Magnetic scattering and electron pair breaking by rare-earth-ion substitution in BaFe2As2 epitaxial films

The effect of electron doping by trivalent charge state rare-earth ion (RE = La, Ce, Pr, and Nd) substitutions on the superconductivity in BaFe2As2 was examined using epitaxial films. Each of the RE substitutions suppressed the resistivity anomaly associated with the magnetic/structural phase transitions, leading to the resistivity drops and superconductivity transitions. Bulk superconductivity was observed at the maximum onset critical temperature (Tconset) of 22.4 K for La-doping and 13.4 K for Ce-doping, while only broad resistivity drops were observed at 6.2 K for Pr-doping and 5.8 K for Nd-doping but neither zero resistivity nor distinct Meissner effect were observed at least down to 2 K. The decrease in Tconset with increasing the number of RE 4f electrons cannot be explained in terms of the crystalline qualities or crystallographic structure parameters of the BaFe2As2 films. It was clarified, based on resistivity-temperature analyses, that magnetic scattering became increasingly significant in the above order of the RE dopants. The negative magnetoresistance was enhanced by the Ce- and Pr-doping, implying that the decrease in Tc originates from magnetic pair breaking by interaction of the localized 4f orbitals in the RE dopants with the itinerant Fe 3d orbitals.


Introduction
In the few months since the report on high critical temperature (T c ) superconductivity at 26 K in the 1111-type iron pnictide LaFeAs(O 1x F x ) [1], 122-type AEFe 2 As 2 (where AE = alkaline earth) [2] has also joined the family of iron-based high-T c superconductors as a parent material. To induce its high-T c superconductivity, both types of doping carrier, i.e. holes and electrons, are typically used by selection of appropriate aliovalent dopants. To date, most such carrier doping processes have been performed by substituting the AE sites with alkali metals with different ion charges, such as K (e.g., in hole-doped (Ba 1x K x )Fe 2 As 2 ) [2], and by substituting the Fe sites with transition metals with excess 3d electron numbers, such as Co (e.g., in electron-doped Ba(Fe 1x Co x ) 2 As 2 ) [3]. The doping sites are categorized into two modes for 122-type AEFe 2 As 2 , i.e., "indirect doping" for doping at sites other than the Fe sites, and "direct doping" for doping at the Fe sites, because the AE and the FeAs layers are spatially separated and the Fermi level is composed mainly of Fe 3d orbitals [4].
The maximum T c value for each parent material of the iron-based superconductors was obtained by indirect doping [2],[57] because direct doping has a major influence on carrier transport in the conducting Fe layer. It was therefore expected that a new indirect electron doping mode for AEFe 2 As 2 , e.g. at the AE sites, would lead to high-T c superconductors, similar to the effects of other indirect doping methods, such as K doping at AE sites (maximum T c = 38 K) [2] and isovalent P doping at As sites (maximum T c = 31 K) [8]. However, indirect electron doping of AEFe 2 As 2 by substituting the AE sites with trivalent rare earth (RE) ions was difficult to perform by conventional solid-state reactions; only Ba [9] and Sr [10] have been reported. This difficulty was attributed to electronic instability arising from the high localized density T. Katase et al. 4 of states at the Fermi level, as predicted for La-doped AEFe 2 As 2 [(AE 1x La x )Fe 2 As 2 ] by density functional theory calculations [11,12].
Under such circumstances, indirect RE doping of AEFe 2 As 2 was achieved by applying a high-pressure synthesis process for (Sr 1x La x )Fe 2 As 2 polycrystals [10] and a melt-growth process using a flux agent for (Ca 1x RE x )Fe 2 As 2 (where RE = La -Nd) single crystals [1316]. Among these materials, it should be noted that a Pr-doped CaFe 2 As 2 single crystal demonstrated a maximum T c of 49 K [14], which is the highest reported T c among the 122-type AEFe 2 As 2 series, although their shielding volume fractions are as low as <1 % at 40 K [13].
