Complex oxide perovskites exhibit a wide range of magnetic properties. Some are attractive for spintronics applications, especially in an oxide electronics platform due to the ease of integration via epitaxial, isostructural growth. Among the perovskites, ferromagnetic manganites have been identified as candidate materials for pure spin current generation, and 4d and 5d transition metal oxides as candidates for pure spin current detection by electrical means. Recent studies [1][2][3][4][5][6][7] have shown that (La 2/3 Sr 1/3 )MnO 3 (LSMO) is particularly attractive for spin current generation with Gilbert damping parameter as low as α ∼ 1 × 10 -3 . This low damping is comparable to damping in ferromagnets such as permalloy and yttrium iron garnet, with typical damping parameters of 6 × 10 -3 and 10 -4 -10 -3 respectively. 8,9 Moreover, integration with a wide variety of oxide perovskites 10 makes LSMO even more promising for efficient generation of pure spin currents.

Promising candidate materials to electrically detect spin currents from LSMO include oxide perovskites with high spin orbit coupling such as the ruthenates and iridates. A previous FMR study on bilayers of LSMO and SrRuO 3 (SRO) 1 has shown successful demonstration of spin pumping from LSMO into SRO with interfacial spin-mixing conductance values comparable to other systems at room temperature. The increased damping observed with increasing SRO thickness may be attributed to proximity-induced magnetism which is plausible as ferromagnetism is stabilized in SRO at low temperature. Other studies of LSMO/SrIrO 3 (SIO) bilayers have also demonstrated enhanced Gilbert damping as well as proximity-induced magnetism in SIO. 3 CaRuO 3 (CRO) is a metallic ruthenate without long-range magnetic order. The smaller lattice parameter of CRO, compared to SRO and SIO, makes it a closer lattice match to LSMO, thus reducing the effects of strain at the coherent interfaces. The smaller Ca ions, compared to Sr, also cause significant distortions to the perovskite crystal structure, changing the orbital overlap, density of states, and exchange interactions within the material. 11 The absence of long range magnetic order in CRO should reduce the effects of proximity-induced magnetism at the interfaces.

In this paper, we study spin pumping in LSMO/CRO bilayers of varying CRO thickness grown on LSAT substrates, using FMR to generate spin currents and probe magnetism. We observe evidence of spin pumping through the increase in linewidth and effective Gilbert damping with the addition of the CRO layer. We also find that the magnetic anisotropy of the LSMO reverses sign with increasing CRO layer thickness, indicating that the CRO overlayer modifies the underlying LSMO. However this evolution in magnetic anisotropy is not observed in the static bulk magnetization measurements. This discrepancy suggests a depth dependent evolution of magnetic anisotropy. The high frequency magnetic properties of these bilayers highlights the importance of interfacial strain between layers and how such unexpected phenomena can be exploited in the future applications.

Bilayers of LSMO (22 nm thick) with varying thickness of CRO on top were grown on (001)pc-oriented (LaAlO 3 ) 0.3 (Sr 2 AlTaO 6 ) 0.7 (LSAT) substrates, purchased from Crystec GmbH and prepared by sequential ultrasonication in acetone, methanol, and isopropyl alcohol. The films were grown by pulsed laser deposition (PLD) using a 248 nm KrF excimer laser. LSMO was deposited at a temperature of 750 ○ C and O 2 pressure of 320 mTorr, using a laser fluence of 1.25 J/cm 2 and repetition rate of 1 Hz. The LSMO was then postannealed at 600 ○ C in 100 Torr O 2 for 10 minutes. After cooling to room temperature, the chamber was pumped back down to base pressure without breaking vacuum. CRO films were then deposited at 650 ○ C in 60 mTorr O 2 , using a laser fluence of 1.14 J/cm 2 and repetition rate of 3 Hz. The CRO deposition was followed by another post-anneal at 600 ○ C in 100 Torr O 2 for 10 minutes before cooling. We also grew a single layer of LSMO without a CRO layer as a reference followed by a 5 minute exposure to the CRO growth conditions. The target-heater distance was 3 inches, with a laser spot size of approximately 8.5 mm 2 . These growth conditions were optimized for conduction of the CRO layer and ferromagnetic resonance (FMR) signal of the LSMO layer. The two-step in situ procedure was developed because conditions required for metallic CRO caused oxygen-deficiency in the LSMO layer and reduced the magnetization and FMR signal.

