Aspects of the Synthesis of Thin Film Superconducting Infinite-Layer Nickelates

The recent observation of superconductivity in Nd0.8Sr0.2NiO2 calls for further investigation and optimization of the synthesis of this metastable infinite-layer nickelate structure. Here, we present our current understanding of important aspects of the growth of the parent perovskite compound via pulsed laser deposition on SrTiO3 (001) substrates, and the subsequent topotactic reduction. We find that to achieve single-crystalline, single-phase superconducting Nd0.8Sr0.2NiO2, it is essential that the precursor perovskite Nd0.8Sr0.2NiO3 thin film is stabilized with high crystallinity and no impurity phases; in particular, a Ruddlesden-Popper-type secondary phase is often observed. We have further investigated the evolution of the soft-chemistry topotactic reduction conditions to realize full transformation to the infinite-layer structure with no film decomposition or formation of other phases. We find that capping the nickelate film with a subsequent SrTiO3 layer provides an epitaxial template to the top region of the nickelate film, much like the substrate. Thus, for currently optimized growth conditions, we can stabilize superconducting single-phase Nd0.8Sr0.2NiO2 (001) epitaxial thin films up to ~ 10 nm. ________________________ akyuho@stanford.edu bdenverli@stanford.edu


Introduction
Low-temperature oxygen deintercalation of the Ruddlesden-Popper (RP) series of nickelates Lnn+1NinO3n+1 (Ln = lanthanides) gives rise to metastable Lnn+1NinO2n+2 structures with layered NiO2 square planes and formal nickel valence of Ni 1+1/n . [1][2][3][4][5][6][7] Notably, an unusual formal nickel valence of Ni + is reached in the n = ∞ infinite-layer nickelate LnNiO2, realizing possible structural and electronic analogs to the undoped parent compound of layered cuprate high-temperature superconductors. [8][9][10][11][12] The synthesis of these metastable infinite-layer nickelates was first reported in 1983, where polycrystalline perovskite LaNiO3 was reduced to LaNiO2 with hydrogen gas as the reducing agent. 1,2 It was later shown in 1999 and onwards that this topotactic reduction process can be achieved more reproducibly at lower temperature by using metal hydrides for reduction. 4,5,13 This technique was further extended to epitaxial nickelate thin films, with the first demonstration in 2009 for LaNiO3 (001) epitaxial thin films using CaH2 as the reducing agent. 13,14 Motivated to explore the analogy to superconducting cuprates, we have recently observed superconductivity in chemically doped Nd0.8Sr0.2NiO2 (001) epitaxial thin films grown on SrTiO3 (001) substrate by pulsed laser deposition (PLD). 15 This finding warrants the systematic investigation of its superconducting and normal state properties, for which establishing a reproducible synthetic route is critical. There are two key issues in stabilizing Nd0.8Sr0.2NiO2 (001) epitaxial thin films. First is the instability of the precursor perovskite phase. While chemical doping by strontium brings the nickel valence of the infinite-layer phase (nominally Ni 1.2+ ) closer to the thermodynamically stable Ni 2+ , it results in a rather extreme formal nickel valence of Ni 3.2+ in the Nd0.8Sr0.2NiO3 perovskite precursor. This chemical instability adds significantly to the existing synthesis challenges of the undoped perovskite NdNiO3, namely the nontrivial fluctuation of the film quality upon subtle changes in growth and post-annealing conditions. [16][17][18][19] In addition, tailoring the substrate choice to minimize lattice mismatch with the infinite-layer phase (-0.4% for the SrTiO3 (001) substrate) 5 forces a large tensile strain (+2.6% with SrTiO3 (001)) 20 on the perovskite nickelate. These factors pose an interesting materials challenge to forming the aimed infinite-layer structure, the crystallographic quality of which is found to be heavily dependent on that of the precursor perovskite structure. 4 Second, previous studies have shown that it is difficult to stabilize uniform, single-crystalline infinite-layer nickelate films from soft-chemistry topotactic reduction of the perovskite. 13,[21][22][23] For example, reduction studies on LaNiO3 have shown that, besides the infinite-layer LaNiO2 (001), phases such as brownmillerite LaNiO2.5 and a-axis oriented LaNiO2 (100) can appear during reduction. 13,21 A previous study on NdNiO3 reduction also indicated that a fluorite defect phase can be introduced on top of the infinite-layer NdNiO2 (001) films under certain annealing conditions. 22 Depending on the reduction conditions, decomposition of the infinite-layer phase at the upper region of the film was also observed. 23 These results indicate the need of careful optimization of the reduction conditions, and, perhaps, adjustments in the structural design of the film to promote single-phase stabilization.
