Designer Core-Shell Nanoparticles as Polymer Foam Cell Nucleating Agents: The Impact of Molecularly Engineered Interfaces.

The interface between nucleating agents and polymers plays a pivotal role in heterogeneous cell nucleation in polymer foaming. We describe how interfacial engineering of nucleating particles by polymer shells impacts cell nucleation efficiency in CO2 blown polymer foams. Core-shell nanoparticles (NPs) with a 80 nm silica core and various polymer shells including polystyrene (PS), poly(dimethylsiloxane) (PDMS), poly(methyl methacrylate) (PMMA), and poly(acrylonitrile) (PAN) are prepared and used as heterogeneous nucleation agents to obtain CO2 blown PMMA and PS micro- and nanocellular foams. Fourier transform infrared spectroscopy, thermogravimetric analysis, and transmission electron microscopy are employed to confirm the successful synthesis of core-shell NPs. The cell size and cell density are determined by scanning electron microscopy. Silica NPs grafted with a thin PDMS shell layer exhibit the highest nucleation efficiency values, followed by PAN. The nucleation efficiency of PS- and PMMA-grafted NPs are comparable with the untreated particles and are significantly lower when compared to PDMS and PAN shells. Molecular dynamics simulations (MDS) are employed to better understand CO2 absorption and nucleation, in particular to study the impact of interfacial properties and CO2-philicity. The MDS results show that the incompatibility between particle shell layers and the polymer matrix results in immiscibility at the interface area, which leads to a local accumulation of CO2 at the interfaces. Elevated CO2 concentrations at the interfaces combined with the high interfacial tension (caused by the immiscibility) induce an energetically favorable cell nucleation process. These findings emphasize the importance of interfacial effects on cell nucleation and provide guidance for designing new, highly efficient nucleation agents in nanocellular polymer foaming.

Introduction

Foaming of polymers is a widely used process to obtain porous polymeric materials with high specific surface area and low density.[1] Due to their porosity, polymer foams have been successfully used in numerous applications, such as packaging, energy absorption, acoustic and thermal insulation, catalyst carriers, and tissue engineering.[2−6] Fabrication of nanocellular foams is of great interest since they exhibit unique mechanical strength combined with unusual properties, such as high thermal insulation capacity, when compared to traditional foams.[7,8] For instance, when closed cell foams feature cell sizes that are smaller than the collision mean free path of the encapsulated gas molecules (e.g., ∼70 nm under standard conditions), the thermal conductivity of the foams can be significantly reduced due to the so-called Knudsen effect.[7,9] Obviously, polymer foams with these properties would be very promising to be used as high-performance thermal insulation materials.[5,9] However, the preparation of foams with such small cells and with high cell density remains a scientific and technological challenge.

Physical foaming using CO2 as blowing agents has become one of the most promising strategies to prepare nanocellular materials,[10,11] due to its industrial scalability, cost-effectiveness, and eco-friendliness.[12] Fabrication of nanocellular foams with small cell size (e.g., <100 nm) and high cell density (e.g., >10[15] cells cm–3) by CO2 requires high cell nucleation efficiency and reduced cell coalescence, which remains a grand challenge for processing. To promote cell nucleation and improve the uniformity and size control of cells, a commonly adopted strategy is to introduce nanostructured heterogeneous phases to act as heterogeneous nucleation sites in the foamed matrix.[13] According to the classical nucleation theory (CNT), heterogeneous cell nucleation would be preferable due to lower nucleation energy barriers when compared to homogeneous nucleation.[14,15] For instance, foaming of polymers containing block (co)polymer micelles[16−21] and (nano)particulate (fillers)[21−32] as heterogeneous nucleation agents have been reported in the open literature.

Altstädt and coworkers[21] described polymer foams obtained by structuring immiscible polymer blends of poly(2,6-dimethyl-1,4-phenylene ether) (PPE) and poly(styrene-co-acrylonitrile) (SAN). The authors demonstrated that these blends can be successfully compatibilized by Janus particles, which leads to diameter reduction and fine dispersion of PPE domains. This resulted in the decrease of cell size and increase of cell density due to the fact that smaller PPE domains enhance the cell nucleation density for foaming. Rodríguez-Pérez and coworkers[16] reported that the addition of triblock copolymer [poly(methyl methacrylate)-block poly(butyl acrylate)-block-poly(methyl methacrylate)] (MAM) in CO2-assisted poly(methyl methacrylate) (PMMA) foaming can decrease the cell size and significantly increase the cell density when compared to neat PMMA foams. The authors ascribed the enhanced cell morphology to the increased cell nucleation on the MAM nanostructures due to a combination of their high CO2-philicity, favorable surface tension, and phase-separated morphology.

Compared to polymer particulates/micelles, silica nanoparticles (NPs) are of particular interest as heterogeneous nucleation agents in polymer foaming due to their low cost, easy preparation, size control, and the ease of employing various functionalization strategies for their surface decoration.[33,34] Goren et al.[24] demonstrated that the addition of silica NPs in polycarbonate prior to foaming resulted in a reduced cell size, increased cell density, and provided a more uniform cell size distribution due to the preferred heterogeneous nucleation on silica particles. In order to increase cell nucleation events on silica NPs, a commonly adopted method is to modify their surface chemistry. For instance, Zhong and coworkers[35] reported that the surface derivatization of silica NPs with poly[2-(methacryloyloxy)ethyl]trimethylammonium tetrafluoroborate (P[MATMA][BF4]) leads to a higher heterogeneous nucleation efficiency in foaming compared to amino-functionalized silica particles, which is ascribed to the high CO2-philicity of (P[MATMA][BF4]). Ozisik and coworkers[24] demonstrated that fluorination of silica NP surfaces can decrease the value of nucleation energy barriers and significantly enhance the cell nucleation efficiency in CO2 PMMA foaming compared to unmodified silica particles. In addition, we recently reported that silica NPs grafted with a thin poly(dimethylsiloxane) (PDMS) shell can significantly decrease cell size and increase cell density in CO2-assisted polymer foaming, due to the high CO2-philicity and low surface energy of the PDMS shell.[22,33]

