Scholarly article on topic 'Size dependence of structural, magnetic, and electrical properties in corundum-type Ti2O3 nanoparticles showing insulator–metal transition'

Size dependence of structural, magnetic, and electrical properties in corundum-type Ti2O3 nanoparticles showing insulator–metal transition Academic research paper on "Materials engineering"

CC BY-NC-ND
0
0
Share paper
Academic journal
Journal of Asian Ceramic Societies
OECD Field of science
Keywords
{"Low-temperature reduction" / "Titanium oxide" / "Ti2O3 " / "Size dependence" / "Insulator–metal transition" / "Corundum structure"}

Abstract of research paper on Materials engineering, author of scientific article — Yoshihiro Tsujimoto, Yoshitaka Matsushita, Shan Yu, Kazunari Yamaura, Tetsuo Uchikoshi

Abstract Corundum-type Ti2O3 has been investigated over the last half century because it shows unusual insulator–metal (I-M) transition over a broad temperature range (420–550K). In this work, we successfully synthesized Ti2O3 nanoparticles (20, 70, 300nm in size) by the low-temperature reduction between precursors of rutile-type TiO2 and the reductant CaH2, in a non-topotactic manner. The reaction time required for obtaining the reduced phase increases with increasing the particle size. Synchrotron X-ray powder diffraction and electron microscopy studies reveal that the symmetry of all the present samples remains the same as that of bulk samples. However, the particle-size reduction results in three important features compared with bulk samples as follows, (i) color shift from dark brown to bluish black, (ii) anisotropic volume contraction involving the shrinkage of Ti–Ti bonds in the ab plane and along the c axis, (iii) reduction of the I-M transition temperature from 420K to 350K. These suggest that the a 1g band broadening caused by the surface strain effects, which favors narrowing of the band gap, may play a critical role in the suppression of IM transition.

Academic research paper on topic "Size dependence of structural, magnetic, and electrical properties in corundum-type Ti2O3 nanoparticles showing insulator–metal transition"

HOSTED BY

ELSEVIER

Contents lists available at ScienceDirect

Journal of Asian Ceramic Societies

journal homepage www.elsevier.com/locate/jascer

Size dependence of structural, magnetic, and electrical properties in corundum-type Ti2O3 nanoparticles showing insulator-metal transition

CrossMark

Yoshihiro Tsujimoto3'*, Yoshitaka Matsushita®, Shan Yuc, Kazunari Yamaura' Tetsuo Uchikoshia

a Materials Processing Unit, National Institute for Materials Science, 1-2-1 Sengen, Tsukuba, Ibaraki 305-0047, Japan b Materials Analysis Station, National Institute for Materials Science, 1-2-1 Sengen, Tsukuba, Ibaraki 305-0047, Japan c Superconducting Materials Unit, National Institute for Materials Science, 1-1 Namiki, Tsukuba, Ibaraki 305-0044, Japan

ARTICLE INFO

Article history:

Received 16 April 2015

Received in revised form 15June 2015

Accepted 26 June 2015

Available online 17 July 2015

Keywords:

Low-temperature reduction Titanium oxide

Size dependence Insulator-metal transition Corundum structure

ABSTRACT

Corundum-type Ti2O3 has been investigated over the last half century because it shows unusual insulator-metal (I-M) transition over a broad temperature range (420-550 K). In this work, we successfully synthesized Ti2 O3 nanoparticles (20,70,300 nm in size) by the low-temperature reduction between precursors of rutile-type TiO2 and the reductant CaH2, in a non-topotactic manner. The reaction time required for obtaining the reduced phase increases with increasing the particle size. Synchrotron X-ray powder diffraction and electron microscopy studies reveal that the symmetry of all the present samples remains the same as that of bulk samples. However, the particle-size reduction results in three important features compared with bulk samples as follows, (i) color shift from dark brown to bluish black, (ii) anisotropic volume contraction involving the shrinkage of Ti-Ti bonds in the ab plane and along the c axis, (iii) reduction of the I-M transition temperature from 420 K to 350 K. These suggest that the aig band broadening caused by the surface strain effects, which favors narrowing of the band gap, may play a critical role in the suppression of IM transition.

© 2015 The Ceramic Society of Japan and the Korean Ceramic Society. Production and hosting by

Elsevier B.V. All rights reserved.

1. Introduction

Oxygen removal from the band-gap insulating oxides with an early transition metal, such as TiO2, V2 O5, and WO6 forms complex long-period structures involving oxygen-vacancy order or crystal-lographic shear [1-5]. In particular, it is known that the binary titanium-oxygen system forms a homologous series of oxygen sub-stoichiometric phases expressed as TinO2n-1 (n > 2), of which the phase with n from 4 to 10 is termed Magneli phase [2,3,6-9]. Fig. 1 shows the crystal structures of rutile TiO2 (n ^ to) and corundumlike Ti2O3 (n = 2), and Ti4O7 (n = 4). The rutile structure is composed of one-dimensional chains of edge-sharing TiO6 octahedra linked by sharing corners. In the corundum structure, each titanium-centered octahedron shares one face along the c axis and three

* Corresponding author. Tel.: +81 29 859 2553. E-mail address: TSUJIMOTO.Yoshihiro@nims.go.jp (Y. Tsujimoto). Peer review under responsibility ofThe Ceramic Society ofJapan and the Korean Ceramic Society.

edges in the ab plane with one another, corresponding to Ti(1)-Ti(2) andTi(1)-Ti(3) bonds, respectively. TinO2n-1 (4 < n <10) possesses structural features common to both the rutile and corundum structures: the rutile-like slabs extending in two dimensions and n octahedral thick in the third dimension are separated at a shear plane {1 2 1} by sharing edges and faces as in the corundum structure.

