Role of the stability of charge ordering in exchange bias effect in doped manganites

In this work we have carried out an elaborate study on the magnetic properties and investigated the exchange bias phenomena of some charge-ordered (CO) manganites. The detailed study of Sm1−xCaxMnO3 (x = 0.5, 0.55, 0.6, 0.65, 0.7) compounds shows that Sm0.4Ca0.6MnO3, which is the most robust charge ordered material studied here, shows significantly large exchange bias field (HE) as compared to the other compounds. Our experimental results and analysis indicate that TCO, which reflects the stability of the charge-ordered state, is one of the key parameters for the exchange bias effect. Similar behaviour is found in other rare-earth analogues, viz., La1−xCaxMnO3 and Pr1−xCaxMnO3 compounds as well. We also found that with increasing stability of CO states in Sm1−xCaxMnO3 compounds, HE enhances due to increase in number and reduction in size of ferromagnetic clusters.

Ferromagnets, under exposure to a cyclic magnetic field, exhibit a magnetic hysteresis loop (M-H loop) resulting in energy storage in the system 1 . Such loops appear due to irreversible domain wall motion/domain rotation caused by various pinning centres like dislocations and inclusions present in the system. In the permanent magnets, M-H loops are generally symmetric in nature with appreciable remnant magnetization and coercive field. Isothermal magnetic hysteresis phenomenon has been utilized in many technological applications, such as data processing, electronic, automobile, aerospace and biosurgical industries etc. 2 . However, for a few material, M-H loops have been found to be asymmetric in magnetic field direction, and such an asymmetry has been assigned to the presence of exchange bias (EB) phenomenon 3,4 . The asymmetry parameter is called exchange bias field (H E ) and is defined as (H C1 − H C2 )/2 where H C1 and H C2 are coercive fields in the positive and negative field axis. EB is commonly observed in systems that contain phase boundaries separating antiferromagnetic (AFM) and ferromagnetic (FM) regions 3 . The exchange interaction at such magnetic phase boundaries would lead to unidirectional pinning of interfacial spins. In this case, reversal of the oriented FM moment needs an additional energy, leading to the shift in the switching field, thereby introducing asymmetry in M-H loop. Systems with large H E have found several application in different magnetic sensors and read-heads for spintronic applications 5,6 . EB is observed in various configurations viz, core-shell nanostructures, layered structures, two phase nanocomposits, inhomogenous materials etc. 4,[7][8][9] . In all these cases, artificially engineered phase boundaries between two or more crystallogaphically or chemically different phases have been used to achieve the magnetic inhomogenity required for introducing EB. However, such multiple-phase-compounds often found to exhibit compositional variations in the physical scale of micron or sub-micron range, which may affect the reproducibility of EB data measured in different batches of same material 10 . There are, nevertheless, some single phase compounds such as manganites, cobaltates, etc., known to exhibit EB, where magnetic inhomogenity are an inherent characteristics originating due to electronic phase separation 11 .