Strategies to enhance polymer foam cell nucleation of silica NPs, for example, by engineering their surface chemistry,[36] have been well described in the open literature; however, few studies have focused on the properties of the modified interface and its influence on cell nucleation. In this work, we aim at elucidating the influence of the choice of the shell material on cell nucleation. Experimentally, we used core–shell NPs with an 80 nm silica core decorated by various thin polymer shell layers [including polystyrene (PS), PDMS, PMMA, and poly(acrylonitrile) (PAN)]. Core–shell particles were synthesized and used as heterogeneous nucleation agents in CO2-assisted PMMA and PS foaming. For surface engineering, polymer shell layers were chemically attached on the silica particles by the “grafting from” surface modification approach, that is, surface-initiated atom transfer radical polymerization (SI-ATRP),[37] as well as by grafting to methods. Molecular dynamics simulations (MDS) were employed to reveal the effect of incorporating different polymer shell layers on gas partitioning into the shell layer and at the interface. The obtained results demonstrate that altering the polymer shell structure not only influences the interfacial compatibility but also has a strong influence on the CO2 density profile at the interfaces, which significantly affects the cell nucleation and cell morphology in polymer foaming.

Results and Discussion

Designer NP Synthesis and Characterization

The preparation steps of the various functionalized NPs are shown in Figure . Stöber silica NPs with a diameter of 80 nm were obtained, followed by their surface decoration with PS, PMMA, PAN, and PDMS grafts, respectively. We chose 80 nm as the diameter of our working NPs because based on our previous work,[22] this is the optimum size for foam cell nucleation when the nucleation density and nucleation efficiency are simultaneously considered at the given batch foaming conditions. The reaction scheme is depicted in Figure a. Silica NPs were prepared via a Stöber reaction (step 1), followed by the hydrolysis of the surface-exposed ethoxy groups to silanol moieties (step 2). Subsequently, the hydrolyzed particles were modified with (3-aminopropyl)-triethoxysilane (APTES), resulting in the formation of amino-functionalized NPs (step 3). Following the reaction with α-bromoisobutyryl bromide (step 4), NPs grafted with a shell layer of PS (SiO2–PS), PMMA (SiO2–PMMA), and PAN (SiO2–PAN) were obtained by SI-ATRP (step 5) of the respective monomers. PDMS grafted core–shell NPs (SiO2–PDMS) were synthesized by the “grafting to” approach of tethering monoglycidyl ether-terminated poly(dimethylsiloxane) (PDMS-G) (step 4′) to the amino-functionalized silica NPs.

Figure 1

(a) Schematic of the NP preparation process. (b) Single reflection attenuated total reflection–Fourier transform infrared (ATR–FTIR) absorbance spectra and (c) non-isothermal thermogravimetric analysis (TGA) thermograms of the bare, SiO2–PS, SiO2–PMMA, SiO2–PAN, and SiO2–PDMS NPs with a (silica core) diameter of 80 nm. The black arrows in the FTIR spectra indicate characteristic FTIR absorbance bands of the (modified) silica NPs.

Figure b shows FTIR absorbance spectra of the (modified) silica NPs. The remaining ethoxy groups following the Stöber reaction of tetraethyl orthosilicate (TEOS) are clearly observed in the FTIR spectra, that is, the CH2/CH3 bending absorption band at 1452 cm–1 and CH2/CH3 absorption band at 2980 cm–1.[38] After hydrolysis, these absorbance bands disappeared (data not shown). The absorption bands at 1452 cm–1 (ascribed to C=C stretching vibrations) and near 3000 cm–1 (ascribed to aromatic and aliphatic C–H stretching) indicate the successful grafting of PS from silica NPs.[39] The absorption bands for CH3 stretching at 2967 cm–1 and for C–H bending at 1263 cm–1 confirm the successful attachment of PDMS to silica NPs.[40] The absorption bands at 1730 and 1271 cm–1, which are ascribed to the stretching of carbonyl group and C–O, respectively, indicate the successful grafting of PMMA.[41] The absorption peaks at 2244 and 2940 cm–1 are assigned to the vibration of the nitrile group and the stretching vibration of the −CH2 groups in PAN, respectively, which confirms the successful grafting of PAN.[42]

TGA was used to determine the amount of polymer grafted. Figure c shows the weight loss versus temperature curves for non-isothermal TGA measurements of bare SiO2, SiO2–PS, SiO2–PMMA, SiO2–PAN, and SiO2–PDMS. The weight percentage of PS, PMMA, PAN, and PDMS covalently bound to SiO2 NPs was determined to be ∼12.0, ∼14.1, ∼17.7, and ∼5.0 wt %, respectively, from mass loss values. Based on the TGA results, the molar mass (measured by GPC) of grafted polymer chains, and the surface area of the used SiO2 NPs (33 m2 g–1), the grafting density of PS, PMMA, PAN, and PDMS was calculated to be ∼0.45, ∼0.42, ∼0.43, and ∼0.91 nm–2, respectively (see Figure S1). The significantly higher grafting density of PDMS, compared to that of the other grafted polymer chains, can be attributed to the high molar concentration of free PDMS chains by using the melt “grafting to” method.[43−45]

Transmission electron microscopy (TEM) was used to confirm the core–shell structure of the hybrid NPs. Figure shows TEM images of bare and polymer-grafted NPs. A clear polymer shell structure around the silica core can be observed (see Figure b–f). From the TEM images, the shell thickness values were estimated to be 7.0 ± 2.1, 8.3 ± 2.4, 9.6 ± 2.8, and 5.0 ± 1.3 nm for SiO2–PS, SiO2–PMMA, SiO2–PAN, and SiO2–PDMS, respectively, as shown in Figure S2. The NPs obtained were subsequently used as heterogeneous nucleation agents for PS and PMMA nanocomposite foaming.

Figure 2

TEM images showing the structure of the bare and the surface-functionalized core–shell NPs. (a) Bare SiO2, (b) SiO2–PS, (c) SiO2–PMMA, (d) SiO2–PAN, and (e,f) SiO2–PDMS with a silica (core) diameter of 80 nm.

