betaFeedback

WorksPhysics and Astronomy

Hidden interfacial magnetism in RuO2_2 heterostructures revealed by a superconducting spin switch

Abstract

RuO2_2, once the flagship altermagnetic candidate, is now believed to be nonmagnetic in the bulk, yet weak, static and spatially inhomogeneous magnetism emerges at its surfaces and interfaces. Where this magnetism resides, what stabilises it, and whether it is connected to these states remain open questions. Here we address these questions at the Nb/RuO2_2 interface using a superconducting spin switch. A GdN/Nb/RuO2_2 trilayer allows interfacial exchange-field alignment to tune the superconducting transition temperature of Nb, probing interfacial magnetism inaccessible to conventional magnetometry. The switch reveals a hidden magnetic subsystem, which we attribute to a thin RuOx_x redox layer formed as Nb scavenges oxygen from RuO2_2. This layer behaves as a weak and magnetically hard ferromagnet with vacancy-induced anisotropy. Furthermore, supercurrent amplifies the switching efficiency, making the probe sensitivity tunable in situ. Superconducting spin switches thus emerge as a quantitative low-temperature probe of hidden magnetism at buried interfaces.

Few materials have had as turbulent a magnetic biography as RuO2\mathrm{RuO_2}. For decades, bulk RuO2\mathrm{RuO_2} was regarded as a Pauli paramagnetic metal[1, 2], before evidence for itinerant antiferromagnetism emerged[3, 4]. RuO2\mathrm{RuO_2} subsequently became the prototypical dd-wave altermagnet, predicted to exhibit anisotropic momentum-dependent spin splitting whose magnitude and sign depend on crystal orientation and Néel vector direction despite the absence of net magnetisation[5]. This picture was supported by resonant X-ray[6], photoemission[7] and transport signatures[8, 9, 10, 11]. However, it has been progressively challenged by null results, particularly in the bulk: muon spin rotation (μ\muSR)[12, 13] and neutron diffraction[14] find no static magnetic order, while angle-resolved photoemission[15, 16] and quantum oscillations[17, 18] reveal a Fermi surface consistent with a nonmagnetic electronic structure. Thin-films studies have likewise reported no magnetic order[19] and transport signatures consistent with a non-altermagnetic origin[20], while similar spin-torque responses in nonmagnetic IrO2\mathrm{IrO_2} undermined the torque-based evidence for altermagnetism[21]. However, magnetism has not disappeared entirely. Low-energy μ\muSR on thin-films reveals weak, static, and spatially inhomogeneous magnetism localised near surfaces[22], strained films exhibit enhanced magnetic responses[23], and interfacial or defect-related moments continue to emerge in heterostructures[24, 25, 26]. Intriguingly, uniquely-strained RuO2\mathrm{RuO_2} films also exhibit superconductivity[27, 28, 29], underscoring how profoundly the electronic ground state can be reconstructed away from the bulk limit. The emerging picture is therefore that magnetism in RuO2\mathrm{RuO_2} is not an intrinsic bulk band property, but an emergent local phenomenon highly sensitive to surfaces, strains, interfaces and stoichiometry[30]. What has been missing is an experimental probe capable of detecting and quantifying such weak magnetism and a microscopic mechanism explaining its most puzzling combination of properties: a moment too small for standard magnetometry, and magnetic order hard enough to survive as a stable, switchable reference.

Here we supply both. We use a superconducting spin switch in which a superconductor is sandwiched between two magnetic layers (M1/S/M2\mathrm{M_1/S/M_2})[31]. When the superconducting thickness is less than its coherence length ξ0\xi_0, the interfacial exchange fields from M1\mathrm{M_1} and M2\mathrm{M_2} add or subtract for parallel (P) or antiparallel (AP) magnetisations, giving distinct superconducting transition temperatures Tc(P)T_\mathrm{c}(\mathrm{P}) and Tc(AP)T_\mathrm{c}(\mathrm{AP}), with ΔTc=Tc(AP)−Tc(P)\Delta T_\mathrm{c}=T_\mathrm{c}(\mathrm{AP})-T_\mathrm{c}(\mathrm{P})[32, 33, 34, 35, 36, 37, 38, 39, 40]. Comparing TcT_\mathrm{c} between these configurations provides a sensitive differential probe of interfacial exchange fields difficult to isolate in S/M bilayers, particularly in the presence of magnetic domains[41, 42, 43]. Using GdN/Nb/RuO2\mathrm{RuO_2}(110) trilayers, we observe reversible TcT_\mathrm{c} switching independent of the crystallographic direction of field-cooling (FC) and stable against applied fields of several hundred millitesla. These observations are inconsistent with a bulk altermagnetic origin of the switching. Control experiments localise the magnetic subsystem responsible for the switching at the Nb/RuO2\mathrm{RuO_2} interface. We propose that its magnetic layer originates from a RuOx\mathrm{RuO_x} redox layer that necessarily forms as Nb scavenges oxygen from RuO2\mathrm{RuO_2}. Theoretical analysis of the redox chemistry suggests that this redox layer is a thin ferromagnet with a small magnetic moment. The resulting weak Zeeman coupling implies that large fields are required for magnetisation reversal. The local anisotropy axes are randomly oriented hard axes, as in amorphous magnets[44], because the oxygen vacancies that create the moments also determine their spin–orbit anisotropy. Their preferential alignment towards the interface normal produces a coherent easy plane, keeping the moments in-plane and making the switching insensitive to crystallographic axes. Meanwhile, the random component is expected to pin the order, and to favour its breakup into in-plane domains beyond a characteristic length L∗L^{*} because random anisotropy defeats exchange on large scale[45, 46].

This layer is therefore weak in moment and hard in orientation, invisible in magnetometry yet stable enough to serve as a magnetic reference for the switch. The magnetisation of the RuOx\mathrm{RuO}_{x} layer induces an effective exchange field at the Nb/RuOx\mathrm{RuO}_{x} interface. We model the superconducting spin switch with the quasiclassical Usadel equation[47, 48, 49], which allows us to extract the effective interfacial exchange field, and accounts for the increase of ∣ΔTc∣|\Delta T_{\mathrm{c}}| with applied supercurrent. Together with zero-field-cooling (ZFC) measurements, this establishes the spin switch as a sensitive transport probe of buried interfacial magnetic order.

Superconducting spin switch

A AlN(10 nm)/GdN(10 nm)/Nb(9.3 nm)/RuO2(110)(10 nm)//TiO2(110)\mathrm{AlN}(10\ \mathrm{nm})/\mathrm{GdN}(10\ \mathrm{nm})/\mathrm{Nb}(9.3\ \mathrm{nm})/\mathrm{RuO}_{2}(110)(10\ \mathrm{nm})//\mathrm{TiO}_{2}(110) heterostructure was deposited by magnetron sputtering as shown in Figure 1a (see Methods). The epitaxial growth of RuO2\mathrm{RuO}_{2} was confirmed by reflection high-energy electron diffraction (RHEED) in-situ and X-ray diffraction (XRD) as shown in Figure 1b, Figure 1c. The thin-film was patterned into four-terminal bars of 10 μm×220 μm10\ \mu\mathrm{m} \times220\ \mu\mathrm{m} for electrical measurements as shown in Figure 1d. Figure 1e shows the magnetisation versus in-plane magnetic field, M(H)M(H), of the unpatterned film at 10 K. GdN is a ferromagnetic insulator with a Curie temperature of ∼40\sim40 K and an effective switching field of ∼6\sim6 mT at 10 K (see Supplementary S1), consistent with previous reports[39, 50]. Its magnetisation can therefore be reversed with millitesla fields while leaving the harder magnetic subsystem unchanged (see Supplementary S1 about M(T)M(T) and M(H)M(H) at different temperatures). No discernible difference was observed between the magnetisation curves after ±7\pm7 T FC from room temperature and ZFC, indicating that conventional magnetometry does not resolve the magnetic subsystem investigated below (see Supplementary S2 for the discussion about the magnetisation of Nb/RuO2\mathrm{RuO}_{2} bilayers). In some literature, the current and FC directions are aligned along the RuO2\mathrm{RuO}_{2} [1] axis, which has been proposed as the Néel vector direction in altermagnetic RuO2\mathrm{RuO}_{2}[51, 52]. In a dd-wave altermagnet, the anisotropic exchange field is expected to produce an angular dependence of ΔTc\Delta T_{\mathrm{c}} on the magnetisation direction at the opposite interface[53]. Here, however, [1] specifies only the experimental geometry and does not imply any magnetic significance. As we show below, the magnetically active subsystem resides in a RuOx_x redox layer at the Nb/RuO2_2 interface and therefore does not inherit the hypothetical Néel vector orientation of the rutile RuO2_2 lattice.

Figure 1. Structural, magnetic and device characterisation of the AlN(10 nm)/GdN(10 nm)/Nb(9.3 nm)/RuO₂(110)(10 nm)//TiO₂(110) superconducting spin switch. a, Cross-sectional schematic of the AlN/GdN/Nb/RuO₂ heterostructure. The GdN layer and the RuOₓ redox layer provide the two magnetic subsystems, with their magnetisations indicated by the red and blue arrows,

The switch is operated as follows. The device is field-cooled from room temperature in ±9\pm9 T along [1], setting the hard magnetic subsystem as shown in Figure 2a,Figure 2b,Figure 2d,Figure 2e; alternatively, it is zero-field-cooled after oscillatory demagnetisation at room temperature in Figure 2c,Figure 2f. At a few kelvins, the GdN magnetisation is saturated at a preset-field μ0Hset=±50\mu_0 H_{\mathrm{set}}=\pm50 mT along [1] and the field is returned to zero, defining the “±0\pm0 mT” states. The superconducting transition is then recorded upon warming at a current density of 5.38×1085.38 \times10^8 A/m2^2, with the current applied along [1]. After FC at +9+9 T, TcT_{\mathrm{c}} splits between GdN magnetisation along the positive field (along the FC direction, P) and negative field (opposite to the FC direction, AP) directions, giving ΔTc=Tc(AP)−Tc(P)=−8\Delta T_{\mathrm{c}}=T_{\mathrm{c}}(\mathrm{AP})-T_{\mathrm{c}}(\mathrm{P})=-8 mK as shown in Figure 2d. Reversing the FC direction to −9-9 T reverses the TcT_{\mathrm{c}} switching with respect to the external field direction while preserving the relative response, ΔTc=−8\Delta T_{\mathrm{c}}=-8 mK in Figure 2e. After ZFC, by contrast, ΔTc\Delta T_{\mathrm{c}} vanishes as shown in Figure 2f. Similar switching is observed with μ0H=±50\mu_0H=\pm50 mT (see Supplementary S3).

