# Fileset

[s41467-026-75232-0 (1).pdf](https://mdr.nims.go.jp/filesets/633f9eb6-303f-4cac-a947-5842ef450112/download)

## Creator

[Sang J. Park](https://orcid.org/0000-0003-1684-4876), [Keisuke Hirata](https://orcid.org/0000-0002-0917-3907), [Hossein Sepehri-Amin](https://orcid.org/0000-0002-7856-7897), [Fuyuki Ando](https://orcid.org/0009-0003-7789-8170), [Takamasa Hirai](https://orcid.org/0000-0002-5577-8018), [Ken-ichi Uchida](https://orcid.org/0000-0001-7680-3051)

## Rights

[Creative Commons BY Attribution 4.0 International](https://creativecommons.org/licenses/by/4.0/)

## Other metadata

[Trans-scale spin Seebeck effect in nanostructured bulk composites based on magnetic insulator](https://mdr.nims.go.jp/datasets/20802368-cfd1-48b5-b343-4a25faea29fd)

## Fulltext

Trans-scale spin Seebeck effect in nanostructured bulk composites based on magnetic insulatorArticle https://doi.org/10.1038/s41467-026-75232-0Trans-scale spin Seebeck effect innanostructured bulk composites based onmagnetic insulatorSang J. Park 1 , KeisukeHirata 2,3, Hossein Sepehri-Amin 1, Fuyuki Ando 1,Takamasa Hirai 1 & Ken-ichi Uchida 1,3The spin Seebeck effect enables thermoelectric conversion through thermallygenerated spin currents inmagneticmaterials, offering a promising transversegeometry for scalable devices. However, conventional spin Seebeck devicesare confined to nanoscale thin-film architectures, with significantly restrictedoutput power due to the intrinsic constraints of spin and magnon diffusionlengths. Here, we demonstrate a trans-scale spin Seebeck effect using nanos-tructured bulk composites composed of Pt-coated yttrium iron garnet pow-ders fabricated via dynamic powder sputtering and low-temperature sintering.The resulting three-dimensional composites exhibit continuous Pt channelsand robust mechanical integrity. Transverse thermoelectric measurementsreveal isotropic spin Seebeck signals at the bulk scale. Power analysis indicatesthat the three-dimensional architecture enables scalable volumetric thermo-electric power generation beyond diffusion-limited thin-film spin Seebeckgeometries. This work establishes a scalable platform for spin Seebeck ther-moelectric conversion, bridging nanoscale spin caloritronics with macro-scopic device integration.The spin Seebeck effect (SSE) enables the generation of electricalvoltage via thermally driven spin currents (or magnons) in magneti-cally ordered materials1–4. In the SSE, a temperature gradient appliedacross a ferro(i)magneticmaterial (FM) excites spin currents,whicharethen pumped into an adjacent normal metal (NM). There, they areconverted into a transverse electric field through the inverse spin Halleffect (ISHE)5,6. Owing to the orthogonal relationship between the heatflux and the induced electric field, the SSE is classified as a transversethermoelectric effect7–10. Since its discovery in 200811, the SSE has beena central topic of interest in both spintronics and thermoelectrics,serving as a platform for investigating fundamental transportphysics1–3,8–20 and as a basis for thermoelectric energy harvestingapplications1–3,7–11,21–27, based on thermally generated spin currents.SSE-based thermoelectric energy harvesting has been proposedas a solution to the limitations of conventional thermoelectric devices,which rely on the longitudinal Seebeck effect to generate a chargecurrent along the direction of a temperature gradient28–31. Theselongitudinal devices typically adopt π-shaped geometries with multi-ple n-type and p-type thermoelectric legs interconnected by electro-des such as solder. Such complex architectures hindermanufacturability, scalability, and overall device-level efficiency7,32. Incontrast, SSE devices utilize simplified transverse configurations thateliminate the need for electrically and thermally sensitive junctions(Fig. 1a)7–9. This architectural simplicity enhances the mechanicalrobustness and scalability of devices, while also enablingmore flexiblematerial design. More importantly, the SSE enables thermoelectricconversion even in ferro(i)magnetic insulators (FMI), which are inac-cessible to conventional thermoelectrics relying on mobile chargecarriers. This unique capability allows SSE-based devices to harvestwaste heat from insulating materials, thereby expanding the range ofusable thermal sources beyond that of traditional approaches. Inparticular, the independent generation and propagation of spin andReceived: 29 August 2025Accepted: 25 June 2026Check for updates1National Institute for Materials Science, Tsukuba, Japan. 2Toyota Technological Institute, Nagoya, Japan. 3Department of Advanced Materials Science,Graduate School of Frontier Sciences, The University of Tokyo, Kashiwa, Japan. e-mail: PARK.SangJun@nims.go.jp; UCHIDA.Kenichi@nims.go.jpNature Communications |         (2026) 17:6389 11234567890():,;1234567890():,;http://orcid.org/0000-0003-1684-4876http://orcid.org/0000-0003-1684-4876http://orcid.org/0000-0003-1684-4876http://orcid.org/0000-0003-1684-4876http://orcid.org/0000-0003-1684-4876https://orcid.org/0000-0002-0917-3907https://orcid.org/0000-0002-0917-3907https://orcid.org/0000-0002-0917-3907https://orcid.org/0000-0002-0917-3907https://orcid.org/0000-0002-0917-3907http://orcid.org/0000-0002-7856-7897http://orcid.org/0000-0002-7856-7897http://orcid.org/0000-0002-7856-7897http://orcid.org/0000-0002-7856-7897http://orcid.org/0000-0002-7856-7897http://orcid.org/0009-0003-7789-8170http://orcid.org/0009-0003-7789-8170http://orcid.org/0009-0003-7789-8170http://orcid.org/0009-0003-7789-8170http://orcid.org/0009-0003-7789-8170http://orcid.org/0000-0002-5577-8018http://orcid.org/0000-0002-5577-8018http://orcid.org/0000-0002-5577-8018http://orcid.org/0000-0002-5577-8018http://orcid.org/0000-0002-5577-8018http://orcid.org/0000-0001-7680-3051http://orcid.org/0000-0001-7680-3051http://orcid.org/0000-0001-7680-3051http://orcid.org/0000-0001-7680-3051http://orcid.org/0000-0001-7680-3051http://crossmark.crossref.org/dialog/?doi=10.1038/s41467-026-75232-0&domain=pdfhttp://crossmark.crossref.org/dialog/?doi=10.1038/s41467-026-75232-0&domain=pdfhttp://crossmark.crossref.org/dialog/?doi=10.1038/s41467-026-75232-0&domain=pdfhttp://crossmark.crossref.org/dialog/?doi=10.1038/s41467-026-75232-0&domain=pdfmailto:PARK.SangJun@nims.go.jpmailto:UCHIDA.Kenichi@nims.go.jpwww.nature.com/naturecommunicationscharge currents in the FM and NM layers, respectively, allow forseparate optimization of each material, circumventing intrinsic trade-offs inmaterial parameters, such as those dictated by theWiedemann-Franz law28. These advantages make SSE-based devices promisingcandidates for scalable and reliable thermoelectric energy harvesting,despite their relatively low thermopower to date.Over the past decades, numerous efforts have been made toenhance the SSE by improving energy conversion efficiencies withinthe FM and NM, and at their interface, for example, through magnontemperature manipulation16, the use of topological materials33, andinterfacial engineering22,34. However, the practical implementation ofSSE remains challenging, primarily due to the limited power outputresulting from the hetero-structured nature of thin-film FM/NM devi-ces (Fig. 1a). The device thickness is fundamentally constrained by thecharacteristic length of spin transport, namely the spin diffusionlength (λs), over which the spin angularmomentumdecays. As a result,the optimal thickness of the NM layer is typically limited to a fewnanometers (e.g., less than 5 nm for sputtered Pt35), leading to highinternal resistance and reduced output power.The effective FM thickness is similarly governed by the magnondiffusion length (λm) particularly in FMI, where spin angular momen-tum is carried solely by magnons, i.e., collective dynamics of