Electronic phase separation (hence onwards referred as phase separation in rest of this paper) has been observed and studied extensively in partially doped manganites [11][12][13] . Such doped manganites R 1−x A x MnO 3 (R = Rare-earth) are obtained by substituting divalent alkaline-earth ions such as A = Ca 2+ , Ba 2+ and Sr 2+ at the trivalent rare earth site of the parent RMnO 3 13 . In the parent compound, the manganese ions exist in Mn 3+ state whereas in doped compounds both Mn 3+ and Mn 4+ ions exist in the ratio of (1-x): x. This mixed-valent character of Mn ions underlies the rich phase diagrams exhibited with different ground states 13 . For example, one may get metallic or insulating state as well as FM, AFM, canted AFM or charge-ordered (i.e. a periodic arrangement of Mn 3+ and Mn 4+ ions in the lattice) states depending on the percentage of doping 13 . The presence of more than one of such electronic phases in an otherwise chemically homogenous system gives rise to phase separation. In phase separated manganites, EB was first reported in charge-ordered (CO) compound Pr 1/3 Ca 2/3 MnO 3 14 . In this CO system, the regular arrangement of Mn 3+ and Mn 4+ ions also gives rise to AFM ordering. This CO/AFM system, sometime found to contain small inclusions of slightly different electronic phases, giving rise to FM behaviour in the form of nanodomains 14 . Since the exchange bias originate from the pinned interface spins between adjacent FM and AFM domains, H E depends on the surface area of the FM clusters dispersed in CO/AFM matrix. Thus, it is expected that in a bulk compound, H E should increase as the surface area of the FM clusters increases. This can be achieved by decreasing their size and increasing their numbers. To have large H E in the bulk system, such large surface area is essential and can be obtained by arresting the growth of FM cluster size. This is possible in a robust CO system where the strong AFM interaction would restrict the size and growth of the FM clusters. The restriction imposed by the charge conservation would lead to an increase in the number of FM clusters of very small sizes throughout the material. Therefore, in this work, we propose that the robustness/stability of CO system, that is defined by the magnetic field required to melt the CO-state, must be an important parameter in deciding the value of H E in bulk CO manganite. Earlier studies on R 1−x Ca x MnO 3 (R = Pr-Sm) compounds showed that the CO melting field and charge ordering temperature (T CO ) increases with heavier rare earth analogues, indicating increase in the stability of CO state for a particular doping concentration 15,16 . The largest value of CO melting field in the R 1−x Ca x MnO 3 has been reported so far in Sm 0.5 Ca 0.5 MnO 3 (~60 T) suggesting it to be the strongest CO system studied till now 16 . Furthermore, for a particular rare earth, the CO melting field i.e. the strength of CO, increases with increasing T CO (see Fig. 13 of ref. 14). So, one would expect even larger CO melting field in Sm 0.4 Ca 0.6 MnO 3 (reportedly having maximum value of T CO in Sm 1−x Ca x MnO 3 series 17 ), though no such studies have been reported in literature. In this work, we focus on Sm 1−x Ca x MnO 3 (x = 0.5, 0.55, 0.6) and study the evolution of H E with x. We show that the bulk Sm 0.4 Ca 0.6 MnO 3 compound which has relatively larger CO stability, indeed exhibit considerably higher value of H E than that observed in Sm 0.5 Ca 0.5 MnO 3 . Furthermore, H E is found to reduce drastically in nanosized material of Sm 0.4 Ca 0.6 MnO 3 , where the CO strength is remarkably weakened. We have also studied H E values of R 1−x Ca x MnO 3 (R = La, Pr) for compositions having maximum T CO and compared these values with that of Sm 0.4 Ca 0.6 MnO 3 compound to test the validity of our proposal.

Results and Discussion
The bulk samples of Sm 1−x Ca x MnO 3 (x = 0.5, 0.55, 0.6, 0.65, 0.7) and nano-sized samples of Sm 0.4 Ca 0.6 MnO 3 (Sm 0.4 Ca 0.6 MnO 3 -nano) have been prepared by sol-gel technique. All the peaks in the X-ray diffraction (XRD) patterns of these compounds at room temperature could be indexed by orthorhombic structure with Pnma space group, suggesting single phase nature of the samples (details are given in the section 1 of Supplementary information part). The lattice parameters obtained for x = 0.5, 0.55 and 0.6 compositions are a = 5.364 Å, 5.355 Å and 5.349 Å, b = 7.561 Å, 7.552 Å and 7.532 Å, c = 5.420 Å, 5.404 Å and 5.387 Å, respectively.