Removing oxygen from TiO2 also leads to the donation of electrons in Ti 3d bands. Of the series ofTinO2n-1, the phases of 2 < n <6 exhibit high electrical conductivity at room temperature comparable to that of carbon [10-12], which provides opportunities for a new class of titania-based electronic devices [13,14]. On the other hand, there has been a long-standing issue on the nanostructuring process of the reduced phases: high-temperature reducing environments required for the preparation of them involve particle growth [15,16]. Nevertheless, Ohkoshi and his collaborators successfully synthesized Ti3O5 nanoparticles with an average size of 25 nm by using the combination of sol-gel and reverse-micelle techniques [8], which has been widely employed to obtain well-dispersed nanocrystals including metal alloys and oxides [17-19].

http://dx.doi.org/10.1016/jjascer.2015.06.007

2187-0764 © 2015 The Ceramic Society ofJapan and the Korean Ceramic Society. Production and hosting by Elsevier B.V. All rights reserved.

Fig. 1. Crystal structures of rutile TiO2 (a), corundum-like Ti2O3 (b), and Ti4O7 (c). Schematic energy diagram of Ti 3d bands near Fermi level across the insulator-to-metal transition proposed byJ.B. Goodenough et al. [33].

They also demonstrated the photo-induced insulator-to-metal (IM) switching phenomenon in the nanoparticles, accompanying structural phase transitions [8]. In addition, a similar technique was reported to yield fiber-like nanostructured Ti4O7 with a high surface area [14], allowing for a high durable fuel cell catalyst support. However, this synthetic method comprises many steps and requires the removal of the silica matrix used against particle growth at the final stage. Therefore, simpler reaction methods toward reduced titania nanocrystals have been highly demanded.

In 1999, Hayward and Rosseinsky reported a new low-temperature reduction technique using binary metal hydride powders as a reduction agent [20], allowing for access to unprecedented reduced phases that cannot be obtained by conventional high temperature reduction reaction. Metal hydrides such as NaH and CaH2 exhibit strong reducing power even at temperatures less than 500 °C, which enables us to design in a topotactic manner complex metal oxides with unusual coordination environments around the metal centers [21-23], for example, a series of square-planar coordinated iron oxides Srn+1FenO2n+1 (n = 1, 2, to) [24-26] and novel oxyhydrides such as LaSrCoO3H0 7 [27] and SrCrO2H [28]. On the other hand, the strong reducing power at low temperatures is beneficial to suppression of particle growth. Yamamoto et al. demonstrated the successful synthesis of nickel and iron nanopar-ticles by reaction between the corresponding metal-organic salts and CaH2 at 140°C [29,30].

Our group recently applied this reduction method to a metal oxide, namely, titanium dioxide nanoparticles (dia. 20 nm) with the rutile structure [31], and found that the reaction of the rutile nanoparticles with CaH2 at 350 °C for several days yielded corundum-type Ti2O3 keeping the nanomorphology of the precursor. This is the first example of Ti2O3 nanoparticles, but its bulk or sintered sample has been extensively studied over the last half century because a sluggish I-M transition takes place in the temperature range from 420 to 550 K [32-37], unlike sharp I-M transitions seen in other n members. As forTi2O3, each titanium ion is subject to a slight trigonal distortion with six oxygen anions, which splits

3-fold degenerate t2g orbitals into a low-lying a1g orbital directed along the c axis and two egw orbitals directed to near-neighbor Ti ions in the basal plane [6,33]. As illustrated in Fig. 1 (d), the strong hybridization between two a1g orbitals within the Ti(1)-Ti(2) bond yields the bonding (a1g) and antibonding (a*1g) bands with a large energy gap, while the egv and egw* bands are partially overlapped and located in between the a1g and a*1g bands [33]. In the insulating state, a small energy gap of 0.1~0.2 meV opens between the fully occupied a1g and unoccupied egw bands [38,39]. The mechanism of the I-M transition ofTi2O3 is still controversial, but a widely accepted view is that a gradual increase of the Ti(1)-Ti(2) length on heating suppresses the separation of the a1g and a*1g bands, finally leading to the breakdown of the semiconducting gap between a1g and eg v bands. The Ti(1)-Ti(3) lengths in the basal plane shrink only slightly across the transition [6].

