Sub-micrometer yttrium iron garnet LPE films with low ferromagnetic resonance losses
Carsten Dubs, Oleksii Surzhenko, Ralf Linke, Andreas Danilewsky, Uwe Brückner, Jan Dellith
SSub-micrometer yttrium iron garnet LPE films with low ferromagneticresonance losses
Carsten Dubs, Oleksii Surzhenko, Ralf Linke, Andreas Danilewsky, Uwe Br¨uckner, and Jan Dellith INNOVENT e.V., Technologieentwicklung, Pr¨ussingstr. 27B, 07745 Jena, Germany Kristallographie, Albert-Ludwigs-Universit¨at Freiburg, Hermann-Herder-Str. 5, 79104 Freiburg,Germany Leibniz-Institut f¨ur Photonische Technologien (IPHT), Albert-Einstein-Str. 9, 07745 Jena,Germany (Dated: 30 August 2016)
Using liquid phase epitaxy (LPE) technique (111) yttrium iron garnet (YIG) films with thicknesses of ≈
100 nmand surface roughnesses as low as 0.3 nm have been grown as a basic material for spin-wave propagationexperiments in microstructured waveguides. The continuously strained films exhibit nearly perfect crys-tallinity without significant mosaicity and with effective lattice misfits of ∆ a ⊥ /a s ≈ − and below. Thefilm/substrate interface is extremely sharp without broad interdiffusion layer formation. All LPE films ex-hibit a nearly bulk-like saturation magnetization of (1800 ±
20) Gs and an ‘easy cone’ anisotropy type withextremely small in-plane coercive fields < < × − .PACS numbers: 81.15.Lm, 75.50.Gg, 76.50.+g I. INTRODUCTION
Magnonics is an increasingly growing new branchof spin-wave physics, specifically addressing the use ofmagnons for information transport and processing .Single crystalline yttrium iron garnet (YIG), which is aferrimagnetic insulator with the smallest known magneticrelaxation parameter , appears to be a superior candi-date for this purpose . As bulk or as thick film mate-rial, which is commonly grown by liquid phase epitaxy(LPE) , it has a very low damping coefficient and allowsmagnons to propagate over distances exceeding severalcentimeters . However YIG functional layers for practi-cal magnonics should be nanometer-thin with extremelysmooth surfaces in order to achieve optimum efficiencyin data processing and dramatic reduction in energy con-sumption of sophisticated spin-wave devices. Therefore,high-quality thin and ultra-thin YIG films were grown us-ing different growth techniques such as LPE, pulsed laserdeposition (PLD) and rf-magnetron sputtering to inves-tigate diverse spin-wave effects and to design YIG waveg-uides as well as nanostructures for spin wave excitation,manipulation and detection in prospective magnonic cir-cuits.From previous reports about sub-micrometer YIGfilms with thicknesses between 100 and 20 nm avail-able microwave and magnetic key parameters were takenand summarized in Table I. Thus, ferromagnetic reso-nance (FMR) data were included which have been ex-tracted from measurements of the absorption curves orabsorption derivative curves versus sweeping magneticin-plane field H at a fixed frequency f or vs. sweep- ing rf-exciting field h rf with an applied in-plane staticmagnetic bias field. The reported FMR linewidths ∆ H and converted peak to peak linewidths ∆ H p − p of thefield derivative values (∆ H = √ H p − p ), which willbe given during the further paper as full-width at half-maximum ∆ H FWHM , varied between 3 Oe and 13 Oe.The Gilbert damping coefficient α were found in therange from 2 × − to 8 × − . Only the lowest given α value of 0 . × − was obtained for a very short fit rangeof about 4 GHz without any given data in the low fre-quency range below 10 GHz and is therefore not reallycomparable with the other reported values. From thiscompilation it is obvious that neither ∆ H FWHM nor α