Metal Oxide Nanoparticles: An Effective Tool to Modify the Functional Properties of Thermally Stable Polyimide Films

A series of polyimide/metal oxide (either ZrO2 or TiO2) nanocomposite films were fabricated based on two polymer matrices. The prepared films were characterized by scanning electron microscopy (SEM), atomic force microscopy (AFM), and X-ray diffraction analysis (XRD), and their thermal and mechanical properties were investigated with the use of thermogravimetric (TGA), differential thermal analysis (DTA), and thermomechanical analysis (TMA). We have found out that functional properties of the obtained materials are determined by a number of factors, not only the type, size, surface functionality, and concentration of the nanofiller, but also the chemical structure of the matrix polyimide. We have demonstrated some trends in the thermal and mechanical behavior of the materials depending on these features. The data could be of great interest in the areas where new materials with improved functional characteristics are needed.


Introduction
The development of modern technologies in a variety of industrial areas from medicine to aerospace necessitates working out new materials with either improved target characteristics or with an extended set of functional properties. Optimization of the design and performance of modern materials requires exploring the possibilities of controlling their physico-chemical characteristics. In the last 10-15 years, researchers from all over the world have focused mostly on the so-called nanocomposite materials, in particular on the polymer-inorganic ones [1][2][3][4][5], because of the relative easiness of application-oriented control of the functional properties of the resulting materials through the alteration of the physico-chemical characteristics of both constituents.
Aromatic polyimides (PIs) have long been used in many industrial applications due to their excellent mechanical and dielectric properties, corrosion resistance, high thermal and radiation resistance as well ass thermal stability. They are capable of replacing metals or glass in many areas such as electrical, mechanical, and aerospace engineering [6][7][8][9]. In recent years, the PI-based nanocomposite materials filled with metal oxide (MO) nanoparticles such as CeO 2 , ZrO 2 , NiO, TiO 2 , etc. have been extensively studied owing to the superiority of their functional properties over those of the unfilled PIs [10][11][12][13]. MO nanoparticles are now considered to form the basis of the diversity of modern "smart" materials and devices due to the possibility of manipulating their physical and chemical properties by changing the synthesis conditions of the nanoparticles and thus varying their characteristics such as size, morphology, surface stoichiometry (in particular, the number of oxygen vacancies), etc. [14,15]. As a result, the chemical activity of nanoparticles can change considerably, especially with regard to redox reactions and oxygen scavenging [16]. This feature of MO nanoparticles has predetermined their wide use in catalytic and sorption processes. However, the same features of the structure can play a key role in the thermal and especially thermo-oxidative stability of PI materials, providing variability of chemical interactions between the nanofiller and polymer matrix. We have already observed a quite intriguing effect of CeO 2 nanoparticles on the thermal properties of PIs determined by the structure of the latter [10,17]. The introduction of nanosized ceria into the polymer matrix substantially increased the thermo-oxidative stability of only those PIs whose macromolecules contained SO 2 -groups, while that of a PI with no sulfonyl groups deteriorated by 30 • C. We suggested a mechanism of the interactions between the PIs and CeO 2 nanoparticles which would explain such a selective impact basing on the hypothesis on the ability of the nanoparticles to take part in the formation of active oxygen species (AOS) on their surface. The AOS may well take part in a number of reactions between the nanoparticles' surfaces and PI fragments affecting thermal properties of the polymers depending on their structure.
In this context, the capability of other transitional metal oxide nanoparticles to improve thermo-oxidative stability of PIs with various structures seems rather interesting, since these nanoparticles are known to possess quite active surfaces with regard to absorption of oxygen molecules from their surroundings and production of AOS [15]. Among the inorganic nanomaterials, TiO 2 and ZrO 2 are non-toxic, low-cost, and widely used nanofillers exhibiting high surface activity and good dispersibility in the matrix. They also possess high thermal and chemical stability, high refractive index, and dielectric constant. Moreover, they are able to crystallize in a number of polymorphs, whose structural parameters and physical properties differ from one another [18,19]. Both types of nanoparticles have shown good application prospects in improving the functional properties of PIs [12,13,[20][21][22][23]. It should be pointed out that the electrical and optical properties of the PI-based nanocomposites doped with the said types of nanoparticles are discussed profoundly in the scientific literature, since the MO are known to possess a quite large band gap [22]. Data on the thermal and mechanical characteristics of the nanocomposites are scarcer and controversial, the same type of MO nanoparticles can affect these properties oppositely [13,21,24,25].