Our previous report onTi2O3 nanoparticles revealed the short a and long c lattice parameters (a = 5.0745 A, c = 13.7516 A) compared with those of bulk Ti2O3 (a = 5.1570A, c =13.610A) [31].The color was bluish black for the nanoparticles but dark brown for the bulk (see in Fig. 5). In general, particle-size reduction into nanometric scale significantly modifies the fractional atomic coordinates in the lattice and the electronic structure [40-42]. Since the I-M transition of Ti2O3 is intimately connected to the structural parameters, it is worth studying how the I-M transition varies with decreasing particle size. In this work, we have systematically investigated for the first time the effect of particle-size reduction on structural, magnetic, and electrical properties of Ti2O3 nanoparticles.

2. Experimental

Samples of nano-sized Ti2O3 were synthesized by the low-temperature reduction with CaH2, as reported previously [31]. Precursors of commercial rutile-type TiO2, which had ellipsoidal shape with an average major diameter of 20 (Ishihara Sangyo Kaisha, Ltd.), 70 (Sakai Chemical, Co., Ltd.), and 300 (Sakai Chemical, Co., Ltd.) nm, were used after preheating overnight at 200 °C.

Hence, each precursor is termed R1, R2, and R3, respectively. TiO2 and CaH2 (90%, Aldrich) were thoroughly mixed in a molar ratio of 1:4 in an argon-filled glovebox. The mixture was pelletized, sealed under vacuum in a Pyrex tube without exposure to air, and then heated at 350°C for several days with intermediate grinding. After reaction, samples were washed with a 0.1 M solution of NH4Cl in methanol to remove residual CaH2 and a byproduct CaO. Complete removal of calcium-containing phases was confirmed by energy-dispersive X-ray analysis. The nanoparticles reduced from R1, R2, and R3 are termed C1, C2, and C3, respectively.

As a reference, sintered sample ofTi2O3, which is termed 'bulk', was prepared by conventional solid-state reaction. Ti metal powders (99%, Kojundo Chemical Lab. Co., Ltd.) and TiO2 (99.9% Rare Metallic Co., Ltd.) were ground together in the stoichiometric ratio, pelletized, and sealed in an evacuated silica tube. The pelletized mixture was heated at 1150 °C for 12 h.

To characterize the obtained products, synchrotron X-ray diffraction data (SXRD, k = 0.65298 A) were collected from them at room temperature, using a Debye-Scherrer camera installed on BL15XU at SPring-8. Samples were put into a glass capillary with an internal diameter of 0.2 mm for the nanoparticles and 0.1 mm for the bulk. The data were recorded in 0.003° increments in a 29 range of 12-50°. Structural refinements were performed using the RIETAN-FP program [43].

Diffuse reflection spectra were recorded on a UV-vis-NIR spectrometer (UV-570, JASCO) with an integrated sphere equipment and barium sulfate as the reference, and then converted from reflection to absorbance by the Kubelka-Munk method. The specific surface areas were measured with a Macsorb surface area analyzer (HM Model-1201, Mountech) by Brunauer-Emmett-Teller (BET) method, using a mixed gas of N2 and He.

High-resolution transmission electron microscopy (HRTEM) and transmission electron microscopy (TEM) observations were performed using aJEL-2100 microscope (JEOL) at an accelerating voltage of 200 kV. Small amounts of samples were dispersed in ethanol by sonication, and the suspension was dropped onto a carbon-coated copper grid.

Electrical resistivity measurements for 20 nm-sized and bulk Ti2O3 were carried out by a four-point probe method in a Physical Property Measurement System (PPMS, Quantum Design). Magnetic susceptibility of the two samples was measured at an applied magnetic field H = 10 kOe using a Magnetic Property Measurement System (MPMS, Quantum Design). A high-temperature oven was used in the magnetometer to collect the magnetic data in the temperature region of 300 < T<600K.

3. Results and discussion

Fig. 2 shows the SXRD data collected from nano-sized and bulk samples of Ti2O3, which revealed a broadening of the diffraction peaks with decreasing particle size. The XRD patterns of the nanoparticles C1 and sintered sample could be readily indexed on the basis of the hexagonal cell with the corundum structure in the space group R-3c [6]. The calculated lattice parameters are a = 5.0826 (3)A and c =13.6864 (14) A for C1, and a = 5.15640 (1)A and c = 13.61003 (3)A for the bulk, which are in good agreement with those reported previously [6,31]. For nanoparticles C2 and C3, the corundum structure was also successfully obtained as the main phase, although some uncharacterized peaks with tiny intensities were found as minor phases in C3 nanoparticles. It is worth noting that the reaction time correlates with the particle size. Low-temperature reduction using CaH2 or related metal hydride proceeds through a solid-solid reaction between the reductant and precursors, and thus physical contact between them is an important key factor for promoting the reductive reaction [44]. In fact,

Fig. 2. Rietveld refinements against the synchrotron X-ray diffraction data of Ti2O3 nanoparticles (20,70, and 300 nm india. from the top panel) and the bulk. The wavelength was 0.65298 A. Observed, calculated, and difference are presented with cross marks, upper and bottom solid lines, respectively. The vertical lines represent the Bragg peak positions. ForC3 sample, the 29 regions where impurity peaks appeared were excluded during the refinements.

