doi.org/10.1002/cnma.202200495 Research Article www.chemnanomat.org Antisite Defects and Chemical Expansion in Low-damping, High- magnetization Yttrium Iron Garnet Films Jose Santiso,[a] Carlos García,[b] Cristian Romanque,[b] Loïc Henry,[c] Nicolas Bernier,[c] Núria Bagués,[d] José Manuel Caicedo,[a] Manuel Valvidares,[e] and Felip Sandiumenge*[f] Abstract: Yttrium iron garnet is widely investigated for its � 10 6 Oe, indicating a very low occurrence of scattering suitability in applications ranging from magneto-optical and centers. Notably, the saturation magnetization, 4πMs = 2.10 microwave devices to magnonics. However, in the few- (� 0.01) kOe, is higher than the bulk value, in consistency nanometer thickness range, epitaxial films exhibit a strong with the suppression of magnetic moment in the minority variability in magnetic behavior that hinders their implemen- octahedral sublattice by the observed antisite defects. tation in technological devices. Here, direct visualization and Analysis of elemental concentration profiles across the spectroscopy of the atomic structure of a nominally stoichio- substrate-film interface suggests that the Y-excess is origi- metric thin film, exhibiting a small damping factor of nated from unbalanced cationic interdiffusion during the 3.0 · 10 4, reveals the occurrence of Y-excess octahedral early growth stages. antisite defects. The two-magnon strength is very small, Γ0 Introduction the latter applications, downscaling is a prerequisite which so far complicates control on magnetic properties due to their Magnetic garnets combine infrared optical transparency, a strong sensitivity to film strain and composition, and demands prominent Faraday effect, insulating behavior, very small spin- a thorough understanding of the growth mechanism. Tradition- wave damping factors, and low dielectric losses,[1] which make ally, YIG films have been grown by liquid phase epitaxy, offering them highly attractive in magneto-optical and microwave excellent epitaxial quality and bulk-like properties.[7–9] Although applications.[2,3] Among these materials yttrium iron garnet films grown by this technique typically have thicknesses (YIG), a ferrimagnetic insulator exhibiting an extremely low comprised in the micrometer to millimeter range, low-damping intrinsic Gilbert damping parameter of α � 3 · 10 5, appears as films below 40 nm have been reported.[10] Nonetheless, advan- an ideal material for the emerging field of magnonics.[4–6] For ces in oxide thin film deposition have allowed the growth of high quality thin films, well below 40 nm, by a number of [a] Dr. J. Santiso, Dr. J. M. Caicedo techniques including sputtering,[11,12] pulsed laser deposition Catalan Institute of Nanoscience and Nanotechnology (ICN2) (PLD),[13–16] and chemical solution deposition.[17] Although damp- CSIC and the Barcelona Institute of Science and Technology (BIST) ing values as small as 7 · 10 5 have been obtained by PLD,[15] Campus de la UAB, 08193 Bellaterra, Catalonia (Spain) compared with liquid phase epitaxial thick films, thin films still [b] Dr. C. García, Dr. C. Romanque Departamento de Física y Centro Científico Tecnológico de Valparaíso – exhibit a strong variability in the magnetic properties. For CCTVal Universidad Técnica Federico Santa María instance, unusually high magnetic anisotropy has been re- Av. España 1680, 2390123, Valparaíso (Chile) ported in PLD films,[13,14] or both magnetization suppression[11] [c] Dr. L. Henry, Dr. N. Bernier Université Grenoble Alpes and exceptionally high magnetization[12] have been reported in CEA, LETI sputtered samples. F-38000 Grenoble (France) The commonly used substrate for YIG is Gd3Ga5O12 (GGG) in [d] Dr. N. Bagués (111) orientation, which provides a very small tensile misfit Center for Electron Microscopy and Analysis The Ohio State University strain of 0.06%, allowing films to grow coherently in the 43212 Columbus, Ohio (USA) absence of misfit dislocations. Thus, in the absence of a [e] Dr. M. Valvidares significant misfit, the scatter of physical behavior observed in ALBA Synchrotron Light Source 08290 Cerdanyola del Vallès, Catalonia (Spain) thin films is to be sought on interdiffusion and defect chemistry [f] Dr. F. Sandiumenge rather than strain effects. In this sense, it has been argued that Institut de Ciència de Materials de Barcelona (ICMAB-CSIC) Fe3 + vacancies cause the unusual magnetic anisotropy,[13,14] or Campus de la UAB, 08193 Bellaterra, Catalonia (Spain) that magnetization suppression in sputtered films results from E-mail: felip@icmab.cat an enrichment in Gd3 + ions,[11] while other studies point to a Supporting information for this article is available on the WWW under https://doi.org/10.1002/cnma.202200495 preferential Ga3 + diffusion into the film causing a dead layer.[16] © 2022 The Authors. ChemNanoMat published by Wiley-VCH GmbH. This is Besides interdiffusion among equivalent sublattices, YIG films an open access article under the terms of the Creative Commons Attribution often exhibit an anomalous lattice stretching along the growth Non-Commercial NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non- direction[18–20] and rhombohedral distortions,[19,20] which very commercial and no modifications or adaptations are made. likely reflect chemical expansion effects caused by point ChemNanoMat 2023, 9, e202200495 (1 of 9) © 2022 The Authors. ChemNanoMat published by Wiley-VCH GmbH 2199692x, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cnma.202200495 by Readcube (Labtiva Inc.), Wiley Online Library on [26/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Research Article defects, particularly antisite. Altogether, these observations suggest a complex interplay between interdiffusion, defect chemistry and growth conditions, that remains poorly under- stood. The general formula of YIG is Y(c)3Fe(a)2Fe(d)3O12 (Ia3d), where letters (c), (a), (d) are Wyckoff sites here denoting dodecahedral, octahedral and tetrahedral coordination, respectively.[21] Atom- istic calculations show a strong tendency for point defect formation, with oxygen and iron vacancies, and antisite defects (excess Y sitting at octahedral sites, Y(a)), exhibiting the lowest formation energies.