Topotactic reduction of oxide nanomaterials: unique structure and electronic properties of reduced TiO₂ nanoparticles

Abstract

Reduced titanium oxide nanomaterials are of great scientific and industrial interest. Herein we analyze corundum-like titanium oxide nanomaterials synthesized by topotactic reduction of TiO₂ nanoparticles. The structure is probed using X-ray pair distribution functions, and the electronic properties are examined using hard X-ray photoelectron spectroscopy and electron energy loss spectroscopy. Topotactically reduced nanoparticles have a composition, TiO_1.83, that is oxygen rich compared with that of (100)-oriented single crystals reduced under the same conditions (TiO_1.48); the latter, however, have unique electronic properties compared with typical Ti₂O₃. The structure of TiO_1.83 was found to contain both the unique Ti₂O₃ and Ti₄O₇, and the conductivity was lower than the single crystal-derived materials due to interparticle resistance. The findings confirm that topotactic reactions can create nanomaterials having unique structures and properties that cannot be realized in bulk materials.

---

Introduction

The development of advanced nanomaterials is necessary for achieving breakthroughs in batteries, fuel cells, photovoltaics and other areas. Post-synthetic treatments of nanoparticles offer a useful strategy for enhancing materials' properties, as long as the nanostructures are retained during the processing. The resulting materials, however, may be inhomogeneous because the reactions often proceed under kinetically controlled conditions. Since very little is known about the impact of post-synthetic treatments on microstructure, we address this issue herein by examining the low temperature reduction of TiO₂ nanomaterials as a case study.

Reduced titanium oxides (e.g., TiO and Ti₂O₃) are useful in photovoltaics,^1,2 fuel cells,^3–5 memory media,⁶ cosmetics etc. on account of their electronic conductivity and visible light absorption. In addition, titanium dioxide has economic advantages, is environmental friendly and chemical stable. Mild reduction treatments of bulk rutile TiO₂ results in a bluish, n-type semiconductor, which has a gap state localised at ∼2 eV above the valence band maximum.^3,7–9 This is formed by Ti³⁺ species associated with oxygen deficiency, interstitial Ti species, or OH formation.⁷ Under stronger reducing conditions, rutile TiO₂ forms the homologous series, Ti_nO_2n−1 (e.g., Ti₅O₉ and Ti₄O₇), which contains ordered oxygen deficient sites as crystallographic shear planes. This reduction proceeds topotactically.‡

In terms of materials' properties, the topotactically synthesized, single crystalline corundum Ti₂O₃ phase exhibits an excellent electronic conductivity (resistance < 6 μΩ m), even at extremely low temperatures.¹⁰ However, when the same treatment is applied to nanoparticles of rutile TiO₂, they do not show such good conductivity, even though the product appears to have the same crystal structure (XRD analyses, Fig. S12†). Moreover, the apparent conductivity is not reproducible because it strongly depends on the conditions of the pellets, as is often reported for reduced titanium oxide particles.⁴ Such discrepancies in physical properties between single crystalline and powder samples are widely recognized, but the underlying cause is not understood. We use X-ray pair distribution functions (PDFs),^11,12 which take into account the diffuse scattering as well as Bragg peaks, in order to probe both the crystalline and amorphous phases in topotactically reduced TiO₂ nanoparticles. We also use hard X-ray photoelectron spectroscopy (HX-PES) with synchrotron radiation and electron energy loss spectroscopy (EELS) to study the electronic structure.

Results and discussion

Single crystal and nanoparticles samples of rutile TiO₂ were reduced at 350 °C, as described in the Methods section. The rice-shaped morphology of the nanoparticles was retained during the reduction treatment, confirmed from the scanning electron microscope images shown in Fig. 1a. The resulting products, referred to as TiO_x-SC and TiO_x-NP, respectively, were studied by X-ray total scattering and the results were compared with both the nanocrystalline TiO₂ starting material and a commercial sample of bulk Ti₂O₃. TiO_x-SC has a 200 nm thick reduced layer on the surface of 0.5 mm thick rutile crystal with a clear boundary between them,¹⁰ and the Ti–O ratio of the surface layer was determined to be 0.675 by HX-PES (that of the commercially available Ti₂O₃ particles is 0.674), confirming the formation of stoichiometric Ti₂O₃.¹⁰ On the other hand, that of TiO_x-NP was found to be 0.547 by both thermogravimetric analysis and HX-PES (this is equivalent to TiO_1.83). This result suggests that TiO_x-NP has a substantial (18%) deficiency of Ti in the corundum structure, unless other phases are present.

