Synthesis of polystyrene template particles
An overview on the explicit particle recipes is shown in Table 1. To control the particle size, the amount of styrene was increased from 6 to 22 mL. As can be seen in Fig. 2a, the particle diameter can be adjusted linearly with the amount of added styrene. The amount of initiator 2,2′-Azobis(2-methylpropionitril) (AIBN) does not influence the final particle size. This can be inferred from particles A–E with 0.3 g of AIBN, and particles F–J with 0.15 g AIBN. Further, we find no influence of the initiator concentration on the conversion of the synthesis within this range. The overall conversion was determined to be 80 to 90% for all syntheses (see Table 1).
Electrostatic stabilization is introduced by the addition of the comonomer 2-Methacryloxyethyltrimethylammoniumchloride (MTC). This introduces a positive charge to the particle surface. For particles A–J, we kept the molar ratio between monomer and comonomer constant with a ratio of 0.8 mol%. This ratio results in a zeta potential of about + 40 mV for all particles. A ratio between monomer and comonomer in the range of 0.8 mol% (particles A–J) up to 1.2 mol% (particles Y) leads to well-functionalized, stable colloids. This is demonstrated by the synthesis of particles X–Z with different amounts of MTC, while styrene and AIBN concentration were kept constant (Fig. 3). Particle X was fabricated without MTC, resulting in a zeta potential ~ 0 mV. Nevertheless, owing to the use of PVP as a steric stabilizer, the dispersion is still stable. Without MTC, the particles exhibit a very smooth surface. Increasing the amount of MTC to 170 μl lead to a rougher surface, while the particle shape remained spherical. Further increasing the amount of MTC to 300 μl lead to aggregated and deformed particles. The particle aggregation is accompanied by a reduction in the overall conversion. The particle diameter decreased by 200 nm from particles X to Y. A potential reason for this deviation from the expected particle diameter is the better solubility of the PS oligomers due to the copolymerization with MTC. This can reduce the tendency for newly formed oligomers to precipitate onto the existing nuclei. As a consequence, the particle growth is less compared to the comonomer free synthesis. Furthermore, newly formed, small nuclei may aggregate on larger particles, resulting in an increasing particle roughness. Overall, using dispersion polymerization, it is possible to prepare polymer particles with standard deviations that are less or equal 5% of the diameter and are, therefore, highly monodisperse (see SI Fig. 2). It provides access to a complementary size range with particle sizes around 1 μm. We want to stress the simplicity of these recipes, where all chemicals, except for AIBN were used without any additional purification.
Synthesis of TiO2 shells
The PS particles were used as seed templates without purification – not even centrifugation after the dispersion polymerization was employed. This procedure is similar to the scalable synthesis of PS@SiO2 core-shell and SiO2 hollow spheres , which provided access to gram-scale amounts of hollow silica spheres. An ethanolic solution (V = 4 mL) of the precursor titanium butoxide (TBT) was added with a concentration of 0.6 molL−1 to the ethanolic particle dispersion using a syringe pump within 30 min.
We first investigated the influence of the template particle surface functionalization on the TiO2 immobilization and shell formation.
Figure 3 demonstrates the necessity to adjust the cationic surface functionalization. Without the addition of MTC granular TiO2 nuclei are immobilized on the polymer surface. For both cases of added MTC (particles Y and Z, respectively) an increasingly smooth shell was observed. The granular appearance of the TiO2 shell is already apparent in the amorphous shell directly after the TiO2 condensation. This can be inferred from the SEM images of the core-shell structures (Fig. 3, second column), where white speckles cover the previously smooth surface. The presence of MTC at the particle surface apparently influences the nucleation and growth mechanism, which we assign to the altered electrostatic environment. Removing the template core by calcination preserves this granularity, which is shown in TEM and TEM tomography images (Fig. 3 right panels). The spherical shape of the hollow sphere is also preserved, which is apparent from the TEM tilt series (see SI gif files). The highest amount of MTC resulted in the most compact and least granular shells. Nevertheless, the TiO2 coated structures were clustered due to the already clustered particles in the PS seed dispersion. Consequently, a balanced adjustment of the particle surface functionality via MTC is crucial for the colloidal stability and successful coating step.
Calcination is the final step to obtain hollow TiO2 particles (Fig. 3 right panel). We want to stress that the core-shell particles can only be transformed into hollow spheres successfully after aging the core-shell dispersion for 24 h prior to purification. This aging process apparently improves the formation of a pre-condensed TiO2 network and proper covalent connectivity among the granular nuclei in the shell. The shell resilience to the thermal decomposition process is improved. Furthermore, the calcination process itself is very important. We used an adapted temperature profile of Schroden et al. , where the template removal is achieved in a two-step process. The first calcination step is undertaken just at the onset of PS decomposition at a relatively low temperature (300 °C). The complete degradation of the PS is then achieved by a second step at 400 °C. To gain a deeper understanding of the calcination process, combined TGA, DSC, and IR measurements have been performed in air and nitrogen (see Fig. 4 a–d and Fig. S4).
