Structural analysis of cellulose acetate and zirconium alkoxide hybrid fibres

Abstract

We investigated the detailed structures of organic–inorganic hybrid fibres composed of cellulose acetate (CA) and zirconium alkoxides (Zr(OR)₄) using attenuated total refraction-Fourier transform infrared spectroscopy (ATR-FTIR), energy-dispersive X-ray spectroscopy (EDS), and X-ray absorption fine structure (XAFS) measurements. The fibres were prepared by an air-gap spinning technique, where the acetone solution of CA was injected into a Zr(OR)₄–acetone bath. The Zr contents in the prepared fibres increased with increasing Zr(OR)₄ bath concentration, but reached steady-state values at Zr(OR)₄ bath concentrations above 10 wt%. In addition, EDS analysis for the cross-section of the fibre showed that Zr distribution in the fibre varied depending on Zr(OR)₄ bath concentration and alkoxide type. ATR-FTIR and XAFS analysis showed that Zr contained in a CA fibre was hexacoordinated, with a local structure similar to that of hydrolysed Zr(OR)₄. This result indicated that the confinement to the CA fibre had little influence on the local structure of Zr.

---

1. Introduction

Cellulose acetate (CA)–Zr hybrid materials have attracted attention for their applications as absorbents for chemical separations or supports for enzyme immobilizations.^1–9 CA can be easily molded into different forms, such as membranes and fibres.¹⁰ Zirconia is known to have great chemical resistance, thermal stability, permeability and biocompatibility compared to silica, alumina and titania,^4,9 and it has phosphate ion absorption ability.⁸ CA can form composite materials with zirconia, via coordination of –OH or –CO groups to zirconium species.^4,8 For example, the report of Rodrigues-Filho et al. shows that CA–Zr composite membranes can act as absorbents for phosphate chemical species, which cause serious algae proliferation and environmental problems.⁸ They prepared the membrane by immersing CA cast film into a zirconium alkoxide (Zr(OR)₄) solution. In contrast, Nakane et al. studied fibrous CA–Zr composites for enzyme immobilization, preparing them by injecting CA solution into a Zr(OR)₄ bath using an air-gap spinning technique.^3–6 CA chains are instantaneously cross-linked by Zr(OR)₄, from which fibrous gels can be obtained. By using Zr(OR)₄, the gel becomes tough enough to resist high pressure in an enzyme reaction column.⁴ In addition, as fibrous materials have larger surface areas than film materials, fibrous CA–Zr composites should be more efficient phosphate ion absorbents and superior supports for enzyme immobilization. In addition, CA–Zr composites can be converted into CA–zirconium phosphate composites,¹⁰ which have been studied as ion-exchange materials.^2,11

However, in spite of the many application studies, it is still unclear how preparation conditions influence the structure of CA–Zr gel fibres. In this study, we systematically investigated the effect of alkyl chain length and Zr(OR)₄ bath concentration on the structure of resultant CA–Zr fibres, using thermogravimetric analysis (TGA), X-ray diffraction (XRD), attenuated total refraction Fourier transform infrared spectroscopy (ATR-FTIR), swelling measurements, energy dispersive X-ray spectroscopy (EDS), and X-ray absorption fine structure (XAFS) measurements. We chose zirconium tetra-n-propoxide (Zr(OPr)₄) and zirconium tetra-n-butoxide (Zr(OBu)₄), which were used in the previous reports.^4,8 Zirconium tetra-ethoxide could not be used, because it does not dissolve in acetone used as a solvent in this investigation. Zr XAFS is a powerful tool for the analysis of local structure around a target element, allowing us to obtain information on coordination number and bond distance. In addition, we compared the structures of hydrolysed Zr(OR)₄ with and without CA fibre, and studied the effect of confinement to the CA fibre on the structure of Zr. This structural knowledge provided useful information for the preparation of CA–Zr composite materials.