Fig. 1 Atomic structure analysis using pair distribution functions (PDFs). Rutile TiO₂ nanoparticles, their reduced product (TiO_x) and corundum Ti₂O₃ particles are compared. (a) X-ray total scattering patterns. The blue region represents diffuse scattering intensity from reduced nanoparticles. The inset scanning electron microscope images show that the morphology was retained during the reduction treatment (scale bar: 30 nm). The inset structural image illustrates the lattice strains determined by the analyses. (b) Rutile TiO₂ structure. During the reduction, Ti atoms move into vacant sites and form face-sharing TiO₆ octahedra (green site). (c) Corundum Ti₂O₃ structure. (d) X-ray PDFs. The experimental plots (blue dots) up to 30 Å are compared with calculated plots (red lines). The PDF of the TiO_x sample was fitted with a mixture of the corundum structure and Ti₄O₇. The green dotted lines show the representative Ti–Ti distance of corundum Ti₂O₃. The numbers show Ti atoms defined in panel ‘c’. (e) Schematic energy bands for Ti₂O₃. e_g bands and a nonbonding t_2g band, which splits into Ti–Ti bonding a_1g bands and a nonbonding e^π_g band.

The X-ray total scattering patterns (Fig. 1a) show that TiO_x-NP exhibits broad diffuse scattering intensity underneath Bragg peaks that are assignable to the corundum structure; this indicates the presence of an amorphous and/or nanocrystalline phase as well as the crystalline corundum phase. Since the PDF of TiO_x-NP does not decay beyond the nearest neighbour peaks located within 4 Å (Fig. 1d), the additional phase is not amorphous, but nanocrystalline. In addition, we observe that amorphous phases can be formed by using a higher reduction temperature of >390 °C (Fig. S3†).

This presence of an additional nanocrystalline phase is not unreasonable, considering the fact that the reduction of (100)-oriented TiO₂ results in a single-crystalline corundum phase, while the reduction of (110)-oriented TiO₂ results in polycrystalline Ti₂O₃ domains surrounded by Ti₄O₇.¹⁰ This is because the crystallographically equivalent directions of [100]_R and [010]_R are geometrically non-equivalent in (100)-oriented TiO₂, which results in unique crystal orientation relationships of [100]_R||[110]_C, [010]_R||[001]_C, and [001]_R||[11̄0]_C, while (110)-oriented TiO₂, whose [100]_R and [010]_R are geometrically equivalent as well, becomes polycrystalline.¹⁰ (An “R” subscript is used for rutile, and “C” is used for corundum). According to the previous investigations,¹³ nanoparticles kept single crystalline domains, thus we consider that the inhomogeneous contact between TiO₂ particles and CaH₂ powder formed Ti₂O₃ nanoparticles and less-reduced nanoparticles as the additional phase, taking also the composition into consideration.