In this study, we have fabricated a set of nanocomposite materials doped with ZrO 2 and TiO 2 . Since it is well-known that the surface activity of nanoparticles is determined by their chemical nature, size, and the presence of additional groups on the surface, we employed a number of synthesis techniques in order to investigate the impact of these factors along with the concentration of the nanofillers on the structure, morphology, and thermal and mechanical properties of aromatic PIs of various structures.

Synthesis of Nanoparticles and R-BAPS Matrix Polymer
To assess the effect of the preparation method of the nanoparticles on the properties on the nanocomposite materials, zirconium dioxide powders were synthesized from precursors of various natures. In the case of using inorganic salt ZrOCl2·*8H2O, the precipitation of zirconium oxyhydroxide was carried out with the use of an aqueous solution of ammonia, followed by washing the formed precipitate by decantation. The precipitate was dried at 100 °C to constant weight and then subjected to hydrothermal treatment. A four-hour isothermal treatment of ZrO(OH)2 under hydrothermal conditions at 250 °C and a pressure of 15 MPa led to the formation of nanocrystalline zirconium dioxide with a large number of hydroxyl groups remaining on their surface ( Figure S1) [26]. The resulting powder is designated ZrO2(hydroph). In some cases, ZrO2(hydroph) nanoparticles were dried at 160 °C for 1 h in order to remove the hydroxyl groups from the surface, the obtained powder being designated ZrO2(dried).
An alternative approach was the high-temperature hydrolysis of organic Zr(C2H7O2)4 under solvothermal conditions [27]. The resulting suspension of nanoparticles was washed with ethanol while stirring and centrifuging, and then dried to a constant weight. To remove residual organic impurities from the surface of nanoparticles and to control their size, zirconia powders were annealed in air at 500 °C for 2 h, and at 800 °C for 10 min and 2 h according to the temperature-time dependencies of their sizes presented in [28]. Nanoparticles of sizes 8, 18, and 28 nm were obtained, and designated ZrO2(8 nm), ZrO2(18 nm), ZrO2(28 nm), respectively.
Titanium dioxide nanoparticles were obtained by solvothermal synthesis. Titanium butoxide was added dropwise to toluene and stirred continuously for 24 h, the reaction mixture was then transferred to a steel autoclave with a Teflon cell with 80% filling factor.

Synthesis of Nanoparticles and R-BAPS Matrix Polymer
To assess the effect of the preparation method of the nanoparticles on the properties on the nanocomposite materials, zirconium dioxide powders were synthesized from precursors of various natures. In the case of using inorganic salt ZrOCl 2 ·*8H 2 O, the precipitation of zirconium oxyhydroxide was carried out with the use of an aqueous solution of ammonia, followed by washing the formed precipitate by decantation. The precipitate was dried at 100 • C to constant weight and then subjected to hydrothermal treatment. A four-h isothermal treatment of ZrO(OH) 2 under hydrothermal conditions at 250 • C and a pressure of 15 MPa led to the formation of nanocrystalline zirconium dioxide with a large number of hydroxyl groups remaining on their surface ( Figure S1) [26]. The resulting powder is designated ZrO 2(hydroph) . In some cases, ZrO 2(hydroph) nanoparticles were dried at 160 • C for 1 h in order to remove the hydroxyl groups from the surface, the obtained powder being designated ZrO 2(dried) .
An alternative approach was the high-temperature hydrolysis of organic Zr(C 2 H 7 O 2 ) 4 under solvothermal conditions [27]. The resulting suspension of nanoparticles was washed with ethanol while stirring and centrifuging, and then dried to a constant weight. To remove residual organic impurities from the surface of nanoparticles and to control their size, zirconia powders were annealed in air at 500 • C for 2 h, and at 800 • C for 10 min and 2 h according to the temperature-time dependencies of their sizes presented in [28]. Nanoparticles of sizes 8, 18, and 28 nm were obtained, and designated ZrO 2 (8 nm), ZrO 2 (18 nm), ZrO 2 (28 nm), respectively.
