*2.3. Characterization*

Film thickness was routinely studied using a J. A. Woollam alpha-SE spectroscopic ellipsometer (J.A. Woollam Co., Lincoln, NE, USA) in the wavelength range 390–900 nm. A Cauchy function was successfully used to model the collected data.

X-ray di ffraction (XRD) measurements were used to investigate the out-of-plane crystalline orientation of the thin films on single crystal substrates. Symmetric θ-2θ-scans were carried out on a Bruker AXS D8 Discover di ffractometer (Bruker AXS, Karlsruhe, Germany) equipped with a LynxEye strip detector (Bruker AXS) and a Ge (111) focusing monochromator, providing CuK α1 radiation.

Chemical composition was analyzed using a Panalytical Axios Max Minerals XRF system (Malvern Panalytical, Malvern, UK) equipped with a 4 kW Rh tube. Omnian and Stratos options were employed for standardless measurements of thin film cation composition.

The chemical state of the cations, particularly cobalt, was investigated by X-ray photoelectron spectrometry (XPS) using a Thermo Scientific Theta Probe Angle-Resolved XPS system (ThermoFisher Scientific, Waltham, MA, USA). The instrument was run with a standard Al K α source (hν = 1486.6 eV), and the analysis chamber pressure was maintained on the order of 10−<sup>8</sup> mbar. Pass energy values of 200 eV and 50 eV were employed for survey scans and detailed scans, respectively. The data were corrected for any drift by setting the binding energy for adventitious carbon to 284.8 eV. Data treatment and fitting were performed within the Avantage software suite (ThermoFisher Scientific). The background was fitted to a Shirley-type pseudostep function.

Cross section SEM images of the deposited films were obtained using a Hitachi SU8230 SEM (Hitachi, Krefeld, Germany) with a cold cathode field emission electron gun. The total voltage was set to 2 kV and the films were imaged by means of secondary and back-scattered electrons.

#### **3. Results and Discussion**

#### *3.1. Deposition of GdCoO3*

The development of ternary deposition processes typically requires insight into the individual growth behavior of the binary components. ALD of Co3O4, using Co(thd)2 as a precursor and ozone as the oxidizing agent, was established by Klepper et al. in the 114–307 ◦C temperature range, with a growth per cycle (GPC) of ≈ 0.20 Å/cycle at 300 ◦C [23]. For a similar Gd(thd)3-based deposition process, Niinistö et al. reported self-limiting growth for Gd2O3 films in the range from 250 to 300 ◦C, with a GPC of ≈ 0.30 Å/cycle at 300 ◦C [24]. Our attempts at deposition of the same binary processes gave reproducible GPCs of 0.16 Å/cycle and 0.37 Å/cycle for the formation of CoOx and Gd2O3, respectively, with no observed thickness gradients.

Based on these results, a series of (Gd, Co)-oxide films were deposited at 300 ◦C. The Gd(thd)3: Co(thd)2 pulsed ratio was varied systematically to identify the conditions required to obtain the desired deposited stoichiometry of GdCoO3. We employed a super cycle approach with a general super cycle, given as:

$$m \times \left[ m \times \left[ Gd(thd)\_3 + O\_3 \right] + l \times \left[ Co(thd)\_2 + O\_3 \right] \right] \tag{1}$$

where *n*, *m*, and *l* were varied to achieve the desired cation ratio.

Figure 1 shows the deposited cation ratio for Gd (cat.% Gd) of the obtained film and the GPC as a function of the pulsed cation ratio (cat.% Gd) at 300 ◦C. The relative amount of deposited Gd increases from 2 to 51 cat.% Gd in the explored pulsed cation range of 33–67 cat.% Gd. The concentration of Gd in the deposited film consistently increases with increasing amounts of pulsed Gd(thd)3, except for the plateau interval observed between 50 and 56 cat.% Gd pulsing ratio, where the Gd concentration in the product takes a constant value at around 32 cat.%. We note that the desired Gd:Co ratio of close to unity is obtained for 67 cat.% of pulsed Gd. The GPC of the deposited films at 300 ◦C increases smoothly with an increased fraction of Gd pulses, in accordance with the higher GPC of Gd2O3, see Figure 1. However, an excess of Gd pulses must be applied to achieve stoichiometric GdCoO3. We do observe a small reduction in overall GPC (0.24 Å/cycle) as compared to a linear combination of the binary oxides [(0.37 + 0.16)/2 = 0.27 Å/cycle], possibly due to either inhibition of growth from Gd(thd)3 on Co-O\* surfaces or by increased growth from Co(thd)2 on Gd-O\* surfaces, or most likely a combination of both judging from the dependency of pulsed to deposited composition in Figure 1. This is an effect seen in several ALD processes, such as reported earlier by our group in the case of LaAlO3 [25]. The GPCs obtained at 300 ◦C for (Gd, Co)-oxides are in good agreemen<sup>t</sup> with the results of Seim et al., who reported an average GPC of 0.35 Å/cycle for LaCoO3 at 350 ◦C following a similar β-diketonate and ozone deposition process [21].