The pure template particles (polystyrene) show three prominent steps. These are directly related to the applied temperature profile. The degradation starts at the end of the first heating ramp (2 K/min) (1). Already 80% of polystyrene decompose in the form of CO2 during the isothermal conditions (300 °C), which took 2 h. The PS decomposition is strongly accelerated during the second heat ramp (2 K/min) to 400 °C (2). Within 300 mins another 13% of PS are decomposed to CO2, whereas the residues remain stable in air during the 5 h isothermal heating step at 400 °C. The last 5% of material is fully decomposed during the last heating ramp up to 700 °C (3). The IR spectra (Fig. 4c) only indicated CO2 as the decomposition product. Single IR spectra of the three steps can be found in the supporting information. Each of the decomposition steps was accompanied by an exothermal event in the DSC signal. This indicates the oxidative decomposition of the polystyrene backbone [51, 52].
The calcination of the core-shell particles shows a couple of interesting deviations from the pure PS decomposition. The degradation starts at the same time/temperature as the pure polystyrene particles (1). Since the IR spectra show exclusively CO2 bands (see Fig. 4d), it is reasonable to assume that only the polymer decomposed at this point. The PS decomposition, however, is significantly slower compared to the neat seed particles. This could be caused by the limited mass transport to the PS core owing to the presence of the TiO2 shell. Only 40% mass has been lost by the end of the first isothermal annealing step at 300 °C. The DSC signal reveals a second event occurring during the isothermal part at 300 °C (2). Xie et al. saw a similar exothermic peak in their differential thermal analysis measurements on TiO2 particles from different phases at 280 °C . They found that this peak is due to the loss of water absorbed at the TiO2 particle surface. Therefore, the peak could be linked to a condensation process in the TiO2 shell and the release of the enclosed water molecules. However, we cannot unambiguously determine the onset of water loss in the IR spectra because the amount of released water is very low and cannot be separated from the background water bands at 3750 cm−1 and 1500 cm−1. The second heating ramp lead to a third prominent step at 400 °C (3). In this case, the degradation product is also CO2. In contrast to the pure PS particles, the last heating step to 700 °C results in a very small mass loss of only 2%. Since the corresponding IR data show no trace of CO2, all PS must already be decomposed at the end of the second isothermal step, and a further condensation reaction of the TiO2 is assumed. Calculations show that 34% material should be left over, which agrees well with the experimental data of 33%. We conclude that the TiO2 shell aids the decomposition reaction of the polymer even though the mass transfer is reduced. In our case, it is not possible to see the transformation of the amorphous to the anatase phase in the DSC curves. Xi et al. and Li et al. saw this event happening at temperatures above 400 °C [53, 54]. Therefore, the effect is likely to be superimposed by the exothermic degradation peak of PS.
These results show the importance of the right temperature profile for the calcination process. The overall particle shrinkage can be estimated by the mass loss of the PS particle at the first heating step. The mass loss translates into a volume shrinkage of the template particle. Assuming an isotropic shrinkage of the template particle, the diameter of the shrunk particle can be recalculated. The shrunk particle size then corresponds to the final hollow-core diameter. Starting with an initial particle with a diameter of 621 nm and considering a mass loss of 40%, this would result in a shrunk particle diameter of 520 nm. This is in reasonable agreement with the experimental data, where the hollow core has a size of ~ 500 nm. Calcinations in inert atmosphere, where the template particle is intact much longer, show less shrinkage of the hollow spheres (see SI Fig. 4). This observation strengthens the interpretation that the second peak in the DSC measurements is connected to the formation of the final and stable TiO2 structure, and therefore to the shell condensation process. Calcination profiles, which omit the mild calcination at 300 °C or which feature too fast heating ramps, sacrifice the shell integrity and result in collapsed structures (see SI Fig. 5).