R1 and R2 with high surface areas of 80.72 and 69.64 m2/g, respectively, were completely reduced to Ti2O3 within 8 days, but 20 days for R3 with much lower surface area of 6.36 m2/g.

Figs. 3(a)-(c) and 4(a)-(c) show typical TEM images of the rutile precursors and their reduced phases, respectively. No apparent particle growth was observed after the reductive reaction in the TEM images, which is consistent with our previous report [31]. However, the BET measurements revealed a reduction in the specific surface areas, by 23 and 44% for C1 and C2 nanoparticles, respectively, in comparison with the corresponding precursors. These results suggest a partial aggregation of nanoparticles. In contrast, the specific surface area of C3 was 7.69 m2/g, which was larger than that of the precursor by 21%. The opposite behavior is likely related

to the fact that particles of R3 broke into pieces during reduction as seen in Fig. 4(c). Figs. 3(d)-(f) and 4(d)-(f) show typical HRTEM images of the precursors and products, and the insets display selected area electron diffraction (SAED) patterns for R1 and C1, and fast Fourier transforms (FFT) patterns for R2, R3, C2, and C3. The SAED or FFT patterns of each sample are consistent with diffraction patterns expected from the rutile or corundum structure. The HRTEM images of C1 and C2 as well as R1 and R2 showed good crys-tallinity and uniformly spaced lattice fringes in each particle, which indicates that each particle maintains the single-domain structure in spite of the drastic structural rearrangements from rutile through Ti4O7 to corundum structure [31]. In contrast, the HRTEM of C3 sample shown in Fig. 4(f) exhibited lattice fringes in the regions A and B running in different directions between each other, which was also confirmed from the FFT patterns. This observation revealed that the single crystalline nature of R3 was lost after the reduction.

Fig. 5 shows the diffuse reflectance spectra for C1, C2, C3, and the bulk. For comparison, the spectral data of 20 nm-sized TiO2 (R1) are

also plotted. In contrast to the absorption only under UV irradiation (À<400nm) for the TiO2 nanoparticles, the bulk Ti2O3 exhibited a strong absorption extending to the NIR region, with two broad bands centered at 378 and 514 nm, corresponding to the a1g-a*1g and a1g-egw* interband transitions, respectively [39,45,46]. These two peaks were gradually broadened with the decrease of particle size, which suggests a semiconducting-gap narrowing. We also see a gradual enhancement of the absorption in the NIR region with decreasing particle size, reflecting the difference in the sample color between the bulk and C1. Therefore, these spectral changes observed for the nanoparticles strongly suggest that the electronic structure of Ti2O3 is modified by size reduction or surface strain effects.

To refine the cell parameters and crystallographic sites of the products, Rietveld analysis was performed using the SXRD data on the basis of the corundum structure. Both the Ti and O sites remained fully occupied within errors, and thus fixed at unity during refinements. Plots of the observed and calculated diffraction data are shown in Fig. 2, and the final refined structural

Fig. 4. TEM and HRTEM images of Ti2O3 nanoparticles with 20 nm (a, d), 70 nm (b, e), and 300 nm (c, f) in diameter, respectively. The insets show selected area diffraction patterns (d) and fast Fourier transformation patterns (e, f).

parameters are tabulated inTable 1. Fig. 6 shows the cell parameters and c/a ratio plotted as a function of the particle size. Refined interatomic distances of Ti(1)-Ti(2) and Ti(1)-Ti(3) bonds are shown in Fig. 7. The lattice volume exhibited an anisotropic contraction

accompanying the particle-size reduction: the a axis monotonically decreased but the c axis expanded, resulting in an increase of the c/a ratio. The unusual size dependence of volume observed for Ti2O3 is quite different from related corundum compounds M2O3 (M = V,

Table 1

Crystallographic data forTi2O3 nanoparticles and the bulk obtained from Rietveld refinement of synchrotron X-ray diffraction data at room temperature. Atom Site C1 (20 nm) C2 (70 nm) C3 (300 nm) Bulk

x/z Biso (A2) x/z Biso (A2) x/z Biso (A2) x/z Biso (A2)

Ti 12c (00z) 0.3425(1) 2.11(4) 0.34277(7) 1.59(2) 0.34414(4) 1.34(1) 3.4465(2) 0.18(1)

O 18e (x 01/2) 0.3210(8) 0.09(7) 0.3183(6) 0.66 (б) 0.3113(2) 1.04(2) 0.3113(2) 0.10(1)

a (A) 5.0826(3) 5.1039(2) 5.13837(9) 5.15443(1)

c (A) 13.6864(14) 13.6759(8) 13.6232(3) 13.60810(2)

V (A3) 306.19(4) 308.52(2) 311.502(12) 313.1051 (7)

Rwp(%) 1.18 1.39 1.90 7.14

Rb(%) 4.14 5.11 3.41 4.20

Rf (%) 2.07 3.11 2.44 2.04

J__- 1 - 1 ' '

200 400 600 800 1000 1200 1400

À (nm)

Fig. 5. UV-vis-NIR absorption spectra of Ti2O3 nanoparticlesand bulk. The insets show photographs ofthe pellets: 20 nm-sizedTiO2 (white), 20nm-sizedTi2O3 nanoparticles (bluish black), and bulk Ti2O3 (dark brown).

