High-resolution Mn K-edge x-ray emission and absorption spectroscopy study of the electronic and local structure of the three different phases in Nd0.5Sr0.5MnO3

PHYSICAL REVIEW B 93, 205108 (2016) High-resolution Mn K-edge x-ray emission and absorption spectroscopy study of the electronic and local structure of the three different phases in Nd0.5Sr0.5MnO3 S. Lafuerza,1,*J. Garc´ ıa,2G. Sub´ ıas,2J. Blasco,2and P. Glatzel1 1ESRF-The European Synchrotron, 71, Avenue des Martyrs, Grenoble, France 2Instituto de Ciencia de Materiales de Arag´ on, Departamento de F´ ısica de la Materia Condensada, CSIC-Universidad de Zaragoza, C/ Pedro Cerbuna 12, 50009 Zaragoza, Spain (Received 4 March 2016; revised manuscript received 11 April 2016; published 9 May 2016) Nd0.5Sr0.5MnO3is particularly representative of mixed-valent manganites since their three characteristic macroscopic phases (charge-ordered insulator, ferromagnetic-metallic, and paramagnetic insulator) appear at different temperatures. We here report a complete x-ray emission and absorption spectroscopy (XES-XAS) study of Nd0.5Sr0.5MnO3as a function of temperature to investigate the electronic and local structure changes of the Mn atom in these three phases. Compared with the differences in the XES-XAS spectra between Nd0.5Sr0.5MnO3 and the single-valent reference compounds NdMnO3(Mn3+)andSr/CaMnO 3(Mn4+), only modest changes have been obtained across the magnetoelectrical transitions. The XES spectra, including both the Mn Kαand Kβ emission lines, have mainly shown a subtle decrease in the local spin density accompanying the passage to the ferromagnetic-metallic phase. For the same phase, the small intensity variations in the pre-edge region of the high-resolution XAS spectra reflect an increase of the p-dmixing. The analysis of these XAS spectra imply a charge segregation between the two different Mn sites far from one electron, being in intermediate valences Mn+3.5±δ/2(δ<0.2e−) for all the phases. Our results indicate that the spin, charge, and geometrical structure of the Mn atom hardly change among the three macroscopic phases demonstrating the strong competition between the ferromagnetic conductor and the charge-ordered insulator behaviors in the manganites. DOI: 10.1103/PhysRevB.93.205108 I. INTRODUCTION Mixed-valent manganese perovskites of the type RE1−xAxMnO3,RE 1−xA1+xMnO4, and RE2−2xA1+2xMn2O7 (RE =rare-earth, A=alkaline-earth), hereafter referred to as manganites, have recently been the subject of an extensive and deep study due to the occurrence of exotic phenomena like colossal magnetoresistance or so-called charge ordering (CO) [1–4]. Despite this broad interest, a complete and comprehensive understanding of the changes in the Mn local electronic structure along with the complex behavior exhibited by these manganites is lacking. Depending on the formal valence state of the Mn atom, they show different magnetic and electrical phases (including ferromagnetic-metal or antiferromagneticinsulator) and also phases identified as charge-orbital ordered. The classical description of this behavior in terms of ionic Mn3+/Mn4+configurations which includes the double exchange mechanism has been challenged by both experimental results and electronic structure calculations [5–11]. Among them, half-doped manganites with general formula (RE0.5A0.5)MnO3present many intriguing features such as real space ordering of charge, spontaneous phase separation, field-induced antiferromagnetic-insulating to ferromagneticmetallic transition resulting in a giant change in resistivity, etc. [4]COofMn 3+/Mn4+and orbital ordering was proposed as the fingerprint of the charge exchange antiferromagneticinsulating low temperature phase [12–20]. The conventional model used to describe this low-temperature phase is a onedimensional zigzag chain of Mn atoms in the ab plane coupled antiferromagnetically to each other with a checkerboard ordering of the Jahn-Teller distorted Mn3+and isotropic Mn4+ *[email protected] ions. In particular, Nd0.5Sr0.5MnO3is very representative since the three macroscopic phases of half-doped manganites (charge ordered insulator, ferromagnetic conductor, and paramagnetic insulator) appear at different temperatures [21]. Above the ferromagnetic ordering temperature TC=255 K it is a paramagnetic insulator, in the range 255 K ⩾T⩾150 K it is a magnetoresistive ferromagnetic metal, and below the antiferromagnetic ordering temperature TN=150 K it shows CO. In short, the proposed mechanism of the phase transitions has been explained as follows: In the paramagnetic phase, the doped holes are randomly localized at the Mn sites being the fluctuation time among the sites larger than the permanence time at each site. Then the ferromagnetic-metal phase is assumed to appear due to the double exchange mechanism and the doped holes are highly delocalized and spread out over the Mn atoms. Finally, the insulator CO phase is described as the spatial localization of the doped holes with a periodic ordering of Mn3+and Mn4+ions. X-ray absorption measurements at the Mn Kedge in the La1−xCaxMnO3series have demonstrated the absence of integer ionic states of the Mn atoms [6,8,9,22–25]. In addition, results of resonant x-ray scattering (RXS) at the Mn Kedge in half-doped manganites have shown that the electronic disproportionation between the different sites in the checkerboard Mn ordering is of the order of 0.2 electrons [5,26–30]. On the other hand, extended x-ray absorption fine structure (EXAFS) data in La1−x(Ca,Sr)xMnO3compounds have proven that the local