In the first section of this paper, we examine the optimization of PLD growth of perovskite Nd0.8Sr0.2NiO3 (001) on SrTiO3 (001) substrate. We discuss two different optimized growth conditions for Nd0.8Sr0.2NiO3, which are based on two different laser fluences. In the second section of this paper, we present studies on the CaH2-assisted topotactic reduction of the Nd0.8Sr0.2NiO3 (001) precursor phase, discussing the effect of a SrTiO3 capping layer on the reduction process and the evolution of the nickelate film as a function of reduction time. Finally, we discuss how the choice of the growth conditions affects the crystallinity and the superconducting transition of Nd0.8Sr0.2NiO2 (001). 5 × 5 mm 2 TiO2-terminated SrTiO3 (001) substrates were pre-annealed for 30 minutes at 930 °C under oxygen partial pressure PO 2 = 5 × 10 -6 Torr to achieve a sharp step-and-terrace surface prior to film growth. Undoped and Sr-doped nickelate films were grown on these substrates by PLD with a KrF excimer laser (λ = 248 nm), using mixed-phase polycrystalline targets of Nd2NiO4 + NiO and mixed-phase polycrystalline targets of (Nd0.8Sr0.2)2NiO4 + NiO, as confirmed by powder x-ray diffraction (XRD), respectively. In this study, we used a uniform 1.25 mm × 1.90 mm laser spot for ablation, formed by imaging an aperture. The targets were prepared by sintering mixtures of stoichiometric amounts of Nd2O3, SrCO3, and NiO powders at 1350 °C for 12 hours, with two intermediate grinding and pelletizing steps after initial decarbonation at 1200 °C for 12 hours.
Details on the PLD growth conditions of SrIrO3 (001) epitaxial films (Figure 7(a)) can be found in Ref. 24.
The deposited films were cut into two pieces of size 2.5 × 5 mm 2 . Loosely wrapped in aluminum foil to avoid direct contact with the reducing agent, each piece was vacuum-sealed (pressure < 0.1 mTorr) with 0.1 g of CaH2 powder in a Pyrex glass tube. The tube was then heated to the desired temperature to perform reduction. The temperature ramping rate was fixed at 10 °C min -1 .
XRD symmetric θ-2θ scans of the nickelate films were measured using a monochromated Cu Kα1 (λ = 1.5406 Å) source. Temperature-dependent resistivity (ρ(T)) measurements were conducted in a four-point geometry using aluminum wire-bonded contacts. In some cases, gold contact pads were evaporated using electron-beam evaporation before wire-bonded contacts were made. Crosssectional scanning transmission electron microscopy (STEM) specimens were prepared using a  Literature on PLD-growth of undoped NdNiO3 reports a relatively flexible range of growth conditions, with substrate temperature Ts ranging from 600 to 750 °C, PO 2 ranging from 100 to 200 mTorr, laser fluence F ranging from 1.5 to 2.1 J cm -2 , and laser frequency f ranging from 2 to 30 Hz. 15-17, 19, 25, 26 Figure 1 shows the XRD symmetric θ-2θ scan of a ~ 20 nm NdNiO3 (001) film grown by PLD on single-crystal SrTiO3 (001) substrate under the growth conditions of Ts = 600 °C, PO 2 = 200 mTorr, F = 1.4 J cm -2 , and f = 4 Hz. Considering the tensile strain induced by the substrate, the extracted film c-lattice constant of 3.77 Å is in good agreement with the pseudocubic bulk lattice constant of NdNiO3 (3.807 Å). 27 The prominent NdNiO3 001 peak and the presence of fringes around the film peaks in the symmetric θ-2θ scan (Figures 1(a) and 1(b)) are indications that the growth conditions are well within the optimal growth window of NdNiO3. 28 In many material systems, it is often the case that doping or partial cationic substitution requires minimal or no change in the PLD growth conditions. [29][30][31][32][33] However, when the identical growth conditions above are employed using the 20 at. % Sr-doped target, the resultant film symmetric θ-2θ scan is far from that of highly-crystalline Nd0.8Sr0.2NiO3 (001) (Figures 1(a) and 1(b)). Namely, the perovskite 001 peak is absent (Figure 1(a)), and the extracted c-lattice constant of 3.80 Å is nontrivially larger than that of the optimized NdNiO3 (001) film (Figure 1(b)). Both of these features have been previously observed in undoped nickelate films. 