issignificantly influenced by the YIG film thickness downto 20 nm. The differences are probably resulted fromadditional ferromagnetic losses due to contributions ofhomogeneous and/or inhomogeneous broadening by mi-crostructural imperfections or magnetic inhomogeneities.In this report we present microstructural, magneticand FMR properties of LPE-grown 100 nm thin YIG andLanthanum substituted (La:YIG) films with low ferro-magnetic resonance losses. Film thicknesses were deter-mined by X-ray reflectometry (XRR) and surface rough-ness by atomic force microscopy (AFM) measurements.Crystalline perfection and compositional homogeneitywere investigated by high-resolution X-ray diffraction(HR-XRD) and X-ray photoelectron spectroscopy (XPS)as well as by secondary ion mass spectroscopy (SIMS).Static and dynamic (microwave) magnetic characteriza-tions were carried out by vibrating sample magnetometry(VSM) and by Vector Network Analysis (VNA), respec-tively. a r X i v : . [ c ond - m a t . m t r l - s c i ] A ug TABLE I: Key parameters reported for thin/ultrathin YIG films on (111) GGG substrates
Growth method Thick- RMS- 4 πM s a H c a ∆ H a f ∆ H α (Reference) ness roughness FWHM FWHM × − (nm) (nm) (kGs) (Oe) (Oe) (GHz) (Oe)LPE
100 - 1.81 - 3.0 7 1.6 2.8LPE (this study) 83–113 0.3–0.8 1.78–1.82 ≤ b 11
79 0.2 1.72 <
23 - 1.60 < c c
22 0.13 1.78 0.4 12 c c
20 0.2 - 0.4 13 c c
20 0.2–0.3 2.10 0.2 3.3 c c a Measurements at RT with the in-plane external magnetic field H b YIG films grown on the (100) GGG substrates c Peak-to-peak value ∆ H p − p of the derivative of FMR absorption transformed into ∆ H FWHM = ∆ H p − p ×√ II. RESULTSA. Microstructural properties
Selected microstructural and magnetic properties ofliquid phase epitaxial grown YIG (sample A-C) andLa:YIG (sample D) films are given in Table II. The con-sistent magnetic as well as microwave properties obtainedfor films deposited during different growth runs demon-strate a high reproducibility of the LPE growth tech-nique. Fig. 1a shows XRR plots of films with thicknessesof about 100 nm which are smaller than the previously re-ported thinnest LPE YIG films . The smallest root-mean-square (RMS) surface roughness of about 0.25 nmobtained for the sample B in Fig. 1b is nearly compa-rable with epi-polished GGG substrate quality of ≈ ω -scan (rocking curve) with a Gaussian-like fit-ted GGG substrate 444 reflection and a second fittedpeak at the right shoulder which corresponds to the YIG444 film reflection. This indicates a tensile stressed YIGfilm because of the smaller film lattice parameter com-pared to the commercially available Czochralski-grownGGG substrate ( a s =1.2382 nm). For La:YIG films we ob-served a perfect pseudo-Voigt fitted substrate peak with-out any additional shoulder (not shown) which indicates d = 97 nmd = 83 nm I n t en s i t y ( c p s ) Position (Cu-KL ) Sample CSample D (a)(b)
FIG. 1: (a) XRR plots of sub-micrometer-thick YIGLPE films. (b) 5 × µm AFM surface topography ofsample B with RMS roughness of 0.25 nm.a perfect lattice match between substrate and LPE film.This is in remarkable contrast to YIG films depositedby various gas phase techniques such as PLD and rf-TABLE II: YIG/La:YIG film properties grown on (111) GGG substrates by LPE technology
Thick- RMS- Relative lattice VSM a FMR a Sample ness roughness misfit ∆ a ⊥ /a s πM s H c πM eff ∆ H FWHM b ∆ H α (nm) (nm) × − (kGs) (Oe) (kGs) (Oe) (Oe) × − A 113 0.8 4.7 1.82 0.10 1.637 1.4 0.5 1.4B 106 0.3 1.8 1.78 0.20 1.658 1.5 0.7 1.2C 83 0.6 0.3 1.82 0.16 1.672 1.4 0.5 1.6D c
