Multiscale Colloidal Assembly of Silica Nanoparticles into Microspheres with Tunable Mesopores

Colloidal assembly of silica (nano)particles is a powerful method to design functional materials across multiple length scales. Although this method has enabled the fabrication of a wide range of silica‐based materials, attempts to design and synthesize porous materials with a high level of tuneability and control over pore dimensions have remained relatively unsuccessful. Here, the colloidal assembly of silica nanoparticles into mesoporous silica microspheres (MSMs) is reported using a discrete set of silica sols within the confinement of a water‐in‐oil emulsion system. By studying the independent manipulation of different assembly parameters during the sol–gel process, a design strategy is outlined to synthesize MSMs with excellent reproducibility and independent control over pore size and overall porosity, which does not require additional ageing or post‐treatment steps to reach pore sizes as large as 50 nm. The strategy presented here can provide the necessary tools for the microstructural design of the next generation of tailor‐made silica microspheres for use in separation applications and beyond.

they can easily be modified with many different surface-active groups. [3] However, not all types of molecules can be separated efficiently with the MSMs that are commercially available today. An important yet notoriously difficult class of molecules to separate are biomacromolecules such as peptides and antibodies, mainly due to their large size and diverse properties. [4,5] Separation of this class of molecules depends upon the availability of MSMs that can be specifically tuned to the size and shape of the macromolecule of interest, thus, requiring new technologically scalable approaches for manufacturing.
To this end, we focus on creating MSMs with highly tunable porosity characteristics, i.e., pore size, pore volume, and surface area, by sol-gel emulsion chemistry that is both versatile, easily reproducible, and scalable. The sol-gel reaction, i.e., the transformation of a suspension of silica nanoparticles (sols) into gels, can easily be trigged upon change of pH, temperature, or ionic strength. [6,7] Typical precursor solutions include tetraethyl orthosilicate [8][9][10][11][12][13] or sodium silicate (waterglass), [14,15] but in this work discrete silica nanoparticles are used as building blocks to form the gel. There are several benefits of using well-defined nanoparticles to form a gel. Gels made from particles do not need additional template molecules to guide the formation of a porous network, have great flexibility in terms of microstructural design and process scalability, and most importantly, allow a high level of control over the reaction.
To attain MSMs, the sol-gel reaction is confined within emulsion droplets. Emulsions are ideal systems to synthesize particles with a well-defined shape and an internal composition defined by the confined nanoparticle building blocks. There are various approaches to use emulsions to create higher order assemblies which can be roughly divided into methods that are dominated by: 1) the external confinement and process conditions, i.e., evaporation-driven colloidal assembly, [16][17][18][19][20][21][22][23][24] and 2) by the interaction forces between the nanoparticles, i.e., entropydriven or gelation-driven colloidal assembly. [25][26][27][28][29] In evaporation-driven assembly, the nanoparticles are forced into close contact by gradually removing the carrier liquid until the (nano) particles assemble into larger assemblies, which are sometimes called supraparticles, supracolloids, or supraballs. [22,24] In gelation-driven assembly, the nanoparticles are assembled into supraparticles by screening of the nanoparticle surface charges and by increasing the frequency of nanoparticle-nanoparticle