**Figure 1.** Deposited cation ratio (cat.% Gd, as measured by XRF) and GPC as a function of the pulsed cation ratio (cat.% Gd) for (Gd, Co)-oxide films deposited at 300 ◦C. The dotted lines refer to deposited cation ratio (cat.% Gd) and GPC as a function of the pulsed cation ratio (cat.% Gd) for Gd2O3 film.

#### *3.2. Deposition of Gd1*−*<sup>x</sup>CaxCoO3*

Based on the results obtained for the ternary (Gd, Co)-oxide system, the quaternary (Gd, Ca, Co)-oxide system was explored in an attempt to target products with the Gd0.9Ca0.1CoO3 composition. ALD was carried out at 300 ◦C following an identical process as for ternary (Gd, Co)-oxide films, with the essential modification of substituting a number of Gd(thd)3-pulses with Ca(thd)2-pulses. The [Gd(thd)3 + Ca(thd)2]: Co(thd)2 pulsed ratio was maintained at 2:1 in order to keep the deposited (Gd + Ca): Co atomic ratio close to unity. The Ca pulsed ratio, using Ca(thd)2 as precursor, was varied from 3 to 7 cat.%. Figure 2a shows the deposited cation ratios and the GPC as a function of the Ca pulsed ratio (cat.%) for depositions at 300 ◦C. The Ca content in the films correlates fairly well with the relative amount of Ca pulses. In a few experiments deviating behavior was observed, which reflects the challenge of controlling the simultaneous growth of three different cation species [26]. However, quite a stable growth situation was obtained for the range around 4–5 cat.% Ca. The A-site (Gd + Ca): B-site (Co) stoichiometry was analyzed as function of the relative amount of Ca-pulses (Figure 2b). With the current pulsing strategy, the target (Gd + Ca): Co ratio close to unity is obtained for films deposited at a Ca pulsed ratio between 3.5 and 5 cat.%. The targeted composition Gd0.9Ca0.1CoO3

is obtained for a Ca pulsed ratio of 4.5 cat.%, for which an equiatomic ratio is maintained between the perovskite A- and B-sites.

**Figure 2.** (**a**) Deposited cation ratio (cat.% Ca, as measured by XRF) and (**b**) GPC deposited cation ratio (cat. % (Gd + Ca) and Co, as measured by XRF) as a function of the pulsed cation ratio (cat.% Ca) for (Gd, Ca, Co) films deposited at 300 ◦C. The dotted line indicates the targeted Ca deposited concentration.