97 0.8 0.0 1.78 0.18 1.712 1.6 0.7 1.7Accuracy ± ± . ± . ± . ± . ± . ± . ± . ± . a VSM and FMR measurements at room temperature with applied in-plane magnetic field b FMR linewidth value at frequency f =6.5 GHz c La:YIG LPE film sputtering on GGG substrates. For those filmsthe YIG reflection has always been detected at consider-ably lower Bragg angles compared to the GGG substrateindicating a significant distortion of the cubic YIG garnetcell with significantly enlarged lattice parameters (com-pressive stress) .The relative effective misfit ∆ a ⊥ /a s = ( a s − a ⊥ YIG ) /a s obtained from strained film lattice parameter in growthdirection a ⊥ YIG and the substrate lattice parameter a s canbe used as a measure for epitaxial induced in-plane ten-sion or strain. Due to YIG Poisson’s ratio of ν P = 0 . bulk lattice parameter a YIG = 1 . should have a relative effective misfit of ∆ a ⊥ /a s =+11 × − (tensile stress). In the case of our sub-micrometer YIG films ∆ a ⊥ /a s has been determined tobe in the range between zero and +5 × − (see Ta-ble II) compared to PLD-grown YIG films with up to∆ a ⊥ /a s = − × − (see e.g. Ref.12). Hence, ourLPE films are under tension but not to the extent whichwe expected for nominally pure YIG material withoutadditional lattice expansion by lattice defects or impu-rities. To find the reason for this, high-resolution re-ciprocal space map (HR-RSM) and XPS investigationswere performed. Fig. 2b shows a HR-RSM plot aroundthe symmetrical 444 Bragg reflection with symmetricaldiffracted intensity for the GGG substrate and asym-metric diffracted intensity toward higher scattering an-gles along Q z (2 θ - ω -Scan) which we attribute to theYIG 444 film reflection. Broadening of the film reflec-tion along Q z is due to the finite coherence lenght ofthe sub-micrometer thin film in growth direction andother broadening mechanisms as for example heteroge-neous strain. The extension of the film reflection upto the substrate peak position suggests that the film iscontinuously strained due to an existing compositionaland/or strain gradient. No peak broadening along the Q x direction ( ω -scan) indicates single crystalline perfec-tion parallel to the film plane without significant mosaic-ity due to tilts of epitaxial regions with respect to oneanother.To evaluate the compositional homogeneity along the YIG 444
FWHM=0.011° I n t en s i t y ( c p s ) Angle ω (deg) Sample B Gaussian fit GGG Gaussian fit YIG Cumulative fitFWHM=0.0058°
GGG 444 (a) -0.002 -0.001 0.000 0.001 0.0025.5905.5925.5945.5965.5985.6005.602 Q X (1/nm) Q Z ( / n m ) (b) FIG. 2: (a) HR-XRD ω -scan around substrate/film 444Bragg reflection of sample B. By fitting procedures theYIG film peak has been extracted. (b) HR-RSM scansaround substrate/film 444 reciprocal point revealasymmetric diffracted intensities towards higher Q z values for sample A.growth direction of the films and to detect expectedimpurities (e.g. Pb from solvent) depth profile analy- substrate I n t en s i t y ( a . u . ) Etch time (s)
Y 3p3 Fe 2p O 1s Gd 3d5 Ga 2p3 Pb 4f < 5 nmYIG/GGGinterfaceYIG film (a)
Ga substrateLa:YIG film La:YIG/GGG interface I n t en s i t y La , G a ( a . u . ) Etch time (s)
La < 11 nm x 100 (b)
FIG. 3: (a) XPS depth profile of sample B reveals avery narrow interface between film and substrate. ThePb 4f signal could not be detected within the detectionlimit of about 0.1 at-%. (b) SIMS depth profile analysisdetects the
