Role of NiO in the nonlocal spin transport through thin NiO films on Y_3Fe_5O_{12}
Geert R. Hoogeboom, Geert-Jan N. Sint Nicolaas, Andreas Alexander, Olga Kuschel, Joachim Wollschläger, Inga Ennen, Bart J. van Wees, Timo Kuschel
RRole of NiO in the nonlocal spin transport through thin NiO films on Y Fe O Geert R. Hoogeboom , Geert-Jan N. Sint Nicolaas , Andreas Alexander , OlgaKuschel , Joachim Wollschl¨ager , Inga Ennen , Bart J. van Wees and Timo Kuschel Physics of Nanodevices, Zernike Institute for Advanced Materials,University of Groningen, Nijenborgh 4, 9747 AG Groningen, The Netherlands, Department of Physics and Center of Physics and Chemistry of New Materials,Osnabr¨uck University, Barbarastraße 7, 49076 Osnabr¨uck, Germany Center for Spinelectronic Materials and Devices, Department of Physics,Bielefeld University, Universit¨atsstraße 25, 33615 Bielefeld, Germany (Dated: December 23, 2020)In spin transport experiments with spin currents propagating through antiferromagnetic (AFM)material, the antiferromagnet is treated as a mainly passive spin conductor not generating noradding any spin current to the system. The spin current transmissivity of the AFM NiO is affected bymagnetic fluctuations, peaking at the N´eel temperature and decreasing by lowering the temperature.In order to study the role of the AFM in local and nonlocal spin transport experiments, we send spincurrents through NiO of various thickness placed on Y Fe O . The spin currents are injected eitherelectrically or by thermal gradients and measured at a wide range of temperatures and magneticfield strengths. The transmissive role is reflected in the sign change of the local electrically injectedsignals and the decrease in signal strength of all other signals by lowering the temperature. Thethermally generated signals, however, show an additional upturn below 100 K which are unaffectedby an increased NiO thickness. A change in the thermal conductivity could affect these signals. Thetemperature and magnetic field dependence is similar as for bulk NiO, indicating that NiO itselfcontributes to thermally induced spin currents. I. INTRODUCTION
The scope for using antiferromagnets (AFMs) as basefor spintronic devices has been unveiled by the possibilityto inject spin angular momentum [1–4] which can be car-ried over long distances [5–10] and by the read-out of themagnetic order [11–13]. AFM dynamics is fast comparedto that of ferromagnets (FMs) due to high eigenfrequen-cies [14, 15]. Further, AFMs have vanishing magneti-zation which increases its robustness against magneticperturbations and reduces the cross talk between AFMs[16]. The manipulation of the magnetic moments thusrequires relatively high magnetic fields as compared toFMs. However, the needed magnetic field strength tocontrol the magnetic alignment in the AFM can be re-duced by combining a thin AFM layer with FM layersdue to the exchange coupling across the interface.Spin current has been sent efficiently normal throughthe AFM | FM bilayers NiO | YIG (yttrium iron garnet,Y Fe O ) in a local geometry for which the relaxationlength in NiO is a few nm [1–3, 17–23]. The spin currentscan be generated with electrical means via the spin Halleffect (SHE) in Pt causing electron spins to accumulateat the Pt | NiO interface, or by a heat gradient throughoutthe magnet via the spin Seebeck effect (SSE). The spincurrent through the interface is established via a finitespin mixing conductance and spins flowing into the Ptare converted into a charge current by the inverse spinHall effect (ISHE).Spin Hall magnetoresistance (SMR) is the local com-bination of the SHE and the ISHE and is sensitive to theapplied spin transfer torque on the magnet. This depends on the direction of the magnetic moments under influenceof a magnetic field. Lowering the temperature of thesedevices results in a sign change in the SMR showing thatthe spin current interacts with magnetic moments thatare 90 ° angular shifted as compared to room temper-ature. Although the