2. Experimental

2.1 Materials

CA and Zr(OBu)₄ (85% solution in 1-butanol) was purchased from Wako Pure Chemical Industries, Ltd. and used as received. The rate of acetylation for CA was 39.8%. Zr(OPr)₄ (70% solution in propanol) and ZrO₂ powder were purchased from Sigma-Aldrich Co. LLC. Acetone used in this study was dehydrated using 13 Å molecular sieves.

2.2 Preparation of the CA–Zr(OR)₄ fibres and comparison samples

The sample preparation method, a so-called air-gap spinning technique, is schematically represented in Fig. 1. Firstly, a 15 wt% CA–acetone solution was prepared using dehydrated acetone. In another bottle, each Zr(OR)₄ compound was dissolved in 15.2 g dehydrated acetone. The CA solution was then set in a syringe with a 21G needle, and ejected by syringe pump into the stirred Zr(OR)₄ solution bath, instantaneously forming a fibrous material. After the determined immersion time, the obtained fibres were repeatedly washed with acetone, followed by distilled water, in order to remove unreacted components. All samples were dried under vacuum and used for measurements. The Zr(OR)₄ bath concentrations were 1, 2.5, 5, 10, 15, 20, and 25 wt%. The immersion times were 1, 5, 10, 30, and 60 min. The nozzle-to-bath distance, spinning solution volume, and spinning rate were 10 cm, 2.5 mL, and 3 mL min⁻¹, respectively.

Fig. 1 Schematic representation of CA–Zr fibre preparation methods.

Two types of Zr(OR)₄ with different alkyl lengths, Zr(OPr)₄ and Zr(OBu)₄, were used in this study. For comparison, we also prepared a pure CA fibre sample (CA–Zr-0), a hydrolysed Zr(OBu)₄ sample (Bu–H), and a hydrolysed Zr(OPr)₄ sample (Pr–H). CA–Zr-0 was prepared by ejecting 15 wt% CA solution into distilled water. Bu–H and Pr–H was prepared by drying each alkoxide solution in a glass dish.

2.3 Characterization

TGA measurements were performed using a thermogravimetric analyser (DTG-60, Shimadzu, Japan) from 30 to 600 °C at a heating rate of 10 °C min⁻¹, in an air atmosphere. XRD measurements were carried out using an X-ray diffractometer (Ultima-IV, Rigaku, Japan), equipped with a sample holder for fibrous specimens. Cu Kα radiation (wavelength, 1.54 Å) was used, and the data were collected in the 2θ range from 5° to 40° at room temperature. ATR-FTIR measurements were carried out at room temperature using an IR spectrometer (IR Affinity-1, Shimadzu, Japan) equipped with a single reflection ATR accessory (MIRacle 10, Shimadzu, Japan) containing a diamond/ZnSe crystal.

2.4 Swelling measurements

Swelling experiments were carried out by immersing the dried fibre sample into acetone for 8 h at room temperature. Each experiment was performed three times, and the results were averaged. The swelling ratio was defined as the ratio of weight after swelling to weight before swelling.

2.5 EDS measurements

EDS was performed using FE-SEM (ULTRA plus, ZEISS Co., Ltd., Germany) combined with an EDS detector (XFlash detector 5030, BRUKER AXS Co., Ltd., USA). The cross-sections of the fibre samples were observed as follows:¹² the dried fibre was broken into a short piece, then dipped into a 1 wt% ionic liquid ethanol solution, in order to add conductivity. The ionic liquid used was 1-ethyl-3-methylimidazolium tetrafluoroborate ([EMI⁺][BF₄⁻]), which was homogeneously dispersed in dehydrated ethanol using ultrasound. The dipped fibre specimen was dried under vacuum for about 30 min. The dried fibre specimen was then set on a sample holder using conductive carbon tape in order to see the fibre cross-section. Before measuring EDS, further drying was performed for 30 min. The acceleration voltage was 5.00 kV.