structure around the Mn atom is distorted in both the charge-ordered and paramagnetic phases, whereas the ferromagnetic-metallic phase appears concomitantly with a collapse of this distribution into a symmetric environment of the nearest-neighboring O atoms [6,31–33]. We here present a complete study of the local electronic and geometrical structure of the Mn atom in the three different 2469-9950/2016/93(20)/205108(11) 205108-1 ©2016 American Physical Society S. LAFUERZA et al. PHYSICAL REVIEW B 93, 205108 (2016) phases of Nd0.5Sr0.5MnO3by high-resolution x-ray emission and absorption spectroscopy (XES-XAS) [34–37]. The detection of the emitted x-rays after Kedge photon absorption as a function of both incident and emitted energies makes the high-resolution XES-XAS techniques unique to detect small electronic and/or geometrical differences among the different phases. Like XAS, XES is also sensitive to the oxidation state of the Mn atoms but it has the advantage that it is less dependent on the ligand environment. We have recorded XES at both the Mn Kαand Kβlines. The Kαlines are the strongest fluorescence lines and arise from 2pto 1stransitions; while about 8 times weaker are the core-to-core (CTC) Kβlines, which result from 3p→1stransitions; and again 50–100 times weaker are the valence-to-core (VTC) Kβlines which are transitions from the valence shell to the 1sstate [35]. We also performed highenergy resolution fluorescence-detected (HERFD) absorption spectroscopy by setting the emission energy to the maxima of the main fluorescence lines and recording the intensity as a function of the energy of the incident photons across the Mn K edge. For the sake of comparison, the same measurements were carried out in parallel on the related single-valent compounds: NdMnO3(Mn3+), SrMnO3(Mn4+), and CaMnO3(Mn4+). NdMnO3and CaMnO3undergo antiferromagnetic transitions at a TNof about 78 K [38] and 125 K [39], respectively, whereas cubic SrMnO3adopts an antiferromagnetic structure between 230 and 260 K depending on the oxygen stoichiometry of the samples [40–41]. We note that because of the interference between the Mn Kedge and the Nd L2edge, for Nd0.5Sr0.5MnO3the EXAFS signal cannot be obtained by means of standard XAS. Interestingly, we have been able to work out the EXAFS spectra of this compound by using the simultaneously recorded total fluorescence and high-resolution data measured at the maximum of the Kα1emission line. In this way, the temperature dependence of the EXAFS signal across the three different phases of Nd0.5Sr0.5MnO3has also been characterized. Our results show that the electronic state of the Mn atom in Nd0.5Sr0.5MnO3is intermediate between the single-valent compounds and changes very little in the three phases: paramagnetic-insulator, ferromagnetic-metal, and antiferromagnetic charge-ordered. The similarity of the Mn electronic and local structural state at the different magnetoelectrical phases explains the strong competition between the ferromagnetic-metallic and the CO phases in this compound and by extension in the manganites. II. EXPERIMENTAL Polycrystalline samples of Nd0.5Sr0.5MnO3,NdMnO 3, SrMnO3, and CaMnO3were synthesized using a sol-gel method by the citrate route. Details about the synthesis are given in Refs. [42,43]. The obtained specimens were single phase as confirmed by conventional x-ray powder diffraction. Magnetic measurements were also performed between 5 and 300 K using a commercial superconducting quantum interference device (SQUID) magnetometer. According to the magnetic susceptibility measurements, Nd0.5Sr0.5MnO3shows two phase transitions at TC≈255 K and at TN≈150 K [5]. The experiment was performed at ID26 beamline of the European Synchrotron Radiation Facility (ESRF) in Grenoble (France). The incident energy was tuned through the Mn K edge by means of a pair of cryogenically cooled Si(311) monochromator crystals. Rejection of higher harmonics was achieved by three Si mirrors working under total reflection (2.5 mrad). A reference Mn metallic foil was used to calibrate the monochromator energy by setting the first inflection point of the Mn Kedge at 6539 eV. The inelastically scattered photons were analyzed using a set of four spherically bent Ge(333) and Ge(440) crystals for the Mn Kαand Kβlines, respectively. The analyzer crystals were arranged with the sample and detector (avalanche photodiode) in a vertical Rowland geometry (R≈1 m). The total experimental broadening, determined as the average of the full width at half maximum (FWHM) of the elastic profiles, was about 0.4 eV (Mn Kα) and 0.7 eV (Mn Kβ). Nonresonant Mn Kαand KβXES scans were recorded at incident energy of 6700 eV. HERFD-XAS spectra at the Mn Kedge were measured at various emission energies while recording simultaneously the total fluorescence yield (TFY) with a diode. X-ray absorption near edge structure (XANES) spectra were collected for all the samples at the maximum of the Mn Kα1emission line while EXAFS spectra were only measured in Nd0.5Sr0.5MnO3. Spinselective XANES spectra were measured at the maximum oftheMnKβ1,3emission line for each sample and also at the maximum of the Mn Kβshoulder. A continuous He-flow cryostat was used for the temperature-dependent measurements. The data were collected on concentrated pellets after finding no difference in the XANES spectra as compared with 5 −10% in mass diluted pellets. Self-absorption effects can be discarded as almost no difference is found between XANES spectra recorded in TFY mode and those measured in transmission mode with