In situ polymerization and polymer grafting to stabilize polymer-functionalized nanoparticles in polymer matrices

The design of hybrid polymer/inorganic nanoparticle materials is motivated by nanoparticles' potential to enhance properties and performance in existing polymer and composite materials.^1–4 Nanoparticles with unique inherent properties add functionality to polymeric materials, or the existing polymer matrix properties can be reinforced through the particle–matrix interface that is magnified when using nanoparticles compared to other composite materials.^5–7 However, reinforcement often requires a maximal interfacial interaction between particles and matrix, which makes optimal particle dispersion imperative.^8–11 

A successful method for controlling particle dispersion in polymer matrices is to functionalize nanoparticle surfaces with polymer chains.^12,13 Polymer-functionalized nanoparticles possess tunable structural parameters [e.g., graft density $(σ)$ and graft chain length (N)] and unique physical confinement effects that influence particle assembly, mobility, and miscibility within the matrix and make them viable components for reinforcing nanocomposites.^2,8,14–18 For example, in the glassy state, polymer-functionalized nanoparticles improve mechanical performance when particles are spatially distributed and strongly bind to the matrix.⁸ In systems with chemically identical matrix and graft chains, these criteria are met by altering entropic forces, specifically by making the graft chain length (N) much larger than the matrix chain length (P) $(α=NP>1)$⁠.^13,19 However, when matrix and graft chains are enthalpically immiscible, blends are driven to phase separate.

To implement nanoparticles into commercially viable materials, it is essential to develop multiple avenues for dispersing polymer-functionalized particles into matrices, including matrices that are immiscible with the particle's graft.^20,21 For these systems, a host of non-equilibrium pathways exist that control the structure formation and behavior of polymeric materials.^22,23 Reaction-induced phase transitions (RIPTs) utilize chemical reactions to promote or trap desired phases, equilibrium or non-equilibrium, and have most recently been used to control and stabilize nanoscale morphologies in block polymer systems.^24–28 In the case of liquid-to-solid phase transitions, the entropic gain for mixing in the liquid state allows components to be homogeneously dispersed prior to phase transformation. During reaction-induced transformation, the rapid formation of a polymer during polymerization prevents the system from macrophase separating. A notable and commercially relevant example of stabilizing immiscible polymer blends with RIPT is the production of high-impact poly(styrene) (HIPS).^29–31 The morphology of HIPS consists of phase separated poly(butadiene) (PBD) droplets in a poly(styrene) (PS) matrix. Stabilization of nano- or micrometer-sized PBD droplets is due to an in situ processing method where the free-radical polymerization of styrene in the presence of PBD results in the formation of a PS homopolymer and PBD grafted with PS chains. The PS grafts on PBD form via the creation of an allylic radical on the PBD backbone, which initiates PS growth from the PBD chain. In addition to liquid-to-solid phase transformation, Hickey and co-workers have recently shown that similar polymer grafting chemistry will drive nanostructural transitions in block polymer and monomer blends.^28,32 By further adapting this processing method for polymer-functionalized nanoparticles in which the polymers on the nanoparticle surfaces contain unsaturated backbones, there is potential for similar RIPT stabilization in traditionally immiscible hybrid systems.^33–35 

The following work presents a new method for dispersing polymer-functionalized nanoparticles in immiscible polymer matrices using HIPS-inspired chemistry.^28–32 Here, we show that silica nanoparticles functionalized with poly(cyclooctadiene) (PCOD) synthesized using surface-initiated ring-opening metathesis polymerization (SI-ROMP) are blended into styrene and methyl methacrylate solutions and undergo a free-radical polymerization to create PS or poly(methyl methacrylate) (PMMA) hybrid materials (Fig. 1). SI-ROMP is the preferred surface-initiated polymerization method to create polymer-functionalized inorganic nanoparticles due to the formation of polymers containing unsaturated backbones, similar to poly(butadiene) used in HIPS processing. The dispersion of particles within the matrix is improved with fast polymerization kinetics during in situ polymerization and simultaneous grafting of PS or PMMA from PCOD-functionalized nanoparticles, akin to traditional HIPS.^29–31 This method adapts commercially relevant chemical processes for the development of well-dispersed hybrid polymer/inorganic nanoparticle materials. The method described is a potential avenue to enhance nanoparticle loadings and levels of particle dispersion needed for solid state material reinforcement. The dispersion of polymer-functionalized nanoparticles into enthalpically incompatible matrices is a challenging problem that requires reduced enthalpic and entropic penalties to create stable materials. Minimization of these thermodynamic considerations due to in situ grafting process will lead to new classes of materials that are not reliant on miscible particle and polymer components.

