Fingerprint of dynamical charge/spin correlations in the tunneling spectra of colossal magnetoresistive manganites
S. Seiro, Y. Fasano, I. Maggio-Aprile, E. Koller, R. Lortz, Ø. Fischer
aa r X i v : . [ c ond - m a t . s t r- e l ] J u l Fingerprint of dynamical charge/spin correlations in thetunneling spectra of colossal magnetoresistive manganites
S. Seiro, ∗ Y. Fasano, † I. Maggio-Aprile, E. Koller, R. Lortz, ‡ and Ø. Fischer D´epartement de Physique de la Mati`ere Condens´ee, Universit´e de Gen`eve,Quai Ernest-Ansermet 24, 1211 Geneva, Switzerland
Abstract
We present temperature-dependent scanning tunneling spectroscopy measurements onLa − x Ca x MnO ( x ∼ .
33) films with different degrees of biaxial strain. A depletion in normal-ized conductance around the Fermi level is observed both above and below the insulator-to-metaltransition temperature T MI , for weakly as well as highly-strained films. This pseudogap-like deple-tion globally narrows on cooling. The zero-bias conductance decreases on cooling in the insulatingphase, reaches a minimum close to T MI and increases on cooling in the metallic phase, followingthe trend of macroscopic conductivity. These results support a recently proposed scenario in whichdynamical short-range antiferromagnetic/charge order correlations play a preeminent role in thetransport properties of colossal magnetoresistive manganites [R. Yu et al ., Phys. Rev. B ,214434 (2008)]. PACS numbers: 71.30.+h,75.47.Lx,68.37.Ef,73.50.-hKeywords: metal-insulator transition, manganites, scanning tunneling microscopy, thin films ∗ Present address: Max Planck Institute for Chemical Physics of Solids, N¨othnitzer Str. 40, 01187 Dresden,Germany. E-mail: [email protected] † Present address: Instituto Balseiro and Centro At´omico Bariloche, Bustillo 9500, 8400 Bariloche, Argentina. ‡ Present address: Department of Physics, The Hong Kong University of Science & Technology, Hong Kong.
1n spite of extensive theoretical and experimental research on colossal magnetoresistivemanganites, the mechanism underlying the transition from insulator to metal-like transportconcomitant to ferromagnetic ordering has not yet been completely understood [1]. In theparamagnetic (PM) phase, tunneling spectroscopic measurements are in agreement withthe presence of a gap in the density of states: Tunneling conductance presents a depletionat low bias voltages [2, 3] and the zero-bias conductance follows a thermally-activated-likebehavior [3, 4]. On cooling into the ferromagnetic state the depletion does not disappear [2,3], in apparent contrast to the macroscopic transport properties. In this work we showthat for La . Ca . MnO (LCMO) films of different strain levels the depletion observed inboth the insulating and metallic regimes is accompanied by a spectral weight redistributionon cooling through T MI . The temperature evolution of the zero-bias conductance (ZBC)accounts for the macroscopic insulator-to-metal transition. Our results support a recenttheoretical study which shows that dynamical nanoscale antiferromagnetic/charge order(AFM/CO) correlations give rise to a pseudogap in the density of states (DOS) aroundthe chemical potential ( µ ), not only above but also below the insulator-to-metal transitiontemperature T MI [5].The films studied in this work have been grown by rf sputtering on (100) SrTiO (STO)and (110) NdGaO (NGO). The growth procedure has been reported in detail in Refs. [2, 6].X-ray diffraction measurements confirmed structural homogeneity, single crystallinity, andthe presence of a single phase. Both LCMO/STO and LCMO/NGO films were found to beunder strain. Reciprocal space mapping for LCMO/STO films confirmed an in-plane pa-rameter equal to that of the substrate and a strongly reduced out-of-plane lattice parameter( c ∼ .
80 ˚A) [2]. The LCMO/NGO film is weakly strained ( c ∼ .
