ynthesis and processing of polyimide – silica hybrid materials †

Inorganic–organic hybrid materials allow for combining features typical of the inorganic component with those of the organic component in one material. Generally, the preparation of organic and inorganic compounds requires considerably different synthesis conditions. Hence, the development of one-pot routes to inorganic– organic hybrid materials is challenging. We herein report a fully green one-pot synthesis of polyimide/silica (PI/SiO2) hybrids. Specifically, we co-condense both components hydrothermally, using nothing but the respective precursors and water. Furthermore, we show that the PI and the SiO2 component can be covalently connected under hydrothermal conditions, using the compatibilizer (3-aminopropyl)triethoxysilane. We thoroughly investigate the effect of different reaction conditions, including temperature, pH, precursor concentration and reaction time on the morphology and crystallinity of the final materials. The polyimide component, poly(hexamethylene pyromellitimide) was chosen for its thermoplasticity, which allows for processing both the PI and the PI/SiO2 via sintering. For being a solvent-free method, sintering qualifies as a green processing technique. This work is the first report of the simultaneous hydrothermal condensation of an inorganic and an organic material.


Introduction
The continuously increasing level of technologization of our modern world generates the need for even more sophisticated materials. Unfortunately, the majority of currently employed syntheses ofespecially organicmaterials, are harsh and toxic. Mankind's growing awareness of human health and the environment creates the demand for less harmful procedures for both producing and processing materials. The development of chemical reactions which minimize environmental impact is summarized by the principles of green chemistry, 1 and is a highly timely topic. Green materials chemistry aims at responding to both the need for highly advanced materials, and the need for benign synthesis approaches. 2 There are a handful of major approaches to introduce a high level of sophistication into materials, amongst which is the preparation of hybrid materials. Hybrids or composites combine at least two chemically different compounds (and thus also their properties) into an intimately mixed or connected novel material. 3 Employing this concept, one can for instance couple the low densities and ease of processing found in organic polymers with the mechanical performance of inorganic materials, such as ceramics. 4 The resulting compound is a socalled inorganic-organic hybrid material. There are four major synthesis approaches towards inorganic-organic hybrid materials, which are (i) mixing the ready-made inorganic and the ready-made organic component; (ii) synthesizing the organic component in the presence of the pre-fabricated inorganic component; (iii) preparing the inorganic component in the presence of the pre-fabricated organic component; and (iv) synthesizing both components simultaneously. 2 In terms of green chemistry, such concurrent one-pot approaches are clearly preferable over the other routes, as the reduced number of synthesis steps automatically minimizes e.g. solvent consumption and purication steps. However, the development of one-pot syntheses of inorganic-organic hybrid materials is intrinsically challenging, since conventional preparation conditions of the two components oen vary considerably from each other. For instance, inorganic compounds oen require polar reaction media or high temperatures, while syntheses of organic materials are mainly carried out in comparatively apolar environments at rather moderate temperatures.
We have recently started to work on transferring a widely used inorganic synthesis route, hydrothermal synthesis, to the preparation of organic polymers. With hydrothermal polymerization (HTP), we have developed a geomimetic technique to synthesize fully aromatic polyimides (PIs) of high crystallinity. [5][6][7][8] The fact that a plethora of inorganic compounds, including for instance metal oxides, carbonates and sulfates, can be generated hydrothermally has been known for over 150 years. 9 Moreover, such processes are a major naturally occurring formation pathway of crystalline minerals. Since HTP also generates high crystallinity in polyimides, it is referred to as a geomimetic technique. HTP is inherently green, as it does not require anything besides H 2 O as the reaction medium, and the starting compounds, i.e. the monomers. This is indeed in stark contrast to conventional routes towards polyimides, which require the use of toxic and high-boiling (and hence costly to recycle) solvents such as dimethylformamide or N-methyl-2pyrrolidone. 10 The outstanding crystallinity generated via HTP is a feature that is highly sought aer for PIs, since it enhances certain material properties that these high-performance polymers are especially desired for, such as thermal or chemical stability. Consequently, hydrothermally synthesized PIs are typically insoluble in virtually any solvent, including conc. H 2 SO 4 at reux. 5 Moreover, PIs generated from HTP are typically devoid of thermally induced soening via a glass or melting transition. 5 Therefore, the advantages of HTP, i.e. the facts that the technique is green and at the same time leads to an enhancement of the properties of materials, go to date hand in hand with the major drawback of non-processability of the obtained PIs.
