Tunable superconductivity in Fe-pnictide heterointerfaces by diffusion control

New Fe-pnictide heterostructures of the type LnOFeAs/BaFe$_2$As$_2$ (Ln = La, Sm) were grown by pulsed laser deposition (PLD) and investigated. Their common structural unit of [Fe$_2$As$_2$] planes allows perfect matching between the different Fe-pnictide unit cells and a coherent and atomically sharp interface. We test the stability of the heterointerface in the presence of Co$^{2+}$ (cations) as well as for excess O$^{2-}$ (anions) and discuss the consequences on the electronic properties: While undoped SmOFeAs/BaFe$_2$As$_2$ remains non-superconducting, a balanced Co-concentration after diffusion across the interface results in superconductivity within Co-substituted variants. In contrast, excess O$^{2-}$ causes the formation of an interfacial layer in SmOFeAs/BaFe$_2$As$_2$ with increased O$^{2-}$/As$^{3-}$ ratio and develops a metal-to-superconductor transition with time. The engineered heterointerfaces may provide a sophisticated pathway to bridge the gap between Fe-pnictides and Fe-chalcogenides.

The monolayer (ML-) FeSe/SrTiO3 heterointerface between an Fe-chalcogenide and a perovskite oxide has recently demonstrated an ability to host high-temperature superconductivity in the range of 40 -75 K [1 -3], although the origin of Cooper pairing is still controversial [4,5]. As van der Waals compounds, FeSe and FeTe have also been employed in the formation of heterointerfaces with topological insulators in the search for Majorana bound states [6,7]. While interface engineering in ML-FeSe/SrTiO3 has attracted worldwide attention, the realization of equivalent experiments for Fe-pnictides is currently less developed and highly challenging due to their polar surfaces.
Uncompensated dipole moments are responsible for reconstruction and charge modulation on a BaFe2As2 surface [20], and may also lead to the observed Fermi surface shifts in BaFe2As2 after surface doping [21].
Like Ba-122/Ba-122 (Fig. 1d), the Ln-1111/Ba-122 interface does not interrupt the sequential stacking of layers with alternating layer charge ±2 and thus avoids a polarization discontinuity (Fig. 1f). We prove below that the commonly shared [Fe2As2] 2layer constitutes a nonpolar (stable) heterointerface after joining Ln-1111 and Ba-122 unit cells with atomic precision as designed in Fig.1e. Besides the recently emphasized geometric design plans [13], we turn the attention towards electrostatic principles with their vital consequences in the engineering of films and interfaces. Both, coherency and stability of the Fe-pnictide heterointerfaces can be disturbed by changes in the local charge distributions and the ionic environment. We have, therefore, tested the interface stability in the presence of Co 2+ supply and excess O 2− by selecting different target compositions for LnO1-yFe1-xCoxAs (Ln = La, Sm and x = 0, 0.15; y = 0, 0.2) after BaFe2As2 deposition. We show results of five heterostructures investigated by X-ray diffraction and reflectivity (XRD, XRR), scanning transmission electron microscopy (STEM), electron energy loss spectroscopy (EELS), electron-dispersive spectroscopy (EDS), Auger electron spectroscopy (AES) and electrical transport measurements. Basic information of the heterostructures and band structure calculations based on density functional theory (DFT) can be found in the Supplementary Information (SI, Tab. S1 and Fig. S1).  Fe2As2] layer (black arrow) [8], b) polar/weakly polar oxide interface in Ba-122/SrTiO3 for TiO2 termination with a shared Ba/Sr plane [12] and c) for SrO termination with the formation of an interfacial BaFeO2+x layer (represented here with reduced thickness) [12]. When both layers are Fe-pnictides, there is the possibility for d) a charge compensated polar/polar heterointerface between two different Ba-122 layers proposed in Refs. [14,33] and e) a charge compensated polar/polar heterointerface between Ln-1111 and Ba-122 with a shared [Fe2As2] layer (this work). Although both layers are polar, they share the same sequence of layer charges resulting in a non-polar interface. f) The expected converging electric potential across Fe-pnictide/Fe-pnictide heterostructures.
Co 2+ ions, supplied during Ln-1111 deposition, do not perturb any layer or interface polarity and diffuse from the top Ln-1111 into the bottom Ba-122 layer down to the substrate (Fig. 2a,b). Codiffusion leads to electron doping of the originally undoped Ba-122 layer and the final Co concentration depends on the Co supply during film growth (i.e. the composition of the LnOFe1-xCoxAs target) and the layer thicknesses. In the relevant heterostructures we have used LnOFe1-xCoxAs targets with x = 0.15. The actual average Co-content in LnOFe1-xCoxAs/Ba(Fe1-xCox)2As2 is x = 0.08 ± 0.02 (in La-1111/Ba-122) and 0.06 ± 0.03 (in Sm-1111/Ba-122) with an almost balanced distribution along the cross section. As a consequence, 3D superconductivity develops in the complete heterostructure, which we will discuss below in more detail.
Upon the supply of excess O 2− the AES depth profile revealed significant differences (Fig. 2d) compared to the clean Sm-1111/Ba-122 heterostructure (Fig. 2c)   Employing EELS we have traced the cation interdiffusion of Ba 2+ and La 3+ ions at least 5 nm deep 8 into the adjacent layers (Fig. 3e). Assuming charge neutrality, the diffusion of two Ln 3+ ions should compensate for the diffusion of three Ba 2+ ions in the opposite direction. The larger ionic radius of Ba 2+ compared to La 3+ could partially explain an increase of interfacial unit cell volumes, however, while La substitution in BaFe2As2 is documented, the structural stability of LaOFeAs is expected to be strongly limited when Ba 2+ substitutes for La 3+ . The presence of Co 2+ does not change the electrostatic considerations and is thus not detrimental to the Fe-pnictide heterointerface.
Lanthanide substitution at Ba sites in Ba-122 could result in a locally confined interfacial electron doping since superconductivity was reported for (Ba1-xLax)Fe2As2 films [22], whereas Smsubstitution in Ba-122 remained unachieved in previous film growth attempts [23]. At present, we can only determine that the Sm gradient into Ba-122 is similar to that of La when comparing AES depth profiles (Fig. 2). We anticipate that we do not find superconductivity in a clean Sm-1111/Ba-122 heterostructure, i.e. without Co 2+ nor excess O 2− (SI, Fig. S1).
Another LaOFe1-xCoxAs/Ba(Fe1-xCox)2As2 heterostructure and a SmOFe1-xCoxAs/Ba(Fe1-xCox)2As2 heterostructure were analyzed by both scanning precession electron diffraction (SPED) and by HAADF-STEM combined with an EDS mapping across their interfaces. Fig. 4 shows the results of this analysis. The SPED data reveals the distinct diffraction patterns found in the different layers and the resulting four dimensional dataset of two real space and two reciprocal space dimensions [24] can then be processed to show the spatial distribution of each exemplar diffraction pattern [25]. These show characteristically that there are three clearly diffraction patterns for both the heterostructures, one for the MgO, one for the Ba-122 structure and one for the Ln-1111 structure. These are all different enough that the spots can be isolated and the spatial distribution of these three patterns can be easily mapped as virtual dark field (VDF) images. Additionally, there is often a region at the top of the Ba-122 layer which is faulted and which produces a complex diffraction pattern that appears to resemble a mixture of the 122 and the 1111 diffraction patterns, as such, this cannot be mapped as a separate pattern. We did not find any evidence that O 2− may diffuse after film growth through the artificially designed grain boundary between Ln-1111 and Ba-122, as we repeated an AES analysis on one heterostructure after six months and could not detect any changes in the element concentrations. Any interlayer formation is, therefore, completed after film growth. O-diffusion from the MgO substrate can be ruled out.