Introduction
Mesoporous silica microspheres (MSMs) have long been used as the stationary phase in high performance liquid chromatography for the separation and purification of molecules due to their versatile and tunable properties. [1,2] Silica microspheres have good mechanical strength, high thermal, and chemical stability and collisions. The main difference between the two methods is that evaporation-driven assembly usually leads to densely packed supraparticles, whereas gelation-driven assembly usually leads to porous supraparticles such as MSMs.
Due to current limitations in tuning the porosity characteristics during MSM synthesis, in many conventionally applied industrially relevant processes the as-gelled silica microspheres have to undergo one or multiple ageing or post-treatment steps. [30,31] A common ageing step that is frequently used is Ostwald ripening. In Ostwald ripening, the gelled microspheres are immersed in a liquid in which they are soluble and are heated to high temperatures for prolonged times in a sealed reactor. Material on the surface is then slowly dissolved and precipitated into regions of negative curvature, i.e., in between the necks between the nanoparticles and inside small pores. The result is a decrease in surface area and an increase in average pore size. [32,33] This is a time-consuming process although successful, it also leads to significantly wider pore size distributions as compared to MSMs that have been directly synthesized. [34] In this work, we describe how to create MSMs with a highly tunable porosity characteristics using a limited set of discrete silica sol particles as building blocks confined within waterin-oil (W/O) emulsion droplets. We show that by the choice of silica sol and mixtures thereof, careful manipulation of the gelation rate of the sol nanoparticles and the processing conditions, we can vary the surface area, pore volume, and average pore diameter of the MSMs over a much wider range than has been shown so far in literature. The presented approach does not require additional post-treatment steps such as ageing and is expected to ensure availability of MSMs for efficient separation of a wide variety of biomacromolecules in future.

General Mechanisms of Microsphere Formation
Perfectly spherical MSMs were synthesized via a protocol inspired by literature. [10,[35][36][37] In general, a W/O emulsion was prepared by adding a colloidal silica sol to an external oil phase containing emulsifier under rapid stirring. After emulsification, the gelation of the sol particles into MSMs was induced by either shrinking the emulsion droplets (dewatering) under vacuum conditions (evaporation-driven assembly) or by minimizing the water uptake of the oil phase and greatly increasing the reaction temperature in combination with high concentrations of salt (gelation-driven assembly). In this work, water uptake is defined as the increase in storage capacity of the oil phase as a function of temperature above its saturation point at room temperature. After gelation, the microspheres were removed from the oil phase via filtration and were subsequently dried and calcined to remove any organic residue still present in the MSMs. More detailed information can be found in the Experimental Section and Section S1 in the Supporting Information.
We hypothesize that MSM formation is dependent on two main parameters: 1) the rate of sol particle gelation r gel modulated by the sol particles used as building blocks and by the reaction conditions and 2) the rate of emulsion droplet shrinkage r shrink due to dewatering of the droplets. A proposed mechanism is shown in Figure 1.
At the start of the process, the colloidal silica particles are randomly distributed throughout the emulsion droplets. Then, depending on how quickly the sol particles within these droplets gel and how quickly water is transported out of the droplets, the resulting microspheres can either become very porous, very dense, or anywhere in between. If the gelation of the sol particles occurs significantly faster than the shrinking of the droplets (r gel >> r shrink ), i.e., when there is no shrinkage of the droplets and the assembly process is completely gelation-driven, the resulting microspheres can be very porous with a porosity as high as φ p = 0.70 and in theory as high as the total fraction of water in the sol. If the droplet shrinkage rate is significantly faster than the gelation rate (r gel << r shrink ), i.e., evaporation-driven assembly, the sol particles are compressed into dense spheres, with a porosity as low as φ p = 0.30. Microsphere porosities in-between these two values are achieved by a careful balance between the gelation rate and droplet shrinkage rate. Regardless of the chosen pathway, all microspheres are perfectly spherical. Scanning electron microscope (SEM) images of the surface of the microspheres reveal that the microspheres are completely assembled of smaller sol particles ( Figure S4, Supporting Information). A representative size distribution of the spheres is shown in Figure S5 in the Supporting Information.

Synthesis of MSMs
Our design strategy is based on three parameters that will be addressed below: 1) choice of the silica building blocks, 2) control over the water uptake of the oil phase, and 3) control over the sol particle mobility and stickiness. Full control over all three parameters is key to tailor the porosity characteristics of the resulting MSMs for different applications (e.g., separation of macromolecules).