La signal of the film as well as the Gasignal of the substrate (sample D) and their changes atthe film/substrate interface.ses were carried out by XPS. Fig. 3a shows a homo-geneous distribution of the YIG matrix elements alongthe film growth direction and a sharp transition at thefilm/substrate interface. The obtained width of the tran-sition layer for sample B is below 5 nm. But the obtaineddepth profile consists of a convolution of the true concen-tration profile with the depth resolution of the XPS sys-tem under the concrete measuring conditions and shouldbe narrower. Therefore, these profiles demonstrate thatno broad interdiffusion layer is formed by element in-termixing at the interface at an early state of epitaxialgrowth or by diffusion of substrate ions into the epitaxiallayer and vice versa during the subsequent growth pro-cess.Whereas XPS surface analysis of the very first atomiclayers (not shown) gives a Pb content of about 0.2 at-%,no Pb signal could be observed during the depth profile analyses within the detection limit of 0.1 at-% . There-fore, it is assumed that the Pb signal corresponds toa surface contamination of condensed PbO vapor fromhigh temperature solution and this contamination is com-pletely removed by the first argon-ion etching step. ForYIG films grown in La O containing solution no La sig-nal could be detected by XPS that give indicates thatthe La content must be below 0.5 at-% . In order toimprove the detection capability additional qualitativeSIMS measurements were carried out. Due to the result-ing sputtering effect and by time-dependent detection ofthe sputtered sample ions one obtains depth profiles ofthe film elements as shown for La in Fig. 3b. Here, thecounts of two separate measurements taken under identi-cal measuring conditions at neighboring sample positionswere added up in order to enhance the statistical signifi-cance. It is clearly visible that the lanthanum signal de-creases at the film/substrate interface whereas substratesignals like . Ga simultaneously increase.
B. Static magnetic measurements
The vibrating sample magnetometry was used to mea-sure the net magnetic moment m of the YIG/GGG sam-ples at room temperature. As a thickness of GGG sub-strates ≈ H and (ii) to preferthe in-plane sample orientation that ensured considerablylower fields H s for the YIG films to attain the saturation.Fig. 4a presents a typical dependence of the total mag-netic moment m vs the in-plane magnetic field H andillustrates the method allowing us to separate the m ’components produced by the YIG film and the GGGsubstrate. Being subsequently normalized to the filmvolume, the YIG component loops yield the followingmaterial parameters – a saturation magnetization M s , acoercivity H c and a saturation field H s , i.e. the field (av-eraged over ascending H ↑ and descending H ↓ branchesof hysteresis loops) where the YIG film magnetizationapproaches 0.9 × M s . In order to estimate the in-planeanisotropy, we have repeated this procedure for the sam-ples rotated around the (cid:104) (cid:105) axis perpendicular to filmsurfaces. Fig. 4b demonstrates such results as polarsemi-log plots vs the azimuthal angle ϕ . A saturationmagnetization M s in Fig. 4b seems independent of ϕ .The obtained 4 πM s values cluster around 1800 Gs usu-ally reported for bulk YIG single crystals. Within anexperimental error (mostly defined by the YIG volumeuncertainity of ± πM s val-ues in other LPE films listed in Table II. The obtainedcoercivity ( H c ≤ . H c values is also registered. In contrast,the azimuthal dependence of the saturation field H s ob- -150 -100 -50 0 50 100 150-800-600-400-2000200400600800 -2 0 2-2000200 m ( µ e m u ) H (Oe)