Temperature dependence of magnetization, M(T), of all these samples studied in the temperature range of 5 K to 330 K under zero field cooled (ZFC) and field cooled (FC) conditions at H = 100 Oe are shown in Fig. 1(a-d). For all the samples, M(T) shows a monotonously increasing behavior as the temperature is lowered from 330 K to ~270 K till a peak is observed at T CO ~ 270 K. It has been argued that although the material remains paramagnetic both above and below T CO , the magnetic interaction changes from FM (T > T CO ) to AFM (T < T CO ) at the charge ordering temperature 18 . This results in a peak-like structure at T CO in the magnetic susceptibility. A close look into the M(T) data of Sm 0.4 Ca 0.6 MnO 3 -nano ( Fig. 1a) reveals that the peak-like structure is very weak, indicating the suppression of CO as well as AFM interaction to a great extent. This results in a relative enhancement of magnetic susceptibility in the Sm 0.4 Ca 0.6 MnO 3 -nano at low temperatures vis-à-vis that of corresponding bulk materials. Charge ordering feature in these manganites is also manifested through the occurrence of an anomaly in dielectric constant around T CO (Inset I: Fig. 1a). Generally, it has also been reported that in similar systems CO is followed by CE-type insulating AFM phase at lower temperature 16 . This is also observed in our systems where the M(T) curves exhibit a discernible hump indicating AFM ordering below T CO (insets I of Figs. 1a,c,d and inset of Fig. 1b). Below the AFM ordering, M(T) under ZFC and FC configuration exhibit irreversible behavior (T irr ~ T N ). With the increasing magnetic field, the magnetic irreversibility reduces. Observation of such magnetic irreversibility in an AFM material suggests the presence of additional FM interaction. Below T N , ZFC M(T) curve for both the bulk and nano samples measured at a very low field of 100 Oe show an increasing trend with decreasing temperature indicating that the FM interactions are spontaneous and not field induced. The magnetization value of Sm 0.4 Ca 0.6 MnO 3 -nano below T N appears to be quite high relative to the bulk compounds. Such an enhancement of magnetization in nano material is due to increasing ferromagnetic component arising from the uncompensated surface spins in comparison with the AFM core.
On further decreasing the temperature, Sm 0.4 Ca 0.6 MnO 3 shows another cusp in the M(T) curve at 35 K (~T f ), much below T N (Fig. 1c). This low temperature cusp, observed earlier in quite a few manganite systems had been argued due to the cluster glass behaviour, where T f represents the spin freezing temperature 19,20 . Among all the compounds studied here, such spin freezing effect, although quite prominent for Sm 0.4 Ca 0.6 MnO 3 , found to be present in other bulk compounds as well. The temperature derivative of M(T) exhibit an anomaly around T f indicating the spin freezing behaviour in the Sm 1−x Ca x MnO 3 (x = 0.5, 0.55) bulk compounds studied here (inset II: Fig. 1c,d). This effect, however, has not been observed in Sm 0.4 Ca 0.6 MnO 3 -nano, within the resolution limit of measurement.
Another important fact should mention here which is the effect of the external magnetic field on the transport properties of robust charge-ordered compound. In Sm 1−x Ca x MnO 3 (x ~ 0.5), since the charge ordering state is Scientific RepoRts | 7: 3220 | DOI:10.1038/s41598-017-03451-z very robust, the insulating state at the low temperature is almost unaffected even in the presence of 90 kOe applied magnetic field. The resistivity as a function of temperature of some selected compounds in the absence and in the presence of magnetic field (90 kOe) is shown in supplementary part.   (Fig. 3c). An important property usually observed in the EB system is the magnetic training effect in which the H E and remanence asymmetry (M E = (M R1 + M R2 )/2) values are known to decreases as the system undergoes successive field cycling at a particular temperature after field cooling 19 . Furthermore, the variation of H E with the loop index number (λ) follows the relation (H E − H E∞ ) ∝ 1/√λ, where H E∞ is the EB field at λ = ∞ 19 . It has been observed that similar behaviour is followed in bulk Sm 0.4 Ca 0.6 MnO 3 compound when field cycled at 5 K (Fig. 4c,d), thus confirming the presence of exchange bias effect in bulk Sm 1−x Ca x MnO 3 compounds studied here. The estimated H E values found to be ~1250 Oe, ~2080 Oe, ~2470 Oe, ~960 Oe and ~537 Oe at 5 K for these bulk samples with x = 0.5, 0.55, 0.6, 0.65 and 0.7 respectively. Thus, H E values of the bulk Sm 1−x Ca x MnO 3 (x = 0.5, 0.55, 0.6, 0.65, 0.7) compounds qualitatively mimics the variation of T CO with x similar to the reported phase diagram 17 (Fig. 4a). Furthermore, these compounds show large H E values. On the contrary, Sm 0.4 Ca 0.6 MnO 3 -nano does not show any loop shift (H E ~ 0 Oe), indicating that H E vanishes as CO is suppressed drastically (Fig. 3d). Furthermore, it is known from previous studies that CO stability and T CO increases from R = La to Sm in R 0.5 Ca 0.5 MnO 3 series 15 , and for a particular R-ion CO stability increases with increasing T CO (see Fig. 13 of ref. 14). Accordingly we have estimated the H E values of La 0.32 Ca 0.68 MnO 3 and Pr 0.4 Ca 0.6 MnO 3 compounds that have the largest T CO (most stable CO for a particular rare earth) in the La 1−x Ca x MnO 3 and Pr 1−x Ca x MnO 3 series respectively 11 . The H E values of these two compounds found to be 625 Oe and 1200 Oe which are much less than that of Sm 0.4 Ca 0.6 MnO 3 (H E ~ 2470 Oe) supplementing our proposal that H E increases with increasing CO stability.
