PHYSICAL REVIEW B 89, 024420 (2014)

Intrinsic antiferromagnetic coupling underlies colossal magnetoresistance effect:Role of correlated polarons

V. Moshnyaga,1,* A. Belenchuk,2 S. Huhn,1 C. Kalkert,1 M. Jungbauer,1 O. I. Lebedev,3 S. Merten,1 K.-Y. Choi,4,5

P. Lemmens,4 B. Damaschke,1 and K. Samwer1

1Erstes Physikalisches Institut, Georg-August-Universitat-Gottingen, Friedrich-Hund-Platz 1, D-37077 Gottingen, Germany2IIEN, Academy of Sciences of Republic Moldova, Strada Academiei 3/3, MD-2028 Chisinau, Republic of Moldova

3Laboratoire CRISMAT, UMR 6508 CNRS-ENSICAEN, 6 Boulevard du Marechal Juin, 14050 CAEN Cedex, France4Institut fur Physik der Kondensierten Materie, Technische Universitat Braunschweig, Mendelssohntrasse 3,

D-38106 Braunschweig, Germany5Department of Physics, Chung-Ang University, Seoul 156-756, Republic of Korea

(Received 22 November 2013; published 30 January 2014)

A commonly believed picture of colossal magnetoresistance (CMR) effect is related to a first-order phasetransition and electronic phase separation with coexisting ferromagnetic metallic and antiferromagnetic insulatingphases. However, the underlying mechanism, i.e., the characteristic energy scale of the interacting phases andtheir spatial extent, is still under debate. Here we present experimental evidence on the existence of an effectiveantiferromagnetic coupling between the ferromagnetic nanodomains in epitaxial thin films of a classical CMRmaterial (La1−yPry)0.67Ca0.33MnO3 with Pr doping, y = 0.375 and 0.4. This coupling yields to peculiar low-fieldCMR behavior with magnetic hysteresis and slow resistance relaxation, both induced by the magnetizationreversal. The coercive field obeys a square-root temperature dependence for T � TC and increases anomalouslyclose to the phase transition. We modeled the magnetic structure within the phase-separation scenario as anassembly of single-domain ferromagnetic nanoparticles, antiferromagnetically coupled (pinned) by correlatedJahn-Teller polarons. The concentration of polarons increases drastically close to phase transition as indicatedby the third harmonic of the electrical conductivity as well as Raman spectroscopy.

DOI: 10.1103/PhysRevB.89.024420 PACS number(s): 75.47.Lx, 71.38.−k, 75.47.Gk

I. INTRODUCTION

The colossal magnetoresistance (CMR) effect, which man-ifested itself as a drastic decrease of electrical resistance inan applied magnetic field, was first observed in single crystalsof perovskite manganites in 1970 [1]. The “rediscovering” ofCMR in thin manganite films [2,3] initiated an enormous boomof research focusing on the intriguing CMR physics as wellas on the potential applications. After almost two decades ofextensive experimental and theoretical studies, a commonlybelieved picture has been established: the fundamentals ofCMR are related to a first-order phase transition and electronicphase separation with coexisting ferromagnetic (FM) metallicand antiferromagnetic (AFM) insulating phases.

The generic phase diagram of CMR manganites [4] withFM and AFM phases underscores a competition betweentwo driving tendencies: electron delocalization, favoring aFM phase, and localization, which stabilizes an AFM groundstate. These competing interactions result in the coexistenceof FM metallic and charge-ordered insulating (COI) phases[5] close to a first-order magnetic transition if the strength ofelectron-phonon coupling is large enough. “Colossal” valuesof CMR = �ρ(H )/ρ(0) � 106% were indeed observed closeto the FM/AFM phase boundary, achieved both by “fillingcontrol” close to x � 0.5 doping in the La1−xCaxMnO3

system [6] and by “bandwidth control” while substitutinga large La cation by a smaller Pr in La5/8−xPrxCa3/8MnO3

(Ref. [7]). Electronic phase coexistence was believed to governthe magnetotransport in manganites close to phase transition

*Corresponding author: vmosnea@gwdg.de

at TC, although the energy scale and the nature of magneticinteractions between the competing phases still remain unclear.Sen et al. [8] have demonstrated theoretically a coexistence ofFM and nm size AFM phases even in the absence of A-sitedisorder [9] but at sufficiently large electron-phonon coupling.Correlated polarons [10,11] (CPs), associated with orbitalpolarization of Mn3+ states and corresponding static Jahn-Teller (JT) distortions of MnO6 octahedrons, may contributeto the nm-scale phase separation. CPs were detected in neutronand x-ray scattering [12–17] as a short-range-ordered latticesuperstructure with a correlation length of δ � 1–2 nm andcharge/orbital ordering (COO) of CE type. Considering theground state of the CE phase to be AFM, one can suggestthat CPs may also possess short-range AFM correlations.However, the estimated very small amount of CPs [15–17]questions their role in the complex magnetic and electric stateclose to TC.

Here we report that the CMR in high quality and strain-free epitaxial films of (La1−yPry)0.67Ca0.33MnO3 (LPCMO)on MgO(100) substrates is characterized by a low-fieldhysteresis and slow relaxation dynamics, both originatingfrom a peculiar magnetic domain structure with nm-sizeFM domains, antiferromagnetically coupled by correlatedJahn-Teller (JT) polarons. Our experimental results confirmthe earlier theoretical models of electronic phase separationwith competing FM metallic and AFM insulating phases [8,18]at the first-order phase transition. Moreover, we demonstratethat even a tiny amount of the polaronic AFM phase stabilizesthe FM nanodomains and induces an effective exchangecoupling between them, thus playing a decisive role in thespin polarized charge transport and in the CMR behavior ingeneral.