Normalised resistance transitions for different magnetic configurations

Figure 2. Tc_c of the switch for each magnetic configuration. All measurements in zero applied field after saturating GdN at ±50 mT. a-c, Normalised R(T) after FC at +9 T, -9 T, and ZFC. d-f, Enlarged transition regions. After FC the two GdN polarities give distinct Tc_c, with sign set by the FC direction; after ZFC the difference vanishes. The insets sketch the relative orientation of the redox-layer magnetisation m (prepared by FC) and the GdN magnetisation.

To identify the magnetic switching field associated with the RuOx_x redox layer, we measured the field dependence of the resistance of the AlN(10 nm)/GdN(10 nm)/Nb(9.3 nm)/RuO2_2(110)(10 nm)//TiO2_2(110) sample measured at 3.1 K, within the superconducting transition, as shown in Figure 3a. Apart from a background associated with the suppression of superconductivity by the magnetic field, pronounced hysteretic anomalies appear around ±400\pm400 to ±500\pm500 mT. In contrast, no corresponding anomaly is observed at 5 K, where the Nb is in the normal state. The disappearance of these features above indicates that they reflect a magnetic modulation of superconductivity rather than ordinary normal-state magnetoresistance. Similar resistance peaks localized near the coercive field have been reported in ferromagnet/superconductor hybrids[42]. This assignment is independently supported by the remanent superconducting response shown in Figure 3b. After field cooling at +9+9 T, the superconducting transition depends strongly on whether the sample is preset to −400-400 or −500-500 mT before returning to zero field. In other words, when the sample is preset to −400-400 mT, the sign of ΔTc\Delta T_{\mathrm{c}} shift is consistent with that in Figure 2a and Figure 2d, whereas presetting to −500-500 mT reverses the sign, making it consistent with Figure 2b and Figure 2e.

Switching field of the interfacial redox layer

Figure 3. Switching field of the interfacial redox layer. a Magnetic-field dependence of the resistance of the AlN(10 nm)/GdN(10 nm)/Nb(9.3 nm)/RuO2_2(110)(10 nm)//TiO2_2(110) sample measured at 3.1 K and at 5 K. The red and blue dots represent field sweeps from positive to negative field and from negative to positive field, respectively, at 3.1 K, while the green dots represent the field sweep measured at 5 K. b Zero-field superconducting transitions measured after field cooling at +9 T and subsequent presetting of the magnetic field. The red and blue curves correspond to measurements at +0 and -0 mT, respectively, after presetting to -400 mT, whereas the green and purple curves correspond to measurements at +0 and -0 mT, respectively, after presetting to -500 mT.

Moreover, magnetic-field-dependent measurements of TcT_{\mathrm{c}} on the AlN(10 nm)/GdN(10 nm)/Nb(9.3 nm)//SiO2_2/Si control sample revealed a positive sign for the exchange field induced by GdN within our model (see Supplementary S4), implying a negative sign for the exchange field induced by RuOx_x. The sign of the exchange field hexh_{\mathrm{ex}} coincides with that of the imaginary part of the spin-mixing conductance, Gi≃πG0NFhexdNbG_i \simeq\pi G_0 N_F h_{\mathrm{ex}} d_{\mathrm{Nb}}, where G0G_0 is the conductance quantum and NFN_F is the density of states per spin at the Fermi level[40, 54, 55]. Unlike conventional superconducting spin switches with two equivalent superconductor/magnet interfaces, where the absolute sign of the exchange field cannot be determined, our asymmetric structure allows this sign to be inferred[31].

This negative exchange field is consistent with the itinerant picture in which the t2gt_{2g} band of Ru3+Ru^{3+} (t2g5t_{2g}^{5}) is more than half filled, so that Nb electrons hybridize with the unoccupied minority-spin states. The same sign follows in the atomic limit, as antiferromagnetic exchange between Nb conduction electrons and the Ru3+Ru^{3+} spins (see Supplementary S11).

The origin of the subsystem is further confirmed by a series of control experiments. (i) Crystallographic orientation: the switching measurements were repeated with current, cooling-field and preset-field applied along [11ˉ0][1\bar{1}0] and along the in-plane direction midway between [001][001] and [11ˉ0][1\bar{1}0], i.e. 45∘45^\circ from both axes, yielding comparable ΔTc\Delta T_{\mathrm{c}} values (Supplementary S5). The switching is therefore insensitive to crystallographic orientation, arguing against an origin associated with the proposed altermagnetic RuO2RuO_{2}. (ii) Orthogonal geometry: with the cooling-field along [001][001] but the preset-field used to switch GdN rotated by 90∘90^\circ, ΔTc\Delta T_{\mathrm{c}} vanishes, as expected for a superconducting switch controlled by the relative angle between two in-plane magnetisations[31, 56] (Supplementary S6). (iii) Nb thickness dependence: increasing the Nb thickness to 1212 nm, above ξ0\xi_{0} (∼11.35\sim11.35 nm), also suppresses ΔTc\Delta T_{\mathrm{c}} to zero. The switching therefore requires the exchange perturbations from the two interfaces to overlap coherently across the Nb layer, as expected for a genuine two-interface exchange field effect[31, 32] (see Supplementary S7 for ξ0\xi_{0} and Supplementary S8 for the switching of the 12-nm-thick Nb device). This thickness dependence rules out alternative mechanisms based on stray fields[57, 58] or spin-polarised quasiparticles[59, 60], which do not require dNb<ξ0d_{\mathrm{Nb}} < \xi_{0}. (iv) No RuO2RuO_{2}: an AlN(10 nm)/GdN(10 nm)/Nb(9.3 nm)//SiO2/SiAlN(10\ \mathrm{nm})/GdN(10\ \mathrm{nm})/Nb(9.3\ \mathrm{nm})//SiO_{2}/Si control sample shows no dependence of TcT_{\mathrm{c}} on the GdN magnetisation direction and, decisively, the same ±9\pm9 T FC protocol produces no evidence of a hard magnetic subsystem. This excludes the GdN/Nb interface as its origin (see Supplementary S9). (v) No GdN: an AlN(10 nm)/Nb(9.3 nm)/RuO2(110)(10 nm)//TiO2(110)AlN(10\ \mathrm{nm})/Nb(9.3\ \mathrm{nm})/RuO_{2}(110)(10\ \mathrm{nm})//TiO_{2}(110) control sample shows no measurable TcT_{\mathrm{c}} response upon reversing the cooling-field direction (see Supplementary S9). (vi) RuO2RuO_{2} thickness dependence: increasing the RuO2RuO_{2} thickness to 5050 nm, beyond the relevant proximity length in RuO2RuO_{2}, preserves the switching polarity observed in the 10-nm-thick RuO2RuO_{2} sample (see Supplementary S10). For a 50-nm-thick RuO2RuO_{2} layer, the TiO2/RuO2TiO_{2}/RuO_{2} interface is far beyond the relevant superconducting proximity length from the Nb layer. Therefore, Cooper pairs are not expected to reach the TiO2/RuO2TiO_{2}/RuO_{2} interface and the persistence of switching in the thicker RuO2RuO_{2} sample supports an origin near the Nb/RuO2RuO_{2} interface. Thus, the hard magnetic subsystem at the Nb/RuO2RuO_{2} interface cannot be independently resolved through TcT_{\mathrm{c}} in the bilayer geometry; GdN provides the second, switchable exchange field required for its differential detection. Together, these controls localise the hard magnetic subsystem to the Nb/RuO2RuO_{2} interface and establish the superconducting spin switch as its readout.

Together, these observations reveal an unusual combination of magnetism. The interfacial magnetic state retains the orientation set during FC and remains stable against external fields up to 400 mT. Despite this large coercive field scale, magnetometry of the unpatterned film stack detects no corresponding remanence and coercive field signal within experimental resolution (see Supplementary S2). The subsystem therefore combines a small net moment with hysteresis and rigidity. The origin of this weak and hard subsystem is addressed in the next section.

A weak and magnetically hard magnet at the Nb/RuO2_2 redox interface

The Nb/RuO2_2 interface is not chemically inert. Per mole of oxygen, the formation enthalpies of the niobium oxides exceed that of RuO2_2 by more than a factor of two[61, 62, 63], providing a strong thermodynamic drive for in-situ-deposited Nb to scavenge oxygen from the RuO2_2 surface. Reduction of Ru4+^{4+} to Ru3+^{3+} at the Nb/RuO2_2 interface, combined with oxygen diffusion at room temperature, therefore can produce a thin self-limited Ru3+^{3+}-rich layer at the interface. The oxygen vacancies play a dual role in generating magnetism. First, reduction creates Ru3+^{3+} sites with one hole in the t2gt_{2g} multiplet (t2g5t^{5}_{2g}), which in the atomic limit forms a Kramers doublet with a robust S=12S = \frac{1}{2} moment, while the loss of Ru-O bonds narrows the magnetically active planar dd-band. The Stoner criterion, not satisfied in stoichiometric RuO2_2, can therefore be reached in the reduced interfacial layer, which orders as an itinerant ferromagnet. Second, the same vacancies generate strong local symmetry breaking. Through spin–orbit coupling, each reduced Ru site acquires a magnetic hard axis associated with its local vacancy environment. These axes are random from site to site, but their ensemble is combed towards the interface normal by the vacancy-concentration gradient. The redox reaction therefore can create both the magnetic order and the random anisotropy governing its macroscopic behaviour.

The resulting state is a thin ferromagnet with random anisotropy, of the type studied in amorphous magnets[44, 45, 46]. The preferential alignment of the local hard axes towards the interface normal produces a coherent easy plane, whereas the random component competes with the exchange stiffness. In this picture, the energy gained by following the random axes grows faster with length scale than the exchange energy cost, leading to the predicted formation of in-plane domains beyond the Imry–Ma length L∗L^{*}[45, 46] (see Supplementary S11). Independently of the domain structure, the small magnetisation results in weak Zeeman coupling to an external field. Therefore, magnetisation reversal, whether by coherent rotation or by depinning of domain walls from the random-anisotropy landscape[64], requires large fields even for moderate anisotropy energy. The same oxygen vacancies that enable the Stoner instability through band narrowing also pin the resulting order through random anisotropy. This naturally produces the experimentally observed weak and rigid state—a magnet with an almost undetectable net moment, yet an orientation stable against large magnetic fields.