17,34 The cross-sectional HAADF STEM image of this film reveals that it is densely populated with vertical RP-type faults ( Figure 1(c)). These defects are formed when an AO rocksalt layer (where A corresponds to the A-site cation) stabilizes in between the perovskite layers. [34][35][36] The frequent inclusion of these rocksalt layers breaks the structural long-range order of the perovskite phase and makes the film analogous to a highly disordered sequence of in-plane oriented RP phases. Indeed, the observed diffraction pattern of the secondary phase film matches well to the (110) oriented trilayer RP phase (Nd0.8Sr0.2)4Ni3O10, which has a 220 diffraction peak aligning well to the observed film peak at 2θ ≈ 47.84°. 37 Also, this phase has no 110 peak by symmetry, 37 consistent with the absence of a film peak near the SrTiO3 001 substrate peak. As will be later shown, this phase with densely populated vertical RP-type faults behaves very differently from the perovskite in terms of topotactic reduction and transport properties. Given the high degree of disorder, for simplicity we denote this as a 'secondary phase'. Overall, these observations indicate that Sr doping significantly reduces the growth window of the perovskite phase, and that further optimization of growth conditions is required. show the Nd0.8Sr0.2NiO3 00l peaks superposed with the secondary phase peak, which has a smaller 2θ value than the Nd0.8Sr0.2NiO3 002 peak in the XRD symmetric θ-2θ scan (Figures 2(a) and 2(b)).
In addition, we find that the population of the two observed phases changes as a function of target history. The XRD symmetric θ-2θ scan of three samples grown consecutively under fixed growth conditions ( Figure 2(b)) shows that the secondary phase gradually dominates over the perovskite phase with increasing target ablation. This is in line with the previous observation of limited film reproducibility and nickel enrichment of the target over time in the PLD study of NdNiO3; 19 the favored ablation of the A-site cations from the target may increasingly promote the stabilization of the A-site-rich secondary phase over time. To avoid ambiguities arising from target history, we subsequently re-polished the target surface after each film growth.
It is interesting to see how these partially optimized mixed-phase films transform upon CaH2assisted topotactic reduction. Figure 2(c) shows the evolution of the XRD θ-2θ peaks of a mixedphase film over the reduction process. The right of the as-grown double peak, which corresponds to the perovskite phase, shifts further rightward upon reduction and saturates at a 2θ value corresponding to c = 3.32 Å, indicating the successful transformation of the perovskite phase to the infinite-layer structure. 5,13 In contrast, the left of the as-grown double peak shifts further leftward towards the SrTiO3 002 peak position. Again assuming the secondary phase can be approximately described as (Nd0.8Sr0.2)4Ni3O10 (110), the corresponding reduced structure (Nd0.8Sr0.2)4Ni3O8 (100) will have a 200 peak near 2θ ≈ 46.32°, which is very close to the SrTiO3 002 peak at 2θ ≈ 46.47°. 3 Note that the change in the crystallographic notation is due to the difference in the conventional unit cell space group of (Nd0.8Sr0.2)4Ni3O10 (P21/a) and (Nd0.8Sr0.2)4Ni3O8 (I4/mmm). 3,37 This reduced structure also has no lower-order peak (i.e. 100 peak) by symmetry. 