1098-0121/2014/89(2)/024420(8) 024420-1 ©2014 American Physical Society

V. MOSHNYAGA et al. PHYSICAL REVIEW B 89, 024420 (2014)

II. SAMPLE PREPARATION AND EXPERIMENTALTECHNIQUES

LPCMO films with Pr doping y = 0.375 and 0.4 wereprepared by a solution-based and vacuum-free metalorganicaerosol deposition (MAD) technique [19]. The precursors,acetylacetonates of La, Pr, Ca, and Mn, have been weightedin appropriate amounts to obtain the experimentally deter-mined molar ratios in the solution, i.e., La:Pr:Ca:Mn =0.57:0.34:0.40:1. Then they have been dissolved in dimethyl-formamide to a concentration 0.02 M calculated on Mnprecursor and sprayed onto the heated MgO(100) substrateby using compressed air at a pressure p = 5 atm. Keeping thefollowing typical MAD processing conditions, like substratetemperature Tsub = 800–900 °C, the solution volume V =2 ml, and deposition time t = 5 min, one gets an LPCMOfilm with the thicknesses, d � 60 nm; the correspondinggrowth rate is v � 12 nm/min. After deposition the films werecooled down to room temperature in 30 min. For comparisonLa0.7Ca0.3MnO3 (LCMO) and La0.7Sr0.3MnO3 (LSMO) weregrown on MgO(100) as described in details in [17].

The structure of the films was characterized at room temper-ature by x-ray diffraction (XRD, �-2� Bragg-Brentano geom-etry, CuKα radiation), small-angle x-ray reflection (XRR), andtransmission electron microscopy (TEM) as well as by scan-ning tunneling microscopy (STM). TEM and high-resolutionTEM (HRTEM) studies were carried out on a cross-sectionsample using a Tecnai G2 30 UT microscope operated at300 kV and having 0.17-nm point resolution. A cross-sectionsample for TEM measurements was prepared using a focusion-beam (FIB) machine (FEI Helios 600 NanoLab DualBeaminstrument in CIC Nano GUNE). The four-probe dc and acresistivity measurements were performed by using a physicalproperty measurement system (PPMS) from Quantum Designin the temperature range T = 2–400 K, and for magnetic fields,μ0H = 0–4 T, applied parallel to the film plane. The amplitudeand the frequency of the ac current were varied in the range J

= 0–400 μA and f = 0–1000 Hz, respectively. Along with theac voltage at the fundamental frequency Uω, a third harmonicsignal U3ω was evaluated by means of Fourier analysisof the measured ac signal. Magnetic measurements werecarried out by using a commercial superconducting quantuminterference device (SQUID) magnetometer magnetic propertymeasurement system (MPMS, from Quantum Design, forT = 10–400 K and fields μ0H = 0–4 T, aligned parallel to thefilm plane. In addition the magnetooptical Kerr effect and Kerrellipticity were measured by using He-Ne laser with polariza-tion modulation at 50 kHz in a closed-circle He cryostat, T =20–300 K, and magnetic fields μ0H = 0–1.5 T, oriented at 45°with respect to the film plane. The same experimental setupwas used to measure the resistance relaxation for magneticfields applied parallel to the film plane. Raman spectroscopywas studied by using a Raman confocal microscope (LabRamHR800-UV) with Ar+-laser excitation at the wavelength, λ =488 nm. The spectra were measured for T = 80–300 K byusing a cryogenic table “Linkam THMS 600” continuouslycooled by liquid N2. The sample temperature was controlledby pumping rate of N2 and by heating of the sample holder.The measured Raman spectra were corrected by subtractingthe base line and by normalizing by the Bose-Einstein factor.

FIG. 1. The θ -2θ pattern of an LPCMO/MgO film with y = 0.4shows the substrate peaks and (00l) peaks (l = 1,2,3,4), from thefilm of a perovskite structure, indicating an out-of-plane epitaxy withc-axis lattice parameter, c = 0.3870 nm. From the periodicity of XRRoscillations (see the inset) the thickness, d = 53.6 nm, was obtained.

III. RESULTS

XRR measurements (see the inset to Fig. 1) indicate a largescale homogeneity of the LPCMO films with the thicknessd = 50–70 nm. XRD analysis (see Fig. 1) reveals a perfectout-of-plane epitaxy and a strain-free state of the films witha pseudocubic lattice parameter c � 0.3870 nm, very closeto the bulk value [7]. The STM image in Fig. 2 shows asurface morphology characteristic for epitaxial growth withatomically smooth terraces of 1 unit cells (u.c.) height andmean-square roughness, rms = 0.6 nm, at the 1-μm2 area.Moreover, an in-plan epitaxy of the film is evident with mosaicblocks oriented along two mutually perpendicular directionsdue to cubic crystalline structure of MgO substrate.

The TEM structural analysis shown in Fig. 3 infers a singlecrystalline character of LPCMO/MgO films and confirms STM

FIG. 2. (Color online) STM image of the studied LPCMO/MgOfilm (y = 0.4) with a mean-square roughness, rms = 0.6 nm.