Tetraethylorthosilicate (TEOS), ammonium hydroxide (28% in water), Grubbs' second-generation catalyst (G2), pentane, 3-bromopyridine (3BP), 1,5-cyclooctadiene (COD), ethyl vinyl ether, styrene, methyl methacrylate (MMA), benzoyl peroxide (BPO), and hydrofluoric acid (49% aq.) were purchased from Sigma Aldrich. The G2 catalyst was converted into Grubbs' third-generation catalyst (G3) by addition of 10 mol equivalent of 3BP and precipitation in pentane.³⁶ Ethanol (EtOH), tetrahydrofuran (THF), and toluene were purchased from ThermoFisher Scientific. THF was purified using a solvent column (JC Meyer). Alumina was purchased from Honeywell. The silane, (5-bicyclo[2.2.1]hept-2-enyl)triethoxysilane (NBES), was purchased from Gelest.

Silica particles with average diameters of 140 nm (d = 140 nm) were synthesized using Stöber processing methods previously reported.³⁷ Modification of the particle surface with (5-bicyclo[2.2.1]hept-2-enyl)triethoxysilane was performed using previously published procedures.^38,39 Particles were separated from the reaction solution using centrifugation (Eppendorf, 10 000 rpm) and redispersed in anhydrous THF prior to degassing and placing in the glovebox.

SI-ROMP of PCOD was performed following modified synthetic procedures from previously published work.³⁹ In a typical experiment, silica nanoparticles functionalized with (5-bicyclo[2.2.1]hept-2-enyl)triethoxysilane were mixed with a THF solution containing G3 (2 ml, 0.05M) and washed via centrifugation (10 000 rpm, 45 min) and redispersed in a THF solution ([3BP] = 0.02 M) three times. After washing unreacted G3 from the nanoparticle solution, the activated particles were injected into a degassed COD solution containing 3BP (0.1M). The polymerization was run for 2 h at 0 °C. The polymerization was terminated by injecting excess ethyl vinyl ether (10 ml) to the reaction mixture and stirred for an additional 30 min before being diluted in toluene and passed through alumina several times. Particles were then separated from the reaction solution by centrifugation (10 000 rpm, 45 min) and redispersed in toluene (4 wt. %). Particles are not dried prior to dispersion in styrene or MMA solution, and are instead solvent switched by three additional washes in the monomer solvent prior to blend preparation. Attempts to dry PCOD-functionalized particles under vacuum lead to cross-linking, resulting in large amounts of insoluble particles.

Styrene and MMA monomer blends were purified and prepared in identical procedures. Monomer solvent (100 ml) was purified by passing through alumina and was collected and stored in a refrigerator prior to use. PCOD-functionalized nanoparticle blends are prepared by solvent switching PCOD-grafted particles from toluene to the monomer solvent (10 wt. %) and diluting solutions to desired particle weight percent (10 wt. %, 5 wt. %, and 1 wt. %). For each ml of monomer, 14.5 mg of benzoyl peroxide (BPO) was added to the solution. Particle solutions were capped with a septum and sonicated for 10 s to fully dissolve the reagents. Argon was then flowed into the headspace of the vials for 5 min to reduce temperature-related degradation. Samples were placed in an oven (90 °C) and polymerized for 21 h. After polymerization, samples were quenched with liquid nitrogen and freeze dried under vacuum for 24 h. The reaction conversion was determined by comparing the mass of blends before polymerization and after freeze drying.