localizedmagnetic moments36–38. For instance, yttrium iron garnet (YIG), awidely studied FMI known for its exceptionally low Gilbert dampingconstant (<10−4)39, exhibits λm of ~10 µm at room temperature40,41,which is significantly longer than that of ferro(i)magnetic metals(FMMs), e.g., 3.5 nm for sputtered Ni42. Nevertheless, these lengthscales remain microscopic, limiting device configurations to quasi-two-dimensional thin-film geometries (Fig. 1a), where only narrowvolumes of FM and NM adjacent to the interface contribute to the SSE(see Fig. 1c). In addition, this planar architecture introduces opera-tional anisotropy; the heatflux andmagnetization of FMmust be in thedirections perpendicular and parallel to the FM/NM interface,respectively, thereby restricting practical applicability. To overcomethese structural limitations and enable scalable SSE-based energyharvesting, a transition from thin-film architectures to bulk-compa-tible, three-dimensional geometries is essential, which we refer to astrans-scaling.One approach to implementing trans-scaling is to engineerextrinsic device parameters. Specifically, a three-dimensional nanos-tructured composite material24,43 can be constructed by embeddingmultiple NM-coated FM domains throughout the bulk (Fig. 1b). Thisdesign effectively extends conventional quasi-two-dimensional SSEdevices into bulk composites, increasing the usable thermal volume,which enhances the overall heat-to-electricity conversion capabilitywithout relying on high-resistance thin-film architectures. This con-cept, referred to as the trans-scale SSE, leverages composite archi-tectures in which NM channels are distributed throughout the FMmatrix, allowing volumetric utilization of spin currents and cir-cumventing the intrinsic limitations imposed by λs and λm (Fig. 1d).To elucidate the physical basis that enables the trans-scale SSE tosurvive such a geometrically complex FM/NM interface network, wefirst recall that in an FM/NM bilayer the ISHE field obeysEISHE / js × m̂ ð1Þwhere js is the spatial direction of the thermally driven spin currentinduced by a temperature gradient ∇T , and m̂ is the magnetization(spin-polarization) direction. For anarbitrarily oriented interfacepatchwith unit normal n̂, the injected spin current flows along n̂, and itsmagnitude scales with the normal projection of ∇T . The local ISHEresponse therefore generalizes toEISHEðn̂Þ / ðn̂ � ∇TÞðn̂ × m̂Þ: ð2ÞBecause both prefactors in Eq. (2) reverse sign under inversion ofthe interface normal (n̂ ! �n̂), their product remains invariant,ensuring that rough, concave, convex, or oppositely oriented facetscontribute constructively rather than canceling.When the orientationsof n̂ are statistically isotropic in three-dimensional space, the orienta-tional average reduces toðn̂ � ∇TÞðn̂× m̂Þ� �=13ð∇T × m̂Þ ð3Þindicating that the emergent macroscopic SSE retains the usual ISHEsymmetry ∇T × m̂ independent of morphological complexity, with ageometric reduction factor of 1/3. A full derivation is provided inSupplementary Note 1.Despite its conceptual promise, the realization and developmentof such three-dimensional SSE devices remain challenging due to dif-ficulties in NM-FM coating and sintering. First, conventional methodsfor coating FMwith NM, such as atomic layer deposition (ALD), rely onchemicalprecursors andoftenproducepoor FM/NM interfacial qualitydue to residual organic contaminants and weak mechanical FM/NMbonding. These interfacial problems significantly degrade the spinmixing conductance; for example, YIG/Pt interfaces prepared by ALDexhibit spin mixing conductance reduced by a factor of ~20 comparedto those prepared by sputtering44. FMMs, while easier to fabricate intoNMmicrochannel structures, exhibit very short λs on the order of a fewnanometers42 and are susceptible to oxidation, which limits theireffectiveness in such composites43. Moreover, electrical shuntingbetween the FMM and NM can substantially reduce the SSE signalgenerated in the NM layer. The anomalous Nernst effect (ANE)10,45FMNMFMNMTrans-scale SSE(Quasi-3D, this study)Conventional SSE(Quasi-2D)∇Volumetric contribution of spin currentsInterfacial contribution of spin currentsHH∇~ λs~ λma bc dFig. 1 | Conceptual comparison between conventional thin-film-based spinSeebeck effect (SSE) and the proposed trans-scale SSE. a Schematic of a typicalquasi-two-dimensional SSE device composed of a ferro(i)magnetic material (FM)and a normalmetal (NM) layer. The output electric field is generated orthogonal tothe applied thermal gradient (∇T) and magnetic field (H). b Schematic of a trans-scale SSE device, realized via a bulk composite of NM-coated FM domains. c In theconventional SSE device, the device length scale is determined by spin diffusionlengthofNM λs andmagnondiffusion length of FM λm.Here, only spin currents andmagnons that reach the FM/NM interface contribute to SSE, limiting device per-formance. d In the trans-scale SSE, distributed NM channels allow efficient localconversion of magnons throughout the FM volume, enhancing effective spin cur-rent utilization.Article https://doi.org/10.1038/s41467-026-75232-0Nature Communications |         (2026) 17:6389 2www.nature.com/naturecommunicationsgenerated in the FMMcan also interfere with the detection of the pureSSE contribution. Therefore, oxide-based FMIs with long λm, such asYIG, are more desirable. However, they are structurally more difficultto process due to their thermodynamic stability, which requires hightemperatures for sintering (e.g., above 900 °C for YIG46). Such hightemperatures can lead to several issues in the thin NM layer47,48,including increased surface roughness, film discontinuity, and dewet-ting, all of which degrade layer quality (see Supplementary Note 2 fordetailed discussion). Overcoming these challenges is thus crucial forrealizing SSE-based energy harvesting devices that can effectively uti-lize heat from insulators at macroscopic scales. While the conceptualstructure of bulk SSE composites was outlined nearly a decade ago43,their practical realization has remained difficult due to the intrinsicconstraints discussed above. The successful demonstration of suchsystems would represent a significant advance toward practical spin-caloritronic applications.ResultsFabrication of YIG-Pt composite materialsHere, we demonstrate the trans-scale SSE using FMI-NM compositematerials composed of multiple YIG-Pt domains distributed through-out the bulk (Fig. 1b). To fabricate this structure, we established adynamic powder sputtering system49 (Supplementary Fig. 1 for pho-tographic setup) that enables uniform, nanometer-scale Pt coating onYIG powders without the use of chemical precursors (Fig. 2 andMethods). UniformPt coatingwas achievedby simultaneously rotatingand vibrating the YIG powders during deposition. Figure 2b shows atransmission electron microscopy (TEM) image of Pt-coated YIGpowders, confirming the formation of a continuous 5-nm-thick Ptlayer. When the Pt thickness was reduced below 5 nm, the coatingbecame non-uniform (Supplementary Fig. 2).This metallic coating not only provides spin-caloritronic func-tionality (i.e., thermal spin pumping), but also improves mechanicaladhesion between powders owing to the ductility and high atomicmobility of metals compared to oxides. As a result, the Pt-coatedpowders could be sintered through metal-mediated ductile channelsat low temperatures, including 300 °C and even at room temperature(Methods), while maintaining the conducting paths in the bulkcomposite. In contrast, pure YIGpowders could not be densifiedunderthe same conditions, highlighting the critical role of Pt layers inenabling densification.Scanning electronmicroscope (SEM) images reveal the formationof micrometer-scale Pt channels, which