YIG+GGG YIG GGG m ( µ e m u ) H (Oe) (a) H C , H S ( O e ) M S ( k G s ) s H s H c (b) FIG. 4: (a) The net VSM magnetic moment m of thesample D as well as its components induced by the YIGfilm and the GGG substrate vs the in-plane magneticfield H parallel to the (cid:104) (cid:105) direction. (b) Azimuthalangle dependencies for the VSM loop parameters ofsample D, i.e. a saturation magnetization M s , asaturation field H s and a coercivity H c . The H s six-foldsymmetry with the mimima along the (cid:104) (cid:105) ‘easy axes’and the maxima along the (cid:104) (cid:105) ‘hard axes’ indicatesthe cubic magnetocrystalline anisotropy.viously reveals the six-fold symmetry which matches thecrystallographic symmetry of YIGs. The H s maxima co-incide with the in-plane (cid:104) (cid:105) projections of the hardmagnetization axes, whereas the H s minima correspondto the (cid:104) (cid:105) crystallographic directions. The (cid:104) (cid:105) ‘easyaxes’ orientation suggests an ‘easy cone’ anisotropy afterUbizskii . He has also demonstrated that relativelysmall in-plane magnetic fields lead to single-domain YIGfilms, although a deviation of magnetization vector fromthe film plane still remains due to finite values of thecubic anisotropy constants.In conclusion, as the demagnetizing factor at the out-of-plane YIG film orientation is 1, the out-of-plane satu-ration field has to be close to the in-plane 4 πM s values.This fact is qualitatively confirmed by our out-of-planemeasurements. Unfortunately, the GGG component of the total VSM signal at fields H ⊥ ≈ . ≈
100 nm (see, for instance, Fig. 4a) and, hence,a reasonable accuracy of ± − . C. FMR absorption
FMR absorption spectra for each of studied YIG filmswere recorded at several values ( H ≤ H = 1 . f FWHM ≈ f ≈ . f of measured spectra andthe corresponding in-plane fields H to estimate the gy-romagnetic ratio γ and the effective magnetization M eff in the Kittel formula f = γ (cid:112) H ( H + 4 πM eff ) . (1)Then, the best fitting pair of γ and M eff allowed us (i) toconvert every frequency spectrum into the magnetic fieldscale, (ii) to fit rescaled spectra with the Lorentz functionand (iii) to evaluate, thereby, the corresponding linewidth∆ H FWHM . The selected results of the described proce-dure – namely, 4 πM eff and ∆ H FWHM at the referencefrequency f = 6 . H FWHM values is
A: 1.4 x10 -4 ; 0.5 Oe B: 1.2 x10 -4 ; 0.7 Oe C: 1.6 x10 -4 ; 0.5 Oe D: 1.7 x10 -4 ; 0.7 Oe ∆ H F W H M ( O e ) f (GHz) Sample: α ; ∆ H f (GHz)1-S FIG. 5: Frequency dependence of FMR absorptionlinewidth ∆ H FWHM for YIG LPE films A–D at variousvalues of the in-plane magnetic field ( H ≤ α are obtained from. Inset shows an example ofFMR absorption spectrum measured for the sample Aat H = 1 . α = 1.2 × -4 α = 0.7 × -4 α = 0.4 × -4 d = 106 nm d = 410 nm d = 3.0 µ m d = 300 µ m ∆ H F W H M ( O e ) f (GHz) α = 0.5 × -4 FIG. 6: Frequency dependencies of the FMR linewidth∆ H FWHM for YIG LPE films of various thickness d andthe YIG sphere with diameter d = 300 µm . The Gilbertdamping factors α are calculated from slopes of the bestlinear fits according to Eq. (2).presented in Fig. 5 vs the FMR frequency. The plots inFig. 5 are known to provide data about the Gilbertdamping coefficient α and the inhomogeneous contribu-tion ∆ H to the FMR linewidth that are mutually relatedby ∆ H FWHM = ∆ H + 2 αfγ (2)As the FMR performance of thin YIG films stronglydepends on the working frequency of future magnonicapplications, we have included various quality parame-ters in Table II, viz. i) the Gilbert damping coefficient α which is mostly responsible for the FMR losses athigh magnetic fields ( H (cid:29) πM eff ), ii) the inhomoge-neous contribution ∆ H that dominates at small fields( H (cid:28) πM eff ) as well as iii) the FMR linewidth at thereference frequency f =6.5 GHz which approximately cor-responds to the case H ≈ πM eff . The latter is estimateddown to ∆ H FWHM =1.4 Oe that is to our knowledge thenarrowest value reported so far for YIG films with a thick-ness of about 100 nm and smaller. The Gilbert dampingcoefficients are estimated to be close to α ≈ × − which is comparable to the best values reported so far(compare with Table I). The zero frequency term ∆ H isfound almost the same for all YIG films including the Lasubstituted one. The obtained value ∆ H ≈ . − . H FWHM = 0 . α = 0 . × − reported by R¨oschmann and Tolksdorf for bulk discsmade of single YIG crystals. III. OUTLOOK AND CONCLUSIONS