Fig. 6. Variation of the a and c lattice parameters, volume, and c/a ratio with particle ratio for single crystals taken from Ref. [6].

size. The inset in the bottom-right panel shows the temperature evolution of the c/a

Fig. 7. Ti-Ti bond lengths in the ab plane and along the c axis as a function of particle size.

Cr) [42,47] where both the a and c lattice parameters shrink with decreasing the particle size.

The c/a ratio is often used as a measure of the I-M transition [6], because the a and c lattice parameters are related to the Ti(1)-Ti(3) bonds in the basal plane and the Ti(1)-Ti(2) bond along the c axis. The inset of Fig. 6(d) shows the temperature dependence of the c/a ratio measured with single crystals by Rice and Robinson in Ref. [6]. The c/a value exhibits a slow increase with increasing temperature up to 400 K, but rapidly increases above the I-M transition temperature (TIM). This behavior is mainly ascribed to the Ti(1)-Ti(2) bond elongation along the c axis. Looking back to the present Ti2O3 nanoparticles, the values of the c/a ratio at room temperature increased as the particle size decreased, and especially for C1 and C2, and are comparable to those in the metallic region of the single crystals. However, both Ti(1)-Ti(2) and Ti(1)-Ti(3) bonds decreased, for example, by -1.60% and -1.74% for C1, respectively, when compared to the bulk values (Fig. 7). These results indicate that the relationship between the c/a ratio and the Ti-Ti bond lengths is different between temperature and particle size.

The electrical resistivity and magnetic susceptibility measurements were performed in order to study the influence of particle-size reduction on the I-M transition. Fig. 8 presents the

temperature dependence of the electrical resistivity (p) of the C1 and bulk. The p of the bulk exhibited a semiconducting behavior in the measured temperature range from 250 to 400 K (<TiM), which is consistent with previous reports [ref]. The resistivity curve of C1 also showed negative temperature coefficient up to 400 K, but began to drop steeply at 350 K. This characteristic behavior is very similar to the resistivity reduction at the onset temperature of I-M transition of bulk samples in early reports [6,11,48]. Thus, the observed change in the resistivity of C1 at 350 K should be derived from the I-M transition. On the other hand, the resistivity of C1 is more than three orders of magnitude greater than that of the bulk sample, which should be attributed to surface oxidation of the nanoparticles, as observed by the XPS measurements [31].

Fig. 9 shows the temperature dependence of the magnetic susceptibility x (=M/T) of the C1 and bulk. The inset displays an enlargement of the data in the range of 300 < T<600K. The magnetic susceptibility of the bulk is consistent with those in previous studies [11,49]. The y-Tcurve at low temperatures (<TjM) exhibited an almost temperature-independent behavior reflecting the nonmagnetic or spin-singlet ground state [50], which was followed by a sudden increase caused by a transition to the paramagnetic metal phase at 420 K. Similar upturn was also observed in the magnetic susceptibility of C1 at 350 K, corresponding to the anomaly observed in the p vs T curve. Therefore, we can conclude that particle-size reduction shifts the TjM of the bulk Ti2 O3 to lower temperatures. The y-Tcurve of C1 monotonically increased even below TIM in contrast to the nearly temperature-independent behavior in the bulk. The difference is probably derived from non-interacting free spins created by the surface oxidation described above.

The electrical resistivity and magnetic susceptibility measurements demonstrated the decrease of TjM by size reduction. The observed TIM reduction is apparently not consistent with the contraction of the Ti(1)-Ti(2) bond. However, the volume contraction involving the shortage the Ti(1)-Ti(2) bond is of advantage in an enhancement of bandwidth of the Ti 3d bands, as seen in various transition metal compounds showing pressure-induced metallization [51]. Indeed, the absorption spectroscopy measurements evidenced the broadening of the a1g bandwidth. Similar volume reduction was observed in bulk Ti2O3 under hydrostatic pressure conditions [52,53], but increasing pressure results in the decrease of both the a and c lattice parameters with the c/a ratio nearly

Fig. 8. Electrical resistivity (p) of Ti2O3 nanoparticles (20 nm) and the bulk sample as a function of temperature. The p for the bulk is multiplied by 103. The insulator-to-metal transition temperature is marked with arrow.

Fig. 9. Magnetic susceptibility of Ti2O3 nanoparticles (20 nm) and the bulk sample as a function of temperature in an applied field of 10k0e. The inset displays the data measured with a high-temperature oven. The insulator-to-metal transition temperature is marked with arrows.

unchanged. This is significantly different from the anisotropic volume change against particle size observed in this study.