The structures of the TiO₂ nanoparticles and the Ti₂O₃ sample were refined well using the PDFs (Fig. 1d, S6 and S9†), but that of TiO_x-NP could not be fitted with Ti₂O₃ alone (Fig. S13†). For example, the peak assignable to the Ti–Ti distance between face-sharing octahedra in Ti₂O₃ (∼2.5 Å, Ti₁–Ti₂ in Fig. 1c) is smaller than those of Ti₂O₃. The PDF can be fitted reasonably well up to 10 Å with a model based upon a mixture of Ti₂O₃ and TiO₂ (Fig. S13†), but the fit is unacceptable in the longer distance region. Accordingly, the PDF pattern was then analyzed on the basis of a mixture of corundum Ti₂O₃ plus an oxygen-rich intermediate phase belonging to the homologous series Ti_nO_2n−1.¹⁰ The best fitting was achieved on the basis of a mixture of Ti₂O₃ and Ti₄O₇ (R_w = 23.5%) as shown in Fig. 1d. Considering the composition (TiO_1.83, which is richer in oxygen than Ti₂O₃ and Ti₄O₇) and the difference between 4 and 9 Å (Fig. 1d), TiO_x-NP probably contains a small amount of an oxygen-richer phase as well; for example, the Ti₂O₃ phase may be Ti deficient as indicated by the Rietveld refinements (cf. Fig. S12†).

The lattice constants of the Ti₂O₃ phase present in TiO_x-NP were determined to be a = 5.09391 Å and c = 13.7947 Å, which is highly compressed in the ab plane and expanded along the c axis (as illustrated in Fig. 1a) compared with a typical Ti₂O₃ (a = 5.15858 Å, and c = 13.6419 Å, refined by PDF, Fig. S6†). According to the Goodenough model,¹⁴ Ti₂O₃ becomes metallic at high temperatures with an increase in the c/a ratio,¹⁵ which closes the semiconductor gap between the a_1g band and the e^π_g band (Fig. 1e). The former band is formed between the face-sharing Ti–Ti dimers along the c axis, and the latter is formed between the edge-sharing octahedra in the ab plane.^16–19 The Ti–O bonds of the corundum phase in TiO_x-NP are shorter (peak at 2.02 Å) than in Ti₂O₃ (2.05 Å), as shown in Fig. S18.† This is consistent with the EELS O K-edge (Fig. 2d) in the high energy region (>535 eV): the peaks of TiO_x-NP and TiO_x-SC are shifted towards higher energies than in Ti₂O₃, suggesting stronger Ti–O bonding of TiO_x-NP and TiO_x-SC compared with Ti₂O₃.²⁰ Thus, the Ti₂O₃ phase in TiO_x-NP is considered to be metallic like the single crystal sample, TiO_x-SC. This view is also supported by the identical optical absorption features (Fig. S23†).

Fig. 2 Electronic structure analyses of rutile TiO₂ single crystal, reduced single crystal (TiO_x-SC), reduced nanoparticles (TiO_x-NP) and Ti₂O₃ particles. (a) Ti 2p HX-PES. (b) O 1s HX-PES with the Fermi edge spectra (insets). The takeoff angle was 88.5°. (c) Ti L-edge EELS. (d) O K-edge EELS. The EELS of Ti₂O₃ are consistent with those in the literature,²⁴ except for the small post-edge peaks of Ti L-edge indicated by ‘*’, which are attributable to the Ti₃O₅ impurities (Fig. S5†).

The HX-PES data (Fig. 2a and b) of TiO_x-NP and TiO_x-SC are identical and close to those of typical corundum Ti₂O₃.²¹ Their main Ti 2p peaks are located at the same energy as TiO₂ (Fig. 2a, inset), suggesting that less-screened final states are dominant. Since the electronic structure models of Ti₂O₃ including electron correlations are still open to debate, here we avoid following a specific one but rather focus on the experimental facts. The lower-energy shoulder of the Ti 2p_3/2 peak (459.4 eV) reflects core–hole screening by Ti 3d electrons, as is often observed in transition metal oxides.^22,23 The Ti 3d electrons exist just below the Fermi level E_F (Fig. 2b, inset: the peaks of TiO_x-SC, TiO_x-NP and Ti₂O₃ are at 0.97, 0.72 and 0.57 eV, respectively.), and these electrons should be in the a_1g band (Fig. 1e). Since the compositions of TiO_x-SC and Ti₂O₃ are the same, the 3d peak intensities probably reflect difference in the ionization cross-sections associated with orbital symmetry change.