increase the effective volumeof both FM and NM materials contributing to the SSE (Fig. 2c). X-raydiffraction (XRD) confirmed that the YIG remains the predominantphase without the formation of secondary phases during fabrication(see Supplementary Note 3 and Supplementary Fig. 3 for detaileddiscussion). A more comprehensive discussion of potential contribu-tions from secondary phases is presented in Discussion, where we alsoexamine the exclusion of possible signal contamination from parasiticthermoelectric effects.Four composite samples were prepared to experimentallydemonstrate the trans-scale SSE. These samples differed in Pt thick-ness (15 nm and 30nm), both of which are well above λs of Pt ( < 5nm35), ensuring clear observation of the SSE without discontinuities inthe Pt channels. In addition, two different sintering conditions wereemployed: pressing at 300 °C and room temperature under highpressure (Methods). The sintered pellets were labeled according totheir Pt thickness and sintering temperature (i.e., 15-300, 15-RT, 30-300, and 30-RT), and they exhibited well-consolidated structures withrelative densities ranging from 67% (15-RT) to 73–75% (15-300, 30-300,and 30-RT), as shown in Supplementary Fig. 4.Electrical and thermal conductivity of nanostructured bulkcompositesWe then evaluated the electrical conductivity (σ) of the bulk compositeby normalizing the volumetric electrical conductance with the geo-metric dimensions of the composite, including YIG and Pt (Methods).The measured σ values range from (5–8) × 102S/m for 15 nm Pt sam-ples and (23–50) × 102S/m for 30nm Pt samples at room temperature(Supplementary Fig. 4a), all of which fall within the “bad metal”regime50. These values are higher than those of typical homogeneoussemiconductors (non-doped Si ~ 1.5 × 10−3S/m and Ge ~ 2 S/m) andoxide-based bulk ANE materials ((3–26) × 102S/m)51–53.Next, we estimated the thermal conductivity (κ) by multiplyingthe measured thermal diffusivity, specific heat, and densityPtYIGNi (protection layer)Ni (protection layer)PtYIGYIGPowderPt coating by sputterYIG Single X’talYIG-Pt bulk compositeLow-T sinteringcabPt surrounding YIG grainsFig. 2 | Fabrication and characterization of YIG-Pt composite materials fortrans-scale spin Seebeck effect. a Schematic illustration of the fabrication pro-cess. YIGpowders are coatedwithPtusingdynamicpowder sputtering, followedbylow-temperature sintering to produce bulk composites. b TEM images showinguniform nm-scale Pt coating on the surface of YIG powders. cMicrostructural andcompositional characterization of the bulk composite. SEM images (SE and BSE)reveal grain size distribution. EDX elemental mappings confirm uniform distribu-tion of Y, Fe, and O, throughout the composite matrix and Pt surrounding YIGgrains.Article https://doi.org/10.1038/s41467-026-75232-0Nature Communications |         (2026) 17:6389 3www.nature.com/naturecommunications(Supplementary Fig. 4). The κ values are 1.46–1.87W/mK, which aresubstantially lower than those reported for polycrystalline YIG (~4.9W/mK) and single-crystalline YIG (~7–8W/mK)46,54. Such low κ, corre-sponding to 30–38% of the polycrystalline reference, cannot beattributed solely to the relative density, for example, 75% for 30-RT.The reduction likely reflects multiple phonon-scattering channelsintroduced by the composite structure, including porosity, grainboundaries, interface defects, and thermal boundary resistance at theYIG/Pt interfaces. In insulating YIG, heat conduction is mainly gov-erned by phonons, with a relatively small magnon contributionexpected at room temperature in polycrystalline samples46,54. There-fore, controlling the YIG/Pt interfacial density, defect structure, andpore morphology may provide an additional route for thermal-conductivity engineering in bulk SSE composites.Observation of trans-scale SSE in nanostructured bulkcompositesWe now present the observation of the trans-scale SSE in the YIG-Ptcomposites under two distinct magnetothermal configurations: dub-bed “Hz-∇xT” (“Hx-∇zT”) configuration measuring voltage into the y-direction, while the orthogonal H is applied to z (x)-direction andtemperature gradient (∇T) is applied to x (z)-direction, as shown inFig. 3a, d). While rotating the magnetothermal configurations are notexpected to generate a consistent SSE signal in conventional FM/NMplanar film structures14,20 due to lack of spin pumping under out-of-plane H, we show that both geometries consistently produce SSE sig-nals in our composites, highlighting a distinctive feature of isotropictrans-scale SSE. In the following, we provide a systematic evaluation torigorously validate the observation by excluding potential parasiticcontributions that could influence voltage measurement and byassessing the effects of secondary phases. These parasitic signalsinclude the Seebeck effect driven by oblique heat flux55,56, the (proxi-mity-induced) ANE14,20,57, and magnetoresistance effects.We first focus on the observed transverse electric field (Ey) in theHz-∇xT configuration (Fig. 3a). The signal was symmetrized withrespect to H to remove field-even components, such as the magneto-Seebeck effect caused by unintended transverse heat gradients (∇yT)in Pt. As shown in Fig. 3b, a clear field-odd signal was observed in the30-RT samplewithin theHz rangeof ± 5 kOe. The lineardependenceofEy on ∇xT was confirmed in Fig. 3c, using Esaty denoting Ey at the satu-rated fields (>2.5 kOe). These results directly confirm that the observedsignal originates from magnetism-induced transverse thermoelectricconversion, with longitudinal thermoelectric and magneto-electriccontributions effectively excluded. The linear slope of Fig. 3c corre-sponds to the spin Seebeck coefficient (SSSE � Esaty =∇xT), shown inSupplementary Fig. 5. Samples sintered at room temperature exhibitedSSSE values of 17.3 nV/K (15-RT) and 14.8 nV/K (30-RT), which are higherthan those of samples sintered at 300 °C (9.2 nV/K for 15-300 and7.5 nV/K for 30-300). The 15-nm-thick Pt samples exhibited slightlyhigher SSSE than the 30-nm-thick Pt samples, qualitatively consistentwith the typical Pt thickness dependence of SSSE in the conventionalYIG/Pt systems due to the small λs of Pt35. Using the estimated σ andSSSE, we calculated the transverse power factor (PFSSE = σSSSE2) drivenby SSE. Supplementary Fig. 5b shows that 30-RT exhibited the largestPFSSE (5.1 × 10−13W/mK2), approximately 630% higher than that of 15-300 (0.7 × 10−13W/mK2), strongly suggesting the importance of opti-mizing the Pt thickness and sintering conditions to enhance thetransverse thermoelectric performance.Next, we demonstrate the SSE in the Hx-∇zT configuration usingthe 30-RT sample by rotating it 90° to the original magnetothermalconfiguration, where the Ey signal was measured under Hx and ∇zT(Fig. 3d). In conventional anisotropic FM/NM thin-film structures, no0 20 4001dezilamroNEy or MHx (kOe)EyM05001000VHally (nV)0.0 2.5 5.0050100Hz (kOe)Ey = Vy /L y()m/VxT [K/mm]5.884.713.532.351.1800 2 4 60255075100125Esaty()m/VT (K/mm)15-30015-RT30-30030-RTHzVyxT30-RT sampleYIGPtHxVyzT or IzzT = 2.64 K/mmxyzYIGPt30-RT samplea b cd e fIz = 100 mA00 5 1002550Ey=Vy /L y()m/VHx (kOe)30-RT sampleFig. 