It is interesting to compare with studies on isostructural V2O3 nanoparticles. Bulk V2O3 is known to undergo the 1st order phase transition from a paramagnetic metal to antiferromagnetic insulator at around 160 K, accompanying a structural phase transition from the hexagonal (R-3c) to the monoclinic (/2/a) structure [47,54]. The trigonal symmetry on the vanadium center splits the manifold t2g bands into non-degenerate a1g and doubly degenerate egw bands. Note that the former is located at higher energy level than the latter, which is opposite to the situation in Ti2O3. In the low-T insulating state, the on-site Coulomb repulsion in vanadium further splits the latter bands into the lower Hubbard band (LHB) and upper Hubbard band (UHB), and an energy gap ~0.5 eV opens between a1g and the low-lying LHB [55]. The V-V bond length along the c axis suddenly shortens across the I-M transition to the metallic state to collapse the energy gap. On the other hand, size reduction into nanometric scale (~10nm) for V2O3 exhibits a uniaxial compression along the c axis through surface strain [47]: both the c axis length and the V-V bond length along the c-axis are contracted without significant changes in the a axis and inplane V-V bond lengths. These structural modification causes not only broadening but also low-energy shift of the a1g band, leading to disappearance of the I-M transition down to low temperatures. The modulation of the a1g energy band in Ti2O3 by size reduction is very similar to that in V2O3 nanoparticles, but the difference of energy levels between the a1g and egw bands obviously dictates their electronic and magnetic ground states.

We would like to emphasize that the reduction in the I-M transition temperature for Ti2O3 has been observed for the first time in this study. The electrical resistivity of bulk Ti2O3 gradually decreases with increasing chemical doping [39] and hydrostatic pressure [56]; however, the Tim remains unchanged regardless of the doping and pressure levels, and the doping and pressure variations of the gap energy between a1g and egw bands scale with that of the Ti(1)-Ti(2) bond length or c/a ratio. These behaviors are in sharp contrast with Ti2O3 nanoparticles. It is apparent that the electronic state of Ti2O3 is perturbed by surface strain effects in a different way than chemical doping and hydrostatic pressure. To gain deeper understanding the causal connection between the surface strain effects and Tim reduction, further investigation into the local structures around Ti and O atoms should be made using other experimental techniques such as extended X-ray absorption fine structure spectroscopy and X-ray emission spectroscopy [37,47].

Recently, Tominaka and his collaborators investigated single crystal samples (2 mm x 3 mm x 0.5 mm) of (1 0 0)-oriented TiO2 reduced under similar reaction conditions described in this study, and found a metallic conductivity in the measured temperature range of 8-300 K [57], which is markedly contrasted with our experimental results. The quality of their samples, however, is not suitable to evaluate the physical properties ofTi2O3 prepared with CaH2: the reduction to Ti2O3 only proceeded 200 nm in depth and the conduction path is not clear due to cracks shown in TEM images. Indeed, we found that (1 0 0)-oriented TiO2 single crystals reduced by CaH2 were quite fragile and easily cleaved perpendicular to (0 1 1) plane because of the non-topotactic reaction involving displacement of titanium atoms and removal of oxygen atoms. These facts highly indicate that the TiO2 layer suffers oxygen removal or stacking faults. Introduction of such oxygen deficiencies or defects into TiO2 is well known to give a metallic behavior [58-60]. And, since the reductive reaction proceeds through Magneli phases such as Ti4O7 [31], the interfacial structure between Ti2O3 and TiO2, which has not been distinctly characterized, likely contributes to the metallic conductivity. In fact, the X-ray photoelectron spec-troscopy measurements for the inhomogeneous sample denied the existence of a finite density of states at the Fermi energy at room

temperature [61], that is, the metallic conductivity observed by the authors is not an intrinsic property for the nanostructured Ti2O3. In addition, Tominaka et al. claimed later that Ti2O3 nanoparticles with 20 nm in diameter, corresponding to C1 sample in this study, were a mixture ofTi2O3 andTi4O7 and the mass fraction of the latter was estimated to be approximately 50% on the basis of pair distribution functions study [60]. However, our experimental results can exclude their argument definitely; we did not see either XRD peaks assignable to Ti4O7 or anomalies caused by I-M transitions at around 150 K in the magnetic susceptibility and electrical resistivity curves of our samples. Therefore, the conclusion presented in their early studies is overly optimistic and misleading.

4. Conclusions

We have successfully synthesized Ti2O3 nanoparticles with different particle sizes (20, 70, and 300 nm in dia.) by a low-temperature reduction, and have systematically investigated the structural, electrical, and magnetic properties. All the present products adopt the corundum structure as bulk Ti2O3, but exhibit a gradual color change from dark brown to bluish black with decreasing particle size because of surface strain effects. We also found unusual reduction in TIM compared with the bulk with decreasing particle size, which has never been observed in the bulk where the Tim is not affected by chemical doping and hydrostatic pressure. The anisotropic volume contraction and the shrinkage of the nearest neighbor Ti-Ti bond length along the c axis, which are followed by broadening and high-energy shift of the a1g band, are also unique features seen only in Ti2O3 nanoparticles. Information on the local structures modified by surface strain effects shed light on the mechanism of the TIM reduction.