024420-2

INTRINSIC ANTIFERROMAGNETIC COUPLING . . . PHYSICAL REVIEW B 89, 024420 (2014)

FIG. 3. (Color online) (a) Bright-field low magnification TEMimages of a LPCMO/MgO film (y = 0.4) and corresponding EDpattern showing a single crystalline epitaxial character of the film.Notice the presence of twin boundaries indicated by white arrow-heads. An orthorhombic Pnma structure is indicated by diffractionspots (1/2, 0, 0) and (0, 1/2, 0) in ED pattern (left, bottom inset)and also twinning character of the film. HRTEM GPA color strainanalysis maps along two orthogonal direction g100 (perpendicular tothe interface) and g010 (along the interface) are given as inset. Noticethe absent of twin boundary in GPA pattern. (b) HRTEM imageof a LPCMO/MgO interface showing flat film/substrate interfaceand confirming coexisting crystallographic Pnma rotational twinningdomains with heteroepitaxial perfect twin boundary indicated bywhite arrows.

morphology data concerning a large scale homogeneity andflatness of the films. The sample in Fig. 3(a) shows a uniformthickness �50 nm in agreement with the nominal one andis characterized by the presence of twin boundaries, alignedperpendicular to the substrate surface. Electron diffraction(ED) analysis elucidates an orthorhombic Pnma structure withthe following lattice parameters: a = 0.544 nm, b = 0.769 nm,c = 0.545 nm at room temperature (no. ICSD 96908). The EDpattern in the inset of Fig. 3(a), being a superposition of theform MgO (substrate) and LPCMO (film), clearly evidencestwo main features of the film: (1) heteroepitaxial growth ofthe LPCMO film on the MgO substrate with the pseudocubicperovskite block ap LPCMO of the film aligned parallel to thatof the substrate, ap LPCMO//aMgO, and (2) twinning structureof LPCMO film. The HRTEM image of the LPCMO film[Fig. 3(b)] confirms the presence of rotational twins andreveals a heteroepitaxial and homogeneous interface. It shouldbe noticed that due to very small orthorhombic distortionof LPCMO structure from a cubic perovskite, the latticeparameters of LPCMO, i.e., a/�2, b/2, and c/�2, are very

FIG. 4. (Color online) General characteristics of the metal-insulator (a) and ferro-paramagnetic (b) phase transitions in a(La0.6Pr0.4)0.67Ca0.33MnO3 film on MgO(100). The insets illustratea temperature hysteresis in resistivity [(a), right inset] and magneti-zation [(b), right inset] as well as magnetic field dependences of theresistivity [(a), left inset] and of the magnetization [(b), left inset] atT = 10 K.

close to each other and can be expressed as ap LPCMO. In thisrespect, the geometric phase analysis (GPA) patterns point outa uniform strain in both normal to the growth direction (g100)and parallel to the LPCMO/MgO interface (g010) directions.No twin boundary can be distinguishing in GPA color strainanalysis maps [see right bottom inset Fig. 3(a)]. In addition, anetwork of misfit dislocations along the substrate/film interfaceactuates the relaxation of the lattice mismatch strain within thefirst two to three monolayers of the growing film, yielding astrain-free state of the rest of the LPCMO film.

The temperature dependences of the resistivity ρ and mag-netization M presented in Fig. 4 demonstrate extremely sharpand coupled metal-insulator (TMI) and magnetic (TC) transi-tions at TMI ∼ TC = 195 K, with the maximal value of log-arithmic derivative of the resistivity αρ = (T/ρ)(dρ/dT ) =140. This confirms a high crystalline quality and chemicalhomogeneity of the films. The measured values of the residualresistivity, ρ(4 K) � 180 μ cm, and of the saturationmagnetization, M(10 K) � 600 emu/cm3 � 3.6μB/Mn [insetin Fig. 4(b)], agree well with the nominal Ca doping x � 0.3.The zooms in ρ(T ) and M(T ) [insets to Figs. 4(a) and 4(b)]reveal a warming/cooling hysteresis of δT � 1 K, clearlyseen for T = 185–195 K, i.e., just below TC. Apparently,CMR = 100%[R(0)−R(4T )]/R(4T ), as shown in Fig. 5(a),

024420-3

V. MOSHNYAGA et al. PHYSICAL REVIEW B 89, 024420 (2014)

FIG. 5. Magnetic field dependencies of the resistivity (a) andmagnetization (b) of an LPCMO (y = 0.4, TC = 195 K) film fortemperatures, 180 K < T < 205 K, where the CMR develops. Theinset to Fig. 2(a) demonstrates the correlation between the coercivefield (SQUID measurement) and the field at which CMR shows amaximum. The inset to Fig. 2(b) illustrates a strong enhancement ofthe saturation field in the vicinity of TC.

develops exclusively within this narrow temperature windowof about 10 K. At low temperatures, no tunneling MR norCMR was observed [see the inset in Fig. 4(a)] thus rulingout the grain-boundary-governed transport in the films understudy. All this indicates that CMR in our LPCMO/MgO filmsis driven by the first-order phase transition, as indicated bythe hysteresis in ρ(T ) and M(T ), and can be viewed as amagnetic-field-induced insulator-to-metal transition.