3 Such clear difference in the structural evolution of the two phases upon reduction translates to the transport properties of the two phases. For the film that dominantly consists of the secondary phase, with essentially no sign of the infinite-layer phase after reduction (Figure 3(a)), no evidence for superconductivity is found down to 2 K. For the mixed-phase film after reduction (Figure 3(b)) a superconducting transition is observed, with an onset at 14.7 K (point of maximum curvature), a midpoint at 12.6 K, and zero resistance at 7.2 K (indistinguishable from the noise floor) ( Figure   3(c)). These observations indicate that the infinite-layer nickelate phase, not the reduced secondary phase, is superconducting. We emphasize that the presence of the perovskite 001 film peak, and the 002 film peak position, are the two strongest and most useful functional indicators for superconductivity; when the 001 peak is not observed and/or the 002 peak 2θ position is below ~ 48°, the subsequently reduced film never exhibits superconductivity. The cross-sectional STEM images of the reduced mixed-phase film ( Figure 4) show a segregation of the two competing phases, where the infinite-layer phase is stabilized in the vicinity of the substrate and the secondary phase sits above the infinite-layer phase. Such preferred stabilization of the infinite-layer structure near the substrate has been observed in previous nickelate reduction studies. 22,23 In particular, this was also observed for films grown by metal organic decomposition, 23 suggesting that the target history effects in the PLD growth of nickelates is not the primary factor for this phenomenon. Rather, this suggests that the epitaxial strain energy provided by the substrate plays an important role in stabilizing the perovskite phase during growth and the infinite-layer phase during the reduction process. Hence, growing thinner films can promote single-phase stabilization of the desired phase. By further empirically optimizing the growth conditions and keeping the film thickness below ~ 15 nm, we were able to obtain Nd0.8Sr0.2NiO3 (001) epitaxial films on SrTiO3 (001) substrates with no visible secondary phase peaks in XRD under two different growth conditions. The first we denote as the 'high-fluence' growth conditions, with Ts = 600 °C, PO 2 = 150 mTorr, F = 2.0 J cm -2 , and f = 4 Hz, while the second is in 'low-fluence' growth conditions, with Ts = 600 °C, PO 2 = 70 mTorr, F = 1.0 J cm -2 , and f = 4 Hz. Figure 5 shows the XRD symmetric θ-2θ scans of six optimized samples with film thickness ranging from 5 to 15 nm in these two growth conditions. All samples show Nd0.8Sr0.2NiO3 00l film peaks with prominent 001 peak intensity and clean single 002 film peaks (Figures 5(a) and 5(b)). While vertical RP-type faults still exist (Figures 5(c) and 5(d)), the density of these defects is much lower than in the secondary phase. These observations indicate that single-phase Nd0.8Sr0.2NiO3 (001) films with a low density of RP-type faults can be synthesized with the two above growth conditions in a reproducible fashion.
II. Optimizing the Reduction Process for Nd0.8Sr0.2NiO2 (001) Stabilization   FIG. 6. (a) XRD symmetric θ-2θ scan of a partially optimized sample (same growth conditions as films in Figure 2(b)) with film thickness of ~ 60 nm and no SrTiO3 capping layer before (red) and after (blue) reduction (240 °C, 5 hours). (b) XRD symmetric θ-2θ scan of a capped sample grown under the high-fluence conditions with film thickness of ~ 11 nm and cap thickness of ~ 25 nm before (red) and after (blue) reduction (9 hours at 260 °C, followed by 3 hours at 280 °C). (c) Evolution of Nd0.8Sr0.2NiOx 002 peak of a high-fluence capped sample with film thickness of ~ 11 nm and cap thickness of ~ 25 nm during the reduction process (from bottom to top: as-grown, 4 hours at 260 °C, additional 3 hours at 260 °C, additional 6 hours at 280 °C). (d) Evolution of Nd0.8Sr0.2NiOx 002 peak of a low-fluence sample with film thickness of ~ 5 nm and 5 unit cells of SrTiO3 (001) capping layer during the reduction process (from bottom to top: as-grown, 0.5 hours at 240 °C, additional 1 hour at 240 °C, additional 1 hour at 240 °C).