3 | Observation of the trans-scale spin Seebeck effect. a Schematic illustra-tions ofHz-∇xT configuration. bMeasured transverse electric field (Ey) under ∇T inHz-∇xT configuration as a function of H. The subscripts x, y, and z indicate thecorresponding directions of the applied temperature gradient (∇T), charge current(I), and magnetic field (H). (c) Saturated electric field (ESaty ) plotted as a function of∇xT . Error bars in (c) represent the standard deviation. d Schematic illustrations ofHx-∇zT configuration. eMeasured transverse electric field (Ey) under ∇T in Hx-∇zTconfiguration. The measured Hall voltage (VHally ) is also presented for reference,alongwith a linearfitting curve. fNormalized Ey andMagnetization (M) of the 30-RTsample under high H. Detailed M-H curves for the other samples are provided inSupplementary Fig. 6.Article https://doi.org/10.1038/s41467-026-75232-0Nature Communications |         (2026) 17:6389 4www.nature.com/naturecommunicationsSSE signal is expected in this rotated magnetothermal configuration,due to the absence of spin pumping at the FM/NM interface. Only(proximity-induced) ANE can be captured in such cases14,20. Figure 3eshowsa clearfield-odd signal, with SSSEmeasured as 14.4 nV/K,which isconsistent with that obtained in the Hz-∇xT configuration (Fig. 3c andSupplementary Fig. 5). This consistency supports the assignment ofthe signal to SSE and demonstrates the reproducibility of the bulk SSEresponse under orthogonal magnetothermal configurations, whileconfirming the directional isotropy of the system and its potential forenergy harvesting applications.Validation of trans-scale SSE and exclusion of parasitic effectsOne may raise concerns about parasitic contributions from the(proximity-induced) ANE, which can arise in both Hz-∇xT and Hx-∇zTconfigurations, if secondary (proximity-induced) FMM phases wereunintentionally formed during fabrication. However, we rule this outbased onmultiple lines of evidence. First, XRDmeasurements shownodetectable formation of FMMs, such as Fe3O4 or FePt (SupplementaryFig. 3b and Supplementary Note 3), thereby excluding their contribu-tions to ANE. This is further supported by the magnetization data: ifsuch FM phases were present, the saturation magnetization wouldincrease, considering that the ferrimagnetic YIG (37.9 emu/g) exhibitsmuch lower values than those of common Fe-based FMs (85–92 emu/gfor Fe3O458, 74-76 emu/g for γ-Fe2O359, and 186-218 emu/g forα-Fe60). Incontrast, we observe a decrease in saturation magnetization upon thedeposition of paramagnetic Pt (Supplementary Fig. 6), and the mag-netization curve shows a single, well-defined hysteresis loop withoutanomalies, confirming the absence of secondary magnetic phases(Supplementary Fig. 6 for detailed low-field data).We also exclude the contribution of proximity-induced ANEbecause the magnetic proximity effect between YIG and Pt and its con-tribution to transverse thermoelectric effect are well known to be neg-ligibly small, as demonstrated by X-ray magnetic circular dichroism andtransport studies14,61–64. Consistently, SQUID-VSM measurements showno additional magnetic component or anomalous magnetic behavior.Even if local strain-induced interfacial modification exists below thedetection limit, itwould affect interfacial spin transparencyor SSE-drivenISHE efficiency65 rather than generate an independent ANE-like voltage.To further confirm the absence of such contributions from secondary(proximity-induced) FMMs, we performed Hall measurements, measur-ing Ey under a charge current Iz and Hx, to sensitively detect any (proxi-mity-induced) FMMs effectively contributing to the transport (Fig. 3e).The Hall voltage (VHally ) shows a linear response to Hx, indicating theabsence of (proximity-induced) FMMs effectively contributing thetransverse signals. Furthermore, the transverse thermoelectric signalexhibits a suppression at high magnetic fields (Fig. 3f), which is a well-establishedhallmarkof theSSE13,15. This suppression arises fromthefield-induced reduction of thermally excitedmagnons and is not expected forANE of FMM impurities, whose transverse voltage simply follows themagnetization and saturates at high fields. This high-field dampingtherefore provides an additional signature that the observed transversevoltage originates from magnon-driven SSE rather than ANE.Finally, to provide an independent and unambiguous confirma-tion, we prepared a YIG–W composite, where W has a negative spinHall angle. Figure 4 shows that the transverse thermoelectric voltageofthe YIG–W composite in the Hz-∇xT configuration exhibits a clear signreversal relative to the YIG–Pt composite under otherwise identicalconditions, despite the higher noise level associated with the largerelectrical resistance of the control sample (Supplementary Note 4).This inversion provides direct evidence that the observed transversevoltage originates from themagnon-driven thermal spin pumping andits ISHE-driven spin-charge conversion in the bulk composite, estab-lishing the trans-scale SSE.We additionally note that the SSE originates from magnons gen-erated in YIG, not from secondary magnetic phases, such asantiferromagnetic α-Fe2O3 and YFeO3. While both are reported toexhibit long-rangemagnon transport66,67, their formation could not bedistinguished in the XRD data shown in Supplementary Fig. 3 and theirmagnon transport is known to be strongly anisotropic and to requirehigh magnetic fields. These characteristics suggest that they are inerttomagnon transport within the field range used in ourmeasurements.Thickness-dependent power scaling enabled by volumetric spinconversionWe evaluate the device-level performance using the maximum outputpower,Pmax =V2oc4Rint, ð4Þwhere Voc is the open-circuit voltage and Rint is the internal resistanceof the device. To isolate the architectural contribution, we fix thelateral device dimensions (Ly and Lz) and the applied heat-fluxboundary condition, while varying only the device or FM thicknesst (Fig. 5).For conventional quasi-two-dimensional interfacial SSE devices(Fig. 1a), theopen-circuit voltage followsV2Doc ðtÞ / S0ð1� e�t=λm Þ. Underfixed NM geometry, the internal resistance remains approximatelythickness independent. Therefore,P2Dmax tð Þ / ð1� e�t=λm Þ2, ð5Þwhich saturates for t≫λm. In the present analysis, a representativeλm= 10 µm, as reported for YIG at room temperature17,68, was adoptedfor the 2D reference calculations.In contrast, for the present quasi-three-dimensional composite,the percolatedmetallic NMnetwork increases the effective conductivecross-section with thickness, resulting in R3Dint / 1=t. Within theexperimentally accessible thickness range (≫λm), the effective SSE isnot limited by interfacial magnon diffusion and therefore does notexhibit the saturation characteristic of two-dimensional systems.−5.0 −2.5 0.0 2.5 5.0−101Ey/Esaty (normalized)Hz (kOe) YIG-Pt (30-RT) YIG-W (30-RT)Fig. 4 | Normalized spin Seebeck voltage for YIG–Pt and YIG–W bulk compo-sites. The two samples exhibit opposite signs of transverse thermoelectric voltage,reflecting the opposite signs of the spin Hall angles of the nonmagnetic metals, Pt(positive) andW (negative). The signal in the YIG–Wcomposite appears noisier dueto the large electrical resistance of the non-optimized composite. See Supple-mentary Note 4 for fabrication details and discussion.Article https://doi.org/10.1038/s41467-026-75232-0Nature Communications |         (2026) 17:6389 5www.nature.com/naturecommunicationsConsequently,P3Dmax tð Þ / t, ð6Þleading to a non-saturating increase of output power with increasingthickness. A detailed derivation of these scaling relations, includingexpressions in terms of experimentally measurable parameters, isprovided in Supplementary Note 5.For the three-dimensional composite, experimentally measuredtransport parameters from the 30-300 and 30-RT samples were usedfor the scaling analysis, including κ = 1.87W/mK,σ = 2.31 × 103S/m, andSSSE = 14.8 nV/K for the 30-RT sample. For the quasi-two-dimensionalreference systems, representative material parameters reported forpolycrystalline NiFe2O4/Pt21, single-crystalline YIG/Pt69, and epitaxialYIG films/WSe2/Pt34 heterostructures were adopted, using literaturevalues of SSSE, κ of the FM layer and σ of the NM layer. Using theseparameters (see Supplementary