Choice of the Colloidal Silica Building Blocks
The first design principle obtained in this work, which will be employed throughout the following sections, is the choice of the colloidal silica sol as building blocks. The choice of the sol has a direct influence on the internal morphology (size of the pores and struts) as well as the external morphology (specific surface area, SSA) of the microspheres. To build a large variety of microspheres, five colloidal silica sols with different particle sizes varying from 4 up to 100 nm were used (see also Sections S2.1-S2.3 and Figures S1 and S2, Supporting Information). With these five sols microspheres were synthesized with SSAs varying from 38 up to 560 m 2 g −1 , and, most notably, independent of the size or the porosity of the microspheres ( Figure S6, Supporting Information).
All sols used in this work are negatively charged and are highly stable, i.e., they do not settle or aggregate spontaneously for up to several months. [6,7] The stability of a sol can be described using the Derjaguin-Landau-Vervey-Overbeek (DLVO) theory. The DLVO theory provides a theoretical framework that describes the interaction potential between two charged surfaces in a liquid medium. [38] For a short summary on the DLVO theory, we refer to Section S2.4 and Figure S3 in the Supporting Information.
On first sight it appears that the smaller the size of the sol particles that are used as building blocks, the higher the porosity of the synthesized microspheres (Figure 2a). Smaller sol particles have a higher mobility due to Brownian motion. The frequency that two particles will collide and rapidly form a loose gel network (high porosity) is therefore higher for smaller particles. Moreover, smaller sol particles have a higher areal density of isolated silanol groups on the nanoparticle surface that can easily be deprotonated and react to form siloxane bridges. [6,39] The number of deprotonated silanol groups is directly related to the number of charges per nanoparticle. [40] To quantify the charge density of the different sols, the electrokinetic charge density versus the average sol particle size is shown in Figure 2b. The electrokinetic charge density is the charge density normalized at the shear plane, which can be derived from the zeta potential (see also Section S2.5 in the Supporting Information). [41,42] Figure 2b shows that the charge density clearly depends on the particle size. [40] The larger the nanoparticle, the lower the charge density. Our results are in good agreement with a theoretical study by Barisik et al. [43] and an experimental study by Shi et al., [44] who showed that the charge density decreases significantly for silica nanoparticles from 4 to 30 nm in size and becomes constant for nanoparticles above 30 nm. This behavior can be attributed to a difference in the type of surface silanol groups that are present on the nanoparticles. Nanoparticles with a diameter less than 30 have relatively high amount of isolated silanol groups that are far apart from each other due to the large particle curvature. These isolated silanol groups can be easily deprotonated, causing a steep drop in the charge density. Nanoparticles with a diameter larger than 30 nm have a relatively high amount of vicinal silanol groups, i.e., H-bridged silanol groups, which are not easily deprotonated. [39] These results show that the choice of the silica sol used as building block has important implications for the assembly behavior of the microspheres and thereby the microsphere porosity. In order to tailor the microsphere properties, it is, however, desirable to be able to tune the microsphere porosity without changing the silica building blocks, as shown in the next two sections.

Evaporation-Driven Assembly
One way of manipulating the microsphere porosity is by controlled shrinkage of the emulsion droplets at elevated temperature and reduced pressure, i.e., evaporation-driven assembly. Figure 2. a) Effect of sol particle size and reaction pressure on porosity. b) Electrokinetic charge density as a function of the particle size. Below a particle size of ≈30 nm the charge density is strongly dependent on the particle size. Above ≈30 nm the charge density becomes independent of the particle size. To illustrate this point, the porosity was compared for microspheres synthesized from the different sols at two different reaction pressures at a constant reaction temperature of T R = 65 °C (Figure 2a). Here, one can clearly see that the lower the reaction pressure during the evaporation step, the lower the microsphere porosity.
Following the classical Clausius-Clapeyron equation a lower absolute pressure equals a lower boiling point of water. [45] Here P 1 is the atmospheric pressure (1013 mbar), P 2 is the reduced pressure, T 1 is the standard boiling point of water, T 2 is the boiling point of water at reduced pressure, R is the universal gas constant, and ΔH vap is the enthalpy of vaporization, which can be calculated from [46] in which A, T c , and n are regression coefficients depending on the chemical compound. For water, A = 54 kJ mol −1 , T c = 647.13 K, and n = 0.34. [46] At P 2 = 200 mbar, T 2 = 60 °C. At P 2 = 100 mbar, T 2 = 45 °C. Therefore, at a lower absolute pressure and at constant reaction temperature T R , more thermal energy ΔT = T R − T 2 is available for the evaporation process. The result is that the larger the evaporation rate, the faster droplet shrinkage and, hence, more densely packed microspheres are obtained. To confirm, the amount of evaporated water over time was measured at both pressures and at a constant reaction temperature T R = 65 °C ( Figure S7, Supporting Information). It can be seen that at P = 100 mbar (ΔT = 20 °C), the evaporation rate is roughly two times larger than at P = 200 mbar (ΔT = 5 °C).