2.6 XAFS measurements

The prepared fibre samples were pulverized with an agate mortar, and put through a 400-mesh sieve, in order to obtain a homogeneous fine powder. These powdered fibre samples were then compressed into 10 mmϕ pellets together with boron nitride, which acted as a binder. Preliminary proper sample weights were calculated using the residual content values obtained from TGA analysis (Fig. 2(c)) and SAMPLEM4M software. Zr K-edge XAFS measurements were carried out in transmission mode using synchrotron radiation through a Si(111) monochromator at BL-12C in Photon Factory, Institute of Materials Structure Science, High Energy Accelerator Research Organization (KEK), Japan. XAFS spectra were analysed by Athena and Artemis software.¹³ Fourier transformation of k³-weighted EXAFS oscillation was performed in the range of the wave vector k = 1.8–14 Å⁻¹.

Fig. 2 Typical TGA curves of samples prepared at various (a) Zr(OBu)₄ and (b) Zr(OPr)₄ bath concentrations. Dependence of residual content on (c) Zr(OR)₄ concentration and alkoxide type (immersion time fixed at 30 min) and (d) immersion time.

3. Results and discussion

3.1 TGA

Fig. 2(a) shows typical TGA curves of the samples prepared using various Zr(OBu)₄ concentrations. The immersion time was fixed at 30 min. Corresponding TGA curves for Zr(OPr)₄ are shown in Fig. 2(b). All samples thermally decomposed in three stages, consistent with a previous report by Hanna et al.:¹⁴ the first decomposition (ca. 140 °C) was attributed to the decomposition of bound water. The second and third decompositions took place at around 280 and 400 °C, respectively. Considering that the amount of decomposition at the second stage decreased with increasing Zr(OR)₄ bath concentrations and residual content, the second stage must correspond with the degradation of CA. The third stage can be attributed to carbonization of the products to ash. The amounts of decomposition at the second and third stages are summarized in Fig. S1 (see ESI†). The degradation temperatures of all Zr samples were similar, but were significantly lower than those of CA–Zr-0. The same phenomenon was observed for other CA–metal oxide systems.^15,16 These results indicated that interactions between the CA molecular chains were weakened by the insertion of Zr(OR)₄ molecules.

In addition, residual contents were similar irrespective of the alkoxide type, indicating that alkyl chain length in Zr(OR)₄ did not affect Zr contents in the fibres (Fig. 2(c)). This result was surprising, because it was expected to be more difficult for larger Zr(OR)₄ compounds to diffuse into the CA chains. Residual content increased with increasing Zr(OR)₄ bath concentration, but with an upper limit (less than 30 wt%).

Conversely, immersion time had different effects on the two types of alkoxide. Fig. 2(d) shows the variation in residual contents obtained from TGA measurements as a function of immersion time. Comparing the two types of alkoxide, a larger initial increase in residual content was observed for samples prepared with Zr(OPr)₄ than with Zr(OBu)₄. This result was attributed to Zr(OPr)₄ being smaller than Zr(OBu)₄, and, thus, being able to diffuse into the fibre more rapidly. In addition, for all samples, the residual content increased steeply immediately after the beginning of the reaction, and took about 30 min to reach equilibrium, suggesting that Zr(OR)₄ reacted rapidly at the surface of the fibre and slowly diffused to the inside of the fibre. This result was the same as that of sodium alginate–CaCl₂ gel beads prepared by a similar method,¹⁶ and was consistent with EDS observations (see Section 3.4). Next, Zr(OPr)₄ baths of 15 wt% and 2.5 wt% were compared. In the 2.5 wt% bath, the residual content, i.e., Zr species contained in the fibre, was lower than in the 15 wt% bath, even after equilibrium was established. This could be because most Zr(OPr)₄ in the 2.5 wt% bath was consumed, causing the reaction to stop before its upper limit.

