**Characterization of the Electronic Structure of Spinel Superconductor LiTi2O4 using Synchrotron X-ray Spectroscopy**

Chi-Liang Chen and Chung-Li Dong

Additional information is available at the end of the chapter

#### **1. Introduction**

Numerous studies on high-transition temperature (*T*c) superconductivity were motivated primarily by the intention to explore the nature of cuprates. Experimental results, particularly those based on hard X-ray absorption spectra, show that Cu ions reveal a mixed valence and are accountable for high-*T*c superconductivity. Recently discovered iron-based superconduc‐ tors with perovskite blocking layers [i.e., LnFeAsO (Ln = lanthanide), BaFe2As2, KFeAs, FeSe, and FeAs] have received considerable attention similar to cuprate superconductivity, which was first explored in the 1980s. New superconducting compounds and non-cuprate super‐ conductors with magnetic elements, particularly 3*d* transition metals, have been widely analyzed by various scholars. Spinel LiTi2O4 (LTO) is a titanium (Ti)-based superconductor that is considered an exotic BCS *s*-wave superconductor with a *T*<sup>c</sup> of ~12 K [21]. This super‐ conductor is regarded as such because it can be described using electron–phonon interaction within the framework of the BCS model [43]. Of the more than 200 typical compounds with an AB2O4 normal spinel structure, in which low-valent-state cation "A" occupies the tetrahe‐ dral (8a, T*d*) interstitial sites and high-valent-state cation "B" occupies the octahedral (16d, O*h*) sites, only a few exhibit superconductivity, including CuV2S4 (*T*c = 4.5 K), CuRhS4 (*T*c = 4.8 K), and CuRhSe4 (*T*<sup>c</sup> = 3.5 K). LTO is the only known spinel oxide superconductor, and it has the highest *T*c among superconductors with a spinel structure. The 3*d* transition metals occupying the O*<sup>h</sup>* sites in spinel oxides may generally exhibit antiferromagnetism, ferromagnetism, and charge/orbital ordering. The nature of magnetic and electronic properties depends on the average valence of the cations. The closely related case of LiV2O4 (LVO), which is an isostruc‐ tural compound, exhibits distinct physical properties. LVO is the closest neighbor to LTO, and it contains a mixed-valence system with an equal ratio of V3+ (*S* = 1) and V4+ (*S* = ½). LVO exhibits

© 2015 The Author(s). Licensee InTech. This chapter is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

strong electronic correlation, resulting in a greatly enhanced effective mass. The resistivity of this compound displays a metallic character. Although LVO is not a superconductor, it exhibits a heavy fermionic behavior typically observed in a Ce-based (4*f* electron) system [2, 22].

Similar to cuprates [6, 33] and pnictides [9, 24], Ti-based superconductors (titanate) present the possibility of a mixed-valence state (with electronic configuration of 3*d*0.5, equal ratio of Ti3+ of spin *S* = 1/2 and Ti4+ with *S* = 0) in the ground state. Given that magnetic impurities can suppress the superconductivity of various materials, an investigation into whether the dilute doping of magnetic ions in LTO results in the complete suppression of superconductivity must be conducted. Such an undertaking may provide valuable insights into the mechanism of superconductivity. A previous research observed a rapidly suppressed superconducting *T*<sup>c</sup> in magnetically doped system LiTi1-*x*M*x*O4 (M = Cr and V) [19]. Among the explored dopants, vanadium (V) was determined to be more effective in decreasing *T*<sup>c</sup> even at small proportions, and it was also observed in other 3*d* transition metals (e.g., Cr, Mn, Fe, etc.). The results of the study not only signified the magnetic influences and other mechanisms for *T*c suppression, but also the charge density wave and structural distortion. For V-doped LTO, the *T*c of LTO decreases with V at 2% proportion from ~13 K to ~5 K [18, 19, 24, 49]. These superconducting activities in such a series may unravel the mechanism of superconductivity. Material study on the atomic/electronic structure has essentially illuminated the understanding of its electron transportation properties. Studies have investigated the electronic structure of LTO [8, 39]. The decrease in superconductivity of V-doped LTO is discussed based on a pair-breaking mecha‐ nism [19] and a crystalline distortion from octahedral symmetry [34]. The change of Ti and V valence states is suggested to be thoroughly linked to their chemical and physical properties. Electron–electron interaction cannot be overlooked because magnetic ions occupy the struc‐ tures [32]. In this event, the mechanism of superconductivity remains unknown.

The atomic and electronic structures of the novel spinel LTO superconductors have been investigated. X-ray spectroscopy, X-ray absorption near-edge structure (XANES) spectrosco‐ py, and resonant inelastic soft X-ray scattering (RIXS) spectroscopy are powerful tools for obtaining information on the local orbital character of a specific element. These approaches are also used to probe occupied or unoccupied electronic states near the Fermi level (*E*F) and the structural symmetry of mixed-oxide systems. However, these spectroscopic techniques are sensitive to atomic symmetry because edge-sharing charge and charge distribution induce electron–electron and electron-orbital interactions, which may be related to the magnetic nature of the transition metal systems.

XANES spectroscopy is a sensitive tool that can be employed to probe unoccupied electronic states above *E*<sup>F</sup> and to analyze the structural symmetry of mixed-oxide systems [10, 12]. It can provide details on the electronic–orbital interaction from the transition metal 3*d*-O 2*p* hybrid‐ ization states and the symmetry of the atomic structure. Therefore, the Ti L- and K-edges can be used to determine the valence states of Ti ions in LiTi2-*x*V*x*O4 (LTVO). The V ions occupying and doped in the O*<sup>h</sup>* site may distort the crystal structure of LTVO because of the bonding of O–Ti–O as well as the unoccupied states in the 3*d* orbitals [34].

RIXS spectroscopic technique is used to explore the electronic structure of materials and to associate the structure with the XANES spectrum. RIXS is a process that describes the deexcitation of the final state of X-ray absorption, providing information on ground state via the excited intermediate states. Consequently, by tuning various excitation energies, certain RIXS features can be enhanced, allowing the separation of different electronic configurations in the mixed ground states [11, 30, 42]. The above conditions indicate that RIXS is a complementary tool compared with XANES. Unlike XANES, RIXS can be used to investigate a forbidden electronic transition (e.g., *d*-*d* or *f*-*f* excitation) because it involves two dipole-allowed transi‐ tions, given that its final state has symmetry similar to that of the initial state of the former. The RIXS of TM L-edge can reflect the 3*d* partial density of states and is regarded as a useful approach for analyzing the electron correlation among strongly correlated materials [31, 36, 45] as well as the charge transfer between TM 3*d* and O 2*p* orbitals.

In this research, XANES and RIXS spectra were applied to understand the influence of V doping on the atomic/electronic properties of the LTVO system and to discuss the basis of the quick suppression of superconductivity. The occupied or unoccupied states near *E*F of a solidstate solution may be a useful starting point in obtaining fundamental information on the complicated behavior of materials, including ferrites, high-*T*<sup>c</sup> superconductors, and strongly correlated and ladder systems. These properties demonstrate that magnetic behavior and atypical physical and chemical characteristics can be synthesized and analyzed. The obtained X-ray spectra reveal that the rapidly suppressed superconductivity is associated with the variation of Ti-O hybridization mainly at *e*<sup>g</sup> bands rather than with the magnetic nature of the substituted V ion.

#### **2. Experiments**

strong electronic correlation, resulting in a greatly enhanced effective mass. The resistivity of this compound displays a metallic character. Although LVO is not a superconductor, it exhibits a heavy fermionic behavior typically observed in a Ce-based (4*f* electron) system [2, 22].

Similar to cuprates [6, 33] and pnictides [9, 24], Ti-based superconductors (titanate) present the possibility of a mixed-valence state (with electronic configuration of 3*d*0.5, equal ratio of Ti3+ of spin *S* = 1/2 and Ti4+ with *S* = 0) in the ground state. Given that magnetic impurities can suppress the superconductivity of various materials, an investigation into whether the dilute doping of magnetic ions in LTO results in the complete suppression of superconductivity must be conducted. Such an undertaking may provide valuable insights into the mechanism of superconductivity. A previous research observed a rapidly suppressed superconducting *T*<sup>c</sup> in magnetically doped system LiTi1-*x*M*x*O4 (M = Cr and V) [19]. Among the explored dopants, vanadium (V) was determined to be more effective in decreasing *T*<sup>c</sup> even at small proportions, and it was also observed in other 3*d* transition metals (e.g., Cr, Mn, Fe, etc.). The results of the study not only signified the magnetic influences and other mechanisms for *T*c suppression, but also the charge density wave and structural distortion. For V-doped LTO, the *T*c of LTO decreases with V at 2% proportion from ~13 K to ~5 K [18, 19, 24, 49]. These superconducting activities in such a series may unravel the mechanism of superconductivity. Material study on the atomic/electronic structure has essentially illuminated the understanding of its electron transportation properties. Studies have investigated the electronic structure of LTO [8, 39]. The decrease in superconductivity of V-doped LTO is discussed based on a pair-breaking mecha‐ nism [19] and a crystalline distortion from octahedral symmetry [34]. The change of Ti and V valence states is suggested to be thoroughly linked to their chemical and physical properties. Electron–electron interaction cannot be overlooked because magnetic ions occupy the struc‐

tures [32]. In this event, the mechanism of superconductivity remains unknown.

nature of the transition metal systems.