Interfacial exchange fields and current-enhanced ΔTc\Delta T_{\mathrm{c}}

For dNb<ξ0d_{\mathrm{Nb}} < \xi_0, one can assume that the superconducting order parameter is uniform across the Nb film, and the two magnetic interfaces can be represented by thickness-averaged effective exchange fields whose combination depends on the relative magnetisation alignment, heff,P(AP)=hGdN±hRuOxh_{\mathrm{eff,P(AP)}} = h_{\mathrm{GdN}} \pm h_{\mathrm{RuOx}}. The parameter hGdN/RuOxh_{\mathrm{GdN/RuOx}} is the effective exchange field averaged over the superconducting layer, induced by each magnetic interface. The critical temperatures of the diffusive Nb film in the respective magnetic configurations, in the presence of a supercurrent, then follow from the self-consistently linearised Usadel theory[48, 65, 66]. Using hGdN=0.13h_{\mathrm{GdN}} = 0.13 meV estimated from Ref. 37, we obtain hRuOxh_{\mathrm{RuOx}} to be −0.004-0.004 meV (see Supplementary S12 about the calculation details). This highlights the superconducting switch as a sensitive probe of interfacial magnetism, capable of detecting an exchange field more than 30 times weaker than that of GdN.

We next examine the current dependence of ΔTc\Delta T_{\mathrm{c}}. Increasing the applied current from 0.05 mA (5.38×1085.38 \times10^{8} A/m2^{2}) to 0.225 mA (2.42×1092.42 \times10^{9} A/m2^{2}) enhances ΔTc\Delta T_{\mathrm{c}} after FC ±9\pm9 T by approximately a factor of two as shown in Figure 4a. In contrast, after ZFC, ΔTc\Delta T_{\mathrm{c}} remains small over entire current range. The quasiclassical model qualitatively reproduces this enhancement and reveals physics beyond the Ginzburg–Landau description. Expanding the digamma function to leading order gives additive contributions from current-induced depairing and exchange pair-breaking. The quasiclassical expression, however, retains higher-order cross terms between the two pair-breaking mechanisms, allowing depairing to enhance the sensitivity of TcT_{\mathrm{c}} to the interfacial exchange field. Figure 4b shows the theoretical calculation of ΔTc\Delta T_{\mathrm{c}} depending on the qξ0q\xi_0. qq is expressed with respect to ξ0\xi_0 and characterises the depairing induced by the supercurrent. Here, ξ0=ℏD/Δ0\xi_0 = \sqrt{\hbar D/\Delta_0}, where DD is the diffusion constant in Nb and Δ0\Delta_0 is the zero-temperature superconducting gap of the reference film in the absence of pair-breaking. The model reveals that increasing depairing enhances ∣ΔTc∣|\Delta T_{\mathrm{c}}|. This behavior arises because TcP>TcAPT_{\mathrm{c}}^{\mathrm{P}} > T_{\mathrm{c}}^{\mathrm{AP}}, such that depairing suppresses TcAPT_{\mathrm{c}}^{\mathrm{AP}} more strongly than TcPT_{\mathrm{c}}^{\mathrm{P}}, thereby increasing ∣ΔTc∣|\Delta T_{\mathrm{c}}|. Establishing a quantitative relation between qq and the total current is beyond the scope of this work, owing to the likely inhomogeneity of the current distribution along the 10 μ\mum dimension of the superconducting region[67] (see Supplementary S12 about further details).

Figure 4. Supercurrent dependence of the difference of ∣Tc∣|T_{\mathrm{c}}|. a, Supercurrent dependence of ∣ΔTc∣|\Delta T_{\mathrm{c}}| under ±0\pm0 mT. The red, blue, and green dots represent the results obtained after FC at +9 T, -9 T, and ZFC, respectively. b, Calculated dependence of ∣ΔTc∣|\Delta T_{\mathrm{c}}| on qξ0q\xi_0, which characterises the

Conclusions and outlook

In summary, we have used a superconducting spin switch to reveal hidden magnetism at the Nb/RuO2\mathrm{RuO_2} interface, which we attribute to a weak and magnetically hard ferromagnetic state in an interfacial RuOx\mathrm{RuO_x} redox layer. The switching provides a sensitive measure of the relative interfacial exchange fields, with its response amplified by supercurrent through the super-additive interplay between exchange fields and current-induced pair-breaking. The magnetic subsystem detected by the switch is not a bulk property of RuO2\mathrm{RuO_2}, consistent with the null results of bulk probes[12, 14, 15, 17], but is associated with an interfacial redox layer whose magnetism is both enabled and pinned by oxygen vacancies. This mechanism accounts for the seemingly contradictory properties that have made RuO2\mathrm{RuO_2} interfacial magnetism so elusive: a weak magnetic moment, undetectable by conventional magnetometry, yet static on μ\muSR timescales[22], spatially inhomogeneous, and remarkably hard against magnetic-field reversal. The resulting picture is of a magnet that is weak in moment and hard in orientation. Although Nb promotes oxygen removal in the present heterostructure, a similar vacancy-driven mechanism is expected to operate in other oxygen-deficient RuO2\mathrm{RuO_2} systems, including commonly studied RuO2\mathrm{RuO_2}/Py heterostructures where oxygen vacancies are present[25]. For magnetism, if its random-anisotropy character is confirmed, it may provide a solid-state realisation of a generalised Imry-Ma model whose phase diagram—including domain morphology, wall chirality, and glassiness, tunable by the redox depth—remains open. More broadly, the interfacial system realised here provides a largely unexplored setting in which random-anisotropy magnetism, interfacial spin–orbit coupling, and ss-wave superconductivity coexist at atomic proximity.

The inverse switching polarity further identifies a negative hex(Gi)h_{\mathrm{ex}}(G_i) at the Nb/RuOx\mathrm{RuO_x} interface. Established spin switches based on two layers of the same magnets have primarily exploited their magnetic configurations. Here, combining a known magnetic reference with an interface of unknown magnetic character enables the switch to probe the sign and magnitude of interfacial exchange itself. For superconductivity, this establishes the switch, operated at finite current, as a quantitative probe of interfacial magnetism invisible to conventional magnetometry, with a sensitivity that is itself current-tunable. This enhanced response may extend the accessible range of weak interfacial magnetism beyond what can readily be resolved by heavy-metal-based spin Hall magnetoresistance. We expect this probe to be transferable to the growing family of compensated and weakly magnetic materials whose interfaces, rather than their bulk, host the magnetism.

The microscopic origin of the inverse superconducting switching polarity observed here remains an important subject for future research. Comparative studies using materials with established negative spin polarisation, such as FeCr[68, 69], could help clarify the relationship between spin polarisation, interfacial exchange and switching polarity. Further experiments, including depth-resolved low-energy μ\muSR and spatially resolved XMCD-PEEM, could characterise the depth distribution and lateral inhomogeneity of the interfacial magnetism. Comparing the resulting magnetic correlation lengths with predictions of the random-anisotropy model would test whether the layer indeed reaches the Imry–Ma regime.

Methods

Film growth and device fabrication

Thin-films are deposited onto pre-cleaned 10 mm×10 mm10\ \mathrm{mm} \times10\ \mathrm{mm} TiO2(110)\mathrm{TiO_2}(110) substrates using an Eiko ES-250S rf magnetron sputtering with a base pressure better than 6×10−6 Pa6 \times10^{-6}\ \mathrm{Pa}. The 1010-nm-thick RuO2(110)\mathrm{RuO_2}(110) films are grown epitaxially on TiO2(110)\mathrm{TiO_2}(110) by reactive sputtering at a substrate temperature of 400∘C400^\circ\mathrm{C}, with Ar and O2\mathrm{O_2} partial pressure of 0.24 Pa0.24\ \mathrm{Pa} and 0.06 Pa0.06\ \mathrm{Pa}, respectively. Nb is deposited in Ar at a pressure of 0.5 Pa0.5\ \mathrm{Pa}, while GdN is grown by reactive sputtering from a Gd target with Ar and N2\mathrm{N_2} partial pressures of 0.46 Pa0.46\ \mathrm{Pa} and 0.04 Pa0.04\ \mathrm{Pa}, respectively. Finally, a 1010-nm-thick AlN capping layer is grown by reactive sputtering from an Al target with Ar and N2\mathrm{N_2} partial pressures of 0.46 Pa0.46\ \mathrm{Pa} and 0.04 Pa0.04\ \mathrm{Pa}, respectively, to protect the structure.

Magnetic measurements

Magnetic moment versus magnetic field measurements are performed using a Quantum Design Magnetic Property Measurement System (MPMS) equipped with a vibrating sample magnetometer superconducting quantum interference device (SQUID). Magnetic fields of up to 7 T7\ \mathrm{T} were applied using the system’s superconducting magnet.

Transport measurements

Low-temperature current-voltage I(V)I(V) measurements are performed in a four-terminal electrical setup using Au wires ultrasonically bonded to 8080-nm-thick Au contact pads deposited on the devices. Measurements above 1.8 K1.8\ \mathrm{K} are carried out in a Quantum Design Physical Property Measurement System (PPMS) with an in-plane magnetic field. Measurements below 1.8 K1.8\ \mathrm{K} are performed in the same system using the He-3 option. TcT_c is defined at half the normal-state resistance unless otherwise noted.

Competing interests

The authors declare no competing interests.

respectively. Because the Nb thickness, 9.3 nm, is smaller than the superconducting coherence length, Cooper pairs in Nb can coherently experience the exchange fields from both the GdN/Nb and Nb/RuOx_x interfaces. b, RHEED image of the RuO2_2(110) layer with the electron beam along the [1] and [11ˉ\bar{1}0] directions. c, XRD image of the heterostructure whereby the x-ray source is CuKα λ=1.54Å). d, Device structure and measurement configuration. e, Magnetisation as a function of in-plane magnetic field measured at 10 K, showing the magnetic hysteresis of the GdN layer. The magnetic field was applied along the [1] direction, which was also the field-cooling direction. Red, blue, and green solid lines show the magnetisation of the unprocessed film after field-cooling ±7 T and ZFC, respectively.

depairing induced by the supercurrent. The result was calculated using Eq. (5) with the estimated values of hRuOxh_{\mathrm{RuO}_{x}}, hGdNh_{\mathrm{GdN}}, τsf\tau_{sf}, and τso\tau_{so}.