optimal thickness to get a jump at the edge of about 1. The applied data treatment procedures are described in the following. The XES spectra have been normalized with respect to the area using the range 5876–5908 eV for the Mn KαXES and the range 6461–6506 eV for the Mn KβXES. The Mn Kα XES spectra have been fitted with two Lorentzian functions to determine the FWHM of the Kα1line [44]. The Mn Kβ VTC region spectra have been further treated by performing a background subtraction in order to get rid of the contribution from the CTC main line tail. For that, we have modelled the background using seven Voigt functions that have been fitted to several data points below (6492–6515 eV) and above (6540– 6565 eV) the VTC features following the procedure described in Ref. [45]. Regarding the x-ray absorption data, the XANES spectra have been normalized to the high-energy part well above the absorption edge (about 100 eV) [46]. The Fourier transforms (FTs) of the EXAFS signals were calculated for akrange of 1.5–12 ˚ A−1usingasinewindow.TheEXAFS structural analysis has been performed using theoretical phases and amplitudes calculated by the FEFF-6 code [47] and fits to the experimental data were carried out in R space with the ARTEMIS programme of the Demeter package [48]. III. RESULTS A. Mn Kαand KβXES spectra Figure 1compares the Mn KαXES spectra of Nd0.5Sr0.5MnO3,NdMnO 3(Mn3+), SrMnO3(Mn4+), and 205108-2 HIGH-RESOLUTION Mn K-EDGE X-RAY EMISSION . . . PHYSICAL REVIEW B 93, 205108 (2016) FIG. 1. Normalized Mn KαXES spectra of Nd0.5Sr0.5MnO3, NdMnO3,SrMnO 3,andCaMnO 3measured at 80 K. Inset: Kα1 line full width at half maximum (FWHM) from the same data as a function of formal valence. The red line is a linear fit considering only NdMnO3and SrMnO3. CaMnO3(Mn4+) measured at 80 K. These spectra consist of two peaks, Kα1and Kα2, which correspond to 2p3/2and 2p1/2final states, respectively. We note that the maximum of the Kα1peak is shifted to higher energies by about 0.2 eV for the Mn3+compound compared to the Mn4+samples while the spectrum of mixed-valent Nd0.5Sr0.5MnO3lies closer to that of NdMnO3. Clearly, the KαXES of Nd0.5Sr0.5MnO3 cannot be described by a 1:1 superposition of the spectra of NdMnO3and SrMnO3. We have fitted the spectrum using the formula σNd0.5Sr0.5MnO3=(1 −x)σNdMnO3+xσSrMnO3and we have found the best fit for the addition 0.8σNdMnO3+ 0.2σSrMnO3. Therefore, the weight needed in the spectra is not simply equal to the doping concentration. As reported in previous works [35,44,49], the Kα1FWHM can be related to the spin state. The inset of Fig. 1shows the evolution of the Kα1line FWHM for the four compounds plotted versus formal valence. We note that the value corresponding to Nd0.5Sr0.5MnO3falls out from the linear dependence between these two parameters considering the single-valent compounds and appears nearer to the one of NdMnO3. We continue with the Mn CTC KβXES spectra of the same samples measured at room temperature (RT), which are shown in Fig. 2. In this case, the spectral features arise from a 3p→1sdecay process and are separated into the strong Kβ1,3 peak and a broad Kβshoulder at lower emitted energy. The presence of the Kβsatellite and the energy splitting between the Kβ1,3and Kβlines are determined by the intra-atomic exchange interaction between the spin of the 3phole and the 3delectrons. The strong Kβ1,3peak (low intense Kβshoulder) is mainly arising from the coupling between the 3phole, which is antiparallel (parallel) to the total spin of the 3delectrons. In this way, compounds with the same spin state show very similar spectra as observed for our samples with formal Mn4+(3d3) ionic states. Nevertheless, systematic changes with Mn valence when increasing from 3+to 4+are observed. The Mn valence FIG. 2. Normalized Mn CTC KβXES spectra of Nd0.5Sr0.5MnO3, NdMnO3,SrMnO 3,andCaMnO 3measured at RT. Inset: Formal spin values derived from the IAD analysis of the same data as described in the text plotted versus formal valence. Error bars have been estimated by comparison with the data points of the two binary oxides and the solid line is the linear dependence considering the latter points. increase is reflected in the shift of the Kβ1,3maximum to lower energy (∼0.4 eV) and strong broadening on its low-energy side, and the intensity loss of the Kβshoulder. On the other hand, the spectrum of Nd0.5Sr0.5MnO3lies in between the Mn3+and Mn4+groups but closer to the spectrum of the Mn3+ sample as found in the KαXES spectra. In this case, the best agreement is found for the addition 0.7σNdMnO3+0.3σSrMnO3. Tyson et al. [8] also reported a nonlinear dependence between the mix weights and the doping concentration in the Mn CTC KβXES data of the related system La1−xCaxMnO3for x< 0.4, being the weights for x=0.3 such that 0.9σLaMnO3+ 0.1σCaMnO3. Besides, we have analyzed the Mn CTC KβXES spectra using the integrated absolute difference (IAD) method [49] in order to retrieve the evolution of the Mn local spin moment both across the different compounds and as a function of temperature. This method allows quantitatively obtaining the relative change in the local 3dspin magnetic moment by integrating the absolute value of the difference between sample spectrum and a reference spectrum. In this analysis, we have also included the data of the binary oxides Mn2O3(Mn3+) and MnO2(Mn4+) and the latter compound has been used as reference. In order to obtain a linear relationship between IAD and spin (S), we have assumed an ionic picture for MnO2and Mn2O3and used their formal spin values which are