In Fig. 5 a close correlation between the ρ(H ) and M(H )dependencies in the vicinity of phase transition is shown.For low fields, 0 < H � 5 kOe, the resistivity [Fig. 5(a)]is hysteretic with two maxima at the coercive field HC [seethe inset to Fig. 5(a)]. A suppression of CMR for T <

180 K is apparently accompanied by the vanishing of theρ(H ) hysteresis. Furthermore, a magnetic zero-field-cooled(ZFC)–FC hysteresis, δM = M(H↑)−M(H ↓) �= 0, also“opens up” within the same temperature interval, 185 K < T

< 195 K, and causes a dramatic increase of Hsat [see the insetin Fig. 5(b)]. Outside the phase-transition region, both for the

FIG. 6. (a) Time dependencies of the resistance in LPCMO film(y = 0.4) close to MI transition, measured after applying andswitching off the external magnetic field, H = 1 kOe. Along with theinstantaneous resistance changes, i.e., the CMR effect, one can seethe “in relaxation”, i.e., the resistance decreases exponentially afterapplying the field and the “out relaxation” (the resistance increasesafter switching off the field). The evaluated relaxation time τ � 200 s;(b) time dependencies of the resistance after applying an in-planemagnetic field, H = 1 kOe, measured at different temperatures in thevicinity of TC = 195 K.

FM and paramagnetic (PM) states, M(H ) and ρ(H ) curvesare evidently not hysteretic. In the PM state for 200 K < T �220 K a metamagnetic transition [Fig. 5(b)] develops forH ∼ 5–20 kOe and results in a very large saturation magneticmoment, MS(205 K) = 5 × 10−4 emu, which consists of 70%of the FM moment, MS(170 K) = 7 × 10−4 emu. The ρ(H )behavior [see Fig. 5(a)], being in close agreement with themetamagnetic transition, deviates strongly from the parabolicdependence, �ρ = ρ(H ) − ρ(0) ∼ −H 2 ∼ −M2, as wouldbe the case for the field-induced PM magnetization [20]. ForT > 220 K no metamagnetic transition was observed forH � 50 kOe.

In Fig. 6 the slow resistance relaxation, measured in thevicinity of the phase transition, is shown. After applying andswitching off a magnetic field, H = 0–1 kOe, in addition tothe instantaneous resistance change (CMR) we observed anexponential relaxation of the resistance, R(t) ∼ exp(−t/τres).Usually the slow resistance relaxation, along with the

024420-4

INTRINSIC ANTIFERROMAGNETIC COUPLING . . . PHYSICAL REVIEW B 89, 024420 (2014)

FIG. 7. (Color online) (a) The square of the measured normalizedcoercive field, h2

c = [Hc(T )/Hc(0)]2, as a function of the normalizedtemperature, t = (T − TC)/TC, for LPCMO films with y = 0.4(closed squares) and y = 0.375 (close triangles), left scale. The fit(dashed line) is the dependence, h2

c ∼ (1 − t/tB), with the blockingtemperature, TB = 1.05TC. On the right scale the temperaturedependencies of the relative amount of correlated polarons, nCP(T ) =U3ω(T )/Uω(T ), (open squares), and of CMR(T) (line), multipliedby a factor of 5 × 10−6 in LPCMO with y = 0.4 demonstrate anenhancement in a number of polarons close to TC as well as the scalingnCP(T ) ∼ CMR(T). The sketch in Fig. 7(a) shows the proposedmagnetic structure near and far below TC; (b) normalized coercivefield, hc = Hc(T )/Hc(0), as a function of normalized temperature,t = (T − TC)/TC, for optimally doped LCMO (closed circles) andLSMO (closed diamonds) films (left scale). One can see (right scale)that ncp(T ) for LCMO (open circles) and LSMO (open squares) areby factor 5 × 103 and 105 smaller than that for LPCMO.

temperature- and high-field-driven M(H ) hysteresis, wasconsidered as a manifestation of a metastable phase-separatedstate [21,22] close to the first-order phase transition. In contrastto the previous experiments [21,22], the slow resistancerelaxation in LPCMO films was observed for low magneticfields H < 1 kOe and, furthermore, R(t) demonstrates theso-called “recovery effect”: after removal of magnetic fieldthe resistance changes back to the initial value at H = 0.The evaluated relaxation time, τres ∼ 200 s, can be modeledwithin the Neel relaxation of the magnetization (see below) andthe superparamagnetic model of Bean and Livingston [23].Importantly, the slow resistance relaxation is limited withinthe above-discussed narrow temperature region close to TC

[see Fig. 6(b)] and is not present both for FM (T � TC) andPM states.

In Fig. 7(a) we present the temperature dependencies ofthe squared normalized coercive field, h2

c = [Hc(T )/Hc(0)]2,as a function of the normalized temperature, t = (T − TC)/T ,