Polymer-functionalized nanoparticles were separated from the as-processed blends to characterize the polymer chains on the surface of the nanoparticles after polymerization. As-processed samples were dissolved in THF and centrifuged for 45 min at 10 000 rpm. The homopolymer solution was decanted into a second centrifugation vial, and the separated particles at the bottom of the centrifuge tube were redispersed in fresh THF. The process was repeated five times, each time redispersing particles in fresh THF and decanting the homopolymer solution into a fresh vial to remove any undesired polymer or particle from the respective solutions. The homopolymer solutions were filtered three times with a 0.2 μm filter (PTFE) prior to SEC analysis. The nanoparticle solutions were separated for HF treatment or solvent switching to toluene for dynamic light scattering (DLS) analysis.

As-polymerized samples were vacuum pressed into bulk films to remove air bubbles. Samples (2.2 g) were broken into small chunks before sandwiching between two sheets of kapton and a pressing mold. PS samples were pressed under vacuum at 120 °C for 20 min, and PMMA samples were pressed under vacuum at 150 °C for 20 min before breaking from the mold.

Micrographs of polymer-functionalized nanoparticles and hybrid polymer/inorganic nanoparticle materials were obtained using a FEI Tecnai G2 Spirit BioTwin TEM. Polymer-functionalized nanoparticles were dropcast from toluene onto TEM grids (Electron Microscopy Sciences, Formvar/Carbon 200 Mesh, Copper) and allowed to dry overnight. Dried hybrid polymer/inorganic nanoparticle material samples were microtomed into 70–90 nm sections using a Leica UC6 ultramicrotome.

The number- and weight-average molecular weight (M_n and M_w, respectively) and dispersity (Đ) for polymer samples were determined using a Tosoh EcoSEC (Tosoh Co.) equipped with a Wyatt Dawn Heleos-II eight-angle light scattering detector (Wyatt Technology Corp.) at concentrations of 5 mg/ml in tetrahydrofuran at 40 °C. Grafted chains are cleaved from the particle surface by treatment with hydrofluoric acid (1 ml, 49%, aq.) and were vented overnight in a fume hood before drying under vacuum.

Particle graft density and mass fraction were determined using a TA Instruments Q800 SDT. Vacuum dried samples (10–15 mg) were placed in an alumina crucible and heated to 100 °C for 20 min to remove residual moisture. Samples were then heated to 800 °C at 10 °C/min, and the particle mass fraction was determined as the difference in weight fraction between 800 °C and 100 °C for blend samples and as the difference in weight fraction between particles before and after SI-ROMP for particle samples.

Glass transition temperatures (T_g) were measured using a TA Instruments Q2000 DSC. Dried samples (10–20 mg) were pressed into aluminum pans and heated (10 °C/min) to 150 °C for 5 min, cooled to 0 °C for 5 min, and heated again to 150 °C for 5 min. The glass transition of nanocomposites is determined on the second heating cycle.

DLS measurements were conducted using a Brookhaven Instruments BI-200SM Research goniometer system with a 637 nm, 30 mW laser and a 100 μm aperture. Dilute solutions (1 mg/ml, toluene) were filtered with a 0.45 μm filter (PTFE) prior to measurement. Average values for the relaxation rate, $Γ$⁠, were obtained at multiple angles using the CONTIN algorithm to fit the autocorrelation function. Polymer-functionalized nanoparticles were separated from the nanocomposite matrix by five washes in toluene.

The PCOD-functionalized nanoparticles used in this study were synthesized via surface-initiated ring-opening metathesis polymerization (SI-ROMP), which were prepared following adapted SI-ROMP procedures [Fig. 2(a)].^21,39,40 The ring-opened structure of the monomer COD resembles poly(1,4-butadiene) and is used as an analogue to the PBD used in HIPS processing.^30,31 The polymer-functionalized nanoparticles as shown in Fig. 2 have a graft density of 0.03 chains/nm², molecular weight of M_w = 13 kg/mol, and dispersity of Ð = 1.51 that were consistent for the two particle reactions ran for MMA and styrene solutions, respectively. The additional elution peaks after the 10 min mark in Fig. 2(c) are attributed to be oligomeric PCOD that undergo secondary metathesis during polymerization, as the peaks show up in both particle reactions, but do not alter M_w.