Structural properties were characterized by a variety of methods. X-ray diffraction (XRD) 2θ -θ and ω scans were measured to investigate crystalline phases and measure film lattice parameters. Xray reflectivity (XRR) was performed to measure layer thickness and interface roughness. XRD and XRR were performed on PANalytical Empyrean and X'Pert Materials Research Diffractometers. Surface roughness and morphology were measured immediately after sample growth by atomic force microscopy (AFM), using a Veeco Dimension 3100 in tapping mode.

Static magnetic properties were measured at varying temperatures and external magnetic fields with the reciprocating sample option (RSO) of a Quantum Design Magnetic Property Measurement System (MPMS). The sample mounting was varied to measure along the (100) (in-plane, IP), (110) (IP-45 ○ ), and (001) (out-of-plane, OOP) sample axes. Hysteresis loops were measured at 300 K and the diamagnetic background from the LSAT substrate was subtracted.

FMR was used both to generate spin currents and to measure dynamic magnetic properties of the material. Broadband FMR measurements were performed at room temperature in a coplanar waveguide geometry 12 with the sample at different orientations relative to the magnetic field. The resulting spectra were fit to the lock-in "derivative" of the sum of the symmetric and antisymmetric components of a Lorentzian, and used to determine the resonant field H FMR , and linewidth ΔH.

XRR measurements indicate that the LSMO layers are on average 22 nm thick, while the CRO layers range between 0 and 11 nm. In addition, the XRR spectra allow us to investigate the interfacial roughness, as well as density differences between the layers. The best-fit models required an additional low density, high roughness surface layer on top of the CRO film, LSMO film, and LSAT substrate. Additionally, we see that the CRO surface becomes smoother as the CRO layer is grown thicker.

These results agree with surface morphology as seen in AFM images. The LSMO films show signs of 3D island growth 13 and have a root-mean-square roughness of 1.9 nm (5 unit cells). These films also contain rectangular pinhole-like features oriented at 90 ○ to one another; this is likely due to the high deposition fluence and fast growth rate, which were optimized for FMR signal strength rather than crystalline or surface quality. Due to the initial surface roughness of these LSMO films, the CRO growth on top of LSMO starts out rough, but becomes smoother as the CRO layer grows thicker. Rough step-like edges form once the CRO film has grown over 5 nm thick, while the pinhole-like features fully smooth over once the CRO film has grown over 10 nm thick.

Despite the large surface roughness, XRD spectra show highly crystalline and epitaxial LSMO and CRO layers on LSAT substrates, with only the (00l)pc (pseudocubic) family of peaks in the out-ofplane direction for both LSMO and CRO layers. This result confirms that both layers are grown with the perovskite crystal structure and (00l) orientation, and no secondary phases, such as RuO 2 , have formed. All films have a rocking curve full-width-at-half-max comparable to the LSAT substrate, indicating excellent crystallinity. Finally, Laue oscillations, a sign of good interface quality, can be seen on either side of the peaks.

Both LSMO and CRO are found to be very well lattice-matched to the LSAT substrate. Figure 1 shows XRD spectra of a bare LSMO film as well as LSMO bilayers with 4.4 nm and 11.1 nm of CRO. As the CRO thickness increases, the LSMO peak, most visibly the (001)pc peak, shifts to higher angles, indicating a contraction of the out-of-plane lattice parameter from 3.894 Å to 3.882 Å, compared to the bulk value of 3.876 Å. 14 The ∼20 nm LSMO layers are thin enough to be compressively strained to the LSAT substrate (a = b = c =3.868 Å), 15 so the change in lattice parameter suggests that the CRO film imposes additional interfacial strain and structural changes in the LSMO layer.

In order to probe whether these structural changes in LSMO caused corresponding magnetic changes, static magnetization measurements were performed. Room-temperature hysteresis loops were taken along the out-of-plane ⟨001⟩ and in-plane ⟨100⟩ and ⟨110⟩ directions, shown in Figure 2. For all samples, the out-of-plane direction is the magnetically hardest one due to shape anisotropy. From the saturation field in the ⟨001⟩ direction, we find that the out-of-plane magnetic anisotropy H ,2 decreases.