S=3/2 and S=2, respectively since Mn is in high-spin configuration. The inset of Fig. 2shows the formal spin obtained in this way plotted versus formal valence for all the compounds. The formal spin state of Mn in Nd0.5Sr0.5MnO3is intermediate between the single-valent samples but it deviates from the corresponding average spin state in agreement with the 0.7:0.3 ratio for the NdMnO3/SrMnO3weights previously obtained. In Fig. 3are shown the spectra of Nd0.5Sr0.5MnO3measured at temperatures corresponding to the three phases (300 K, 205108-3 S. LAFUERZA et al. PHYSICAL REVIEW B 93, 205108 (2016) 6470 6480 6490 6500 0 20 40 60 80 100 120 140 Nd 0.5 Sr 0.5 MnO 3 [T=300K]-[T=200K] [T=200K]-[T=80K] Emitted Energy (eV) Norm. Intensity (arb. units) -6 -4 -2 0 2 4 6 K K ' T = 300 K T = 200 K T = 80 K Difference (arb. units) FIG. 3. Comparison between the normalized Mn CTC KβXES spectra of Nd0.5Sr0.5MnO3measured at 300 K, 200 K, and 80 K (left axis) and differences between the same spectra (right axis). 200 K, and 80 K). Despite the significant changes in the macroscopic properties, these spectra are almost identical (see also the corresponding IAD values in the inset of Fig. 2), indicating that the electronic states at the Mn atom are very similar for the three different phases. The difference spectra of Nd0.5Sr0.5MnO3formed by subtracting the data at two temperatures are also displayed. The differences [T=300 K]-[T= 200 K] and [T=200 K]-[T=80 K] reflect the changes that take place when entering the ferromagnetic-metallic and antiferromagnetic charge-ordered phases, respectively. We find small but clearly perceptible changes and the signs of the difference signals are reversed. No changes are observed in the difference [T=300 K]-[T=80 K] and the spectra of NdMnO3, SrMnO3, and CaMnO3measured at 80 K are also alike the data at 300 K (not shown). Similarly to Qian et al. [50] and HerreroMart´ ın et al. [51], we have also evaluated the difference signal resulting from a change in the Mn formal valence state by doing σSr/CaMnO3–σNdMnO3,σNd0.5Sr0.5MnO3–σNdMnO3, and σMnO2–σMn2O3. However, the spectral changes induced by a Mn formal valence increase do not resemble those in the differences [T=300 K]-[T=200 K] and [T=200 K]-[T= 80 K]. The observed thermal behavior could thus be interpreted as a change in the local spin density between the ferromagnetic phase and the paramagnetic and antiferromagnetic phases. In order to discern whether this change corresponds to either an increase or decrease of the local spin density, we have computed the first moment (E=j(EjIj)/jIj)ofthe Kβ1,3peak in the spectra at the different temperatures using the area 6485–6495 eV. While this method is less robust to detect small changes than the IAD values, it has the advantage that it maintains the sign of the shift (which is lost in the IAD analysis). We obtain Eequal to 6489.81 and 6489.82 eV for 300 K and 80 K, respectively, and E=6489.80 eV for 200 K. Since the spin state and the first moment of the Kβ1,3peak shift in a linear fashion as shown previously for other Mn compounds [35], we can then conclude that our Mn CTC KβXES spectra of Nd0.5Sr0.5MnO3as a function of temperature reflect a subtle decrease of the local spin density in the ferromagnetic-metallic phase as expected from the charge delocalization. FIG. 4. Normalized Mn VTC KβXES spectra of Nd0.5Sr0.5MnO3, NdMnO3,SrMnO 3,andCaMnO 3measured at RT after background removal. While the CTC Kβemission lines are sensitive to the local magnetic moment, VTC Kβemission lines mainly probe the occupied metal p-density of states (DOS) up to 25 eV below the Fermi energy. The x-ray emission process is dominated by dipole transitions and therefore the spectral features correspond to valence band states with metal psymmetry. Since the metal p-DOS is strongly mixed with ligand orbitals, the intensities of the two observed features, Kβ shoulder and Kβ2,5line, reflect the amount of hybridization between the metal poccupied electronic levels and ligand sand porbitals, respectively [45]. Figure 4shows the VTC KβXES spectra of all the samples taken at RT after background removal. The compounds with higher Mn formal valence state, Sr/CaMnO3, show the strongest Kβ and Kβ2,5features, indicating a larger hybridization as expected from the shorter Mn-O distance (the average Mn-O distance is 2.05 ˚ AforNdMnO 3[52] while for Sr/CaMnO3is 1.90 ˚ A[31,53,54], in agreement with a larger ionic radii for Mn3+compared with Mn4+). As occurred in the case of the Mn Kαand CTC Kβemission lines, the spectrum of Nd0.5Sr0.5MnO3is very similar to that of NdMnO3,even though in this case the Kβ2,5maximum appears slightly shifted to lower energy contrary to the expected position from its oxidation state. A shift of 1 eV is observed between NdMnO3and Sr/CaMnO3, consistent with the difference in one oxidation state unit [55]. We also measured the VTC Kβ XES spectra at temperatures corresponding to the three phases in Nd0.5Sr0.5MnO3and at 80 K for the other compounds. However, the spectra do not change with temperature outside the experimental error for any of the samples (not shown). The main conclusion from the Mn Kαand KβXES measurements is that while significant changes in the electronic structure are observed between the different reference compounds, only very small variations are detected for Nd0.5Sr0.5MnO3as a function of temperature despite the different physical properties. The electronic state of the Mn atom seems to be similar in all phases and cannot be described 205108-4 HIGH-RESOLUTION Mn K-EDGE