87 ˚A) and compressed onaverage in the plane [6]. As estimated from Laue oscillations, the thickness of the films iswell above the dead-layer thickness [7, 8].The resistivity ( ρ ) of the films was measured in a four-point configuration, with the currentflowing parallel to the film plane, along one of the main pseudocubic axis. Both kind of filmsexhibit a transition from insulator- to metal-like behavior at a temperature T MI lower thanthat of bulk compounds, as expected for films under biaxial strain [6, 7, 9, 10, 11, 12].For LCMO/STO, T MI =154 K while for LCMO/NGO, T MI =235 K. The residual resistivitiesare 5 mΩ cm and 0.35 mΩ cm, respectively. By taking advantage of the shift in T MI due tosubstrate-induced strain [6, 7, 9, 10, 11, 12], we accessed the metallic and insulating phases2n a wide range of temperatures.Since three-dimensional perovskites lack an easy cleaving plane, we thoroughly cleanedthe surface of the films with isopropanol in an ultrasonic bath prior to entering the samplein a variable-temperature home-made scanning tunneling microscope. This procedure hasbeen widely applied to non-cleavable manganites [2, 13, 14]. Topographs and current vs.voltage, I ( V ), maps were measured as a function of temperature. Topographs were acquiredat a constant current of 0.2-0.6 nA and a bias voltage of 1-3 V. Spectroscopy was measured ata fixed tip-sample separation (0.5 nA at 0.5 V bias) by recording the current when rampingthe bias voltage. The tip electrode made of electrochemically etched Ir was grounded andthe bias voltage V was applied to the sample. Figure 1: Topographic and spectroscopic properties of weakly-strained LCMO/NGO. (a) Top:800 ×
800 nm topograph taken at 258 K for a bias voltage of 1.5 V and a tunnel current of 0.2 nA.Bottom: Topographic profile along the dashed line. (b) Top: 450 ×
450 nm zero-bias conductancemaps taken at different temperatures for a junction impedance of 1 GΩ (0.5 V, 0.5 nA). The colorscale used is the same for all maps. Bottom: Distribution of zero-bias conductance values. Thefull lines are Gaussian fits to the experimental distribution. Topographs reveal flat terraces separated by growth steps multiple of the out-of-planepseudocubic lattice parameter for both LCMO/NGO and LCMO/STO films [2, 15]. In3ig. 1(a) we show a typical topograph for the 42 nm thick LCMO/NGO film. ZBC mapsover an area covering a few topographic steps exhibit some dispersion of conductance valuesbut no particular spatial pattern, as shown in Fig. 1(b) for LCMO/NGO and in Refs. [15, 16]for LCMO/STO. For both LCMO/NGO and LCMO/STO films the ZBC distribution isGaussian at all temperatures, complementing our previous results at higher energies onLCMO/STO films [2, 15]. Note that the ZBC level of a pixel in Fig 1(b) is not the averageof ZBC values over the pixel suface, but the point value at the pixel center. If there weresub-pixel regions of very high and very low ZBC, as observed for Pr-doped LCMO films at1.5-2 eV energies [17], they should be manifest in the distribution of conductance values witha statistical weight proportional to their area fraction. The absence of a bimodal distributionof conductance values rather appears to indicate that there is no static phase separation inregions with strongly differentiated electronic properties. Assuming that a “different” phasehas been missed by the discrete positioning of the tip, the surface fraction of it would belimited to roughly 10 − , casting a doubt on the role it plays in transport properties. Theabsence of regions with strongly differentiated tunneling spectroscopic properties (i.e. ofa bimodal ZBC distribution) on LCMO films has also been reported for LCMO/NGO atT MI [18] and for cation ordered LCMO/MgO ( x = 1 /
4) at 115 K and 294 K [13].For both strain states, not only the ZBC values but also the whole I ( V ) curves at anygiven temperature display no bimodal distribution over the field of view. As an example,Fig. 2(a) shows local LCMO/STO I ( V ) curves acquired over a 350 nm-long path at temper-atures below and above the transition. The current vs. voltage curves in the metallic and in-sulating phases present a highly non-linear character and are remarkably alike. At first sight,one might be tempted to link these observations to insulating behavior (DOS( µ ) = 0). How-ever, non-linear I ( V ) characteristics can appear also in metals (DOS( µ ) = 0), in particular ifthe DOS is non-constant in the vicinity of µ , or if the tunneling barrier is energy-dependent(e.g. for bias voltages that are a non-negligible fraction of the apparent workfunction [19]).Further information on the DOS( µ ) can be gained by a close inspection of the behavior ofzero-bias conductance as a function of temperature.Figure 2(b) presents the temperature evolution of macroscopic resistivity and the inverseof the ZBC obtained from the spatially-averaged local I ( V ) curves, for both weakly andhighly-strained films. We found that 1/ZBC is finite at all measured temperatures androughly follows the behavior of macroscopic resistivity, increasing on warming in the low-4emperature phase and decreasing on warming in the high-temperature phase. For a gappedDOS, a thermally-activated ZBC is expected, the activation energy yielding an estimate ofthe gap. In the high temperature phase, the activation energy obtained from the slope of ourlog(ZBC) data vs. 1 / T is comparable to the activation energy estimated from the slope oflog( ρ ) vs. 1 / T, (0 . ± .