Hidden interfacial magnetism in RuO2_2 heterostructures revealed by a superconducting spin switch

Tetsuma Mandokoro; Hisakazu Matsuki; Shutaro Karube; Rodrigo de las Heras; Andrei Mazanik; Akihiro Yamada; Fugo Tokoro; Susumu Hirata; Ryusuke Hisatomi; Yoichi Shiota; Vitaly N. Golovach; F. Sebastian Bergeret; Teruo Ono

Temperature dependence of the GdN magnetisation

Figure S1 shows the temperature dependence of the magnetisation for the GdN magnetisation in AlN(10 nm)/GdN(10 nm)/Nb(9.3 nm)/RuO2(10 nm)//TiO2(110)\mathrm{AlN}(10\ \mathrm{nm})/\mathrm{GdN}(10\ \mathrm{nm})/\mathrm{Nb}(9.3\ \mathrm{nm})/\mathrm{RuO_2}(10\ \mathrm{nm})//\mathrm{TiO_2}(110) heterostructures. The Curie temperature of GdN is approximately 40 K. Figure S1 also shows the magnetic field dependence of magnetisation at 2 K, 4 K, and 6 K. Although the ferromagnetism of GdN strongly depends on the temperature as shown Figure S1, both the coercive field and remanence magnetisation remain nearly unchanged within the temperature range investigated in this study. These results are consistent with the properties of GdN reported in previous studies. [S1, 2].

Temperature and magnetic field dependence of GdN magnetisation

Figure S1. (a) Temperature dependence of GdN magnetisation. (b) Magnetic field dependence of GdN. The red, blue, and green solid lines indicate the data measured at 2, 4, and 6 K, respectively.

Magnetometry of the AlN/Nb/RuO2_2//TiO2_2 and variability of the TiO2_2 substrate-background

To examine whether the magnetisation of RuOx_x redox layer inferred from the superconducting transport measurements is accompanied by a detectable macroscopic magnetic moment, we performed magnetic-field-dependent magnetisation measurements by SQUID. Figure S2 compares the magnetic responses of AlN(10 nm)/GdN(10 nm)/Nb(9.3 nm)/RuO2_2(10 nm)//TiO2_2 and AlN(10 nm)/Nb(9.3 nm)/RuO2_2(10 nm)//TiO2_2 samples. Although the contribution from the GdN layer is clearly visible in the AlN/GdN/Nb/RuO2_2//TiO2_2, almost no magnetic component attributable to the AlN/Nb/RuO2_2//TiO2_2 sample can be resolved. Figure S2 shows the magnetic-field dependence of the AlN/Nb/RuO2_2//TiO2_2 sample alone, again revealing no clear macroscopic hysteresis corresponding to the switching-field scale identified by superconducting transport. Here, the background linear components have been subtracted from the data. A quantitative extraction of only RuOx_x magnetic contribution is complicated by the magnetic background of the TiO2_2 substrate. As shown in Figure S2, the M−HM-H response of the same TiO2_2 substrate changes appreciably between measurements performed two days after baking and one week after baking. This thermal-history and time dependence prevents a reliable subtraction of the substrate contribution at the level required to isolate the weak magnetic signal from the RuOx_x redox layer. We therefore refrain from assigning an absolute magnetic moment or coercive field to the RuOx_x magnetic state from the SQUID measurements. Nevertheless, the absence of a coercivity hysteretic signal in the magnetometry indicates that the switching observed in superconducting transport is not associated with a readily detectable macroscopic ferromagnetic moment. The magnetic moment of the RuOx_x layer can remain below the sensitivity imposed by the substrate background even though its local magnetic configuration strongly affects the adjacent superconducting Nb layer.

Magnetic-field dependence of the magnetic moment per area for AlN/GdN/Nb/RuO2//TiO2 and AlN/Nb/RuO2//TiO2 samples

Figure S2. (a) Magnetic-field dependence of the magnetic moment per area for the AlN/GdN/Nb/RuO2_2//TiO2_2 (red) and AlN/Nb/RuO2_2//TiO2_2 (blue) samples. (b) Enlarged magnetic-field dependence of the AlN/Nb/RuO2_2//TiO2_2 sample. (c) M−HM-H curves of the same TiO2_2 substrate measured 2 days after baking (green) and one week after baking (purple).

Temperature dependence of the normalized Nb resistance under ±50\pm50 mT with current applied along [001]

Figure S3–Figure S3 show the temperature dependence of the normalized resistance for the Nb 9.3-nm-thick sample measured with a current applied along the [1] direction under an external magnetic field of ±50\pm50 mT. Figure S3–Figure S3 show enlarged views of the superconducting transition region. Similar to the results obtained at ±0\pm0 mT in the main text, the superconducting transition temperature (TcT_c) changes depending on the relative orientation between the exchange field from RuOx_x and the magnetisation of GdN. Moreover, the sign of this shift is reversed by reversing the field-cooling (FC) direction. In contrast, no discernible difference in the transition temperature is observed after zero-field cooling (ZFC). Compared with the ±0\pm0 mT results, the difference in TcT_c becomes smaller under the applied magnetic field presumably because the contribution from the RuOx_x exchange field is partially masked by the external magnetic field.

Temperature dependence of normalized resistance and enlarged superconducting transition regions

Figure S3. (a–c) Temperature dependence of the normalized resistance for the AlN\mathrm{AlN} (10 nm)/GdN\mathrm{GdN} (10 nm)/Nb\mathrm{Nb} (9.3 nm)/RuO2\mathrm{RuO}_{2}(110) (10 nm)//TiO2\mathrm{TiO}_{2}(110) sample measured with a current applied along the [001][001] direction under an external magnetic field of ±50\pm50 mT after FC at +9 T, FC at −9 T, and ZFC, respectively. (d–f) Enlarged views of the superconducting transition regions in (a–c), respectively. Red and blue solid lines indicate the result of +50 mT and −50 mT, respectively. The accompanying schematics indicate the relative orientation between the exchange field from RuOx\mathrm{RuO}_{x} and the magnetisation of GdN.

In-plane magnetic field dependence of TcT_c in GdN/Nb

Figure S4 shows the in-plane magnetic field dependence of the superconducting transition temperature TcT_c of an AlN(10 nm)/GdN(10 nm)/Nb(9.3 nm)//SiO2/Si\mathrm{AlN}(10\ \mathrm{nm})/\mathrm{GdN}(10\ \mathrm{nm})/\mathrm{Nb}(9.3\ \mathrm{nm})//\mathrm{SiO}_2/\mathrm{Si} heterostructure. The external magnetic field is parallel to the current. The inset shows an enlarged view of the low-field region. For a thin superconducting film under an in-plane magnetic field, the suppression of TcT_c is approximated within the two-dimensional Ginzburg–Landau framework as [S3],

Tc(H)=Tc(0)[1−(HHc2(0))2].(1)T_c(H) = T_c(0)\left[1-\left(\frac{H}{H_{c2}(0)}\right)^2\right]. \tag*{(1)}

In a superconductor/ferromagnetic-insulator bilayer, an additional exchange field can be induced in the superconductor through the magnetic proximity effect. When this proximity-induced exchange field is parallel to the Zeeman field generated by the applied magnetic field, the total spin splitting in the superconductor increases with increasing magnetic field, leading to a monotonic suppression of TcT_c. In contrast, when the induced exchange field is antiparallel to the applied-field contribution, the two fields can partially compensate each other. Such exchange-field compensation can result in a nonmonotonic Tc(H)T_c(H) dependence and, in the appropriate regime, reentrant superconducting behavior [S4, 5].

As shown in Figure S4, TcT_c decreases monotonically with increasing magnetic field. This behavior is therefore consistent with the exchange field induced by GdN in Nb being aligned such that it enhances, rather than compensates. A small cusp is observed in the low-field region, as highlighted in the inset. We tentatively attribute this feature to a small component of the GdN magnetisation that does not fully follow the external magnetic field during magnetisation reversal.Note that, because the superconducting transition becomes broader with increasing magnetic field, TcT_c was defined at R=0.25RNR = 0.25R_{\mathrm{N}}.

Magnetic field dependence of $T_c$ of AlN(10 nm)/GdN(10 nm)/Nb(9.3 nm)//SiO$_2$/Si heterostructures

Figure S4. Magnetic field dependence of TcT_c of AlN(10 nm)/GdN(10 nm)/Nb(9.3 nm)//SiO2_2/Si heterostructures. The inset shows an enlarged view of the low-field region.

Crystallographic-orientation dependence of the superconducting switch

Figures S5a and b show the results obtained after FC along the [11ˉ0][1\bar{1}0] direction, with the subsequent magnetic field, and hence the GdN magnetisation, also aligned along the same direction. Figures S5c and d show the corresponding measurements for field cooling and magnetic-field application along the 45∘45^\circ direction, which is defined here as the in-plane direction halfway between [001][001] and [11ˉ0][1\bar{1}0]. For both orientations, a clear modulation of TcT_c is observed, comparable to that obtained for the [001][001] direction.

Within the proposed altermagnetic picture of RuO2_2, the Néel vector has been reported to preferentially align along the crystallographic [001][001] axis, and recent first-principles calculations predict a strong magnetocrystalline anisotropy stabilizing this orientation [S6, 7]. If the observed ΔTc\Delta T_c originated predominantly from such an altermagnetic order, one would therefore expect the effect to depend strongly on the field orientation and to be substantially reduced when the field is applied away from the preferred Néel-vector direction. Instead, a comparable ΔTc\Delta T_c is observed for the [001][001], [11ˉ0][1\bar{1}0], and 45∘45^\circ directions. This weak angular dependence disfavors an interpretation based on the conventional altermagnetic order of RuO2_2 and instead suggests that the RuO2_2-side magnetic state relevant to the superconducting switching is nearly isotropic with respect to the in-plane field direction.

Temperature dependence of normalized resistance measurements

Figure S5. (a,b) Temperature dependence of the normalized resistance measured after FC ±9\pm9 T along the [11ˉ0][1\bar{1}0] direction, with the subsequent magnetic field, respectively. (c,d) Corresponding measurements for FC ±9\pm9 T and magnetic-field application along the 45∘45^\circ direction, respectively. Red and blue solid lines represent the measurements at +0+0 and −0-0 mT, respectively. The insets show enlarged views of the superconducting transitions.