X-RAY EMISSION . . . PHYSICAL REVIEW B 93, 205108 (2016) in terms of a 1:1 superposition of Mn3+/Mn4+(neither temporally nor spatially), in agreement with the homogenous mixed-valent character of Nd0.5Sr0.5MnO3. In general, for all the emission spectra, Nd0.5Sr0.5MnO3shows a behavior closer to NdMnO3(Mn3+) than to Sr/CaMnO3(Mn4+). However, we find very small temperature changes in the XES spectra of Nd0.5Sr0.5MnO3accompanying the transition to the ferromagnetic-metallic phase that are interpreted as a decrease in the local spin density and are almost completely reversed when passing to the low-temperature charge-ordered insulating state. B. High-resolution and spin-selective XANES spectra Figure 5(a) shows the comparison between the TFYXANES and the HERFD-XANES spectra measured at the maximum of the Kα1line for the Nd0.5Sr0.5MnO3sample at RT. The HERFD-XANES shows sharper spectral features (at both the pre-edge structure and white line) than the TFY-XANES. In Fig. 5(b) the HERFD-XANES spectra at the Kα1line are shown for all the compounds. The spectra are shifted following the expected trend with the formal valence state of the Mn atom [56]. The absorption edge appears at a higher energy for the Mn4+compounds (Sr/CaMnO3) and at a lower energy for NdMnO3(Mn3+). The energy separation, also referred to as a chemical shift, is about 3.1 eV as deduced from the maximum of the derivatives of the NdMnO3and SrMnO3 spectra. For Nd0.5Sr0.5MnO3, the absorption edge position lies in between the two extremes suggesting an intermediate valence. The pre-edge structure also evolves in spectral shape and intensity between the samples [see Fig. 5(b), inset]. The pre-edge of Sr/CaMnO3exhibits three features: a strong peak at 6542 eV (a2), a weaker shoulder around 2 eV below (a1) and a third peak about 2 eV above (a3). For both NdMnO3 and Nd0.5Sr0.5MnO3the pre-edge shows two resolved peaks a1 and a2 of comparable intensities that are shifted to lower energies relative to those in Sr/CaMnO3. Spin-selected x-ray absorption spectroscopy stems from the assumption that the Kβ(3p→1s) x-ray emission spectrum can be separated into two internally referenced spin-up and spin-down parts [34,57–59]. The spin-up and spin-down contributions refer to a parallel and antiparallel alignment respectively of the 3phole spin and the local 3dmoment. One can achieve a local spin selectivity in the Kedge absorption spectra by setting the emission energy to either the Kβ1,3main line to get the spin-down contribution, or the Kβsatellite to obtain the spin-up contribution, and measure the emitted photons as a function of the incident x-ray energy. Figure 5(c) shows the spin-selective XANES spectra measured at the Kβ1,3 (spin-down) and Kβ(spin-up) lines for Nd0.5Sr0.5MnO3at RT. The XANES at the Kβline is shifted to lower energy by about 0.8 eV with respect to the spectrum measured at the Kβ1,3line as deduced from the maximum of the derivatives. Similar results were found for the other samples measured (not shown). Moreover, the prepeaks in the spin-selective XANES at the Kβ1,3and Kβlines show the same shift in analogy with previous works in related manganese perovskites [58,60,61]. HERFD-XANES spectra of the different compounds were also measured as a function of temperature. The detailed temperature dependence of the HERFD-XANES spectra was FIG. 5. Normalized XANES spectra at the Mn K edge measured at RT. (a) Comparison between the HERFD-XANES at the Kα1line and TFY-XANES of Nd0.5Sr0.5MnO3. (b) HERFD-XANES at the Kα1line of Nd0.5Sr0.5MnO3,NdMnO 3,SrMnO 3,andCaMnO 3.The inset shows an expansion of the pre-edge region. (c) Spin-selective XANES spectra measured at the Mn Kβ1,3(spin-down) and Kβ (spin-up) lines for Nd0.5Sr0.5MnO3. 205108-5 S. LAFUERZA et al. PHYSICAL REVIEW B 93, 205108 (2016) FIG. 6. Difference HERFD-XANES data measured at the Mn Kα1line for (a) Nd0.5Sr0.5MnO3between 292 K and each T, (b) NdMnO3 between 260 K and each T, (c) SrMnO3between 260 K and each T, and (d) CaMnO3between 260 K and each T. followed between 292 K and 80 K (T ≈20 K) at the Kα1 emission line energy because of the larger intensity of the Kα1line and the consequently higher signal-to-noise ratio. The difference signal obtained by subtracting the HERFDXANES data at the highest temperature from all the data files (at different temperatures upon cooling down) has been evaluated to better elucidate the possible changes. Figure 6 shows the difference signals obtained for Nd0.5Sr0.5MnO3, NdMnO3,SrMnO 3, and CaMnO3. The main changes occur at the prepeaks and rising edge. Regarding the evolution at the main edge, an increase of the slope is observed when cooling down. This effect is common for all the samples and can be explained as due to the freezing of the thermal vibrations [62,63]. We note that the weaker variation occurs for NdMnO3where the Mn atom is surrounded by a tetragonaldistorted oxygen octahedron which implies a less sharp rising edge and thus obscures the thermal vibrations. Concerning the prepeaks, NdMnO3,SrMnO 3, and CaMnO3exhibit a monotonic evolution with decreasing temperature, whereas in Nd0.5Sr0.5MnO3the structures in the difference signal show a change of sign. An upward peak appears at 6540 eV below 240 K when the paramagnetic-insulator to ferromagneticmetallic transition occurs that becomes downward below 140 K, which corresponds to the insulating charge-ordered low temperature phase. Conversely, a downward peak occurs at 6542 eV (T<240 K) that becomes upward below 140 K. This result suggests that the p-dhybridization increases in the