3.2 XRD

Fig. 3(a) shows the XRD results of the CA–Zr(OBu)₄ series. XRD results for Zr(OPr)₄ are shown in Fig. 3(b). Although CA–Zr-0 contained a large amorphous region, two peaks were observed at around 8.7° and 17.2°, which should correspond to (110) and (200), respectively, as previously reported.¹⁷ However, for samples containing Zr, the (110) peak disappeared and the (200) peak became smaller with increasing Zr content, irrespective of the type of alkoxide. These results indicated that Zr(OR)₄ intercalated between the CA chains, and disturbed the crystallization of CA.

Fig. 3 XRD profiles for samples prepared in (a) Zr(OBu)₄ and (b) Zr(OPr)₄ baths of various concentrations.

3.3 FT-IR

Fig. 4(a) shows the results of FT-IR measurements for samples prepared in Zr(OBu)₄ solutions. The corresponding results for Zr(OPr)₄ are shown in Fig. 4(b). At first, we explain the results of the fibre samples (i.e. except Bu–H and Pr–H). A peak appeared at ∼660 cm⁻¹ for all of fibre samples, except CA–Zr-0. This peak corresponded to the superposition of metal–oxygen stretching vibrations, indicating the formation of Zr–O bonds.^2,7 The broad peak at 1560 cm⁻¹ indicated interactions between C=O and Zr,¹⁰ suggesting the interaction of Zr with hydroxyl groups or carbonyl groups of CA. The sharp peaks at 1228 and 1044 cm⁻¹ were attributed to C–O single bond stretching mode;¹⁸ these peaks became smaller with increasing Zr(OBu)₄ concentrations. Furthermore, the sharp peak at 1734 cm⁻¹, which was attributed to the ester carbonyl in CA,² became smaller with increasing Zr(OBu)₄ concentration. These results indicated that the vibrations of CA molecules were suppressed crosslinks formed by Zr(OBu)₄.

Fig. 4 Variation in ATR FT-IR spectra as a function of (a) Zr(OBu)₄ and (b) Zr(OPr)₄ bath concentration.

In the case of Bu–H, a peak also appeared at 660 cm⁻¹, corresponding to a Zr–O bond. A peak at 1560 cm⁻¹ would usually be assigned to interactions between C=O and Zr, as mentioned above. However, considering that Bu–H should not contain C=O, and that there were no peaks at 1770 cm⁻¹ for C=O bonds,¹⁹ the peak at 1560 cm⁻¹ peak was attributed to Zr–O–C bonds.²⁰ In addition, there was a broad peak at around 3300 cm⁻¹, corresponding to an O–H bond. These results suggested that Bu–H contained a Zr–OH structure. The results for Pr–H were similar to those for Bu–H. We were unable to observe a peak at 460 cm⁻¹, assigned to Zr–O–Zr, due to noise in the data. However, considering that hydrolysed Zr alkoxides generally have a Zr–O–Zr structure, as well as Zr–OH,²¹ we surmised that Bu–H and Pr–H would also both have Zr–O–Zr and Zr–OH structures. This data is further discussed in Section 3.6.