Temperature dependence of the normalized Nb resistance in the orthogonal magnetic configuration

In a superconducting switch, no difference in TcT_c is expected between the relative magnetic configurations of 90° and 270°. To examine this situation, we measured TcT_c of the 9.3-nm-thick Nb sample in the orthogonal magnetic configuration. Figure S6 shows the device structure and measurement configuration. To avoid possible geometrical effects associated with applying the magnetic field along the short-axis direction of the wire, we used a device geometry rotated by 90° with respect to the device shown in Figure 1d of the main text. For the FC process, a magnetic field of ±9\pm9 T was applied along the [1] direction. Figure S6 schematically illustrates the relative configuration between the RuOx_x exchange field and the GdN magnetisation. Figure S6–Figure S6 show the temperature dependence of the normalized resistance measured under magnetic fields applied along the [11ˉ0][1\bar{1}0] direction. For both ±0\pm0 and ±50\pm50 mT, almost no difference in TcT_c is observed between the two magnetic-field directions as shown in Figure S6–Figure S6. These results are consistent with the superconducting switch model, in which TcT_c is unchanged for orthogonal magnetic configurations. The small residual difference is likely attributable to a slight misalignment of the magnetisation direction and the device geometry. Note that under the ±0\pm0 mT condition, the difference of TcT_c is sensitive to the magnetic-field angle, and a finite difference can appear even for a small angular deviation from the ideal orthogonal configuration. This sensitivity may be related to the difference between the saturation and remanent magnetisation states of GdN, although further investigation is required to clarify its microscopic origin.

Device geometry, exchange-field configuration, and normalized resistance measurements

Figure S6. (a) Device geometry rotated by 90∘90^\circ with respect to the device shown in Figure 1d of the main text and the measurement configuration. (b) Relative configuration between the RuOx\mathrm{RuO}_{x} exchange field and the GdN magnetisation. (c,d) Normalized resistance as a function of temperature for the 9.3-nm-thick Nb sample measured under ±0\pm0 mT after FC at +9+9 T and −9-9 T, respectively. The yellow and purple solid lines represent the data measured at +0+0 mT and −0-0 mT, respectively. The insets show enlarged views near the superconducting transition temperature. (e,f) Normalized resistance of the 9.3-nm-thick Nb sample as a function of temperature measured under ±50\pm50 mT after FC at +9+9 T and −9-9 T, respectively. The yellow and purple solid lines represent the data measured at +50+50 mT and −50-50 mT, respectively. The insets show enlarged views near the superconducting transition temperature.

Superconducting coherence length of Nb

Figure S7a shows the out-of-plane magnetic-field dependence of the normalized resistance for the AlN(10 nm)/Nb(9.3 nm)/RuO2_2(10 nm)//TiO2_2(110) sample measured at various temperatures. Here, the upper critical field, μ0Hc2\mu_0H_{c2}, was defined as the external magnetic field at which the normalized Nb resistance is 0.5. Figure S7b shows the temperature dependence of μ0Hc2\mu_0H_{c2} on the left axis and the temperature dependence of the normalized resistance at zero magnetic field on the right axis. The fitting was performed in the temperature range close to TcT_c using the relation [S3, 8].

μ0Hc2(T)=μ0Hc2(0)(1−T/Tc).(2)\mu_0H_{c2}(T) = \mu_0H_{c2}(0)(1 - T/T_c). \tag*{(2)}

The superconducting coherence length, ξ\xi, was then estimated from

ξ=Φ02πμ0Hc2(0),(3)\xi= \sqrt{\frac{\Phi_0}{2\pi\mu_0H_{c2}(0)}}, \tag*{(3)}

where Φ0\Phi_0 is the quantum superconducting flux [S9]. From the fitting, μ0Hc2(0)\mu_0H_{c2}(0) was estimated to be 2.56±0.0082.56 \pm0.008 T, yielding a superconducting coherence length (ξ\xi) of 11.35±0.0211.35 \pm0.02 nm. These results indicate that the 9.3-nm-thick Nb sample used in the main text is thinner than ξ\xi, whereas the 12-nm-thick Nb sample in S8 is thicker than ξ\xi.

Out-of-plane magnetic-field dependence and temperature dependence of the normalized resistance and upper critical field

Figure S7. (a) Out-of-plane magnetic-field dependence of the normalized resistance for the AlN(10 nm)/Nb(9.3 nm)/RuO2_2(110) (10 nm)//TiO2_2(110) sample measured at various temperatures. (b) Temperature dependence of μ0Hc2\mu_0H_{c2} (left axis and red dots) and normalized resistance at zero magnetic field (right axis and blue dots). The orange solid line represents the fitting result for the temperature dependence of μ0Hc2\mu_0H_{c2}.

Superconducting switching for an Nb thickness exceeding the coherence length

In the superconducting switch, when the superconducting layer thickness exceeds ξ\xi, Cooper pairs cannot effectively experience the exchange fields from both the upper and lower magnetic layers simultaneously. Therefore, the difference of TcT_{\mathrm{c}} originating from the relative magnetic configuration is significantly reduced compared with the case where the superconducting layer thickness is smaller than ξ\xi [S10]. Figure S8–Figure S8 show the temperature dependence of the normalized resistance for the AlN(10 nm)/GdN(10 nm)/Nb(12 nm)/RuO2_{2}(110) (10 nm)//TiO2_{2}(110) sample. Both the FC field and the magnetic field applied were oriented along the [1] direction. The Nb thickness of 12 nm is thicker than the estimated ξ\xi in S7. Consistently, TcT_{\mathrm{c}} shift depending on the relative configuration between the RuOx_{x} exchange field and the GdN magnetisation is strongly suppressed under both the ±0\pm0 mT conditions and the ±50\pm50 mT conditions as shown in Figure S8. These results further support that the TcT_{\mathrm{c}} difference observed in Figure 2 of the main text is not caused by extrinsic effects such as a residual magnetic field or a stray field of GdN but rather originates from the exchange field across the Nb layer.

Normalized Nb resistance as a function of temperature for the 12-nm-thick Nb sample

Figure S8. (a–d) Normalized Nb resistance as a function of temperature for the 12-nm-thick Nb sample. The measurements were performed under ±0\pm0 mT in (a,b) and under ±50\pm50 mT in (c,d). The red and blue solid lines represent the data measured at +0+0 mT and −0-0 mT in (a,b), and at +50+50 mT and −50-50 mT in (c,d), respectively. The insets show enlarged views near the superconducting transition temperature. The schematic illustrations indicate the relative configurations of the exchange fields.

Temperature dependence of the normalized Nb resistance in control samples

The results presented in the main text originate from the combined effect of the RuOx\mathrm{RuO}_{x} exchange field and the GdN exchange field; therefore, a similar TcT_{c} difference is not expected when either contribution is absent.

Figures S9a and b show the temperature dependence of the normalized resistance for the AlN(10 nm)/GdN(10 nm)/Nb(9.3 nm)//SiO2/Si\mathrm{AlN}(10\ \mathrm{nm})/\mathrm{GdN}(10\ \mathrm{nm})/\mathrm{Nb}(9.3\ \mathrm{nm})//\mathrm{SiO}_{2}/\mathrm{Si} control sample after FC at +9 T+9\ \mathrm{T} and −9 T-9\ \mathrm{T}, respectively. In both FC conditions, no clear change in TcT_{c} depending on the GdN magnetisation direction is observed. This control experiment allows us to exclude a possible superconducting switching mechanism arising from additional magnetic regions at the GdN/Nb interface. In particular, structural disorder or amorphous patches at the GdN/Nb interface could, in principle, host a magnetic state distinct from the bulk GdN magnetisation, resulting in an effective SC/FM(FerroMagnet)/FM-type structure. If the relative orientation between such interfacial magnetism and the GdN magnetisation produced a superconducting spin-valve effect, reversing the FC direction would be expected to induce a measurable change in TcT_{c}. The absence of a clear TcT_{c} difference between the +9 T+9\ \mathrm{T} and −9 T-9\ \mathrm{T} FC conditions indicates that this scenario does not account for the superconducting switching observed in the main sample.

Figures S9c and d show the temperature dependence of the normalized resistance for the AlN(10 nm)/Nb(9.3 nm)/RuO2(10 nm)//TiO2(110)\mathrm{AlN}(10\ \mathrm{nm})/\mathrm{Nb}(9.3\ \mathrm{nm})/\mathrm{RuO}_{2}(10\ \mathrm{nm})//\mathrm{TiO}_{2}(110) control sample under ±50 mT\pm50\ \mathrm{mT} after FC at +9 T+9\ \mathrm{T} and −9 T-9\ \mathrm{T}, respectively. In both cases, no clear change in TcT_{c} depending on the external magnetic-field direction is observed. In principle, a TcT_{c} difference could arise from the relative orientation between the RuOx\mathrm{RuO}_{x} exchange field and the external magnetic field in this control sample. However, such a shift was not clearly resolved in the present measurements, suggesting that any TcT_{c} difference originating solely from the RuO2\mathrm{RuO}_{2} exchange field, if present, is below the experimental detection limit. These control experiments indicate that the coexistence of the RuOx\mathrm{RuO}_{x} exchange field and the GdN magnetisation is essential for the TcT_{c} difference observed in the main text.

Temperature dependence of normalized resistance for control samples

Figure S9. (a,b) Temperature dependence of the normalized resistance of the AlN(10 nm)/GdN(10 nm)/Nb(9.3 nm)//SiO2_2/Si control sample after FC at (a) +9 T and (b) −9-9 T. The red and blue solid lines represent the measurements at +0 mT and −0-0 mT, respectively. (c,d) Temperature dependence of the normalized resistance of the AlN(10 nm)/Nb(9.3 nm)/RuO2_2(10 nm)//TiO2_2(110) control sample measured under ±50\pm50 mT after FC at (c) +9 T and (d) −9-9 T. The red and blue solid lines represent the measurements under +50 mT and −50-50 mT, respectively.

Superconducting switching in the 50-nm-thick RuO2\mathrm{RuO}_{2} sample

RuO2\mathrm{RuO}_{2} is known to exhibit metallic transport except in the ultrathin limit [S11]. Therefore, in the present experiment, the possibility that Cooper pairs penetrate into metallic RuO2\mathrm{RuO}_{2} cannot be excluded. To rule out the possibility that a small magnetic moment at the TiO2/RuO2\mathrm{TiO}_{2}/\mathrm{RuO}_{2} interface contributes to the superconducting switch effect, we performed superconducting switch measurements using an AlN(10 nm)/GdN(10 nm)/Nb(9.3 nm)/RuO2(50 nm)//TiO2\mathrm{AlN}(10\ \mathrm{nm})/\mathrm{GdN}(10\ \mathrm{nm})/\mathrm{Nb}(9.3\ \mathrm{nm})/\mathrm{RuO}_{2}(50\ \mathrm{nm})//\mathrm{TiO}_{2} heterostructure. For a RuO2\mathrm{RuO}_{2} thickness of 50 nm, considering the coherence length estimated in Figure S8, the Cooper pair is not expected to extend as far as the TiO2/RuO2\mathrm{TiO}_{2}/\mathrm{RuO}_{2} interface. The measurements were performed at a current density of 4.30×108 A/m24.30 \times10^{8}\ \mathrm{A}/\mathrm{m}^{2}.