Titanium dioxide nanoparticles were obtained by solvothermal synthesis. Titanium butoxide was added dropwise to toluene and stirred continuously for 24 h, the reaction mixture was then transferred to a steel autoclave with a Teflon cell with 80% filling factor. The parameters of isothermal treatment were chosen in accordance with the method presented in [29], the TiO 2 nanoparticles obtained being of size 14 nm.
In order to synthesize R-BAPS polyimide the diamine BAPS was dissolved in NMP at 20 • C under argon. Then, an equimolar amount of dianhydride R was added to the resulting solution while stirring. The reaction was carried out in an argon flow. Stirring was continued for five hours to complete the polycondensation process.

Preparation of Pristine and Nanocomposite Films
PI/MO nanocomposites were fabricated by a standard solution casting technique [30,31]. The proper amounts of MO nanoparticles in NMP were sonicated for 1 h and then mixed with corresponding PAA solutions. The nanocomposite solutions were stirred for 24 h in order to form a quasi-homogeneous system. The compositions obtained as well as the reference bare PAA solutions were cast on glass supports and dried for 4 h at 80 • C. The final processing of the films obtained was performed by gradual heating up to 365 • C for PMDA-ODA and 300 • C for R-BAPS, followed by curing at these temperatures for 30 min. A list of the prepared nanocomposite samples is presented in Supplementary (Table S1).

Characterization Techniques
In order to investigate the structural and morphological features of the nanocomposite samples obtained as well as their constituents, a number of techniques were used.
Scanning electron microscopy (SEM) was performed on the cryo-cleavages of unfilled and nanocomposite films using a SUPRA-55VP microscope (Carl Zeiss, Oberkochen, Germany) equipped with a secondary electron detector. The specimens were fixed on microscope holders with special glue and covered with a thin layer of platinum.
Atomic force microscopy (AFM) data were obtained using an SPM-9700HT scanning probe microscope (Shimadzu, Kyoto, Japan). The AFM images were taken in air at room temperature. For the instrument operation, we chose a tapping mode, using NCHR-30SS Silicon tips with curvature radius less than 2 nm. The images 1024 × 1024 points were obtained.
IR spectra of unfilled and nanocomposite films were recorded on a Vertex 70 IR Fourier spectrometer (Bruker, Billerica, MA, USA) with the ATR (Attenuated Total Reflection) reflector (Pike Technologies, WI, USA) at room temperature in the range of 4000-500 cm -1 (resolution 4 cm -1 , number of scans 30) with a ZnSe working element. When registering the ATR spectra, a correction was introduced that takes into account the penetration depth depending on the wavelength. The equipment was employed to confirm the presence of functional groups on the surface of ZrO 2(hydroph) nanoparticles.
X-ray phase analysis of the samples was carried out on a Rigaku SmartLab 3 diffractometer (Rigaku Corporation, Tokyo, Japan) with CuKα radiation. The diffraction (XRD) patterns were taken in the range of angles 2θ = 10-60 • at a speed of 1 • /min. The phase composition of the nanoparticles was determined with the use of PDWin 4.0 software (NPO "Burevestnik", Russia) employing profile analysis of XRD patterns, the results of the analysis being compared with the ASTM database. The crystallite size was calculated from the broadening of X-ray diffraction lines according to the Scherrer equation [32].
Thermal properties of the materials were investigated with the use of a DTG-60 setup (Shimadzu, Kyoto, Japan) capable of simultaneous thermogravimetric (TGA) and differential thermal analysis (DTA). The specimens were heated up to 600 • C at a rate of 5 • C/min in air flow (80 mL/min). Thermal characteristics τ 10 and τ 50 (the temperatures at which a polymer or a composite loses 10 and 50% of its initial weight, respectively, because of thermal destruction) were determined using TGA curves. Data on glass temperature, T g , of the samples as well as on their supramolecular structure were obtained from DTA curves.
A thermomechanical analysis (TMA) was employed in order to investigate the mechanical behavior of the film samples under heating. A TMA 402 F1 Hyperion thermal Analyzer (NETZSCH, Selb, Germany) was used. The specimens were heated at a rate of 5 • C/min in argon flow (70 mL/min). T g values were determined from TMA curves.