The high-energy tail of the O 1s peak probably reflects charge deficiency in oxygen ions, rather than surface water molecules or hydroxyl groups to which the tail is usually assigned; because the detectable depth of the HX-PES method is ∼21 nm.§ Interestingly, the O 1s tail was reported to be prominent in the metallic phase of Ti₄O₇ but not in the semiconducting phase.²² From these O 1s HX-PES peaks and the O K-edge EELS peaks (Fig. 2d), the e^π_g conduction band of TiO_x-SC, TiO_x-NP and Ti₂O₃ are respectively calculated to be at ∼0.27, ∼0.41 and ∼0.33 eV above E_F. This is supported by the optical absorption measurements (Fig. S23†). Thus, the excellent conductivity of TiO_x-SC does not originate from a narrow semiconductor gap as expected on the basis of the Goodenough model.

The Ti L-edge EELS (Fig. 2c) of TiO_x-SC is unique compared with typical corundum Ti₂O₃, whose peaks are 2 eV lower than that of TiO₂.²⁴ This spectral behaviour is similar to Ti⁴⁺ ions in ilmenites (e.g., NiTiO₃),¹⁶ which has a corundum structure composed of M²⁺/Ti⁴⁺ alternating layers along the c-axis, but TiO_x-SC should consist of Ti³⁺ ions in view of the stoichiometric composition. That is, Ti³⁺ ions in TiO_x-SC appear to behave like Ti⁴⁺ ions during the electron excitations. This apparent 3d electron deficiency and the smaller 3d peak in the HX-PES may result from different electron correlations between TiO_x-SC and Ti₂O₃, probably originating from lattice strains. These different electronic structures are considered to be the origin of the metallic conduction of TiO_x-SC, though further detailed analyses are needed.

The EELS spectra of TiO_x-NP appear to contain Ti₄O₇, which has broad peaks in a similar range to Ti₂O₃,²⁴ together with the unique Ti₂O₃ of TiO_x-SC. From the PDF analysis, the mass fraction of the Ti₄O₇ phase is estimated to be 46%. The discrepancy in the conductivities between the TiO_x-SC and TiO_x-NP samples may originate from the Ti₄O₇ phase, which is expected to be less conducting, or from surface layer and/or interparticle resistance. The reduced (110)-oriented single crystal TiO₂, mentioned earlier, also contains Ti₄O₇, but shows comparable conductivity to the single crystalline Ti₂O₃ obtained from (100)-oriented single crystal TiO₂ (Fig. S25†). Thus, the presence of the Ti₄O₇ phase is not definitive in explaining the conductivity difference. It can be concluded that this is probably due to interparticle resistance.

Conclusions

Topotactic reactions are attractive post-synthetic treatments for nanomaterials because of their negligible impact on the nanostructure as well as possible creation of unique electronic properties. As exemplified by the rutile topotactic reduction, which is too slow (200 nm per day at the optimum conditions)¶ to be applied to bulk materials though feasible for nanomaterials (e.g., less than a few hundred nanometer-size particles), there may be plenty of topotactic reactions applicable only to nanomaterials. Thus, topotactic reactions have great potential to create nanomaterials having unique structures and properties that cannot be realized in bulk materials.

Since these reactions tend to proceed under kinetically controlled conditions, the products may contain nanocrystalline phases as well as crystalline phases, as we found for the reduced nanoparticle sample. Syntheses of nanomaterials are, generally, more or less kinetically controlled, thus careful analyses, with the presence of additional nanocrystalline or amorphous phases in mind, are needed for the discussion of their properties.

Methods

Synthesis

TiO_x-SC and TiO_x-NP were respectively synthesized by reducing mirror-polished, (100)-oriented rutile TiO₂ single crystals (0.5 mm thick, 5 mm × 5 mm pieces) and rutile nanoparticles (10–30 nm) using CaH₂ at 350 °C, as reported previously.^10,13

X-ray structure

The X-ray total scattering data were collected with Mo Kα (λ = 0.7107 Å) operated at 50 kV, 40 mA for 22 h (RINT RAPID-S, Rigaku). Background subtraction, X-ray polarization correction, absorption correction, and Compton scattering correction were performed using the PDFgetX2 program.²⁵ The structure functions (Q_max = 17.0 Å⁻¹, spatial resolution of 0.37 Å) were converted into PDFs. The PDF data were analyzed by curve fitting using the PDFgui program²⁶ with a spherical particle model.²⁷ Structures were visualized using the VESTA program.²⁸ The composition was determined by thermogravimetric analysis (SII EXSTAR6000) scanned at 1 °C min⁻¹ in air.