In this study, we present the rst example of a one-pot, fully green synthesis of an inorganic-organic hybrid material. Specifically, we selected a semialiphatic PI as the organic component, and a metal oxide silica as the inorganic component. The target PI was chosen, as semialiphatic PIs are known thermoplastics, 11 which we anticipated to allow for processing via sintering. In fact, sintering can be considered a green technique as it does not employ any organic solvents. Moreover, we expected semialiphatic PIs to be colorless, since the conventional orangebrownish color of PIs is believed to be related to those PIs typically comprising alternating electron-rich and electron-poor aromatic rings and their intermolecular packing into chargetransfer complexes. 12 The latter is unwanted in e.g. aerospace applicationswhere PIs are the external coating materials of choice, a consequence of their high resistance to cosmic radiationsince the yellowish coloring generates high solar absorptivities. 13 SiO 2 was selected as the inorganic component for (i) being the archetype of a metal oxide that can be prepared hydrothermally, and (ii) its white color, from which the target PI/ SiO 2 materials were expected to additionally prot.

Hydrothermal synthesis of PI6
Since HTP of semialiphatic PIs has to date not been reported, we decided to concentrate our efforts on one system only: PI6the polyimide resulting from the comonomers hexamethylene diamine (HMDA) and pyromellitic acid (PMA), see Fig. 1. The choice of PI6 was based on (i) the fact that it is a well-known, high-melting thermoplastic with a reported melting point T m of 454 C, 11 and (ii) the low-cost of both comonomers (both are used as starting compounds in industrially relevant syntheses: PMA or its dianhydride is one of the two comonomers of Kapton™; HMDA is one of the two comonomers of nylon 6,6).
To rst generally check the feasibility of PI6 synthesis via HTP, a dispersion containing HMDA and PMA in 1 : 1 ratio at c ¼ 0.1 mol L À1 in H 2 O was prepared. This dispersion was lled into a non-stirred autoclave (see the ESI † for the set-up), heated to the reaction temperature T R z 200 C, and kept there for a reaction time of t R ¼ 24 h. Aer t R , the autoclave was rapidly cooled back to RT by quenching with tap water. The obtained white dispersion was ltered, rinsed with deionized water, and dried at RT and reduced pressure (see the ESI † for experimental details). The ATR-FTIR spectrum of hydrothermally prepared PI6 (Fig. 2) agrees well with the literature. 14 The most prominent modes areñ s,C]O z 1765 cm À1 andñ as,C]O z 1700 cm À1 . Also, the imide ring's (O)C-N stretching mode at z1370 cm À1 and the H 2 C-N stretching mode at z1400 cm À1 are clearly present. Moreover, in the higher wavenumber region of 4000-2500 cm À1 , solely aromatic and aliphatic C-H modes are present (C Ar -H:  3101 and 3035 cm À1 ; C alkyl -H: 2941, 2858 cm À1 ). Any heteroatom-H modes (O-H, N-H), including H-bonding, would occur in this region. Hence, we conclude from ATR-FTIR analysis that neither of the possible end-groups NH 2 , NH 3 + or CO 2 H is present. Two further end-groups are conceivable, namely CO 2 À and cyclic anhydrides. Neither the characteristic carboxylate C]O mode (z1570 cm À1 ) nor the characteristic C]O vibration of pyromellitic anhydride (z1820 cm À1 ) is present, which also excludes these end-groups. Hence, it can be stated that PI6 is fully condensed as judged from ATR-FTIR spectra. Note that n as,C]O z 1700 cm À1 occurs at almost the same wavenumber as the C]O vibration in free PMA would (1698 cm À1 , cf. Fig. 2).
Consequently the presence of a strong vibration at 1700 cm À1 cannot be viewed as indicative of PMA consumption and imide formation by itself, but must be considered together withñ s,C]O (z1765 cm À1 ). Powder X-ray diffraction (PXRD) of PI6 conrms very high crystallinity (Fig. 3A). This is indeed expected, since HTP is a polymerization technique that generates outstandingly crystalline polyimides. [5][6][7] The diffractogram shows barely any amorphous background and the major reections are rather sharp, cf. Moreover, we nd six smaller and rather broad reections between 10 and 35 (2q, Cu-Ka), labelled E-J (Fig. 3A). To our knowledge, to date, there is only one article which reported on the PXRD analysis of PI6. 15 The authors prepared PI6 by polycondensation in ethylene glycol (at 130 C) and by solid-state polycondensation of the monomer salt (at 180 C). Both polymerizations lead to PI6 of rather poor crystallinity, with diffractograms showing only two reections, at z7 and z19 (2q, Cu-Ka), respectively. We speculate that these reections correspond to the most intense reections in PI6 prepared by HTP (A and B in Fig. 3A).
Scanning electron microscopy (SEM) evinces a highly homogeneous sample (cf. ESI †), exclusively composed of ribbons of z50 mm in length, z1-3 mm in width, and z100 nm in thickness. These ribbons are additionally covered with highly oriented, smaller ribbons (or scales) growing away from the parent ribbons almost perpendicularly (Fig. 3B). These structures are indicative of dendritic growth.