The sample was first field-cooled under a magnetic field of 9 T, after which the magnetic state on the RuO2\mathrm{RuO}_{2} side was controlled by applying saturation fields of ±1.5\pm1.5 T. Figure S10 and Figure S10 show the temperature dependence of the normalized resistance measured at ±0\pm0 mT after saturation at ±1.5\pm1.5 T. Although TcT_{\mathrm{c}} is reduced compared with the 10-nm-thick RuO2\mathrm{RuO}_{2} sample, likely due to the increased RuO2\mathrm{RuO}_{2} thickness, a clear difference in TcT_{\mathrm{c}} depending on the saturation-field direction and the magnetisation direction of GdN is still observed, similar to the case of 10-nm-thick RuO2\mathrm{RuO}_{2}. Importantly, the polarity of the superconducting switching is also the same.

Temperature dependence of normalized resistance

Figure S10. (a,b) Temperature dependence of the normalized resistance of the AlN(10 nm)/GdN(10 nm)/Nb(9.3 nm)/RuO2_{2}(50 nm)//TiO2_{2} heterostructure measured after saturation at ±\pm1.5 T, respectively. The red and blue solid lines represent the measurements at +0 and −0 mT, respectively. The insets show enlarged views of the superconducting transitions.

These results indicate that the superconducting switch effect discussed in this work originates from magnetism at the RuO2/Nb\mathrm{RuO}_{2}/\mathrm{Nb} interface rather than from magnetism at the TiO2/RuO2\mathrm{TiO}_{2}/\mathrm{RuO}_{2} interface.

Origin of the magnetism in the RuOx_x redox layer

Reduction of Ru4+^{4+} at the Nb/RuO2_2 interface produces Ru3+^{3+}, although further local reduction to Ru2+^{2+} may also occur. However, oxygen vacancies in RuO2_2 are known to induce electron accumulation at neighboring Ru sites and stabilize substantial fractions of low-valent Ru3+^{3+} species [S12, 13]. XPS studies have directly resolved enhanced Ru3+^{3+} components in oxygen-deficient RuO2_2, while Ru K-edge XANES measurements yield average Ru valences as low as approximately +3.4, consistently interpreted as a mixed Ru4+^{4+}/Ru3+^{3+} state [S12–14]. Although a fraction of the interfacial Ru sites may be further reduced to lower-valence states, a substantial population of Ru3+^{3+} sites is nevertheless expected to persist. For the magnetic physics considered here, the relevant reduced species is Ru3+^{3+}, which carries one hole in the t2gt_{2g} multiplet (t2g5t_{2g}^{5}), forming in the atomic limit a Kramers doublet with a robust S=1/2S = 1/2 moment, in contrast to Ru4+^{4+}, whose moment formation relies on the weak Hund’s coupling and is quenched by hybridization in stoichiometric RuO2_2.

The reduction of one quarter of the oxygens is not a dilute perturbation: each oxygen is shared by three Ru atoms, so 25% oxygen deficiency leaves on average 1.5 missing ligands per RuO6_6 octahedron, and the loss of a single ligand removes an antibonding shift of order 0.3–0.5 eV from the t2gt_{2g} orbital pointing along the broken bond—comparable to the bandwidth. Organizing the problem by symmetry rather than amplitude, one finds a conclusion independent of the octahedron orientation and of the RuO2_2 polymorph: at each reduced site the orbital doublet with lobes along the local vacancy axis n^v\hat{\mathbf{n}}_v is lowered, and the magnetically active hole resides in the remaining planar orbital, whose band is narrowed by the missing ligands. Two consequences follow. First, the Stoner criterion UνF≳1U\nu_F \gtrsim1, far from satisfied in stoichiometric RuO2_2, becomes attainable in the redox layer: the same vacancies that create the Ru3+^{3+} holes carrying the magnetic moment also narrow the band in which they order. The order is ferromagnetic because the layer is metallic and mixed-valent, while structural disorder frustrates any coherent antiferromagnetic pattern. Second, through the spin-orbit coupling each vacancy axis becomes a local magnetic hard axis: a single-site anisotropy energy K(m⋅n^v)2K(\mathbf{m} \cdot\hat{\mathbf{n}}_v)^2 with K>0K > 0, where m(r)\mathbf{m}(\mathbf{r}) is the coarse-grained Stoner order parameter of the layer.

The vacancy axes are random in direction, but not isotropically so: oxygen leaves through the Nb interface, so the distribution of n^v\hat{\mathbf{n}}_v is biased toward the normal e^z\hat{\mathbf{e}}_z, with a quadrupolar alignment parameter Pv=⟨P2(n^v⋅e^z)⟩>0P_v = \langle P_2(\hat{\mathbf{n}}_v \cdot\hat{\mathbf{e}}_z) \rangle> 0. The anisotropy energy of the layer then splits into a coherent part, K‾Pvmz2\overline{K}P_v m_z^2—an easy plane, since the mean hard axis is the normal—and a zero-mean random-axis part. Together with the exchange stiffness of the itinerant magnet, this makes the layer a random-anisotropy magnet of the type studied in amorphous systems [S15], with additional Dzyaloshinskii–Moriya and Rashba-type couplings generated by the parity-odd part of the vacancy fields.

This picture is consistent with the phenomenology of RuOx_x magnetisation. In such a magnet, the exchange stiffness favours uniform order, whereas the random anisotropy favours local alignment with the random axes. As argued by Imry and Ma [S16, 17], the energy gained by following the random axes over a region of size LL grows faster with LL than the exchange cost of rotating the magnetisation, in any dimension below four. Long-range order is therefore expected to be destroyed beyond a characteristic length, the Imry–Ma length L∗L^*, set by the exchange stiffness and by the strength and correlation length of the vacancy disorder, which are not determined here; the layer would then break into nearly collinear in-plane domains whose orientation wanders beyond L∗L^*. The total moment of a nanometer-thick layer is small, and if the layer breaks into such domains, their partial alignment by field cooling would reduce the domain-averaged moment mˉ\bar{m} further, so that it falls below standard magnetometry resolution: the magnet is fragile in moment, and M(H=0)≈0M(H = 0) \approx0 is expected rather than paradoxical. Its rigidity follows from the same smallness: an external field couples to the layer only through a small Zeeman energy, so that reversal—whether by coherent rotation or by depinning of domain walls from the random-anisotropy landscape [S18]—requires large fields even for a moderate anisotropy energy. The same oxygen vacancies that enable the Stoner instability (band narrowing) also pin its order parameter (random axes). A fragile yet rigid magnet is thus a natural outcome of interfacial redox chemistry.

Two further properties matter for the switch. First, the easy plane is a gradient effect, strongest for thin redox layers; a thick redox layer would approach an isotropic random-axis glass with no switchable moment, acting on the superconductor only as a spin-disorder pair breaker. The observation of a hysteretic, switchable subsystem is thus consistent with the thin, self-limited redox depth. Second, the layer is in-plane isotropic—mˉ\bar{m} is slaved to the cooling field, not to crystal axes. This distinguishes the redox scenario sharply from any altermagnetic or surface-crystalline magnetism, which would imprint the symmetry of the (110) surface on the angular response, and it is consistent with the observed insensitivity of the switch amplitude to the FC angle in the plane (see S6 for the 45° data for in-plane anisotropy).

The electronic structure of the reduced layer also accounts for the negative sign of the exchange field inferred in the main text. In the t2g5t_{2g}^{5} configuration of Ru3+^{3+} the t2gt_{2g} band is more than half filled: the majority-spin t2gt_{2g} states are occupied, and the only empty t2gt_{2g} states near the Fermi level are minority-spin states. Hybridization of Nb conduction electrons with these empty states lowers the energy of minority-spin Nb electrons, inducing a spin polarization in Nb antiparallel to the RuOx_x magnetisation, i.e. an antiferromagnetic coupling between Nb conduction electrons and the Ru moments. The same sign follows in a localized picture (atomic limit), since a Nb electron can virtually hop into the Ru3+^{3+} hole only if its spin is antiparallel to the local moment. The mechanism yields the opposite sign for GdN, where the lowest empty Gd 5d5d states are majority-spin states owing to their exchange splitting by the 4f74f^{7} moment, in agreement with the positive exchange field found for the GdN control sample (Supplementary S5).

Theory II: Quasiclassical theory about ΔTc\Delta T_c

To determine the critical temperature of the diffusive superconducting film, we solve the Usadel equation [S19] subject to the appropriate boundary conditions for superconductors in contact with ferromagnetic insulators [S20].

Near the superconducting transition temperature, the Usadel equation can be linearized. For Nb thicknesses satisfying dNb≲ξd_{\mathrm{Nb}} \lesssim\xi, the superconducting order parameter can be approximated as uniform across the film thickness. Averaging the linearized equation over the film thickness then yields the following self-consistency relation for TcT_c:

ln⁡Tc0Tc=(1+δ)ψ(z+)+(1−δ)ψ(z−)2−ψ(12)(4)\ln\frac{T_{c0}}{T_c} = \frac{(1+\delta)\psi\left(z_+\right)+(1-\delta)\psi\left(z_-\right)}{2}-\psi\left(\frac{1}{2}\right) \tag*{(4)}

where ψ(z)\psi(z) is the digamma function and Tc0T_{c0} is the superconducting transition temperature of the reference Nb film in the absence of interfacial magnetic depairing. The parameters δ\delta and z±z_{\pm} are defined as:

δ=τsf−1−τso−1(τsf−1−τso−1)2−(4hℏ)2(5)\delta= \frac{\tau_{\mathrm{sf}}^{-1}-\tau_{\mathrm{so}}^{-1}}{\sqrt{\left(\tau_{\mathrm{sf}}^{-1}-\tau_{\mathrm{so}}^{-1}\right)^2-\left(\frac{4h}{\hbar}\right)^2}} \tag*{(5)}
z±=12+2Dq⃗ 2+2τsf−1+τso−1±(τsf−1−τso−1)2−(4hℏ)28πkBTc/ℏ.(6)z_{\pm}=\frac{1}{2}+\frac{2D\vec{q}^{\,2}+2\tau_{\mathrm{sf}}^{-1}+\tau_{\mathrm{so}}^{-1}\pm\sqrt{\left(\tau_{\mathrm{sf}}^{-1}-\tau_{\mathrm{so}}^{-1}\right)^2-\left(\frac{4h}{\hbar}\right)^2}}{8\pi k_{\mathrm{B}}T_c/\hbar}. \tag*{(6)}