Microcellular and Nanocellular Foams

Prior to foaming, the NPs were melt-blended with PS and PMMA. Subsequently, the nanocomposites were compression molded to form films with a thickness of typically 200 μm (for comparison, we kept the volume number density of the particles with different surface chemistry constant at the value of 7.5 × 1013 particles cm–3).

PS and PMMA nanocomposites with bare and core–shell NPs were foamed after saturation (for 4 h) with CO2 at 55 bar over a period of 30 s; the foaming temperature for PS was 115 °C and for PMMA, it was 40 °C.

Figure shows scanning electron microscopy (SEM) images of cross-sectioned PS and PMMA foams with/without particles after foaming. For both foam matrices, after incorporation of NPs, the cell size decreases and the cell density increases compared with the foams containing no particles. From Figure b–f, it is obvious that for PS nanocomposite foams, the incorporation of SiO2–PMMA, SiO2–PAN, and SiO2–PDMS NPs can significantly decrease the cell size and increase the cell density compared with untreated silica and SiO2–PS particles. PMMA nanocomposite foams that contain SiO2–PS, SiO2–PAN, and SiO2–PDMS particles have smaller cell sizes and higher cell densities compared to that of the foams with SiO2–PMMA and SiO2 (as shown in Figure h–l). For a quantitative comparison, the values of the cell size and cell density of representative PS and PMMA foams were determined and the results are shown in Figure .

Figure 3

SEM images showing the microstructures of the cross-sectioned PS nanocomposite foams containing (a) no NP, (b) bare SiO2, (c) SiO2–PS, (d) SiO2–PMMA, (e) SiO2–PAN, and (f) SiO2–PDMS. SEM images of the PMMA-based nanocomposite foams containing (g) no NP, (h) bare SiO2, (i) SiO2–PMMA, (j) SiO2–PS, (k) SiO2–PAN, and (l) SiO2–PDMS.

Figure 4

(a) Cell size, (b) cell density, and (c) nucleation efficiency of PS nanocomposite foams nucleated via the designer NPs. In the second row, we zoom in on the graphs of the first row. (d) Cell size, (e) cell density, and (f) nucleation efficiency of PMMA nanocomposite foams obtained by the designer NPs.

For comparison, the cell size and cell density values for PS foams without particles were approximately 26 μm and 1.7 × 106 cells cm–3, respectively, and for pristine PMMA foams, these were approximately 13 μm and 3 × 108 cells cm–3, respectively, as shown in Figure S3. Thus, PS and PMMA foams without added NPs have a larger cell size and lower cell density compared to the foams containing NPs, as shown in Figure , which indicates that the addition of the NPs has a substantial nucleation effect.

It can be clearly observed in Figure a,b that PS foams that contain SiO2–PDMS have the smallest cell size (∼390 nm) and the highest cell density (2.13 × 1013 cells cm–3). In addition, SiO2–PMMA particles give rise to smaller cell sizes and higher cell densities compared to SiO2 and SiO2–PS. Thus, the question arises: how does the choice of the shell material influence foam cell size and how cell size variation effects can be explained? The decrease in cell size and the increase in cell density with SiO2–PDMS and SiO2–PMMA can be ascribed to the enhanced CO2-philicity of PDMS and PMMA shells, which agrees well with our previous reported results.[22,33] Strikingly, PS foams with SiO2–PAN feature smaller cell size and higher cell density values compared to bare and PS-grafted SiO2. Since PAN is CO2-phobic (the measured CO2 absorption for PAN saturated at 55 bar for 4 h is nearly 0%), SiO2–PAN is expected to be less efficient as a nucleation agent compared to SiO2–PS. This result will be discussed in the next section.

Compared with PS-based foams, PMMA foams nucleated by SiO2, SiO2–PS, SiO2–PMMA, and SiO2–PAN NPs show higher cell density and smaller cell size (Figure ). Moreover, the impact of the variation of the induced polymer shell layers on the foam cell morphology is significantly reduced (Figure ). This is ascribed to a higher CO2-philicity of PMMA (∼18 wt % at 55 bar and room temperature) compared to PS (∼7 wt %), leading to a lower nucleation energy barrier in foam cell nucleation. Similar to PS foams, SiO2–PDMS-nucleated PMMA featured the smallest cell size (∼480 nm) and highest cell density (∼1.5 × 1013 cells cm–3). This is attributed to the high CO2-philicity (∼75 wt %) and low surface tension of the PDMS layer.[36,46−48] For PMMA foams nucleated by SiO2 and SiO2–PMMA, cell size and cell density values are comparable and in the range of 1 μm and 3.5 × 1012 cells cm–3, respectively. Foams with SiO2–PS and SiO2–PAN have a smaller cell size and higher cell density values compared to the SiO2 and SiO2–PMMA-nucleated cases. For example, SiO2–PAN exhibits cell size and cell density values of approximately 560 nm and 1.3 × 1013 cells cm–3, which is significantly enhanced compared to SiO2 and SiO2–PMMA. Considering the higher CO2-philicity of PMMA compared to PS and PAN (∼0 wt % absorption), the striking effect of SiO2–PS and SiO2–PAN on cell dimensions is ascribed to the influence of interfacial interactions, which will be discussed in the following section.

Figure c and 4f show the nucleation efficiency of NPs in PS and PMMA foaming, respectively. The nucleation efficiency value is defined as the ratio of the number of cells per cm3 of the originally unfoamed polymer to the number of NPs per cm3 of unfoamed polymer. (We consider here unfoamed material as the cell number considered here does not include the foam expansion factor). It is assumed that (i) there is no cell coalescence during foaming, and (ii) that every particle provides one potential nucleation site. However, we note that the number of nucleation sites per particles is not limited to one. In principle, there are no physical restrictions that prevent the occurrence of more than one nucleation event per particle, that is, nucleation efficiencies exceeding unity are possible.[46]

The nucleation efficiency of PS foams with SiO2–PDMS is significantly higher compared to that of the other types of particles, as shown in Figure c. For instance, a nucleation efficiency of ∼0.28 was obtained for the PDMS-decorated silica, which is 100 folds higher compared to the values observed for the SiO2, SiO2–PS, SiO2–PAN, and SiO2–PMMA (nucleation efficiency <2.84 × 10–3). The SEM images, as shown in Figure f, reveal that SiO2–PDMS featured significant smaller cell size and higher cell density compared to PS foams with other particles (Figure b–e), which is due to the high nucleation efficiency of SiO2–PDMS. Additionally, PS foams containing SiO2–PMMA and SiO2–PAN feature nucleation efficiency values of 2.84 × 10–3 and 2.30 × 10–3, respectively, which are much higher compared to that of SiO2–PS (3.8 × 10–5) and SiO2 (2.9 × 10–5). The higher nucleation efficiency of SiO2–PAN compared to SiO2–PS is ascribed to the weaker interfacial interaction between PAN and PS, which will be discussed later.