ferromagnetic-metallic phase and decreases in the insulating phase. The hybridization increase in the ferromagnetic phase is consistent with the reduction in local spin moment derived from the Mn CTC KβXES spectra. C. EXAFS spectra The EXAFS signal at the Mn Kedge of Nd0.5Sr0.5MnO3 cannot be obtained with a sufficiently long range in kfor quantitative analysis by conventional XAS measurements due to interference with the Nd L2edge (6722 eV). We have 205108-6 HIGH-RESOLUTION Mn K-EDGE X-RAY EMISSION . . . PHYSICAL REVIEW B 93, 205108 (2016) overcome this problem by combining the HERFD-EXAFS at the Mn Kα1line with the simultaneously measured TFYEXAFS. We note that these EXAFS spectra suffer from Nd self-absorption at the L2edge energy which appears as a positive contribution in the TFY and as a negative contribution in the HERFD as discussed by Bianchini et al. [64]. In Fig. 7(a) we show that the EXAFS signal free of Nd self-absorption can be obtained from the weighted addition of the HERFD and TFY spectra. The weight of each spectrum is fitted to suppress the intense Nd L2white line. In any case, the EXAFS signal free of Nd self-absorption does not significantly change for a variation of 5% of the weighted addition. The thus obtained EXAFS signals and the corresponding FTs are shown as a function of temperature in Figs. 7(b) and 7(c), respectively. We observe a weak thermal dependence of the EXAFS signal. The spectra were fitted including the first (O atoms) and second shell contributions (Nd and Sr atoms). The analysis shows that the Mn atom is surrounded by six O atoms at the same distance, i.e., nearly a regular octahedron, with a relatively high Debye-Waller (D-W) factor at 292 K. The Mn-O distance hardly changes with temperature showing simply a small increase at the onset of the ferromagnetic-metallic phase, where it is also observed a minimum in the D-W factor indicating a less distorted octahedron [see inset of Fig. 7(c)]. These results show that the interatomic distance does not follow a normal expansion but remains nearly constant. On the other hand, the Debye-Waller factor does not follow a standard thermal variation (Debye or Einstein models) and the high value at low temperatures manifests a small structural distortion. IV. DISCUSSION AND CONCLUSIONS The combination of high-resolution XES-XAS spectroscopies applied to the study of the different magneto-electrical phases in Nd0.5Sr0.5MnO3has provided valuable information on the electronic state and local geometry of the Mn atom. Several works were reported earlier on XES experiments in perovskite manganites [8,48,50,51,58,60,62]buttheywere limited to only one de-excitation channel (mainly Mn CTC KβXES) and lack exhaustive temperature dependence. Conversely, our XES study includes the Mn Kαand Kβemission lines and in the latter both the CTC (Kβ1,3and Kβlines) and VTC (Kβ2,5and Kβ lines) regions as a function of temperature. We obtain that the Mn KβXES spectra are almost identical in the paramagnetic, metallic-ferromagnetic, and CO phases of Nd0.5Sr0.5MnO3. This finding indicates that the Mn oxidation and spin state are not significantly changed by the different magnetic and electrical behavior. Only modest variations are observed accompanying the passage to the ferromagnetic-metallic phase when looking at the thermal difference signals of the Mn CTC Kβspectra, which are reversed when entering the charge-ordered one. These changes agree with a small decrease of the local spin density in the ferromagnetic-metallic phase compared with the paramagnetic and antiferromagnetic insulating phases. The absence of major changes seems to be controversial in regards to the CO phase of Nd0.5Sr0.5MnO3where there are two different Mn sites (conventionally described as Mn3+and Mn4+within a simple 6.46.56.66.76.86.97.07.1 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 (a) Norm. Absorption (arb. units) Incident Energy (keV) TFY-XAS HERFD-XAS 57% TFY-XAS+ 43% HERFD-XAS 024681012 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 (b) 292 K 200 K 180 K 160 K 140 K 120 K 100 K 80 K k2(k) k (Å-1) * 50 100 150 200 250 300 1.85 1.90 1.95 2.00 T (K) R Mn-O (Å) 0.000 0.002 0.004 0.006 2 Mn-O (Å 2 ) 01234 0.0 0.5 1.0 1.5 292 K 200 K 180 K 160 K 140 K 120 K 100 K 80 K (c) R (Å) Modulus of the FT FIG. 7. EXAFS data of Nd0.5Sr0.5MnO3. (a) Comparison between the HERFD-EXAFS at the Mn Kα1, TFY-EXAFS and derived corrected EXAFS spectrum as described in the text. (b) k2χ(k) EXAFS signals between 292 K and 80 K. Asterisk indicates the position of the Nd L2edge white line. (c) Moduli of the FTs of the same signals. Inset: Temperature dependence of the refined Mn-O distance (closed squares, left axis) and Debye-Waller factor (open squares, right axis) derived the from EXAFS data of Nd0.5Sr0.5MnO3. The dashed line indicates the onset of the ferromagnetic-metallic to antiferromagnetic-CO transition. 