magnetic moments of YIG alignparallel to the magnetic field, those of NiO tend to alignperpendicular to the magnetic field to lower the Zeemanenergy and due to exchange bias. Therefore, NiO can bethe source of this angular shifted, negative SMR [21, 23]which angular dependence resembles bulk NiO [11, 12].In addition to spin transfer torque which is dampedout over short distances, the spin accumulation in Ptalso causes magnetic excitation in the magnetic bilayer;magnons. These quasiparticles carry the spin angularmomentum over long distances. Furthermore, the cre-ation of magnons with Joule heating from the chargecurrent in Pt causes the magnons to flow from the hottowards the cold part. This results in a negative magnonchemical potential µ m near the injector and a positive µ m at some distance from it [24].All these different sources and forms of spin currentare influenced by the temperature. The positive SMRindicates that the NiO spin transmissivity is high at hightemperatures. The interactions with magnetic momentsin YIG dominate the effect on the spin accumulation overthose in NiO. The SSE also shows this transmissivity ef-fect as a peak at the N´eel temperature. These observa-tions might be explained by magnetic fluctuations, giv-ing the magnetic moments in NiO a component along theYIG magnetization and allowing the transport of spinsalong this direction through the AFM. a r X i v : . [ c ond - m a t . m e s - h a ll ] D ec m m YIGPtJ e NiOJ e M Pt V V αd data adatomlayer2.8nm 2.6nm 0.6nm3.6nm 3.6nm 1.4nm4.7nm 4.7nm 1.6nm6.9nm 7.0nm 1.5nm i n t en s i t y ( a r b . u . ) scattering vector q (Å -1 ) NiO NiOt c)a) b)
FIG. 1. a) X-ray reflectivity data and fits of exemplary samples with various NiO thicknesses t prior to surface treatment andPt device deposition. The data is vertically shifted for clarity. The black lines are fits. The dashed lines consider a singlehomogeneous NiO layer while the solid lines allow the existence of an additional adatom layer. The resulting fit parameters forthe thickness of NiO and the adatom layer are included in the inset table for both fit methods which result in almost the sameNiO thicknesses. All fits consider the YIG layer to be infinite and the GGG substrate to be negligible due to the relative thickYIG layer of 260 nm. b) TEM image of the NiO(4.9 nm ∗ ) device. The cross section shows the Pt injector and NiO thin filmbeing polycrystalline as well as the single crystalline YIG. c) Illustration of the device structure including the electrical injectionmeasurement scheme. The YIG magnetization M is rotated in-plane by a magnetic field with angle α . By the exchange biaswith YIG, the NiO magnetic moments in both sublattices align perpendicular to the magnetic field. The SHE-induced spinaccumulation in Pt causes transfer of spin angular momentum into the magnetic bilayer by the spin-flip mechanism. Theinjected spin current is carried by propagating magnons through the NiO | YIG bilayer and detected as a voltage via the ISHEin a second Pt strip at distance d . In this article, we investigate the temperature depen-dence of different spin currents through the NiO | YIG bi-layers in both local and nonlocal geometry. Most observa-tions can be explained by a decrease in spin current trans-missivity of the NiO layer below the N´eel temperature .Both local and nonlocal SSE signals show damping effectswhile lowering the temperature. The nonlocal electricallyinjected magnons pass the NiO layer twice and are there-fore damped out stronger. Below 50 K, however, all SSEsignals show an upturn with decreasing temperatures.The temperature dependence of the thermal conductivityof the materials involved could change the thermal profileat low temperature. However, the low-temperature SSEsignals are not affected by NiO thickness and thereforeseem unrelated to the NiO transmissivity. In addition,the increase with magnetic field strength at low temper-ature is similar to that in bulk NiO, indicating that NiOitself contributes to the SSE.