3.4 EDS measurement

Next, we performed EDS measurements for the cross-section of the fibre in order to clarify the distribution of Zr inside the fibre. Fig. 5(b) shows C and Zr mapping images for fibre samples prepared at various Zr(OPr)₄ bath concentrations. For fibres prepared at low Zr(OPr)₄ bath concentrations (1 wt%), Zr existed only at the fibre surface. However, those prepared at higher concentrations had homogeneously distributed Zr. These results can also be seen from Fig. 6, which shows the distribution of C and Zr along arrows in the figure. For the fibre prepared in the 1 wt% bath (Fig. 6(a)), the C profile was nearly constant, while the Zr profile had a concave upward curve, indicating Zr locally distributed at the surface of the fibre, with little Zr at the centre of the fibre. However, when prepared in the 10 wt% bath (Fig. 6(b)), both C and Zr profiles were nearly constant, indicating the homogeneous distribution of Zr. The same results were also observed by point analyses (Fig. S2, see ESI†), where more detailed elemental information was obtained. These results suggested that the reaction proceeded from the outside of the fibre. In the case of 1 wt%, the Zr(OPr)₄ concentration might have been too low to penetrate inside the fibre, resulting in local Zr distribution at the fibre surface only. Fig. 5(a) also shows C and Zr mapping of the samples prepared in Zr(OBu)₄. Although Zr contents were similar irrespective of the alkoxide used, we did not observe local Zr distribution in the 1 wt% Zr(OBu)₄ sample. This difference might be attributed to the influence of Zr(OR)₄ reactivity and diffusion speed; considering that resistance to hydrolysis increases for longer alkyl chains,²² the reactivity of Zr(OBu)₄ might be lower than that of Zr(OPr)₄. Therefore, Zr(OPr)₄ would react quickly on the outside of the fibre, and slowly diffuse into the fibre, whereas Zr(OBu)₄ could enter the centre of the fibre at the beginning of the reaction, and then slowly react with CA.

Fig. 5 EDS elemental maps of fibre sample cross-sections prepared in (a) Zr(OBu)₄ and (b) Zr(OPr)₄ baths of various concentrations. The upper and lower figures indicate C-mapping and Zr-mapping, respectively. Samples were processed in ionic liquid solution and fixed on conductive carbon tape.

Fig. 6 C and Zr distribution along arrows across the cross-section of the fibre shown in the bottom figures. Fibre samples were prepared in (a) 1 wt% and (b) 10 wt% Zr(OPr)₄ baths.

3.5 Swelling measurements

Fig. 7 shows the results of swelling measurements as a function of Zr(OR)₄ bath concentration. The swelling ratio decreased with increasing Zr(OR)₄ bath concentration, suggesting that incorporated Zr compounds acted as cross-linking points. No significant difference was observed between the two alkoxides used, indicating that the formed network structures were similar irrespective of the type of alkoxide.

Fig. 7 Variation in the swelling ratio for fibre samples as a function of Zr(OR)₄ bath concentration. Swelling ratios were evaluated from the ratio of weight after swelling in acetone to weight in the dry state.

3.6 XAFS measurements

Finally, we present the results of Zr-K edge XAFS measurements. Fig. 8 shows normalized XANES spectra for the Zr(OBu)₄ series, Zr(OPr)₄ series, Bu–H, Pr–H, ZrO₂, and Zr foil. No differences were observed among the Zr(OBu)₄ series, Zr(OPr)₄ series, Bu–H, and Pr–H, indicating that their local structures were similar. Fig. 9 shows the Fourier transformed EXAFS spectra for ZrO₂, Zr(OBu)₄ fibre samples, and Bu–H. The fibre samples and Bu–H produced similar curves, giving strong peaks at 1.6 Å, assigned to the Zr–O bond,²³ and a small peak at 3.1 Å, assigned to the Zr–Zr distance.²⁴ Both peak positions did not change with increasing Zr(OR)₄ concentration, indicating that the distance was not affected by Zr content, even at the upper limit. Therefore, we concluded that Zr(OR)₄ was incorporated in its hydrolysed form, where some Zr species were polymerized and the distance between neighbouring Zr atoms was not affected by the Zr content of the whole fibre. The EXAFS curves were similar to those of polymeric Zr hydroxide,^24,25 suggesting the formation of Zr(OH)₄ derivatives in the fibre sample and Bu–H, which was consistent with ATR-FTIR results (see Fig. 4). In addition, there were no large oscillations in the high length region (above 4 Å) for Zr(OBu)₄ fibre samples and Bu–H, in contrast with ZrO₂. This result indicated that Zr in the fibre sample and Bu–H did not have a long-period structure, as found in ZrO₂, which was consistent with the XRD results (see Fig. 3).