01) eV for LCMO/STO and (0 . ± .
01) eV for LCMO/NGO. How-ever, the thermal activation here may be only apparent. The computational study in Ref. [5]proposes the occurrence of dynamic nanoscale AFM/CO correlations in the vicinity of theFM-AFM/CO phase boundary. If the characteristic lifetime of these correlations is shorterthan the timescale of tunneling experiments, they will not be visualized in ZBC images.Nevertheless the presence of these correlations yields a finite DOS ( µ ) at all temperaturesfor long Monte Carlo time scales in the calculation [5], with a temperature dependence thatfollows the trend of macroscopic conductivity. This is indeed observed in our films, suggest-ing that the observed activated-like behavior does not come from a thermal excitation ofcarriers over a band gap but is due to dynamical CO/AFM correlations.The DOS in Ref. [5] is not gapped above T MI , but presents a pseudogap-like depletion in Figure 2: (a) Local I ( V ) curves acquired along a 350 nm trace on highly-strained LCMO/STOfor the metallic (120 K) and insulating (170 K) phases. (b) Local (open symbols) vs. macroscopic(lines) electronic properties vs. temperature for LCMO/NGO (black) and LCMO/STO (red).Temperatures are expressed relative to the respective insulator-to-metal transition temperatures:T ∗ MI = T MI for resistivity curves, while for zero-bias conductance T ∗ MI = T SMI , the temperature atwhich ZBC reaches a minimum. Since scanning tunneling spectroscopy probes the DOS at thesurface, T SMI is slightly different to T MI . µ with a pseudogap energy scale that globally decreases on cooling. We candirectly test these predictions by calculating the normalized or logarithmic conductance,( dI/dV ) / ( I/V ), a method that attenuates the dependence on the tunneling barrier andyields a quantity proportional to the local DOS [20]. Normalized conductance (NC) curvesare presented in Fig. 3 for both highly (left panel) and weakly (right panel) strained films.At high temperatures NC curves present a depletion around zero bias. This depletionsurvives down to low temperatures, but NC peaks (in the case of LCMO/STO) or kinks(for LCMO/NGO) become increasingly marked upon cooling. These features have beenpreviously ascribed to the spectral signature of polarons [2, 21], the distance between thepeaks being related to the polaron binding energy. In this context, a ‘polaron’ describes anobject that is not just a charge coupled to a local lattice distortion, but which has a complexspin and orbital structure, rather like the AFM/CO correlations considered in Ref. [5].An important result of Fig. 3 is that there is neither a discontinuity nor an abruptchange in the shape of spectra across the insulator-to-metal transition. No hard gap openson warming through T MI but the DOS is depleted in the vicinity of µ at all measuredtemperatures. However, subtle and gradual changes in the spectral shape are observedas a function of temperature. In the insulating phase, on cooling towards the transition,ZBC decreases as peaks/kinks develop at the flanks of the depletion, indicating a spectralweight transfer from low to high energies. On further cooling, in the metallic phase spectral Figure 3: Normalized conductance curves at different temperatures both above (full symbols)and below (open symbols) the insulator-to-metal transition for LCMO/STO (left panel) andLCMO/NGO (right panel). et al. [5] SS / ( T S M I ) (T-T MI )/T MI LCMO/NGO LCMO/STO(28 nm) LCMO/STO(31 nm)
Figure 4: Pseudogap energy (relative to its value at T
SMI ) vs. reduced temperature for LCMO/STOin red (circles are data from this work, squares from Ref. [2]), LCMO/NGO (orange) and resultsfrom the simulations in Ref. [5] (gray-shaded zone, covers the error). The temperatures are givenrelative to T
SMI for all experimental data (see caption of Fig. 2), and relative to the temperatureof the local DOS minimum for the simulation. weight increasingly builds up at the chemical potential and the height of conductance peaksslightly decreases, see Fig. 3(a). In terms of the findings of Ref. [5], these results can beinterpreted as follows: In the insulating phase the AFM/CO correlation lifetime increasesupon cooling, but decreases as ferromagnetic order sets in at
