of 50 nm (see Fig. 3). MMDs reach the theoretical value of 2 expected from the Flory distribution. Note the scattering in the MMD data ascribed to the tailing of the surfactant peak on the polymer one (Fig. 2).

A simple reaction scheme based on these preliminary observations and previous cationic cyclosiloxane polymerization studies [6,10–14] is proposed in Scheme 1 . Rapid hydrolysis and condensation of short disilanol in the water phase lead to cycles by end-biting of small dihydroxy oligomers. The strong acidity catalyses both the rapid condensation and end-biting reactions [30]. The cyclosiloxanes are so hydrophobic (solubility of D₄ in water is 2.4 × 10^–7 mol l^–1) [31] that they gather into droplets stabilized against coalescence by the electrosteric surfactant and against Ostwald ripening by the few polymeric chains they contain. The surfactant also catalyses interfacial ring-opening polymerization of cyclosiloxane and PDMS polycondensation [13,32]. The last step, according to previous studies, is likely to occur in the bulk of the particles by phase transfer catalysis [8].

3.2 Detailed kinetics and colloidal studies

A model recipe was chosen and carried out in three experiments that differ in volume, reactor type and monomer addition conditions (see details in Table 1, runs 1–3).

The overall content of cyclosiloxanes was obtained from polymer and cycle SEC peak integration, after refractive index correction (see experimental part). Since SEC resolution is good enough to differentiate peaks of the various cyclosiloxanes (at least D₄, D₅ and D₆) (Fig. 2), these were deconvoluted into a series of Gaussian peaks, as presented in Fig. 4 on an ideal case. D₄, D₅ and D₆ cyclic contents, derived from the Gaussian curves integration, are presented in Fig. 5 as a function of time. D₄ is created in high yields at the beginning, but rapidly disappears. On the other hand, a majority of D₅, and to a lesser extent D₆, are gradually formed and consumed as a function of time. Such a trend is very close to cyclosiloxane cationic ROP in miniemulsion, where 5- or 6-unit cycle contents pass through a maximum before being almost quantitatively consumed at long polymerization times [14]. This behaviour is quite different from the one observed in anionic process, where 4-unit cycles are always in excess compared to the less strained cycles. These latter gradually build up in the mixture with time for never disappearing [5,8].

Fig. 6 shows the evolution of the number average molar mass and MMD as a function of time. $\overline{M_n},$ s smoothly evolve with time to reach 50 000 g mol^–1 after seven days, in parallel with $\overline{M_w}/\overline{M_n}$, which rises up to a value of 2. Polycondensation is a slow motion process here compared to experiments where dodecyl benzene sulphonate acid (DBSA) is used as a catalyst [32]. This result is a consequence of the weak ability of the proton to penetrate the surfactant ethoxylated layer at the shell of the particles. The equilibrium molar mass values are three times less than those involving a DBSA-catalysed PDMS condensation, values that depend on the organic phase content in the recipe [14].

Fig. 7 reports the average particle size (z-average diameter in nm) and polydispersity index as a function of time. The particle size decreases down to 50 nm, whereas the polydispersity index first increases and then decreases with time. Since small cycles and long-chain PDMS exhibit comparable densities (0.95 and 0.976, respectively), a decrease in particle size is related to an increase of the number of particles during the reaction. A tentative scheme based on emulsion micellar particle nucleation [33] is proposed to account for these results (Scheme 2 ). Cyclosiloxanes formed in situ are rapidly entrapped by the surfactant to produce large swollen particles containing few polymer chains. The primary diameter of 170 nm found here actually corresponds to typical sizes found in polymer-stabilized monomer droplets prepared in miniemulsion; these entities will be thus referred to as ‘monomer’ droplets (see Scheme 2). Two processes compete to deplete these droplets: (i) nucleation of a new particle via monomer diffusion towards an empty micelle; (ii) monomer diffusion towards the smaller particles to compensate for their rapid consumption by interfacial polymerization, which is faster in particles exhibiting large specific surface. Both processes decrease the average particle size of the dispersion, the second however to a lesser extent, since differences between particle sizes are less important. On the other hand, polydispersity increases with continuous nucleation, which prevails at the beginning of the polymerization, whereas it decreases with particle swelling, dominant at longer time when micelles have vanished. The fact that the particle size equilibrium is reached after long times (typically 4 days, see also Fig. 5) is explained by the considerably slow rates of diffusion of cyclosiloxanes [17], especially for those of 5 and 6 D units [11], in such a high ionic strength medium.

3.3 Influence of specific parameters

The influence of the concentration of surfactant was shown from experiments carried out at DCMS addition rates of 6.64 ml h^–1 (runs 1,4–7,11) and 9.4 ml h^–1 (runs 2,3,9,10). Particle sizes decrease with surfactant content, but to a low extent. Average molar masses decline with increasing surfactant concentration, which is again attributed to a shielding of the surface by the bulky surfactant that slows down the polycondensation reaction. Cycle contents also slightly decrease with increasing surfactant content (from 9.7 to 7.8 wt% at an addition rate of 6.64 ml h^–1; Table 1, runs 1,4–7,11), but this discrepancy may not be significant regarding the scattering in SEC integration.

The same recipe was reproduced at different volume and addition rates (Table 1, runs 1-3,5,11,12). Slow addition of DCMS profits to a fine decrease of the final particle size as well as polydispersity index. This result indicates that a better repartition of cyclosiloxanes towards the micelles is exerted by slowly adding the monomer, in other words less big droplets form during the process [14, 34]. Batch D₄ polymerization in similar surfactant and acidity conditions was also carried out in this study to check out the effect of ‘infinitely rapid’ addition rate. The particle size was much larger (172 nm) and the polydispersity index smaller (1.08) in these conditions. Nucleation of micelles is a faster step in this experiment since only D₄, which diffuses more readily than bigger cyclosiloxanes [11], is injected here. Polydispersity index has then time to decrease while particles grow according to the second step in Scheme 2. Variations in molar masses as well as cycle contents with addition rate are difficult to observe (see Table 1) and would require generating further data to show a possible trend.

3.4 Unsolved problem

The formation of an oil layer at the top of the reactor was systematically observed here, the content of which varies within the experiments. Depending on the surfactant content, a turbid layer made of large droplets or an oily film was observed. It was also noticed that this layer gradually vanishes at long times and accordingly the average particle size increases, revealing the slow but real motion of the polymerization/polycondensation reactions. Such phase separation is recurrent of emulsion polymerization of cyclosiloxanes, but is so far poorly understood [15,34]. One way to avoid the formation of an oil layer is to generate by sonication droplets of finite size and high specific surface that rapidly polymerize. Another possibility is to use a PDMS seed [17] previously synthesized in the presence of similar electrosteric surfactant to avoid the formation of big monomer droplets. We are currently testing this second possibility.