Electronic structure

HX-PES were collected with synchrotron hard X-rays (BL15XU SPring-8, 5.95 keV). The spectra were corrected through background subtraction with the Shirley method, the Savitzky–Golay smoothing process, and intensity normalization based on the peak areas of Ti 2p, O 1s or valence band. The EELS data were collected using a transmission electron microscope (JEM-2100F, JEOL) at 200 kV. The Ti L-edge and O K-edge spectra were corrected via a model-based deconvolution process using the EELSModel3.0 program.²⁹ In the model, a power law background and hydrogenic cross-sections of Ti L-edge and O K-edge were assumed. The energy-scales were calibrated by setting the O K-edge peak position of TiO₂ to be at 531.0 eV.³⁰

Acknowledgements

This work was supported by the World Premier International Research Center Initiative (WPI-MANA), MEXT, Japan. The NIMS Beamline at SPring-8 is acknowledged for beamtime (2011B4612). The authors are grateful to HiSOR, Hiroshima University, and JAEA/SPring-8 for the development of HX-PES at BL15XU of SPring-8. The authors thank Y. Tsujimoto (NIMS) for reduction treatment using CaH₂; A. Yamamoto (NIMS) for data processing support on the synchrotron X-ray PDF measurements; S. Ueda (NIMS), Y. Yamashita (NIMS), and S. Ishimaru (SPring-8 Service) for technical support on the HX-PES measurements; Y. Nemoto (NIMS) and I. Yamada (NIMS) for help with the EELS measurements; and T. Narushima (NIMS) for help in FIB sampling.