II. METHOD
The YIG films of 260 nm thickness are obtained com-mercially and grown on a gadolinium gallium garnet(111) substrate. These were covered by NiO films of var-ious thicknesses by reactive molecular beam epitaxy at asample temperature of 250 ° C. Directly after deposition,the correct stoichiometry and chemical cleanliness of thefilms was checked in situ via x-ray photoelectron spec-troscopy. To determine the exact thicknesses t of the NiOfilms, the bilayer magnets were characterized by x-ray re-flectivity (XRR) in a Philips X’Pert Pro diffractometerwith a Cu K α source. Exemplary curves are shown inFig. 1 a). The NiO optical parameters are fixed for the fits based on the Parratt algorithm [25] and taken fromthe Henke tables [26]. The fit is improved when allowingan adatom layer with free optical parameters. However,this does not affect the NiO thickness significantly.To study the effect of magnetic order or a possi-ble adatom layer on the interface spin transmissivity,some samples were etched by an Ar ion plasma etch at200 W for 10 seconds which is indicated as NiO( t nm ∗ ).Sputter-deposited Pt strips were grown on top of thesebilayer magnets by electron beam lithography using a4% PMMA and an aquasave spincoat. Transmissionelectron microscopy (TEM) images have been takenin a JEOL JEM 2200FS operated at 200 kV using aGatan OneView CMOS camera. The TEM specimenhas been cut out from the original structured sampleby focused ion beam preparation (FEI Helios NanoLabDualBeam). Figure 1 b) shows a TEM image of thePt(8 nm) | NiO(4.9 nm ∗ ) | YIG(260 nm) sample indicating aclean interface, a uniform thickness and a polycrystallineNiO structure.The Pt strips have dimensions of 20 µ m ×
100 nm × Ω range. A high resistance is an indication of a rel-atively proper stoichiometry of the NiO [27] and confirmsany nonlocal signals are spin current related phenomenaas opposite to amorphous YIG [28]. Voltage spikes, how-ever, occasionally resulted in a conductive film, makingthe devices unusable. Figure 1 c) shows a schematic il-lustration of the resulting device structure including theelectrical measurement setup. An alternating current of100 µ A was sent through the injector strip and both thelocal and nonlocal voltage response is measured while ap-plying a magnetic field in various in-plane directions α .The measurements are performed with a superconducting (a) (b) (c) Δ R l ω R l - R ( Ω ) ω Δ ρ l / ρ ω angle (deg) FIG. 2. First harmonic local resistivity changes R ωl -R obtained for the Pt | NiO(4.9 nm ∗ ) | YIG sample at 2 T a) as a functionof the rotation angle for various temperatures. R is defined as the fitted background resistance and is zero at α = 90 ° . Minorbackground signals due to temperature drift as a function of the time are subtracted. The SMR modulation of the resistanceshows an angular shift of 90 ° (negative SMR) by decreasing the temperature. b) The amplitude of the sinusoidal fit ∆ ρ ωl /ρ is depicted as a function of the temperature for the samples with various thicknesses of NiO showing the sign reversal attemperatures of 150 K or above. c) ∆ R ωl as a function of the magnetic field strength. The signals show quadratic magneticfield dependence as the corresponding lines of the fits show. For thick films, the quadratic increase becomes more negative,similar to films without YIG [12]. For thin films, the quadratic increase is positive, showing increased spin conductivity withlarger magnetic fields. magnet system with a variable temperature inlet (VTI)and an ac lock-in method to distinguish electrical injected(first harmonic) and heat related (second harmonic) sig-nals [29]. III. BACKGROUND
Magnetic moments on neighboured (111) planes in NiOalign antiparallel due to the relatively strong superex-change H e of 968 T via the non-magnetic O − [15, 30].Due to the applied magnetic field, the canting angle ofthe magnetic sublattices θ = arcsin( H/ H e ) (cid:28) ° [31],resulting in a small gain in the Zeeman energy. When theZeeman energy is comparable to the magnetic anisotropyenergy, the magnetic moments can align with respect tothe magnetic field, which is easiest within the magneticeasy plane of one of the many possible domains. Eventhough SMR measurements in bulk NiO show a satura-tion at 6 T [11], indicating that the majority of the mag-netic moments