Phase engineering of rare earth nickelates by digital synthesis

Chemical doping is a very effective way of introducing new functionality into bulk compounds resulting in a plethora of fascinating many-body phenomena, including metal-insulator transition, high temperature superconductivity, exotic magnetism, and multiferroicity to name a few.^1–3 Inevitably, such a random chemical doping also introduces chemical disorder. The local structure of a material around the dopants can be significantly altered compared to the average long-range structures.⁴ As a result, short-range structural and chemical fluctuations can significantly impact the overall electronic and magnetic behavior of the material.^5,6 Despite the tremendous progress in various synthesis methodologies of complex oxide materials, it is still a major challenge to achieve an arbitrary amount of ionic substitution. In some extreme cases, dopant ions can even lead to the mesoscopic or microscopic phase separation, which often remains unnoticed in conventional diffraction experiments.^7,8 Accordingly, the development of alternative ways of materials engineering remains a paramount importance for the materials design community. Ex-situ doping of various ions is also being attempted to realize new functional behavior of correlated materials in recent times.^9,10 On the other hand, the epitaxial growth of ultra-thin superlattices (SLs) of complex oxides with the variation in the SL period below a characteristic length-scale, referred as “digital synthesis,”^6,11–14 is another promising route to overcome several challenges of conventional chemical doping.

Among the various 3d transition metal oxides, rare earth nickelates RENiO₃ (RE = La, Nd, Pr, Sm, Eu, Y, etc.) constitute an important class of materials which exhibits an interesting phase diagram with a strong dependence of electronic, magnetic, and structural transition temperatures on the choice of the RE ion.^15,16 For instance, NdNiO₃ and PrNiO₃ show simultaneous metal-insulator transition, orthorhombic to monoclinic structural transition, and paramagnetic to $E′$ type antiferromagnetic transition (⁠$E′$-AFM), whereas the other members of the series with a smaller RE ion, e.g., Eu, Sm, Y, etc., first undergo metal-to-insulator transitions (MIT) and structural transition at high temperature followed by an onset of antiferromagnetism at lower temperatures. Moreover, the synthesis of bulk RENiO₃ crystals requires high temperature and high oxygen pressure to stabilize the unusual +3 oxidation state of Ni.^15,16 Apart from LaNiO₃,¹⁷ only micron sized single crystals of RENiO₃ have been obtained so far due to their low thermodynamic stability.¹⁸ Recent progress in the epitaxial stabilization of thin films and heterostructures of RENiO₃^19,20 offers an exciting opportunity to investigate the evolution of electronic and magnetic transitions in this family through digital synthesis on high quality samples having a large area.

In this letter, we report the electronic and magnetic properties of a series of m uc EuNiO₃/p uc LaNiO₃ heterostructures (uc = unit cells, in pseudocubic notation). The superlattices with m = 3, 2, and 1 for p = 1 and m = 1 for p = 2 (labeled as mENO/pLNO) have been grown by pulsed laser deposition (PLD). The obtained superlattices can be viewed as ordered analogues of Eu_0.75La_0.25NiO₃, Eu_0.67La_0.33NiO₃, and Eu_0.50La_0.50NiO₃, Eu_0.33La_0.67NiO₃ compositions, respectively [Figs. 1(a) and 1(b)]. The desired superlattice structure of each sample has been confirmed by synchrotron X-ray diffraction. The electronic and magnetic properties have been investigated by dc-transport measurement and resonant soft X-ray scattering. While 3ENO/1LNO, 2ENO/1LNO, and 1ENO/1LNO SLs exhibit first order metal-to-insulator transitions (MIT) and long-range antiferromagnetic ordering, the transition temperatures are strongly altered compared to the bulk analogs Eu_1–xLa_xNiO₃ (x = 0.25, 0.33, and 0.5).²¹ Surprisingly, unlike the bulk doped composition, the electronic and magnetic transitions in the digital alloys are entirely suppressed in the [1ENO/2LNO] film, whereas a simultaneous transition around 100 K is expected from the phase diagram of the bulk Eu_1–xLa_xNiO_3.²¹ 

Ultra-short period [3ENO/1LNO] × 9, [2ENO/1LNO] × 12, [1ENO/1LNO] × 18, and [1ENO/2LNO] × 12 SLs have been grown on single crystalline NdGaO₃ (NGO) (1 1 0)_or [=(0 0 1)_pc, where or and pc denote orthorhombic and pseudo-cubic settings, respectively] by pulsed laser deposition, equipped with high pressure RHEED (reflection high energy electron diffraction).²² The layer by layer growth of both ENO and LNO layers in all SLs have been monitored and confirmed by RHEED.^22,23 Each growth cycle has been started with an ENO layer and finished with a LNO layer as illustrated in Fig. 1(a). XRD measurements have been performed with a six-cycle diffractometer at the 6-ID-B beamline of the Advanced Photon Source. dc transport measurements have been carried out in a Quantum Design Physical Property Measurement System (PPMS). The appearance of $E′$-antiferromagnetic ordering (⁠$E′$-AFM) has been probed at the resonant soft X-ray beamline 4.0.2 of the Advanced Light Source.