evaluated from the magnetooptic and SQUID measurementsof the LPCMO films with y = 0.375 and 0.4. At low tem-peratures (T � TC) hc decreases with increasing temperatureas expected for a FM material. The data can be fitted verywell by the formula hc = (1 − T/TB)1/2, which describesa classic behavior of single-domain ferromagnetic particles[23]. The fitting parameter, TB = 1.05TC, is the so-calledblocking temperature, above which particles behave as an idealsuperparamagnet (hc = 0). The average radius of magneticdomains (particles) can be estimated from the expressionKUV/kBTB ∼ 25, which is obtained from the Neel relaxationtime, τobs(T ) = τ0exp(KUV/kBTB), for the observable relax-ation times, τobs ∼ 100 s, and τ0 ∼ 10−9 s (see Ref. [23]). Withthe uniaxial anisotropy constant, KU = 3.6 × 104 J/m3, mea-sured for different manganites [24–26], including LPCMO,and the obtained TB = 220 and 205 K for films with y =0.375 and 0.4, respectively, the averaged radius of domains isRFM = (75kBTB/4πKU)1/3 � 6–8 nm. Such a small domainsize illustrates a weakening of FM exchange and an increase ofthe electron-phonon coupling [27]. Moreover, by approachingTC from below [see Fig. 7(a)] the coercive field hc increasesanomalously for LPCMO before vanishing for T > TC. Thisand the drastic increase of saturation field of the magnetization[Fig. 5(b)] are well-known fingerprints of exchange-coupledinhomogeneous magnetic systems. The classical examplesare artificially layered systems: (1) Co/Cu/Co [28] with anonmagnetic Cu layer, actuating a thickness dependent (AFMor FM) coupling of Ruderman-Kittel-Kasuya-Yosida (RKKY)type; (2) Fe/Fe1−xMnx bilayers [29], with the FM layer (Fe)AFM coupled to the Fe1−xMnx (0.15 < x < 0.3) layer; and(3) oxide superlattices [30] of (La2/3Ba1/3MnO3./LaNiO3)N .As shown for comparison in Fig. 7(b), the hc(t) dependence inoptimally doped LCMO/MgO and LSMO/MgO films is ratherlinear, pointing out the existence of weak-pinning centers [31].Finally, no anomalous increase of coercive field close to TC

was observed both for LSMO and LCMO.Thus, the low-field hysteretic CMR in epitaxial

LPCMO/MgO films close to the phase transition is intimatelyrelated to a peculiar magnetic state with FM nanodomainscoupled to each other by AFM exchange. A hint to the possibleorigin of the exchange coupling mediator can be found inFig. 7(a) (right scale), which demonstrates that CMR(T) scalesnicely with the temperature dependence of electrical thirdharmonic coefficient, K3ω(T ) = log10[U3ω(T )/Uω(T )], hereUω and U3ω are measured ac voltages at fundamental (ω =117 Hz) and third harmonic frequencies. As was discussedin detail in Ref. [17], the origin of 3ω signal is ascribedto correlated polarons, which, due to a CE-type ordering ofMn3+/Mn4+ ions, can be considered as electric quadrupolemoments. The latter are well-known sources of the nonlinearcoupling to an electric field, Q = χE2 (see Ref. [32]). Thenonlinear resistance, R3ω = dU3ω/dJ , being also proportionalto the square of the current or E field, provides then a measureof the number of quadrupoles, i.e., CPs, in the sample. Thequotient K3ω ∼ R3ω/Rω is then proportional to the relativeamount of CPs, nCP = NCP/N0, to the whole number chargecarriers N0 induced by Ca doping. Close to phase transition theamount of polarons in LPCMO increases dramatically up tonCP � 0.5% at 195 K and seems to be able to mediate the AFMcoupling. Remarkably, the corresponding values of K3ω in

024420-5

V. MOSHNYAGA et al. PHYSICAL REVIEW B 89, 024420 (2014)

FIG. 8. The inset shows the temperature evolution of Ramanspectra for LPCMO film (y = 0.4) for temperatures, T = 80–300 K.Spectra were shifted by 100 units along vertical axis for clarity. Themain panel demonstrates the temperature dependence of the intensityratio between the Raman shift at 620 cm−1 due to JT phonons(stretching of MnO6 octahedra) and the Raman shift at 440 cm−1

(internal vibrations). A sharp peak close to TC is in qualitativeagreement with the behavior of correlated polarons in Fig. 7(a).

LCMO and LSMO films [see Fig. 7(b)] are much smaller, i.e.,about 10−4% and 10−7%, respectively. These films comparedto LPCMO reveal linear hc(t) dependencies [see Fig. 7(b)]as well as no anomalous increase of hc close to the phasetransition. Apparently, the amount of correlated polarons inLCMO and LSMO is not sufficient for AFM coupling, andthis seems to be the reason for the absence of low-field CMRas well as for moderate and very small CMR values for LCMOand LSMO films [17], respectively.

Another evidence for the enhancement of the polaronicphase at the phase transition could be obtained from Ra-man spectra, shown in Fig. 8. One can see characteristictemperature-dependent Raman shifts at 440 and 620 cm−1,identified [33,34] as Eg (internal vibrations or bending)and B2g (in-phase stretching) modes of MnO6 octahedra,respectively. Due to symmetry considerations Eg is one of fiveRaman active modes (A1g + 4Eg) of the rhombohedral (R-3c)structure and B2g is the characteristic JT phonon mode withinthe orthorhombic (Pnma) structure, which has 24 active modes(7Ag + 5B1g + 7B2g + 5B3g). The intensity I620 increases asthe temperature decreases from 300 K down to TC ∼ 195 K andby further cooling this line becomes suppressed. In contrast,I440 is suppressed by decreasing temperature down to TC,but it increases significantly in the FM state for T � TC.The observed temperature evolution of Raman spectra reflectsthe competition between electron localization via JT distor-tions or CPs, which are compatible with a Pnma structure,and electron delocalization which favors a more symmetricR-3c structure without JT distortions. The intensity ratio,

γ (T) = I620(T )/I440(T ) (Fig. 8), illustrates this competition,underscoring a lattice aspect of the phase transition manifestedby the enhancement of JT distortions at TC.