Fig. 8 Normalized XANES spectra for Zr(OBu)₄ and Zr(OPr)₄ series, Bu–H, Pr–H. As a comparison, results of Zr foil and ZrO₂ were also shown.

Fig. 9 Fourier-transformed k³-weighted Zr K-edge EXAFS spectra for CA–Zr fibre samples prepared at different Zr(OBu)₄ bath concentrations, Bu–H, and ZrO₂.

Next, we performed curve fitting analysis of the EXAFS oscillations for the first coordination sphere of Zr(OBu)₄ and Zr(OPr)₄ fibre samples. The obtained coordination numbers and Zr–O distances are summarized in Fig. 10. The coordination numbers were near 6, and were not affected by the type of alkoxide or concentration. The obtained coordination numbers for Bu–H and Pr–H were also ca. 6, in agreement with the report of Bradley et al.²⁶

Fig. 10 (a) Coordination numbers and (b) Zr–O distances evaluated from fitting analysis of EXAFS results.

Finally, we have depicted the structure of the CA–Zr fibre in Fig. 11. It should be noted that this schematic representation is different from that reported previously.⁴ Considering the XAFS and ATR-FT-IR results, the Zr contained in the CA fibre is hexacoordinate, and has the same structure as that of hydrolysed Zr alkoxide, containing Zr–OH. We surmised that the Zr alkoxides in the fibre were polymerized, but the degree of polymerization of hydrolysed Zr alkoxide was assumed to be small, in accordance with the literature.²⁶ In addition, considering that condensation reactions should be slower than hydrolysis and gelation reactions,²¹ we assumed that, after one Zr(OR)₄ molecule forms a cross-linking point with CA chains, a few other Zr(OR)₄ molecules become polymerized with that cross-linking Zr(OR)₄. The Zr alkoxide type and concentration did not influence the local structure of the cross-linking point.

Fig. 11 Schematic representation of the structure of CA–Zr fibres surmised from ATR-FTIR and XAFS results. The local structure of the cross-linking point, which consisted of several hydrolysed Zr(OR)₄ molecules, was not affected by the Zr(OR)₄ type or concentration, but the number of the cross-linking point increased with increasing Zr content.

4. Conclusions

In this study, we investigated the effects of Zr(OR)₄ bath concentration and alkoxide-type on the structure of the CA–Zr fibres obtained. The type of alkoxide did not affect the local structure of CA–Zr fibres, but influenced their gelation speed (faster for Zr(OPr)₄). In addition, there was an upper limit to Zr content in the fibre (less than 30 wt%). EDS analysis revealed that for high Zr(OPr)₄ bath concentration, Zr was homogeneously distributed inside the fibre, while for low bath concentrations, Zr existed locally at the fibre surface. These results suggested that the gelation reaction proceeded from outside of the fibre. However, in the case of Zr(OBu)₄, homogeneous Zr distribution was observed for all Zr(OBu)₄ concentrations. These results might be attributed to the differences in reactivity and diffusion speed between the two alkoxides used. Finally, ATR-FTIR and XAFS measurements revealed detailed structural information on the CA–Zr fibres, showing that the Zr species were similar to hydrolysed Zr(OR)₄, and that the local structure-forming cross-linking point was not altered by Zr content and alkoxide type.

Acknowledgements

The authors would like to thank Prof. Jae-ho Kim for his helpful advice. This work was performed under the approval of the Photon Factory Program Advisory Committee (Proposal No. 2015P009).