are coherently controlled by the magneticfield, a NiO film of 120 nm thickness shows no sign ofmagnetic saturation up to 18 T [12]. Thin films are moresubject to crystallographic defects, resulting in pinning ofthe domain walls [32] and thereby requiring larger mag-netic fields to be manipulated.Domain walls also influence spin currents passingthough NiO by reflection and absorption and they giverise to bound magnons. AFM domain walls can be engi-neered, induced and controlled via exchange bias [33, 34].These effects are more pronounced at low temperaturesfor which the activation energy to move the domain wallis larger than the thermal energy [35]. The control overthe magnetic moments in NiO can be increased usingthe exchange bias with YIG. At low temperatures, theexchange bias is large and comparable in size to the co- ercivity of YIG [19, 23].The SMR signal in bulk NiO increases with the etchstep by a factor of two [11] which could be caused bya change in the magnetic order. Velez et al. reportedan etching step on a YIG sample leading to larger SMRsignals [36], similar as with NiO. However, at low tem-peratures, the etching of YIG lead to a sign change of itsSMR. This shows that the magnetic moment directionsare aligned off the magnetic field direction which is, onaverage, larger than 45 ° .The parameters for diffusion of spin currents throughNiO are temperature dependent as well since this re-lies on thermally excited magnetic fluctuations [37].The magnetic fluctuations could also explain FM reso-nance measurements as a function of the NiO thickness[1, 37, 38]. At low temperatures, an increase in the SSEsignal from YIG has been observed, suggesting a strongcorrelation between magnon and phonon transport [39].In Pt | YIG | NiO | YIG | GGG structures a similar increasein SSE has been attributed to an increased NiO trans-missivity and an increased contribution of the GGG sub-strate [40]. When replacing the easy-plane AFM thinfilm with the uniaxial Cr O , normal spin currents canbe blocked by the AFM film when the N´eel vector liesperpendicular to the YIG magnetization [41]. Magnonscrossing the interface of NiO | YIG might not experiencea similar blocking mechanism as NiO is affected differ-ently by the exchange bias and has many magnon modesextending to low frequencies [42], already populated at afew Kelvin. In addition to incoherent magnon transport,GHz magnons have been driven coherently through easy-plane AFM films, observed by means of element- andtime-resolved x-ray pump-probe measurements [43, 44].Spin transport theories consider the NiO as being aninactive, opaque layer. However, NiO is known to be re-sponsible to create a SSE signal on its own. A SSE signal a) b) c)
FIG. 3. Second harmonic local resistivity changes a) obtained for the Pt | NiO(4.9 nm ∗ ) | YIG sample at 2 T and its fits as afunction of the rotation angle for various temperatures. The SSE modulation of the resistance shows a phase shift of 180 ° at 40 and 80 K equivalent to a sign change, while the 2 K data shows a similar sign as the data obtained at temperatures at120 K and above. b) The amplitude of the sinusoidal fit ∆ R ωl is depicted as a function of the temperature. c) The local SSEshows no magnetic field dependence at 300 K. At 5 K, however, an increasing magnetic field response for the 2.6 nm samples isobserved which is similar to the SMR results. from a NiO layer of 200 nm thickness on FM permalloy in-creases with increasing temperature starting from about150 K [45]. A magnon diffusion theory for the SSE inAFMs shows that the SSE is expected to go to zero atlow temperatures in NiO [46]. However, a SSE signal wasestablished on a µ m length scale in bulk NiO with a non-local geometry at low temperatures [8]. A magnetic fieldlifts the degeneracy of low frequencies magnon modeswith opposite spin, creating an imbalance in their popu-lation [8, 47]. The observation of electrically injected spincurrents through a thin film of α -Fe O whose relaxationlength is governed by domain configurations [48] showeda proof of principle of spin currents through single layerAFM thin films. The nonlocal geometry is employedfor the NiO | YIG bilayer to investigate long-distance spintransport through thin NiO film as a function of the mag-netic order affected by the temperature and the magneticfield.