IV. DISCUSSION

Remarkably, very different electric, nCP(T ), magnetic,hc(T ), and structural, γ (T ), characteristics as well as CMR(T)itself display a pronounced sharp peak close to the phasetransition. This undoubtedly points out a strong couplingbetween electron, spin, and lattice (phonons) degrees of free-dom in an electronically/structurally inhomogeneous LPCMO.We consider this classic CMR material as an “intrinsicexchange coupled system”, sketched in Fig. 7(a). Here, FMnanodomains are AFM exchange coupled (pinned) by corre-lated JT polarons (CPs), nucleated at the domain walls, andassumed to possess AFM correlations. A preferred formationof the insulating COO phase at surfaces, interfaces, or grainboundaries is well documented in the literature [35–37] andcan be understood within the weakening of FM exchangeat the two-dimensional defects. Due to a single crystallinecharacter of our LPCMO/MgO(100) films (see structuraldata in Figs. 1–3) and their relatively large thickness, d �50–70 nm � RFM � 6–8 nm, grain boundaries and interfacesas locations points of CPs can be ruled out. Thus, domainwalls provide a unique intrinsic possibility to host CPs.Phenomenological Ginzburg-Landau approaches predict theformation of a charge-ordered [38] and AFM state [39] withinthe domain wall, especially when the bulk (domain) phaseis located not far away from the FM/AFM boundary in thephase diagram [4]. Our LPCMO films [27] with Pr doping, y

� 0.4, and reduced TC = 195 K � 370 K in comparison tothe double exchange FM metallic LSMO, fit nicely the aboveconditions to locate the AFM polaronic phase at the domainwalls. The T scale for low-field hysteretic CMR and for theunderlying AFM coupling in LPCMO is limited from aboveby the appearance of charge ordering at TCO ∼ 220 K [7],which breaks the symmetry of the PM phase and enhances theformation of CPs. Within the narrow temperature interval,180 K < TC < TB < TCO, a mixed phase with AFM layersand coupled FM nanodomains exists, yielding extremely largefield-induced resistance changes and metamagnetic transition.Note a similarity between the blocking temperature, TB < TCO,and the T ∗ scale [18], at which the “preformed magneticclusters” (domains) can be substantially influenced by anapplied magnetic field. A strain-free state of the LPCMOfilm and extremely sharp MI transition as an indicator forthe absence of A-site quench disorder [9] both infer that sucha mixed phase could be a true thermodynamic phase as wassuggested earlier [38]. However, a more detailed inspection ofthe nature of mixed phase goes out of the scope of this paper.

Our results, being in line with the electronic phase sepa-ration model of Burgy et al. [18], specify further the natureof competing phases, i.e., FM vs AFM. They also provideevidence for the location of the AFM phase within the domainwalls, where the FM order parameter becomes zero, changingfrom +M to −M . Indeed, a polaronic AFM phase [Fig. 7(a)]apparently resembles the “collinear” phase in Ref. [18][Fig. 2(c), “green phase”], which separates two domains withstaggered “up” and “down” magnetizations. Remarkably, even

024420-6

INTRINSIC ANTIFERROMAGNETIC COUPLING . . . PHYSICAL REVIEW B 89, 024420 (2014)

a small amount of CPs can actuate the AFM coupling betweenFM nanodomains, thus playing an important role in CMR. Themaximal concentration of CPs in LPCMO, estimated fromnonlinear electric measurements [see Fig. 7(a)], is nCP(TC)� 0.5%, and taking a homogeneous distribution of CPs withthe size δ � 1.5 nm [13,14], the average distance betweenthem will be ACP = (nCP)−1/3δ ≈ 9 nm. This fits nicely thesize of FM domains, RFM � 6–8 nm, evaluated from magneticmeasurements. Considering ACP being a few nm large, one cansuggest that an elastic (strain) mechanism [40] might play arole in actuating AFM coupling via correlated polarons. Mis-matching lattice volumes of the polaronic, Pnma, and polaron-free, R-3c, phases may further support such strain scenario.For T � TC in the globally FM metallic phase, the amountof CPs in LPCMO is strongly reduced, nCP ∼ 10−6, and thedistance between them becomes large, ACP � 100 nm � RFM,thus allowing a strain relaxation and the absence of coupling.

The coupling constant JAFM can be estimated from theslow relaxation of resistivity (Fig. 6), which actually meansthat after field removal the JAFM acts as an effective fieldto restore the AFM coupling between FM nanodomains. Forlow magnetic fields, H = 0–1 kOe, and for T ∼ TC, theresistance obeys a square field dependence, �R ∼ −H 2, butmagnetization is a linear function of the field, �M � H .Taking H as a parameter one gets �R ∼ �M2. ConsideringNeel relaxation of the magnetization, �M(t) ∼ exp(−t/τmag),with magnetic relaxation time, τmag = τ0exp(KUV/kBTB), onegets an exponential relaxation of the resistance, �R(t) ∼�M2 ∼ exp(−2t/τmag) ∼ exp(−t/τres), with τres = τmag/2.The relaxation time of the resistance in the presence of theZeeman energy, μμ0H , and AFM coupling energy of the formEAFM = JAFMA, can be written as τres(T ) = τ0/2exp[(kUV −μμ0H + EAFM)/kBT ]. For T = 197 K, the magnetic momentof the domain, μ = nμB, with n the number of Mn spinsin the volume, V = 4/3π (RFM)3, a magnetic field, μ0H =0.1 T, and the area, A = 4π (RFM)2, of a domain (RFM�7nm), the AFM coupling constant, JAFM = 4.5 × 10−4 J/m2,and the AFM energy per Mn spin, EAFM/Mn = 0.21 meV =2.4 K, were obtained. This coupling being much weaker than atypical AFM exchange energy, e.g., 10–12 meV (TN ∼ 140 Kin LaMnO3), agrees with a short-range character of the COOpolaronic phase. The value estimated above exceeds theinterlayer coupling, 0.1 erg/cm2 = 10−4 J/m2, observed inartificial FM/AFM exchange-bias manganite heterostructures[41], but is comparable with the RKKY coupling in oxidesuperlattices [30], JRKKY � 3 × 10−4 J/m2.