Table 1), Fig. 5 presents Pmax as afunction of thickness under identical device area (10mm × 10mm)and heat-flux (jQ = 1.5 × 104W/m2) boundary conditions. For the three-dimensional composite, the scaling analysis is shown only in theregime where the thickness exceeds both λm and the experimentallyaccessible range (>0.1mm), corresponding to the volumetric scalingregime. The quasi-three-dimensional composite architecture exhibitsan approximately linear increase in output power with thickness,whereas conventional quasi-two-dimensional devices exhibit satura-tion behavior governed by λm. This contrast demonstrates that thepresent architecture removes the intrinsic interfacial thickness lim-itation characteristic of thin-film SSE systems and establishes a bulk-operating scaling regime enabled by volumetric spin-to-charge con-version. The dashed lines in Fig. 5 represent projected scaling behaviorassuming higher intrinsic SSSE comparable to those reported in opti-mized thin-film systems (on the order of 1 µV/K), whilemaintaining thesame three-dimensional architecture and boundary conditions.Because the maximum output power in the volumetric regime followsP3Dmax / S2SSEt, increasing the intrinsic SSSE enhances the slope of thelinear thickness scaling without altering the linear thicknessdependence.DiscussionIn summary, we have demonstrated the observation of trans-scale SSEusing YIG-Pt bulk composite materials. The composites were fabri-cated via a dynamic powder sputtering method followed by low-temperature sintering. These bulk materials exhibit clear transversethermoelectric signals at the bulk scale, consistent with volumetricspin-to-charge conversion. Through rigorous signal and materialcharacterization, we confirm that the observed signals originate purelyfrom the SSE of YIG, with negligible parasitic contributions.We believe this study represents a key step toward the develop-ment of practical SSE-based thermoelectric devices by bridging thetechnical gap between conventional nanoscale SSE effects, which arelimited by the spin and magnon diffusion lengths, and bulk deviceplatforms. From the intrinsic (materials) perspective, furtherimprovement in output power can be achieved by optimizing FM/NMinterfaces21,22, for example by inserting topological materials34,70 ororganic semiconductors71 at the interface, as well as engineeringmagnon transport in YIG independently from phonons16,21. In addition,replacing YIG with alternative high-performance FMI with lower Gil-bert damping constants and longer magnon diffusion lengths, andadopting NM materials with a giant inverse spin Hall effect, such astopological insulators and Weyl semimetals72–77, could substantiallyincrease conversion efficiency at bulk scale. These NM materials offerlong spin diffusion lengths (up to 2 µmat low temperature78,79 and evenat room temperature80) and spin Hall conductivities that exceed thoseof Pt by orders of magnitude, providing a direct route to boost spinSeebeck coefficient and the resulting output power. Many of thesestrategies have already been demonstrated independently in thin-filmspintronics and spin caloritronics, where significant performanceimprovements compared to Pt-based systems have been achieved.Translating these advances into bulk architectures could thereforeaccelerate the development of practical SSE-based energy harvesters.Moreover, hybridization with other transverse thermoelectric phe-nomena, such as the ANE7,10,43,81, which has recently shown remarkableperformance enhancements in topological materials82–86 and structu-rally heterogeneous systems45,87–89,mayprovide anadditional route forenhancing bulk-scale performance. Hybrid transverse thermoelectricresponses combining SSE and ANE have been reported in YIG/FMMbilayer systems, for example using Ni as the ferromagnetic metal9,where anomalous Nernst response and spin-to-charge conversioncoexist within the same device geometry. In this context, employingFMMs that exhibit both a sizable ANE and efficient spin-to-chargeconversion, including anomalous ISHE90, could enable constructivecoexistence of ANE- and SSE-driven transverse voltages within thesame device geometry. Because both effects share identical symmetrywith respect to magnetization and thermal gradient, theirbaLy = 10 mmLz = 10 mmjQ = 1.5×104 W/m2λm = 10 µm0 5 10 15 200246810Pmax = V2 oc/4Rint (pW)t (mm)kluBmliF kluBmliF0.001 0.01 0.1 1 10 1000.0010.010.1110100Pmax = V2 oc/4Rint (pW)t (mm)3D devices (this work)      30-300                           30-RT                           SSSE=1 �V/K (ref.)    2D devices NFO poly/Pt YIG single/Pt Epi-YIG/WSe2/PtFig. 5 | Thickness-dependent scaling of maximum output power in transversespin Seebeck devices. a Linear-scale plot and b logarithmic-scale plot of themaximum output power Pmax as a function of ferromagnetic layer thickness. Pmaxwas calculated under identical device area (Ly = Lz = 10mm) and heat-flux(jQ = 1.5× 104W/m2) conditions using Pmax =V2oc=4Rint, where Voc is the open circuitvoltage andRint is the internal resistance. A representativemagnon diffusion lengthof λm= 10 µmwas used for the 2D reference calculations. The 2D reference systemswere modeled using representative material parameters from polycrystallineNiFe2O4 (NFO poly)/Pt21, single-crystalline YIG (YIG single)/Pt69, and epitaxial YIGfilms (Epi-YIG)/WSe2/Pt34 heterostructures reported in the literature. For the 3Darchitecture, the thickness-dependent scaling is shown only for thicknessesexceeding both λm and the experimentally accessible range (>0.1mm), therebyrepresenting the volumetric scaling regime. Detailed assumptions and parametervalues are provided in Supplementary Note 5.Article https://doi.org/10.1038/s41467-026-75232-0Nature Communications |         (2026) 17:6389 6www.nature.com/naturecommunicationscontributions may, in principle, superpose and yield synergisticenhancement when their signs are aligned.From the extrinsic (structural) perspective, our symmetryanalysis in Eq. (3) reveals that the trans-scale SSE amplitude in thepresent composites contains a geometric reduction factor of 1/3arising from the statistically isotropic distribution of local FMI/NMinterface normals. Structural engineering that induces anisotropicor partially aligned interface orientations, for example, via con-trolled grain-shape anisotropy, magnetic-field-assisted forming, ordirectional sintering, can bias the distribution of interface normalsand increase the geometric contribution toward the planar-filmlimit (i.e., the reduction factor is 1). This identifies structuralorientation control as an additional, previously unexplored degreeof freedom for amplifying trans-scale SSE signals in 3Darchitectures.Altogether, the demonstration of trans-scale SSE in a bulkcomposite platform establishes a previously inaccessible regime oftransverse thermoelectric conversion that extends well beyond theconstraints of conventional thin-film architectures. By enablingvolumetric spin-current generation and macroscopic ISHE conver-sion within a mechanically robust 3D network, this work provides apractical and scalable route for implementing spin-based thermo-electric functionality at bulk scale. Unlike interfacial thin-film SSEsystems, the present architecture supports bulk-scalable geome-tries in which output power increases with device thickness withoutinterfacial saturation, thereby expanding the accessible designspace for transverse thermoelectric energy conversion. Theseresults position trans-scale SSE composites as a promising platformfor next-generation transverse thermoelectrics based on spincurrents.MethodsSample preparationThe Pt-coated YIG powders were prepared using a customizeddynamic sputtering system for powder coating (SUGA Co., Ltd., SSP-1500B) with a