205108-7 S. LAFUERZA et al. PHYSICAL REVIEW B 93, 205108 (2016) ionic picture). We argue that the electronic difference among the two sites is too small to be detected or, in other words, the charge segregation between the two Mn sites is very small and hence the two sites are very similar. The largest differences in the Mn Kαand KβXES data have been observed across the different compounds (mixed-valent Nd0.5Sr0.5MnO3and single-valent NdMnO3,Sr/CaMnO 3). In all cases, the spectra of the single-valent samples can be grouped in two according to the Mn formal valence (Mn3+and Mn4+) while the spectra of Nd0.5Sr0.5MnO3show an intermediate behavior. Both the Mn Kα1FWHM and Mn Kβ1,3and Kβlines position and intensity ratio depend on the local spin. Our results indicate that the spin value is in agreement with the mixed-valent state of the Mn atom intrinsic to half-doped Nd0.5Sr0.5MnO3. We now proceed to the discussion of the HERFD-XANES spectra, where we analyze the main edge and the prepeak structure separately. As noted in related works [6,8,9,22–25,47], the energy position of the main edge is correlated with the formal valence state of the Mn atom. In a first approximation, based on an ionic model, it can be stated that Mn in Nd0.5Sr0.5MnO3is composed by 50% Mn3+and 50% Mn4+, which would correspond to a formal valence 3.5+.Ifthis were the case, then the HERFD-XANES should agree with the 1:1 addition of the spectra of NdMnO3and SrMnO3.Thisis justified by the fact that the main difference between XANES spectra of compounds with similar local structure geometry but different formal valence states comes from the energy shift of the main edge while the spectral shape remains largely unchanged [56]. However, we find that the HERFD-XANES spectrum of Nd0.5Sr0.5MnO3strongly disagrees with the 1:1 linear superposition of NdMnO3and SrMnO3spectra keeping their original chemical shift E =3.1 eV as shown in Fig. 8. Since the interaction time for the Kedge photoabsorption process is about 10−15 s, this finding means that there is neither spatial nor temporal distribution of purely ionic states Mn3+ and Mn4+. This result seems to be at odds with the conventional description of the so-called CO phase at low temperatures in FIG. 8. HERFD-XANES at the Mn Kα1line of Nd0.5Sr0.5MnO3 at RT compared with simulations based on 1:1 superpositions of different reference spectra and E values as described in the text. Inset: Expansion of the pre-edge region. termsofaMn 3+/Mn4+bimodal distribution. X-ray diffraction and RXS experiments have reported a checkerboard ordering in the CO phase with two different Mn sites and noninteger valence states Mn(3.5±δ/2)+where δ=0.16 ±0.02 [5]. The latter corresponds to E =0.6 eV as follows from the linear relationship empirically found between chemical shift and charge segregation [56]. In Fig. 8two more simulations of the Nd0.5Sr0.5MnO3spectrum in terms of 1:1 additions have been carried out taking into account this chemical shift value, either using the NdMnO3and SrMnO3spectra (Mn3+/Mn4+)or adding up the Nd0.5Sr0.5MnO3spectrum after applying shifts of ±E/2=0.3eV(Mn 3.42+/Mn3.58+). The latter simulation fits nicely the HERFD-XANES of Nd0.5Sr0.5MnO3confirming in this way that in the CO phase the charge segregation is very small between the two different Mn sites, which are in intermediate valence states Mn(3.5±δ/2)+as obtained in RXS experiments. Moreover, it is well known that the white line width increases with the lattice distortions [52]. We note that for NdMnO3where the local structure is tetragonal distorted, the white line width is larger than that of the Sr/CaMnO3compounds with regular structure. For Nd0.5Sr0.5MnO3it is similar to Sr/CaMnO3, indicating that the Mn atom is surrounded by an almost regular oxygen octahedron. This result is consistent with the lack of pure ionic Mn3+and Mn4+valences and corroborates the presence of Mn intermediate valence states with very small charge disproportionation. On the other hand, the spin-selective XANES spectra measured at the Kβ1,3and Kβlines are shifted in energy by about 0.8±0.2eVfor the four samples as deduced from the maximum of the derivatives. This shift arises from the exchange splitting whereby the spin-up states have lower energy than the spin-down states [59], being nearly independent of the Mn formal valence state. The evolution obtained in the prepeak structures is more difficult to interpret. On one hand, the prepeaks have been associated to 1s→3dquadrupole transitions. However, the low intensity predicted for the quadrupole channel indicates that its contribution is very limited. A dipolar contribution to the prepeaks has been proposed due to the strong hybridization of Mn 4porbitals with dorbitals of adjacent Mn atoms through the oxygen 2porbitals by Elfimov et al. [65]in LaMnO3and Joly et al. [66]inTiO 2. Overall, it seems that the pre-edge features contain information on the hybridization with the surrounding atoms and the splitting of the empty Mn dband. As follows from our results [see inset of Fig. 5(b)], the total intensity of the prepeaks increase with the Mn formal valence which can be interpreted as due to the enhancement of the 3d–2p–3dcovalence (Mn–O–Mn) when going from Mn3+-NdMnO3to Mn4+-Sr/CaMnO3.In line with the XES results, the HERFD-XANES data do not show large variations with temperature. However, the difference spectra of the HERFD-XANES measured at the Kα1line reveal changes in the pre-edge region of mixed-valent Nd0.5Sr0.5MnO3(see Fig. 6) that reflect an increase of the p-dhybridization in the ferromagnetic-metallic phase. This result nicely connects with the reduction in the local spin density deduced from the Mn CTC KβXES spectra. Figure 9 shows the temperature dependence in the prepeak region of the spin-selective XANES measured at the Kβ1,3(spin-down) and Kβ(spin-up) lines at 300, 200, and 80 K for all the compounds. We first point out that there is no temperature dependence 205108-8 HIGH-RESOLUTION Mn K-EDGE X-RAY EMISSION . . . PHYSICAL REVIEW B 93, 205108 (2016) 6535 6540 6545 0.0 0.2 0.4 0.6 0.8 1.0 RT 200 K 80 K Ca Sr Nd 0.5 Sr 0.5 Norm. Absorption (arb. units) Incident Energy (eV) Nd a3 a2 a1 FIG. 9. Expansion of the pre-edge region of the