Aer this rst proof-of-concept, we studied the effect of three parameters on the hydrothermal synthesis of PI6: T R , t h and pH. Aiming at lower T R and t R is highly interesting in terms of green chemistry (lowering the energy consumption). The pH was additionally studied, since polycondensationsthe formation reaction of both polyimides and SiO 2can typically be sped up by either acidic or basic catalysis. However, PIs do not subsist basic conditions: both mineral and organic bases tend to attack the cyclic imide moiety and generate oligo-/poly(amic acid)s and oligo-/poly(amic carboxylates). 16,17 Therefore, we only investigated acidic conditions. Specically, we studied two T R , 140 and 200 C; four different t R , 2, 4, 24, and 48 h; and additional acidication with HCl to pH ¼ 1. Note that a dispersion of the comonomers PMA and HMDA is already acidic (pH ¼ 4), due to the acidity of the tetracarboxylic acid PMA. When studying PI6 formation at the signicantly lower reaction temperature T R z 140 C, from ATR-FTIR spectra (ESI †) it becomes clear that PI6 does not fully form even aer 48 h for pH ¼ 1, and appears nearly complete aer 48 h at pH ¼ 4 (non-acidied samples). Interestingly, at pH ¼ 1 and T R ¼ 140 C, a large number of broad modes at 3750-2250 cm À1 is found for any reaction time. These modes are attributed to protonation of e.g. CO 2 À and NH 2 groups, in connection with H-bonding between these groups (cf. ESI †). We hypothesize that these samples contain considerable amounts of protonated low-molecular weight compounds, i.e. monomers, monomer salts and eventually dimers, that did not yet react to the polyimide. This hypothesis is further strengthened by PXRD data: samples prepared at T R ¼ 140 C at pH ¼ 1 either show many sharp reections (indicative of the presence of crystalline small molecules) that do not t the major reections of PI6 (at t R ¼ 2 h, 4 h), or show both reections associated with PI6 and many sharp reections. SEM analysis of all the products of all attempts to synthesize PI6 at 140 Cno matter if pH ¼ 1 or pH ¼ 4corroborates the hypothesis that crystalline small molecule species form rather than the desired PI. Specically, we nd numerous angular or otherwise regularly intergrown objects of forms indicative of crystallinity, of considerable size (up to 0.5 mm). These morphologies are clearly too big to be polyimide crystallites. As PMA is a well- known tecton for supramolecular solid-state assemblies, such as co-crystals and salts, 18,19 we speculate that several different compounds of this type form hydrothermally at 140 C. Consequently, we did not further pursue low T R . For T R ¼ 200 C, we only investigated shorter t R , since full condensation and high crystallinity had already been obtained at t R ¼ 24 h (Fig. 2 and 3). ATR-FTIR analysis indicates full condensation even aer t R ¼ 2 h and 4 h (ESI †), for both pH ¼ 1 and pH ¼ 4. From PXRD we nd high crystallinity in samples obtained at pH ¼ 1: all major (A-D) and minor reections (E-J) found for t R ¼ 24 h at pH ¼ 4 ( Fig. 3A) are present for both 2 h and 4 h. However, PI6 obtained aer 2 h shows one additional reection at approx. 10 (2q, Cu-K a ), which is most likely due to the presence of shorter oligomers. 8 PI6 obtained at pH ¼ 1 at T R ¼ 200 C appears fully condensed from ATR-FTIR at all tested t R (2 h, 4 h, 24 h)the spectra (ESI †) t the literature and the spectrum of PI6 obtained aer 24 h at pH ¼ 4 ( Fig. 2) well. The PXRD pattern shows the presence of the major reections (A-D, Fig. 3A). In addition, there are however many strong reections (t R ¼ 2 h) that we attribute to monomer salts/oligomers and weak (t R ¼ 4 h, 24 h) reections that are not present in PI6 synthesized at pH ¼ 4 (see the ESI †). To our surprise, we must conclude that the PI6 formation is slower and changed at pH ¼ 1 as compared to pH ¼ 4. This effect could be due to the increased acidity, but could also be related to the presence of Cl À ions in the system. Morphology-wise, we nd ribbon morphologies at pH ¼ 4 (cf. Fig. 3B and the ESI †), and platelettype morphologies at pH ¼ 1 (cf. Fig. 3C and the ESI †), for all t R . To investigate the differences in morphology and crystal structure of PI6 as a function of the pH, we performed energy dispersive X-ray spectroscopy (EDAX) analysis on samples synthesized at T R ¼ 200 C, t h ¼ 24 h, at pH ¼ 4 and pH ¼ 1, respectively. From EDAX spectra (cf. ESI †), it becomes clear that PI6 synthesized at pH ¼ 4 contains exclusively C, N and O (note that H cannot be detected via EDAX), while PI6 synthesized at pH ¼ 1 contains Cl. Therefrom, we infer that despite thorough washing of PI6 samples with aq. dest. prior to drying, HCl is retained by the polyimide. While further investigations would be necessary to clearly identify the nature of the retention, we speculate that HCl forms adducts with either imide carbonyl oxygen or imide nitrogen atoms: for the presence of lone pairs these parts of the PI chains are the only conceivable loci in the polymer where an adduct with HCl can be formed. For the preparation of PI6/SiO 2 hybrid materials, we refrained from pursuing HCl addition any further.