18 Superconductors – New Developments

O–Ti–O as well as the unoccupied states in the 3*d* orbitals [34].

The atomic and electronic structures of the novel spinel LTO superconductors have been investigated. X-ray spectroscopy, X-ray absorption near-edge structure (XANES) spectrosco‐ py, and resonant inelastic soft X-ray scattering (RIXS) spectroscopy are powerful tools for obtaining information on the local orbital character of a specific element. These approaches are also used to probe occupied or unoccupied electronic states near the Fermi level (*E*F) and the structural symmetry of mixed-oxide systems. However, these spectroscopic techniques are sensitive to atomic symmetry because edge-sharing charge and charge distribution induce electron–electron and electron-orbital interactions, which may be related to the magnetic

XANES spectroscopy is a sensitive tool that can be employed to probe unoccupied electronic states above *E*<sup>F</sup> and to analyze the structural symmetry of mixed-oxide systems [10, 12]. It can provide details on the electronic–orbital interaction from the transition metal 3*d*-O 2*p* hybrid‐ ization states and the symmetry of the atomic structure. Therefore, the Ti L- and K-edges can be used to determine the valence states of Ti ions in LiTi2-*x*V*x*O4 (LTVO). The V ions occupying and doped in the O*<sup>h</sup>* site may distort the crystal structure of LTVO because of the bonding of

RIXS spectroscopic technique is used to explore the electronic structure of materials and to associate the structure with the XANES spectrum. RIXS is a process that describes the deAll the LTVO samples were synthesized through conventional two-step solid-state reactions with highly pure oxides of TiO2 (99%), Li2CO3 (99.99%), and V2O3 (99%) [19]. The samples were dried at 150 °C for at least 2 h. Additional Li2CO3 (5 mol%) was added into the samples to compensate for lithium evaporation at high temperature. These processes were performed in a dynamic vacuum environment. In the first step, a mixture of Li2CO3 and TiO2 was pulverized and calcined at 750 °C for 8 h in ambient air. The mixture was pulverized and calcined again at 800 °C for 12 h to form the intermediate compound Li2Ti2O5. In the second step, the stoi‐ chiometric powders of Li2Ti2O5, Ti2O3, and V2O3 were mixed and ground and then cold-pressed into pellets. The pellets were wrapped with gold foil and sintered in an alumina crucible at 880 °C for 24 h in a dynamic vacuum environment at less than 10−5 torr. The final products were stored in an Ar-filled glove box or vacuum desiccator to prevent aging. Powder X-ray diffraction (XRD) patterns [19] with X-ray (Cu, Kα =1.5418 Ǻ) radiation in a diffractometer (Philips PW3040/60) confirmed the phase purity of the product and showed the variation of the lattice parameter as a function of V dopant. Different proportions of the dopant up to 2% were selected to understand the effect of dilute doping. Oxygen stoichiometry was expected to remain constant for all small concentrations given that the preparations were all the same.

XANES spectra were obtained from the National Synchrotron Radiation Research Center (Taiwan) and were operated at 1.5 GeV with a current of 360 mA. The K-edge of Ti and V K- edge of LTVO were measured with a wiggler beamline BL17C1 equipped with Si (111) crystal monochromators. The absorption spectra were recorded in the fluorescence-yield mode with a Lytle detector [27]. These hard X-ray absorption spectra, which could provide information on the unoccupied states with transition metal *p* state, were normalized to a unit step height in the absorption coefficient from below to above the edges. The oxide powders, namely, TiO2, VO2, Ti2O3, and V2O3, and standard metal foils were used for energy calibration and for comparing the electronic valence states. The XANES spectra at the Ti L- and O K-edges were determined at beamline BL20A in the total-electron-yield mode using an ultrahigh-vacuum chamber with pressure of ~5×10−9 torr. All spectra were normalized with the standard proce‐ dure.

The experiments for X-ray emission spectroscopy (XES) were conducted at beamline 7.0.1 at the Advanced Light Source, Lawrence Berkeley National Laboratory. The beamline is armed with a spherical grating monochromator and an undulator (period 5 cm in 99 poles) [3]. The RIXS spectra were recorded with a high-resolution grating spectrometer at grazing incidence with a 2D detector [17]. The resolution for Ti L emission spectra was about 0.4 eV. The monochromator resolution was set up similar to that of the emission measurements.

X-ray absorption at the Ti L-edge determined that the electron in the Ti 2*p* core level was excited to the unoccupied Ti 3*d* and 4*s* orbitals. XES recorded the signal from the decay process related to the partial densities of Ti 4*s* and 3*d* states. RIXS spectra were acquired by tuning different excitation energies according to the x-ray absorption spectral profile to measure XES. Sample current mode was employed to record XANES with an energy resolution of ~0.15 eV. Mean‐ while, a high-resolution grazing-incidence grating spectrometer with a 2D multi-channel plate detector was used to record XES at a resolution of 0.6 eV [36].

#### **3. Results and discussion**

#### **3.1. Structure of LTVO**

The crystal structure of cubic spinel LTO, which belongs to the Fd3m space group, is illustrated in Fig. 1(a). The structure demonstrates that the lattice parameter of LTO is 8.404 Å (a = b = c) and contains eight AB2O4 units per unit cell. In general, a spinel structure has a total of 56 ions, 32 anions, and 24 cations per unit cell. The Li, Ti, and oxide ions are located at tetrahedral (T*d*) A-sites, octahedral (O*h*) B-sites, and 32e sites, respectively. The XRD patterns of LiTi2-*x*V*x*O4 (*x* = 0, 0.05, 0.01, 0.015, and 0.02) are shown in Fig. 2(a). The result of the chemical analysis (determined with ICP-AES) is consistent with the nominal composition [19, 18, 24, 49]. These patterns reflect that the crystal structure is of pure spinel structure without an impure phase. The lattice parameter of LTVO samples displayed in Fig. 2(b) linearly decreases with the increasing concentration of the dopant. Doping V, which has a smaller ionic radius, induces a change in the lattice parameter when in a Ti site. This lattice shrinkage can be attributed to the fact that the ionic radius of V3+ (0.64 Å) is smaller than that of Ti3+ (0.67 Å) in the O*h* site [29]. The rate of shrinkage of the lattice parameter is ~3.6 Å/at% based on the occupancy of O*<sup>h</sup>* sites. This rate is also related to the bond length between the cations (Ti and V) and oxygen. Electronic structure strongly depends on the hybridization states of the transition metal 3*d*-O 2*p* orbitals nearby EF when electronic exchange normally occurs. Thus, the changed electronic structure around the transition metal ions greatly affects the oxidation state and physical properties of the latter. Considering these physical properties in AB2O4 with mixed-valence states require wide-ranging knowledge of their electronic structure, particularly the *p-d* hybridization and spin-orbital symmetry. Several experiments and theories show that AB2O4 exhibits strong electron–hole correlations because of the TM ions located in various sites [8, 39].

**Figure 1.** Crystal structure of LTO [Li (green), Ti (blue), and O (grey)].

Fig. 2. (a) Powder XRD patterns of LTVO crystals with *x* = 0, 0.005, 0.01, 0.015, and 0.02. The patterns are fitted to the Fd3m space group and are indexed. (b) Chemical composition of V **Figure 2.** (a) Powder XRD patterns of LTVO crystals with *x* = 0, 0.005, 0.01, 0.015, and 0.02. The patterns are fitted to the Fd3m space group and are indexed. (b) Chemical composition of V dependence of the lattice constant of LTVO.