Such system with intrinsic AFM coupling, originated froma phase coexistence, is an insulator, which shows a resistivitymaximum at H ∼ Hc due to the antiparallel orientationof the magnetizations of adjacent nanodomains. For fields,Hc < H < Hsat, the magnetization of the nanodomains flipsalong the field direction, but the “interface” spins still remainAFM coupled, yielding the hysteretic ρ(H ) and M(H )

behavior [Figs. 2(a) and 2(b)]. Larger fields, H > Hsat �10–30 kOe, finally destroy the AFM correlations, resultingin a closure of the ρ(H ) and M(H ) curves. For the FMstate the resistance contribution due to spin scattering ofthe charge carriers at the domain wall, ρ ∼ cosθij, becomessmall. Indeed, in the absence of AFM coupling the angle θij

between magnetizations of adjacent domains decreases. Asa result epitaxial LPCMO films show a typically low residualresistivity value, ρres(4.2 K) � 10−4 cm, similar to that of thedouble-exchange LSMO [20], which contains no CPs. One cansee a striking analogy between the observed intrinsic low-fieldCMR in LPCMO films close to TC and the extrinsic interfaciallow-field tunneling magnetoresistance (TMR) in granularmanganites [42] or in the LSMO/STO/LSMO trilayers [43].They both show two resistance maxima at ±Hc originatingfrom AFM coupling between FM metallic domains (CMR) orelectrodes (TMR). However, in the TMR systems a symmetrybreak of the MnO6 network at the interface leads [35] tothe orbital reconstruction at the interface, stabilizing a highlyisolating COO-CE phase in very thin manganite films [36] andmanganite/titanite superlattices [37]. In contrast, the MnO6

frame remains continuous in an intrinsic CMR system andCPs, nucleated at the interfaces between FM domains, showpresumably short-range CO-AFM correlations, which canbe destroyed at relatively low magnetic fields of few kOe.Interesting for applications is the fact that low-field CMR wasobserved at relatively high temperatures �200 K, indicatinga persistence of high spin polarization in FM domains closeto TC.

V. CONCLUSIONS

CMR in epitaxial strain-free (La0.6Pr0.4)0.67Ca0.33

MnO3/MgO films is demonstrated to be a low-field andhysteretic effect, controlled by peculiar magnetic domainstructure with FM nanodomains, AFM coupled by correlatedJahn-Teller polarons nucleated at the domain walls. Theobtained small size of FM nanodomains, RFM ∼ 6–8 nm,taken together with the estimated coupling constant,JAFM = 4.5 × 104 J/m2, provide a rational explanation of thelong-standing issue in CMR physics, i.e., how a tiny amountof correlated polarons can be responsible for extremely largemagnetic-field-induced resistance changes.

ACKNOWLEDGMENTS

The authors thank Professor W. Felsch for fruitful com-ments and discussions, and Professor A. L. Chuvilin for FIBspecimen preparation. Financial support from the DeutscheForschungsgemeinschaft (DFG) via SFB 602 (TP A2) andSFB 1073 (TP B1 and B4), the Leibniz Program, and EUFP7 (IFOX Project) is acknowledged. K.Y.C. acknowledgesfinancial support from the Humboldt Foundation and the NRFof Korea (Grant No. 2009-0093817).

[1] C. W. Searle and S. T. Wang, Can. J. Phys. 48, 2023 (1970).[2] R. von Helmolt, J. Wecker, B. Holzapfel, L. Schultz, and

K. Samwer, Phys. Rev. Lett. 71, 2331 (1993).

[3] K. Chahara, T. Ohno, M. Kasai, and Y. Kozono, Appl. Phys.Lett. 63, 1990 (1993).

[4] N. Mathur and P. Littlewood, Phys. Today 56(1), 25 (2003).

024420-7

V. MOSHNYAGA et al. PHYSICAL REVIEW B 89, 024420 (2014)

[5] E. Dagotto, T. Hotta, and A. Moreo, Phys. Rep. 344, 1 (2001).[6] P. Schiffer, A. P. Ramirez, W. Bao, and S.-W. Cheong, Phys.

Rev. Lett. 75, 3336 (1995).[7] M. Uehara, S. Mori, C. H. Chen, and S.-W. Cheong, Nature

(London) 399, 560 (1999).[8] C. Sen, G. Alvarez, and E. Dagotto, Phys. Rev. Lett. 98, 127202

(2007).[9] V. Moshnyaga, L. Sudheendra, O. I. Lebedev, S. A. Koster,

K. Gehrke, O. Shapoval, A. Belenchuk, B. Damaschke, G. vanTendeloo, and K. Samwer, Phys. Rev. Lett. 97, 107205 (2006).

[10] C. S. Nelson et al., Phys. Rev. B 64, 174405 (2001).[11] A. S. Alexandrov and A. M. Bratkovsky, Phys. Rev. Lett. 82,

141 (1999).[12] Md. M. Seikh, C. Narayana, L. Sudheendra, A. K. Sood, and

C. N. R. Rao, J. Phys.: Condens. Matter 16, 4381 (2004).[13] C. P. Adams, J. W. Lynn, Y. M. Mukovskii, A. A. Arsenov, and

D. A. Shulyatev, Phys. Rev. Lett. 85, 3954 (2000).[14] P. Dai, J. A. Fernandez-Baca, N. Wakabayashi, E. W.

Plummer, Y. Tomioka, and Y. Tokura, Phys. Rev. Lett. 85, 2553(2000).