During our soft-chemistry topotactic reduction experiments on the partially optimized films, we found the same challenges of film degradation that were observed in previous studies of undoped nickelates. 22,23 Namely, only a portion of the perovskite film is converted to the infinite-layer structure, which is identified from the significantly reduced film peak intensity in the XRD symmetric θ-2θ scan after the reduction process (Figure 6(a)).
There are several potential factors which can contribute to film degradation during reduction. If the reduction temperature Tr is too high, the films can degrade before successfully forming the infinite-layer structure. 4,5 This has been observed in the previous reduction study of undoped polycrystalline NdNiO3 samples, where decomposition to Nd2O3 and Ni occurred when Tr higher than 200 °C was employed with NaH as the reducing agent. 5 It is also possible that the infinitelayer phase is not accessible regardless of the value of Tr because the reducing agent is not reactive enough; such is the case for the reduction of NdNiO3 with hydrogen gas. 5 Therefore, the choice of an appropriate reducing agent along with careful optimization of Tr and reduction time are required to achieve the highest crystallinity infinite-layer phase. We again note the structural support at the boundaries of the film. While epitaxial strain and structural support is provided by the substrate at the bottom of the film, promoting the infinite-layer phase, this is not the case for the top of the film away from the interface, which can lead to partial film degradation and the formation of impurity phases. 22,23 These factors suggest that capping the perovskite Nd0.8Sr0.2NiO3 (001) film with SrTiO3 may be helpful for the topotactic reduction to the infinite-layer structure in various ways. The capping layer can act as a protective barrier to prevent direct exposure of the film to the reducing agent, With these considerations in mind, we grew a SrTiO3 (001) capping layer epitaxially on the Nd0.8Sr0.2NiO3 (001) film under the same Ts and PO 2 as during the nickelate film growth, keeping the film thickness below ~ 15 nm. Indeed, the XRD θ-2θ film peak intensity of the capped sample after reduction is much more prominent than that of the uncapped sample ( Figure 6); in fact, the reduced film peak intensity is almost comparable to that of the as-grown film (Figures 6(b) -6(d)).
A direct quantitative measure of how much of the film has reduced to the infinite-layer structure is the comparison between the total thickness of the perovskite phase in the as-grown film and the total thickness of the infinite-layer phase in the reduced film. X-ray reflectivity (XRR) is a standard ex situ measurement technique for obtaining film thickness. 38 However, due to the small electron density contrast between the infinite-layer phase and the secondary phase, the XRR measurements alone are unable to provide a good estimate of the infinite-layer phase thickness. 38 In addition, the presence of the SrTiO3 (001) capping layer complicates the thickness extraction from XRR.
Instead, as an approximate measure, the Scherrer equation where K is the Scherrer constant, λ is the x-ray wavelength, b is the full width at half maximum intensity of the film peak in the symmetric θ-2θ scan, and θ is the Bragg angle, can be employed to estimate how much of the film has converted to the infinite-layer phase. [39][40][41] The numerical value of the Scherrer constant K is often approximated to be 0.9, 41 but this value can vary nontrivially upon the geometric factors (i.e. size, shape, and orientation) of the crystallites. 39 approximation should also be applicable to the infinite-layer phase with reasonable accuracy.