#### dependence of the lattice constant of LTVO. **3.2. Resistive properties**

edge of LTVO were measured with a wiggler beamline BL17C1 equipped with Si (111) crystal monochromators. The absorption spectra were recorded in the fluorescence-yield mode with a Lytle detector [27]. These hard X-ray absorption spectra, which could provide information on the unoccupied states with transition metal *p* state, were normalized to a unit step height in the absorption coefficient from below to above the edges. The oxide powders, namely, TiO2, VO2, Ti2O3, and V2O3, and standard metal foils were used for energy calibration and for comparing the electronic valence states. The XANES spectra at the Ti L- and O K-edges were determined at beamline BL20A in the total-electron-yield mode using an ultrahigh-vacuum chamber with pressure of ~5×10−9 torr. All spectra were normalized with the standard proce‐

The experiments for X-ray emission spectroscopy (XES) were conducted at beamline 7.0.1 at the Advanced Light Source, Lawrence Berkeley National Laboratory. The beamline is armed with a spherical grating monochromator and an undulator (period 5 cm in 99 poles) [3]. The RIXS spectra were recorded with a high-resolution grating spectrometer at grazing incidence with a 2D detector [17]. The resolution for Ti L emission spectra was about 0.4 eV. The

X-ray absorption at the Ti L-edge determined that the electron in the Ti 2*p* core level was excited to the unoccupied Ti 3*d* and 4*s* orbitals. XES recorded the signal from the decay process related to the partial densities of Ti 4*s* and 3*d* states. RIXS spectra were acquired by tuning different excitation energies according to the x-ray absorption spectral profile to measure XES. Sample current mode was employed to record XANES with an energy resolution of ~0.15 eV. Mean‐ while, a high-resolution grazing-incidence grating spectrometer with a 2D multi-channel plate

The crystal structure of cubic spinel LTO, which belongs to the Fd3m space group, is illustrated in Fig. 1(a). The structure demonstrates that the lattice parameter of LTO is 8.404 Å (a = b = c) and contains eight AB2O4 units per unit cell. In general, a spinel structure has a total of 56 ions, 32 anions, and 24 cations per unit cell. The Li, Ti, and oxide ions are located at tetrahedral (T*d*) A-sites, octahedral (O*h*) B-sites, and 32e sites, respectively. The XRD patterns of LiTi2-*x*V*x*O4 (*x* = 0, 0.05, 0.01, 0.015, and 0.02) are shown in Fig. 2(a). The result of the chemical analysis (determined with ICP-AES) is consistent with the nominal composition [19, 18, 24, 49]. These patterns reflect that the crystal structure is of pure spinel structure without an impure phase. The lattice parameter of LTVO samples displayed in Fig. 2(b) linearly decreases with the increasing concentration of the dopant. Doping V, which has a smaller ionic radius, induces a change in the lattice parameter when in a Ti site. This lattice shrinkage can be attributed to the fact that the ionic radius of V3+ (0.64 Å) is smaller than that of Ti3+ (0.67 Å) in the O*h* site [29]. The rate of shrinkage of the lattice parameter is ~3.6 Å/at% based on the occupancy of O*<sup>h</sup>* sites. This rate is also related to the bond length between the cations (Ti and V) and oxygen. Electronic

monochromator resolution was set up similar to that of the emission measurements.

detector was used to record XES at a resolution of 0.6 eV [36].

**3. Results and discussion**

**3.1. Structure of LTVO**

dure.

20 Superconductors – New Developments

8 The transport properties of the samples are specified in Fig. 3(a) with V doping of less than 0.025. The findings indicate that the superconducting *T*c is suppressed with an increasing proportion of the dopant. Fig. 3(c) reveals that the resistivity of the samples gradually increases. The un-doped LTO has a superconducting *T*<sup>c</sup> of ~12 K with a transition width (△*T*c) of ~0.4 K. Fig. 3(b) indicates that when the Ti ion is replaced with V ions, *T*<sup>c</sup> exhibits an abrupt but linear suppression with respect to the level of doping. Given the magnetic pair-breaking effects [47], the doped magnetic ions suppress superconductivity. Considerable research has been devoted to exploring cuprate [48]; Tarascon et al., 1987) and pnictide [41] superconducting systems based on resistivity measurements. For the LTO series, the effect of doping with both non‐ magnetic and magnetic impurities at both T*d* and O*h* sites on (Li1*-x*A*x*)(Ti2*-y*B*y*)O4 [A = Mg and Mn; B = Al and Cr] compounds has been previously reported [26]. The substitution of non‐ magnetic ions (e.g., Mg2+ at the T*d* site and Al3+ at the O*h* site) slightly suppresses supercon‐ ductivity similar to the case when Li+ ions occupy the T*d* site. However, the effect of superconductivity suppression is greater when the magnetic Mn2+ ions occupy the T*d* site. The suppression of superconductivity is attributed to pair-breaking when the Cooper pairs are scattered with magnetic impurities. The effect of substituting Ti3+ with Cr3+ at the O*h* site is even more significant. samples. By distorting the local symmetry of the spinel structure, V substitution will not only alter the valence of Ti, but the hole/electron effect as well. This phenomenon is discussed in the next section.

Fig. 3. (a) Transport properties of LTO and LTVO samples at low temperature without magnetic field. **Figure 3.** (a) Transport properties of LTO and LTVO samples at low temperature without magnetic field. (b) Chemical composition of V-dependence of superconducting Tc. (c) Resistivity at room temperature.

10 (b) Chemical composition of V-dependence of superconducting Tc. (c) Resistivity at room temperature. The resistivity near 300 K increases as *x* increases [Fig. 3(c)], indicating the extreme sensitivity of the transport properties to the electronic effect of 3*d* transition metals. The itinerant carrier density can be attributed to the cation (Ti3+) in LTO. Similar to the reports of a previous study on Al-doped LTO [16], the results of the current study clarify the effect of carrier density on superconductivity, such that the resistivity increases with the level of doping in the normal state. Except for the semiconducting behavior at the normal state, a similar *T*c has been determined in LiTi1.7, Al0.3O4, and LTO samples [19]. This observation implies that *T*c is slightly affected by substituting a considerable proportion of the trivalent non-magnetic ions. The ratio Ti4+/Ti3+ in O*<sup>h</sup>* symmetry can then be modulated with impurity doping to vary the supercon‐ ducting properties of LTO. Electron–electron interaction may be a key factor in causing changes in the transport properties of the samples. By distorting the local symmetry of the spinel structure, V substitution will not only alter the valence of Ti, but the hole/electron effect as well. This phenomenon is discussed in the next section.

#### **3.3. Electronic structure results based on x-ray spectra**

the doped magnetic ions suppress superconductivity. Considerable research has been devoted to exploring cuprate [48]; Tarascon et al., 1987) and pnictide [41] superconducting systems based on resistivity measurements. For the LTO series, the effect of doping with both non‐ magnetic and magnetic impurities at both T*d* and O*h* sites on (Li1*-x*A*x*)(Ti2*-y*B*y*)O4 [A = Mg and Mn; B = Al and Cr] compounds has been previously reported [26]. The substitution of non‐ magnetic ions (e.g., Mg2+ at the T*d* site and Al3+ at the O*h* site) slightly suppresses supercon‐

superconductivity suppression is greater when the magnetic Mn2+ ions occupy the T*d* site. The suppression of superconductivity is attributed to pair-breaking when the Cooper pairs are scattered with magnetic impurities. The effect of substituting Ti3+ with Cr3+ at the O*h* site is even

samples. By distorting the local symmetry of the spinel structure, V substitution will not only alter the valence of Ti, but the hole/electron effect as well. This phenomenon is discussed in the next section.

10

Fig. 3. (a) Transport properties of LTO and LTVO samples at low temperature without magnetic field. (b) Chemical composition of V-dependence of superconducting Tc. (c) Resistivity at room

**Figure 3.** (a) Transport properties of LTO and LTVO samples at low temperature without magnetic field. (b) Chemical

The resistivity near 300 K increases as *x* increases [Fig. 3(c)], indicating the extreme sensitivity of the transport properties to the electronic effect of 3*d* transition metals. The itinerant carrier density can be attributed to the cation (Ti3+) in LTO. Similar to the reports of a previous study on Al-doped LTO [16], the results of the current study clarify the effect of carrier density on superconductivity, such that the resistivity increases with the level of doping in the normal state. Except for the semiconducting behavior at the normal state, a similar *T*c has been determined in LiTi1.7, Al0.3O4, and LTO samples [19]. This observation implies that *T*c is slightly affected by substituting a considerable proportion of the trivalent non-magnetic ions. The ratio Ti4+/Ti3+ in O*<sup>h</sup>* symmetry can then be modulated with impurity doping to vary the supercon‐ ducting properties of LTO. Electron–electron interaction may be a key factor in causing changes in the transport properties of the samples. By distorting the local symmetry of the spinel structure, V substitution will not only alter the valence of Ti, but the hole/electron effect

composition of V-dependence of superconducting Tc. (c) Resistivity at room temperature.

as well. This phenomenon is discussed in the next section.

ions occupy the T*d* site. However, the effect of

ductivity similar to the case when Li+

22 Superconductors – New Developments

temperature.

more significant.

With the aim to investigate the effects of V dilute doping on LTO, this research measures XANES and RIXS at Ti L-edges. As shown in the inset in Fig. 4(a), the Ti L-edge XANES spectra exhibit several well-resolved features because of the excitations of a 2*p* core electron into the Ti 3*d* empty states, that is, a transition from the ground state with configuration 2*p*<sup>6</sup> 3*dn* to an excited electronic configuration 2*p*<sup>5</sup> 3*dn*+1 with numerous multiplet excitations. When the spinorbital coupling is in the transition metal 2*p* state, the spectra reveal two prominent features located at the energy ranges of 455 eV to 461 eV and 461.2 eV to 468 eV, corresponding to the absorptions L3 (2*p*3/2 → 3*d*) and L2 (2*p*1/2 → 3*d*) edge respectively. These prominent features are due to a strong Coulombic interaction between the poorly screened Ti 3*d* electrons and the Ti 2*p* core hole [12]. *L*2 edge features are normally broadened compared with the L3 edge because of the lesser lifetime of the 2*p*1/2 core hole (i.e., a radiationless electron transition from energy level 2*p*3/2 to the 2*p*1/2), accompanied by the promotion of a valence electron into the unoccupied states (conduction band). In the O*<sup>h</sup>* crystal field, the 3*d* band splits into *t*2g (*dxy*, *dxz*, and *dyz*) and *e*g (*dx 2 -y 2* and 3*d3z2 -r 2* ) subbands [∆ = (*e*g) − (*t*2g) = 10 Dq]. Given the O*<sup>h</sup>* crystal-field splitting, the Ti L3-edge feature possesses *t*2g and *e*<sup>g</sup> bands. The RIXS Ti 3*d* spectra of LTO and doped LTO (V 2 %) are exhibited with energy-loss scales in Fig. 4(a). Four distinct spectral features are observed in RIXS spectra. These features are as follows: elastic peak at the energy loss at zero, *d–d* excitations at about 4 eV, a broad-band feature at around 7 eV, and a large energy dispersed-feature above 10 eV. Letters *a* to *k* denote the different exciting energies based on the XANES spectra, as displayed in the inset of Fig. 4(a). For example, XES spectrum *b* is collected by tuning the incident photon energy at 457 eV in XANES. The inelastic scattering features ranging from 5 eV to 10 eV originate from the complicated charge-transfer excitation from O 2*p* to Ti 3*d t*2g and *e*g subbands [1, 31]. The low-energy inelastic scattering features absent in TiO2 are observed at energy of less than 5 eV [36]. The spectra within this energy range significantly differ from that of TiO2. The appearance of Ti3+ implies that an electron occupies the empty 3*d t*2g orbitals, thus inducing the energy-loss features. Strong *d*-electron correlation is revealed because the relative intensities of peaks A4 and B4 (*d-d* excitations) are observed to be markedly different at varying excitation energies. An enhanced energy-loss peak is observed at 1 eV in spectrum *a*, and the intensity of this peak intensively drops when the excitation energy is tuned to a higher energy (peak *b* in XANES). With the increase of excitation energy at *c*, another intense peak at energy loss of about 2.9 eV is observed. An excitation energy tuned below the L3 *t*2g peak in the spectrum (letter *a*) enhances inelastic scattering A4, which is uncommon. This enhanced inelasticity is not observed at either *t*2g (*b*) or *e*<sup>g</sup> (*d*) resonant energy. This phenomenon can be explained by considering the presence of Ti3+, as revealed by the constant-initial-state absorption spectrum [3]. The spectral shapes in XANES spectra are similar for both LTO and LTVO, wherein spectral differences are slightly small. Given that the spectral change of XANES is small with V doping, the spectral deconvolution and specific electronic state can be measured separately from RIXS [11]. The result at a specific energy (*c*) in Figs. 4(b) and 4(c) indicates that an enhanced trivalent Ti contribution is observed at a dip quadrivalence within the spectrum (between *t*2g and *e*g). The Ti3+ *t*2g- and *e*g-resonance energies correspond to the excitation energies 456 and 457.5 eV respectively. [3].

**3.3 Electronic structure results based on x-ray spectra** 

Fig. 4. (a) Ti L-RIXS spectra of LTVO (*x* = 0 and 0.02) recorded with several excitation energies labeled with letters *a* to *k* in the Ti XANES L-edge spectrum. Ti L3 RIXS spectra at (b) Ti3+ t2g-resonance and at (c) Ti3+ eg-resonance. The insets of (b) and (c) display the energy diagram of the **Figure 4.** (a) Ti L-RIXS spectra of LTVO (*x* = 0 and 0.02) recorded with several excitation energies labeled with letters *a* to *k* in the Ti XANES L-edge spectrum. Ti L3 RIXS spectra at (b) Ti3+ t2g-resonance and at (c) Ti3+ eg-resonance. The insets of (b) and (c) display the energy diagram of the *d*-*d* excitation.

11 *d*-*d* excitation. With the aim to investigate the effects of V dilute doping on LTO, this research measures XANES and RIXS at Ti L-edges. As shown in the inset in Fig. 4(a), the Ti L-edge XANES spectra exhibit several well-resolved features because of the excitations of a 2*p* core electron into the Ti 3*d* empty states, that is, a transition from the ground state with configuration 2*p*<sup>6</sup> 3*d<sup>n</sup>* to an excited electronic configuration 2*p*<sup>5</sup> 3*dn*+1 with numerous multiplet excitations. When the spin-orbital coupling is in the transition metal 2*p* state, the spectra reveal two prominent features located at the energy ranges of 455 eV to 461 eV and 461.2 eV to 468 eV, corresponding to the absorptions L3 (2*p*3/2 3*d*) and L2 (2*p*1/2 3*d*) edge respectively. These prominent features are due to a strong Coulombic interaction between the poorly When excitation energy is tuned to the *t*2g resonance, the low-energy-excited feature C4 at ~1 eV demonstrated in Figs. 4(b) and 4(c) is resonantly enhanced. This feature corresponds to the electron-hole pairs within the *t*2g band. When the spectrum is acquired with energy at about *e*<sup>g</sup> resonance energy [Fig. 4(c)], the intensity of peak D4 increases. Such an increase can be attributed to the transition between the occupied *t*2g and the unoccupied *e*g states. The presence of this RIXS feature indicates a strong electron–electron association. The *d-d* excitation peak at an energy loss of about 2.9 eV corresponds to the crystal-field splitting (i.e., 10 Dq) and refers to the ground state without the core hole. The comparison of the spectra of each set shows that the intensity of the loss feature decreases with the increasing concentration of dopant (*x* = 0.02). Hence, fewer electrons are distributed in the *t*2g band. In a previous study on TiO2 [31], no obvious feature was exhibited in the region located below the elastic signal (peak). The change in intensity is due to the variation of the incompletely filled *t*2g band arising from the V doped effect. This result forcefully indicates that the Ti valence is enhanced when V is doped. The active Ti ions in LTO have a formal oxidation state trivalence and quadrivalence, which display nearly 0.5 electrons in the 3*d* orbital, and possess a lesser electronic density. Doping with V reduces the number of Ti 3*d* electrons, suggesting an increase in the formal oxidation number of the Ti ion. The Ti ions show a mixed-valence state and contain some *t*2g electrons in LTO. With no *t*2g electron, *d*-*d* excitation is absent because no electron subsists in *t*2g to be excited to the *t*2g/*e*g unoccupied state. A little variation in valence of Ti is then reflected in the RIXS spectra. The diminution or absence of features in the energy range from 0 eV to 5 eV indicates a decrease in *d*–*d* transitions. More electrons occupy the *t*2g orbital when the oxidation number of Ti changes from trivalence to quadrivalence. The probability of exciting the *t*2g electron into the unoccupied states (*t*2g or *e*g) is therefore increased. Only RIXS can observe the *d*-*d* excitations (*t*2g-*t*2g and *t*2g-*e*g) in the spectra. The significance of observing this *d*-*d* excitation is that the *t*2g occupation number changes with V doping, whereas the number of *t*2g electrons decreases with V doping. Therefore, the significant RIXS spectrum can observe this kind of valence change.

Fig. 5. (a) XANES Ti L3,2-edge spectra for LTVO (x = 0.005 to 0.02), LTO (*x* = 0), Ti2O3, and anatase TiO2. (b) Integrated area ratio of A5/B5. This ratio increase implies the presence of 3*d*<sup>0</sup> states. (c) Total **Figure 5.** (a) XANES Ti L3,2-edge spectra for LTVO (x = 0.005 to 0.02), LTO (*x* = 0), Ti2O3, and anatase TiO2. (b) Integrat‐ ed area ratio of A5/B5. This ratio increase implies the presence of 3*d*<sup>0</sup> states. (c) Total area of L3-edge (A5+B5) decreases with the increase of V concentration

14 area of L3-edge (A5+B5) decreases with the increase of V concentration Complementary information is also acquired for this research by thoroughly analyzing the XANES Ti L3,2-edge spectra. The spectra of LiTi2-*x*V*x*O4 (or LTVO) (*x* = 0 to 0.02) are presented in Fig. 5(a). The white line shapes of the spectra of LTO (undoped) and slight V doping (*x* = 0.005) in LTO are similar. The A5 and C5 features (B5 and D5) are previously assigned as the *t*2g (*e*g) states of the 10 Dq crystal-field-split in 3*d* orbitals. This O*h* crystal-field splitting is ~1.8 eV at LTO, but increases from 1.8 eV to 2.0 eV when the V concentration increases from *x* = 0.005 to *x* = 0.02. This phenomenon reveals the distortion of O*h* symmetry when Ti is replaced with slight V doping. This finding is consistent with the powder XRD results and the theoretical calculations (De Groot et al., 1990, Finkelstein et al., 2002; De Groot et al. 1993). Moreover, the results specify that B5 peak broadens with its gradual V doping. As previously reported (De Groot et al., 1990, Finkelstein et al., 2002; De Complementary information is also acquired for this research by thoroughly analyzing the XANES Ti L3,2-edge spectra. The spectra of LiTi2-*x*V*x*O4 (or LTVO) (*x* = 0 to 0.02) are presented in Fig. 5(a). The white line shapes of the spectra of LTO (undoped) and slight V doping (*x* = 0.005) in LTO are similar. The A5 and C5 features (B5 and D5) are previously assigned as the *t*2g (*e*g) states of the 10 Dq crystal-field-split in 3*d* orbitals. This O*<sup>h</sup>* crystal-field splitting is ~1.8 eV at LTO, but increases from 1.8 eV to 2.0 eV when the V concentration increases from *x* = 0.005 to *x* = 0.02. This phenomenon reveals the distortion of O*h* symmetry when Ti is replaced with slight V doping. This finding is consistent with the powder XRD results and the theoretical calculations [12, 13, 15]. Moreover, the results specify that B5 peak broadens with its gradual V doping. As previously reported [12, 13, 15], and according to the Jahn-Teller distortion with ∆*E*J-T energy splitting in the *e*g band, the *e*g orbitals point directly toward the Ti 2*p* orbitals of the octahedrally coordinated O atoms. In general, *e***g** band is sensitive to the local environment, producing a changed bonding distance and a O–Ti–O angle in the presence of V doping. These *e*g-related peaks are also broader than the *t*2g peak because of the larger hybridization between *e*g orbitals and O ligand states, and because of the associated effects of solid-state broadening [12]. The *e*g feature appears as a narrow and symmetric profile in LTO, representing symmetric octahedrally coordinated Ti–O bonds. This feature becomes broad and asymmetric upon doping, suggesting a distortion. In particular, distortion may arise from the uneven Ti-O bonds in the O*<sup>h</sup>* symmetry. The *e*<sup>g</sup> peak at the side with greater energy originates from the short Ti-O bonds because of the increase in hybridization (relative to the long Ti-O bonds); hence, the intensity ratio of the high-energy to low-energy *e*g peaks is increased with V doping. This result implies that long Ti–O bonds become shorter [17]. An enhancement in the integrated area under A5 peak implies the increase in unoccupied states of Ti 3*d* states, thereby indicating that the Ti valence increases with slight V doping. From the Ti–O bond, the electronic configuration of Ti exhibits a combination of 3*d*<sup>0</sup> (*t*2g0 *e*g 0 ) and 3*d*<sup>1</sup> (*t*2g1 *e*0 ) in the ground state. Fig. 5(b) shows

11

**3.3 Electronic structure results based on x-ray spectra** 

**<sup>455</sup> <sup>460</sup> <sup>465</sup> <sup>470</sup> Energy(eV)**

**Intensity (a. u.)**

(a) L-RIXS

**Intensity (arb. units)**

24 Superconductors – New Developments

*d*-*d* excitation.

of (b) and (c) display the energy diagram of the *d*-*d* excitation.

2*p*<sup>5</sup>

a b c d e f gh i <sup>j</sup> <sup>k</sup> LiTi2-xVxO4 L-XAS

**-20 -15 -10 -5 0 Energy Loss(eV)**

is, a transition from the ground state with configuration 2*p*<sup>6</sup>

**Intensity (a. u.)**

**Intensity (a. u.)**

Fig. 4. (a) Ti L-RIXS spectra of LTVO (*x* = 0 and 0.02) recorded with several excitation energies labeled with letters *a* to *k* in the Ti XANES L-edge spectrum. Ti L3 RIXS spectra at (b) Ti3+ t2g-resonance and at (c) Ti3+ eg-resonance. The insets of (b) and (c) display the energy diagram of the

**Figure 4.** (a) Ti L-RIXS spectra of LTVO (*x* = 0 and 0.02) recorded with several excitation energies labeled with letters *a* to *k* in the Ti XANES L-edge spectrum. Ti L3 RIXS spectra at (b) Ti3+ t2g-resonance and at (c) Ti3+ eg-resonance. The insets

When excitation energy is tuned to the *t*2g resonance, the low-energy-excited feature C4 at ~1 eV demonstrated in Figs. 4(b) and 4(c) is resonantly enhanced. This feature corresponds to the electron-hole pairs within the *t*2g band. When the spectrum is acquired with energy at about *e*<sup>g</sup> resonance energy [Fig. 4(c)], the intensity of peak D4 increases. Such an increase can be attributed to the transition between the occupied *t*2g and the unoccupied *e*g states. The presence of this RIXS feature indicates a strong electron–electron association. The *d-d* excitation peak at an energy loss of about 2.9 eV corresponds to the crystal-field splitting (i.e., 10 Dq) and refers to the ground state without the core hole. The comparison of the spectra of each set shows that the intensity of the loss feature decreases with the increasing concentration of dopant (*x* = 0.02). Hence, fewer electrons are distributed in the *t*2g band. In a previous study on TiO2 [31], no obvious feature was exhibited in the region located below the elastic signal (peak). The change in intensity is due to the variation of the incompletely filled *t*2g band arising from the V doped effect. This result forcefully indicates that the Ti valence is enhanced when V is doped. The active Ti ions in LTO have a formal oxidation state trivalence and quadrivalence, which display nearly 0.5 electrons in the 3*d* orbital, and possess a lesser electronic density. Doping with V reduces the number of Ti 3*d* electrons, suggesting an increase in the formal oxidation number of the Ti ion. The Ti ions show a mixed-valence state and contain some *t*2g electrons in LTO. With no *t*2g electron, *d*-*d* excitation is absent because no electron subsists in *t*2g to be excited to the *t*2g/*e*g unoccupied state. A little variation in valence of Ti is then reflected in the RIXS spectra. The diminution or absence of features in the energy range from 0 eV to 5 eV indicates a decrease in *d*–*d* transitions. More electrons occupy the *t*2g orbital when the oxidation number of Ti changes from trivalence to quadrivalence. The probability of exciting the *t*2g electron into the unoccupied states (*t*2g or *e*g) is therefore increased. Only RIXS can observe the *d*-*d* excitations (*t*2g-*t*2g and *t*2g-*e*g) in the spectra. The significance of observing this *d*-*d* excitation is that the *t*2g occupation number changes with V doping, whereas the number of *t*2g electrons decreases with V doping. Therefore, the significant RIXS spectrum can observe this kind of valence change.

With the aim to investigate the effects of V dilute doping on LTO, this research measures XANES and RIXS at Ti L-edges. As shown in the inset in Fig. 4(a), the Ti L-edge XANES spectra exhibit several well-resolved features because of the excitations of a 2*p* core electron into the Ti 3*d* empty states, that

3*dn*+1 with numerous multiplet excitations. When the spin-orbital coupling is in the transition metal 2*p* state, the spectra reveal two prominent features located at the energy ranges of 455 eV to 461 eV and 461.2 eV to 468 eV, corresponding to the absorptions L3 (2*p*3/2 3*d*) and L2 (2*p*1/2 3*d*) edge respectively. These prominent features are due to a strong Coulombic interaction between the poorly

3*d<sup>n</sup>*

(a) (b) (c) (d) (e) (f) (g) (h) (i) (j) (k)

<sup>B</sup> A4 <sup>4</sup>

x=0.02 x=0

x=0 x=0.02

x=0.02 x=0



to an excited electronic configuration

(b) Ti3+ t2g-resonance

(c) Ti3+ eg-resonance

C4

C4 D4

the integrated area of ratio A5/B5, which clearly increases with V doping. Consequently, the more intense A5 (*t*2g) feature implies a higher oxidation number and exhibits the presence of Ti4+ (3*d*<sup>0</sup> ) [28, 40]. Moreover, the area of the L3-edge that corresponds to (A5+B5) progressively decreases when V concentration increases [Fig. 5(c)] because V replaces Ti. This trend is in good agreement with the XRD results [18]. A small but actual increase in Ti valence with slight V doping is observed in XANES Ti L3,2-edge spectra. Such an increase is significantly obvious in the Ti L-RIXS spectra.

The pre-edge features of Ti K-edge XANES spectra are displayed in Fig. 6(a). These spectral features at K-edge are due to the transitions from the Ti 1*s* core level to the 4*p*-derived final states based on the dipole selection rule. The pre-edge features are a combination of strongly hybridized Ti 4*sp* and 3*d* and O 2*p* orbitals. Quadrupole-allowed transitions generally occur at the pre-edge region in the transition metal oxides, which correspond to the contribution from 3*d* orbitals through 4*sp*-3*d* hybridization [14, 25, 37]. The inset of Fig. 6(a) reveals that a Gaussian function is subtracted from the original Ti K-edge spectrum for a detailed comparison of the pre-edge spectra. Fig. 4(a) shows the Ti K-edge spectra of LTO and LTVO (0.005 to 0.02). As marked by the black arrow in Fig. 6(a), the spectra of LTVO with a small concentration of *x* from 0.005 to 0.01 are similar to that of LTO with the same photon energy at the main peak A6. The Ti valence performs to maintain +3.5 owing to the smaller concentration of V doping. Nevertheless, the intensity of the pre-peak spectrum increases relative to that of pure LTO as the doping level is increased to *x* = 0.015 and *x* **=** 0.02. Meanwhile, as directed by the red arrow in Fig. 6(a), chemical shift is also observed as V concentration increases. Fig. 6(b) displays the pre-edge region between 4969 and 4977 eV, as well as the three main features (i.e., C6, D6, and E6) in XANES Ti K-edge spectra of TiO2 (anatase, Ti4+) and LTVO (*x* = 0, 0.02) samples. The origin of the splitting of pre-peaks is caused by local excitations (1*s* to 3*d t*2g and *e*g) [5, 14, 38, 46]. Conversely, the greatest contribution to this splitting is suggested to be the corner- and edge-sharing Ti octahedra that yield non-local, intersite hybrid excitations. The next-nearest neighbor 3*d* states (*t*2g and *e*g) are related Ti 4*p* states that absorb atom via the anion O 2*p* states [7, 23, 44]. The local structure of Ti for LTO and LTVO also possesses O*<sup>h</sup>* symmetry; hence, the pre-peak region (A6 and B6) can be described in terms of a similar scenario [23, 44, 7]. The intensity of these pre-peaks varies with V concentration. In the V-doped LTO, the substitution of V for Ti slightly decreases the lattice parameter [18]. A reduced bond length (Ti–O–Ti) subsequently increases the overlap of the first-nearest-neighbors and the absorbing atom (e.g., Ti 4*p*-Ti 3*d* orbitals) mediated by O ion. Therefore, these pre-peaks are closely related to the increase in the number of the first-nearest-neighbor unoccupying the Ti 3*d* states. These intersite hybrid peaks intensities increase with the increase of V concentration. The inset of Fig. 6(b) shows that the variation in the area under the pre-peak suggests that the 3*d* unoccu‐ pied states are altered via the interaction of the Ti 4*p*-O 2*p*-Ti 3*d* states. These results strengthen the conclusion that Ti is in mixed-valence states between LTO and LTVO [14, 24]. The analysis of the spectra of LTO and LTVO (*x* = 0.02) reveals that pre-peaks A6 and B6 are stronger in LTVO than in LTO, attended by a chemical shift of the main peak to higher energy. Therefore, Ti valence is increased when V is doped. The average valence of Ti in the case of doped LTVO (*x* = 0.015 and 0.02) is approximately +3.6, which is obtained with a simple calculation and a fit of the ratio of areas under the spectral lines.

The O K-edge XANES spectra [Fig. 6(c)] provide useful information about the unoccupied density of states in TMO because of the covalent mixing between O and TM ions. The two preedge features are at about 530 eV to 536 eV because of the strong hybridizations in Ti 3*d*–O 2*p* states, and correspond to the *t*2g and *e*g states of Ti 3*d* orbitals respectively. The intensity of the *t*2g and *e*g peaks at the O K-edge changes is similar to an increased peak ratio *t*2g/*e*g in Ti Ledge XANES. In sum, the above results specified by the XANES and RIXS spectra strongly confirm that the valence of Ti is increased when the doping level of *x* increases. Thus, the Ti-O hybridization and electron–electron correlation are modified when a slight doping of V ions completely suppresses the superconductivity of LTO.

the integrated area of ratio A5/B5, which clearly increases with V doping. Consequently, the more intense A5 (*t*2g) feature implies a higher oxidation number and exhibits the presence of

decreases when V concentration increases [Fig. 5(c)] because V replaces Ti. This trend is in good agreement with the XRD results [18]. A small but actual increase in Ti valence with slight V doping is observed in XANES Ti L3,2-edge spectra. Such an increase is significantly obvious

The pre-edge features of Ti K-edge XANES spectra are displayed in Fig. 6(a). These spectral features at K-edge are due to the transitions from the Ti 1*s* core level to the 4*p*-derived final states based on the dipole selection rule. The pre-edge features are a combination of strongly hybridized Ti 4*sp* and 3*d* and O 2*p* orbitals. Quadrupole-allowed transitions generally occur at the pre-edge region in the transition metal oxides, which correspond to the contribution from 3*d* orbitals through 4*sp*-3*d* hybridization [14, 25, 37]. The inset of Fig. 6(a) reveals that a Gaussian function is subtracted from the original Ti K-edge spectrum for a detailed comparison of the pre-edge spectra. Fig. 4(a) shows the Ti K-edge spectra of LTO and LTVO (0.005 to 0.02). As marked by the black arrow in Fig. 6(a), the spectra of LTVO with a small concentration of *x* from 0.005 to 0.01 are similar to that of LTO with the same photon energy at the main peak A6. The Ti valence performs to maintain +3.5 owing to the smaller concentration of V doping. Nevertheless, the intensity of the pre-peak spectrum increases relative to that of pure LTO as the doping level is increased to *x* = 0.015 and *x* **=** 0.02. Meanwhile, as directed by the red arrow in Fig. 6(a), chemical shift is also observed as V concentration increases. Fig. 6(b) displays the pre-edge region between 4969 and 4977 eV, as well as the three main features (i.e., C6, D6, and E6) in XANES Ti K-edge spectra of TiO2 (anatase, Ti4+) and LTVO (*x* = 0, 0.02) samples. The origin of the splitting of pre-peaks is caused by local excitations (1*s* to 3*d t*2g and *e*g) [5, 14, 38, 46]. Conversely, the greatest contribution to this splitting is suggested to be the corner- and edge-sharing Ti octahedra that yield non-local, intersite hybrid excitations. The next-nearest neighbor 3*d* states (*t*2g and *e*g) are related Ti 4*p* states that absorb atom via the anion O 2*p* states [7, 23, 44]. The local structure of Ti for LTO and LTVO also possesses O*<sup>h</sup>* symmetry; hence, the pre-peak region (A6 and B6) can be described in terms of a similar scenario [23, 44, 7]. The intensity of these pre-peaks varies with V concentration. In the V-doped LTO, the substitution of V for Ti slightly decreases the lattice parameter [18]. A reduced bond length (Ti–O–Ti) subsequently increases the overlap of the first-nearest-neighbors and the absorbing atom (e.g., Ti 4*p*-Ti 3*d* orbitals) mediated by O ion. Therefore, these pre-peaks are closely related to the increase in the number of the first-nearest-neighbor unoccupying the Ti 3*d* states. These intersite hybrid peaks intensities increase with the increase of V concentration. The inset of Fig. 6(b) shows that the variation in the area under the pre-peak suggests that the 3*d* unoccu‐ pied states are altered via the interaction of the Ti 4*p*-O 2*p*-Ti 3*d* states. These results strengthen the conclusion that Ti is in mixed-valence states between LTO and LTVO [14, 24]. The analysis of the spectra of LTO and LTVO (*x* = 0.02) reveals that pre-peaks A6 and B6 are stronger in LTVO than in LTO, attended by a chemical shift of the main peak to higher energy. Therefore, Ti valence is increased when V is doped. The average valence of Ti in the case of doped LTVO (*x* = 0.015 and 0.02) is approximately +3.6, which is obtained with a simple calculation and a

) [28, 40]. Moreover, the area of the L3-edge that corresponds to (A5+B5) progressively

Ti4+ (3*d*<sup>0</sup>

in the Ti L-RIXS spectra.

26 Superconductors – New Developments

fit of the ratio of areas under the spectral lines.

Previous studies have indicated that LTO is a spinel superconductor, whereas spinel LVO reveals a heavy fermionic behavior with a Curie-Weiss spin susceptibility and a large electronic specific heat. Conduction arises on a Ti sublattice via the *t*2g orbitals with a narrow bandwidth (2 eV to 3 eV), displaying possible strong electronic correlations. The origin of the physics and electronic properties of the spinel structure are yet to be resolved. Our current research demonstrates the role of 3*d* electrons in the properties of LTO and LTVO. The full solid solution of LiTi2−*<sup>x</sup>*V*x*O4 (0 ≤ *x* ≤ 2) has been investigated [24]. The replacement of V ions for Ti site modifies the bandgap and *d* electron/hole exchange in Ti energy levels in order to maintain electrical neutrality. Several mechanisms for understanding the quickly suppressed superconductivity with a slight doping of LTVO have been proposed. The slight doping of LTVO may incur two major electronic effects. The first one is a simple pair-breaking effect with electron spin *S* = 1 on superconductivity [20, 24]. The second one refers to the effect of carrier doping on electron correlation via band filling. The 3*d* electrons of V are assumed to hybridize the Ti conduction electrons, making them itinerant. A simple magnetic pair-breaking of electron spins induced by additional impurity is commonly assumed to be responsible for suppressing superconduc‐ tivity based on the localized moments of 3*d* electrons of V [24]. The localized magnetic moment per V atom is estimated to be 1.7 μB in the V4+ state with *S* = 1/2 [24]. The XANES measurements at V K-edges plotted in Fig. 7 show that V is more likely to be in state 3+ (Bordage et al., 2110), which contains a *d*<sup>2</sup> electronic configuration and is supposed to offer a weak magnetic moment probably because of the antiparallel orientation of the electron spins. The decreased V magnetic moment is also obvious in LiZnV2O4 [35]. A second possible reason for a weak magnetic moment is the decreased charge carriers at Ti *t*2g bands. In this case, V electrons are expected to be localized at the Ti conduction electrons. If the Ti ions in LTO possess oxidation numbers 3+ and 4+, then ~0.5 (*t*2g) electrons are expected to be in the conduction band. This condition results in the observation of a *d*-*d* excitation within the energy range of 0 eV to 5 eV owing to the presence of electrons in the *t*2g occupied state that are excited to either the *t*2g or *e*g unoccu‐ pied state. The number of electrons occupying the *t*2g orbital is expected to decrease if the oxidation number of Ti increases from trivalence to quadrivalence because of the doping of V ions. The spectra in RIXS reveal a significant variation in the *d*-*d* excitation feature, which suggestively decreases when LTO becomes LTVO, even with slight doping of V ions. This situation shows the decreased Ti 3*d* occupied states with the increase of *x* concentration. The Ti *t*2g occupation number is then altered upon V doping. This valence variation is small in XANES Ti K- or L-edges spectra, but weighty in RIXS spectra. The results, particularly the reflection of the decreased intensities of *t*2g-*t*2g, and *t*2g-*e*<sup>g</sup> transitions from the RIXS spectra,

eg)-O 2*p* states.

support the role of these electronic states in rapidly suppressing *T*<sup>c</sup> in slight V doping. Accordingly, the findings of this research support the conclusion that the decreased density of the states of Ti 3*d* electrons at *E*F is responsible for rapidly suppressing superconductivity.

Fig. 6. (a) XANES Ti K-edge spectra of LTVO (*x* = 0 to 0.2) and Anatase-TiO2 in the pre-edge region. The inset shows that a Gaussian function was subtracted from the original TiO2 spectrum. (b) Detailed comparison of the pre-edge region for LTO and LTVO (*x* = 0.02). (c) XANES O K-edge spectra **Figure 6.** (a) XANES Ti K-edge spectra of LTVO (*x* = 0 to 0.2) and Anatase-TiO2 in the pre-edge region. The inset shows that a Gaussian function was subtracted from the original TiO2 spectrum. (b) Detailed comparison of the pre-edge re‐ gion for LTO and LTVO (*x* = 0.02). (c) XANES O K-edge spectra shown with two main peaks in the pre-edge region corresponding to the hybridization of Ti 3*d* (t2g and eg)-O 2*p* states.

shown with two main peaks in the pre-edge region corresponding to the hybridization of Ti 3*d* (t2g and

Fig. 7. XANES V K-edge of LTVO (*x* = 0.02) compared with the larger doping of LTVO (*x* = 0.05 **Figure 7.** XANES V K-edge of LTVO (*x* = 0.02) compared with the larger doping of LTVO (*x* = 0.05 and 0.1). No signifi‐ cant variation is observed in the spectra

and 0.1). No significant variation is observed in the spectra

20

#### **4. Conclusion**

support the role of these electronic states in rapidly suppressing *T*<sup>c</sup> in slight V doping. Accordingly, the findings of this research support the conclusion that the decreased density of the states of Ti 3*d* electrons at *E*F is responsible for rapidly suppressing superconductivity.

19

5460 5470 5480 5490 5500 5510

Fig. 7. XANES V K-edge of LTVO (*x* = 0.02) compared with the larger doping of LTVO (*x* = 0.05

**Figure 7.** XANES V K-edge of LTVO (*x* = 0.02) compared with the larger doping of LTVO (*x* = 0.05 and 0.1). No signifi‐

**Photon Energy (eV)**

20

Fig. 6. (a) XANES Ti K-edge spectra of LTVO (*x* = 0 to 0.2) and Anatase-TiO2 in the pre-edge region. The inset shows that a Gaussian function was subtracted from the original TiO2 spectrum. (b) Detailed comparison of the pre-edge region for LTO and LTVO (*x* = 0.02). (c) XANES O K-edge spectra shown with two main peaks in the pre-edge region corresponding to the hybridization of Ti 3*d* (t2g and

**Figure 6.** (a) XANES Ti K-edge spectra of LTVO (*x* = 0 to 0.2) and Anatase-TiO2 in the pre-edge region. The inset shows that a Gaussian function was subtracted from the original TiO2 spectrum. (b) Detailed comparison of the pre-edge re‐ gion for LTO and LTVO (*x* = 0.02). (c) XANES O K-edge spectra shown with two main peaks in the pre-edge region

eg)-O 2*p* states.

28 Superconductors – New Developments

**Normalized Absorption (Arb. Units)**

cant variation is observed in the spectra

corresponding to the hybridization of Ti 3*d* (t2g and eg)-O 2*p* states.

*x***=0.1**

V K-edge

**LiTi2-***<sup>x</sup>* **V***x* **O4**

*x***=0.05**

*x***=0.02**

and 0.1). No significant variation is observed in the spectra

The results on the effect of V doping on the atomic and electronic structures of LTVO, XAS, and RIXS spectra reveal that the valence of V ions remains constant, but the hybridization of Ti-O considerably varies. In particular, the results show the mixed valency nature of Ti ions and a significant variation in the hybridization of Ti 3d-O 2p states. The sharp decrease in the superconductivity of V-doped samples is ascribed to the electron–orbital interaction arising from the hybridization of TM and O orbitals. Low-energy excitation because of d-d excitation indicates that electron correlation and Ti oxidation number are enhanced, which support the XAS observation. Meanwhile, the observed properties of superconductivity are attributed to an altered density of the states of Ti 3d electrons and Ti-O hybridization. The Ti electronic configuration and lattice distortion induce the rapidly suppressed superconductivity and are attributed to hybridization rather than to the magnetic nature of the substituted ion. This research also demonstrates that RIXS is a powerful tool for investigating the electronic states and electron correlations of Ti compounds, in which the XANES spectral features are subdued.