The structures of the polystyrene, core-shell, and hollow particles Y (see Fig. 4e) were further characterized by SAXS and SEM/TEM. The measured SAXS data exhibit only weak features, which prevent a thorough fitting analysis. Furthermore, PS seed particles are too large to identify the radius in the experimentally reachable q range. Thus, based on the TEM result, we calculated the form factor of homogeneous spheres with a diameter of 621 nm and compared it to our experimental data – both agree well. The model has a Gaussian size distribution with a standard derivation of 10%, which is slightly higher compared to the SEM images due to instrumental smearing effects. The measurement of the PS@TiO2 core-shell particles is shown in Fig. 4e (green symbols). The scattering of PS@SiO2 can be described by the model of a homogeneous core-homogeneous shell . The sharp and well defined boundary between core and (monodisperse) shell leads to significant oscillations in the scattering data. In contrast to PS@SiO2 the shell of PS@TiO2 is less dense and highly particulate (see Fig. 3). As consequence of such a fractal-like shell morphology, the corresponding form factor scattering miss such pronounced oscillations. The main features are a q−4 scaling at intermediate q (ca. 0.008–0.05 Å−1) and a q−2 power law for q > 0.057 Å−1. A very weak oscillation around 0.017 Å−1 hints toward the expected dimension of the shell thickness of about 30–40 nm before calcination. The q−2 power law for q > 0.057 Å−1 is indicative of strong scatters with a mainly 2D-structure. This is in contrast to scattering patterns of similar (silica-based) core-shell systems [55, 56]. These systems exhibit a pronounced form factor and no q−2 scaling law at high q. The deviation of our system from these findings can be understood by the scattering contrast situation (PS ≈ 9.51∙10−6 Å−2, TiO2 ≈ 31.8∙10−6 Å−2, no solvent) and the lower bending curvature due to the large template particle diameter. At high q, the q−2 term seems to simply add up to the scattering of a pure PS sphere (q−4) underlining the fact that the shell is particulate. For comparison, the scattering of homogeneous core-homogeneous shell spheres is given (d = 621 nm, 10% Gaussian distribution and dshell = 35 nm, 20% Gaussian distribution; green dotted line PS@TiO2). During the calcination process, the amorphous TiO2 shells undergo a transformation into anatase phase (Fig. 4f). During this process the overall size of the particle shrinks. The experimental SAXS scattering pattern of these hollow spheres exhibits a minimum at q ≈ 0.026 Å−1. The corresponding correlation length of about 24 nm agrees well with the thickness of the TiO2 shell obtained from TEM analysis. The q−4 behavior at intermediate and high q reflects the contrast situation for hollow spheres. This is corroborated by the calculation of homogenous hollow spheres (blue dotted line; dinner = 621 nm, 10% Gaussian with zero contrast, dshell = 24 nm, 25% Gaussian with contrast 31.8∙10−6 Å−2). The q−2 power law at low q is attributed to the particulate shell, since rough surfaces can be considered as fractals.
As stated above, the TiO2 shell undergoes a phase transition during the calcination procedure. This can be seen in X-ray diffraction measurements in Fig. 4f. The core-shell particles (green line) do not show any features except for an amorphous halo. This pattern is caused by the TiO2 and the amorphous polymer core. After calcination, distinct peaks are visible (blue line). The inset shows the normalized data that agree very well with the expected diffraction pattern of anatase.
Size series of TiO2 hollow particles
We now want to highlight the robustness of our synthetic protocol. Therefore, TiO2 shells have been synthesized on PS particles A–E. Setting the MTC/Styrene ratio to 0.8mol% during the seed synthesis and the amount of TBT to 5.5∙10−4 to 6.5∙10−4 mol/m2 particle surface during the coating step, it is possible to fabricate TiO2 core-shell particles without secondary nucleation or particle clustering (Fig. 5). When using 800 μL of TBT, shell thicknesses between 45 and 60 nm could be achieved. Stable and smooth TiO2 shells could be immobilized on the PS seed particles. Also, the template particles themselves remain colloidally stable and are coated as individual objects. One may expect that the shell thickness decreases with increasing template particle diameter when employing the same amount of TBT precursor. This, however, would only be true, if the same particle concentration was used. Owing to our scalable process, we directly use the as-synthesized polymer dispersion, where both, particle diameter and particle concentration vary with the initial monomer concentration. Both parameters cancel each other out leading to comparable shell thicknesses for each batch.
Calcination of the particles in air leads to an isotropic shrinkage up to 20% compared to the core-shell size (see STEM, Fig. 5). For particles > 700 nm we find no systematic variation of the degree of shrinkage to the particle size. As a consequence, the TiO2 hollow spheres B, C, and D all have the same particle diameter of ~ 900 nm; particle E is slightly larger with 1025 nm. Yet, despite the considerable shrinkage, the particles retain their spherical shape and shrink in an isotropic fashion. This is even more remarkable as the ratio t/D between shell thickness (t) and particle diameter (D) is very small. t/D ranges from 4% for the smallest hollow spheres to 2% for the largest one. With decreasing t/R ratio the mechanical stability of the TiO2 hollow spheres decreases, which is also known for their silica shell counterparts . We, consequently, observed an increased portion of fractured or buckled hollow spheres from particles C to E.