Acknowledgements

This work was supported in part by the Japan Society for the Promotion of Science (JSPS) through a Grant-in-Aid for Scientific Research (25289233 and 50584075). Y.T. acknowledges Y. Katsuya, M. Tanaka, and O. Sakata for supporting the SXRD experiments (2012A4508, 2014A4504).

References

[1] S. Andersson and A.D. Wadsley, Nature, 211,581-583 (1966).

[2] L.A. Bursill and B.G. Hyde, Prog. Solid State Chem., 7,177-253 (1972).

[3] S. Andersson, A. Sundholm and A. Magneli, Acta Chem. Scand., 13, 989-997 (1959).

[4] S. Yamazaki, C. Li, K. Ohoyama, M. Nishi, M. Ichihara, H. Ueda and Y. Ueda, J. Solid State Chem., 183,1496-1503 (2010).

[5] M.M. Dobson and R.J.D. Tilley, Acta Cryst., B44,474-480 (1988).

[6] C.E. Rice and W.R. Robinson, Acta Cryst., B33,1342-1348 (1977).

[7] S. Asbrjnk and A. Magneli, Acta Cryst., 12,575-581 (1959).

[8] S. Ohkoshi, Y.Tsunobuchi,T. Matsuda, K. Hashimoto, A. Namai, F. Hakoe and H. Tokoro, Nat. Chem., 2,539-545 (2010).

[9] M. Marezio and P.D. Dernier, J. Solid State Chem., 3,340-348 (1971).

[10] J.R. Smith and F.C. Walsh, J. Appl. Electrochem., 28,1021-1033 (1998).

[11] A. Kitada, G. Hasegawa, Y. Kobayashi, K. Kanamori, K. Nakanishi and H. Kageyama, J. Am. Chem. Soc., 134,10894-10898(2012).

[12] H. Ueda, K. Kitazawa, H. Takagi and T. Matsumoto, J. Phys. Soc. Jpn., 71, 1506-1510(2002).

[13] T. Ioroi, Z. Siroma, N. Fujiwara, S. Yamazaki and K. Yasuda, Electrochem. Commun., 7, 183-188(2005).

[14] C. Yao, F. Li, X. Li and D. Xia, J. Mater. Chem., 22,16560-16565 (2012).

[15] T. Ioroi, H. Kageyama, T. Akita and K. Yasuda, Phys. Chem. Chem. Phys., 12, 7529-7535 (2010).

[16] W.-Q. Han and Y. Zhang, Appl. Phys. Lett., 92, (2008), 203117-1-203117-3.

[17] T. Li, M. Moon, A.A. Morrone, J.J. Mecholsky, D.R. Talham and J.H. Adair, Lang-muir, 15, 4328-4334 (1999).

[18] S. Yamamoto, Y. Moritomo, T. Ono and M. Takano, Appl. Phys. Lett., 87, (2005), 032503-1-032503-3.

[19] J.Jin, S. Ohkoshi and K. Hashimoto, Adv. Mater., 16, 48-51 (2004).

[20] M.A. Hayward, M.A. Green, M.J. Rosseinsty and J. Sloan, J. Am. Chem. Soc., 121, 8843-8854(1999).

[21] M.A. Hayward, Semicond. Sci. Technol., 29,0664010 (2014).

[22] C. Tassel and H. Kageyama, Chem. Soc. Rev., 41,1989-2520 (2012).

[23] T. Yamamoto and H. Kageyama, Chem. Lett., 42,946-953 (2013).

[24] Y. Tsujimoto, C. Tassel, N. Hayashi, T. Watanabe, H. Kageyama, K. Yoshimura, M. Takano, M. Ceretti, C. Ritter and W. Paulus, Nature, 450,1062-1065 (2007).

[25] C. Tassel, L. Seinberg, N. Hayashi, S. Ganesanpotti, Y. Ajiro, Y. Kobayashi and H. Kageyama, Inorg. Chem., 52, 6096-6102 (2013).

[26] H. Kageyama, T. Watanabe, Y. Tsujimoto, A. Kitada, Y. Sumida, K. Kanamori, K. Yoshimura, N. Hayashi, S. Muranaka, M. Takano, M. Ceretti, W. Paulus, C. Ritter and G. André, Angew. Chem. Int. Ed., 47,5740-5745 (2008).

[27] M.A. Hayward, E.J. Cussen, J.B. Claridge, M. Bieringer, M.J. Rosseinsky, C.J. Kiely, S.J. Blundell, I.M. Marshall and F.L. Pratt, Science, 295,1882-1884 (2002).

[28] C. Tassel, Y. Goto, Y. Kuno, J. Hester, M. Green, Y. Kobayashi and H. Kageyama, Angew. Chem. Int. Ed., 53,10377-10380 (2014).

[29] L. Seinberg, S. Yamamoto, R. Gallage, M. Tsujimoto, Y. Kobayashi, S. Isoda, M. Takano and H. Kageyama, Chem. Commun., 48, 8237-8238 (2012).