#### *3.3. Characterization of GdCoO3 and Gd0.9Ca0.1CoO3 Thin Films*

#### 3.3.1. X-Ray Diffraction (XRD)

The as-prepared GdCoO3 and Gd0.9Ca0.1CoO3 films deposited at 300 ◦C are X-ray amorphous. Crystallization is achieved upon annealing at 650 ◦C for 30 min in air on LAO and YAP single crystals, resulting in preferential orientation depending on the substrate type and orientation. Figure 3a,b shows XRD patterns of post-annealed GdCoO3 and Gd0.9Ca0.1CoO3 films deposited on LAO(100)pc. The diffractograms for the crystalline films on LAO(100)pc can be indexed as orthorhombic GdCoO3 (*Pbnm*, SG# 62; Z = 4) with a preferred (010) growth orientation. The orthorhombically distorted GdCoO3 perovskite relates to the ideal cubic perovskite structure (*Pm-3m*; Z = 1) as *ao* = √2 × ac, mboxemphbo = 2 × *bc*, *co* = √2 × cc with dimensions *ao* = 5.380 Å, *bo* = 7.437 Å and *co* = 5.210 Å. The (rhombohedral) LAO substrate exhibits a pseudo cubic structure *ac* = 3.79 Å (note √2 × *ac* = 5.36 Å). The growth of GdCoO3-based perovskites onto LAO is favored in the (010) orientation as the *a*and *c*-axis of the film match the diagonals of the cube faces of the substrate. In this configuration, GdCoO3 will experience a lattice expansion of 2.5% in order to match the diagonal by diagonal area of the LAO substrate (ALAO = *ac<sup>2</sup>* = 28.73 Å2 and AGCO = *ao* × *co* = 28.02 Å2). The position of the (020) and (040) reflections indicate that *bGCO*||*LAO(100)* = 7.42 Å (strain−0.2%), which indicates a small compression compared to the theoretical orthorhombic structure. This is in good agreemen<sup>t</sup> with the expected expansion in *a*. We used Scherrer's formula on the well-defined GdCoO3 (040) reflection (Supporting Figure S1) to estimate a crystallite size of 24.8 nm, which indicates that the crystallites traverse from the substrate to the film surface. A higher degree of crystallinity is observed for GdCoO3, which exhibits sharper and more intense (020) and (040) reflections than Gd0.9Ca0.1CoO3. This is in good agreemen<sup>t</sup> with Bretos et al., who reported a slower crystallization process for Ca-substituted perovskites [27].

**Figure 3.** XRD patterns of (**a**) 30 nm GdCoO3 and (**b**) 30 nm Gd0.9Ca0.1CoO3 films grown on LAO(100), (**c**) 30 nm GdCoO3 grown on YAP(100) and (**d**) 30 nm GdCoO3 grown on YAP(001), post-annealed for 30 min at 650 ◦C. Bragg reflections originating from the substrate are marked with a star; film reflections are marked with their designated plane of reflection.

Figure 3c,d shows the measured XRD patterns from crystalline GdCoO3 films deposited on YAP (100) and YAP (001), respectively, after post annealing at 650 ◦C in air for 30 min. Both YAP and GdCoO3 are orthorhombic perovskites and exhibit quite similar unit cell dimensions; for YAP, *aYAP* = 5.330 Å, *bYAP* = 7.375 Å and *cYAP* = 5.180 Å. GdCoO3 deposited on YAP(001) grows with a preferred (001) orientation, whereas GdCoO3 deposited on YAP(100) exhibits a preferential (100) growth orientation. Thus, the preferential (001) growth orientation of GdCoO3 onto an oriented YAP(001) substrate is favored due to a minimized lattice compressive stress of 1.7% in this configuration (VYAP(001) = *aYAP* × *bYAP* = 39.31 Å2 and VGCO = *ao* × *bo* = 39.97 Å2). On the other hand, a (100) orientation of the substrate results in the growth of GdCoO3 in a preferential (100) orientation with a lattice compression of 1.2% (VYAP(100) = *bYAP* × *cYAP* = 38.20 Å2 and VGCO = *bo* × *co* = 38.64 Å2). The close lattice match means that the film reflections are observed as a broadening of the substrate peaks, making it difficult to analyze the diffraction in terms of crystallite size or strain. The Gd0.9Ca0.1CoO3 thin films deposited on YAP substrates exhibited too poor a crystallinity, even after annealing, to be properly indexed (see Supplementary Figure S2).

By use of this appropriate selection of substrates, we have demonstrated the preferred crystallization along all three crystallographic axes. The effect of surface structure on catalytic activity is well known, so the ability to select the growth orientation of crystalline films may be of high importance.

The physical properties of gadolinium cobaltites depend, inter alia, on the temperature and cation substitutions, type, and concentration, which in turn may have a profound effect on the catalytic performance. For instance, an expanded lattice triggered by the substrate may stabilize the high-spin Co(III) configuration at temperatures lower than 800 K, i.e., the transition temperature for bulk GdCoO3 [28]. Such scenarios are interesting from an ALD perspective, since key physical and chemical performance properties may be tuned by the choice of appropriate lattice-matching substrates.