Strikingly, in PMMA foaming, SiO2–PMMA shows a comparable nucleation efficiency to SiO2 with values of ∼0.047, which is lower when compared to SiO2–PS and SiO2–PAN that featured values of ∼0.07 and ∼0.17, respectively. This is mainly due to the weak interactions and the incompatibilities at the interface between the polymer shell (i.e., PS and PAN) and PMMA matrix. Similar to PS foams, PMMA foams with SiO2–PDMS shows the highest nucleation efficiency values of ∼0.21 compared to that of the other types of particles. This further confirms the energetically favorable nucleation on SiO2–PDMS NPs due to the low surface energy of the PDMS shell and the higher local CO2 concentration (∼75 wt %)[48] in the shell. In addition, in a recent study, we introduced phase-separated PDMS domains in a PMMA matrix by blending PMMA–PDMS–PMMA triblock copolymers with PMMA. The several tens of nm sized phase-separated PDMS domains, that is CO2-philic domains, were extremely effective in nucleating foam cells in the PMMA matrix (see Supporting Information, Figure S4). This further confirms that CO2-philic domains locally enhance CO2 concentration, which is favorable for nucleation, even without the close proximity of a nucleating surface.

Furthermore, from the above it is clear that cell nucleation is energetically unfavorable on silica NPs that feature a polymer shell layer of the same chemical composition as the polymer matrix. This contrast in nucleation efficiency is due to the different interfacial properties.[21]

MDS and Impact of Molecularly Engineered Interfaces on Foam Cell Nucleation

To obtain a microscopic interpretation of the effect of the CO2-philicity of the shell and the matrix as well as the molecular interactions between the shell and matrix on gas partitioning, we performed MDS. To do so, we set up a generic model that incorporates dispersed Lennard-Jones (LJ)[49] beads as gas molecules (representing CO2), a polymer brush representing the shell layer (polymer length N = 30 and grafting density 0.45 chains/σ2) in contact with a polymer matrix (degree of polymerization P = 100) using a bead-spring model.[50] This model captures the physicochemical behavior of polymers and can be universally applied to different types of polymers.[50] We can mimic particular brush, matrix, and gas combinations by tuning the relative affinity of the different components via the strength of their interactions. Since the brush thickness is much smaller than the particle radius, we can neglect curvature effects on density profiles.[51,52] Therefore, we study local density distributions for brushes attached to flat walls. To mimic contact between the shell and a large matrix, we chose conditions with a constant gas density of 0.015 σ–3 in the polymer matrix. To mimic situations for NPs with different shell structures as nucleation agents, the miscibility of both the gas molecules and the polymer matrix with the different NP shell layers was tuned by altering the strength of the molecular interactions via ε in the LJ potential, such that they can be compared qualitatively to the experimental systems. The interactions between the gas molecules and the matrix were kept constant. Details on the exact interactions and simulation setup can be found in the Materials and Methods section. The results for four chemically different NP shell layers are shown in Figure .

Figure 5

Effect of interfacial compatibility on the density profiles for CO2-mimicking gas particles obtained from MDS for: (a) NPs engineered with a thin polymer shell layer featuring the same chemical composition as that of the polymer matrix (comparable to the PS shell in the PS matrix). (b) NPs with a polymer shell that features higher gas-philicity (compared with the polymer matrix) and reduced compatibility with the polymer matrix (comparable to the PMMA shell in the PS matrix). (c) NPs designed with a thin polymer shell layer that has purely repulsive interactions with both the gas molecules and the polymer matrix (mimicking the PAN shell in the PS/PMMA matrix). (d) Further increase of the gas-philicity of the shell [based on case (b)] and purely repulsive interactions between polymer matrix of the polymer shell layer (mimicking the PDMS shell in the PS/PMMA matrix). Schemes in (a′), (b′), (c′), and (d′) visualize the local distribution of gas and the interfacial compatibility between the shell layer and polymer matrix for the four different cases described in (a), (b), (c), and (d), respectively. Schemes in (e), (f), (g), and (h) indicate potential locations and types of foam cell nucleation for the four different cases described in (a), (b), (c), and (d), respectively.

Figure a shows the situation in which the shell layer has the same chemical composition as the polymer matrix. It is clear from the overlap between the polymer matrix (green) and shell layer (blue) that the shell layer and matrix are miscible and that they mix in the top layer of the brush alone. This reduced compatibility between chemically identical polymer brushes and matrices might seem surprising, but it has the same origin as autophobic dewetting[53−55] and has been observed in simulations by others performed under similar conditions as well.[56] The reason for this is that the relative long matrix polymers do not gain enough translational entropy by mixing to compensate for the entropic penalty for stretching the brush polymers at these relatively high grafting densities. Experimentally, we have conditions (PS grafting density of 0.45 nm–2 and molar mass of the PS matrix polymers Mw = 230,000 g mol–1) that result in autophobic dewetting.[53]

The red line in Figure a represents the density of gas molecules, and it shows that the gas is nearly homogeneously distributed in the material at a constant density of 0.015 σ–3, despite that the matrix and the shell only partly mix. This uniform distribution implies that during cell nucleation, the core–shell NPs coated by shells that are chemically identical to the matrix show similar CO2 partitioning as bare silica NPs. This explains why SiO2–PS NPs in PS matrices and SiO2–PMMA NPs in PMMA matrices have comparable nucleation efficiency values to their bare counterparts in PS and PMMA matrices, respectively.