Gelation-Driven Assembly
While it is easy to synthesize microspheres with a low porosity by simply shrinking the emulsion droplets, it is not as straightforward to synthesize microspheres with a high porosity, especially if the sol particles are relatively large, i.e., larger than 20 nm. Therefore, in the next section, we do not shrink the droplets, but instead control the microsphere porosity via the reaction temperature and addition of salt. To illustrate this point, the resulting porosity for sol 3 is plotted at different reaction temperatures at atmospheric pressure and at a salt concentration of C salt = 0.19 ± 0.02 m (Figure 3a). It can be seen that the porosity of the MSMs increases with an increasing temperature, which can be explained by an increased particle mobility. However, above T R = 65 °C the porosity no longer increases even though the mobility of the sol particles still increases. The reason for this is that water uptake in the oil phase, phenethyl alcohol (PEA), also increases with increasing temperature (Figure 3b, inset). [47] This causes a shrinkage of the emulsion droplets during the gelation process and subsequently leads to more densely packed microspheres than one would expect based solely on the increased particle mobility.
To minimize the effect of water uptake, i.e., to minimize the droplet shrinkage during the gelation process, a mixed organic phase was used containing a nonpolar organic compound that has no affinity to water, i.e., mesitylene (MST). By adding a nonpolar organic compound to the oil phase, the uptake of water becomes adjustable (Figure 3b). Note that if the concentration of MST is too high, i.e., ≥ 75% of the total mixture, the emulsifier precipitates and no stable emulsion can be formed. The addition of a nonpolar organic compound, reducing water uptake and minimizing shrinkage of the aqueous droplets during gelling, allows the synthesis of MSM with significantly higher levels of porosity and pore sizes.
A high reaction temperature by itself is, however, not enough to access the whole range of microsphere porosities, especially if the particles are forced into close contract due to shrinkage of the droplets. If the electric double layer around each sol particle is too large, i.e., the particles are too negatively charged, the particles will not stick together and instead repel each other upon collision. A small amount of salt is required to reduce repulsion of the sol particles and to induce spontaneous gelation over time. [48,49] Higher concentrations of salt lead to a more compressed double layer (shorter Debye screening length) and therefore faster gelation of the sol particles (more stickiness). To illustrate this effect, the concentration of added salt was plotted versus the porosity of microspheres synthesized from sol 4 (Figure 4a). Figure 4 shows that the microsphere porosity initially increases rapidly with an increasing salt concentration. Above a certain salt concentration, in this case C salt > 0.20 m, the microsphere porosity no longer increases significantly. We hypothesize that once the majority of the electrical charges are shielded, additional salt has no effect on the probability that particles will successfully stick together upon particle-particle collision. Moreover, above C salt > 0.40 m the emulsion itself will become unstable and particles start to precipitate out of solution and form large aggregated structures. [50] The optimum salt concentration is dependent on the sol used as building blocks and requires some experimental optimization. The order of magnitude can, however, be estimated from macroscopic gelation experiments of the sols at room temperature or calculated using DLVO theory (Figure 4b). The best results are obtained when the macroscopic gelation time of the sol particles is roughly equal to the duration of the experimental procedure, which is typically a few hours.