The in-plane magnetically easy axis for each sample is the ⟨110⟩ axis, as expected for LSMO films grown on (001)-oriented LSAT substrates. 16,17 The addition of the CRO increases the coercive field, which then decreases as the CRO layer becomes thicker. The differences between the ⟨100⟩ and ⟨110⟩ orientations -in coercive field and remanence -are slight, suggesting a small in-plane anisotropy H ∥,4 .

As the presence of the CRO layer does not appear to adversely affect the magnetic properties of LSMO, FMR measurements were performed on the LSMO/CRO bilayers as well as the LSMO reference thin film in order to study spin pumping in this system. With the magnetic field pointing in the out-of-plane direction, the FMR linewidth and resonance field (Figure 3a,b) were measured as a function of frequency. The linewidths ΔH( f ) display a linear dependence on frequency 1 with the slope proportional to α, the Gilbert damping parameter.

The Gilbert damping parameters for each sample were found to be very sensitive to deposition conditions. It should be noted that the we observe damping parameters as low as 6.9 ×10 -4 , comparable to the lowest LSMO damping found in the literature, 2 and as high as 1 ×10 -3 , still comparable to other high-quality LSMO in the literature. 3,6 As the CRO layer thickness increases, the total linewidth increases as seen in Figure 3b, despite the variation in α. Between comparable samples, the increase in α with increasing CRO thickness matches the change expected due to spin pumping from LSMO into CRO.

The frequency dependence of the FMR linewidth measured with an in-plane field provides us with additional insight into spin scattering mechanisms within LSMO. The non-linear behavior of the frequency dependence (Figure 4a) can be explained in terms of two-magnon scattering, often attributed to defects that can cause spins to scatter and process out of phase with one another. [18][19][20] This non-linear scattering process significantly dominates over Gilbert damping as the primary spin scattering mechanism with in-plane field. Differences between ΔH( f ) with the field along the in-plane ⟨100⟩ and ⟨110⟩ directions indicate an anisotropy in the twomagnon scattering process, with more scattering occurring along the ⟨100⟩ axes. This symmetry is consistent with rectangular defects (for example, pits or islands) oriented primarily along the ⟨100⟩ rather than the ⟨110⟩ axes; 20 evidence of these types of pits on the LSMO surface are visible in the surface morphology of LSMO and of bilayers with thin CRO layers.

The frequency dependence of the out-of-plane FMR resonance field measurements enables us to deduce the effective magnetization, M eff , from the y-intercept of H FMR ( f ). M eff = MS -H ,2 depends on the total saturation magnetization of the material MS as well as the out-of-plane magnetic anisotropy H ,2 . As the CRO layer thickness increases, M eff gradually increases (Figure 3d), suggesting that MS is increasing, H ,2 is decreasing, or both. Static magnetometry measurements indicate that the increase in M eff can be largely attributed to a change in out-of-plane anisotropy.