Superconductivity in Fe-pnictide heterointerfaces
Electrical transport measurements revealed superconductivity in Ln-1111/Ba-122 heterostructures with Co 2+ and excess O 2− , but neither for a clean Sm-1111/Ba-122 heterostructure nor in films of BaFe2As2 film (20 nm) on MgO(100) or in LnOFe1-xCoxAs films on MgO(100), even in the presence of a doping agent like Co-substitution. This result was attributed to strain [26]. Our DFT calculations for the La-1111/Ba-122 heterointerface (SI, Fig. S7) with an atomic arrangement such as experimentally observed in Fig. 3b predicts a smooth   In the presence of Co 2+ the measured superconducting transition temperature exceeds the maximum Tc found for LnOFe1-xCoxAs compounds, that is 13 -18 K (for Ln = La -Sm). This result is easily understood in terms of Co-diffusion from the LnOFe1-xCoxAs layer into the BaFe2As2 layer.
The diffusion process during film growth balances the Co-content in the individual Fe-pnictide layers: The Co-content in the top LnOFe1-xCoxAs layer decreases while the Co-content increases towards optimal doping in the initially undoped Ba-122 at the bottom. The average Co-contents, x = 0.08 ± 0.02 (in La-1111/Ba-122) and 0.06 ± 0.03 (in Sm-1111/Ba-122) are close to optimal doping in Ba-122, which is consistent with the observed Tc,90 of 16.5 -20.5 K (Fig. 5 a,c). Since superconductivity develops in the whole heterostructure, a 3-dimensional (3D) superconducting state is traced by the angular dependence of the upper critical field (Fig. 5 b,d)

Sm-1111/Ba-122 heterostructures with excess O 2− and an interfacial layer formation develop an
induced metal-to-superconductor transition over time (Fig. 5e, SI, Fig. S1). We attribute this induced superconductivity to electronic/structural modifications caused in the interfacial layer with an increased O 2− /As 3− ratio. The as-grown heterostructure is initially non-superconducting.
A superconducting transition emerges after several weeks with Tc increasing, as displayed in Fig. 5e, up to 240 days after growth showing Tc,90 = 27.5 K and a complete superconducting transition.
Compared to the Co-substituted variants, the larger Tc,90 indicates less disorder at the Fe-sites but sufficient doping. The angular dependence of µ0Hc2 deviates from a 3D AGL fit with an anisotropy of γ = 4.1 and approaches a two-dimensional (2D)-like behavior described by Tinkham's formula [27] (Fig. 5f). This indicates, that superconductivity appears in a locally confined region close to the interface which is in agreement with the formation of an interfacial layer showing strong O 2− /As 3− imbalance.