Solution blending of PCOD-functionalized nanoparticles and monomer solvent minimizes the matrix length P during mixing and reduces the entropic penalties that prevent homogeneous mixtures. Particle blends were prepared through sonication and vortex mixing to fully disperse the particles in monomer, and the particles remain in suspension after storing in a freezer for several days. The suspension of polymer-functionalized nanoparticles in the monomer solvent was quantified through multi-angle DLS measurements as shown in Fig. 2 and in the supplementary material. The distribution of hydrodynamic diameters (d_H) in each respective solvent is monomodal, and the average d_H in toluene (177 ± 2 nm), styrene (179 ± 2 nm), and MMA (174 ± 4 nm) does not deviate significantly enough to suspect particle aggregation due to immiscibility.

To create hybrid PS/silica nanoparticle materials, BPO was added to PCOD-functionalized nanoparticle/styrene blends of three different loadings (1, 5, and 10 wt. % by polymer-functionalized nanoparticle weight). After heating for 21 h at 90 °C, there is an observable phase change in the reaction blend. Before polymerization, the nanoparticle/styrene mixtures are liquid, and after polymerization, the samples are solid. Conversion of the blends is measured by comparing the blend mass before polymerization and after freeze drying and was consistently above 90% conversion for all loadings.

The morphology of the hybrid PS/silica nanoparticle materials after polymerization was determined by performing TEM on dried microtomed samples. The micrographs of the hybrid PS/silica nanoparticle samples are shown in Fig. 3. As seen in Fig. 3, there exist individually dispersed particles within all three particle loadings. As particle loading increases, particle strings and submicron clusters become increasingly evident. These results are promising, as particles at 1 wt. % are individually dispersed using the in situ polymerization method described here. The mixture of individually dispersed particles and string-like formations at increased particle loadings suggests that preferred equilibrium phases are potentially arrested during the phase change and that phases such as strings, sheets, or particle clusters may be accessible, similar to a previous report.¹³ Additional experiments exploring reaction conditions and systemic parameters are needed to understand the underlying phase behavior and are currently underway.

In addition to PS, hybrid PMMA/silica nanoparticle materials created using in situ polymerization of MMA mixtures were also successful (Fig. 4). The solution blends were prepared and run under identical conditions as the PS-based materials. The main notable difference is that the free-radical polymerization of PMMA has faster polymerization kinetics. During the free-radical polymerization of MMA, the solutions boil and become increasingly viscous within the first hour of heating, which suggests that the reaction leads to a runaway polymerization of MMA at 90 °C. This is consistent with previous reports of bulk PMMA polymerization kinetics using free-radical initiators.^41,42 To prevent the vial from building up pressure, a needle was used to release the pressure during the polymerization.

Similar to the PS blends, PMMA/silica nanoparticle hybrid materials were dried and microtomed for TEM imaging. The TEM micrographs show a range of particle dispersions, as seen in Fig. 4. Unlike the PS blends, samples with 1 wt. % silica nanoparticles had large sectioned areas with no observable particles within the matrix. Additionally, large voids were occasionally seen throughout the sectioned sample [Fig. 4(a)], which possibly suggests that polymer-functionalized particles weakly interact with the PMMA phase, and at low weight percentages either phase separate during the runaway polymerization or dislodge from the matrix during microtoming. However, when the particle loading increased to 5 wt. %, large sectioned areas of the sample showed uniform particle dispersion throughout the matrix. At 10 wt. %, particles remain dispersed but begin to show signs of microphase separation or self-assembly into a preferred phase. It is evident that the fast polymerization kinetics of PMMA are more successful at trapping particles in their liquid-like disperse state at moderate loadings (5 wt. %).

The glass transition temperature of the as-processed polymer/silica nanoparticle hybrid materials were measured using DSC. In addition, homopolymer PMMA and PS were synthesized under identical polymerization conditions with no silica nanoparticles (0 wt. %) and their T_g's measured. The glass transitions of these control samples were compared to the glass transition temperatures of the hybrid materials. T_g values were obtained on the second heating cycle, and the results can be seen in Fig. 5 with T_g values included in Table I.