Evaluation of the Pore Network
The SSA of the MSMs together with the porosity determines the average pore diameter of the microspheres and, therefore, the potential application of the synthesized MSMs. In Figure 5a, the average pore diameter of all synthesized MSMs from the five silica sols versus the microsphere porosity is shown. It can be clearly seen that the higher the porosity, the larger the average pore diameter. In addition, the larger the size of the sol nanoparticles used as building block, the larger is also the average pore diameter of the microspheres. From an application point of view, the width of the pore size distribution is, however, just as important as the average pore diameter. For example, in chromatography, the narrower the spread of the pore size distribution, the higher the peak resolution that can be obtained. [51] The obtained width of the pore size distri-bution always had a coefficient of variation of CV = σ/μ = 0.4, regardless of the total porosity or the silica sol used ( Figure S8, Supporting Information), indicating good reproducibility and control during synthesis.
To evaluate the gas adsorption behavior between samples, microspheres with the same porosity (φ p = 0.67) but constructed from different sols were compared. The results are shown in Figure 5b. Every sample displayed IUPAC type IVa isotherms, which is characteristic for adsorption behavior inside mesoporous adsorbents. [52] At low pressure p/p° there is a region of mono-multilayer adsorption on the mesopore walls, followed by pore condensation and finally a horizontal saturation plateau, indicating completely filled mesopores. The onset of the condensation increases, and the length of the saturation plateau decreases, for microspheres synthesized from sol 1 to sol 4. This is logical because the average pore diameter of the microspheres of sol 1 is smaller than that of sol 4, resulting in earlier and faster saturation of the complete pore network. The pore condensation regions of samples 2 to 4 have a partially common pressure range (p/p° = 0.85-0.90) and partially overlapping pore size distributions (5-15 nm). This is because pores of the same size have the same condensation pressure. [53] Pore condensation is followed by a hysteresis loop where the desorption branch becomes parallel to the adsorption branch. The hysteresis loops become sharper and steeper ranging from sol 1 to sol 4. In general, the steeper the loop, the faster the nitrogen can evaporate from the pore network. Interpretation of the shape of a hysteresis loop is, however, not straightforward because the desorption branch is dependent on various network effects and pore blocking. [52] The pore size distributions of the compared samples show some overlap at the base of the profiles because the silica nanoparticles used as building blocks are not monodisperse and share partially similar size fractions (Figure 5c). Each sample, however, displays a narrow pore size distribution, which is significantly smaller than pore size distributions that can be obtained when additional ageing steps such as Ostwald ripening are used. [34] SEM shows that microspheres that were treated with an Ostwald ripening process have a much more uneven distribution of pores over the surface ( Figure S9, Supporting Information). MSMs that have been formed via direct nanoparticle assembly have an even distribution of pores across the surface.
These results clearly show that microspheres obtained via the presented method have better defined porosity characteristics and that the process in general is significantly more controllable than comparable process that requires post-treatment steps, e.g., to tune pore sizes.

Fine Tuning via Sol Particle Mixing
In above sections, the general principles of tuning pore size and surface area of MSMs have been discussed. Above methodology can be further extended using mixtures of silica sols to very precisely tune the microsphere properties. By mixing two sols with different particle sizes, the resulting microspheres will have an average surface area based on the ratio of the mixture. By tuning the surface area, the pore size distribution of the microspheres can be shifted without changing the total porosity. As an example, microspheres from a 1:2 mixture of sol 2 and sol 3 were synthesized. The resulting pore size distribution of the mixture is located precisely in between the two original sols (Figure 6a). The distribution profile overlaps at the base with the two original sols because the microspheres are build-up from both particle size fractions. However, the majority of the pores clearly have a pore diameter in-between that of the two original sols.
Furthermore, in select cases mixing two sols can yield MSMs with properties that are otherwise unobtainable. The largest average pore diameter that was obtained from one building block, i.e., sol 5, was 35 nm with a microsphere porosity of φ p = 0.42. Attempts to increase the average pore diameter further have not been successful because the larger the nanoparticle building block, the more difficult it becomes to obtain high porosities. By adding a small portion of another sol, microspheres can be obtained with higher porosities and subsequently larger pore sizes. To illustrate this point, 2 wt% of sol 1 was mixed with sol 5. The resulting microspheres have a porosity of φ p = 0.52 and an average pore diameter of 50 nm, which is significantly larger than what could be obtained from just one sol (Figure 6b). The microspheres are still perfectly spherical but are not as densely packed as without the addition of sol 1 ( Figure S10, Supporting Information). This shows that, with a limited set of silica nanoparticles, a large variety of MSMs can be synthesized with precise control over pore size, porosity, and SSA as needed for a specific application.