Using this approach, we estimate that the infinite-layer phase of 8.5 nm in thickness is stabilized within the capped reduced film shown in Figure 7(b). Although slightly underestimating, this value is in reasonable agreement with the infinite-layer phase thickness of 9.3 nm measured from the cross-sectional HAADF-STEM image (Figure 7(b)). This demonstrates that the Scherrer estimate is a useful method for monitoring the crystalline film thickness ex situ non-destructively during the reduction process with reasonable accuracy. Furthermore, we note that the infinite-layer phase thickness nearly approaches the maximum possible reduced film thickness dmax of 9.7 nm, extracted from the as-grown perovskite film thickness of 10.7 nm. This corresponds to approximately 3 unit cells of unconverted Nd0.8Sr0.2NiO3 (001), which can be attributed to the interfacial layers as previously observed. 22,23 In comparison to the partial decomposition of the uncapped film upon reduction (Figure 6(a)), the crystallinity of the film with SrTiO3 capping layer shows significant improvement, with essentially the entire film transformed to the infinite-layer phase. The optimal reduction condition varies as a function of film crystallinity and the thickness of the capping layer, which appears to act as a diffusion barrier to oxygen deintercalation. Highly crystalline samples with ~ 25 nm of SrTiO3 (001) capping layer show gradual XRD peak shifts at Tr > 240 °C (Figure 8(a)), while in the limit of no capping layer a complete transition to the infinitelayer phase along with partial film degradation can occur with only 2 hours of reduction at Tr = 240 °C (Figure 8(a)). For given crystallinity and capping layer thickness, Tr should be low enough such that the film does not decompose, but also high enough such that the duration of the film exposure to reducing conditions is minimized. As a conservative approach, we performed With Nd0.8Sr0.2NiO3 nominally optimized in these two different growth conditions, we examine how the difference in the growth conditions affects the crystallinity and the superconducting transition of the resultant Nd0.8Sr0.2NiO2. Figure 9 shows the XRD symmetric θ-2θ scans of two capped samples: one grown under the high-fluence conditions (Figure 9(a)) and the other grown under the low-fluence conditions (Figure 9(b)). Both samples show prominent 001 perovskite film peaks with no double-peak feature in the 002 film peak, suggesting that the films are dominantly single-phase Nd0.8Sr0.2NiO3 (001) films.
However, we observe multiple signatures indicating that the reduced low-fluence sample has limited crystallinity compared to the high-fluence sample. First, the film peaks of the low-fluence sample are less symmetric and triangular in shape (Figure 9(b)), indicating the presence of nontrivial disorder in the film. Second, upon reduction the low-fluence sample peak intensity decreases (Figure 9(b)), in contrast with the high-fluence sample (Figure 9(a)). This decrease in the peak intensity is also an indication of limited crystallinity in the precursor perovskite phase, resulting in the degradation of film quality during topochemical reduction. This is confirmed from the HAADF STEM images of the two samples after reduction (Figure 9(c)), which show that the infinite-layer region of the low-fluence sample is much less coherent than that of the high-fluence sample. In particular, we observed inclusions and precipitates in the low-fluence sample ( Figure   9(d)), similarly to previous reports on partially optimized undoped NdNiO3 thin films. 18 The continuous propagation of disorder into the capping layer again suggests that this disorder originates from the as-grown state before reduction. The higher magnification views of these HAADF STEM images (Figure 9 ρ-T measurements on the two reduced samples reveal that the low-fluence sample has a higher superconducting transition temperature Tc than the high-fluence sample (Figures 9(e) and 9(f)).
For the high-fluence sample shown here, the superconducting transition occurs at an onset of 6.7 K, a midpoint at 5.3 K, and zero resistance at 2. There are some observations worth discussing at this point. First, the wide sample-to-sample variation in Tc in the first report 15 has now been reproducibly narrowed in this study, controlled in part by the use of precise imaging conditions for ablation. For the two perovskite phase growth conditions optimized and studied here, the high-fluence samples have significantly better crystallinity in the reduced phase. Therefore, it is somewhat surprising that the low-fluence samples show systematically higher Tc. While the origin of this distinction is yet unclear, we note the difference in the c-lattice constant of the two groups of samples, which may indicate that the distance between Ni-O planes is highly relevant for Tc. On the other hand, for further systematic studies on superconductivity and normal state properties, high-fluence conditions may be preferable given the more uniform crystallinity.

Conclusion
In summary, we have investigated the synthesis of infinite-layer nickelate Nd0.8Sr0.2NiO2 (001) epitaxial thin films. The two principal technical issues we identified were the stabilization of the doped perovskite phase, and the balance between complete topotactic reduction versus subsequent decomposition.
We emphasize that the current conditions presented may not be the global optimum for PLD growth, given the many parameters and potentially competing factors for synthesis of the perovskite phase and the reduction to the infinite-layer phase. Nevertheless, we hope the current work will be valuable to the community interested in this system. We further note that high-quality