To reveal the origin of EB effect and the increase in the values of H E with increasing CO stability and x value in Sm 1−x Ca x MnO 3 (x = 0.5, 0.55, 0.6) system, the temperature dependence of H E has been studied (Fig. 4b). H E obtained for Sm 0.4 Ca 0.6 MnO 3 compound up to a temperature of 150 K shows that above T irr , H E becomes temperature independent with a negligible value. Below T irr , H E increases slowly with decreasing temperature and then rises at a much faster rate below the spin freezing temperature T f ~ 35 K. Similar variation of H E is also observed for Sm 0.5 Ca 0.5 MnO 3 compound (Fig. 4b) which too exhibits a signature of spin freezing in the M(T) curve as shown in inset II of Fig. 1d. Increase in H E below 35 K was also observed previously in Pr 1/3 Ca 2/3 MnO 3 manganite system (which do not exhibit any prominent cusp at low temperature) and was ascribed to the glassy behavior of the compound 14 . This indicates that the EB effect originates below T irr (~T N ), i.e. with the formation of FM clusters and increases drastically below freezing temperature in all the bulk Sm 1−x Ca x MnO 3 (x = 0.5, 0.55, 0.6) compounds. To confirm the presence of ferromagnetic clusters and glassy behavior further measurements, as described below, have also been carried out. Magnetic relaxation experiments on Sm 1−x Ca x MnO 3 (x = 0.5, 0.55, 0.6) were carried out to confirm the presence of FM clusters. We have measured the isothermal remnant magnetization (IRM) as a function of time at T = 5 K, after cooling the sample in ZFC condition, followed by cycling in an external field of 0 Oe → 1 kOe → 0 Oe. M(t) data for Sm 1−x Ca x MnO 3 (x = 0.5, 0.55, 0.6) shown in Fig. 5(a-c), relaxes as M(t) = M 0 + M 1 exp(−t/ t 1 ) + M 2 exp(−t/t 2 ) + M 3 exp(−t/t 3 ), where M 0 is the time independent and M 1 , M 2 , M 3 are three time dependent fractions relaxing with time constants t 1 , t 2 , and t 3 , respectively 21 . It can therefore be said that our relaxation experiment confirms the presence of multiple relaxation times, as expected for a collection of FM clusters.
The presence of FM clusters in an antiferromagnetic matrix is expected to exhibit glassy behavior. To confirm the glassy nature of the system, the ac susceptibility measurement has been carried out for the compound Sm 0.4 Ca 0.6 MnO 3 at different frequencies. The ac susceptibility exhibit a peak around 35 K, and with increase in frequency the peak temperature shifts towards higher temperatures ( Fig. 6(a)). This type of frequency dependence of magnetic transition is typical for glassy systems and the corresponding peak temperature as f → 0 is defined as the freezing temperature. In typical glassy system, the relative shift in freezing temperature per decade of frequency is determined by ref. 22 The calculated value of δT f is found to be ~0.12 for the compound, which is typically in the range of that found in cluster glass systems 23 . The frequency dependence of the glassy systems also follows conventional power-law divergence of critical slowing down where τ is relaxation time corresponding to the measured frequency (τ = 1/ν), τ 0 is single spin flip relaxation time, T SG is the freezing temperature corresponding to zero frequency and zν′ is the dynamic critical exponent. The estimated value of zν′ is 4.2 and τ 0 ≈ 10 −10 sec [ Fig. 6(b)] from the ac susceptibility data again found to be in the same range to those of cluster glass compounds 24 .