Conclusions
We have shown that the colloidal assembly of silica nanoparticles into MSMs can be precisely manipulated by actively and passively controlling the reaction conditions and environment in which the sol-gel reaction takes place. With these insights into the underlying mechanisms of colloidal assembly in hand, we introduced a strategy to the design of perfectly spherical MSMs with precisely tunable porosity characteristics across multiple length scales that are highly reproducible and scalable using only a limited set of discrete colloidal silica sols as building blocks. In contrast with conventionally applied industrially relevant processes, we demonstrated that we can synthesize MSMs with highly tunable porosity characteristics without the need for additional ageing or post-treatment steps and/or additional template molecules to guide the formation of the porous network. The results shown here are expected to have a significant impact and ensure availability of MSMs for use in different applications in future, e.g., for the separation of a wide variety of biomacromolecules. [54]

Experimental Section
Materials: The colloidal silica sols used in this work were provided by Nouryon Pulp and Performance Chemicals AB, Sweden and consist of colloidal silica nanoparticles (sol) of different sizes and concentrations in water ( Table 1). All sols were ammonium stabilized, which means they had been brought to a pH of 8-10 by the addition of ammonia (25% w/w, Scharlau). The sols 3A, 4A, 4B, and 5A were diluted with deionized water that contained sufficient ammonia in order to ensure the concentration of ammonium ions was the same as the pre-diluted sol. A detailed characterization of the nanoparticles is shown in Section S2 in the Supporting Information. PEA (99%) and MST (97%) as the oil phase and hydroxypropyl cellulose (HPC, average M w 1 00 000 g mol −1 ) as a stabilizer and emulsifier were purchased from Acros Organics. A 5 m stock solution of salt was prepared of ammonium acetate (Merck). All chemicals were used as received without further purification. The water used in this work was deionized by a Milli-Q Advantage A10 system (Merck Millipore) and had an electrical resistivity of 18.2 MΩ cm at 25 °C.
Solvent Characteristics: The water content in the saturated mixed organic phase was determined by means of the Karl-Fischer volumetric method using a Mettler-Toledo DL38 Karl Fischer titrator. Milli-Q water and organic phase (PEA and MST) were vigorously stirred in a 100 mL glass vial for 12 h and subsequently settled at room temperature for at least 24 h. Then the saturated solution was decanted in a separatory funnel and samples were taken for analysis.
Synthesis Route: Three different experimental protocols were carried out to produce silica microspheres. A detailed description of the protocol, amounts of materials used, and properties of the resulting microspheres are listed in Section S1 in the Supporting Information.
Microsphere Characterization: SSAs, pore volumes, and pore size distributions of the synthesized samples were determined from nitrogen sorption isotherms (Micromeritics TriStar 3000). The isotherms were measured at −196 °C. The SSA was calculated from the monolayer adsorbed gas quantity in the pressure interval p/p° = 0.05-0.22 using the Brunauer-Emmett-Teller (BET) equation. [55] The pore volume and pore size distribution were calculated from the desorption isotherm using the Barrett-Joyner-Halenda model. [52,56] Every sample displayed IUPAC type IVa isotherms, indicating adsorption behavior inside mesoporous adsorbents. [52] The microsphere porosity was calculated from the total pore volume of the microspheres and the density of amorphous silicon dioxide [57] φ ρ = + 1 pore SiO pore where V pore is the total pore volume of the particles and ρ SiO2 is the density of amorphous SiO 2 , which is assumed as 2.2 g cm −3 . [6] SEM images of select microspheres were obtained using an FEI Quanta 3D equipped with a field emission electron gun operating at 5 kV. The particles were deposited on an SEM-stub and sputter-coated with a 20 nm layer of gold (Emitech K550) to prevent charging.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.