[22,23] Antisite defects can also be of disorder type, formed by the exchange between (c) and (a) -site cations according to the disorder reaction Y(c) + Fe(a)!Y(a) + Fe(c), keeping the nominal stoichiometry unaltered. The energy of this reaction is 0.8 eV per defect,[24] which is comparatively low enough to favor the occurrence of Y(a) antisites in stoichiometric crystals as native defects, in agreement with nuclear magnetic resonance spectroscopy studies.[25–27] Given that YIG ferrimag- netism arises from the unbalanced antiferromagnetic coupling between 16 octahedral and 24 tetrahedral sites hosting the Fe3 + ions, magnetic dilution by Y antisites or Fe vacancies in the different sublattices may either increase or suppress net magnetic moment.[23,28,29] Note, however, that since (a) and (c) sites have parallel spin coupling,[30] disorder-type antisites alone leave the net moment unaltered. In this work we use aberration-corrected scanning trans- mission electron microscopy (STEM) imaging and spectroscopy, to elucidate the formation mechanism and spatial distribution of point defects. We show that local off-stoichiometry resulting from unbalanced cationic interdiffusion profiles across the substrate-film interface leads to the accumulation of excess- type antisite defects within the interdiffusion zone and a background defect concentration throughout the film volume. Despite this chemical complexity, ferromagnetic resonance (FMR) measurements yield a small Gilbert damping of 3.0 · 10 4, comparable with best values obtained by PLD,[15] and an almost negligible two-magnon scattering (TMS) strength parameter of Figure 1. a) High resolution 2θ/ω scans about the 444 reflection of GGG and Γ0 = 10 6 Oe, indicating magnetic homogeneity. The saturation YIG. Fringes indicate a high crystal coherence and a film thickness of 25 nm. magnetization determined from FMR, 4πMs = 2.10 (� 0.01) kOe, b) Reciprocal space map around the 880 reflection. Vertical alignment of the GGG and YIG reflections indicates perfect in-plane matching. c) Fe-L2,3 XAS is higher than the bulk value: 1.73–1.75 kOe,[31] which is and XMCD spectra acquired under an applied field of 2T. The measured attributed to the suppression of magnetic moment in the energy difference E(L2)–E(L3) = 13.4 eV indicates trivalent iron (see Supple- minority octahedral sublattice by non-magnetic Y(a) antisites. mental Information S1), and main spectral features: Oh, T1d and T2d, are consistent with the expected YIG magnetic structure. Results and Discussion cubic interaxial angle of 89.38°, in consistency with a rhombo- Figure 1a shows a high resolution XRD scan around the 444 hedral distortion resulting from the stretching along the cubic reflection of both film and substrate. From the film peak body diagonal. By analyzing the Fe-L2,3 edge XAS signal position an interplanar spacing along the film normal of d444 = (Figure 1c) we find that Fe ions are in a trivalent state, at least 1.8075 Å is obtained, which indicates an anomalous lattice within the first 2–5 nm beneath the film surface sampled by the stretching of 1.10% relative to the bulk value, d444 = 1.78776 Å X-rays (see Supplemental Information S1 and references there- (a = 12.376 Å).[21] The XRD fringes indicate a high crystal in). Figure 1c also shows the corresponding XMCD spectrum coherence of the film and the period of the oscillations obtained under a magnetic field of 2 T. The main features are a corresponds to a film thickness of 25 nm, in good agreement prominent positive maximum at 709.1 eV bounded by two with STEM observations. XRD reciprocal space mapping around negative minima at 707.9 eV and 709.7 eV, which can be the asymmetric 880 YIG reflection indicates perfect in-plane assigned to the octahedral (Oh) and tetrahedral (Td) sublattices, matching with the substrate (see Figure 1b), and a reduced ChemNanoMat 2023, 9, e202200495 (2 of 9) © 2022 The Authors. ChemNanoMat published by Wiley-VCH GmbH 2199692x, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cnma.202200495 by Readcube (Labtiva Inc.), Wiley Online Library on [26/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Research Article confirming the expected YIG magnetic structure and the from geometrical phase analysis (GPA,[34]). A striking feature is trivalent state of Fe ions. the layer of high (yellow) contrast adjacent to the interface, Similar distortions, with stretching values comprised be- which manifests as a prominent peak in the strain profile tween 1.0 and 1.5%, and rhombohedral angles of about 89.5°, depicted in Figure 2c (bottom panel). The corresponding have been observed by several authors.[18–20] In fact, from HAADF contrast profile is depicted in Figure 2c (upper panel). reported diffraction data one observes that this phenomenon is We note that due to the large difference between the Gd (Z = more common than manifested in the literature. For instance, 64) and Y (39), Ga (31) and Fe (26) atomic numbers, the normal strains ranging from 0.42% to 0.11% are observed in apparent width of the contrast transition region can be films with thicknesses comprised between 16 nm and considered to mostly reflect the interdiffusion length of Gd3 + 164 nm,[12] and a value of 0.96% is inferred from data ions. Comparing the contrast and strain profiles one observes corresponding to a 250 nm thick film in ref. [11] Interestingly, that the thickness of the stretched layer is similar to that of the these distortions are markedly enhanced up to 2.7% (stretching) substrate to film transition region, both indicated by yellow and 88.8–89.1° (rhombohedricity) in Y-excess films exhibiting a shadowing. Remarkably, beyond this feature, instead of recover- high concentration of Y(a) antisites.