Besides the efforts to avoid growth defects as well asinterface roughness and to reduce impurity incorpora-tion during the LPE deposition process further high-resolution investigations are necessary to gain more in-sight into the YIG microstructure and to identify theproperties which play an essential role for its FMR per-formance. Therefore, in future studies we will carry outHR-RSM scans with asymmetrical reflections to deter-mine in-plane and axial strain, respectively, the Time-of-Flight (ToF) SIMS analysis technique using elementstandards to precisely quantify the La substitution con-centration as well as to detect impurity elements from thehigh-temperature solutions in our sub-micrometer LPEfilms. Furthermore, angular dependent measurements ofthe resonance field and of the FMR linewidth will be in-tended to determine the influence of uniaxial magneticanisotropies on the ferromagnetic resonance losses.In conclusion, liquid phase epitaxy has the potential toprovide sub-micrometer YIG films with outstanding crys-talline and magnetic properties to meet the requirementsfor future magnon spintronics with ultra-low effectivelosses if a drastic miniaturization down to the nanometerscale is possible. First sub-100 nm lateral sized structureshave presently been prepared which could be the nextstep to LPE-based microscaled spintronic circuits. Thedevelopment of YIG LPE films with thicknesses below100 nm is now in progress and remains a big challengefor the classical thick-film LPE technique. IV. METHODSA. Sample fabrication
YIG films were grown from PbO-B O based high-temperature solutions resistively-heated in a platinumcrucible at about 900 ◦ C using standard dipping LPEtechnique. During different growth runs nominally pureYIG films were grown on one-inch (111) gadolinium gal-lium garnet (GGG) substrates to check the reproducibil-ity of the sub-micrometer liquid phase epitaxial growth.For La substituted films La O was added to the al-ready used high-temperature solution. To remove solu-tion remnants from the sample surfaces the holder hadto be stored in a hot acidic solution after room tem-perature cooling. Afterwards the reverse side layer wasremoved by mechanical polishing from the double-sidegrown samples. Chips of different sizes were prepared bya diamond wire saw and sample surfaces were cleanedusing ethanol, distilled water and acetone. The LPE filmthickness was determined by X-ray reflectometry using aPANanalytical/X-Pert Pro system. B. Microstructural properties
The root-mean-square surface roughness was deter-mined by AFM measurements for each sample at threedifferent regions over 25 µm ranges using a Park Scien-tific Instruments, M5. HR-XRD studies were performedby a five-crystal diffraction spectrometer of Seifert (3003PTS HR) equipped with a four-fold Ge 440 asymmetricmonochromator using CuK α radiation. The resolutionlimit was 1 × − deg. GGG substrate lattice param-eters were obtained by the Bond method. Depth pro-file analyses were carried out by an Axis Ultra DLD