References

1. J. E. Kiefer and G. Touey, I&EC Prod. Res. Dev., 1965, 4, 253–256.
2. V. K. Gupta, D. Pathania, P. Singh, B. S. Rathore and P. Chauhan, Carbohydr. Polym., 2013, 95, 434–440.
3. K. Nakane, S. I. Suye, T. Ueno, Y. Ohno, T. Ishikawa, T. Ogihara and N. Ogata, J. Appl. Polym. Sci., 2010, 116, 2901–2905.
4. K. Nakane, T. Ogihara, N. Ogata and Y. Kurokawa, J. Mater. Res., 2003, 18, 672–676.
5. K. Nakane, T. Ogihara, N. Ogata and Y. Kurokawa, J. Appl. Polym. Sci., 2001, 81, 2084–2088.
6. Y. Kurokawa, Polym. Gels Networks, 1996, 4, 153–163.
7. S. A. Nabi and M. Naushad, Colloids Surf., A, 2008, 316, 217–225.
8. U. P. Rodrigues-Filho, Y. Gushikem, M. C. Goncalves, R. C. Cachichi and S. C. Castro, Chem. Mater., 1996, 8, 1375–1379.
9. G. Arthanareeswaran and P. Thanikaivelan, Sep. Purif. Technol., 2010, 74, 230–235.
10. C. A. Borgo, A. M. Lazarin and Y. Gushikem, Sens. Actuators, B, 2002, 87, 498–505.
11. C. A. Borgo and Y. Gushikem, J. Colloid Interface Sci., 2002, 246, 343–347.
12. Y. Ishigaki, Y. Nakamura, T. Takehara, T. Shimasaki, T. Tatsuno, F. Takano, Y. Ueda, Y. Motoo, T. Takegami, H. Nakagawa, S. Kuwabata, N. Nemoto, N. Tomosugi and S. Miyazawa, Microsc. Res. Tech., 2011, 74, 1024–1031.
13. B. Ravel and M. Newville, J. Synchrotron Radiat., 2005, 12, 537–541.
14. A. A. Hanna, A. H. Basta, H. El-Saied and I. F. Abadir, Polym. Degrad. Stab., 1999, 63, 293–296.
15. T. Uesaka, N. Ogata, K. Nakane, K. Shimizu and T. Ogihara, J. Appl. Polym. Sci., 2002, 83, 1750–1758.
16. T. Yotsuyanagi, T. Ohkubo, T. Ohhashi and K. Ikeda, Chem. Pharm. Bull., 1987, 35, 1555–1563.
17. M. Gopiraman, K. Fujimori, K. Zeeshan, B. S. Kim and I. S. Kim, eXPRESS Polym. Lett., 2013, 7, 554–563.
18. D. Murphy and M. N. de Pinho, J. Membr. Sci., 1995, 106, 245–257.
19. H. Hayashi, H. Suzuki and S. Kaneko, J. Sol-Gel Sci. Technol., 1998, 12, 87–94.
20. S. Sakka, Handbook of sol–gel science and technology. Processing characterization and applications. Characterization and properties of sol–gel materials and products, Springer, New York, 2005, vol. 2.
21. M. J. Percy, J. R. Bartlett, J. L. Woolfrey, L. Spiccia and B. O. West, J. Mater. Chem., 1999, 9, 499–505.
22. D. C. Bradley, R. C. Mehrotra and D. P. Gaur, Metal Alkoxides, Academic Press, London, 1978.
23. T. Tobase, A. Yoshiasa, L. Wang, H. Hong, H. Isobe and R. Miyawaki, J. Mineral. Petrol. Sci., 2015, 110, 1–7.
24. C. Walther, J. Rothe, M. Fuss, S. Buchner, S. Koltsov and T. Bergmann, Anal. Bioanal. Chem., 2007, 388, 409–431.
25. H. R. Cho, C. Walther, J. Rothe, V. Neck, M. Denecke, K. Dardenne and T. Fanghanel, Anal. Bioanal. Chem., 2005, 383, 28–40.
26. D. C. Bradley and D. G. Carter, Can. J. Chem., 1961, 39, 1434–1443.

---

Footnote

† Electronic supplementary information (ESI) available: Additional TGA and EDS data. See DOI: 10.1039/c6ra08974g

---