IV. RESULTS AND DISCUSSION
Figure 2a) shows the angular dependent SMR modula-tion of the Pt | NiO(4.9 nm ∗ ) | YIG device at 2 T for varioustemperatures. A positive SMR is observed above 150 K,while a 90 ° angular shift (negative SMR) can be iden-tified for lower temperatures. The fitted amplitude forthe devices with different NiO thicknesses as a functionof temperature is shown in Fig. 2b). The N´eel temper-ature is an indication for the magnetic order and can bedetermined by a peak in the SMR or SSE temperature de-pendence [2, 21]. This is not observed within the temper-ature range except for the NiO(4.9 nm ∗ ) sample. A highN´eel temperature is an indication of a high magnetic or-der. Bulk NiO has a N´eel temperature of 525 K and thin-ner films are expected to have a lower N´eel temperature.However, the NiO(4.9 nm ∗ ) sample has a lower SMR signchange temperature than the 250 K of the other samples. The etch step thus either lowers the order and therebythe NiO SMR contribution, or increases the transmissiv-ity for spin currents towards and from YIG. The lack ofsuch SMR peaks within the measured temperature rangeindicates that the N´eel temperature is equal or higherthan that of comparable thick NiO layered devices in lit-erature [2, 20, 21].Figure 2c) shows the SMR amplitude for different fieldstrengths at either 5 K and 300 K. The amplitudes of allsamples are fitted with an offset and a quadratic magneticfield dependence. The offset is a result of the exchangebias with YIG as all applied field strengths are largerthan the sub mT coercivity of YIG, while the magneticfield strength required to manipulate a thin NiO film is3 orders of magnitude higher [12]. The exchange biasof YIG therefore enables some control of the NiO mag-netic moments even at the smallest applied magnetic fieldstrengths.Both the Pt | NiO(7.5 nm) | YIG andPt | NiO(7.5 nm ∗ ) | YIG sample show a negative quadraticSMR increase by further increasing the magnetic field,similar to the initial increase in SMR with field strengthobserved in Pt on bulk NiO [11] and on thin NiO films[12]. With higher magnetic field strengths, the domainsize increases and the Zeeman energy becomes largerthan the magnetic anisotropy, allowing more control overthe direction of the magnetic moments. The negativeSMR increase with increasing magnetic field strengthusing AFMs is caused by the increased influence of theN´eel vector in all magnetic domains [11, 12]. ThinnerNiO | YIG films, however, show a positive quadraticSMR increase with increasing magnetic field strength.Remarkably, a positive SMR sign is retrieved above 2 Tfor the Pt | NiO(2.6 nm ∗ ) | YIG device at 5 K.Since the positive SMR in Pt | NiO | YIG samples is at-tributed to the interaction of the accumulated spins atthe Pt | NiO interface with YIG being dominant over theinteraction with NiO [20]. Thus, we can conclude that a) b)
FIG. 4. Nonlocal electrically injected signal (1st harmonic) amplitude of different samples with a distance d between the Ptbars of about 750 nm. The voltage is divided by the length of the Pt strips (20 µ m) as well as the injector current (100 µ A).a) The amplitude decreases by lowering the temperature showing a lowered transmissivity of NiO at low temperatures. b)As a function of the magnetic field strength, the room temperature signal slightly decreases, roughly following the same fielddependence as for devices without the NiO interlayer. by increasing the magnetic field, not only the controlover the magnetic moments increases, but also the spincurrent transmissivity becomes more efficient and the in-teraction with YIG becomes dominant. A higher mag-netic order created with the magnetic field might in-crease the transmissivity. On top of this positive con-tribution, a positive quadratic increase could originatefrom the Hanle magnetoresistance, but this contributionis too small to be fully responsible for the sign change[49–51].Figure 3a) shows the angular dependence of the lo-cal SSE signal in the Pt | NiO(4.9 nm ∗ ) | YIG device at 2 Tand various temperatures. Most curves follow a regular E ∝ J s × M dependence, except