Hydrothermal synthesis of PI6/SiO 2 hybrid materials
The silica precursors used for the synthesis of PI6/SiO 2 hybrid materials are tetraethoxysilane (TEOS) and (3-aminopropyl)triethoxysilane (APTES). The condensation reaction equations of both TEOS and PI6 in the presence of APTES and TEOS are depicted in Fig. 4. TEOS is a standard sol-gel precursor of SiO 2 , and APTES is a standard network-modier that has been used to (i) introduce amino-functionalities in SiO 2 , 20 and (ii) as a compatibilizer between SiO 2 and PIs in the preparation of PI/SiO 2 hybrids. 21,22 Note that other approaches to covalently link PIs to SiO 2 include (i) employing epoxy-functionalized silicon alkoxides, 23 (ii) the use of PI monomers with pendant OH groups able to co-condense with TEOS, 24 (iii) employing pre-formed PMMA-functionalized SiO 2 particles, where the PMMA-functionalization allows for crosslinking with the PI via transesterication, 25 (iv) to decorate PIs with pendant silane side-groups that are able to co-condense with TEOS, 26 or to employ APTES end-capped poly(amic acid) species. 27 However, all the PI/SiO 2 hybrids and composites reported to date have been prepared employing classical, non-green routes, i.e. by polycondensation in toxic solvents such as DMF or NMP via poly(amic acid) intermediates. [21][22][23][24][25][26][27] From our studies of HTP of PI6, we restricted the employed reaction parameters to T R ¼ 200 C, t R ¼ 24 h and the autogenous pH ¼ 4, generated by the acidity of PMA. We investigated three types of precursor mixtures: (i) HMDA/PMA/TEOS (experiments #1-3) (ii) HMDA/PMA/TEOS/APTES (experiments #4-6), and (iii) HMDA/PMA/APTES (experiments #7-9), all summarized in Table 1.
The rst set of experiments (#1-3) was aimed at a rst proofof-concept that PI6 and SiO 2 can be synthesized simultaneously. ATR-FTIR spectra of all three obtained hybrids show all major modes of PI6 and lack modes of unreacted monomers or endgroups (see the ESI †), as discussed previously and shown in Fig. 2. In addition, we nd the presence of the characteristic siloxane (Si-O-Si) asymmetric vibration at 1080 cm À1 (ref. 28) that clearly grows in intensity with increasing molar amount of the supplied TEOS (cf. ESI †). However, we do not nd silanol modes, which would be expected at approx. 3500-3000 cm À1 (SiO-H) and approx. 960 cm À1 (Si-OH). 28 It should be mentioned that the Si-OH mode at ca. 960 cm À1 would be hard to assign, since it overlaps with several peaks in the ngerprint region of the PI6 backbone. Note that the 960 cm À1 peak as well as the characteristic 800 cm À1 (Si-O bond bending vibration) 28 are present in SiO 2 that we prepared hydrothermally from TEOS at pH ¼ 4 for comparison (see the ESI †). EDAX further conrms the From SEM images, it becomes clear that the morphology of PI6 is retained at all three concentrations of TEOS: as for pure PI6, we nd ribbon morphologies (see Fig. 5C and the ESI †). In addition, the presence of spherical nanoparticles (approx. 100-200 nm) that are agglomerated (Fig. 5B) and oen cover the PI6 ribbons, especially at the high initial amount of TEOS (n TEOS ¼ 7.5 mmol; exp. #3, SEM see the ESI †) is also evidenced. From SEM it becomes apparent that the number of particles increases with increasing initial TEOS concentration. EDAX evinces that these nanoparticles contain mostly Si and O, which allows for the conclusion that they are composed of SiO 2 . ‡ Furthermore, the diffractograms of all three PI6/SiO 2 materials show the presence of all (major and minor) reections of PI6 (Fig. 5C, and compare Fig. 3A). Moreover, with increasing n TEOS , especially evident for n TEOS ¼ 3.0 mmol and 7.5 mmol, a growing amorphous halo becomes apparent, which we attribute to SiO 2 . Note that at n TEOS ¼ 1.5 mmol, the diffractogram shows two small but sharp reections at approx. 30 and 32 2q, Cu-K a (highlighted by orange arrows in Fig. 5A). We speculate that these reections are related to a small amount (as not apparent from ATR-FTIR spectra) of either unreacted monomers or low-molecular weight oligomers, which might have been enclosed in freshly formed PI6. From the analysis of experiments #1-3, we can draw the conclusions that TEOS, HMDA and PMA can simultaneously undergo polycondensation to a mixture of amorphous, spherical colloidal SiO 2 particles and crystalline PI6 ribbons. As expected, the products are mixtures of SiO 2 and PI6, which are not covalently bound to each other.