Here, τsf\tau_{\mathrm{sf}} denotes the effective spin-relaxation time associated with the ferromagnetic-insulator interfaces. By contrast, τso\tau_{\mathrm{so}} characterizes spin relaxation arising from spin–orbit coupling in Nb. The parameter hh is the effective exchange field averaged over the superconducting layer, induced by the two magnetic interfaces. For the parallel (P) and antiparallel (AP) magnetic configurations, the effective exchange fields are:

hP/AP=hGdN±hRuOx.(7)h_{\mathrm{P/AP}}=h_{\mathrm{GdN}}\pm h_{\mathrm{RuOx}}. \tag*{(7)}

The gauge-invariant superfluid wavevector is defined as q⃗=∇⃗φ−2eℏA⃗\vec{q}=\vec{\nabla}\varphi-\frac{2e}{\hbar}\vec{A} where φ\varphi is the superconducting phase and A⃗\vec{A} is the vector potential. In the limit q→0q\to0, corresponding to the absence of depairing associated with finite condensate momentum, Eq. (4) formally resembles the expression describing the suppression of superconductivity by magnetic impurities [S21]. The underlying physical mechanisms are nevertheless distinct. Here, the effective exchange field and spin-relaxation processes arise from the spin-mixing conductances at the superconductor/ferromagnetic-insulator interfaces, rather than from a bulk magnetic field or randomly distributed magnetic impurities. Accordingly, hh and τsf\tau_{sf} should be interpreted as effective parameters characterizing the interfacial exchange and spin-relaxation effects, respectively.

For Nb, we take the spin–orbit relaxation time to be τso≈0.2\tau_{so} \approx0.2 ps, as estimated from Ref. [S22]. To estimate the effective exchange field induced by GdN, we use the experimental results reported in Ref. [S23], which investigated a symmetric GdN/Nb/GdN spin valve. For a Nb thickness of dNb=10d_{\mathrm{Nb}} = 10 nm, the reported transition temperatures were Tc0≈6.7T_{c0} \approx6.7 K for the bare film and TcAP≈3.9T_{c}^{AP} \approx3.9 K and TcP≈3.7T_{c}^{P} \approx3.7 K for the valve. Using these values in Eq. (4), we obtain the effective spin-relaxation time and exchange field:

τsf≈1.74 ps,hGdN≈0.12 meV.(8)\tau_{sf} \approx1.74\ \mathrm{ps}, \qquad h_{\mathrm{GdN}} \approx0.12\ \mathrm{meV}. \tag*{(8)}

The relation between the thickness-averaged field hGdNh_{\mathrm{GdN}} and the interfacial exchange JGdNJ_{\mathrm{GdN}} is [S24]:

JGdN=dNbaNbhGdN≈3.63 meV(9)J_{\mathrm{GdN}} = \frac{d_{\mathrm{Nb}}}{a_{\mathrm{Nb}}}h_{\mathrm{GdN}} \approx3.63\ \mathrm{meV} \tag*{(9)}

where aNb=0.33a_{\mathrm{Nb}} = 0.33 nm is the lattice parameter for Nb.

For the GdN/Nb/RuOx_x spin valve investigated here, the Nb thickness is dNb=9.3d_{\mathrm{Nb}} = 9.3 nm. Assuming that the interfacial exchange field JGdNJ_{\mathrm{GdN}} is unchanged, we obtain the corresponding thickness-averaged contribution:

hGdN=aNbdNbJGdN≈0.13 meV.(10)h_{\mathrm{GdN}} = \frac{a_{\mathrm{Nb}}}{d_{\mathrm{Nb}}}J_{\mathrm{GdN}} \approx0.13\ \mathrm{meV}. \tag*{(10)}

We define Tc0T_{c0} using a reference AlN(10 nm)/Nb(9.3 nm)/SiO2_2/Si sample. As shown in Figure S12, its superconducting transition temperature is Tc0=6.00T_{c0} = 6.00 K.

Superconducting transition temperature of the AlN(10 nm)/Nb(9.3 nm)//SiO2/Si sample

Figure S12. Superconducting transition temperature of the AlN(10 nm)/Nb(9.3 nm)//SiO2/Si sample.

Under an applied current density of 5.38×1085.38 \times10^{8} A/m2^{2}, the measured transition temperatures of our spin valve are TcAP≈3.186T_{c}^{AP} \approx3.186 K and TcP≈3.194T_{c}^{P} \approx3.194 K. Using these values in Eq. (4), together with the estimated GdN contribution, we obtain:

τsf≈1.8 ps,hRuOx≈−0.004 meV.(11)\tau_{sf} \approx1.8\ \mathrm{ps}, \qquad h_{\mathrm{RuO_x}} \approx-0.004\ \mathrm{meV}. \tag*{(11)}

Finally, using the same thickness-scaling relation as in Eq. (9), we estimate the interfacial exchange field associated with the RuOx_x layer:

JRuOx=dNbaNbhRuOx≈−0.11 meV.(12)J_{\mathrm{RuO_x}} = \frac{d_{\mathrm{Nb}}}{a_{\mathrm{Nb}}}h_{\mathrm{RuO_x}} \approx-0.11\ \mathrm{meV}. \tag*{(12)}

[S1] F. Leuenberger, A. Parge, W. Felsch, K. Fauth, and M. Hessler, Phys. Rev. B 72, 014427 (2005).

[S2] R. Vidyasagar, S. Kitayama, H. Yoshitomi, T. Kita, T. Sakurai, and H. Ohta, The European Physical Journal B 86, 52 (2013).

[S3] M. Tinkham, Introduction to Superconductivity (Courier Corporation, 2004).

[S4] H. W. Meul, C. Rossel, M. Decroux, Fischer, G. Remenyi, and A. Briggs, Physical Review Letters 53, 497 (1984).

[S5] D. Maryenko, M. Kawamura, I. V. Maznichenko, S. Ostanin, D. Zhang, M. Kriener, V. K. Dugaev, E. Y. Sherman, A. Ernst, and M. Kawasaki, Science AdvAnceS (2026).

[S6] C. He, Z. Wen, J. Okabayashi, Y. Miura, T. Ma, T. Ohkubo, T. Seki, H. Sukegawa, and S. Mitani, Nature Communications 16, 8235 (2025).

[S7] Y. Guo, A. Chen, Z. Zeng, T. An, Q. Cui, Y. Ga, D. Zheng, X. Deng, X. Zhang, M. Tang,