the curves at 80 K and40 K. These curves show a 180 ° angular shift in the signal,equivalent to a sign change. SSE sign changes have beenreported in literature for ferrimagnetic gadolinium irongarnet, attributed to different contribution of the Gd andFe sublattices at different temperatures [52]. This expla-nation does not hold for the NiO | YIG bilayers since theNiO moments are perpendicular to those of YIG [23, 44].The amplitude of the sinusoidal fits with a period-icity of 360 ° is shown in Fig. 3 b) for the differentsamples. Until about 150 K, most of these amplitudesdecrease with decreasing temperature. This decreasingtransmissivity trend below the N´eel temperature is ob-served in both SSE [2] and SMR [20, 21] measurements.The NiO(3.6 nm) and NiO(4.7 nm) samples on the otherhand show a more flat signal with temperature and havea smooth peak around 150 K. Chen et al. observed asimilar peak for a 5 nm interlayer, and attributed it tothe combination of magnon population and relaxation inYIG, interface effects and an enhancements of spin cur-rents near the blocking temperature around 30 K [40] and50 K [19]. No SSE signal is acquired within the noise forthe NiO(7.0 nm) | YIG device at room temperature, possi-bly because the layer is too thick for the exchange bias toaffect the top part of the film. The NiO(4.9 nm ∗ ) sample, on the other hand, exhibits the same behavior as both2.6 nm samples.When further lowering the temperature, however, allsamples show a recurring SSE signal. The size of thesesignals is comparable to the signal at room temperature.A higher thermal conductivity of the magnetic bilayeror a larger SSE coefficient could play a role in creatinglarger thermally created spin currents for the same givencharge current in Pt. For bilayers including NiO, a SSEin NiO could be another source of enhanced spin currentat low temperature.To study the sensitivity of these low temperature SSEsignals to the magnetic order, field dependent measure-ments have been performed and compared to room tem-perature dependence, as shown in Fig. 3 c). At roomtemperature, no SSE increase is observed contrary to theSMR signal at these temperatures. The increased mag-netic order related to the SMR increase thus seems tohave little effect on the SSE induced spin current. At5 K there is, similar to the SMR signal, an increase withmagnetic field.Several phenomena might affect these low temperatureSSE signals; magnetic pinning by crystallographic de-fects, the enhanced exchange bias with YIG or NiO beinga source of spin currents. To determine whether the spintransport at low temperatures has been improved or toestablish that NiO is a spin current source, we have tocompare these results with data of the nonlocal measure-ments.The room temperature nonlocal spin transportthrough the bilayer shows similar transport properties aswithout the NiO layer in terms of angular dependence,signal strength and field dependence [53, 54]. All signalminima and maxima correspond to the same direction ofthe magnetic field as observed in Pt | YIG systems withcomparable Pt strip distances and are defined as posi-tive. The electrical injection signals of the NiO(3.6 nm)and NiO(4.9 nm ∗ ) samples in Fig. 4 a) show a signal a) b) FIG. 5. Nonlocal thermally injected signal (2nd harmonic) amplitude for different samples. The voltage is divided by thelength of the Pt strips (20 µ m) as well as the square of the injector current (100 µ A). a) When lowering the temperature belowthe N´eel temperature, the signal decreases. Similar to the local SSE signal, the NiO(3.6 nm) and NiO(4.7 nm) samples showan additional peak around 150 K. With further lowering of the temperature, all the samples show a recurring signal which isattributed to the thin NiO film. b) As a function of the field strength, the room temperature signal remains constant or slightlyincreases, while the 5 K measurements show a larger increase with increasing field. In the etched sample this trend is amplified. which is a factor of 3 smaller than the signals obtainedin YIG without the NiO film [53]. The NiO(4.7 nm)sample shows