In a second set of experiments (#4-6), we employed APTES (Fig. 4) as an additional compatibilizer. APTES has been used as a compatibilizer in several studies on PI/SiO 2 hybrids. 21,22 On one hand, the presence of Si-OEt moieties allows for hydrolysis to Si-OH, and subsequent co-condensation with Si-OH from hydrolyzed TEOS to Si-O-Si functions. On the other hand, the NH 2 functionality of the aminopropyl moiety is able to cocondense with 1,2-dicarboxylic acid moieties of PMA to form imide functions (Fig. 4). Therefore, the use of APTES is known to enable covalent bonding between PI and SiO 2 which should allow for evolution from a mere PI/SiO 2 mixture to a class II hybrid material (dened as a hybrid in which the different components are connected via covalent or iono-covalent bonds). 3 When additionally employing APTES, we adjusted the stoichiometric ratio of reactants such as n(NH 2 ) : n(CO 2 H) ¼ 1 : 2, which corresponds to equimolarity of amino and 1,2dicarboxylic acid moieties. This is extremely important, since polyimides are polycondensates obeying Carothers' law, which states that high conversion and degrees of polymerization can only be obtained at 'ideal' equimolarity of reacting functions. In experiments employing PMA/HMDA/TEOS/APTES, we tested different n APTES . Specically the reactant ratio was adjusted to Moreover, we do not nd any indication for linking functions besides the imide moiety. Other linking functions that would be conceivable are diamide or amic acid moieties (cf. Fig. 4): no amide modes are observed in ATR-FTIR spectra. PXRD results in completely different patterns than those obtained for PI6/SiO 2 mixtures. In all three diffractograms, the minor reections of Table 1 Overview of PI6/SiO 2 hybrid materials syntheses. All experiments were carried out at T R ¼ 200 C, t R ¼ 24 h, pH ¼ 4, V ¼ 15 mL; n is given in [mmol] in all columns # n PMA n HMDA n APTES n TEOS P n(NH 2 ) P n(CO 2 H) P n(Si) PI6 (E-J in Fig. 3) are fully absent (Fig. 6). At the lowest employed APTES amount (n APTES ¼ 0.6 mmol), the major reections of pure PI6 (A-D, compare Fig. 6) are clearly visible (Fig. 6A). Note that A-D are situated on top of an amorphous underground, which can be described by two overlapping amorphous halos, centered around approx. 9 and 19 (2q, Cu-K a ), respectively (orange arrows in Fig. 6). Note that the position of these halos is signicantly different from amorphous SiO 2 prepared by hydrothermal condensation of TEOS for comparison (grey diffractogram in Fig. 6). Moreover, three new relatively sharp reections are present (labelled K, L and M in Fig. 6). At n APTES ¼ 0.9 mmol, K, L and M are also present and found on top of two overlapping amorphous halos, however at changed relative intensities and the major reections of PI6 (A-D) are fully absent. At n APTES ¼ 1.2 mmol, no sharp reections are present anymorenothing but the two amorphous halos are visible (Fig. 6A,  domains have some degree of ordering that is however different from pure PI6. We believe that this segregation is a consequence of the different polarities (and hence solubilities in hightemperature water) and speeds of condensation of HMDA and PMA, and TEOS and APTES, respectively. Moreover, our experimental set-up, i.e. the use of non-stirred autoclaves (see the ESI †), further supportsor at least does not disturbthe segregation processes. SEM reveals very interesting morphological changes. At n APTES ¼ 0.6 mmol, spheres of approx. 1-5 mm in diameter coexist with intergrown platelets of approx. 100-200 nm thickness and several microns in lateral extension (Fig. 6B). The amount of spherical objects is rather small with respect to the amount of intergrown platelets (Fig. 6B and ESI †for overview images). At higher amounts of APTES, i.e. n APTES ¼ 0.6 and 0.9 mmol, approx. 1-5 mm-sized spheres coexist with partly bundled and interwoven bent bers of ca. 100 nm in diameter and several microns in length. The relative number of spheres with respect to bers increases with growing n APTES (Fig. 6B and ESI †). EDAX results show that the relative amount of Si is higher in the spherical objects than in the bers (see the ESI †). The fact that spherical, Si-rich morphologies coexist with platelets (n APTES ¼ 0.6 mmol) and bers (n APTES ¼ 0.9 mmol and 1.2 mmol) further substantiates the hypothesis of segregated domains in the hydrothermal syntheses of these hybrid materials. Note that there are striking differences in the SiO 2 -rich morphologies between the absence of APTES and its presence (e.g. #1 vs. #4). While SiO 2 nanoparticles form in the absence of APTES, its presence leads to the formation of spherical microparticles. We have currently no explanation for these differencestheir understanding would require a more detailed micromorphological study. The factors that might play a role are (i) the decreased apolarity of APTES compared to TEOS (and especially Si(OH) 4 ), (ii) different hydrolysis and condensation rates of APTES vs. TEOS, and (iii) different cocondensation rates of APTES with PI6 vs. APTES with TEOS. In a last set of experiments (#7-9, Table 1), we investigated the co-condensation of PMA, HMDA and APTES in the absence of TEOS. ATR-FTIR