The study on the mechanical characteristics of the unfilled and nanocomposite PI films was conducted with the use of an AGS-X 5kN (Shimadzu, Kyoto, Japan) setup. Mechanical tests were carried out in a uniaxial extension mode at room temperature. Strip-like samples 2 × 30 mm 2 in size were stretched at a rate 10 mm/min. The Young's modulus E, yield stress σ y , break stress σ b , and the ultimate deformation ε b for each sample were determined. For each composition 8-10 strips were tested, and the average and the standard deviation were calculated.

Investigation of Structure and Morphology
To assess the distribution of the nanofillers within polymer matrices and to investigate the morphology of the materials obtained, SEM and AFM images of the samples were analyzed (Figures 2 and 3). Mechanical tests were carried out in a uniaxial extension mode at room temperature. Strip-like samples 2 × 30 mm 2 in size were stretched at a rate 10 mm/min. The Young's modulus E, yield stress σy, break stress σb, and the ultimate deformation εb for each sample were determined. For each composition 8-10 strips were tested, and the average and the standard deviation were calculated.

Investigation of Structure and Morphology
To assess the distribution of the nanofillers within polymer matrices and to investigate the morphology of the materials obtained, SEM and AFM images of the samples were analyzed (Figures 2 and 3).     (Figure 2c). In contrast, ZrO2(8 nm) tends large (up to 1 µ m) aggregates in the matrix. The distribution of these aggr the PI matrix was still uniform and the polymer seemed to get adhered to Comparing Figure 3d,e we also found that drying of the as-prepared zirconia powder had a direct impact on the surface topography and roughness of the material. Introduction of ZrO 2(hydroph) caused the formation of clearly visible nodular topography (Figure 3d) while the surface of the PMDA-ODA/5%ZrO 2(dried) film demonstrated smooth morphology with the emergence of uniformly distributed hills (Figure 3e), probably clusters of ZrO 2 . The values of surface roughness were estimated to be Rq = 9.0 nm and 2.0 nm for PMDA-ODA/5%ZrO 2(hydroph) and PMDA-ODA/5%ZrO 2(dried) samples, respectively. Figure 4 shows IR spectra of the bare and nanocomposite PI film samples. The spectrum of the initial PMDA-ODA film exhibited the characteristic peaks at 1777 cm −1 (symmetric C=O stretching vibrations), 1714 cm −1 (antisymmetric C=O stretching vibrations), 1372 cm −1 (C-N stretching vibrations), and 725 cm −1 (deformation vibrations of C=O in the cycle), attributed to the imide cycle (Figure 4a), which indicates the conversion of PAA to the corresponding PI. When zirconium dioxide nanoparticles were introduced into the PI matrix, a change in the shape of the band of antisymmetric stretching vibrations of the C=O group of the imide ring was observed (Figure 4b). A slight shift to the highfrequency region, which is a typical sign of the weakening of the hydrogen bond network, took place. In contrast, the noticeable broadening of the same band to the low-frequency region was recorded for a PMDA-ODA filled with TiO 2 nanoparticles, which may indicate the development of a network of hydrogen bonds. The values of surface roughness were estimated to be Rq = 9.0 nm and 2.0 nm for PMD ODA/5%ZrO2(hydroph) and PMDA-ODA/5%ZrO2(dried) samples, respectively. Figure 4 shows IR spectra of the bare and nanocomposite PI film samples. The spe trum of the initial PMDA-ODA film exhibited the characteristic peaks at 1777 cm −1 (sy metric C=O stretching vibrations), 1714 cm −1 (antisymmetric C=O stretching vibration 1372 cm −1 (C-N stretching vibrations), and 725 cm −1 (deformation vibrations of C=O in t cycle), attributed to the imide cycle (Figure 4a), which indicates the conversion of PAA the corresponding PI. When zirconium dioxide nanoparticles were introduced into the matrix, a change in the shape of the band of antisymmetric stretching vibrations of t C=O group of the imide ring was observed (Figure 4b). A slight shift to the high-frequen region, which is a typical sign of the weakening of the hydrogen bond network, took pla In contrast, the noticeable broadening of the same band to the low-frequency region w recorded for a PMDA-ODA filled with TiO2 nanoparticles, which may indicate the dev opment of a network of hydrogen bonds.  