[23] While in those off- ing the bulk lattice dimensions, the film still retains an stoichiometric films such large distortions can be explained by oscillating positive background strain. The average strain value the ionic radii difference between Fe3 + and Y3 + ions in within the homogeneous region is < ɛ⊥ > = 0.007 with a octahedral coordination, 0.645 Å and 0.900 Å, respectively,[32] in standard deviation σ = 0.003. Including the interfacial region nominally stoichiometric films the origin of the observed yields < ɛ⊥ > = 0.010, similar to the averaged value derived distortion remains unclear. from XRD results, 0.011. Figure 2a shows a cross-section high resolution annular We used EELS at the Fe-L2,3 edge and EDS mapping to dark-field (HAADF) image viewed along the [0 1 1] zone axis. determine the spatial homogeneity of the oxidation state of Fe Since this is an incoherent image with contrast proportional to and the elemental distribution over the windows marked in Z1.7 (Z is the atomic number),[33] the GGG substrate appears Figure 3a. The EEL spectra, depicted in Figure 3b, were analyzed brighter than the YIG film. In agreement with the XRD data, the using three components: L3, L2, and a third one, labelled *, to image shows perfect lattice continuity across the interface. It account for the low intensity shoulder on the high energy side can be observed, however, that the transition from the of the L3 peak, as illustrated in the inset, and results are substrate to the film is not abrupt, but exhibits an apparent depicted in Figure 3c. The L3 peak is located at 709.2 eV in good width of about 4 nm (area framed with dashed line), signifi- agreement with XAS and, with the exception of the outermost cantly larger than the rms roughness of the substrate surface, spectrum (distance = 25 nm), neither its position nor the energy suggesting a chemically diffuse interface. difference relative to the L2 signal do change through the Figure 2b shows a mapping of the normal strain (ɛ⊥) over thickness of the film, thus indicating a homogeneous trivalent the corresponding image shown in the upper panel, derived state of Fe ions. Under the assumption of cation stoichiometry, Figure 2. a) Cross-section HAADF image of the GGG/YIG interface viewed along the [01-1] direction. The framed region shows a contrast transition region consistent with a chemically diffuse interface. b) GPA mapping of the normal strain component (ɛ⊥) over the image shown in the upper panel, comprising the whole film thickness. The film surface is indicated by a dashed line. c) Upper panel: HAADF intensity profile integrated over the image shown in b). Lower panel: Normal strain profile integrated over the GPA map shown in b). Yellow shadowings emphasize the correspondence between the widths of the contrast transition region marked in a) and the highly strained interfacial layer. The arrowed peak corresponds to a strain maximum out of the Gd diffusion length (see text). ChemNanoMat 2023, 9, e202200495 (3 of 9) © 2022 The Authors. ChemNanoMat published by Wiley-VCH GmbH 2199692x, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cnma.202200495 by Readcube (Labtiva Inc.), Wiley Online Library on [26/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Research Article Figure 3. a) Cross section HAADF image illustrating the areas used for EELS and EDS analyses. b) Through thickness sequence of EEL spectra obtained from the corresponding area marked in a). The inset illustrates the fitting procedure employed to analyze the spectra. c) Position dependence along the film normal of the energies corresponding the L3, L2 and * spectral features. A dashed line marks the position of the interface. d) EDS elemental maps acquired from the corresponding area marked in a). e) Elemental concentration profiles across the substrate-film interface as determined from integration over the corresponding maps shown in d). f) Normalized elemental concentration profiles. this result is consistent with the absence of oxygen vacancies. particularly relevant to the defect chemistry of the interfacial However, we cannot rule out their presence as they could also region. The first one concerns the marked asymmetry between form to compensate for cation vacancies. the diffusion behavior of Fe/Y and Ga/Gd ions. While from the As shown in Figure 3d, EDS results indicate a homogeneous substrate side Gd and Ga draw rather similar profiles, from the cation distribution within the film and signal blurring in the film side the Y profile is much steeper than the Fe one and vicinity of the interface, consistent with the contrast transition clearly reveals an excess peak ~ 20 at% above the stoichiometric zone observed in HAADF images. Corresponding concentration level. Moreover, the Y profile shows fluctuations above the profiles, shown in Figure 3e, indicate that larger ions, Gd3 + and stoichiometric level throughout the bulk film. The differential Y3 +, exhibit smaller diffusion lengths, � 4 nm, than Ga3 + and behavior between Y and Fe ions in fact suggests that the Fe3 +, � 8.5 nm. Values of interdiffusion lengths reported in the formers are replacing iron at octahedral or tetrahedral sites literature for this system are controversial. Liu et al. report forming antisite defects. Recalling that the HAADF intensity HAADF images with a sharp contrast change between the GGG profile shown in Figure 2 is dominated by the Gd atomic substrate and the YIG film, and EDS elemental profiles number, it can be inferred that the position of the Y-excess exhibiting chemical intermixing over 2.3 nm which they attrib- layer coincides with the position of the (arrowed) shoulder ute to delocalization of the X-ray emission signal.[35] Gallagher adjacent to the main peak in the strain profile, which lies out of et al. report HAADF images with a smoother contrast transition the contrast transition region. Hence, these observations allow and chemical intermixing over 4 nm, as determined by EDS, to associate the main strain peak with the Gd interdiffusion which also attribute to delocalization effects.[12] On the other region and its shoulder with the position of an adjacent Y- hand, Mitra et al. observe a smooth transition region in HAADF excess layer. Thus, these strain peaks would be consistent with images and a 6 nm wide intermixing region, as determined by the chemical expansion expected for the occurrence of Gd and EELS, which they consider representative of true interdiffusion Y antisite defects. effects.