spectra show both characteristic imide and siloxane modes, and the absence of modes related to potential end-groups or unreacted monomeric species (see the ESI †). Note that as seen for the previous set of experiments (#4-6), the intensity of Si-O-Si modes increases with growing amount of APTES, which is expected. As for experiments in the presence of APTES and TEOS, we do not nd any indication for amide (1,2diamide or 1-amide-2-carboxylic acid) linkages. PXRD patterns show the same trends as observed in samples #4-6: at the lowest n APTES (0.6 mmol), we still nd the major reections A-D of PI6, in coexistence with the new reections K and L, and the previously observed two amorphous halos (see the ESI †). Furthermore, the morphologies observed from SEM images show features as noted for the previous series of hybrids (#4-6): (i) in all three samples #7-9, we nd the coexistence of interwoven, bundled bers with spherical particles of approx. 5 mm in diameter; and (ii) the number of spheres relative to bers increases with increasing n APTES (see the ESI †). EDAX shows that the spherical morphologies are by far richer in Si than in the bers. From the characterization of hybrids in the synthesis of which APTES was used (#4-9), we can conclude that APTES allows for covalently linking SiO 2 and PI6, as underlined by the presence of imide and Si-O-Si modes, and the absence of monomer/end-group-related modes in ATR-FTIR spectra. Moreover, employing APTES has important consequences for crystallinity and morphology, while an impact of TEOS on both features is not detected.
2.3 Upscaling of the hydrothermal PI6/SiO 2 synthesis and solvent-free processing by sintering To this point, we have shown that PI6 and SiO 2 can be synthesized simultaneously in nothing but hot water from the comonomers HMDA, PMA and the silica source TEOS. Moreover we showed that the well-known compatibilizer APTES also efficiently links PIs and SiO 2 under hydrothermal synthesis conditions. However, both crystallinity and morphology of PI6 and SiO 2 are strongly inuenced when APTES is employed, which is supposed to be additionally supported by using nonstirred autoclaves where the absence of mechanically mixing the reaction mixtures reinforces segregation phenomena.
In a last set of experiments, we investigated the upscaling of the hydrothermal PI6/SiO 2 synthesis, as well as the processing of the obtained powders into macroscopic objects, specically into pellets, by solvent-free sintering. Scaled-up syntheses were performed in a 1 L stirred batch autoclave (see the ESI † for the experimental set-up). The reactor was lled with a dispersion containing HMDA, PMA, TEOS and APTES in 400 mL H 2 O. The used concentrations and reactant ratios were the same as in experiment #5 (cf. Table 1), with resulting employed molar amounts of n PMA ¼ 40 mmol, n HMDA ¼ 24 mmol, n APTES ¼ 32 mmol, and n TEOS ¼ 32 mmol. Therefore, the PI6/SiO 2 synthesis was effectively scaled up by a factor of 26.
6. For comparison, we also scaled up the HTP of pure PI6 (see the ESI †for experimental details). As for the small scale experiments, hydrothermal syntheses were run for 24 h at 200 C and the autogenous pH ¼ 4. As for the smaller scale experiments, pure PI6 and PI6/SiO 2 were obtained as white powders. ATR-FTIR spectra show the presence of imide modes and lack modes related to unreacted monomers/end-groups for upscaled PI6an overlay of the spectra of PI6 from small scale synthesis and upscaled PI6 shows virtually no differences in spectra (see the ESI †). For upscaled PI6/SiO 2 , we also nd both imide and Si-O-Si modes, and the absence of unreacted end-groups/ monomers in the ATR-FTIR spectrum. In fact, the spectrum perfectly coincides with the one of the corresponding small scale synthesis (ESI †). PXRD of upscaled PI6 generates exactly the same diffractogram as in the small scale: all major and minor reections of PI6 are present (see Fig. 7A). For the upscaled hybrid material there are slight differences with respect to the small scale PI6/SiO 2 . While only three of the four major PI6 reections persist (B, C, D), the new reections K, L, M are all present (see Fig. 7B). Moreover, the amorphous underground is now composed of only one halo centered around 19 (2q, Cu-K a ). In addition, the reections are less intense and sharper than for the small scale synthesis. This underlines the effect of stirring, leading to less segregation into PI6 and SiO 2 -rich domains, respectively, and therefore less crystallinity. SEM analysis reveals that the upscaled pure PI6 is still composed of ribbon structures (see Fig. 7E). However, they are not covered with smaller ribbons (which was the case for small scale PI6, cf. Fig. 3), and seem less globally oriented (see Fig. 7F). The upscaled PI6/SiO 2 shows two coexisting morphologies: bers and spherical particles of approx. 1 mm in diameter. Compared to the small scale PI6/SiO 2 , the upscaled sample is more homogeneous and the observed spherical particles are smaller and less disperse in size. This again underlines the effect of stirring. EDAX of the upscaled PI6/SiO 2 evinces that the spherical objects are richer in Si than the bers (see the ESI †).