Another dynamical law for glassy systems is the Vogel-Fulcher law, where ν 0 is the attempt frequency, E a is the activation energy and T 0 is the Vogel-Fulcher temperature. The best fitted experimental data yields E a /K B = 38.9 K and T 0 = 33.6 K [ Fig. 6(c)].
The glassy feature may also be demonstrated through dc magnetization measurement viz. memory effect as well as magnetic relaxation. Memory effect has been observed in bulk Sm 0.4 Ca 0.6 MnO 3 compound [see Supplementary for full information]. The presence of cluster glass state in this compound is further proved by the analysis of temperature dependent magnetic relaxation data. According to Ulrich et al. 25 , the relaxation rate of remanent magnetization in a system consisting of magnetic clusters obey the decay law, W(t) = Bt −n , t ≥ t 0 , where B is a constant, t 0 is crossover time and 'n' is a measure of the dipolar interaction strength among the magnetic clusters involved in the relaxation process.
In case of canonical spin glass, the value of 'n' remain almost constant, whereas it changes in case of cluster glass systems. In our case, 'n' is found to be sensitive to both temperature as well as applied magnetic field (Fig. 7a,b). Such dynamic behaviour 'n' is also found for other two bulk compounds studied here. The fraction of glassy component in this compound thus certainly not static but found to be dependent on both temperature and magnetic field, thus confirming the presence of cluster glass state in bulk Sm 1−x Ca x MnO 3 compounds.
All these results firmly establish cluster glass behaviour with the freezing of FM spins below T f in the bulk Sm 1−x Ca x MnO 3 (x = 0.5, 0.55, 0.6) compounds. The interface exchange coupling between these FM clusters and the CO AFM background lead to unidirectional shift of the hysteresis loops i.e. EB effect. Additionally, we have also seen that H E depends strongly on x where H E(x=0.4) ~ 2 H E(x=0.5) at 5 K. To understand the increase of H E with x, we have estimated the relative FM and AFM/PM fraction from the isothermal magnetization data using the relation, M(H) = P tanh(QH) + SH, where, P, Q and S are constants. The fraction remains almost conserved in all the three compounds suggesting that the FM volume fraction is not responsible for the increase of H E with increasing x. The above result leads to investigate the number and size distribution of the clusters with variation of x. Since H E is known to depend primarily on the FM and AFM interface area rather than the FM volume fraction, a system consisting of smaller clusters in larger number is expected to exhibit larger H E than systems with larger clusters with smaller number.
The average size of short range ferromagnetic clusters can be generally estimated following the formula given by Niebieskikwiat and Salamon in several phase separated manganite and cobaltite systems 14 . According to them, the cooling field (H cool ) dependence of H E can be expressed by the relation, where J i is the interface exchange constant, g ≈ 2 is the gyromagnetic factor, μ B is the Bohr Magneton, μ 0 ≈ 3μ B is the magnetic moment of the Mn core spin, L is the Langevin function, μ = N ν μ 0 is the magnetic moment of ferromagnetic cluster, N ν is the number of spins in a cluster and T f is the temperature below which glassy state exists. The first term in relation 4 dominates for small H cool and H E depends on J i 2 . However the second term dominates for large H cool and for J i < 0 the value of H E decreases for higher H cool . M(H) measured at different cooling fields (H cool ) for bulk Sm 1−x Ca x MnO 3 (x = 0.5, 0.55, 0.6) samples (Fig. 8) shows that H E initially increases with H cool and thereafter exhibits a decreasing tendency above H cool ~ 8 T. Figure 8 show the H E as a function of H cool along with the fit to Eq. 4 with an overall scale factor and J i , N ν as adjustable parameters for all the bulk compounds. The exchange constant obtained from the best fit to the experimental data yields to be negative for all the samples, indicating the existence of FM clusters in AFM host (similar to that discussed earlier). For Sm 0.5 Ca 0.5 MnO 3 compound, we have deduced the value of N ν ≈ 14. We have also estimated the value of N ν from the earlier reported H EB vs. H cool data for this particular system 19 9(a-c). For more generalization we have plotted the H E as a function of Ca-doping even for higher Ca