[11] In the present case, the differential behavior exhibited The location of the Y-enriched layer at the boundary of the by the different elements, together with the HAADF contrast Gd diffusion length strongly suggests a dominant role of Gd transition between the substrate and the film, strongly support ions diffusing from the substrate during the early stages of the interpretation of our EDS results in terms of true chemical nucleation and growth. The partial occupancy of dodecahedral intermixing. (c) sites by Gd may cause an accumulation and displacement of In order to compare in more detail the behavior of the Y ions on the surface of the growing film that would eventually different ions, Figure 3f shows the diffusion profiles normalized cease when its thickness reaches the Gd diffusion length, to the nominal cation stoichiometry, i. e. the rate of occupation leading to the observed Y peak immediately ahead of the reach of dodecahedral sites by Gd and Y, and of the sum of of Gd ions. octahedral and tetrahedral sites by Ga and Fe. Two features are ChemNanoMat 2023, 9, e202200495 (4 of 9) © 2022 The Authors. ChemNanoMat published by Wiley-VCH GmbH 2199692x, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cnma.202200495 by Readcube (Labtiva Inc.), Wiley Online Library on [26/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Research Article The second feature referred to above is that all concen- average atomic number of column 2, < Z > = 32.5, is larger tration curves approximately intersect at a value of 10 at% (see than the Fe atomic number, Z = 26, the intensity of the central Figure 3e), in agreement with the nominal total-cation versus peak in each triplet should be lower than that of the oxygen ratio. This indicates that at the crossing point there is neighboring ones. Careful inspection of the experimental 5 at% excess of Y and Gd atoms relative to the number of image, however, reveals that in many triplets within the YIG available dodecahedral (c) sites. This interdiffusion unbalance film the central column is more intense than the side ones across the interface should cause (c)-site cation excesses which, (examples are indicated by circles). In order to inquiry into the according to atomistic calculations on the energetics of cation origin of this contrast inversion, Figure 4b shows the calculated disorder in garnets are expected be accommodated through concentration dependence of the 3/2 intensity ratio (see also the formation of antisite defects.[36] inset in Figure 4c). It can be readily seen that the introduction Figure 4a provides a closer look to the YIG/GGG interface. In of Y(a) defects in the octahedral sublattice causes an apparent this projection cationic columns draw a honeycomb-like pattern relative increase of the intensity of the central column (3). We elongated along the [011] direction. As shown in the simulated note, however, that there are factors not taken into account in image (right panel), the long sides of the hexagonal motif are the simulation that might alter the image contrast and there- formed by three closely located, 1.1 Å apart, cationic columns, fore an estimation of the antisite defect concentration: (i) the labelled 2 or 3 depending on their composition and structure: occurrence of Gd(a) antisite defects within the interdiffusion Along column 2, dodecahedral Y(c) sites alternate with tetrahe- zone, which may lead to an overestimation of the actual dral Fe(d) sites, while column 3 is composed by octahedral Fe(a) concentration of Y(a) ones, (ii) local lattice distortions associated chains. The other two columns, alternating along rows with the with the occurrence of the defects and, (iii) inhomogeneous 2-3-2 triplets, are formed by chains of dodecahedral Y(c) (bright) distribution of the defects along the atomic columns.[49] Hence, and tetrahedral Fe(d) (darker) coordination, respectively. this analysis is intended as a qualitative appraisal of the image The theoretical image contrast clearly reflects the atomic contrast dependence on Y(a) concentration. number difference between different columns. Both 2 and 3 In order to estimate the strain associated to the observed -type columns have the same atomic density and therefore population of Y(a) defects one needs to consider the misfit their HAADF intensities can be directly compared. Since the constraint imposed by the substrate. According to atomistic Figure 4. a) High resolution HAADF image viewed along the [01-1] zone axis illustrating in detail the contrast in the vicinity of the YIG/GGG interface. Red line shows the path of the intensity profile shown in c). Right panel: Image simulation for stoichiometric YIG identifying different types of atomic columns with numbers, and corresponding HAADF intensity profile. The central column (3) has a lower intensity than neighboring 2-type columns according with its smaller average atomic number. Examples deviating from this rule are indicated by circles in the experimental image. b) Simulated 3/2 intensity ratio as a function of defect concentration. c) Experimental HAADF intensity profiles obtained through the 2-3-2 triplets along the [011] direction. The inset shows image simulations corresponding to increasing concentrations of Y(a) antisite defects, and corresponding intensity profiles. d) Variation of the 3/2 intensity ratio with distance from the substrate along the film normal, [111]. The solid line corresponds to a Savitzky-Golay type smoothing of the experimental data. Inset illustrates the fitting procedure. ChemNanoMat 2023, 9, e202200495 (5 of 9) © 2022 The Authors. ChemNanoMat published by Wiley-VCH GmbH 2199692x, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cnma.202200495 by Readcube (Labtiva Inc.), Wiley Online Library on [26/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Research Article calculations, inserting one Y(a) excess defect per unit cell (8 f. u.) experimental data.[37] It can be observed that the transition from causes a 0.72% increase in volume.[23] This indicates that an Y the substrate (shadowed area) to the film is characterized by an excess as small as 0.62 at% increases the YIG lattice parameter abrupt increase of the 3/2 (3/[(2 + 2’)/2]) intensity ratio, from 12.376 Å to 12.406 Å, switching the misfit strain from consistent with the concomitant compositional variation of the tensile to compressive on GGG (a = 12.380 Å). In our case, the atomic columns, and reaches a first maximum within the Gd occurrence of larger Gd(a) defects in the interdiffusion region diffusion length which is thought to reflect the occurrence of should induce even larger dilations. It appears that the Gd(a) excess antisite defects. Beyond this maximum the intensity observed uniaxial stretching does not only reflect the chemical ratio decays drawing an oscillation similar to that observed in expansion associated with the defects, but is further enhanced the normal strain profile (see Figure 2c), suggesting a possible by the Poisson