In contrast to the AE = Ca and Sr systems, there had been no reports on RE-doped BaFe 2 As 2 , probably because of the large difference between the ion radii of Ba 2+ (142 pm for 8-fold coordination) and RE 3+ (111116 pm) [17]. However, we recently succeeded in indirect La doping at the Ba sites in BaFe 2 As 2 by using a film-growth process [18]. This success was attained by using the highly non-equilibrium nature of thin film deposition via the vapor phase. The maximum T c (22.4 K) of the (Ba 1x La x )Fe 2 As 2 films was much lower than that of the RE-doped CaFe 2 As 2 single crystal, and its electronic phase diagram was almost the same as that of directly-doped Ba(Fe 1x Co x ) 2 As 2 . On the other hand, according to the scenario in (Ca 1x RE x )Fe 2 As 2 crystals where a smaller RE dopant than La, such as Pr, leads to a higher T c , we expected that substitution of smaller RE ions with open-shell 4f electrons for the Ba sites would provide a higher T c than that attained for (Ba 1x La x )Fe 2 As 2 .

Experimental
(Ba 1x RE x )Fe 2 As 2 thin films were grown directly on MgO (001) single crystals with a PLD system using a second harmonic of a Nd:YAG laser as the excitation source [19]. Bulk polycrystalline samples of RE (La, Ce, Pr, Nd, and Sm)-containing BaFe 2 As 2 were used as PLD targets (see supplementary Fig. S1 for the synthesis of the bulk polycrystals). Powder X-ray diffraction (XRD, using CuK anode radiation) analyses confirmed that the crystalline phases of the bulk polycrystals were mixtures of

Results
A. Growth of (Ba 1x RE x )Fe 2 As 2 thin films Figure 1 shows the out-of-plane XRD patterns for films of (a) (Ba 1x Ce x )Fe 2 As 2 , and only the segregation of SmAs impurities was observed ( Fig. 1(d)), indicating that the incorporation of the smaller Sm ions into BaFe 2 As 2 was unsuccessful.
B. Structural characterization with x film = 0.28), while the shrinkage of the a-axis length was very small (the largest a/a was ~ -0.3%). Consequently, V decreases monotonically as x film increases. These results, which are similar to those obtained for (Ba 1x La x )Fe 2 As 2 [18], substantiate the fact that the RE 3+ ions substitute the Ba 2+ sites in the epitaxial films. It is noteworthy that the c-axis shrinkage increases in the order of Nd, Pr, and Ce dopants, when compared with the same x film , but this result is inconsistent with the differences in the ion radii of the RE 3+ ions (their radii decrease in the order of Ce 3+ (114 pm), Pr 3+ (113 pm), and Nd 3+ (111 pm) because of the lanthanide contraction) [17]. Additionally, the variation in the a-axis length was independent of the ion radii.  gradually increased as x film increased, which may originate from the structural strains and/or distortions in the films, probably because of large ion-size mismatches between the Ba 2+ ion and the doped RE 3+ ions. However, it is safely concluded that the structural quality is similar for all RE dopants.
C. Transport and magnetic properties comparison. The  of an undoped BaFe 2 As 2 film decreases with decreasing T from 300 K, and falls rapidly from ~150 K, whose resistivity anomaly is associated with magnetic/structural transitions [23]. As seen in supplementary Fig. S3, a d/dT -T curve provides a clear peak and resistivity anomaly temperatures (T anom ). It should be noted that the anomalous temperature range around T anom for the undoped BaFe 2 As 2 epitaxial film is broader than that of a single crystal [24]. However, the crystalline quality of this film ( of the 002 diffraction ~ 1 deg.) is almost the same as that of the single crystal ( of the 002 diffraction = 0.7 deg.) [24]. Further, it is reported that a sharp d/dT curve similar to that of the single crystal is observed even in a polycrystal [25]. These results indicate that the broader magneto-structural transition of this film does not originate from a crystalline quality issue. It is reported that a small in-plane stress applied to the BaFe 2 As 2 single crystal broadens the structural transition due to the de-twinning of the crystals [26]; therefore, we speculate that a lattice-strain effect at the epitaxial film -substrate interface would be an origin of the broadening. In all dopant cases,  at 300 K gradually decreased, and T anom shifted to lower temperatures as x film increased.