base pressure of less than 5 × 10−4Pa. The initial YIGpowders were prepared by crushing YIG single crystals (FerrisphereInc.). The sputtering system vibrated the powders at a frequency of22.5 Hz and rotated the powders at approximately 17 rpm to stir them.Pt was sputtered from a 99.99%-purity Pt target with an Ar gas flow of6.0 sccmat room temperature. The plasmapowerwas set to 50W. The5-nm Pt sample shown in Fig. 2 was deposited for 30min, and the Ptthicknesses of themain samples (15 nmand 30nm)were labeled basedon the corresponding deposition durations.The Pt-coated YIG powders were then sintered into bulk form byhot pressing under two low-temperature, high-pressure conditions: (1)300 °C at 350MPa and (2) room temperature at 500MPa. Sinteringwas performed using a hot-pressing apparatus (AS ONE, H400-15) inair, employing a tungsten carbide die and punches with a diameter of10mm. The samples were cut using a diamond wire saw (DiamondWireTec, DWS.100) for subsequent materials characterization. Allproperties reported in this study were measured from samples pre-pared in a single batch.Structural characterizationStructural analysis was performed using a Rigaku MiniFlex600 X-raydiffractometer with Cr-Kα radiation (wavelength = 0.22897 nm). Thediffraction angleswere converted to commonly usedCu-Kα equivalent(wavelength = 0.15406 nm) for presentation in Supplementary Fig. 3.The X-ray tube was operated at 40kV and 15mA.Microstructural observations were performed using a Carl ZeissCrossBeam 1540EsB scanning electron microscope (SEM) equippedwith energy-dispersive X-ray spectroscopy (EDS). Transmission elec-tron microscopy (TEM) was carried out with a Titan G2 80-200microscope equipped with a probe aberration corrector. TEMspecimens were prepared using the lift-out technique with an FEIHelios 5UX dual-beam focused ion beam system.Measurement of thermal and electrical conductivitiesThe electrical conductivity (σ) was measured using a standard four-probe method with a current source (Keithley, 2450) and a nanovolt-meter (Keithley, 2182a). A linear voltage–current response was con-firmed by measuring 9 points with reverse polarities of the current.The thermal conductivity (κ) was estimated by multiplying ther-mal diffusivity, heat capacity, and density. Owing to the statisticallyrandom orientation of Pt-coated YIG particles in the sintered compo-site, the macroscopic thermal and spin transport properties are trea-ted as effectively isotropic in the device-level analysis. The thermaldiffusivitywas obtained using a laserflashmethod (LINSEIS, LFA1000),specific heat was measured by DSC (Rigaku, Thermo plus EVO2), anddensity was calculated from the sample mass and volume at roomtemperature. TheDSCmeasurementwas conductedwith analuminumpan over a temperature range of 5 °C to 45 °C with a ramp rate of 1 °C/min. Alumina (Al2O3) powder was used as a reference for the DSCmeasurement.Measurement of the spin Seebeck effectThe SSE was measured in both Hz-∇xT and Hx-∇zT configurations. ForHz-∇xT configuration, two Cu wires were attached along the y-direc-tion using silver epoxy and the sample was sandwiched between twothermally conductive, electrically insulating sapphire blocks serving asheat baths. Thermal grease was used at the interfaces to ensure auniform temperature gradient. The x-directional temperature gradientapplied was controlled using a PID-based temperature controller withmeasured temperaturedifferences usingdifferential thermocouples. Itis worth noting that themeasured temperature difference includes notonly the drop across the sample but also the interfacial drops. Thisleads to an overestimation of the applied gradient across the samplesand thus anunderestimationof the reported spin Seebeckcoefficients.Although this effect has been reported to be appreciable even at roomtemperature for YIG single crystals (despite the use of thermalgrease)91, we expect the error here to be smaller because the thermalconductivity of our samples is lower (1.46–1.87W/mK), i.e., about18–27%of that of single-crystalline YIG (~7–8W/mK46,54). Consequently,a larger fraction of the total temperature drop occurs within thesample rather than at the interfaces. A current source (ADVANTEST,R6243) and a multimeter (Keithley, 2000) were used for thermal gra-dient control. The voltage was measured using a nanovoltmeter(Keithley, 2182a)while applyingmagneticfieldswith anelectromagnet.TheHx-∇zT configuration was conducted in a similar manner, with the30-RT sample bridged by two Cu blocks to apply a z-directional tem-perature gradient, as described in refs. 87,92. Adhesive thermal tapes(3M, VHR0601-03) were used to electrically insulate the sample fromthe Cu blocks.Magnetic property analysisThe magnetic properties of the samples were investigated using aSQUID magnetometer (Quantum Design, MPMS). The measurementswere conducted at T = 300K with a sweep rate of 100Oe/s. For thefine-scan measurement in Supplementary Fig. 6c, the magnetizationwas recorded at each field after stabilizing the magnetic field.Measurement of the Hall effectThe Hall effect shown in Fig. 3e wasmeasured underHx and Iz (same astheHx-∇zT configuration) at room temperature using theDC resistivityoption of a He cryostat (Cryogenic Limited, CFMS). Two electrodeswere attached to the sample ends with silver epoxy to apply a currentusing a current source (Keithley, 2450) and two additional electrodeswere attached orthogonally to capture the Hall voltage using a nano-voltmeter (Keithley, 2182a). The input current was 100mA. TheArticle https://doi.org/10.1038/s41467-026-75232-0Nature Communications |         (2026) 17:6389 7www.nature.com/naturecommunicationsmeasurement was conducted over a magnetic field range of ± 10 kOe,consistent with the Hx-∇zT SSE measurement, using a sweep rate of50Oe/s. The data in Fig. 3ewere symmetrized to remove themagnetic-field-even dependent components (e.g., magnetoresistance), follow-ing the same procedure used for the SSE data.Data availabilityAll data that support the findings of this study are available within thearticle and Supplementary Information. Source data underlying theplots in themain and Supplementary Figs. are providedwith this paperas a Source Data file. Source data are provided with this paper.References1. Uchida, K. et al. Observation of longitudinal spin-Seebeck effect inmagnetic insulators. Appl. Phys. Lett. 97, 172505 (2010).2. Bauer, G. E. W., Saitoh, E. & Van Wees, B. J. Spin caloritronics. Nat.Mater. 11, 391–399 (2012).3. Boona, S. R., Myers, R. C. & Heremans, J. P. Spin caloritronics.Energy Environ. Sci. 7, 885–910 (2014).4. Uchida, K. et al. Spin Seebeck insulator. Nat. Mater. 