Upon increasing the gas-philicity of the shell layer by increasing the interaction strength between the gas and shell units ε from 1 to 1.5 and by reducing the miscibility between the shell and the matrix by decreasing their interaction strength ε from 1 to 0.5, the matrix-brush overlap in the interface region and gas density profile change significantly, as shown in Figure b. This system mimics conditions comparable to PMMA shells in a PS matrix. The overlap between the brush and the matrix has reduced to being only one molecular diameter σ due to weaker interaction between the shell and the matrix. Interestingly, accumulation of gas molecules at the interface was observed in this case (see Figure b), resulting in a local density of more than 0.07 σ–3. However, the gas density peak appears to reside mainly in the more gas-philic shell phase. According to the CNT, a high local gas concentration will decrease the height of the nucleation energy barrier at the interface. Furthermore, the weaker affinity between shell layer and matrix means higher interfacial tension at the interface, which further decreases the nucleation energy barrier and facilitates bubble nucleation and growth at the interface.[21] This is in agreement with our experimental observation of a higher nucleation efficiency of SiO2–PMMA in PS foams compared to SiO2–PS and SiO2 since the PMMA shell layer is more CO2-philic and less compatible with the PS matrix compared to the PS shell layer.

Figure c shows the situation in which the grafted polymer shell layer features purely repulsive interactions with the gas as well as the polymer matrix, which imitates the case of PS (or PMMA) foams nucleated by SiO2–PAN. Due to the incompatibility between the shell and the matrix (see Supporting Information, Figure S5), there is a clear boundary/interface with an overlap of around one σ between them. Moreover, a high accumulation of gas is observed at the interfaces on the matrix side (peak-height of 0.12 σ–3, see Figure c) and no gas penetrates into the shell layer. In the experiments, the presence of such an interface will decrease the nucleation energy barrier and accelerate the CO2 bubble nucleation rate from the interfacial area due to enriched CO2 concentration. This explains that SiO2–PAN featuring higher nucleation efficiency, when compared to SiO2–PS and SiO2–PMMA in PS and PMMA foams, respectively.

In order to mimic the interactions between SiO2–PDMS, CO2, and polymer matrix (e.g., PS), we increased the strength of the interaction between the shell layer and the gas molecules to ε =2 and we made the interactions between the shell layer and the polymer matrix purely repulsive by excluding attractive interaction within the interaction potential. Strikingly, a substantial reduction of CO2 at the interface area was observed and the CO2 molecules accumulated densely in the shell layer, as shown in Figure d. The increased gas entrapment indicates a further reduction of the nucleation barrier and allows for the growth of cell nucleation efficiency values. This can explain the superior performance observed for PDMS shells in PS/PMMA matrices compared to the other shells (see Figure ). Figure a′–d′ sketches local distribution of gas and the interfacial compatibility for the four cases, as depicted in Figure a–d, respectively.

Based on the simulation results and on the CNT,[20] we propose a specific cell nucleation behavior in foamed composites. This includes the forming of molecularly engineered interfaces, as described in Figure a–d. The results are shown in Figure e–h, respectively. When the NPs are decorated with a shell layer featuring the same chemical composition as that of the matrix (e.g., PS- and PMMA-based foams nucleated with SiO2–PS and SiO2–PMMA, respectively), a fuzzy interface and homogeneous distribution of CO2 will be achieved (Figure a), and the foam cells will be nucleated directly from the silica core surface due to the energetically favorable heterogeneous nucleation, as shown in Figure e. This is similar to the case of foams nucleated by bare silica NPs, and the same amount as for homogeneous nucleation is expected due to the uniformly distributed CO2. For the case described in Figure b (e.g., PS foams nucleated by SiO2–PMMA), foam cell nucleation is expected to occur at the interface between the shell layer and matrix (Figure f), which is ascribed to the local accumulated CO2 (Figure b). Figure g exhibits foam cell nucleation from the interface but outside the shell layer, indicating the cell nucleation behavior of the case described in Figure c (e.g., PS and PMMA foams nucleated by SiO2–PAN), owing to the high accumulation of CO2 at the interface and little CO2 inside the shell layer. Figure h shows foam cell nucleation inside the shell layer from the silica core surface, corresponding to the case of Figure d (e.g., PS- and PMMA-based foams nucleated by SiO2–PDMS). Compared to the case of Figure e, this specific type of foam cell nucleation is much more energetically favorable due to the high CO2 concentration inside the shell layer and explains the high cell nucleation efficiency of SiO2–PDMS.

Conclusions

Surface designed core–shell NPs with an 80 nm silica core and different thin polymer shell layers were exploited as heterogeneous nucleation agents in both PMMA and PS foaming. Following the synthesis and characterization of the core–shell structure NPs, the influence of particle surface chemistry on cell nucleation and cell morphology were studied. MDS were employed to describe the interface composition and its influence of cell nucleation. It was found that NPs decorated with a thin PDMS shell exhibited a higher nucleation efficiency in both PMMA and PS foams, when compared to the bare and other polymer shell-grafted NPs. This is ascribed to both the high CO2-philicity of PDMS shell and its high immiscibility with the polymer matrix, which resulted in a high local CO2 concentration in the PDMS shell and a high interfacial tension caused by phase separation. Thus, the energy barrier for CO2 embryo nucleation inside the PDMS shell was reduced. Due to the same chemical composition of the shell layer and polymer matrix, SiO2–PS and SiO2–PMMA showed a high miscibility with the PS and PMMA matrix, respectively, which decreased the heterogeneity of the nucleation agents and increased the cell nucleation energy barrier. This well explains the comparable nucleation efficiency of SiO2–PS and SiO2–PMMA to their bare counterparts in PS and PMMA foams, respectively. The relatively high nucleation efficiency of SiO2–PAN and its high incompatibility with the polymer matrix further confirm that the high immiscibility between the shell layer of NPs and matrices can significantly reduce the cell nucleation energy barrier and promote cell nucleation at the interfaces. In addition, the CO2 distribution was found to be significantly influenced by the composition of the interface, which also affects the cell nucleation. The deeper fundamental insights obtained by MDS provide additional guidance for the design of new highly efficient nucleation agents in nanocellular polymer foaming.