According to X-ray diffraction data, a powder of zirconium dioxide nanopartic obtained from an inorganic precursor consisted of a mixture of tetragonal and monoclin polymorphs in a ratio of 80:20 (Figure 5a). The crystallite size of the monoclinic phase w 21 ± 3 nm, and that of the tetragonal phase was 17 ± 2 nm. In the case of using the organ chelate-forming compound zirconium acetylacetonate as a precursor in the synthesis nanoparticles, only crystallites in the monoclinic modification were formed. The size m-ZrO2 crystallites was assessed to be 8 ± 1 nm. Subsequent annealing of the powder of 800 °C for 10 and 120 min resulted in the removal of the residues of organic compoun from the surface of nanoparticles and led to an increase in the average size of the cohere scattering region (CSR) up to 18 ± 2 ( Figure 5b) and 28 ± 2 nm, respectively. According to X-ray diffraction data, a powder of zirconium dioxide nanoparticles obtained from an inorganic precursor consisted of a mixture of tetragonal and monoclinic polymorphs in a ratio of 80:20 (Figure 5a). The crystallite size of the monoclinic phase was 21 ± 3 nm, and that of the tetragonal phase was 17 ± 2 nm. In the case of using the organic chelate-forming compound zirconium acetylacetonate as a precursor in the synthesis of nanoparticles, only crystallites in the monoclinic modification were formed. The size of m-ZrO 2 crystallites was assessed to be 8 ± 1 nm. Subsequent annealing of the powder at of 800 • C for 10 and 120 min resulted in the removal of the residues of organic compounds from the surface of nanoparticles and led to an increase in the average size of the coherent scattering region (CSR) up to 18 ± 2 ( Figure 5b) and 28 ± 2 nm, respectively.
A broad amorphous halo was registered on the XRD pattern of the PMDA-ODA film in the region of diffraction angles of 13-40 • (Figure 5a-c, blue curve). In the case of composite films, one can also observe reflections corresponding to the initial MO nanoparticles which the PI was filled with (Figures 3 and 4, blue, gray, and orange curves). The shape and position of PMDA-ODA halo remained intact implying that the amorphous structure of the PI did not change by MO nanoparticles. nanoparticles, only crystallites in the monoclinic modification were formed. m-ZrO2 crystallites was assessed to be 8 ± 1 nm. Subsequent annealing of the of 800 °C for 10 and 120 min resulted in the removal of the residues of organic c from the surface of nanoparticles and led to an increase in the average size of t scattering region (CSR) up to 18 ± 2 ( Figure 5b) and 28 ± 2 nm, respectively. A broad amorphous halo was registered on the XRD pattern of the PMDA in the region of diffraction angles of 13-40° (Figure 5a-c, blue curve). In the c posite films, one can also observe reflections corresponding to the initial MO cles which the PI was filled with (Figures 3 and 4, blue, gray, and orange c shape and position of PMDA-ODA halo remained intact implying that the structure of the PI did not change by MO nanoparticles. Figure 6 and Table 1 demonstrate the effect of the MO nanofillers on properties of PI-based nanocomposites. All the TGA curves revealed excell  Figure 6 and Table 1 demonstrate the effect of the MO nanofillers on the thermal properties of PI-based nanocomposites. All the TGA curves revealed excellent thermal stability of the samples up to~420 • C. At the first stage (up to 300 • C), all the residues of low-molecular weight compounds and surface water were removed from the samples, the weight loss at this step being ca. 2 wt.%. This implies the complete imidization of PAAs and the formation of PIs during the film fabrication. The TGA curves of all the investigated samples showed a single degradation step. However, the impact of the MO nanoparticles on the thermal behavior of the nanocomposites differs depending on the PI matrix used. A PI containing sulfonyl groups (R-BAPS) has its thermal stability enhanced being filled with MO nanoparticles. On the other hand, one may notice contradictory trends in the thermal behavior of a PI with no SO 2 -groups (PMDA-ODA) doped with these nanoparticles depending on their characteristics. Such a discrepancy between the two matrices agrees well with the data we reported earlier [10,17] for another type of MO nanoparticle, namely CeO 2 . The mechanism of PI-MO nanofiller interactions causing the enhancement/deterioration of PIs' thermal stability is discussed in [17]. In brief, active oxygen species (oxide, peroxide, superoxide) (AOS) formed on the surfaces of MO nanoparticles during PI heating in air are supposed to take part in a number of chemical processes, inducing thermo-oxidative decomposition of the nanocomposite material.  It is well known that the smaller the nanoparticles, the more defects responsib adsorption of molecular oxygen and formation of the AOS are present on their su  It is well known that the smaller the nanoparticles, the more defects responsible for adsorption of molecular oxygen and formation of the AOS are present on their surface [15]. It is obvious from Table 1 that addition of ZrO 2 nanoparticles of 8 nm in size in PMDA-ODA leads to a noticeable decrease in thermal stability of the polymer, while that of PMDA-ODA-based nanocomposites containing larger nanoparticles (18 and 28 nm) is even higher compared to the pristine matrix. Taking into account the AFM data (Figure 3c), demonstrating a more homogeneous surface topography of the sample containing ZrO 2 (18 nm), and enhanced mechanical properties of this nanocomposite (discussed below in Section 3.3), one would suggest that larger nanoparticles are less prone to aggregation and they facilitate the formation of additional links between PI macromolecules hindering the diffusion of free radicals and retarding the degradation of the nanocomposites.