In the Ce-doping case (a), a resistivity drop without zero resistivity was observed for x film = 0.09, but the resistivity anomaly was still observed at T anom = 70 K. With a further increase in x film , the resistivity anomaly was not detected in the -T curves, and T c onset for a superconductivity transition appears at x film  0.09. T c onset reached a maximum value of 13.4 K at x film = 0.15. The resistivity transition width (defined by where T c offset is the offset critical temperature determined by extrapolating a T curve to zero resistivity) of this film is 4.5 K, which is doubly larger thanT c = 2.7 K of La-doped films [18] although the crystalline qualities are almost the same as seen in ,  and 2 in Figs. 2(b,c), and the dopant distribution is homogeneous, which is confirmed both by EPMA (see supplementary Fig. S2(a)) and XRD (peak shift and broadening are not observed by Ce doping). Thus, the broad resistivity transition reflects a wide vortex-liquid-phase due to strong vortex pinning centers, similar to that of the Co-doped BaFe 2 As 2 epitaxial film [27,28]. The wide liquid phase is due to disorders in the films, which may also the case in the Ce-doped films.  D. Electronic phase diagrams Figure 5 summarizes the electronic phase diagrams for the (Ba 1x RE x )Fe 2 As 2 epitaxial films. The electronic phase diagram of (Ba 1x La x )Fe 2 As 2 epitaxial films is also shown in the figure for comparison [18]. For all RE dopants, the antiferromagnetic (AFM) transition at T anom is suppressed as x film increases, but the suppression rate of T anom is different for different dopants; i.e., it decreases in the order of RE = La, Ce, Pr, and Nd. The vertical arrows, which indicate the extrapolated intersections at the x film axis by using a phenomenological fit to T anom , indicate this trend more clearly. The x film , as shown in Fig. 2(a). The inset figure in Fig. 5 replots T anom and T c onset against the c-axis lattice parameters for the (Ba 1x RE x )Fe 2 As 2 epitaxial films. The suppression curves of T anom converge to a single line, and the peak position of the superconducting dome converges to the same c-axis length of 1.285 nm. This result implies that both chemical pressure and carrier doping are effective for suppression of the AFM transition and induction of superconductivity in the BaFe 2 As 2 system [29].

Discussion
First, we should remind that the amounts of the Fe impurity in the deviation of the AsFeAs bonding angle from a regular tetrahedron is highly correlated with the maximum T c values achieved [30]. Theoretical works have also shown that the Fermi surface topology is highly sensitive to the height of the As atom from the Fe sheet [31]. BaFe 2 As 2 has a tetragonal crystal structure with a space group of I4/mmm, in which only the z-coordinate of the As (z As , at the Wyckoff position 4e) site is a variable parameter. The height of the As atom from the Fe sheet (h As = c(z As 0.25)) and the As-Fe-As angle ( = 2tan 1 (a/2h As )) can therefore be estimated if the a-axis and c-axis lattice parameters and z As are determined. It has been reported that z As can be estimated from the ratio of the integrated intensities of 00l diffractions of epitaxial films in the out-of-plane XRD patterns [32]. In the Ce-doping case, it was found that z As only changed from z As = 0.3545 at x = 0 to 0.352, even when x film =0.29, because the intensity ratios of the 00l diffractions showed little change, as shown in Fig. 1(a), which indicates that the angle  and h As primarily depend on the lattice parameters [33]. Table I summarizes the superconducting and structural parameters of the following optimally Another plausible origin is the different electronic properties that originate from the different numbers of the RE 4f electrons. It should be noted that the -T curves in Fig. 3 have a minimum at T min (shown by closed triangles) and that the resistivities show an upturn in the T region lower than T min for the (Ba 1x RE x )Fe 2 As 2 epitaxial films with x film larger than the critical x. We used conventional power-law behavior, where  =  0 + AT n (where  0 is the residual resistivity), for these -T curves at T>T min to evaluate the electron scattering. The fitting results, indicated by the black lines in Fig. 3, agree well with the experimental -T curves. The resistivity upturn can be explained either by carrier localization with disorder scattering [34] or by magnetic scattering [35].