[15] V. Kiryukhin, T. Y. Koo, H. Ishibashi, J. P. Hill, and S.-W.Cheong, Phys. Rev. B 67, 064421 (2003).

[16] J. Tao, D. Niebieskikwiat, M. Varela, W. Luo, M. A. Schofield,Y. Zhu, M. B. Salamon, J. M. Zuo, S. T. Pantelides, and S. J.Pennycook, Phys. Rev. Lett. 103, 097202 (2009).

[17] V. Moshnyaga, K. Gehrke, O. I. Lebedev, L. Sudheendra,A. Belenchuk, S. Raabe, O. Shapoval, J. Verbeeck, G.VanTendeloo, and K. Samwer, Phys. Rev. B 79, 134413 (2009).

[18] J. Burgy, M. Mayr, V. Martin-Mayor, A. Moreo, and E. Dagotto,Phys. Rev. Lett. 87, 277202 (2001).

[19] V. Moshnyaga, I. Khoroshun, A. Sidorenko, P. Petrenko, A.Weidinger, M. Zeitler, B. Rauschenbach, R. Tidecks, and K.Samwer, Appl. Phys. Lett. 74, 2842 (1999).

[20] A. Urushibara, Y. Moritomo, T. Arima, A. Asamitsu, G. Kido,and Y. Tokura, Phys. Rev. B 51, 14103 (1995).

[21] N. A. Babushkina, L. M. Belova, D. I. Khomskii, K. I. Kugel,O. Y. Gorbenko, and A. R. Kaul, Phys. Rev. B 59, 6994 (1999).

[22] P. Levy, F. Parisi, M. Quintero, L. Granja, J. Curiale, J. Sacanell,G. Leyva, G. Polla, R. S. Freitas, and L. Ghivelder, Phys. Rev.B 65, 140401 (2002).

[23] C. P. Bean and L. D. Livingston, J. Appl. Phys. 30, S120 (1959).

[24] J. W. Lynn, R. W. Erwin, J. A. Borchers, Q. Huang, A. Santoro,J. L. Peng, and Z. Y. Li, Phys. Rev. Lett. 76, 4046 (1996).

[25] Y. Murakami, H. Kasai, J. J. Kim, S. Mamishin, D. Shindo, S.Mori, and A. Tonomura, Nat. Nanotechnol. 5, 37 (2010).

[26] N. D. Mathur, M.-H. Jo, J. E. Evets, and M. Blamire, J. Appl.Phys. 89, 3388 (2001).

[27] V. Moshnyaga and K. Samwer, Ann. Phys. (Berlin) 523, 652(2011).

[28] S. S. P. Parkin, R. Bhadra, and K. P. Roche, Phys. Rev. Lett. 66,2152 (1991).

[29] W. Pan, N.-Y. Jih, C.-C. Kuo, and M.-T. Lin, J. Appl. Phys. 95,7297 (2004).

[30] K. R. Nikolaev, A. Y. Dobin, I. N. Krivorotov, W. K. Cooley,A. Bhattacharya, A. L. Kobrinskii, L. I. Glazman, R. M.Wentzovitch, E. D. Dahlberg, and A. M. Goldman, Phys. Rev.Lett. 85, 3728 (2000).

[31] P. Gaunt, Philos. Mag. 48, 261 (1983).[32] D. S. Bethune, R. W. Smith, and Y. R. Shen, Phys. Rev. A 17,

277 (1978).[33] M. N. Iliev and M. V. Abrashev, J. Raman Spectrosc. 32, 805

(2001).[34] K. Y. Choi, P. Lemmens, T. Sahaouri, G. Guntherodt, Y. G.

Pashkevich, V. P. Gnezdilov, P. Reutler, L. Pinsard-Gaudart, B.Buchner, and A. Revcolevschi, Phys. Rev. B 71, 174402 (2005).

[35] L. Brey, Phys. Rev. B 75, 104423 (2007).[36] A. Tebano et al., Phys. Rev. Lett. 100, 137401 (2008).[37] K. Gehrke, V. Moshnyaga, K. Samwer, O. I. Lebedev, J.

Verbeeck, D. Kirilenko, and G. VanTendeloo, Phys. Rev. B 82,113101 (2010).

[38] G. C. Milward, M. J. Calderon, and P. B. Littlewood, Nature(London) 433, 607 (2005).

[39] M. S. Rzchowski and R. Joynt, Europhys. Lett. 67, 287 (2004).[40] K. H. Ahn, T. Lookman, and A. R. Bishop, Nature (London)

428, 401 (2004).[41] A. L. Kobrinskii, A. M. Goldman, M. Varela, and S. J.

Pennycook, Phys. Rev. B 79, 094405 (2009).[42] H. Y. Hwang, S.-W. Cheong, N. P. Ong, and B. Batlogg, Phys.

Rev. Lett. 77, 2041 (1996).[43] Y. Lu, X. W. Li, G. Q. Gong, G. Xiao, A. Gupta, P. Lecoeur,

J. Z. Sun, Y. Y. Wang, and V. P. Dravid, Phys. Rev. B 54, R8357(1996).

024420-8