Thermal Properties
An increase in ZrO 2 (8 nm) concentration up to 5 wt.% causes a substantial increase in thermal stability (see Figure 6a and Table 1) which likely results from the agglomeration of the nanoparticles with high surface energy [33] and the reduction in the number of the centers producing the AOS. The agglomeration is also confirmed by SEM images (compare Figure 2e,f) as well as by the gradual deterioration in mechanical properties of the nanocomposites (discussed below in Section 3.3) with higher contents of nanoparticles, since it is well known that the aggregates cause inhomogeneity of a resulting material reducing its mechanical strength [2,34] In the case of R-BAPS possessing SO 2 -groups in its macromolecules, we presumed the following mechanism of PI-nanofiller interactions [17]. Sulfonyl radicals (probably formed at the first step of the PI's thermal decomposition) bearing a positive charge on a sulfur atom can be electrostatically attached to the MO nanoparticles with negatively charged oxygen species adsorbed on their surfaces, causing a decrease in the number of volatile products in a certain temperature interval and thereby improving thermal stability of the R-BAPS-based nanocomposites (see Figure 6b and Table 1). One should notice that the enhancement was observed regardless of the type, preparation method, size, and concentration (at least in the investigated range) of the nanoparticles.
Analyzing the effect of zirconia preparation technique on the thermal stability, one can observe no substantial difference in the trends in the thermal behavior of the nanocomposites fabricated with the use of either zirconyl chloride or zirconium acetylacetonate as a precursor for nanoparticles. However, the nanoparticles prepared from zirconyl chloride with the subsequent drying of the obtained powder influenced the values τ 10 and τ 50 (both for PMDA-ODA and R-BAPS nanocomposites) more pronouncedly compared to the as-prepared nanoparticles (see Table 1). This is also likely explained by the surface features of the nanoparticles, since the drying leads to the removal of OH-groups from the surface and provides better distribution of the filler within a polymer matrix (see Figure 3d,e) and more sites for adsorption of oxygen molecules participating in the formation of the AOS.
Considering nanocomposites containing titania, it is evident that this nanofiller improves their thermal stability. Apparently, it relates to the quite large size (14 nm) of the nanoparticles providing fewer catalytically active centers detrimental for PMDA-ODA matrix, but their amount is still enough for reactions with the sulfonyl radicals in the R-BAPS-based nanocomposites. Moreover, titania nanoparticles are better distributed in the PI matrices (see Figure 2c) and provide substantial enhancement of the mechanical properties (discussed below in Section 3.3) of the films. The nanoparticles are assumed to act as linkers between the polymer macromolecules, thereby inhibiting the thermal decomposition of the materials.