9,894–897 (2010).5. Saitoh, E., Ueda, M., Miyajima, H. & Tatara, G. Conversion of spincurrent into charge current at room temperature: Inverse spin-Halleffect. Appl. Phys. Lett. 88, 182509 (2006).6. Sinova, J., Valenzuela, S. O.,Wunderlich, J., Back, C. H. & Jungwirth,T. Spin Hall effects. Rev. Mod. Phys. 87, 1213–1260 (2015).7. Uchida, K. & Heremans, J. P. Thermoelectrics: From longitudinal totransverse. Joule 6, 2240–2245 (2022).8. Boona, S. R., Jin, H. & Watzman, S. Transverse thermal energyconversionusing spin and topological structures. J. Appl. Phys. 130,171101 (2021).9. Uchida, K. et al. Thermoelectric generation based on spin Seebeckeffects. Proc. IEEE 104, 1946–1973 (2016).10. Adachi, H. et al. Fundamentals and advances in transverse ther-moelectrics. Appl. Phys. Express 18, 090101 (2025).11. Uchida, K. et al. Observation of the spin Seebeck effect.Nature455,778–781 (2008).12. Xiao, J., Bauer, G. E. W., Uchida, K., Saitoh, E. & Maekawa, S. Theoryof magnon-driven spin Seebeck effect. Phys. Rev. B Condens.Matter Mater. Phys. 81, 214418 (2010).13. Kikkawa, T. et al. Critical suppression of spin Seebeck effect bymagnetic fields. Phys. Rev. B Condens. Matter Mater. Phys. 92,064413 (2015).14. Kikkawa, T. et al. Longitudinal spin Seebeck effect free from theproximity Nernst effect. Phys. Rev. Lett. 110, 067207 (2013).15. Jin, H., Boona, S. R., Yang, Z., Myers, R. C. &Heremans, J. P. Effect ofthe magnon dispersion on the longitudinal spin Seebeck effect inyttrium iron garnets. Phys. Rev. B Condens. Matter Mater. Phys. 92,054436 (2015).16. Park, S. J. et al. Enhancing spin pumping by nonlocal manipulationof magnon temperature. Matter 7, 4332–4341 (2024).17. Kehlberger, A. et al. Length scale of the spin Seebeck effect. Phys.Rev. Lett. 115, 096602 (2015).18. Wu, S. M. et al. Antiferromagnetic spin Seebeck effect. Phys. Rev.Lett. 116, 097204 (2016).19. Kimling, J. et al. Picosecond spin Seebeck effect. Phys. Rev. Lett.118, 057201 (2017).20. Bougiatioti, P. et al. Quantitative disentanglement of the spin See-beck, proximity-induced, and ferromagnetic-induced anomalousNernst effect in normal-metal-ferromagnet bilayers. Phys. Rev. Lett.119, 227205 (2017).21. Kim, M. Y., Park, S. J., Kim, G. Y., Choi, S. Y. & Jin, H. Designingefficient spin Seebeck-based thermoelectric devices via simulta-neous optimizationof bulk and interface properties. Energy Environ.Sci. 14, 3480–3491 (2021).22. Kim, M., Park, S. J. & Jin, H. Enhancing the spin Seebeck effect bycontrolling interface condition in Pt/polycrystalline nickel ferriteslabs. J. Appl. Phys. 127, 085105 (2020).23. Ramos, R. et al. Thermoelectric performance of spin Seebeck effectin Fe3O4/Pt-based thin film heterostructures. APL Mater. 4,104802 (2016).24. Lee, K. D. et al. Thermoelectric signal enhancement by reconcilingthe spin Seebeck and anomalous Nernst effects in ferromagnet/non-magnet multilayers. Sci. Rep. 5, 10249 (2015).25. Imamura, M. et al. Enhancement of the spin Seebeck effect owingto bismuth substitution in thermoelectric generators fabricatedfrom LPE Bi-substituted YIG films. AIP Adv. 11, 035143 (2021).26. Venkat, G. et al. Magnon diffusion lengths in bulk and thin filmFe3O4 for spin Seebeck applications. Phys. Rev. Mater. 4,075402 (2020).27. Uchida, K., Nonaka, T., Ota, T. & Saitoh, E. Longitudinal spin-Seebeck effect in sinteredpolycrystalline (Mn,Zn)Fe2O4.Appl. Phys.Lett. 97, 062504 (2010).28. Snyder, G. J. & Toberer, E. S. Complex thermoelectric materials.Nat. Mater. 7, 105–114 (2008).29. Zhang, Q., Deng, K., Wilkens, L., Reith, H. & Nielsch, K. Micro-thermoelectric devices. Nat. Electron. 5, 333–347 (2022).30. Park, S. J. et al. Adaptive thermoelectric cooling system for energy-efficient local and transient heat management. Appl. Therm. Eng.216, 119060 (2022).31. Park, S. J. et al. Thermoelectric hotspot cooling using thermallyconductive fillers. Appl. Therm. Eng. 232, 120994 (2023).32. Yan, Q. & Kanatzidis, M. G. High-performance thermoelectrics andchallenges for practical devices. Nat. Mater. 21, 503–513 (2022).33. Jiang, Z. et al. Enhanced spin Seebeck effect signal due to spin-momentum locked topological surface states. Nat. Commun. 7,11458 (2016).34. Lee, S. K. et al. Enhanced spin Seebeck effect in monolayer tung-sten diselenide due to strong spin current injection at interface.Adv. Funct. Mater. 30, 2003192 (2020).35. Sagasta, E. et al. Tuning the spin Hall effect of Pt from the moder-ately dirty to the superclean regime. Phys. Rev. B 94,060412(R) (2016).36. Serga, A. A., Chumak, A. V. &Hillebrands, B. YIGmagnonics. J. Phys.D Appl. Phys. 43, 264002 (2010).37. Pirro, P., Vasyuchka, V. I., Serga, A. A. & Hillebrands, B. Advances incoherent magnonics. Nat. Rev. Mater. 6, 1114–1135 (2021).38. Chumak, A. V., Vasyuchka, V. I., Serga, A. A. & Hillebrands, B.Magnon spintronics. Nat. Phys. 11, 453–461 (2015).39. Liu, Q. B. et al. Unusual anomalous Hall effect in perpendicularlymagnetized YIG films with a small Gilbert damping constant. Phys.Rev. B 101, 174431 (2020).40. Cornelissen, L. J. et al. Long-distance transport of magnon spininformation in amagnetic insulator at room temperature.Nat. Phys.11, 1022–1026 (2015).41. Shan, J. et al. Criteria for accurate determination of the magnonrelaxation length from the nonlocal spin Seebeck effect. Phys. Rev.B 96, 184427 (2017).42. Ko, K. H. & Choi, G. M. Optical method of determining the spindiffusion length of ferromagnetic metals. J. Magn. Magn. Mater.510, 166945 (2020).43. Boona, S. R., Vandaele, K., Boona, I. N.,McComb,D.W. &Heremans,J. P. Observation of spin Seebeck contribution to the transversethermopower in Ni-Pt and MnBi-Au bulk nanocomposites. Nat.Commun. 7, 13714 (2016).44. Schlitz, R. et al. Spin-Hall-active platinum thin films grown viaatomic layer deposition. Appl. Phys. Lett. 112, 242403 (2018).45. Park, S. J. et al. Designing flexible hardmagnetic materials for zero-magnetic-field operation of the anomalous Nernst effect. ActaMater. 301, 121422 (2025).Article https://doi.org/10.1038/s41467-026-75232-0Nature Communications |         (2026) 17:6389 8www.nature.com/naturecommunications46. Miura, A. et al. Probing length-scale separation of thermal andspin currents by nanostructuring YIG. Phys. Rev. Mater. 1,014601 (2017).47. Golosov, D. A. et al. Stability of the platinum electrode during hightemperature annealing. Thin Solid Films 661, 53–59 (2018).48. Sui, M. et al. Effects of annealing temperature and duration on themorphological and optical evolution of self-assembled Pt nanos-tructures on c-plane sapphire. PLoS One 12, e0177048 (2017).49. Schmid, G. H. S. & Eisenmenger-Sittner, C. A method for uniformlycoating powdery substrates by magnetron sputtering. Surf. Coat.Technol. 236, 353–360 (2013).50. Jaramillo, R., Ha, S. D., Silevitch, D. M. & Ramanathan, S. Origins ofbad-metal conductivity and the insulator-metal transition in therare-earth nickelates. Nat. Phys. 10, 304–307 (2014).51. Kim, M. Y. et al. Oxygen vacancy defect engineering for transversethermoelectric enhancement: a novel extrinsic pathway beyondintrinsic approaches. Adv. Sci. 12, 2502892 (2025).52. Suryanarayanan, R., Gasumyants, V. & Ageev, N. Giant Nernst effectin La0.88MnO3 and La0.7Ca0.3MnO3. J. Magn. Magn. Mater. 211,226–231 (2000).53. Ghosh, A., Das, R. & Mahendiran, R. Skew scattering dominatedanomalous Nernst effect in La1-xNaxMnO3. J. Appl. Phys. 125,153902 (2019).54. Boona, S. R. & Heremans, J. P. Magnon thermal mean free path inyttrium iron garnet. Phys. Rev. B Condens. Matter Mater. Phys. 90,064421 (2014).55. Bang, K. M., Park, S. J., Yu, H. & Jin, H. Large transverse thermo-power in shape-engineered tilted leg thermopile.Appl. Energy368,123222 (2024).56. Tanaka, H. et al. Roll-to-roll printing of anomalous Nernst thermo-pile for direct sensing of perpendicular heat flux. Adv. Mater. 35,2303416 (2023).57. Valvidares, M. et al. Absence of magnetic proximity effects inmagnetoresistive Pt/CoFe2O4 hybrid interfaces. Phys. Rev. B 93,214415 (2016).58. Li, Q. et al. Correlation between particle size/domain structure andmagnetic properties of highly crystalline Fe3O4 nanoparticles. Sci.Rep. 7, 9894 (2017).59. Cao, D. et al. High saturation magnetization of γ-Fe2O3 nano-particles by a facile one-step synthesis approach. Sci. Rep. 6,32360 (2016).60. Omelyanchik, A. et al. High-quality α-Fe nanoparticles synthesizedby the electric explosion of wires. J. Magn. Magn. Mater. 484,196–200 (2019).61. Miao, B. F., Huang, S. Y., Qu, D. &Chien, C. L. Absenceof anomalousNernst effect in spin Seebeck effect of Pt/YIG. AIP Adv. 6,015018 (2016).62. Geprägs, S. et al. Static magnetic proximity effects and spin Hallmagnetoresistance in Pt/Y3Fe5O12 and inverted Y3Fe5O12/Pt bilay-ers. Phys. Rev. B 102, 214438 (2020).63. Geprägs, S. et al. Investigation of induced Pt magnetic polarizationin Pt/Y3Fe5O12 bilayers. Appl. Phys. Lett. 101, 262407 (2012).64. Uchida, K. et al. Longitudinal spin Seebeck effect: From funda-mentals to applications. J. Phys. Condens.Matter26, 389601 (2014).65. Solis, L. M. et al. Strain-induced enhancement of spin pumping inPt/YIG bilayers. J. Phys. D Appl. Phys. 59, 025302 (2026).66. Lebrun, R. et al. Tunable long-distance spin transport in a crystallineantiferromagnetic iron oxide. Nature 561, 222–225 (2018).67. Das, S. et al. Anisotropic long-range spin transport in cantedantiferromagnetic orthoferrite YFeO3. Nat. Commun. 