concentration than x = 0.6. Our experimental study clearly reveals that with the increasing of the Ca-doping up to x = 0.6 H E increases, then decreases for x > 0.6 compositions qualitatively similar as T CO (x). Furthermore, using the size of FM clusters, exchange anisotropy constant can be calculated using the expression 14 where ν 0 ~ 10 9 s −1 is the switching attempt frequency, τ ~  Fig. 3. In the compounds studied here, the value of the anisotropy constant estimated at a temperature 5 K are 2.3 × 10 6 erg/cm 3 , 3.3 × 10 6 erg/cm 3 and 4.1 × 10 6 erg/cm 3 for x = 0.5, 0.55 and 0.6, respectively. These values are considerably larger than that reported for bulk FM manganites 26 . The effect of the variation of the size of FM clusters, i.e., the effect of magnetic anisotropy can be qualitative obtained using Meiklejohn and Bean model 7 , estimated for FM/AFM thin films using the relation originally proposed for PM/AFM multilayer thin films. According to this model, H E can be expressed as where J is the exchange integral across the FM/AFM interface per unit area, S AFM and S FM are the interface magnetizations of the AFM and FM layers, respectively, and t FM and M FM are, respectively, the thickness and magnetization of the FM layer. By replacing the FM layer thickness with FM cluster size, the same model has been extensively used in phase separated manganite systems 27 . In our case, where FM clusters are dispersed in AFM matrix, as the size of FM clusters decreases, the term t FM decreases. Furthermore, the decrease of cluster size with increasing x leads to the decrease in magnetization of the FM clusters. As both the terms in the denominator of Eq. (5) decreases with increasing x value, H E increases. Hence, it can be shown qualitatively that with the decreasing size of FM clusters with corresponding increase in values of x in Sm 1−x Ca x MnO 3 (x = 0.5, 0.55, 0.6) compounds, H E increases.
To conclude, we have correlated exchange bias field (H E ) in CO manganites with the stability of charge ordering. For this, we have undertaken a detailed study of magnetic properties of CO Sm 1−x Ca x MnO 3 (x = 0.5, 0.55 and 0.6) bulk compounds and found that with increasing x value the T CO (and hence stability of CO) and the corresponding H E increases. Due to suppression of CO in nano form of Sm 0.4 Ca 0.6 MnO 3 , H E is reduced drastically as compared to that in bulk. We have estimated the size and density of the FM clusters (which introduces EB in the CO AFM system) and found that with increasing x, exchange anisotropy constant and density of clusters increases while their size decreases. We have also found that H E mimics the strength of CO stability in R 1−x Ca x MnO 3 as R changes from La to Sm. From our observation it appears that the increase in H E with increasing CO stability may be a general behaviour for all CO systems and can be utilized in engineering large exchange bias materials.

Methods
Polycrystalline samples of Sm 1−x Ca x MnO 3 (x = 0.5, 0.55, 0.6) have been prepared by well known sol-gel technique. Sm 2 O 3 , CaCO 3 and MnO 2 were used as starting materials. Appropriate amount of oxides were separately dissolved in a HNO 3 solution (Oxalic acid is also added in case of MnO 2 ). These solutions were mixed, and an amount of citric acid equivalent to the total number of moles of metal ions was added under moderate heating and stirring conditions. Subsequently mixture was slowly evaporated at 80-90 °C in water bath resulting in the formation of gel which was heated to 250 °C to remove the organic matter and decompose the nitrates of the gel. The black ash thus obtained was grinded and further heated to 550 °C for 5 h to kick off the remaining organic matter and then pelletized. The pellets then finally sintered for 24 hours at 1250 °C to obtain bulk samples. Nanoparticles of Sm 0.4 Ca 0.6 MnO 3 (Sm 0.4 Ca 0.6 MnO 3 -900) has been prepared by sintering the pellets at 900 °C for 3 hours. X-ray diffraction (XRD) study was carried out, with Rigaku diffractometer in Brag-Brentano geometry, using Cu-Kα source having wavelength 1.54 Å. Magnetization was measured using VSM-SQUID magnetometer (Quantum Design). Some of the Magnetization measurements were carried out on a vibrating sample magnetometer (VSM) coupled to a 9 Tesla physical property measurement system (PPMS) (Quantum Design, USA) at UGC-DAE Consortium for Scientific Research, Mumbai Centre.