contribution. Using a Poisson ratio along the fluctuation of the chemical expansion along the growth [111] direction of ν[111] = 0.29, one obtains an additional direction. At the second peak the intensity ratio adopts a value elongation of 0.0015 (see Figure S2). Therefore, one Y(a) antisite of 0.96 which, according to image simulations would corre- defect per unit cell would account for a net stretching of 0.40%, spond to an Y excess of 2.9 at%. We note that this value is in which is more than half the 0.7% strain background measured qualitative agreement with the local Y excess observed at this by GPA in the bulk film. This shows that very low defect position in the EDS elemental profile (Figure 3e). The position of concentrations, even close or below the resolution limit of the this second peak coincides with that of the second strain peak, EDS analysis, can induce significant strain levels. arrowed in Figure 2c, in agreement with the accumulation of Y We now analyze the spatial distribution of antisite defects in at the boundary of the Gd diffusion zone. Beyond this Y- more detail. Figure 4d shows an experimental HAADF intensity enriched layer, the 3/2 intensity ratio is kept above the profile through the 2–3-2 triplets (along the [011] direction). As stoichiometric value indicating the persistence of a background can be seen, the relative intensity of the central peaks increases concentration of Y(a) defects in consistency with Y-excess towards the substrate, before the background slope increases fluctuations seen in EDS (Figure 3f), and the residual strain due to the increase in Gd concentration. It can also be observed observed by GPA (see Figure 2). that in the bulk film many triplets exhibit an asymmetric In order to determine the saturation magnetization, Ms, and intensity distribution, which evidences unequal occupancies of the Gilbert damping constant, we conducted frequency- the mixed side columns (denoted by 2 and 2’, see inset in dependent FMR measurements (see Figure S3). As argued by Figure 4d). The origin of this disorder is not straightforward due Gallagher et al. from a comparative study using various to possible competition of other defects, which according to measuring approaches, this technique is particularly suited for their relative formation energies could be tetrahedral yttrium the determination of Ms in thin films, where uncertainties may antisites (Y(d)), non-excess Y(a) resulting from the disorder arise from the estimation of the film thickness, particularly in reaction Y(c) + Fe(a)!Y(a) + Fe(c), and tetrahedral vacancies[22–24], the presence of chemically diffused interfaces, and from the although in agreement with the resulting intensity variations paramagnetic contribution of the GGG substrate.[12] Figure 5a their contribution should be secondary compared with excess shows the excitation frequency, f, as a function of the resonance Y(a) defects. field, H0. From the p fit ffiffiof ffiffiffiffiffiffithe ffiffiffiffiffiffiffiffiexperimental ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi data using the Kittel g To obtain statistical trends, Figure 4d shows results from a formula,[38,39] f ¼ 2p H0 ðH0 þ 4pMeff Þ, one obtains the reduced large number of [011] atomic rows, where the solid curve gyromagnetic ratio γ/2π = 2.850 (� 0.003) GHz/kOe and the corresponds to a Savitzky-Golay type smoothing of the effective magnetization 4πMeff = 2.18 (� 0.01) kOe. The magnet- Figure 5. a) Ferromagnetic resonance frequency as a function of applied magnetic field. The blue circles represent the resonance field for a given frequency and the solid red line the Kittel fit. From the linear fit we extract the value of the gyromagnetic ratio (γ) and the effective magnetization (4πMeff). b) Frequency dependence of FMR peak-to-peak linewidth. The open blue circles show the experimental data and the red solid line the fit according to Equation 2. ChemNanoMat 2023, 9, e202200495 (6 of 9) © 2022 The Authors. ChemNanoMat published by Wiley-VCH GmbH 2199692x, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cnma.202200495 by Readcube (Labtiva Inc.), Wiley Online Library on [26/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Research Article ization at saturation, Ms, is related to Meff by 3.0 � 0.6 · 10 4, which is comparable with the lowest values so 2jK j 4jK j 4pMeff ¼ 4pMs Hani , where Hani ¼ MS1 þ MS2 is the anisotropy far reported in the literature for PLD grown YIG/GGG nano- field. Using the first and second order anisotropy constants of meter-thick films.[15] Any increase in linewidth contributed by YIG, K1 = 6100 erg/cm3 and K2 = 260 erg/cm3,[43] one obtains the TMS mechanism is a measure of the scattering rate of the 4πMs = 2.10 (� 0.01) kOe, about 20% larger than the bulk one, uniform precession magnon into other spin-wave modes.[46] As 1.73 1.75 kOe.[31] Similar high values have been reported for can be observed in Figure 5b, the film under study exhibits a other YIG thin films grown on GGG by magnetron sputtering, highly linear behavior of the linewidth and, accordingly, a very 4πMs = 2.172 2.141 kOe,[12] and PLD, 4πMs = 2.100 kOe.[41,42] low value of the TMS strength, Γ0 � 10 6 Oe, which reveals a Although, unfortunately, those reports do not address the highly ordered film, essentially free of defects or inhomogene- defect chemistry of the films and, therefore, it is not possible to ities that could act as scattering centers. rationalize their magnetic behavior in terms of site-selective dilution effects, according to the present observations this increased value is in qualitative agreement with a selective Conclusion suppression of the magnetic moment in the minority octahedral sublattice induced by the observed non-magnetic Y(a) defects. These results highlight the ease of formation of Y(a) antisite Interestingly, one finds in the literature that films exhibiting defects in nominally stoichiometric YIG films, particularly under different magnetism also differ in the shape of their interdiffu- conditions where unbalanced chemical interdiffusion across sion profiles. For instance, in contrast with our results, Mitra film-substrate interface favors the formation of excess Y ions et al. report magnetization suppression in films exhibiting two relative to the number of dodecahedral (c) sites of the garnet crossing points, one at � 12 at% for Fe/Ga