Materials and Methods

Materials

TEOS ≥99.0%, APTES 99%, 2-propanol 99.5%, and PS (Mw = 230,000 g·mol–1, ρ = 1.05 g·cm–3) were purchased from Aldrich (Milwaukee, WI, USA). PMMA was acquired from Arkema (VM100, i.e., a PMMA-co-EA polymer, ρ = 1.18 g cm–3) (La Garenne-Colombes, France). Ammonium hydroxide solution 28–30%, triethylamine (TEA) 99.5%, copper(II) bromide 99%, copper(I) bromide 98%, copper(I) chloride ≥99%, copper(II) chloride 99%, α-bromoisobutyryl bromide ≥99%, hydrochloric acid 37%, aluminum oxide (for chromatography), hydrofluoric acid (48%), acrylonitrile (≥99%), methyl methacrylate (≥99%), and monoglycidyl ether-terminated PDMS-G (Mw = 1000 g mol–1) were purchased from Sigma-Aldrich (St. Louis, MO, USA). Dihydroxy PDMS with a molar mass of 10,000 g mol–1 was obtained from Gelest (Morrisville, PA, USA). Anhydrous magnesium sulfate (>98%) was bought from Fluka (Morris plains, NJ, USA). Sodium bicarbonate was purchased from Church & Dwight (Ewing, NJ, USA). Absolute N,N-dimethylformamide (DMF), toluene, dichloromethane, chloroform, methanol, and tetrahydrofuran (THF) were purchased from Biosolve (Valkenswaard, the Netherlands). Ethanol absolute for analysis was purchased from Merck (Darmstadt, Germany). N,N,N′,N′,N″-Pentamethyldiethylenetriamine (PMDETA) 98% was purchased from Acros Organics (Geel, Belgium). Styrene, acrylonitrile, and methyl methacrylate were passed through an aluminum oxide column prior to polymerization to remove the inhibitor used. Copper(I) bromide and copper(I) chloride were purified by stirring appropriate amounts in water-free acetic acid for 24 h, followed by filtration, washing with ethanol for three times, and subsequent vacuum drying for at least 12 h. Milli-Q water was produced by a Millipore Synergy system (Billerica, MA, USA). Unless otherwise mentioned, all other chemicals were used as received.

NP Synthesis

To prepare Stöber silica NPs (SiO2) with a diameter of ∼80 nm, 168 mL of ethanol was mixed with 28 mL of Milli-Q water and 30 mL of TEOS in the presence of 2 mL of ammonium hydroxide while stirring at 500 rpm at room temperature. After 1.5 h, the obtained SiO2 dispersion was centrifuged at 10,000 rpm for 30 min. Subsequently, the collected SiO2 were redispersed in 2-propanol and centrifuged again. This washing step was repeated two more times followed by vacuum drying of the SiO2 NPs collected, at room temperature for 12 h.

Hydrolysis

To introduce silanol groups on the surface of the SiO2 NPs, the particles were redispersed in Milli-Q water by sonication (Branson 2510, Canada) for 1 h. Subsequently, hydrochloric acid was added to the dispersion while stirring at 500 rpm until the pH of the solution reached a value of approximately 1. After 4 h, the dispersion was centrifuged at 10,000 rpm for 30 min. The collected NPs were redispersed in Milli-Q water and repeatedly centrifuged. This washing step was repeated two more times followed by drying the silanol functional NPs (SiO2–OH) in vacuum at room temperature for 12 h.

APTES Modification

3.0 g SiO2–OH NPs were redispersed in 100 mL of ethanol followed by the addition of 15 mL of APTES. The dispersion was stirred at 500 rpm at room temperature for 17 h. The APTES-functionalized NPs (SiO2–NH2) were collected by centrifugation at 10,000 rpm for 30 min and redispersed in ethanol and centrifuged again. This washing step was repeated two more times followed by drying the collected SiO2–NH2 NPs in vacuum at room temperature for 12 h.

“Grafting to” of the PDMS-G to Silica NPs

1.0 g of SiO2–NH2 NPs was redispersed in 20.5 mL of THF and 15 g of PDMS-G while stirring at 500 rpm for 1 h followed by sonication for 1 h. Subsequently, THF was removed by rotary evaporation and the resulting silica NP dispersion in PDMS-G was immersed in an oil bath, thermostated at 80 °C for 17 h. Following cooling to room temperature, the reaction mixture was washed with THF and centrifuged at 10,000 rpm for 30 min. This washing step was repeated two more times, followed by vacuum drying the PDMS-G-grafted silica NPs at room temperature for 12 h.

Initiator Immobilization

1.5 g of SiO2–NH2 was redispersed in 75 mL of DMF by sonication for 30 min. The mixture was cooled to 0 °C with an ice bath, followed by dropwise addition of 15 mL of TEA and 5 mL of α-bromoisobutyryl bromide within 30 min while stirring at 700 rpm. The mixture was stirred for 17 h at room temperature, followed by centrifugation at 10,000 rpm for 30 min. The collected particles were redispersed in ethanol and centrifuged again to remove unreacted TEA, α-bromoisobutyryl bromide, and the salt formed by TEA and HBr. This washing step was repeated two more times, followed by vacuum drying the ATRP initiator functional NPs (SiO2–Br) at room temperature for 12 h.