Mechanical Properties
The mechanical properties of the PI-based nanocomposites were compared with those of the pristine PI matrices and summarized in Table 2. The stress-strain curve of the PMDA-ODA matrix film is given in Figure 7a. The film stretched uniformly, showing no localized deformation. In contrast, a well-defined maximum indicating the yield point of R-BAPS is clearly evident from its deformation curve (Figure 7b). Such a feature is typical of flexible PIs containing several bridge groups [35], a so-called "neck" being formed on a film. The growth of the "neck" throughout a sample is responsible for its deformation beyond the yield point.  Tensile properties of the nanocomposite PMDA-ODA-as well as R-BAPS-based f exhibit quite similar trends (see Table 2). For instance, at definite combinations of size concentration, the MO nanofillers augmented the Young's modulus of both PIs, the y point also being higher or at least unchanged compared to that of the pristine polym However, the ultimate elongation of the composite films always decreased when the noparticles were added to the pristine PIs. Such an impact of the nanoparticles on mechanical properties of not only PIs but also other polymers is quite consistent with literature data [20,23,36]. The reinforcing effect of MO nanoparticles has been attribu to a certain degree of linking between PI macromolecules and the nanofillers ensuring external loading is transferred from the continuous polymer phase to the filler. The p tive impact of titania was observed to be more pronounced against that of zirconia. Tak into account the SEM images (Figure 2) of the nanocomposites, one could relate this to less pronounced aggregation and more homogeneous distribution of the former nano ticles in the PI matrix. These factors along with compatibility of components and t structure are known to be crucial to mechanical characteristics of nanocomposite mate [2]. Tensile properties of the nanocomposite PMDA-ODA-as well as R-BAPS-based films exhibit quite similar trends (see Table 2). For instance, at definite combinations of size and concentration, the MO nanofillers augmented the Young's modulus of both PIs, the yield point also being higher or at least unchanged compared to that of the pristine polymers. However, the ultimate elongation of the composite films always decreased when the nanoparticles were added to the pristine PIs. Such an impact of the nanoparticles on the mechanical properties of not only PIs but also other polymers is quite consistent with the literature data [20,23,36]. The reinforcing effect of MO nanoparticles has been attributed to a certain degree of linking between PI macromolecules and the nanofillers ensuring that external loading is transferred from the continuous polymer phase to the filler. The positive impact of titania was observed to be more pronounced against that of zirconia. Taking into account the SEM images (Figure 2) of the nanocomposites, one could relate this to the less pronounced aggregation and more homogeneous distribution of the former nanoparticles in the PI matrix. These factors along with compatibility of components and their structure are known to be crucial to mechanical characteristics of nanocomposite materials [2].
It is also evident from Table 2 that an increase in the content of ZrO 2 nanoparticles having the same size led to deterioration in the mechanical characteristics of both PIs (compare the values of Youngs moduli of the PMDA-ODA-based nanocomposites with 3 and 7 wt.% of ZrO 2 (8 nm) and those of the R-BAPS-nanocomposites with 3 and 5 wt.% of ZrO 2 (18 nm)). Obviously, the more nanoparticles are introduced into a PI matrix, the more aggregates, i.e., defects, they form.
Comparing the stiffness and yield point of the nanocomposite materials containing the same amount of zirconia, one may notice that doping both matrix PIs with the larger nanoparticles (either 18 or 28 nm instead of 8 nm) resulted in an enhancement of their mechanical properties ( Table 2). This is probably connected with the tendency of the smallest nanoparticles possessing the highest surface energy to agglomerate [33] and to act as stress concentrators, while bigger nanoparticles provide better distribution of the nanofiller within the polymer matrix. This corresponds well with AFM results (compare Figure 3b,c).
It should be pointed out that mechanical characteristics of the nanocomposite materials are strongly affected by the preparation method of the nanoparticles. According to Table 2, the materials containing zirconia nanoparticles obtained from inorganic zirconyl chloride ZrO 2(hydroph) have their stiffness and strength drastically deteriorated if the as-prepared zirconia powder was not additionally dried. It supposedly related to the presence of OH-groups on the zirconia surface ( Figure S1). On the one hand, these groups can take part in the hydrogen bonding with imidic C=O groups in PIs improving compatibility with the polymer matrix and thereby enhancing mechanical properties of nanocomposite material [37]. On the other hand, the presence of active hydroxyl groups on the surface of the nanoparticles can cause their aggregation [38] in the PAA solution, the aggregates acting as defects in the resulting PI films (compare Figure 3d,e). Moreover, OH-groups may well facilitate adsorption of water molecules on the nanoparticles' surface thereby damaging the PAA because of the hydrolysis of the acid [39,40]. In contrast, nanocomposites (based on both PMDA-ODA or R-BAPS) filled with preliminarily dried ZrO 2(hydroph) nanoparticles turned out to exhibit enhanced mechanical properties.