and another at � 5 at structure. The occurrence of these defects is easily recognized % for Gd/Y, indicating a strong deficiency in Y and Gd ions at through their chemical expansion, which under the constraint the interface[11]. In strong contrast, Gallagher et al. report high of the substrate results in an important lattice stretching along magnetization, similar to that observed in the present work, in the [111] growth direction of 2.8% (rhombohedral distortion). films also exhibiting a unique crossing point at � 10 at%, and More specifically, we observe a large concentration of Gd(a) and anomalous lattice stretching along the [111] direction[12], Y(a) antisite defects within a ca. 4 nm thick interdiffusion zone, reminiscent of the chemical expansion by Y(a) defects observed and accumulation of the latter in an adjacent � 2 nm thick in this work. This observation suggests, in fact, a strong impact layer. Beyond this interfacial layer, the YIG still sustains a of interdiffusion behavior on the magnetization of the films. homogeneous stretching of 0.7% caused by a background Unbalanced interdiffusion results in stoichiometric deviations concentration of Y(a) antisite defects populating the entire film and the formation of antisite defects which may alter the volume. Despite lattice distortions and cationic disorder, the ferrimagnetic balance among the different sublattices in differ- film exhibits a high structural coherency, and FMR yields a low ent ways. Thus, most importantly, this behavior underlines a Gilbert damping parameter, α = 3.0 � 0.4 · 10 4, and an almost challenging need for precise control of interdiffusion in this negligible two-magnon scattering strength parameter, Γ0 complex system. � 10 6 Oe. The saturation magnetization is higher than the bulk To precisely determine the Gilbert damping parameter, α, value, in consistency with the suppression of magnetic moment the inhomogeneous linewidth broadening ΔH0 and the from the minority octahedral sublattice by replacement of Fe3 + strength of the two-magnon scattering, Γ0, we fit the linewidth by nonmagnetic Y3 +. In the light of these results, the recurrent versus excitation frequency data using the following equation: observation of unexplained interfacial stretching by other authors strongly suggests that Y(a) defects are a common 8paf feature in this system. These results emphasize a challenging DH ¼ DH0 þ pffiffi þ DHTMS (1) 3g need for control on interdiffusion to overcome the irreproduci- bility issue associated with thin YIG/GGG epitaxial films. where the first term corresponds to the inhomogeneous broadening, and the second term is the intrinsic contribution to the linewidth which is proportional to the frequency with a Experimental Section slope determined by Gilbert damping parameter, α. For a Film growth: Epitaxial Y3Fe5O12 (yttrium iron garnet, YIG) films of homogeneously in-plane magnetized thin film, the third term, about 25 nm thickness were deposited by PLD onto isostructural corresponding to the TMS broadening, can be approximately Gd3Ga5O12 (GGG) single crystal substrates with (111) orientation expressed as,[43–45] (Crystec GmbH, roughness < 1 nm). The YIG target was prepared by conventional solid state reaction from Y2O3 and Fe2O3 powders � 2 � (above 99.9% purity from Sigma Aldrich). The conditions for the 2 1 ½f þ ðf 0 =2Þ2 � f 0 =2 1=2 DHTMS ¼ pffiffi G 0 sin (2) growth were: KrF excimer UV laser (λ = 248 nm), laser energy 3 ½f 2 þ ðf 0 =2Þ2 � þ f 0 =2 fluence about 4 J/cm2, 1 Hz pulse repetition rate, deposition temperature T = 750 °C, oxygen pressure pO2 = 1.5 mTorr, and where f 0 ¼ ðg=2pÞ4pMeff . When the orientation of the applied target-to-substrate distance of about 45 mm. After growth films field (and the magnetization) is in-plane, Γ0 can be considered were cooled down under the same pO2 at a rate of 10 °C/min. constant.[43] Figure 5b shows the FMR linewidth as a function of excitation frequency. The obtained damping parameter is α = ChemNanoMat 2023, 9, e202200495 (7 of 9) © 2022 The Authors. ChemNanoMat published by Wiley-VCH GmbH 2199692x, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cnma.202200495 by Readcube (Labtiva Inc.), Wiley Online Library on [26/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Research Article X-ray diffraction: High-resolution XRD measurements (2θ/ω scans collection semi-angle of ~ 100 mrad, a drift tube voltage of 400 V and reciprocal space maps) were carried out using a Malvern- (in dual-EELS mode for zero-loss centering) and a dispersion of Panalytical X’Pert-PRO MRD system with a 4-angle goniometer, and 0.25 eV/pixel. Hyperspectral mappings comprised of 218 · 30 pixels using Cu Kα1 radiation (λ = 1.5406 Å, using a 2 · Ge(220) monochro- with a pixel size of ~ 0.2 nm were acquired with a dwell time of mator). 0.1 ms and 100 ms for the low-loss and high-loss regions, respectively. Using subpixel scanning, the above-mentioned acquis- Ferromagnetic resonance: Broadband FMR measurements were ition parameters correspond to EDS and EELS dose-per-frame of performed utilizing a NanOsc PhaseFMR spectrometer and 200 μm- 9.75 · 102 and 1.91 · 103 e-/Å 2. Note also that STEM images were wide coplanar waveguide (CPW). The dc magnetic field H, acquired before and after any EDS or EELS acquisition to verify the generated by the electromagnet, is modulated using a time varying overall structural integrity. field produced by a Helmholtz coil, hac(t), parallel to H with a frequency of 490 Hz and a field strength of 1 Oe. The YIG sample The intensity of the atomic columns in the HAADF image was was placed face-down on the CPW and a RF microwave signal of analyzed extracting a number of separate linear scans (more than frequency f is injected from the spectrometer, therefore producing 30) parallel to the [011] crystallographic direction passing through a RF microwave field hrf perpendicular to the modulated field the peak center of mass with a certain line width of several pixels H + hac(t). The measurements were performed at room temperature, (10) to integrate most of the column intensities. Each linear scan at a fixed frequency in the range from 4 GHz to 17 GHz, while was then fitted to independent Gaussian functions for each atomic sweeping the field through the resonance condition up to 5 kOe. column to extract their position and intensity. For simplicity of