Therefore, the origin of the weak resistivity-upturn behavior for the non-magnetic Therefore, a possible mechanism for breaking of the superconducting electron-pair is the interaction between the localized RE 4f electrons and the Fe 3d conduction electrons [36]. The contribution of the magnetic RE ions to the electron scattering can be estimated by comparing the electronic properties of the RE-doped samples with those of the nonmagnetic reference (Ba 1x La x )Fe 2 As 2 epitaxial films, because their structural differences are negligible, as shown in Fig. 2 and Table I.   As noted at the last of section 3 C, the variation of T c with the RE dopant is opposite between the present (Ba 1x RE x )Fe 2 As 2 epitaxial films and the (Ca 1x RE x )Fe 2 As 2 single crystals; i.e., the maximum T c of the (Ba 1x RE x )Fe 2 As 2 epitaxial films decreases in the order of RE = La, Ce, Pr and Nd, while that of the (Ca 1x RE x )Fe 2 As 2 single crystals increases in this order [1316]. The origins of this difference would be classified to an intrinsic effect and an extrinsic effect. To clarify the extrinsic effect originating from the different crystal structure parameters and crystalline quality between the thin films and the single crystals, we should use epitaxial films of RE-doped CaFe 2 As 2 epitaxial films; however, unfortunately, epitaxial thin film growth of CaFe 2 As 2 by PLD has not been attained [21]. Here, we like to discuss plausible intrinsic origins. One possible origin of this discrepancy may be the structural transition in (Ca 1x RE x )Fe 2 As 2 . It should be noted that (Ca 1x RE x )Fe 2 As 2 transits easily to a collapsed tetragonal structure by replacing the Ca site with a smaller RE dopant, such as Pr and Nd [13]. In the thin film samples used here, the local structures of the (Ba 1x RE x )Fe 2 As 2 films remained almost unchanged for different RE, implying that the film's electronic structures are not changed drastically by the RE doping. Therefore, it would be possible that the structural flexibility to a collapsed structure in (Ca 1x RE x )Fe 2 As 2 contributes to the high T c .

Conclusions
A non-equilibrium PLD process was used to stabilize nonmagnetic/magnetic RE          Table. II. Summary of the fitting parameters using the model [39] for the magnetoresistance of (Ba 1x La x )Fe 2 As 2 epitaxial films. Errors are indicated in the values in the parentheses. Figure 1.

I. Synthesis of bulk polycrystal samples of rare earth (RE) containing BaFe 2 As 2
Bulk polycrystal samples of RE-(= Ce, Pr, Nd, and Sm) containing BaFe 2 As 2 were synthesized for the PLD targets by two-step solid-state reactions. All preparation procedures except for the annealing process were carried out in an Ar-filled glove box (the O 2 impurity concentration was < 1 ppm and the dew point was < -100°C). A mixture of fine pieces of Ba and RE metals along with the powders of Fe and As were mixed in a stoichiometric atomic ratio of Ba : RE : Fe : As = 1-x : x : 2 : 2 and sealed in a stainless-steel tube under a pure Ar atmosphere, followed by a reaction at 700°C for 10 h. The resulting powders were grounded thoroughly and pressed into pellets. The pellets were placed in stainless-steel tubes and heated at 900°C for 16 h. It is important to carefully avoid contamination with atmospheric impurities such as oxygen-and water-related molecules in the PLD targets in order to obtain the high-purity (Ba 1x RE x )Fe 2 As 2 films; i.e., we could not succeed in the RE doping at the early stages of this study when a small amount of a REFeAsO impurity phase was observed in the PLD targets. Further, it should be noted that the REFeAsO phase was always produced when we synthesized (Ba 1x RE x )Fe 2 As 2 using an agate mortar and evacuated silica glass tubes similar to the case to synthesize Ba(Fe,Co) 2 As 2 in ref. S1. We, therefore, used a fully-dried silica-glass mortar to mix the powders and then sealed the pressed pellets into stainless tubes before the reactions.