XPSsystem (Kratos Analytical Ltd.) using a mono-atomicargon-ion etching technique. Qualitative SIMS (HidenAnalytical) measurements were carried out. Here, a filmarea of 500 × µm is irradiated by 5 keV oxygen ions. C. Magnetic properties
The vibrating sample magnetometer (MicroSenseLLC, EZ-9) was used to register the in-plane hystere-sis loops of the YIG/GGG samples at room tempera-ture. The external magnetic field H was controlled withan error of ≤ ◦ ≤ ϕ ≤ ◦ were measured with an angular step of3 ◦ . The FMR absorption spectra were registered with avector network analyzer (Rohde & Schwarz GmbH, ZVA67) attached to a broadband stripline. The sample wasdisposed face-down over a stripline and the transmissionsignals ( S & S ) were recorded. During the measure-ments, a frequency of microwave signals with the inputpower of −
10 dBm (0.1 mW) was swept across the res-onance frequency, while the in-plane magnetic field H was constant and measured with an accuracy of 1 Oe.Each recorded spectrum was fitted by the Lorentz func-tion and allowed us to define the resonance frequencyand the FMR linewidth ∆ H FWHM corresponding to theapplied field H . V. V. Kruglyak and R. J. Hicken, “Magnonics: Experiment toprove the concept,” J. Magn. Magn. Mater. , 191–194 (2006),cond-mat/0511290. S. Neusser, B. Botters, and D. Grundler, “Localization, confine-ment, and field-controlled propagation of spin waves in Ni Fe antidot lattices,” Phys. Rev. B , 054406 (2008). V. V. Kruglyak, S. O. Demokritov, and D. Grundler, “Magnon-ics,” J. Phys. D: Appl. Phys. , 264001 (2010). R. L. Stamps, S. Breitkreutz, J. ˚Akerman, A. V. Chumak,Y. Otani, G. E. W. Bauer, J.-U. Thiele, M. Bowen, S. A.Majetich, M. Kl¨aui, I. Lucian Prejbeanu, B. Dieny, N. M.Dempsey, and B. Hillebrands, “The 2014 Magnetism Roadmap,”J. Phys. D: Appl. Phys. , 333001 (2014), arXiv:1410.6404[cond-mat.mtrl-sci]. R. C. LeCraw, E. G. Spencer, and C. S. Porter, “FerromagneticResonance Line Width in Yttrium Iron Garnet Single Crystals,”Phys. Rev. , 1311–1313 (1958). A. A. Serga, A. V. Chumak, and B. Hillebrands, “YIG magnon-ics,” J. Phys. D: Appl. Phys. , 264002 (2010). A. V. Chumak, A. A. Serga, and B. Hillebrands, “Magnon tran-sistor for all-magnon data processing,” Nat. Commun. , 4700(2014). A. V. Chumak, V. I. Vasyuchka, A. A. Serga, and B. Hillebrands,“Magnon spintronics,” Nat. Phys. , 453–461 (2015). E. A. Giess, J. D. Kuptsis, and E. A. D. White, “Liquid phaseepitaxial growth of magnetic garnet films by isothermal dippingin a horizontal plane with axial rotation,” J. Cryst. Growth ,36–42 (1972). P. Pirro, T. Br¨acher, A. V. Chumak, B. L¨agel, C. Dubs,O. Surzhenko, P. G¨ornert, B. Leven, and B. Hillebrands, “Spin-wave excitation and propagation in microstructured waveguidesof yttrium iron garnet/Pt bilayers,” Appl. Phys. Lett. ,012402 (2014), arXiv:1311.6305 [cond-mat.mes-hall]. M. C. Onbasli, A. Kehlberger, D. H. Kim, G. Jakob, M. Kl¨aui,A. V. Chumak, B. Hillebrands, and C. A. Ross, “Pulsed laserdeposition of epitaxial yttrium iron garnet films with low Gilbertdamping and bulk-like magnetization,” APL Mater. , 106102(2014). B. M. Howe, S. Emori, H.-M. Jeon, T. Oxhol, J. G. Jones, K. Ma-halingam, Y. Zhuang, N. X. Sun, and G. J. Brown, “Pseudomor-phic yttrium iron garnet thin films with low damping and inho-mogeneous linewidth broadening,” IEEE Magn. Lett. , 3500504(2015). H. Chang, P. Li, W. Zhang, T. Liu, A. Hoffmann, L. Deng,and M. Wu, “Nanometer-thick yttrium iron garnet films withextremely low damping,” IEEE Magn. Lett. , 6700104 (2014). H. Wang,