13,6140 (2022).68. Wesenberg, D., Liu, T., Balzar, D., Wu, M. & Zink, B. L. Long-distancespin transport in a disordered magnetic insulator. Nat. Phys. 13,987–993 (2017).69. Uchida, K., Kikkawa, T., Miura, A., Shiomi, J. & Saitoh, E. Quantitativetemperature dependence of longitudinal spin Seebeck effect athigh temperatures. Phys. Rev. X 4, 041023 (2014).70. Lee, W. Y. et al. Role of ferromagneticmonolayerWSe2 flakes in thePt/Y3Fe5O12 bilayer structure in the longitudinal spin Seebeckeffect. ACS Appl. Mater. Interfaces 13, 15783–15790 (2021).71. Kalappattil, V. et al. Giant spin Seebeck effect through an interfaceorganic semiconductor. Mater. Horiz. 7, 1413–1420 (2020).72. Khang, N. H. D., Ueda, Y. & Hai, P. N. A conductive topologicalinsulator with large spin Hall effect for ultralow power spin–orbittorque switching. Nat. Mater. 17, 808–813 (2018).73. Şahin, C. & Flatté, M. E. Tunable giant spin Hall conductivities in astrong spin-orbit semimetal: Bi1-xSbx. Phys. Rev. Lett. 114,107201 (2015).74. Zhou, J., Qiao, J., Bournel, A. & Zhao, W. Intrinsic spin Hall con-ductivity of the semimetals MoTe2 and WTe2. Phys. Rev. B 99,060408(R) (2019).75. Li, X. et al. Large and robust charge-to-spin conversion in sputteredconductive WTex with disorder. Matter 4, 1639–1653 (2021).76. Dc, M. et al. Room-temperature high spin–orbit torque due toquantum confinement in sputtered BixSe(1–x) films. Nat. Mater. 17,800–807 (2018).77. Shi, S. et al. All-electric magnetization switching andDzyaloshinskii–Moriya interaction in WTe2/ferromagnet hetero-structures. Nat. Nanotechnol. 14, 945–949 (2019).78. Xiu, F. et al. Manipulating surface states in topological insulatornanoribbons. Nat. Nanotechnol. 6, 216–221 (2011).79. Dufouleur, J. et al. Enhancedmobility of spin-helical Dirac fermionsin disordered 3D topological insulators. Nano Lett. 16,6733–6737 (2016).80. Hoque, M. A., Sjöström, L., Khokhriakov, D., Zhao, B. & Dash, S. P.Room temperature nonlocal detection of charge-spin interconver-sion in a topological insulator. NPJ 2D Mater. Appl. 8, 10 (2024).81. Mizuguchi, M. & Nakatsuji, S. Energy-harvesting materials based onthe anomalous Nernst effect. Sci. Technol. Adv. Mater. 20,262–275 (2019).82. Sakai, A. et al. Iron-based binary ferromagnets for transverse ther-moelectric conversion. Nature 581, 53–57 (2020).83. Chen, T. et al. Large anomalous Nernst effect and nodal plane in aniron-based kagome ferromagnet. Sci. Adv. 8, eabk1480 (2022).84. Pan, Y. et al. Giant anomalous Nernst signal in the antiferromagnetYbMnBi2. Nat. Mater. 21, 203–209 (2022).85. Asaba, T. et al. Colossal anomalous Nernst effect in a correlatednoncentrosymmetric kagome ferromagnet. Sci. Adv. 7,eabf1467 (2021).86. Wen, J. et al. UltrahighanomalousNernst thermopower and thermalHall angle in YbMnBi2. Commun. Mater. 6, 138 (2025).87. Park, S. J. et al. Enhanced transverse electron transport viadisordered composite formation. Phys. Rev. Lett. 136,056301 (2026).88. Gautam, R. et al. Creation of flexible spin-caloritronic material withgiant transverse thermoelectric conversion by nanostructureengineering. Nat. Commun. 15, 2184 (2024).89. Park, S. J. et al. High-throughput development of flexible amor-phous materials showing large anomalous Nernst effect via auto-matic annealing and thermoelectric imaging. Mater. Today 92,416–424 (2026).90. Wu, H. et al. Separation of inverse spin Hall effect and anomalousNernst effect in ferromagnetic metals. J. Magn. Magn. Mater. 441,149–153 (2017).91. Iguchi, R., Uchida, K., Daimon, S. & Saitoh, E. Concomitantenhancement of the longitudinal spin Seebeck effect and thethermal conductivity in a Pt/YIG/Pt system at low temperatures.Phys. Rev. B 95, 174401 (2017).Article https://doi.org/10.1038/s41467-026-75232-0Nature Communications |         (2026) 17:6389 9www.nature.com/naturecommunications92. Park, S. J., Bang, K. M., Park, J. & Jin, H. High heat-flux sensitivity ofthe planar coil device based on the anomalous Nernst effect. Appl.Therm. Eng. 265, 125555 (2025).AcknowledgementsWe thank W. Zhou for helpful discussions and experimental supportduring the early stage of this work. This work was supported by ERATO“Magnetic ThermalManagementMaterials” (grant no. JPMJER2201) fromJST, Japan.Author contributionsS.J.P. fabricated the Pt-coated YIG powders, measured and analyzed theVSM, XRD, electrical and Hall conductivities, thermal diffusivity, DSCand SSE data. K.H. synthesized the bulk composites at low-temperatureand high-pressure conditions. H.S.A. performed the SEM and TEMmeasurements. F.A. and T.H. assisted with the SSE and thermal diffu-sivity measurements, respectively. K.U. conceived and supervised theproject. S.J.P. wrote the manuscript with input from all authors.Competing interestsThe authors declare no competing interests.Additional informationSupplementary information The online version containssupplementary material available athttps://doi.org/10.1038/s41467-026-75232-0.Correspondence and requests for materials should be addressed toSang J. Park or Ken-ichi Uchida.Peer review information Nature Communications thanks Guoying Gaoand the other anonymous reviewer(s) for their contribution to the peerreview of this work. A peer review file is available.Reprints and permissions information is available athttp://www.nature.com/reprintsPublisher’s note Springer Nature remains neutral with regardto jurisdictional claims in published maps and institutionalaffiliations.Open Access This article is licensed under a Creative CommonsAttribution 4.0 International License, which permits use, sharing,adaptation, distribution and reproduction in any medium or format, aslong as you give appropriate credit to the original author(s) and thesource, provide a link to the Creative Commons licence, and indicate ifchanges were made. The images or other third party material in thisarticle are included in the article's Creative Commons licence, unlessindicated otherwise in a credit line to the material. If material is notincluded in the article's Creative Commons licence and your intendeduse is not permitted by statutory regulation or exceeds the permitteduse, you will need to obtain permission directly from the copyrightholder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.© The Author(s) 2026Article https://doi.org/10.1038/s41467-026-75232-0Nature Communications |         (2026) 17:6389 10https://doi.org/10.1038/s41467-026-75232-0http://www.nature.com/reprintshttp://creativecommons.org/licenses/by/4.0/http://creativecommons.org/licenses/by/4.0/www.nature.com/naturecommunications Trans-scale spin Seebeck effect in nanostructured bulk composites based on magnetic insulator Results Fabrication of YIG-Pt composite materials Electrical and thermal conductivity of nanostructured bulk composites Observation of trans-scale SSE in nanostructured bulk composites Validation of trans-scale SSE and exclusion of parasitic effects Thickness-dependent power scaling enabled by volumetric spin conversion Discussion Methods Sample preparation Structural characterization Measurement of thermal and electrical conductivities Measurement of the spin Seebeck effect Magnetic property analysis Measurement of the Hall effect Data availability References Acknowledgements Author contributions Competing interests Additional information