The structural quality of these SLs has been confirmed by a l scan along the (0 0 2)_pc truncation rod using synchrotron X-ray diffraction. Figure 1(c) shows a representative scan for the 1ENO/2LNO SL. As seen, the diffraction pattern consists of a very sharp substrate peak, broad film peak, and a series of thickness fringes, which are related to the different optical path of X-rays. Importantly, the presence of satellite peaks (denoted as +1 and –1) confirms the desired superlattice period of 3 unit cell (1 uc ENO + 2 uc LNO) for 1ENO/2LNO SL. The out-of-plane lattice parameter (c_pc) is found to be 3.818 Å, 3.807 Å, 3.793 Å, and 3.788 Å for 1ENO/2LNO, 1ENO/1LNO, 2ENO/1LNO, and 3ENO/1LNO, respectively.

As bulk LNO and ENO have different structural symmetry (rhombohedral vs. orthorhombic/monoclinic) and even in ultra-thin geometry RENiO₃ prefers to maintain the bulk-like symmetry,²⁴ a strong competition between these symmetries is possible. The specular (0 0) and off-specular (0 ± 1) reflections with the streak patterns of the RHEED pattern [Figs. 2(a)–2(d)] signify the presence of smooth flat terraces in all samples. Furthermore, 3ENO/1LNO, 2ENO/1LNO, and 1ENO/1LNO SLs show additional half order reflections (0 ± 1/2), as highlighted by the white arrows. Such half order reflections, also observed in NdNiO₃ and EuNiO₃ films, arise from the in-plane doubling of the unit cell and infer the orthorhombic/monoclinic symmetry.²⁵ On the other hand, this half order peak is absent for 1ENO/2LNO SL, suggesting the presence of the rhombohedral symmetry, akin to bulk LaNiO₃. We also note that during the growth of this 1ENO/2LNO SL, the half order peak vanishes after the growth of the second LNO layer in each period, and it reappears after the deposition of the ENO layer in the next repeat. The emergence and evanescence of the half-order peaks as a function of ENO and LNO layer numbers emphasize the strong competition between the symmetry of the constituent layers.

Since RHEED is a surface sensitive technique, we have further investigated the question of the symmetry of the SLs by bulk-sensitive synchrotron diffraction. A-sites in ABO₃ perovskites with the orthorhombic or monoclinic symmetry show antiparallel displacement,²² resulting in half order diffraction peaks with the index (odd/2 even/2 even/2)²⁶ in the Glazer notation.²⁷ Based on this result, the presence of (0 1/2 2)_pc and the absence of (1/2 0 2)_pc peaks in 1ENO/1LNO SL [upper panel of Fig. 2(e)] confirm that the 1ENO/1LNO film is orthorhombically distorted,²² with similar an in-plane orientation as the NGO substrate. The absence of both diffraction peaks for 1ENO/2LNO SL [lower panel of Fig. 2(e)] is consistent with the rhombohedral symmetry inferred from the in-situ RHEED pattern. In short, our combined characterizations using XRD and RHEED implies that the symmetry of these short-period mENO/pLNO SLs is controlled by ENO for m ≥ p and by LNO for m < p.

After confirming the high structural quality of these SLs and probing the structural symmetry across these SLs, we now discuss their electronic and magnetic properties. First, we recall that the bulk ENO undergoes MIT around 460 K and paramagnetic-to-antiferromagnetic phase transition around 200 K,²¹ while the rhombohedral LNO remains metallic and paramagnetic down to low temperature. Figure 3(a) shows dc resistivities of the SLs. As reported earlier,²² 1ENO/1LNO and 2ENO/1LNO SLs exhibit first order MIT around 150 K and 245 K, respectively, with several orders of magnitude changes in resistivity. With the introduction of another ENO layer in each period while keeping p = 1, T_MIT is enhanced to 330 K. The hysteresis region of these SLs implies co-existence of phase separated metallic and insulating regions.²⁸ The hysteresis between the heating and cooling curves decreases with the increasing T_MIT as the higher thermal energy causes faster transformation dynamics between these phases.¹⁶ On the other hand, 1ENO/2LNO SL remains metallic down to 2 K without any detectable transition into the insulating phase.