temperaturedependent spin-selective XANES spectra measured at the Mn Kβ1,3 (spin-down, solid lines) and Kβ(spin-up, open symbols) lines for Nd0.5Sr0.5MnO3,NdMnO 3,SrMnO 3,andCaMnO 3. of the pre-edge spin-selective XANES for NdMnO3down to 80 K. This result can be expected because in the studied temperature range NdMnO3remains paramagnetic-insulator (TN≈78 K) [38] and thus no changes are predicted in the valence, spin, or covalency of the Mn atoms. Nevertheless, a significant temperature and spin dependence of the pre-edge spectra is observed for Nd0.5Sr0.5MnO3and Sr/CaMnO3.In Nd0.5Sr0.5MnO3(TC≈250 K and TN≈150 K) upon cooling from room temperature to 200 K the a2 peak intensity is slightly reduced for the spin-down channel, whereas a strong enhancement of this intensity occurs with further cooling down to 80 K. A distinct decrease of the intensity of the a1 peak is also observed on cooling from 200 K to 80 K, primarily in the spin-up channel. A similar temperature variation of the pre-edge structure has been found in La0.5Ca0.5MnO3 when entering the antiferromagnetic CO phase [60]. In the Mn4+-compounds studied, similar changes are found when crossing TN. Let us focus in the following on CaMnO3 (TN≈125 K [39]) as representative example. In the spin-up channel below 80 K, the a2 peak intensity increases while the a3 peak intensity decreases. In the spin-down channel the temperature dependence of these pre-edge structures is reversed below 80 K. Similar trends are seen for SrMnO3 (TN≈230–260 K [40–41]), but in this case the changes in temperature occur between room temperature and 200 K in agreement with the higher TN. We can then conclude that the main changes in the electronic state of Mn atoms across the different magnetoelectrical phases are related to changes in the Mn 3d-O 2pcovalency that are also intimately connected with changes in the magnetic ordering. The EXAFS results on Nd0.5Sr0.5MnO3have shown that the Mn atom is surrounded by six O atoms at the same distance. The Mn-O interatomic distance is slightly larger at the onset of the ferromagnetic-metallic phase, coincident with a minimum of the D-W [see Fig. 7(c)]. As previously reported for the La1−xCaxMnO3series [9], in Nd0.5Sr0.5MnO3the MnO6octahedra are dynamically distorted in the paramagnetic high-temperature phase. With decreasing temperature, the system tends to stabilize by decreasing the local distortion due to the presence of long-range ferromagnetic interactions. But, below 170 K, when entering the statically ordered CO phase, this local distortion increases again and becomes temperature independent instead of collapsing as occurs for the magnetoresistive manganites [22,31]. To the general behavior of half-doped manganites, our EXAFS results in Nd0.5Sr0.5MnO3 add the finding that the local structural differences at the Mn atom between the ferromagnetic-metallic and CO phases are also very small. In summary, this XAS-XES study performed in Nd0.5Sr0.5MnO3, where the three characteristic phases of the hole-doped manganites are present in separated temperature ranges, demonstrates that the local structure and the electronic state at the Mn atom hardly change among the different phases. However, they clearly differ between the half-doped system and the single-valent Mn3+and Mn4+parent compounds. In this way, the appearance of ferromagnetic-metallic and antiferromagnetic-CO phases must arise from the reported very small differences. Similarly, Fe Kαand KβXES measurements reported no differences across the metal-insulator transition in Fe3O4, the archetypal CO system [67]. These results impose serious constraints on the theoretical models used to describe the mechanisms responsible for the appearance of magnetoresistance and CO phenomena. Conventionally, the theoretical models that describe the behavior of half-doped manganites adopted the ionic picture following the pioneer work of Goodenough et al. [68,69]. The latter considers the existence of Mn3+and Mn4+ions temporally separated but randomly distributed in the lattice for the paramagneticinsulator and ferromagnetic-metallic phases and spatially ordered Mn3+and Mn4+ions in a checkerboard arrangement for the CO phase. This simple model has been already questioned by experimental XAS [5,6,8,9,22,23], RXS [27–30], and bond valence sum analysis [70,71] results showing that the Mn electronic and local structure in the mixed-valent compounds cannot be explained in terms of single-valent integer states. In the so-called CO phases, the charge disproportionation between the distinct Mn sites is nearly indiscernible. Most of the models are based on the uniformity of the doping and take as a basis the description of the electronic states of the Mn atom in an octahedral crystal field. It is, however, crucial to note that the structure is not periodic in the doped oxides. The divalent metal and the rare-earth atoms are randomly distributed in such a way that the crystal field at each of the Mn atoms, and also the local valence, change from site to site. For example, in Nd0.5Sr0.5MnO3the Mn atoms can be surrounded by any combination of (8-n)Nd+nSr neighbors. It has been experimentally demonstrated that the properties of divalent metal rare-earth ordered compounds show a markedly different behavior compared with the disordered ones [72,73]. The key point is that the electronic state at the Mn atom is uniform across the crystal, demonstrating that a strong electronic mixing occurs in the solid. Therefore, the Mn electronic state at each site corresponds to the geometrical average. 205108-9