T < T MI . Although the shortlifetime of the correlations hindered us to directly image them in conductance maps, theyare manifest through the pseudogapped-like depletion of tunneling conductance and thetemperature evolution of the ZBC. It is important to note that the calculations of Ref. [5]were performed in the clean limit. Quenched disorder is known to enhance the colossalmagnetoresistance and extends the region of the phase diagram where it occurs [22], but ourresults call for a systematic study of its effect on the AFM/CO correlations dynamics.The width of the depletion or pseudogap, ∆, was obtained for the curves in Fig. 3 asthe half-distance between the two peaks (kinks) in the case of LCMO/STO(NGO) films.The pseudogap values are summarized in Fig. 4 for all samples [23]. The behavior of ∆ asa function of temperature is in qualitative agreement with the results of Ref. [5], globallyincreasing on warming. In the vicinity of T MI , the pseudogap energy changes its slope but∆ further increases in the high-temperature phase.In conclusion, we have carried out a detailed analysis of the tunnel spectra of manganite7lms that sheds light into the puzzling relation between the local electronic properties andmacroscopic transport across the insulator-to-metal transition. Although both the insulat-ing and metallic phases present a pseudogapped normalized conductance, spectral weightredistributes as a function of temperature in such a way that the temperature evolution ofthe macroscopic conductivity is tracked by the density of states at the chemical potential.A similar behavior was predicted in Ref. [5], where nanoscale spin/charge correlations werefound to be increasingly stable on cooling towards T MI but their lifetime is reduced withthe onset of ferromagnetic order. In addition, the predicted temperature evolution of thepseudogap energy [5] is consistent with the measured ∆( T ). This behavior is observed forhighly and weakly strained films. Our results strongly support that dynamical spin/chargecorrelations play a preeminent role in the transport properties of colossal magnetoresistivemanganites.The authors thank Y. Rong for useful discussions and the Swiss National Science Foun-dation/MaNEP for financial support. [1] For a review, see for example: Y. Tokura, Colossal Magnetoresistive Oxides , Gordon andBreach Science Publishers, Amsterdam (2000), E. Dagotto,
Nanoscale Phase Separation andColossal Magnetoresistance , Springer-Verlag, Berlin (2003), T. Chatterji,
Colossal Magnetore-sistive Manganites , Kluwer Academic Publishers, Dordrecht/Boston/London (2004).[2] S. Seiro et al. , Phys. Rev. B , 020407(R) (2008).[3] U. R. Singh et al , Phys. Rev. B , 014404 (2008).[4] H. M. Rønnow et al. , Nature , 1025 (2006).[5] R. Yu et al. , Phys. Rev. B , 214434 (2008).[6] S. Seiro, E. Koller, Y. Fasano, and Ø. Fischer, Appl. Phys. Lett. , 091913 (2007).[7] M. Bibes et al. , Phys. Rev. B , 134416 (2002).[8] K.-H. M¨uller et al. , J. Magn. Magn. Mater. , 447 (2002).[9] B. Vengalis et al. , J. Magn. Magn. Mater. , 35 (2000).[10] J. H. Song, J.-H. Park, Y. H. Jeong, and T. Y. Koo,
Physica B , 729 (2002).[11] M. Ziese, H. C. Semmelhack and K. H. Han,
Phys. Rev. B , 134444 (2003).[12] S. Valencia, L. Balcells, B. Mart´ınez, and J. Fontcuberta, J. Appl. Phys. , 8059 (2003).
13] L. Sudheendra et al. , Phys. Rev. B , 172407 (2007).[14] S. R¨oßler et al. , E. Phys. Lett. , 17009 (2008).[15] S. Seiro et al , J. Magn. Magn. Mater. , e243 (2007).[16] S. Seiro, PhD Thesis 3995, University of Geneva (2008).[17] J. X. Ma, D. T. Gillaspie, E. W. Plummer, J. Shen,
Phys. Rev. Lett. , 237210 (2005).[18] J. Mitra et al , Phys. Rev. B , 094426 (2005).[19] In our experiments the apparent workfunction ranged between 1 and 1.5 eV.[20] R. M. Feenstra, Phys. Rev. B , 4561 (1994).[21] J. Y. T. Wei, N.-C. Yeh, R. P. Vasquez, Phys. Rev. Lett. , 5150 (1997).[22] E. Dagotto et al. , J. Phys.: Condens. Matter , 434224 (2008).[23] The energy scale of the depletion in our films, comparable to the activation energy in resistivity,is well below the 1.5 eV observed by Fuchigami et al . in Phys. Rev. Lett. , 066104 for the √ × √ / Ca / MnO /Nb-STO film. This indicates that thesurface of our films has not undergone this kind of reconstruction./Nb-STO film. This indicates that thesurface of our films has not undergone this kind of reconstruction.