Polymer Chains Grafted via SI-ATRP

1.0 g of the SiO2–Br NPs were redispersed in 10 mL of DMF by sonication for 30 min. Two other flasks were prepared, one with 156 mg of CuBr and 24.3 mg of CuBr2 and another one with 16.87 mL of DMF, 12.5 mL of styrene, and 459 μL of PMDETA. All three flasks were equipped with magnetic stirrers and sealed with a rubber septum. The flasks were purged with argon for 1 h. Subsequently, the styrene solution was added to the CuBr/CuBr2 mixture, followed by the addition of SiO2–Br NP dispersion to the resulting mixture. The reaction flask was submerged into a 90 °C thermostated oil bath and was stirred at 500 rpm for 17 h under an argon atmosphere. To purify the core–shell NPs, the reaction mixture was washed with DMF and centrifuged at 10,000 rpm for 30 min. This washing step was repeated two more times after which the collected SiO2–PS was vacuum dried at room temperature for 12 h. In order to determine the molar mass of the PS brushes, the SiO2 core of a ∼100 mg of sample dispersed in 2 mL of THF was etched with HF for overnight followed by drying the residual polymer. Subsequently, the molar mass was measured with GPC to be 5.5 × 103 g mol–1. The SI-ATRP of methyl methacrylate was the same as styrene. For the PMMA brushes, a molar mass value of 7.1 × 103 g mol–1 was obtained by GPC. For SI-ATRP of acrylonitrile, the CuCl/CuCl2 mixture was used instead of CuBr/CuBr2, while the other reaction conditions were the same as that of styrene. For the PAN brushes, a molar mass value of 9.1 × 103 g mol–1 was determined by GPC.

PMMA–PDMS10–PMMA Triblock Copolymer Preparation[57]

First, a difunctional PDMS-based ATRP macro initiator was synthesized. 18.0 g of dihydroxy PDMS10 (10,000 g mol–1), 0.63 mL of TEA, and 113 mL of dry toluene were mixed in a round bottom flask equipped with a stirrer bar (500 rpm). The flask was sealed by a rubber septum and placed in an ice bath while being purged with a gentle nitrogen stream for 30 min. Subsequently, 0.63 mL of BiBB was added to the solution under continuous stirring at 500 rpm at ∼0 °C. The reaction was left to stir for 18 h to gradually reach ambient temperature (21 °C). The bromide salt formed was removed by vacuum filtration over a borosilicate filter with a pore size of 10–16 μm. Next, the solvent was extracted by rotary evaporation under reduced pressure. The resulting oil, yellow in color, was diluted with 200 mL of dichloromethane and washed twice with 100 mL of a saturated sodium bicarbonate solution. The PDMS containing layer was isolated (separation funnel) and dried over anhydrous magnesium sulfate. The magnesium sulfate was removed by filtration, and the remaining volatiles were extracted by rotary evaporation under reduced pressure. The obtained Br–PDMS10–Br was used as a bifunctional macro initiator in the ATRP of MMA to yield a PMMA–PDMS10–PMMA triblock copolymer. In a typical reaction, CuCl and a magnetic stirrer bar were placed in a round bottom flask and degassed for at least 30 min by purging with nitrogen. In a separate flask solutions of PMDETA, MMA and Br–PDMS10–Br were purged with nitrogen for 30 min. Subsequently, this monomer solution was added by using degassed syringes to the reaction flask under constant stirring (300 rpm). The molar ratio of [MMA]/[PMDETA]/[CuCl]/[Br–PDMS10–Br] was 200:4:2:1. In a typical reaction, 3 g of Br–PDMS10–Br was used. Following the addition of all ingredients to the reaction flask, the reaction mixture was heated to 80 °C and left to react for 1.5 h. Upon completion of the reaction, 10 mL of chloroform was used to dilute the reaction mixture prior to precipitation in 100 mL of methanol. The precipitated product was collected by filtration and washed several times with methanol. The PMMA–PDMS10–PMMA triblock copolymer obtained was dried under vacuum at 50 °C to constant weight.

Nanocomposite Film Preparation

Nanocomposites were prepared by dispersing 4 wt %, based on the bare silica NP weight, (functional) silica NPs in PS or PMMA using a mini extruder (DSM Xplore, The Netherlands). We selected a particle loading of 4 wt % since this allowed us to add a reasonably high number of particles (7.5 × 1013 particles cm–3) with good foamability of the nanocomposite. In fact, significantly increasing the particle loading would eventually result in no foaming of the respective nanocomposites (data not shown). In a typical procedure, a dry blend of NPs and PS (or PMMA) was fed to the extruder followed by internal mixing for 3 min. The barrel temperature was set to 155 °C, and the screw speed was 100 rpm. Subsequently, the nanocomposite was collected and left to cool to room temperature. A hot press (Fortijne, the Netherlands) was used to press ∼0.2 mm thick nanocomposite films in a mold (4 × 3 cm). The press temperature, applied load, and press time were 180 °C, 250 KN, and 10 min, respectively. In the Supporting Information, SEM images of cross-sectioned nanocomposite films are shown (see Supporting Information, Figure S5), revealing the good particle distribution and the absence of severe particle agglomeration for the nanocomposites used in this work.

Batch Foaming of Nanocomposite Films

The nanocomposite PS films obtained were saturated with carbon dioxide (55 bar) in an autoclave for 3 h at room temperature followed by rapid depressurization. Subsequently, the PS films were foamed by immersion in a glycerol bath, which was thermostated at 100 °C for 30 s. Next, the samples were quenched to room temperature in a 50:50 water/ethanol bath followed by immersion in ethanol for 1 h. Finally, the foams were left to dry in air for at least 12 h prior to further analysis. For the foaming of PMMA nanocomposite films, a CO2 saturation pressure and time of 55 bar and 3 h were used, respectively. Following quick depressurization, the polymer films were foamed by immersion in a water bath set at 40 °C for 30 s after which the samples were quenched in an ice bath for 30 min. Subsequently, the samples were left to dry in air for at least 12 h prior to further analysis. The Supporting Information contains a scheme of the used foaming setup (see Supporting Information,Figure S6). We note that the foaming conditions reported in this work provided the lowest cell size and highest cell density within a range of foaming temperatures (0–110 °C) and times (few seconds to 5 min) (data not shown), and thus they were selected as our standard conditions throughout this work.

CO2 Sorption and Desorption Measurements

The CO2 wt % absorbed in the polymers (a measure for the CO2-philicity) was determined by saturating polymer films with 55 bar CO2 for 3 h followed by releasing the pressure (t = 0) and measuring the weight loss due to CO2 desorption as a function of time. Extrapolating the desorption curves to t = 0 gives the values for the amount of CO2 absorbed.

Fourier Transform Infrared Spectroscopy

FTIR spectra were collected with a Bruker ALPHA single ATR FTIR spectrometer equipped with an ATR single reflection crystal (Bruker Optic GmbH, Ettlingen, Germany). The spectra were collected in the range of 400–4000 cm–1 (spectral solution of 4 cm–1, 1280 scans). Background spectra were recorded against air.