Powder XRD patterns of the (Ba 1x RE x )Fe 2 As 2 bulk polycrystal samples with nominal x = 0.1 are shown in Fig. S1(a), and the nominal x dependence of the lattice parameters in Fig. S1(b). No peak shift was observed in the diffraction angles of the (Ba 1x RE x )Fe 2 As 2 samples from the calculated peak positions of undoped BaFe 2 As 2 . The lattice shrinkage, which is considered to originate from the substitution of Ba with the RE ions having smaller ion radii, was not observed, even though x was varied up to 0.25.
Simultaneously, the 200 diffractions for REAs impurities were observed distinctly for all the cases as indicated by the arrows in Fig. S1(a). These phase-separated bulk polycrystals (i.e., BaFe 2 As 2 + REAs) were used as the PLD targets. II. Chemical composition analysis of (Ba 1x RE x )Fe 2 As 2 epitaxial films Figure S2(a) shows the RE chemical composition mapping images for the optimally doped epitaxial films of (Ba 0.85 Ce 0.15 )Fe 2 As 2 , (Ba 0.82 Pr 0.18 )Fe 2 As 2 , and T. Katase et al. 37 (Ba 0.87 Nd 0.13 )Fe 2 As 2 examined with an electron-probe microanalyzer (EPMA). The EPMA mapping images ensured homogeneous RE distribution. Figure S2(b) presents the measured doping concentration x Film in the (Ba 1x RE x )Fe 2 As 2 epitaxial films measured by EPMA as a function of the nominal x. The good linearity in the relationship between the measured x and the nominal x demonstrates that the chemical compositions of the PLD targets were transferred linearly to the films. The solubility limit lines, which shifted to lower x film as RE changes in the order of Ce, Pr, and Nd, were determined as the chemical composition region where an impurity phase was not detected by the XRD measurements in Fig. 1.  Figure S3 shows the temperature derivatives of resistivities (d/dT) for the films in Fig. 3 [i.e., (Ba 1x Ce x )Fe 2 As 2 with x film = 00.12, (Ba 1x Pr x )Fe 2 As 2 with x film = 00.18, and (Ba 1x Nd x )Fe 2 As 2 with x film = 00.13]. The resistivity anomaly temperature (T anom ), defined as the peak position in the d/dT curves, shifts to lower T with increasing the doping concentrations (x film ).

Figure S3
d/dT curves of (a) (Ba 1x Ce x )Fe 2 As 2 with x film = 00.12, (b) (Ba 1x Pr x )Fe 2 As 2 with x film = 00.18, and (c) (Ba 1x Nd x )Fe 2 As 2 with x film = 00.13. The doping concentration x film is indicated on right-bottom of each figure. The positions of T anom are indicated by the red vertical arrows. Figure S4 shows the magnetoresistance (MR) of (Ba 0.72 Ce 0.28 )Fe 2 As 2 epitaxial film measured at (a) 30K, (b)10 K, and (c) 2 K. The MR were measured under magnetic field parallel to the c-axis and the ab-plane, respectively. With decreasing the temperature, the negative MR steeply increases and the transition magnetic field from the H 2 dependence to the -logH dependence shifts to a lower H. In addition, the negative MR with H//ab is much smaller than that with H//c for all the temperatures. This result indicates that the spin ordering of Ce should be antiferromagnetic parallel to the c-axis and ferromagnetic parallel to the ab-plane.