For both PS and PMMA, the measured T_g of the homopolymer controls range on the low end of the expected T_g values. This is attributed to the reaction conditions used to synthesize PS and PMMA. After the initial mixing of BPO into the reaction solution, the solutions are not stirred and are freeze dried after polymerization to remove unreacted monomer. The procedure likely results in oligomeric and low molecular weight chains within the blends, which will plasticize the system and reduce the T_g, similar to previously published results using a similar polymerization and purification process.⁴³ Under this consideration, all particle loaded samples have similarly low measured T_g values compared to their homopolymer control. No obvious trends are observed as particle loading increases in either PS or PMMA matrix. Notably, 1 wt. % in PS and 5 wt. % in PMMA that showed optimal particle dispersions in Figs. 3(a) and 4(b) have the highest measured T_g for their respective matrix.

While polymerization kinetics appear to influence particle dispersion, it is hypothesized that in situ polymerization of the matrix also grafts chains from the unsaturated PCOD chains that are on the nanoparticle surface. To quantify the extent of polymer grafting from PCOD-functionalized nanoparticles, SEC-MALS, DLS, and TEM were performed on separated polymer-functionalized nanoparticle samples, and their results are summarized in Figs. 6 and 7.

For SEC-MALS measurements, as-polymerized polymer/silica nanoparticle hybrid materials were dispersed in THF and treated with HF to dissolve all silica prior to SEC-MALS measurements. These matrix and grafted polymer ensembles for PS and PMMA samples are shown in Figs. 6(a) and 6(d), respectively, and are compared to the homopolymer PS and PMMA controls (0 wt. %) for respective reference. PS ensembles possess a low molecular weight shoulder persisting at later elution times for all four samples, indicating that there is likely a distinctly different polymer population. The low molecular weight shoulder in the 0 wt. % PS sample is believed to be a result of the synthetic conditions, as solution samples are not stirred during polymerization. However, it is notable that the low molecular weight shoulder increases in scattering intensity as particle loading increases. PMMA homopolymer, by comparison, is monomodal at 0 wt. %. It is only with the presence of PCOD-functionalized nanoparticles that a second, high molecular weight shoulder appears at earlier elution times. Similar to PS ensembles, this shoulder increases as particle content increases.

With the matrix and grafted ensembles, it is difficult to attribute the individual populations to a specific group of chains. For instance, it is not easily discernable if the shoulders in the elution curves are a result of a single population of matrix chains and a single population of grafted chains, or if the matrix population is bimodal and the grafted population is negligible. To directly determine the elution profiles of matrix and grafted populations, the matrix homopolymer was separated from as-polymerized blends by dispersing materials in THF and collecting polymer-grafted particles through centrifugation. The homopolymer solution is then decanted and not treated with HF prior to SEC-MALS. The separated PS homopolymer shown in Fig. 6(b) still exhibits a shoulder, suggesting that the matrix is comprised of two size populations of PS, which is consistent with the PS control sample. For separated PMMA homopolymer as shown in Fig. 6(e), all peaks are monomodal, which differ from the SEC profiles of the as-polymerized samples but are consistent with the control sample that is shown in Fig. 6(d). These results suggest that the large molecular weight shoulders in Fig. 6(d) are grafted to the nanoparticles. The molecular weights and dispersity of the separated homopolymer samples are reported in Table I.

Finally, the polymer-functionalized nanoparticles that were separated from the as-polymerized materials were thoroughly cleaned by repeated centrifugation cycles, each time being redispersed in fresh THF. These particles were then treated with HF to cleave the grafted chains from the particle surface for SEC-MALS measurements. The resulting SEC-MALS profiles are shown in Figs. 6(c) and 6(f) and are compared to the respective chromatographs of cleaved PCOD prior to in situ polymerization. For grafted chains that were separated from a PS phase, all three samples shift to earlier elution times indicating an increase in hydrodynamic size. This is taken to be evidence that PS is grafting from PCOD grafts, increasing the degree of polymerization, and increasing the hydrodynamic size as a result. Since all three particle loadings elute at the same time, it is not expected that particle loading influences grafting using the experimental conditions described here.