[30] L. Seinberg, S. Yamamoto, M. Tsujimoto, Y. Kobayashi, M. Takano and H. Kageyama, Chem. Commun., 50,6866-6868 (2014).

[31] S. Tominaka, Y. Tsujimoto, Y. Matsushita and K Yamaura, Angew. Chem. Int. Ed., 50, 7418-7421 (2011).

[32] F.J. Morin, Phys. Rev. Lett., 3,34-36 (1959).

[33] L.L. Van Zandt, J.M. Honig and J.B. Goodenough, J. Appl. Phys., 39, 594-595 (1968).

[34] A. Tanaka, J. Phys. Soc. Jpn., 73,152-162 (2004).

[35] A.I. Poteryaev, A.I. Lichtenstein and G. Kotliar, Phys. Rev. Lett., 93, (2004), 086401-1-086401-4.

[36] L.F. Mattheiss, J. Phys.: Condens. Matter., 8,5987-5995 (1996).

[37] H. Sato, A. Tanaka, M. Sawada, F. Iga, K. Tsuji, M. Tsubota, M. Takemura, K. Yaji, M. Nagira, A. Kimura, T. Takabatake, H. Namatame and M. Taniguchi, J. Phys. Soc. Jpn., 75, (2006), 053702-1-053702-4.

[38] T.C. Chi and R.J. Sladek, Phys. Rev. B, 15, 5080-5085 (1973).

[39] M. Uchida, J. Fujioka, Y. Onose and Y. Tokura, Phys. Rev. Lett., 101, (2008), 066406-1-066406-4.

[40] G. Li,J.J. Boerio-Goates, B.F. Woodfield and L. Li, Appl. Phys. Lett., 85,2059-2061 (2004).

[41] X.G. Zheng, H. Kubozono, H. Yamada, K. Kato, Y. Ishiwata and C.N. Xu, Nat. NanotechnoL, 3, 724-726 (2008).

[42] D. Tobia, E. De Biasi, M. Granada, H.E. Troiani, G. Zampieri, E. Winkler and R.D. Zysler, J. Appl. Phys., 108,(2010), 104303-1-104303-8.

[43] F. Izumi and K. Momma, Solid State Phenom., 130,15-20 (2007).

[44] Y. Kobayashi, Z. Li, K. Hirai, C. Tassel, F. Loyer, N. Ichikawa, N. Abe, T. Yamamoto, Y. Shimakawa, K. Yoshimura, M. Takano, O.J. Hernandez and H. Kageyama, J. Solid State Chem., 207,190-193 (2013).

[45] S.H. Shin, F.H. Pollak, T. Halpern and P.M. Raccah, Solid State Commun., 16, 687-690(1975).

[46] S.S.M. Lu, F.H. Pollak and P.M. Raccah, Phys. Rev. B, 17,1970-1975 (1978).

[47] Y. Ishiwata, S. Suehiro, T. Kida, H. Ishii, Y. Tezuka, H. Oosato, E. Watanabe, D. Tsuya, Y. Inagaki, T. Kawae, M. Nantoh and K. Ishibashi, Phys. Rev. B, 86, (2012), 035449-1-035449-5.

[48] C.N.R. Rao, R.E. Loehman and J.M. Honig, Phys. Lett., 27A, 271-272 (1968).

[49] A.D. Person, J. Phys. Chem. Solids, 5,316-327 (1958).

[50] R.M. Moon, T. Riste and W.C. Koehler, J. Appl. Phys., 40,1445-1447 (1969).

[51] M. Imada, A. Fujimori and Y. Tokura, Rev. Mod. Phys., 70,1039-1263 (1998).

[52] D. Nishio-Hamane, M. Katagiri, K. Niwa, A. Sano-Furukawa,T. OkadaandT. Yagi, High Press. Res., 29, 379-388 (2009).

[53] S.V. Ovsyannikov, X. Wu, A.E. Karkin, V.V. Shchennikov and G.M. Manthilake, Phys. Rev. B, 86, (2012), 024106-1-024106-14.

[54] P.D. Dernier and M. Marezio, Phys. Rev. B, 2,3771-3776 (1970).

[55] S.Y. Ezhov, V.I. Anisimov, D.I. Khomskii and G.A. Sawatzky, Phys. Rev. Lett., 83, 4136-4139(1999).

[56] B. Viswanathan, S.U. Devi and C.N.R. Rao, Pramana, 1,48-52 (1973).

[57] S. Tominaka, Inorg. Chem., 51,10136-10140 (2012).

[58] T. Nakajima, T. Tsuchiya and T. Kumagai, J. Solid Stat Chem., 182, 2560-2565 (2009).

[59] A. Kitada, S. Kasahara, T. Terashima, K. Yoshimura, Y. Kobayashi and H. Kageyama, Apex, 4, (2011), 035801-1-035801-3.

[60] D.S. McLachlan, Phys. Rev. B, 25, 2285-2296 (1982).

[61] S. Tominaka, H. Yoshikawa, Y. Matsushita and A.K. Cheetham, Mater. Horiz., 1, 106-110(2014).