the To extract the resonance field and peak-to-peak linewidth at a fitting process the Gaussian peaks were assumed to have a specific frequency, each of the resonance spectra is fitted with the common width for all columns of the same type. Intensity ratios first derivative of an asymmetric Lorentzian function. were measured for adjacent columns and plotted against their corresponding average positions along the growth direction. The X-ray absorption spectroscopy and X-ray magnetic circular dichroism: data were averaged by applying a Savitzky-Golay type smoothing XAS and XMCD measurements were carried out at the BOREAS to a single line profile in a window of 200 points, and polynomials beamline of the ALBA Synchrotron Light Source[47] in total electron up to order two. yield (TEY) mode. Fe spectra were recorded across the L2,3 absorption edges. The XMCD spectrum was calculated from the HAADF image simulations were performed using the quantum difference between the XAS spectra for right (σ +) and left (σ ) excitation of phonons model in μSTEM.[48] A simple crystal model almost 100% circularly polarized light. The XAS and XMCD measure- was prepared using the YIG unit cell projected along the[110] ments were performed at room temperature under an external direction. Structural and thermal parameters were obtained from magnetic field of 2 Tesla (T), which was applied along the out-of- ref. [21]. Qualitative simulations were performed considering the plane direction of the films (i. e., along the beam direction). employed HAADF-STEM imaging experimental conditions (alpha angle 20mrad, collection angles: inner 77mrad and outer 230mrad, Transmission electron microscopy: Samples for transmission electron and foil thickness 80 nm). The image contrast may be affected by microscopy (TEM) were prepared by Ga + Focused Ion Beam (FIB) the location and distribution of the antisites along the atomic milling using a FEI Helios 450 machine. Silicon oxide and tungsten columns.[49] Here, we considered a homogenous distribution of the layers were deposited to protect the regions of interest from the Y-antisite defects along the Fe octahedral columns (column 3 in the tails of the ion beam. A 16 kV operation voltage for the initial text), namely, the antisites were introduced by fractionating the thinning was used and then a low beam energy in the range 5– atom (nFe + (1-n)Y = 1) in the unit cell. The concentration of defects 8 kV to reduce FIB-induced damage. TEM experiments were was simulated from 0 up to 80% in 20% steps. performed on a probe-corrected ThermoFisher Scientific Titan Themis operated at a high voltage of 200 kV. High Angle Annular Dark Field (HAADF) images were acquired in Velox software for a multiframe acquisition allowing for an effective electron dose partitioning and a scan error free acquisition. The HAADF acquis- Acknowledgements ition was done with a probe current of ~ 80 pA and a dwell time of 200 ns corresponding to a dose of 1.27 · 102 e-/Å 2 (considering a We thank Audrey Jannaud for FIB preparation of the TEM 1 Å probe size). The incident convergence angle was set to lamellae. We thank Prof. Dave McComb for discussions and ~ 20.7 mrad, while inner and outer HAADF collection angles were access to Ohio Supercomputer Center. Part of this work was equal to ~ 77.8 mrad and ~ 230 mrad, respectively. EDS and EELS carried out through the access to HRTEM and XAS/XMCD spectra were collected in Gatan Microscopy Suite® (GMS) software. EDS spectra were acquired using the Super-X system that comprises facilities granted by the NFFA-Europe Infrastructure (proposals four 30 mm2 windowless silicon drift detectors placed at an ID 678 and ID 571) under Horizon 2020 EU Funding Program. elevation angle of 18° from the horizontal with a symmetrical This work was supported in part by the Spanish Ministry of distribution along the beam axis (45°, 135°, 225° and 315° Economy, Industry and Competitiveness (project PID2019- azimuthal angles) and a 0.64 (� 0.06)° total solid angle. EDS 108573GB-C21), and by the Catalan AGAUR agency (project: mappings were obtained with the resolution of 100 · 50 pixels (pixel 2017-SGR-579). ICN2 is funded by the CERCA programme/ size of ~ 0.32 nm) using a dwell time of 100 ms. The quantification was performed using the k-factors implemented in GMS. As k- Generalitat de Catalunya and by the Severo Ochoa programme factors do not take into account the absorption and the specific (SEV-2017-0706). ICMAB-CSIC is funded by the Severo Ochoa orientation of the lamella with respect to the detectors, k-factors programme (CEX2019-000917-S). CG and JS acknowledge from Fe2O3, Ga2O3, Gd2O3 and Y2O3 were adjusted such as to match funding from the European Union’s Horizon 2020 research and the composition far from the interface with the theoretical one. innovation program under the Marie Sklodowska-Curie grant This method is validated from the derived oxygen content profile which fits very well the theoretical composition (k-factors are agreement No. 101007825 (ULTIMATE-I Project). MV acknowl- relative to the oxygen concentration and therefore the oxygen edges funding by PID2020-116181RB-C32/MCIN/AEI/10.13039/ content cannot be adjusted). EELS spectra were acquired on a post- 501100011033. C. G. and C. R. acknowledge the financial filter CCD US1000 camera mounted onto a GIF Quantum ER, with a support received by ANID FONDECYT/REGULAR 1201102. This ChemNanoMat 2023, 9, e202200495 (8 of 9) © 2022 The Authors. ChemNanoMat published by Wiley-VCH GmbH 2199692x, 2023, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cnma.202200495 by Readcube (Labtiva Inc.), Wiley Online Library on [26/10/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Research Article work was also supported by grant ANID PIA/APOYO AFB180002. [20] N. S. Sokolov, V. V. Fedorov, A. M. Korovin, S. M. Suturin, D. A. Baranov, S. V. Gastev, B. B. Krichevtsov, K. Y. Maksimova, A. I. Grunin, V. E. Bursian, F. S. acknowledges the Spanish Ministry of Industry, Economy L. V. Lutsev, M. Tabuchi, J. Appl. Phys. 2016, 119, 023903. and Competitiveness (project no. RTI2018-098537-B-C21). [21] S. Geller, M. A. Gilleo, J. Phys. Chem. Solids. 1957, 3, 30–36. [22] S. Tan, W. Zhang, L. Yang, J. Chen, Z. Wang, J. Appl. Phys. 2020, 128, 183904. [23] T. Su, S. Ning, E. Cho, C. A. Ross, Phys. Rev. Mater. 