References

  1. [1]Fletcher, J. M., Gardner, W. E., Greenfield, B. F., Holdoway, M. J. & Rand, M. H. Magnetic and other studies of ruthenium dioxide and its hydrate. J. Chem. Soc. A 3, 653–657 (1968).DOI
  2. [2]Ryden, W. D. & Lawson, A. W. Magnetic susceptibility of IrO₂ and RuO₂. J. Chem. Phys. 52, 6058–6061 (1970).DOI
  3. [3]Berlijn, T. et al. Itinerant antiferromagnetism in RuO₂. Phys. Rev. Lett. 118, 077201 (2017).
  4. [4]Zhu, Z. H. et al. Anomalous antiferromagnetism in metallic RuO₂ determined by resonant X-ray scattering. Phys. Rev. Lett. 122, 017202 (2019).DOI
  5. [5]Šmejkal, L., Sinova, J. & Jungwirth, T. Emerging research landscape of altermagnetism. Phys. Rev. X 12, 040501 (2022).arxiv.org/abs/2204.10844
  6. [6]He, C. et al. Evidence for single variant in altermagnetic RuO₂(101) thin films. Nat. Commun. 16, 8235 (2025).DOI
  7. [7]Fedchenko, O. et al. Observation of time-reversal symmetry breaking in the band structure of altermagnetic RuO₂. Sci. Adv. 10, eadi4883 (2024).DOI
  8. [8]Karube, S. et al. Observation of spin-splitter torque in collinear antiferromagnetic RuO₂. Phys. Rev. Lett. 129, 137201 (2022).DOI
  9. [9]Tschirner, T. et al. Saturation of the anomalous Hall effect at high magnetic fields in altermagnetic RuO₂. APL Mater. 11, 101103 (2023).DOI
  10. [10]Zhang, Y. et al. Electrical manipulation of spin splitting torque in altermagnetic RuO₂. Nat. Commun. 16, 5646 (2025).arxiv.org/abs/2412.17013
  11. [11]Chen, H. et al. Spin-splitting magnetoresistance in altermagnetic RuO₂ thin films. Adv. Mater. 37, 2507764 (2025).DOI
  12. [12]Hiraishi, M. et al. Nonmagnetic ground state in RuO₂ revealed by muon spin rotation. Phys. Rev. Lett. 132, 166702 (2024).DOI
  13. [13]Kesler, P. et al. Absence of magnetic order in RuO₂: insights from μSR spectroscopy and neutron diffraction. npj Spintron. 2, 50 (2024).arxiv.org/abs/2405.10820
  14. [14]Kiefer, L. et al. Crystal structure and absence of magnetic order in single-crystalline RuO₂. J. Phys. Condens. Matter 37, 135801 (2025).DOI
  15. [15]Osumi, T. et al. Spin-degenerate bulk bands and topological surface states associated with Dirac nodal lines in RuO₂. Phys. Rev. B 113, 085116 (2026).DOI
  16. [16]Liu, J. et al. Absence of altermagnetic spin splitting character in rutile oxide RuO₂. Phys. Rev. Lett. 133, 176401 (2024).DOI
  17. [17]Wu, Z. et al. Fermi surface of RuO₂ measured by quantum oscillations. Phys. Rev. X 15, 031044 (2025).DOI
  18. [18]Kim, J. H. et al. Definitive Fermi-surface mapping of high-purity RuO₂ via angle-dependent quantum oscillations. Curr. Appl. Phys. 88, 29–35 (2026).DOI
  19. [19]Wu, X. et al. Magnetic uniformity and limits on room-temperature magnetization in strained RuO₂/TiO₂ films. Phys. Rev. Appl. 23, 054082 (2025).DOI
  20. [20]Wang, Y.-C. et al. Absence of transport altermagnetic spin-splitting effect in RuO₂. Nano Lett. 26, 2548–2554 (2026).arxiv.org/abs/2503.07985
  21. [21]Patton, M. et al. Symmetry control of unconventional spin–orbit torques in IrO₂. Adv. Mater. 35, 2301608 (2023).DOI
  22. [22]Akashdeep, A. et al. Surface-localized magnetic order in RuO₂ thin films revealed by low-energy muon probes. Appl. Phys. Lett. 128, 022406 (2026).DOI
  23. [23]Jiang, H. et al. Strain-induced giant enhancement of magnetism in RuO₂ films. Adv. Funct. Mater. 36, e28172 (2026).DOI
  24. [24]Han, Z. et al. Oxygen vacancy-mediated stabilization of antiferromagnetic order in RuO₂ thin films. Small 22, e74165 (2026).DOI
  25. [25]Fields, S. S. et al. Non-altermagnetic origin of exchange bias behaviors in incoherent RuO₂/Fe bilayer heterostructures. ACS Appl. Mater. Interfaces 18, 6166–6179 (2026).DOI
  26. [26]Jeong, S. G. et al. Emergence of unconventional magnetic order in strain-engineered RuO₂/TiO₂ superlattices. Adv. Sci., e77914 (2026).arxiv.org/abs/2601.10518
  27. [27]Ruf, J. P. et al. Strain-stabilized superconductivity. Nat. Commun. 12, 59 (2021).DOI
  28. [28]Uchida, M., Nomoto, T., Musashi, M., Arita, R. & Kawasaki, M. Superconductivity in uniquely strained RuO₂ films. Phys. Rev. Lett. 125, 147001 (2020).DOI
  29. [29]Wadehra, N. et al. Strain-induced superconductivity in RuO₂(100) thin-films. Commun. Mater. 6, 135 (2025).DOI
  30. [30]Choi, I. H., Jeong, S. G., Jalan, B. & Lee, J. S. Exploring altermagnetism in RuO₂: from conflicting experiments to emerging consensus. Nano Converg. 13, 1 (2026).DOI
  31. [31]De Gennes, P. G. Coupling between ferromagnets through a superconducting layer. Phys. Lett. 23, 10–11 (1966).DOI
  32. [32]Tagirov, L. R. Low-field superconducting spin switch based on a superconductor/ferromagnet multilayer. Phys. Rev. Lett. 83, 2058–2061 (1999).
  33. [33]Gu, J. Y. et al. Magnetization-orientation dependence of the superconducting transition temperature in the ferromagnet-superconductor-ferromagnet system: CuNi/Nb/CuNi. Phys. Rev. Lett. 89, 267001 (2002).pubmed.ncbi.nlm.nih.gov/12484851
  34. [34]Bergeret, F. S., Volkov, A. F. & Efetov, K. B. Odd triplet superconductivity and related phenomena in superconductor-ferromagnet structures. Rev. Mod. Phys. 77, 1321–1373 (2005).DOI
  35. [35]Buzdin, A. I. Proximity effects in superconductor-ferromagnet heterostructures. Rev. Mod. Phys. 77, 935–976 (2005).DOI
  36. [36]Moraru, I. C., Pratt, W. P. & Birge, N. O. Magnetization-dependent T_c shift in ferromagnet/superconductor/ferromagnet trilayers with a strong ferromagnet. Phys. Rev. Lett. 96, 037004 (2006).pubmed.ncbi.nlm.nih.gov/16486758
  37. [37]Li, B. et al. Superconducting spin switch with infinite magnetoresistance induced by an internal exchange field. Phys. Rev. Lett. 110, 097001 (2013).pubmed.ncbi.nlm.nih.gov/23496737
  38. [38]Gu, Y., Halász, G. B., Robinson, J. W. A. & Blamire, M. G. Large superconducting spin valve effect and ultrasmall exchange splitting in epitaxial rare-earth-niobium trilayers. Phys. Rev. Lett. 115, 067201 (2015).pubmed.ncbi.nlm.nih.gov/26296128
  39. [39]Zhu, Y., Pal, A., Blamire, M. G. & Barber, Z. H. Superconducting exchange coupling between ferromagnets. Nat. Mater. 16, 195–199 (2017).DOI
  40. [40]Matsuki, H. et al. Realisation of de Gennes’ absolute superconducting switch with a heavy metal interface. Nat. Commun. 16, 6574 (2025).arxiv.org/abs/2404.16264
  41. [41]Rusanov, A. Yu., Hesselberth, M., Aarts, J. & Buzdin, A. I. Enhancement of the superconducting transition temperature in Nb/Permalloy bilayers by controlling the domain state of the ferromagnet. Phys. Rev. Lett. 93, 057002 (2004).DOI
  42. [42]Flokstra, M. & Aarts, J. Domain-wall enhancement of superconductivity in superconductor/ferromagnet hybrids: Case of weak ferromagnets. Phys. Rev. B 80, 144513 (2009).DOI
  43. [43]Miao, G. X., Mascaro, M. D., Nam, C. H., Ross, C. A. & Moodera, J. S. Domain wall induced magnetoresistance in a superconductor/ferromagnet nanowire. Appl. Phys. Lett. 99, 032501 (2011).DOI
  44. [44]Harris, R., Plischke, M. & Zuckermann, M. J. New model for amorphous magnetism. Phys. Rev. Lett. 31, 160–162 (1973).DOI
  45. [45]Chudnovsky, E. M., Saslow, W. M. & Serota, R. A. Ordering in ferromagnets with random anisotropy. Phys. Rev. B 33, 251–261 (1986).pubmed.ncbi.nlm.nih.gov/9937904
  46. [46]Imry, Y. & Ma, S. Random-field instability of the ordered state of continuous symmetry. Phys. Rev. Lett. 35, 1399–1401 (1975).DOI
  47. [47]Larkin, A. & Ovchinnikov, Y. N. Quasiclassical method in the theory of superconductivity. Zh Eksp Teor Fiz 55, 2262–2272 (1969).
  48. [48]Usadel, K. D. Generalized diffusion equation for superconducting alloys. Phys. Rev. Lett. 25, 507–509 (1970).DOI
  49. [49]Hijano, A. et al. Coexistence of superconductivity and spin-splitting fields in superconductor/ferromagnetic insulator bilayers of arbitrary thickness. Phys. Rev. Res. 3, 023131 (2021).DOI
  50. [50]Yao, Y. et al. Probe of spin dynamics in superconducting NbN thin films via spin pumping. Phys. Rev. B 97, 224414 (2018).DOI
  51. [51]Noh, S. et al. Tunneling magnetoresistance in altermagnetic RuO₂-based magnetic tunnel junctions. Phys. Rev. Lett. 134, 246703 (2025).arxiv.org/abs/2502.13599
  52. [52]Dai, J. et al. Thermal imaging of 180° domain contrast in altermagnetic RuO₂. Natl. Sci. Rev. nwag203, (2026).DOI
  53. [53]Mazanik, A. A., De las Heras, R. & Bergeret, F. S. Superconductivity as a probe of altermagnetism: Critical temperature, field, and current. Phys. Rev. B 114, L260501 (2026).arxiv.org/abs/2603.06188
  54. [54]Geng, Z. et al. Superconductor-ferromagnet hybrids for non-reciprocal electronics and detectors. Supercond. Sci. Technol. 36, 123001 (2023).DOI
  55. [55]Jeon, K.-R. et al. Interferometric Evidence of Nonvolatile Anomalous Phase Shifts in Exchange-Spin-Split Josephson Supercurrent Diodes. ACS Nano 20, 4384–4392 (2026).DOI
  56. [56]Bhakat, S., Samanta, S., Mahapatra, S. & Pal, A. Bistable and absolute switching driven by superconducting exchange coupling. Nat. Commun. 16, 9609 (2025).DOI
  57. [57]Stamopoulos, D., Manios, E. & Pissas, M. Stray-fields-based magnetoresistance mechanism in Ni₈₀Fe₂₀–Nb–Ni₈₀Fe₂₀ trilayered hybrids. Phys. Rev. B 75, 184504 (2007).DOI
  58. [58]Nemes, N. M. et al. Origin of the inverse spin-switch behavior in manganite/cuprate/manganite trilayers. Phys. Rev. B 78, 094515 (2008).DOI
  59. [59]Rusanov, A. Yu., Habraken, S. & Aarts, J. Inverse spin switch effects in ferromagnet-superconductor-ferromagnet trilayers with strong ferromagnets. Phys. Rev. B 73, 060505 (2006).DOI
  60. [60]Komori, S., Suzuki, S., Imura, K. & Taniyama, T. Sign control of the critical-temperature switching in oxide superconducting spin valves. Phys. Rev. Appl. 23, L061003 (2025).DOI
  61. [61]Daccà, A., Gemme, G., Mattera, L. & Parodi, R. XPS analysis of the surface composition of niobium for superconducting RF cavities. Appl. Surf. Sci. 126, 219–230 (1998).DOI
  62. [62]Liu, H. et al. Eliminating over-oxidation of ruthenium oxides by niobium for highly stable electrocatalytic oxygen evolution in acidic media. Joule 7, 558–573 (2023).DOI
  63. [63]Yuan, B. et al. Synergistic niobium and manganese co-doping into RuO₂ nanocrystal enables PEM water splitting under high current. Nat. Commun. 16, 4583 (2025).DOI
  64. [64]Proctor, T. C., Chudnovsky, E. M. & Garanin, D. A. Scaling of coercivity in a 3d random anisotropy model. J. Magn. Magn. Mater. 384, 181–185 (2015).DOI
  65. [65]Abrikosov, A. A. & Gor’kov, L. P. Contribution to the theory of superconducting alloys with paramagnetic impurities. Zh. Eksptl. Teor. Fiz 39, 1781–1796 (1960).
  66. [66]Maki, K. Gapless superconductivity. in Superconductivity vol. 2 1035 (1969).
  67. [67]Cirillo, C., Rusanov, A., Bell, C. & Aarts, J. Depairing current behavior in superconducting Nb/Pd₈₁Ni₁₉ bilayers. Phys. Rev. B 75, 174510 (2007).DOI
  68. [68]Thi, N. A. N., Thi, H. Y. N., Yi, H., Joo, S.-J. & Shin, K.-H. Inverse giant magnetoresistance due to spin-dependent bulk scattering in Fe₁₋ₓCrₓ/Cu/Co. Phys. Status Solidi A 204, 3954–3957 (2007).DOI
  69. [69]Suto, H. et al. Study on FeCr thin film for a spintronic material with negative spin polarization. J. Magn. Magn. Mater. 557, 169474 (2022).

Paper details

vlnc:W_jremm5xq
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 9

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents