Novel electronic structures in transition metal oxides

Abstract

In this chapter, we investigate the detailed electronic structure of Ca1?xSrxVO3\text{Ca}_{1-x}\text{Sr}_x\text{VO}_3Ca1?xSrxVO3, with x=0.0,0.2,0.7,x = 0.0, 0.2, 0.7,x=0.0,0.2,0.7, and 1.01.01.0, by means of various electron spectroscopic techniques as well as band structure calculations. We observe that the experimental spectra exhibit extensive modifications with the change in excitation energy, arising from the different surface and bulk electronic structure in this system. The origin of such spectral changes is shown to arise from a charge disproportionation at the surface due to the stronger electron correlation strength. This observation underlines the importance of separating the surface electronic structure from the experimentally observed spectra, before a suitable comparison can be made between theoretical results describing the bulk properties and the experimental spectra. We devise a novel algorithm to separate the surface and the bulk components of the spectra using experimental spectra with various photon energies. Thus, obtained spectral functions represent the true surface and bulk electronic structures independent of the techniques involved. The surface electronic structure exhibits an insulator-to-metal transition with an increase in xxx. The bulk electronic structures correspond to the metallic phase for all the compositions with a large coherent feature, as predicted by theory. The reduction of U/WU/WU/W with the increase in WWW in this series is manifested as a transfer of spectral weight from the incoherent to coherent feature. The bulk photoemission spectra of this series firmly establish the efficacy of the apparently simple Hubbard model treated within the dynamical mean field theory using local self-energies in the limit of infinite dimensions, to describe even quantitatively the properties of such systems. The Ca1?xSrxVO3\text{Ca}_{1-x}\text{Sr}_x\text{VO}_3Ca1?xSrxVO3 series was ideally suited for this goal. Apart from this excellent agreement of the experimental observations and the existing theoretical model, an important open question still lies ahead and possibly within our reach: to find out whether the standing conflict between the theoretical and experimental results on the systematic evolution of doped Mott-Hubbard systems can also be resolved within the existing paradigms of electronic structure theories and this newly devised experimental technique to obtain excitation energy independent spectral functions. This problem will be dealt with in Chapter 4. LaVO? is an antiferromagnetic insulator with an optical gap about 1.1 eV. This gap is much larger than the activated gap (~0.1 eV) obtained in the resistivity measurements. The magnetic measurements exhibit antiferromagnetic order below 125 K (T?). Doping of hole states in LaVO? introduces an antiferromagnetic insulator to a Pauli paramagnetic metal transition at x?0.2x \approx 0.2x?0.2. The compositions with x<0.2x < 0.2x<0.2 exhibit VRH in the transport measurements at low temperatures, suggesting a dominant role of disorder in the Mott-Anderson transition. While the composition close to the MI transition exhibit disordered Fermi liquid behavior, the compositions away from the MI transition show a dependence of resistivity. The enhancement factor for the susceptibility does not change significantly across the series. Ab initio band structure calculation converges to a metallic ground state in LaVO?. This suggests the dominant role of correlation effects in driving the insulating phase of this system. The excitation spectra obtained from these calculations provide a good description of the spectral function away from the Fermi level. The analysis of the valence band spectra in conjunction with the band structure results establishes that LaVO? is a Mott-Hubbard insulator. The band gap obtained from the electron removal and electron addition spectra is about 1 ± 0.2 eV, in agreement with the measured optical gap (1.1 eV). The photoelectron spectroscopic results for the doped compounds reveal the surface electronic structure to be qualitatively different from the bulk one for the metallic compositions, the surface layer being insulating despite extensive doping of charge carriers. This realization of a surface modification in the electronic structure of metallic compositions is essential in order to critically discuss and evaluate the experimental electronic structure in terms of the existing many-body theories and various bulk-sensitive low-energy properties. We have extracted the bulk-related spectra from the total spectra using photon energy-dependent measurements. It is established that all known theoretical results based on the single-band Hubbard model as well as the present multiband model, are clearly insufficient to explain the experimentally observed bulk spectral functions for any value of U/WU/WU/W. It is then remarkable and significant that a simple additive description of the spectral functions of the end members, LaVO? and CaVO?, though empirical and speculative, is qualitatively and quantitatively successful for all the compositions without using any adjustable parameters and therefore deserves serious consideration in terms of rigorous microscopic theories for such systems. This conclusion is further supported by the core level spectra as well as the Bi spectra. The V 2p spectra exhibit distinct signatures of two features in the main signal arising from the t2gt_{2g}t2g and ege_geg sites. The intensity of these features scales linearly with the composition of the compounds studied. A direct consequence of these observations would be that the electronic structure of such a system is intrinsically inhomogeneous, arising from the effect of disorder and cannot be described within a homogeneous model, such as the Hubbard model. This is in contrast to the usual practice in recent times to interpret the properties of doped Mott insulators in terms of the Hubbard model. In this context, it is important to note that considerations of disorder within a single-band Hubbard-like model provide a similar description for the evolution of the spectral function with doping in the presence of extensive inhomogeneity in the compositions over a length scale considerably larger than the atomic length scale. While it is reasonable to expect this description of electronic structure to be valid even in the multiband case, it is not clear at present what would be the implications of this theoretical model concerning the transport and magnetic properties. In conclusion, we have presented high-energy spectroscopic results for La1?xSrxVO3\text{La}_{1-x}\text{Sr}_x\text{VO}_3La1?xSrxVO3 with x=0.0,0.1,0.2,0.3,x = 0.0, 0.1, 0.2, 0.3,x=0.0,0.1,0.2,0.3, and 0.5. The core level spectrum of LaCrO? has been analyzed in terms of the configuration interaction model, and the parameter strengths controlling the electronic structure of this system have been estimated. The valence band spectra corresponding to both occupied and unoccupied states have been analyzed with the help of ab initio band structure calculations; these spectra suggest that the intrinsic gap in this material is about 3.0 eV. We have shown that the oxygen KKK-edge X-ray absorption spectrum arising from the Cr³? state in LaCrO? can be interpreted in terms of transitions into states with a1ga_{1g}a1g, t2gt_{2g}t2g, and ege_geg symmetries on the basis of band structure results. The analysis of the Bi and X-ray absorption spectra suggests a crystal field splitting to be about 2.1 eV and an intra-atomic exchange interaction strength of 0.7 eV in LaCrO?. Substitution of Sr²? in place of La³? dopes holes localized primarily at the Cr sites, generating Cr?? states arising from the Mott-Hubbard insulating ground state of the parent oxide. This is clearly evidenced by the systematic changes in the photoemission, Bi, as well as in the Cr X-ray absorption spectra. These changes have been interpreted in terms of the formation of local Cr?? states. The doping of hole states introduces a new spectral feature with t2gt_{2g}t2g character within the Mott-Hubbard gap with increasing Sr content. Such a localized Cr?? state is expected to be accompanied by a local lattice contraction and the consequent formation of a small polaron. The lowest energy charge excitation in such a system will consist of transferring one electron from a Cr³? site with t2g3eg2t_{2g}^3 e_g^2t2g3eg2 to a nearby Cr?? site with t2g2eg3t_{2g}^2 e_g^3t2g2eg3 configuration. This will effectively interchange the Cr³? and Cr?? sites, corresponding to a hopping of the small polaron. Thus, it appears that the insulating nature of the doped LaCrO? compounds arises from the stability of the Cr³? ion with its half-filled narrow t2gt_{2g}t2g configuration as well as from strong electron-phonon interactions that favor the formation of small polarons. Energy scale over which the resistivity measurements were carried out (< 300 K). In such a situation, the low-temperature (< 300 K) resistivity can be dominated by a very low density of states, below the level of photoemission sensitivity, if such states are present in the vicinity of the Fermi energy. Besides the existence of such low density of states across EFE_FEF, influencing the low-temperature transport properties, the electronic structure of the doped systems is clearly dominated by the formation of the mid-gap doped hole states in agreement with the results obtained by Dagotto et al. [53] based on the dynamical hole approach. However, the prediction of a finite uuu precursor of Drude peaks in the optical conductivity comprising the hole states delocalized over a sizable, but finite length scale is in contrast to the observation of a large distinct gap (~0.6 eV) in the doped compounds. These observations suggest that disorder-induced Anderson localization or the electron correlation effects alone are not sufficient to explain the changes in the electronic structure of the doped compounds. Strong correlation effects, which drive the gap formation, will have to be taken into account simultaneously with the disorder effects. In this context, it is interesting to note that a recent treatment of the simultaneous presence of disorder and electron-electron interaction has suggested [54] characteristic signatures of disorder in terms of localized states across EFE_FEF and an overall gap-like structure driven by correlation. 6.4 Conclusions We have investigated the transport and magnetic properties of the Haldane system, Y2BaNiO5Y_2BaNiO_5Y2BaNiO5, and the hole-doped compositions via substitution of Ca²? by Sr²?. The electronic transport in this system is found to be dominated by variable range hopping of the localized carriers at the Fermi level. The magnetic susceptibility as a function of temperature appears to be consistent with the presence of a spin excitation gap (~18.3 meV) in this system. Interestingly, the doped compounds also show the presence of a finite gap up to 40% hole doping. However, the magnitude of the gap reduces to about 12.5 meV and 4.4 meV with doping, indicating the appearance of in-gap states as observed in the neutron diffraction studies. The band structure calculations show a one-dimensional character of the electronic structure of Y2BaNiO5Y_2BaNiO_5Y2BaNiO5. The non-magnetic calculations exhibit substantial density of states (DOS) at EFE_FEF, suggesting a metallic ground state. The antiferromagnetic ordering along the chain direction depletes the DOS at EFE_FEF; however, it still exhibits a finite DOS at EFE_FEF, in contrast to the insulating property, suggesting a dominant role of electron-electron interactions in the electronic structure of these compounds. We observed that the ground state with antiferromagnetic arrangements of the Ni²? sublattice has a lower energy compared to that of the non-magnetic solution. The calculated magnetic moment in this system is found to be smaller compared to the expected value for the S = 1 state of Ni²?. We have also investigated the electronic structure of Y2BaNiO5Y_2BaNiO_5Y2BaNiO5 and the effect of hole doping by various electron spectroscopic techniques. Y2BaNiO5Y_2BaNiO_5Y2BaNiO5 is found to be a correlation-driven insulator with a charge excitation gap of about 2.3 eV. The doping reduces the overall bandwidth slightly in the occupied part of the electronic structure. More pronounced changes are observed in the spectra corresponding to the unoccupied part of the electronic structure. Doped hole states are formed in the charge gap of the parent compound; the intensity of this feature grows systematically with anomalous spectral weight transfer from the Hubbard bands, similar to that observed in the higher-dimensional divalent nickelates (NiO and La?NiO?). The doped compounds also exhibit a distinct gap of about 0.6 eV, though the transport measurements exhibit variable range hopping instead of an activated behavior. The presence of a finite gap for all the compositions, along with the evidence of low, but finite density of states from resistivity measurements, suggests an interplay of disorder and correlation effects to be important in determining the electronic structure of these compounds. In conclusion, we have investigated the electronic structure of the one-dimensional cuprates Ca2CuO3\text{Ca}_2\text{CuO}_3Ca2CuO3 and Sr2CuO3\text{Sr}_2\text{CuO}_3Sr2CuO3 using band structure calculations and various high-energy spectroscopies. The dispersions of the valence band exhibit an essentially one-dimensional character of the charge carriers. The density of states (DOS) and partial density of states (PDOS) obtained in Ca2CuO3\text{Ca}_2\text{CuO}_3Ca2CuO3 and Sr2CuO3\text{Sr}_2\text{CuO}_3Sr2CuO3 suggest a small value of (cd?Cp)(cd - Cp)(cd?Cp) compared to that in higher-dimensional cuprates. LDA calculations converge to the correct ground state magnetic structure and the magnetic moments, though they do not result in an energy gap for Ca2CuO3\text{Ca}_2\text{CuO}_3Ca2CuO3 and provide only a small gap in Sr2CuO3\text{Sr}_2\text{CuO}_3Sr2CuO3 (? 18 meV). Thus, the experimentally observed large band gaps clearly suggest the importance of electron correlations in determining the electronic structure. While band structure results cannot explain the existence of the satellite features appearing at high binding energies in the valence band photoemission spectra, comparisons between the experimental and calculated spectra allow us to interpret the origins of all other spectral features in the valence and conduction band regions. We estimate different electron interaction strengths from the analysis of the spectra in terms of model many-body calculations. It is found that the charge-transfer energy, AAA, is generally smaller in these compounds compared to other cuprates, such as La2CuO4\text{La}_2\text{CuO}_4La2CuO4 and CuO\text{CuO}CuO, though all other interaction parameters are similar. One possible reason for such a change in AAA as a function of dimensionality may be related to the reduced Madelung potential in these systems. Such reduced values of AAA in one-dimensional cuprates suggest the possibility of an unusual electronic structure with the bare upper Hubbard band within the oxygen 2p bandwidth. This is particularly so for Sr2CuO3\text{Sr}_2\text{CuO}_3Sr2CuO3, which has an even lower AAA than Ca2CuO3\text{Ca}_2\text{CuO}_3Ca2CuO3. This indicates that such low-dimensional systems can provide rare examples of correlated covalent insulators, which have been predicted theoretically a long time ago [75]. Doping of hole states in these systems enhances the local character of the charge carriers. The doped hole states are possibly localized due to the shrinkage of the CuO cluster unit, leading to the formation of small polarons. The doping in Ca2CuO3\text{Ca}_2\text{CuO}_3Ca2CuO3 exhibits the transfer of spectral weight as usually observed in the strongly correlated systems. Interestingly, Sr2CuO3\text{Sr}_2\text{CuO}_3Sr2CuO3 provides a contrasting scenario, transferring the spectral weight from the low energy to the higher energy regions. The present study provides the first example of the evolution of the spectral functions as a function of doping in a correlated covalent insulator, which is unusual in its character compared to the observation in any strongly correlated system. On the atomic potentials, it is influenced considerably by the crystal lattice potential. In order to investigate the role of the Madelung potential on the charge transfer energy, we have calculated the Madelung potential for each of the cases studied here. The charge transfer energy is related to the difference in Madelung potentials at Ni and oxygen sites (?VNi-O\Delta V_{\text{Ni-O}}?VNi-O) [77]. Following Ref. [80], ?VNi-O\Delta V_{\text{Ni-O}}?VNi-O can be computed for a given lattice under the assumption of fully ionized atomic sites (ionic limit) in the lattice. This way, we obtain the value of ?Vm\Delta V_{\text{m}}?Vm for 3-dimensional NiO, to be 48.36 eV. The values of ?Vm\Delta V_{\text{m}}?Vm in 2-dimensional La2NiO4\text{La}_2\text{NiO}_4La2NiO4 range between 48.62 and 49.48 eV arising from nonequivalent oxygen sites in this compound. ?Vm\Delta V_{\text{m}}?Vm decreases to 48.14 - 49.08 eV for 1-dimensional Y2BaNiO5\text{Y}_2\text{BaNiO}_5Y2BaNiO5 and to 46.14 - 47.09 eV for 0-dimensional Lu2BaNiO5\text{Lu}_2\text{BaNiO}_5Lu2BaNiO5. Thus, the overall decreasing trend in AAA with decreasing dimensionality does not follow the trend obtained in ?Vm\Delta V_{\text{m}}?Vm in these compounds, though the 0-dimensional Lu2BaNiO5\text{Lu}_2\text{BaNiO}_5Lu2BaNiO5 with its lowest value of AAA also has the lowest ?Vm\Delta V_{\text{m}}?Vm among the compounds studied here. It is important to note here that the values of ?Vm\Delta V_{\text{m}}?Vm are calculated assuming the ionic configurations of all the elements in the solid. However, the effective charge state of each atom in the solid is generally very different from that in the ionic limit, and unfortunately, there is no reliable way to obtain the correct charge state in the solid. The present study within the cluster approximation shows that the average Ni 3d occupancy, nNin_{\text{Ni}}nNi, increases with the decrease in dimensionality of the crystal structure, thereby decreasing the charge state. This is then expected to reduce the Madelung potential from the value calculated within the ionic limit; this reduction will be more pronounced in the lower dimension due to the increasing trend in nNin_{\text{Ni}}nNi. Thus, it appears that the changes in AAA between the various compounds are likely driven by subtle changes in the Madelung potential arising from the changing dimensionality of the crystal structure, though it is not possible to simulate these changes within a simple ionic model. ________________________________________ 8.6 Conclusions In conclusion, we provide a detailed study of the electronic structure of divalent nickelates with different Ni-O-Ni networks. The electronic structure of these compounds exhibits substantial changes with the change in the dimensionality of Ni-O-Ni connectivity. We analyzed the electronic spectra within the cluster approximation, evaluating both Ni 3d and O 2p contributions on the same footing, thereby providing a quantitative description of the electronic structure in each case. It is found that the changes in the electronic structure are primarily due to a systematic variation in the charge transfer energy (AAA). Since the effective coupling between the Ni sites determining the 3d-bandwidth is controlled by Ni-O-Ni hoppings, a reduction of the dimensionality is expected to reduce the bandwidth; on the other hand, a concomitant reduction in AAA enhances the effective coupling between the Ni sites, tending to increase the 3d-bandwidth. The change in AAA possibly arises from changes in (cd?cp)(c_d - c_p)(cd?cp), the bare energy difference due to changes in the Madelung potential; however, calculations of Madelung potentials within the ionic model are not sufficient to describe the variation in AAA. Another important consequence of decreasing AAA is to increase the average 3d-occupancy, nNin_{\text{Ni}}nNi, with decreasing dimensionality in these divalent nickelates. This change in AAA is also responsible for changing the character of the satellite feature in the core-level photoemission from a primarily poorly-screened d8d^8d8 state in NiO to a dominantly overscreened d9d^9d9 state in the compounds with lower-dimensional Ni-O-Ni networks. The results clearly establish a systematic undermining of the influence of the non-local screening channel on the spectral shape with decreasing Ni-O-Ni connectivity.