Thermogravimetric Analysis

The weight loss of the (modified) particles as a function of temperature was measured with a TGA400 (PerkinElmer, Inc., Waltham, MA, USA). A sample weighing ∼3–6 mg was loaded into the platinum pan and set to 50 °C to stabilize. Subsequently, the sample was heated to 900 °C at a heating rate of 20 °C min–1. The applied N2 flow was 25 mL min–1.

Transmission Electron Microscopy

To investigate the core–shell structure of the functionalized NPs, a FEI/Philips CM300 transmission electron microscope (Eindhoven, the Netherlands) was used. Diluted particle dispersions in THF were deposited on the carbon side of a carbon/copper grid (HC200-Cu) (EMS, Germany). Images were obtained in the bright field mode with a 300 kV acceleration voltage.

Scanning Electron Microscopy

To investigate the morphology of the unfoamed/foamed nanocomposite films, a high-resolution scanning electron microscope (JEOL Field Emission JSM-633OF, JEOL Benelux, Nieuw-Vennep, the Netherlands) was used. The typical electron acceleration voltage used was 5 keV. Prior to analysis, the nanocomposite films and foams were freeze fractured after cooling in liquid nitrogen for 5 min and the obtained cross sections were sputter coated (JEOL JFC-1300 Auto Fine Coater, JEOL Benelux, Nieuw-Vennep, the Netherlands) with gold under an argon atmosphere for 40 s at a current of 30 mA.

Atomic Force Microscopy

The cross-sectioned surfaces of a cryo-microtomed PMMA–PDMS10–PMMA triblock copolymer (10 wt %) PMMA blend film was investigated with atomic force microscopy (AFM) in order to reveal its morphology. To this end, a MultiMode 8 AFM instrument operated with a JV vertical engage scanner and retrofitted with a NanoScope V controller (Bruker) was used in the PeakForce Quantitative Nanomechanical Mapping (QNM) mode. Medium soft (2 N/m nominal spring constant, 7 nm tip radius) cantilevers (OMCL-AC240TS, Olympus) enabled performing both the indentation of the sample surface as well as monitoring the relevant cantilever deflection induced by the tip–sample contact in order to capture images representing the surface mechanical compliance (elastic modulus) and topology. Data were collected following a sine-wave sample-tip trajectory with a frequency of 2 kHz and utilizing a peak-force amplitude value of 150 nm with feedback loop control of 25. The ScanAsyst optimization algorithm was set to “on” to acquire high-resolution images at the lowest applied normal forces. Data were collected in air at controlled temperature (21 °C) and relative humidity (∼40%). Image processing and data analysis were conducted with the NanoScope (ver. 9.10) and the NanoScope Analysis software (ver. 2.00), respectively.

Calculation of Cell Density

The cell size and cell density were obtained by analyzing the SEM cross sections. Cell density (Nv) of the foams was calculated by Kumar’s theoretical approximation.[58] No direct measurements of cell dimensions over the micrograph are required by this method, only the micrograph area (A) and the total number of cells (n) contained therein are measured. Together with the magnification factor of the micrograph (M), Nv can be calculated according to eq

By combining Nv with the volume expansion ratio (B) of nanocomposite films after foaming, the cell numbers per cm3 of unfoamed materials (N) can be calculated according to eq .

Molecular Dynamics Simulations

In the simulations, the polymers are represented by the Kremer–Grest (KG) model,[50] which is a well-established coarse-grained bead-spring model that has been shown to successfully reproduce the static[59] and dynamic[60] properties of polymer brushes. The non-bonded interactions are calculated using the LJ potential

The LJ parameters ε and σ define the units of energy and distance, respectively, and set to unity per default, unless stated differently. The cut-off is set to 2.5 σ for attractive interaction and to 1.12246 σ when purely repulsive interactions are employed.

The reduced LJ units employed in this article can be converted to real units [e.g., poly(ethylene), using ε = 30 meV and σ = 0.5 nm].[50] However, a quantitative comparison between experiments and simulations is not possible and only qualitative effects can be compared. In the KG model, the bonded interaction-potential VKG that acts between connected repeat units is described bywhere the stiffness k = 30 ε/σ2, the maximum extension R0 = 1.5 σ, and ε and σ are set to unity.

In our box, there are 400 polymers of degree of polymerization (N = 30) end-anchored to a surface of 30 × 30 σ2 consisting of LJ beads on a hexagonal lattice, resulting in grafting density α = 0.45 chains/σ2. The grafting density is approximately 15 times the critical grafting density for brush formation α* = 1/(πRgyr2), where Rgyr is the radius of gyration of the polymer free in solution (Rgyr = 3.15 σ). This grafting density is chosen because it is close to the grafting densities typically obtained with SI-ATRP.[33,61] The brush is in contact with at least 1500 polymers of degree of polymerization (P = 100), representing the polymer matrix. Moreover, gas-molecules are dispersed in the system, keeping the density in the matrix phase constant at 0.015 σ–3 by iteration. The total density in the matrix phase is kept constant at approximately 0.87 σ–3 ± 0.01 σ–3via a piston represented by a repulsive mathematical wall.

The affinity between the shell (brush), gas molecules, and the matrix is altered by varying the strength of the LJ potential via ε. The self-interactions for the shell and the matrix are described by the default settings, while the self-interaction for the gas molecules as well as the interaction between the gas and the wall is set to be purely repulsive (cut-off at 1.12246 σ). For the exact values of the different interactions (deviations from the default), we refer to the main text.

The positions and velocities of all particles in the simulation cell are updated using the Verlet algorithm (implemented in LAMMPS[62]) using a time-step of 0.005 τ. The temperature is kept constant at T = 1.0 ε by a Langevin thermostat (time-constant 1 τ), which is implemented via the wall-atoms alone to minimize interference with the system. The final production runs are performed in the NVT ensemble for at least 1,000,000 timesteps. These runs are employed to extract average density profiles for the shell, matrix, and gas particles from the simulations.