Discussion
The diffusion-based design by the incorporation of impurity cations/anions into Fe-pnictide heterointerfaces provides valuable insight into the tunability of their electronic properties. The heterointerface itself (i.e. the formation of either an abrupt coherent interface or an interfacial layer) is strongly governed by electrostatic arguments.
Diffusion processes have been previously used in fabrication routes for Fe-pnictide films that make use of an ionic exchange [28,29], and they occurred between Fe-pnictide films and substrates [30,31]. Co-diffusion from a Ba(Fe1-xCox)2As2 film into an Fe buffer layer was reported previously [30,32]. As shown in Fig. 1a, the Ba-122/Fe heterointerface constitutes one example for a coherent heterointerface. Enhanced superconductivity in a Ba(Fe1-xCox)2As2/Fe heterostructure (by more than ~3 K from ~26 to 29.4 K) was first shown in Ref. [9]. Ba-122/SrTiO3 heterointerfaces (Fig. 1b,c), that depend on the surface termination and can be either abrupt (TiO2 termination) or show a BaFeO2+x interfacial layer formation (SrO termination) [12], are also  storing them only several hours in air or water vapor [34]. There, the formation of (Fe or Sr) vacancies was suggested as crucial mechanism [35]. Another candidate would be As vacancies, that generate local magnetic moments that can coexist with superconductivity in Ln-1111 [36].
Although Ref. [37] mentioned the possibility to suppress the magnetic ordering (spin density wave) by intercalation of H2O into SmOFeAs, we do not have yet any indication for this scenario and the true origin of the induced superconducting state is currently under investigation.

Conclusions
The above results indicate the significant role of ion valencies (Co 2+  The precession and scanning of the beam were controlled by a NanoMEGAS DigiSTAR system (NanoMEGAS SPRL, Brussels, Belgium) with the scan areas and acquisition handled by their Topspin software. In this case, the diffraction patterns were acquired using a prototype system using a Merlin for EM direct electron detector (Quantum Detectors Ltd., Harwell, UK) instead of the normal CCD camera pointed at the focusing screen [40], as recently benchmarked and more fully described in MacLaren et al. [41]. Virtual dark field (VDF) imaging of the different phases was then accomplished using the method described by Paterson et al. [25] where the data files are read into python and processed using the fpd python library (https://fpdpy.gitlab.io/fpd/fpd.html) to numerically integrate the intensity in each diffraction pattern within different periodic arrays of apertures set to correspond to the diffraction spots unique to each of the crystalline phases (and avoiding diffraction spots common to more than one phase).
 Auger electron spectroscopy (AES) was carried out on an ULVAC-Phi 710 Auger electron spectrometer integrated in a scanning electron microscope (SEM) with a primary electron beam of 10 kV. An incident electron current of 10 nA on scanned areas between 9 µm² and 150 µm in diameter resulted in current densities between 57 µA⋅cm -2 and 111 mA⋅cm -2 on the film surfaces. For analysis clean areas without droplets were chosen. AES depth profiles were obtained by Ar + ion sputtering (1 kV on an area of 2×2 mm²). In order to decrease the effectively sputtered film area to ~0.78 mm², the film surfaces were covered by an Al foil having a hole of ~1 mm in diameter. Auger spectra, N(E), were recorded in a kinetic energy range of E = 30 -1500 eV with a step size ∆E = 1 eV. The data was processed using Phi MultiPak software. The quantitative analysis of the element concentrations in the heterostructures is based on relative sensitivity factors using peak-to-peak heights of the differentiated spectra E⋅dN/dE (S5D5).
 Electrical transport measurements (four probe) measurements were carried out in a Quantum Design Physical Property Measurement System (PPMS) equipped with a sample rotator stage in external magnetic fields up to µ0H = 9 T and down to 2 K. A constant current of 1 µA was used in all R(T) and R (H,θ) measurements. For the electrical contacts Cu wires (∅ = 0.01 mm) were attached on the thin film surface using a commercial silver paste. Critical temperatures and upper critical fields, µ0Hc2(T,θ), were evaluated using a resistivity criterion (90% or 50% Rn).
 Electronic structure calculations were performed within density functional theory and generalized gradient approximation [42] for the exchange correlation functional in the projector-augmented plane wave (PAW) formalism [43] as implemented in the Vienna ab-initio Simulation package (VASP) [44]. The energy cutoff was set to 500 eV, and the convergence criterion for the electronic density was chosen as 10 -8 eV. For the bulk BaFe2As2 and LaFeAsO we adopted the experimental structures with the tetragonal I4/mmm and P4/nmm space groups, respectively, as reported in [45].