Similarly, all three samples separated from a PMMA phase shift to earlier elution times. Specifically, the elution peaks match up identically to the high molecular weight shoulder observed in as-processed PMMA ensembles. This is taken to be evidence that PMMA also grafts from PCOD grafts. Additionally, a second population of chains consistently elutes at 10 min in PMMA grafted samples. Since this is a later elution time than the PCOD elution peaks prior to in situ polymerization, it is assumed that this low molecular weight population grafts from oligomeric PCOD observed at late elution times in Fig. 2(c). These results suggest that the in situ polymerization of the matrix phase in the presence of polymers with unsaturated backbones such as PCOD results in polymer grafting.

To support the conclusions developed from SEC-MALS measurements, DLS was performed on separated polymer-functionalized nanoparticle samples dispersed in toluene and compared to the PCOD-functionalized nanoparticle size prior to in situ polymerization. In situ grafting of matrix chains to the PCOD-functionalized nanoparticles would increase the degree of polymerization of grafted chains and as a result increase the hydrodynamic size of polymer-functionalized particles from their initial size in toluene as reported in Fig. 2,. As seen in Fig. 7 and Table I, all six polymer-functionalized nanoparticle samples increased in size after in situ polymerization. Particles separated from PMMA increased in size to 200 nm and particles separated from PS increased in size to 250 nm as per Table I. Given the magnitude of increase in the hydrodynamic size, we claim that the increase in brush height is due to the presence of grafted matrix chains. Additional grafts will effectively increase the degree of polymerization, which scales with polymer brush height and is consistent with the SEC results in Fig. 6. Additionally, particle hydrodynamic sizes do not deviate significantly in size with increased particle loading for PS or PMMA blends. This is consistent with the elution profiles of the cleaved chains discussed above.

Furthermore, particles measured via DLS were also dropcast from the solution onto carbon copper grids for TEM [Figs. 7(b) and 7(c)]. A visible polymeric shell can be seen in between particles, indicating that the polymer is present. The combination of SEC-MALS, DLS, and TEM results leads us to conclude that in situ grafting from PCOD-functionalized nanoparticles does occur with the reported in situ polymerization method.

The in situ polymerization method described here enables high conversions of well-dispersed nanomaterials. All samples reported were polymerized from 6 ml solution blends that yielded several grams of material after freeze drying. To determine whether these materials could potentially be characterized for bulk applications, samples (2.2 g) were hot pressed under vacuum into bulk films (mold length = 64 mm; width = 64 mm; depth = 0.4 mm) to remove air bubbles from as-processed blends. Photographs of bulk films can be seen in Fig. 8.

As is evident in Fig. 8, increasing particle loading percentage increased the amount of light scattered from bulk films. While this is typically a sign of particle aggregation in films, we note that the particle size (140 nm) and contrast in refractive index with the matrix is enough to scatter visible light [as seen similarly prior to polymerization in inset photos in Figs. 2(d)–2(f)]. To ensure that scattering is not due to hot press-induced aggregation, pressed samples were microtomed for TEM. The resulting micrographs can be found in the supplementary material. It is believed that the pressing conditions described are mild enough to preserve particle dispersion but are not taken as evidence that particles are in a preferred equilibrium state. Further investigation of these considerations is currently under way.

Stabilization of immiscible polymer blends is a commercially and an industrially relevant problem, and the incorporation of nanoparticles into these systems opens new avenues of material design and reinforcement. Here, we demonstrate that in situ polymerization of monomer in the presence of PCOD-functionalized nanoparticles leads to well-dispersed nanoparticles in PS and PMMA matrices. A combination of the polymerization kinetics and in situ polymer grafting from unsaturated PCOD chains is believed to influence particle dispersion and morphology. Additional work on the equilibrium phases of these materials and the influence of initial polymer-functionalized nanoparticle structure is currently under way.

See the supplementary material for description on hydrodynamic size measurements obtained using DLS, TGA measurements, and TEM of pressed samples.

This work was supported by the National Science Foundation (NSF), Division of Materials Research Polymers Program (CAREER Proposal No.: DMR-1942508) and start-up funds from the Pennsylvania State University. TEM measurements were taken at the Materials Characterization Lab (MCL) in the Materials Research Institute (MRI) at the Pennsylvania State University. We are grateful to Missy Hazen for her help with microtoming the polymer samples for TEM.