#### **Author details**

Chi-Liang Chen1,2 and Chung-Li Dong1,2

1 Institute of Physics, Academia Sinica, Nankang, Taipei, Taiwan

2 National Synchrotron Radiation Research Center (NSRRC), Hsinchu, Taiwan

#### **References**


[17] Harada, Y.; Kinugasa, T.; Eguchi, R.; Matsubara, M.; Kotani, A.; Watanabe, M.; Ya‐ gishita, A. & Shin, S. (2000). Polarization dependence of soft-x-ray Raman scattering at the L edge of TiO2. *Physical Review B*, 61: 12854.

[5] Brydson, R.; Sauer, H.; Engel, W.; Thomass, J. M.; Zeitler, E.; Kosugi, N. & Kuroda, H. (1989). Electron energy loss and x-ray absorption spectroscopy of rutile and ana‐

[6] Bussmann-Holder; Müller, K. A. & Keller, H. (2007). High *T*<sup>c</sup> Superconductors and

[7] Cabaret, D; Bordage, A; Juhin, A; Arfaoui, M. & Gaudry, E. (2010). First-principles calculations of x-ray absorption spectra at the K-edge of 3*d* transition metals: an elec‐ tronic structure analysis of the pre-edge. *Phys. Physical Chemistry Chemical Physics*, 12:

[8] Cava, R. J.; Murphy, D. W.; Zahurak, S.; Santoro, A. & Roth, R. S. (1984). The crystal structures of the lithium-inserted metal oxides Li0.5TiO2 anatase, LiTi2O4 spinel, and

[9] Chen, C. L. & Dong, C. L. (20012). SUPERCONDUCTOR-Materials, Properties and Applications, 2012 ISBN: 979-953-307-798-6 Chapter 2 'X-ray spectroscopy studies of

[10] Chen, C. L.; Yeh, K. W., Huang, D. J., Hsu, F. C.; Lee, Y. C., Huang, S. W., Guo, G. Y.; Lin, H. J.; Rao, S. M. & Wu, M. K. (2008). Orbital polarization of the unoccupied

[11] Dallera, C.; Grioni, M.; Shukla, A.; Vankó, G.; Sarrao, J. L.; Rueff, J. P. & Cox, D. L. (2002). New Spectroscopy Solves an Old Puzzle: The Kondo Scale in Heavy Fer‐

[12] De Groot, F. M. F.; Fuggle, J. C.; Thole, B. T. & Sawatzky, G. A. (1990). 2*p* x-ray ab‐ sorption of 3d transition-metal compounds: A atomic multiplet descriptiom includ‐ ing the crystal field. *Physical Review B*, 42: 5459. (1990). L2,3 x-ray-absorption edges of

[13] De Groot, F. M. F.; Faber, J. C.; Michile, J. J. M.; Czyzyk, M. T.; Abbate, M. & Fuggle, J. C. (1993). Oxygen 1*s* x-ray absorption of tetravalent titanium oxides: A comparison

[14] Durmeyer, O.; Kappler, J. P.; Beaurepaire, E.; Heintz, J. M. & Drillon, M. (1990). TiK XANES in superconducting LiTi2O4 and related compounds. *Journal of Physics: Con‐*

[15] Finkelstein, L. D.; Zabolotzky, E. I.; Korotin, M. A.; Shamin, S. N.; Butorin, S. M.; Kur‐ maev, E. Z. & Nordgren, J. (2002). Vacant states of TiO2 with rutile structure and their reflection in different-type x-ray absorption spectra. *X-ray Spectrometry*, 31: 414-418.

[16] Fazileh, F.; Gooding, R. J.; Atkinson, W. A. & Johnston, D. C. (2006). Role of Strong Electronic Correlations in the Metal-To-Insulator Transitionin Disordered LiAlyTi2-

, Ca2+, Sc3+, and Ti4+ in Oh (octahedral) symmetry. *Physical Review B*,

tase: a test of structural sensitivity. *Journal of Physics: Condensed Matter*, 1: 797.

Related Transition Metal Oxides ISBN 3540710221.

Li2Ti2O4. *Journal of Solid State Chemistry*, 53: 64-75.

states in multiferroic LiCu2O2. *Physical Review B*, 78: 214105.

with single-particle calculations. *Physical Review B*, 48: 2074

the iron chalcogenides', 21-44.

mions. *Physical Review Letters*, 88: 196403.

5619-5633.

30 Superconductors – New Developments

*d*0

41: 928.

compounds: K+

*densed Matter*, 2: 6127

yO4. *Physical Review Letters*, 96: 046410.


lineshapes in titanium oxides: Influence of valence-band profile and of Coster-Kronig processes. *Physical Review B*, 69: 155421.


[40] Richter, J. H.; Henningsson, A.; Sanyal, B.; Karlsson, P. G.; Andersson, M. P.; Uvdal, P.; Siegbahn, H.; Eriksson, O. & Sandell, A. (2005). Phase separation and charge local‐ ization in UHV-lithiated anatase TiO2 nanoparticles. *Physical Review B*, 71: 235419.

lineshapes in titanium oxides: Influence of valence-band profile and of Coster-Kronig

[29] Lide, D. R. (2007-2008) CRC handbook of Chemistry and Physics, 88th edition, Taylor

[30] Maddox, B. R.; Lazicki, A.; Yoo, C. S.; Iota V.; Chen, M.; McMahan, A. K.; Hu, M. Y.; Chow, P.; Scalettar, R. T. & Pickett, W. E. (2006). 4*f* delocalization in Gd: Inelastic x-

[31] Matsubara, M.; Uozumi, T.; Kotani, A.; Harada, Y. & Shin, S. (2002). Polarization de‐ pendence of resonant x-ray emission spectra in 3*d*n transition metal compounds with

[32] Massidda, S.; Yu, J. & Freeman, A. J. (1988). Electronic structure and properties of su‐

[33] Merz, M.; Nücker, M.; Schweiss, P.; Schuppler, S.; Chen, C. T.; Chakarian, V.; Free‐ land, J.; Idzerda, Y. U,; Kläser M.; Müller-Vogt, M. & Wolf, Th. (1998). Site-Specific X-Ray Absorption Spectroscopy of Y1−xCaxBa2Cu3O7−y: Overdoping and role of apical oxygen for high temperature superconductivity. *Physical Review Letters*, 80: 5192. [34] Millis, A. J. (1998). Lattice effects in magnetoresistive manganese perovskites. *Nature*,

[35] Muhtar; Takagi, F.; Kawakami, K. & Tsuda, N. (1988). Magnetic Susceptibility of Metal-Insulator System LixZn1-xV2O4. *Journal of the Physical Society of Japan*, 57:

[36] Nordgren, J.; Bray, G.; Cramm, S.; Nyholm, R.; Rubensson, J. E. & Wassdahl, N. (1989). Soft x‐ray emission spectroscopy using monochromatized synchrotron radia‐

[37] Poumellec, B.; Marucco, J. F. & Touzelin, B. (1987). X-ray-absorption near-edge struc‐ ture of titanium and vanadium in (Ti,V)O2 rutile solid solutions. *Physical Review B*,

[38] Parlebas, J. C.; Khan, M. A.; Uozumi, T.; Okada, K. & Kotani, A. (1995). Theory of many-body effects in valence, core-level and isochromat spectroscopies along the 3*d* transition metal series of oxides. *Journal of Electron Spectroscopy and Related Phenom‐*

[39] Ra, W.; Nakayama, M.; Uchimoto, Y. & Wakihara, M. (2005). Experimental and com‐ putational study of the electronic structural changes in LiTi2O4 Spinel Compounds upon Electrochemical Li insertion reactions. *The Journal of Physical Chemisity B*, 109(3):

ray scattering at ultrahigh pressure. *Physical Review Letters*, 96: 215701.

n = 0, 1, 2, 3. *Journal of the Physical Society of Japan*, 71: 347-356.

perconducting LiTi2O4. *Physical Review B*, 38: 11352.

tion (invited). *Review of Scientific Instruments*, 60: 1690.

processes. *Physical Review B*, 69: 155421.

& Francis Group.

32 Superconductors – New Developments

392: 147.

3119-3127.

35: 2284.

*ena*, 71: 117.

1130-1134.