From the upscaled syntheses, we obtained ca. 10 g of PI6 and PI6/SiO 2 respectively. 5 g of each was used to prepare pellets by sintering in a hot press, aka compression molding, a technique well suited for the compaction of polymer-based particulate composites. 29 Sintering of polyimides is an established technique, which is conventionally used to process semicrystalline polyimides such as Kapton™. 30,31 Moreover, sintering techniques have been employed for preparing PI composites with e.g.
graphite akes or microsilica as dispersed reinforcement. 32,33 Before sintering, the samples were rst nely ground using a cryo-ball mill where powder freezing was performed using liquid N 2 . Compression molding was subsequently performed for 2 h at 300 C applying a pressure of 80 MPa, yielding cylindrical specimens with a diameter of 40 mm and a height of around 2.5 mm (see the ESI † for information on the experimental set-up). Photographs of the pure PI6 pellet and the PI6/SiO 2 pellet are shown in Fig. 7C and D. Both pellets were shiny and reective, which is a rst indication that (i) the samples had crept during pressing, and (ii) that successful compaction had taken place. Note that the color of both (however more pronounced for the PI6 pellet) changed from white to off-white/grayish. While one might suspect thermooxidative phenomena to be responsible for the darkening, ATR-FTIR analysis of the pellets shows spectra that are unchanged as compared to pristine upscaled PI6 and PI6/SiO 2 , respectively (see the ESI †). Hence, we conclude that the sintering procedure did not lead to major chemical changes. Instead, we speculate that the compaction could have generated weak charge-transfer complexes, e.g. between the electron poor aromatic ring in PI6 and a C]O function of a neighboring chain, by bringing the polymer chains closer together. PXRD analysis of the pellets shows that the average order in both materials is retained. The diffractogram of the PI6 pellet shows all major and minor PI6 reections. However, the reections are less sharp than before sintering. The diffractogram of the PI6/SiO 2 pellet has changed more signicantly. The overall shape of the diffraction curve is retained as compared to PI6/SiO 2 before sintering. However, the reections K, L, M and B, C, D are barely visible anymore. SEM analysis of the PI6 pellets ( Fig. 7G and the ESI †) shows major morphological changes. The pellet is composed of highly parallel bers having near-circular cross-sections of 100-200 nm in diameter. For the compactness of the sample, it is difficult to comment on the length of the bers. Given these important morphological changes, which further corroborate that the sample crept efficiently, it is even more impressive that the crystallinity of PI6 is retained. The PI6/SiO 2 pellet shows less pronounced morphological changes than PI6. While the sample is much denser than before sintering and shows similar brous structures as the sintered PI6 pellet, the spherical SiO 2 -rich particles have not merged with the brous structures and with each other (Fig. 7H). We believe that this result is a consequence of the sintering temperatures being too low for effective consolidation of the SiO 2 -containing compound. Indeed, SiO 2 sintering is typically carried out at temperatures around 1000 C. 34 Finally, we additionally investigated the chemical nature of upscaled PI6 and PI6/SiO 2 hybrids before and aer sintering via solid-state magic angle spinning nuclear magnetic resonance spectroscopy (MAS NMR). Specically, we performed 13 C, 15 N and 29 Si MAS NMR of all four samples (see the ESI † for measurement details). It becomes clear from comparing the spectra that there is virtually no chemical difference between PI6 and PI6/SiO 2 before and aer sintering (see the ESI † for all MAS NMR spectra). Representative 13 C, 15 N and 29 Si MAS NMR spectra of PI6 and PI6/SiO 2 pellets (aer sintering) are depicted in Fig. 8. The 13 C MAS NMR spectrum of PI6 (Fig. 8A) allows for clear assignment of all carbon atoms. The carbonyl C (labelled A) of the imide ring is the most intense signal found at 160.19 ppm, the aromatic C-atom signals are found at 130.16 ppm (B) and 112.76 ppm (C), the aliphatic Catoms in a to the imide nitrogen are found at 32.36 ppm (D), and the C-atoms in b and g give rise to a signal at 20.80 ppm. These shis are in agreement with the literature. 