We also probed the in-plane magnetization dynamics of these bilayers as a function of in-plane field direction to further understand the role of magnetic anisotropy in this system. Figure 4b shows the in-plane resonance field for all samples taken at different angles at a fixed frequency f = 5 GHz. The resonance field exhibits a clear four-fold symmetry within the plane, showing a cubic (biaxial) inplane magnetic anisotropy H ∥,4 and no apparent uniaxial anisotropy within the plane. This symmetry is consistent with the four-fold symmetry of the crystal structure when grown in the (001) orientation, and has been seen in the LSMO/SrRuO 3 system grown on LSAT 1 but not LSMO grown on other substrates. 7,21 At a fixed angle, such as along ⟨110⟩, the resonance field varies between samples due to the differences in anisotropy strength. The resonance field is minimized along the easy axis, with the sign convention that a positive (negative) H ∥,4 corresponds to the easy axis of the sample pointing along the ⟨100⟩ (⟨110⟩) direction. In Figure 4b, starting with bare LSMO (blue) the magnetism within the plane appears quite isotropic. As the CRO layer grows thicker, up to 3.4 nm (orange, purple, red), the four-fold symmetry appears to develop and increase in magnitude, indicating an anisotropy field H ∥,4 that is growing more negative. This indicates that the easy axis of these samples is along ⟨110⟩, which has been well-established for LSMO films grown on LSAT. 22 As even more CRO is deposited (cyan), the magnitude of the anisotropy decreases again, until it switches sign (green). This switch in the sign of the anisotropy field indicates a change in the easy axis of the film to ⟨100⟩, though this is not visible in static magnetometry. Since CRO is not magnetic at room temperature where FMR scans are taken, the evolution in magnetic anisotropy must be occurring in the LSMO layer. One possible explanation for the difference in in-plane magnetic anisotropy determined by static magnetization loops versus FMR measurements is a thickness dependent structural change imposed by the CRO layer. This change would vary continuously with CRO thickness, in contrast to an interfacial effect due to proximity with CRO. While SQUID magnetometry measures the total moment of the film, thereby taking a mean average over the sample, FMR focuses on the most prominent peak in the spectrum, measuring a mode-like average over the sample. A magnetic anisotropy that varies through the depth of the sample may result in conflicting results between these two methods.

Due to the close lattice match of LSMO, CRO, and LSAT, the depth-dependent lattice distortion in the LSMO is likely due to oxygen octahedra distortions. Previous studies have found that growth of LSMO on NdGaO 3 substrates (orthorhombic, c + a -a -, with a bond angle of 154 ○ ) causes (i) a rotation of the oxygen octahedra into the LSMO, (ii) a decrease in TC, (iii) a decrease in MS, and (iv) a change in the in-plane easy axis. 22 Bond-angle changes have also led to changes in the out-of-plane magnetic anisotropy of LSMO. Others have shown that another orthorhombic material, SrIrO 3 (a -b + a -, 154.1 ○ ) 23,24 can even cause perpendicular magnetic anisotropy (easy axis out-of-plane) in a unit cell of LSMO 25 as well as the rotation of the in-plane easy axis. 26 Through rotations of the oxygen octahedra, orbital overlap and exchange coupling in the LSMO can be significantly changed, thus giving rise to changes in the magnetic anisotropy and moment of LSMO.

In our case, the LSAT substrate is cubic: LSMO is rhombohedral in the bulk, with an a -a -a -rotation pattern -neighboring octahedra along all axes rotate in opposite directions (out of phase) -with an Mn-O-Mn bond angle of 166.3 ○ ; 22 CRO is orthorhombic in the bulk, with an a -a -c + rotation pattern -neighboring octahedra rotate the same amount in opposite directions along the a and b axes, but along the c axis, the octahedra rotate together (in phase) -with a Ru-O-Ru bond angle of 149.8 ○ along a and b, and 149.6 ○ along c. 27 CRO has a much larger bond-angle deviation from LSMO than either NdGaO 3 or SrIrO 3 and therefore could also give rise to distortions and magnetic changes in LSMO. These lattice distortions are a result of competition between the LSMO and CRO structures. When the CRO layers are very thin compared to the LSMO, the CRO will take on the rotation pattern of the underlying LSMO lattice. Thicker CRO overlayers may increasingly distort the LSMO surface layers, modifying the measured magnetic anisotropy in FMR measurements more than in SQUID measurements. The differences in magnetic anisotropy measured by FMR are observed once the CRO layer is sufficiently thick, in this case around half the thickness of the 20 nm LSMO layer.

In summary, we developed LSMO films with low intrinsic damping and combined them with CRO layers. The increase in the effective Gilbert damping with CRO thickness is indicative of spin pumping into the CRO. Most significantly, the addition of the CRO layer can change the magnetic properties of LSMO, reducing the out-of-plane anisotropy and rotating the magnetic easy axis within the plane. These results are consistent with a structural change (as seen by XRD) imposed on the LSMO by the CRO layer as it becomes thick enough to influence the underlying magnetic layer. Such structurally-induced magnetic changes may be tunable -for example through magnetoionics 28,29 or integration with ferroelectric materials [30][31][32] -and are potentially important to the development of these architectures. Therefore the epitaxial integration of different oxide materials offers additional degrees of freedom which may be exploited to engineer functional devices.