Thermomechanical Properties
In order to investigate mechanical behavior of the nanocomposites while temperature increases, TMA was carried out. It is noteworthy that the two PI matrices under study differ in their thermomechanical performance (see Figure 8), therefore various stress values were applied to the samples in TMA. In the case of semi-flexible PMDA-ODA, and the corresponding nanocomposites, a standard 0.5 MPa loading was used. In flexible R-BAPS, on the other hand, a quite steep transition to a plastic state occurred even under little stress (see Figure 8b), the polymer stretching far beyond the deformation range registered by the setup. Thus, a smaller value of stress, viz. 25 kPa, was applied to R-BAPS-based materials. The analysis was performed in an inert atmosphere so as to prevent the formation of the so-called destruction crosslinks [41,42] between PI macromolecules caused by thermooxidative destruction of a polymer (for example, PDMA-ODA begins to decompose in air rather close to T g ). Such crosslinks retard polymer chain mobility after glass transition, thereby complicating the determination of T g .
The TMA curves of PDMA-ODA-based samples are shown in Figure 8a. The unfilled PI film demonstrated a noticeable deformation above T g . When the temperature reached 440 • C, an initial stage of the polymer thermal degradation accompanied with the formation of destruction crosslinks might have taken place since the elongation rate decreased gradually. To assess the crosslinking degree, a value of deformation, ∆ε, in the temperature range between T g and a temperature of a maximum deformation on the TMA curve was determined ( Table 3).
formation of the so-called destruction crosslinks [41,42] between PI macromolec caused by thermo-oxidative destruction of a polymer (for example, PDMA-ODA beg to decompose in air rather close to Tg). Such crosslinks retard polymer chain mobility a glass transition, thereby complicating the determination of Tg.   The effect of different types of nanofillers on the thermomechanical behavior of PMDA-ODA has been determined. It is evident from Figure 3 and Table 3 that regardless of their type, size, and concentration, all the nanoparticles caused a decrease in T g of this PI. Since in some cases the stiffness of PMDA-ODA was augmented by the introduction of MO nanoparticles at room temperature (see Section 3.3, Table 2), implying that the nanofiller can act as a linker between the polymer macromolecules, one may suppose that the links may well weaken at higher temperatures (near T g ) or be more flexible compared to the polymer-polymer bonds. Moreover, the nanoparticles and their aggregates may become a hindrance to the inter-as well as intramolecular interactions between PMDA-ODA macrochains, thereby decreasing its T g .
The introduction of MO nanoparticles also affected the shape of TMA curve of PMDA-ODA at temperatures above T g , the thermomechanical behavior of the PI depending on the size and concentration of the nanofiller. At the smallest content of ZrO 2 (8 nm), the nanoparticles were supposed to have an impact on segmental mobility of the PI macromolecules, decreasing its flexibility in this temperature range. We have already observed the similar respond of this PI on the introduction of nanoparticles of other types (CeO 2 and nanocarbon) [35]. Further investigations are needed in order to elucidate the reasons for such an effect. From the practical viewpoint, this means that in the case of the applied mechanical loading not being very high, the polymer can maintain quite high stiffness even above its T g , which is likely to widen the temperature range of its applicability. The higher the concentration of ZrO 2 (8 nm) nanoparticles, i.e., the more aggregates preventing polymer-polymer interactions are formed, the more flexible behavior PMDA-ODA demonstrates after glass transition. The same corresponds to an increase in size of the nanoparticles from 8 to 28 nm at their constant concentration (5 wt.%) (see Table 3). Considering titania-containing nanocomposites, one should mention that according to mechanical The values of T g in the unfilled and nanocomposite film samples were determined. It was found that, regardless of their type, size, and concentration, MO nanoparticles caused a decrease in T g of both matrix PIs. The effect of MO nanoparticles on the behavior of PI at temperatures above T g is demonstrated. Introduction of MO nanofiller into the semiflexible PMDA-ODA can lead to an increase in the rigidity of PI beyond glass transition, which can potentially expand the temperature range of its applicability. In the case of flexible R-BAPS, the introduction of MO nanoparticles caused a significant decrease in T fl of materials. Such an effect can be used in reducing the molding temperature of the PIs with a similar structure.
The data presented in this work can be useful for materials science and engineering since they provide the opportunity for the manipulating the functional properties of polyimide nanocomposites by varying the polymer composition as well as the concentration, size, and functionality of MO nanoparticles. This would allow for the design of PI-based materials with desirable functional properties, expanding the scope of their applications.