Understanding of Pure Spin Transport in a BroadRange of Y F e O -based Heterostructures , Ph.D. thesis, TheOhio State University (2015). O. d’Allivy Kelly, A. Anane, R. Bernard, J. Ben Youssef,C. Hahn, A. H. Molpeceres, C. Carr´et´ero, E. Jacquet, C. Der-anlot, P. Bortolotti, R. Lebourgeois, J.-C. Mage, G. de Loubens,O. Klein, V. Cros, and A. Fert, “Inverse spin Hall effect innanometer-thick yttrium iron garnet/Pt system,” Appl. Phys.Lett. , 082408 (2013), arXiv:1308.0192 [cond-mat.mtrl-sci]. V. Castel, N. Vlietstra, B. J. van Wees, and J. B. Youssef, “Fre-quency and power dependence of spin-current emission by spinpumping in a thin-film YIG/Pt system,” Phys. Rev. B , 134419(2012), arXiv:1206.6671 [cond-mat.mtrl-sci]. C. Hahn, G. de Loubens, O. Klein, M. Viret, V. V. Naletov, andJ. Ben Youssef, “Comparative measurements of inverse spin Halleffects and magnetoresistance in YIG/Pt and YIG/Ta,” Phys.Rev. B , 174417 (2013), arXiv:1302.4416 [cond-mat.mes-hall]. L. J. Cornelissen, J. Liu, R. A. Duine, J. B. Youssef, and B. J.van Wees, “Long-distance transport of magnon spin informationin a magnetic insulator at room temperature,” Nat. Phys. ,1022–1026 (2015), arXiv:1505.06325 [cond-mat.mes-hall]. S. A. Manuilov, R. Fors, S. I. Khartsev, and A. M. Grishin, “Sub-micron Y Fe O Film Magnetostatic Wave Band Pass Filters,”J. Appl. Phys. , 033917–033917 (2009). Y. Sun, Y.-Y. Song, H. Chang, M. Kabatek, M. Jantz, W. Schnei-der, M. Wu, H. Schultheiss, and A. Hoffmann, “Growth andferromagnetic resonance properties of nanometer-thick yttriumiron garnet films,” Appl. Phys. Lett. , 152405 (2012). S. A. Manuilov, S. I. Khartsev, and A. M. Grishin, “Pulsed laserdeposited Y Fe O films: Nature of magnetic anisotropy I,” J.Appl. Phys. , 123917–123917 (2009). R. Hergt, H. Pfeiffer, P. G¨ornert, M. Wendt, B. Keszei, andJ. Vandlik, “Kinetic Segregation of Lead Impurities in GarnetLPE Films,” Phys. Stat. Sol. (a) , 769–776 (1987). x Pb ≈ .
02 formular units in stoi-chiometric Y − x Pb x Fe O . y La ≈ . − y La y Fe O . G. Winkler, “Magnetic garnets,” in
Vieweg tracts in pure andapplied physics; Volume 5 (Friedrich Vieweg & Sohn Verlag,Braunschweig, Wiesbaden, 1981) Chap. 2, pp. 75–79. S. B. Ubizskii, “Orientational states of magnetization in epitaxial(111)-oriented iron garnet films,” J. Magn. Magn. Mater. ,575–582 (1999). S. B. Ubizskii, “Magnetization reversal modelling for (111)-oriented epitaxial films of iron garnets with mixed anisotropy,”J. Magn. Magn. Mater. , 127–141 (2000). C. Kittel, “On the Theory of Ferromagnetic Resonance Absorp-tion,” Phys. Rev. , 155–161 (1948). V. Lauer, D. A. Bozhko, T. Br¨acher, P. Pirro, V. I. Vasyuchka,A. A. Serga, M. B. Jungfleisch, M. Agrawal, Y. V. Kobljanskyj,G. A. Melkov, C. Dubs, B. Hillebrands, and A. V. Chumak,“Spin-transfer torque based damping control of parametricallyexcited spin waves in a magnetic insulator,” Appl. Phys. Lett. , 012402 (2016), arXiv:1508.07517 [cond-mat.mes-hall]. P. R¨oschmann and W. Tolksdorf, “Epitaxial growth and anneal-ing control of FMR properties of thick homogeneous Ga substi-tuted yttrium iron garnet films,” Mat. Res. Bull. , 449–459(1983). T. L¨ober, A. V. Chumak, and B. Hillebrands, Unpublished re-sults.
V. ACKNOWLEDGEMENTS
We acknowledge the partial financial support byDeutsche Forschungsgemeinschaft (DU 1427/2-1). Wethank M. Frigge for EPMA analysis, Ch. Schmidt forXRR measurements and R. Meyer and B. Wenzel fortechnical support.
VI. AUTHOR CONTRIBUTIONS STATEMENT
C.D. conceived the experiments, prepared all samplesand analyzed the data. O.S. performed VSM and FMRmeasurements and analyzed the data. R.L. performedthe XPS experiments. J.D. and U.B. performed the SIMSexperiments. A.D. conducted the XRD experiments andanalyzed the data. C.D. and O.S. wrote the manuscript.All authors contributed to scientific discussions and themanuscript review.