As for the temperature dependence, the rhombohedral metallic phase of bulk LaNiO₃ behaves like a Fermi liquid (FL) at low temperature [ρ(T) = ρ₀ + ATⁿ with n = 2, where ρ₀ is the residual resistivity and A represents the strength of electron-electron scattering].^17,29 Surprisingly, the transport properties of orthorhombic metallic phases are non-Fermi liquid (NFL, n < 2).²⁹ The NFL exponent n can be tuned from 1 to 4/3, 3/2, and 5/3 by the application of pressure²⁹ and epitaxial strain.^30,31 The strong T dependence of dρ_xx/dT² [inset of Fig. 3(b)] for 1ENO/2LNO SL emphasizes the NFL behavior in-spite of its rhombohedral structure. In order to further probe the nature of the metallic phase of 1ENO/2LNO SL, we have plotted log(dρ/dT) vs log(T) in Fig. 3(b), and different regions of this curve have been fitted linearly. Such a linear fitting in the 2 K–35 K range [shown by the red line in Fig. 3(b)] finds n = 1.47 which is the close NFL exponent of 3/2. Similar linear fitting (not shown) results n = 1.34 and 1.39 (both are close to n = 4/3) in 35 K–85 K and 100 K–210 K, respectively, and the exponent changes to 1.07 (i.e., close to the linear T dependence of ρ_xx) in the 210 K–295 K range. Further study is required to understand such a change in the NFL exponent as a function of T.³⁰ Similarly, 1ENO/1LNO SL also shows the NFL transport.²³ We note that due to the limited temperature range for the metallic phase, a similar analysis has not been carried out for 3ENO/1LNO and 2ENO/1LNO SLs.

Now we discuss the magnetic properties of our SLs. The spin structure of the RENiO₃ family can be viewed as ↑0↓0 stacking of (1 1 1) pseudo-cubic planes and is characterized by the (1/2 0 1/2)_or [(1/4 1/4 1/4)_pc] magnetic vector.^22,32–37 Such an $E′$-AFM phase can be probed by examining the (1/4 1/4 1/4)_pc diffraction peak, using resonant soft X-ray scattering with the photon energy tuned near the Ni L₃ edge.^{22,30,33,38–41} The inset of Fig. 3(c) shows the temperature dependence of (1/4 1/4 1/4)_pc in 3ENO/1LNO SL. The integrated area of this curve as a function of T [Fig. 3(c)] yields T_N ∼ 200 ± 5 K for the 3ENO/1LNO SL. Similar measurements have revealed T_N of 220 ± 5 K and 155 ± 5 K for 2ENO/1LNO and 1ENO/1LNO SL, respectively.²² The non-monotonic dependence of T_N on the number of ENO layers (m) is very likely to be related to the change in the underlying $∠Ni−O−Ni$ bond angle.¹⁶ No such diffraction peak has been found for 1ENO/2LNO SL at low temperature. To further confirm the absence of the $E′$-AFM phase in this sample, we have measured the Hall effect at different T (not shown). Previous studies⁴² demonstrate that the Hall coefficient (R_H) shows a sign change (from hole-like to electron-like) across paramagnetic to $E′$-AFM phase transition. No sign change in R_H has been found for the 1ENO/2LNO SL, thus emphasizing the absence of the $E′$-AFM phase in this film down to at-least 2 K.

Finally, we have compared electronic and magnetic transition temperature of the present set of films with the bulk doped series Eu_1–xLa_xNiO₃. As seen in Fig. 4, T_MIT and T_N of these SLs are lower compared to the corresponding bulk composition. These differences can be attributed to several factors, including single crystallinity of the film versus the polycrystalline, granular structure of the bulk, strain (tensile strain can decrease T_N³⁰) and confinement effects due to the finite film thickness, and most importantly, the absence of the disorder in the superlattice that is present in bulk samples. Moreover, 1ENO/2LNO does not show any detectable signs of MIT and magnetic transition while a T_MIT = T_N ∼ 100 K is anticipated in Eu_0.67La_0.33NiO₃ from the bulk phase diagram. We also point out that in the past such a suppression of MIT was only observed for the compressive strain in the RENiO₃ thin films and superlattices.^{23,30,31,39,40,43} This is not the case for our SL since the in-plane lattice constant of the NdGaO₃ substrate (3.858 Å) is larger compared to ENO (3.806 Å) and similar to LNO (3.855 Å). Reduced T_MIT and T_N in the digitally synthesized composition point to the governing role played by the hetero-interface in selecting ground state electronic and magnetic structures of RENiO₃.

In summary, we have grown a series of ENO/LNO superlattice on NGO substrates to investigate the nature of the ground state of rare-earth nickelates through digital synthesis. The desired ultra-short period layer-by-layer growth was confirmed by RHEED and XRD. Electronic and magnetic transition temperatures of these films are markedly modified from analogous randomly doped bulk Eu_1–xLa_xNiO₃. This work demonstrates the utility of digital synthesis for effective doping of complex functional oxides without conventional ionic disorder associated with chemical substitution.

S.M. was supported by IISc start up grant and ISRO-IISc Space Technology Cell. D.M. was supported by the U.S. Department of Energy, Office of Basic Energy Sciences, Early Career Award Program under Award No. 1047478. X.L. acknowledges the support of DE-SC 00012375 grant for synchrotron work. J.C. was supported by the Gordon and Betty Moore Foundation EPiQS Initiative through Grant No. GBMF4534. S.M. also acknowledges the support of EPiQS Initiative No. GBMF4534 for sponsoring a trip to Rutgers University. This research used resources of the Advanced Photon Source, a U.S. Department of Energy Office of Science User Facility operated by Argonne National Laboratory under Contract No. DE-AC02-06CH11357. This research used resources of the Advanced Light Source, which is a Department of Energy Office of Science User Facility under Contract No. DE-AC02-05CH11231.