35 Note that the signals attributed to A, B and C give rise to symmetrical spinning sidebands (labelled A*, B* and C*), which is a consequence of the crystallinity of PI6. The 13 C MAS NMR spectrum of PI6/SiO 2 (Fig. 8B) from sintering shows all the same peaks as pure PI6. Moreover, we nd one additional signal in the aliphatic region assigned to the C-atom in a to Si (F), hence arising from covalent binding of APTES to a polyimide segment. Note that the 13 C MAS NMR spectrum of PI6/SiO 2 also shows spinning side bands. This underlines our previous conclusions drawn from PXRD on PI6/ SiO 2 still showing higher ordering/crystallinity than a fully  Si MAS NMR spectrum of PI6/SiO 2 (Fig. 8C) shows the presence of both Q 4 (Si in bulk silica, i.e. Si connected to four OSi groups) and Q 3 /Q 2 (Si connected to three/two OSi groups and one/two OH groups), which arise from the condensation of TEOS. Moreover, we nd a strong T band (Si connected to three SiOH/SiOR groups and to one alkyl moiety), which arises from the condensation of APTES. Most interestingly, the observed T band is accompanied by two symmetrical spinning side bands, which can only arise in a (at least to some extent) crystalline sample. Therefrom, we can conclude that Si is covalently linked to crystalline/ordered PI6 chains. Note that a rst conjecture that the peak at ca. 0 ppm might be due to a D band (two alkyl moieties), which could arise from APTES containing some extent of di-(3-aminopropyl)diethoxysilane, was disproved by 29 Si NMR of the used APTES (see the ESI †). The chemical shis of T and Q signals are in agreement with the literature. 36 29 Si MAS NMR clearly shows that OH groups are still present in PI6/SiO 2 . These might arise from incomplete condensation of hydrolyzed APTES or TEOS, but could also fully arise from surfacial Si-OH. 36 The observed T signal cannot be so clearly assigned as to conclude if nonhydrolyzed OEt groups are still present or not. However, we would expect to nd corresponding signals in 13 C MAS NMR in this case. Lastly, neither of the MAS NMR experiments of the sintered materials gives an indication of the side-products of potential thermo-oxidation.

Conclusion
In this contribution, we reported the (i) rst example of the hydrothermal synthesis of a semialiphatic polyimide, PI6; (ii) the rst simultaneous hydrothermal synthesis of an inorganicorganic hybrid material, PI6/SiO 2 , including its upscaling and (iii) the solvent-free processing of both PI6 and PI6/SiO 2 via hot pressing. Our investigations show that PI6 and SiO 2 are obtained as a mixture when synthesized from the comonomers HMDA and PMA, and TEOS. As the resulting product is a mere mixture of the polyimide and silica, neither the crystallinity nor morphology of either of the components is inuenced. Moreover, the organically modied silicon alkoxide APTES can be employed hydrothermally as a compatibilizer. The addition of APTES efficiently allows for generating class II hybrid materials, but hampers the crystallinity of the PI6 phase and has major morphological implications. The more APTES is added, the more PI6 phases change to brous morphologies, and SiO 2 phases change from nanoparticles to colloidal particles of 1-5 mm. Both the hydrothermal PI6 and PI6/SiO 2 synthesis were upscaled from an initial reaction volume of 15 mL to 400 mL using a stirred 1 L batch reactor. Stirring has a negative effect on the morphology of pure PI6, which shows more features indicating crystallinity from non-stirred syntheses, but has a positive effect on the homogeneity of PI6/SiO 2 . PI6 and PI6/SiO 2 from upscaling were used for processing experiments. These rst results of PI6 and PI6/SiO 2 sintering show that (i) PI6 can be nicely compacted with retention of the initial crystallinity, and (ii) that PI6/SiO 2 hot pressing results in densication but lacks sufficient creep of the SiO 2 particles. This is attributed to the fact that the employed sintering temperature of 300 C is most likely too low for silica. Beyond this rst proof of concept, further optimization is necessary and will be performed in the near future. All hybrid materials were fully characterized using ATR-FTIR, SEM, EDAX, PXRD and solid-state MAS NMR. The results of these analyses strongly support the drawn conclusions.
Both the synthesis of PI6 and PI6/SiO 2 can be qualied as green, as solely water is used aside the starting compounds. Moreover, conventional hydrothermal SiO 2 syntheses demand for the use of adding mineral acids or bases as condensation promotors. In the reported one-pot route, additional promotors are not necessary, since the comonomer PMA provides sufficient acidity. Processing by sintering also qualies as green, since no solvents are used at all. As we show in this contribution with the example of PI6/SiO 2 , hydrothermal synthesis is a potent means of synthesizing inorganic-organic hybrid materials in one pot.