Notes and references

1. X. Chen, L. Liu, P. Y. Yu and S. S. Mao, Science, 2011, 331, 746–750.
2. H. M. Liu, W. S. Yang, Y. Ma and J. N. Yao, Appl. Catal., A, 2006, 299, 218–223.
3. D. C. Cronemeyer, Phys. Rev., 1959, 113, 1222–1226.
4. T. Ioroi, H. Kageyama, T. Akita and K. Yasuda, Phys. Chem. Chem. Phys., 2010, 12, 7529–7535.
5. T. Ioroi, H. Senoh, S. I. Yamazaki, Z. Siroma, N. Fujiwara and K. Yasuda, J. Electrochem. Soc., 2008, 155, B321–B326.
6. M. Kim, D. H. Kwon, K. M. Kim, J. H. Jang, J. M. Jeon, M. H. Lee, G. H. Kim, X. S. Li, G. S. Park, B. Lee, S. Han and C. S. Hwang, Nat. Nanotechnol., 2010, 5, 148–153.
7. C. Di Valentin, G. Pacchioni and A. Selloni, J. Phys. Chem. C, 2009, 113, 20543–20552.
8. V. M. Khomenko, K. Langer, H. Rager and A. Fett, Phys. Chem. Miner., 1998, 25, 338–346.
9. R. N. Blumenth, J. Coburn, J. Baukus and W. M. Hirthe, J. Phys. Chem. Solids, 1966, 27, 643–654.
10. S. Tominaka, Inorg. Chem., 2012, 51, 10136–10140.
11. T. Egami and S. J. L. Billinge, Underneath the Bragg peaks: structural analysis of complex materials, Elsevier, Oxford, UK, 2003.
12. S. J. L. Billinge and I. Levin, Science, 2007, 316, 1698.
13. S. Tominaka, Y. Tsujimoto, Y. Matsushita and K. Yamaura, Angew. Chem., Int. Ed., 2011, 50, 7418–7421.
14. L. L. V. Zandt, J. M. Honig and J. B. Goodenough, J. Appl. Phys., 1968, 39, 594–595.
15. C. E. Rice and W. R. Robinson, Mater. Res. Bull., 1976, 11, 1355–1359.
16. G. Radtke, S. Lazar and G. A. Botton, Phys. Rev. B: Condens. Matter Mater. Phys., 2006, 74, 155117.
17. L. F. Mattheiss, J. Phys.: Condens. Matter, 1996, 8, 5987–5995.
18. C. E. Rice and W. R. Robinson, Mater. Res. Bull., 1976, 11, 1355–1360.
19. L. L. Vanzandt, J. M. Honig and J. B. Goodenough, J. Appl. Phys., 1968, 39, 594–595.
20. M. Yoshiya, I. Tanaka, K. Kaneko and H. Adachi, J. Phys.: Condens. Matter, 1999, 11, 3217–3228.
21. T. Uozumi, K. Okada, A. Kotani, Y. Tezuka and S. Shin, J. Phys. Soc. Jpn., 1996, 65, 1150–1153.
22. M. Taguchi, A. Chainani, M. Matsunami, R. Eguchi, Y. Takata, M. Yabashi, K. Tamasaku, Y. Nishino, T. Ishikawa, S. Tsuda, S. Watanabe, C. T. Chen, Y. Senba, H. Ohashi, K. Fujiwara, Y. Nakamura, H. Takagi and S. Shin, Phys. Rev. Lett., 2010, 104, 106401.
23. K. Horiba, M. Taguchi, A. Chainani, Y. Takata, E. Ikenaga, D. Miwa, Y. Nishino, K. Tamasaku, M. Awaji, A. Takeuchi, M. Yabashi, H. Namatame, M. Taniguchi, H. Kumigashira, M. Oshima, M. Lippmaa, M. Kawasaki, H. Koinuma, K. Kobayashi, T. Ishikawa and S. Shin, Phys. Rev. Lett., 2004, 93, 236401.
24. E. Stoyanov, F. Langenhorst and G. Steinle-Neumann, Am. Mineral., 2007, 92, 577–586.
25. X. Qiu, J. W. Thompson and S. J. L. Billinge, J. Appl. Crystallogr., 2004, 37, 678.
26. C. L. Farrow, P. Juhas, J. W. Liu, D. Bryndin, E. S. Bozin, J. Bloch, T. Proffen and S. J. L. Billinge, J. Phys.: Condens. Matter, 2007, 19, 335219.
27. R. C. Howell, T. Proffen and S. D. Conradson, Phys. Rev. B: Condens. Matter Mater. Phys., 2006, 73, 094107.
28. K. Momma and F. Izumi, J. Appl. Crystallogr., 2011, 44, 1272–1276.
29. J. Verbeeck and S. Van Aert, Ultramicroscopy, 2004, 101, 207–224.
30. L. A. Grunes, R. D. Leapman, C. N. Wilker, R. Hoffmann and A. B. Kunz, Phys. Rev. B: Condens. Matter Mater. Phys., 1982, 25, 7157–7173.
31. J. R. Gunter and H. R. Oswald, Bull. Inst. Chem. Res., Kyoto Univ., 1975, 53, 249–255.
32. M. P. Seah and W. A. Dench, Surf. Interface Anal., 1979, 1, 2–11.
33. S. Tominaka, Chem. Commun., 2012, 48, 7949–7951.

---

Footnotes

† Electronic supplementary information (ESI) available: Experimental details, HX-PES, PDF, TGA, optical properties. See DOI: 10.1039/c3mh00087g

‡ A component(s) of the parent crystal is exchanged with the surroundings while retaining one or several crystallographically equivalent orientations to the product.³¹

§ Three times as deep as photoelectron inelastic mean free paths, ∼7 nm for both Ti 2p and O 1s.³²

¶ The reaction rate is 20 nm per day for the reduction treatment using CaH₂ at 350 °C,¹⁰ and 200 nm per day for the treatment using NaBH₄ at 375 °C.³³ Further stronger treatments resulted in low crystallinity (cf. Fig. S3†).

---