2021, 5, 094403. Conflict of Interest [24] H. Donnerberg, C. R. A. Catlow, J. Phys.: Condens. Matter 1993, 5, 2947– 2060. [25] P. Novák, J. Englich, H. Stepankova, J. Kohout, H. Lütgemeier, K. Wagner, The authors declare no conflict of interest. W. Tolksdorf, J. Phys. IV France Colloque C1 1997, 7, 283–286. [26] V. Chlan, H. Štĕpánková, V. Procházka, J. Englicha, J. Kohouta, D. Nižňanský, J. Buršík, J. of Magn. Magn. Mat. 2005, 290–291, 993–996. [27] M. Kučera, K. Nitsch, H. Štěpánková, M. Maryško, P. Reiche, phys. stat. Data Availability Statement sol. 2003, 198, 407–414. [28] W. Noun, E. Popova, F. Bardelli, Y. Dumont, R. Bertacco, A. Tagliaferri, M. The data that support the findings of this study are available Tessier, M. Guyot, B. Berini, N. Keller, Phys. Rev. B 2010, 81, 054411. [29] Y. Dumont, N. Keller, E. Popova, D. S. Schmool, M. Tessier, S. from the corresponding author upon reasonable request. Bhattacharya, B. Stahl, R. M. C. Da Silva, M. Guyot, Phys. Rev. B 2007, 76, 104413. [30] G. Winkler, Magnetic Garnets. Vieweg Tracts in Pure and Applied Physics, Keywords: yttrium iron garnet · thin film · antisite defects · Vol. 5. Springer-Verlag, Braunschweig/Wiesbaden, 1981. Gilbert damping · magnetization [31] B. Lax, K. J. Button, Microwave ferrites and ferrimagnetics. Lincoln Laboratory publications, McGraw-Hill, New York, 1962. [32] R. D. Shannon, Acta Crystallogr. Sect. A 1976, 32, 751–767. [33] S. J. Pennycook, D. E. Jesson, Phys. Rev. Lett. 1990, 64, 938–9041. [1] E. J. J. Mallmann, A. S. B. Sombra, J. C. Goes, P. B. A. Fechine, Solid State [34] M. J. Hÿtch, Microsc. Microanal. Microstruct. 1997, 8, 41–57. Phenom. 2013, 202, 65–96. [35] M. Liu, L. Jin, J. Zhang, Q. Yang, H. Zhang, P. Gao, D. Yu, AIP Adv. 2018, [2] J. Douglas Adam, L. E. Davis, G. F. Dionne, E. F. Schloemann, S. N. Stitzer, 8, 085117. IEEE Trans. Microwave Theory Tech. 2002, 50, 721–737. [36] Ch. Milanese, V. Buscaglia, F. Maglia, U. Anselmi-Tamburini, Chem. Mater. [3] B. J. H. Stadler, T. Mizumoto, IEEE Photon. J. 2014, 6, 0600215. 2004, 16, 1232–1239. [4] A. A. Serga, A. V. Chumak, B. Hillebrands, J. Phys. D 2010, 43, 264002. [37] A. Savitzky, M. J. E. Golay, Anal. Chem. 1964, 36, 1627–1639. [5] H. Yu, O. d’Allivy Kelly, V. Cros, R. Bernard, P. Bortolotti, A. Anane, F. [38] C. Kittel, Phys. Rev. 1948, 73, 155–161. Brandl, R. Huber, I. Stasinopoulos, D. Grundler, Sci. Rep. 2014, 4, 6848. [39] S. S. Kalarickal, P. Krivosik, M. Wu, C. E. Patton, M. L. Schneider, P. Kabos, [6] V. V. Kruglyak, S. O. Demokritov, D. Grundler, J. Phys. D 2010, 43, T. J. Silva, J. P. Nibarger, J. Appl. Phys. 2006, 99, 093909. 264001. [40] Y. Sun, M. Wu, in Recent Advances in Magnetic Insulators – From [7] S. M. Rezende, R. L. Rodríguez-Suárez, M. M. Soares, L. H. Vilela-Leão, D. Spintronics to Microwave Applications, Vol. 64, (Eds: M. Wu, A. Hoffmann) Ley Domínguez, A. Azevedo, Appl. Phys. Lett. 2013, 102, 1–4. Solid State Physics, Academic Press 2013. [8] M. B. Jungfleisch, V. Lauer, R. Neb, V. Chumak, B. Hillebrands, Appl. Phys. [41] O. d’Allivy Kelly, A. Anane, R. Bernard, J. Ben Youssef, C. Hahn, A. H. Lett. 2013, 103, 2011–2015. Molpeceres, C. Carrétéro, E. Jacquet, C. Deranlot, P. Bortolotti, R. [9] Y. Sun, H. Chang, M. Kabatek, Y.-Y. Song, Z. Wang, M. Jantz, W. Lebourgeois, J.-C. Mage, G. de Loubens, O. Klein, V. Cros, A. Fert, Appl. Schneider, M. Wu, E. Montoya, B. Kardasz, B. Heinrich, S. G. E. Phys. Lett. 2013, 103, 082408. Te Velthuis, H. Schultheiss, A. Hoffmann, Phys. Rev. Lett. 2013, 111, [42] C. Hahn, V. V. Naletov, G. de Loubens, O. Klein, O. d’Allivy Kelly, A. 106601. Anane, R. Bernard, E. Jacquet, P. Bortolotti, V. Cros, J. L. Prieto, M. [10] C. Dubs, O. Surzhenko, R. Thomas, J. Osten, T. Schneider, K. Lenz, J. Muñoz, Appl. Phys. Lett. 2014, 104, 152410. Grenzer, R. Hübner, E. Wendler, Phys. Rev. Mater. 2020, 4, 024416. [43] J. Lindner, I. Barsukov, C. Raeder, C. Hassel, O. Posth, R. Meckenstock, P. [11] A. Mitra, O. Cespedes, Q. Ramasse, M. Ali, S. Marmion, M. Ward, R. M. D. Landeros, D. L. Mills, Phys. Rev. B 2009, 80, 224421. Brydson, C. J. Kinane, J. F. K. Cooper, S. Langridge, B. J. Hickey, Sci. Rep. [44] K. Lenz, H. Wende, W. Kuch, K. Baberschke, K. Nagy, A. Jánossy, Phys. 2017, 7, 11774. Rev. B 2006, 73, 144424. [12] J. C. Gallagher, A. S. Yang, J. T. Brangham, B. D. Esser, S. P. White, M. R. [45] R. Arias, D. L. Mills, Phys. Rev. B 1999, 60, 7395–7409. Page, K. Y. Meng, S. Yu, R. Adur, W. Ruane, S. R. Dunsiger, D. W. [46] A. Conca, S. Keller, M. R. Schweizer, E. T. Papaioannou, B. Hillebrands, McComb, F. Yang, P. C. Hammel, Appl. Phys. Lett. 2016, 109, 072401. Phys. Rev. B 2018, 98, 214439. [13] S. A. Manuilov, S. I. Khartsev, A. M. Grishin, J. Appl. Phys. 2009, 106 [47] A. Barla, J. Nicolas, D. Cocco, S. M. Valvidares, J. Herrero-Martin, P. 123917. Gargiani, J. Moldes, C. Ruget, E. Pellegrin, S. Ferrer, J. Synchrotron Radiat. [14] S. A. Manuilov, A. M. Grishin, J. Appl. Phys. 2010, 108, 013902. 2016, 23, 1507–1517. [15] C. Hauser, T. Richter, N. Homonnay, C. Eisenschmidt, M. Qaid, H. Deniz, [48] L. J. Allen, A. J. D’Alfonso, S. D. Findlay, Ultramicroscopy 2015, 151, 11– D. Hesse, M. Sawicki, S. G. Ebbinghaus, G. Schmidt, Sci. Rep. 2016, 6, 22. 20827. [49] B. D. Esser, A. J. Hauser, R. E. A. Williams, L. J. Allen, P. M. Woodward, [16] S. M. Suturin, A. M. Korovin, V. E. Bursian, L. V. Lutsev, V. Bourobina, N. L. F. Y. Yang, D. W. McComb, Phys. Rev. Lett. 2016, 117, 176101. Yakovlev, M. Montecchi, L. Pasquali, V. Ukleev, A. Vorobiev, A. Devishvili, N. S. Sokolov, Phys. Rev. Mater. 2018, 2, 104404. [17] I. Lucas, P. Jiménez-Cavero, J. M. Vila-Fungueiriño, C. Magén, S. Sangiao, J. M. deTeresa, L. Morellón, F. Rivadulla, Phys. Rev. Mater. 2017, 1, 074407. [18] J. M. Gomez-Perez, S. Vélez, L. McKenzie-Sell, M. Amado, J. Herrero- Martín, J. López-López, S. Blanco-Canosa, L. E. Hueso, A. Chuvilin, J. W. A. Robinson, F. Casanova, Phys. Rev. Appl. 2018, 10, 044046. Manuscript received: November 3, 2022 [19] B. B. Krichevtsov, S. V. Gastev, S. M. Suturin, V. V. Fedorov, A. M. Korovin, Revised manuscript received: December 7, 2022 V. E. Bursian, A. G. Banshchikov, M. P. Volkov, M. Tabuchi, N. S. Sokolov, Accepted manuscript online: December 9, 2022 Sci. Technol. Adv. Mater. 2017, 18, 351–363. Version of record online: December 19, 2022 ChemNanoMat 2023, 9, e202200495 (9 of 9) © 2022 The Authors. ChemNanoMat published by Wiley-VCH GmbH