considerable lower signals, possibly be-cause the N´eel temperature of the NiO layer is higher.When lowering the temperature of the NiO(3.6 nm) andNiO(4.9 nm ∗ ) samples, a sharp signal decrease is ob-tained.Figure 4 b) shows the nonlocal electrically injected sig-nals as a function of magnetic field strength. A simi-lar lowering is observed in YIG [54]. In the easy-planeanisotropy state, Hematite also shows a magnetic fielddependence due to the rotation of the pseudo-spin, in-duced by the Dzyaloshinskii-Moriya interaction (DMI)which causes a net magnetization [9]. Although DMI isnot present in NiO, a rotation of the pseudo-spin mightstill occur with distance due to the easy-plane anisotropy[55]. The exchange bias might even act as a DMI as itresults in a net magnetization. This would be fairly smallsince the exchange field is less that 1 mT [23]. The smalleffect of the magnetic field shows that diffusion of spincurrents through the NiO also does not strongly dependon magnetic ordering of NiO at room temperature.The nonlocal SSE also shows a decrease in signal withdecreasing temperature (Fig. 5 a)), although not as pro-nounced as the electrically injected signal. This canbe explained by the different path of these magnons;the electrically injected magnons at the Pt | NiO interfacepasses the NiO twice while the largest part of the heatinduced spin current will be created within the thickerYIG layer. Because the gradient is radially distributed,most of the spin current is generated in the vicinity ofthe injector, after which the NiO only needs to be passedonce in order to be detected.At low temperatures, the nonlocal SSE shows similarrecurring signals as measured locally. The size of thesignals is even larger at 5 K than at 300 K. Figure 5 b)shows the nonlocal thermally generated signals as a func- tion of the magnetic field strengths at 5 K and 300 K. Atroom temperature there is little effect of the magneticfield strength, but the small effect shows increasing SSEsignals as a function of the magnetic field strength. Atlow temperatures, however, a considerable increase withmagnetic field strength is observed.Generally, there is a decrease in spin transport throughNiO by lowering the temperature from room temperatureto about 150 K, which is not observed in samples withouta NiO interlayer [56]. This can be explained by the low-ered transport governed by diffusion in this temperatureregion [46] due to a lower amount of magnetic fluctua-tions. At low temperatures, however, the nonlocal SSEsignal deviates from the present understanding of loweredtransmissivity at low temperatures with an inactive NiOlayer. First of all, there are some observations of negativeSSE values around 40 K to 80 K locally and nonlocally.Even though the transmissivity for spin transfer torqueinduced spin currents decreases at low temperatures, theSSE results of both local and nonlocal geometries showan increase in signal at low temperatures which increasewith increasing field.Unlike the room-temperature SSE and the nonlocalelectrically injected magnons, SMR signal strength seemsto increase with increasing magnetic field strength. Theincreasing magnetic order with larger magnetic fieldstrength might influence the transmissivity of the NiOlayer, increasing the spin transfer torque exerted on theYIG. The effect of domain wall pinning by defects isstronger at low temperatures and larger fields might berequired for control over the NiO magnetic moments.Contrary to the local SSE signals, the room temperaturenonlocal SSE signals do increase with increasing field.Moreover, the field dependence at 5 K is strong enoughto make the SSE signal exceed the 300 K signals at largefields for both the local and nonlocal geometries.The effect of magnetic order on the spin transport issignificant and influenced by etching. In addition to theSMR sign change with field at 5 K, the nonlocal SSE ofthe NiO(2.6 nm ∗ ) | YIG device shows an increase of morethan one order of magnitude in signal strength while thesignal of the non-etched sample remains more constantwith field. The etch step affects the interface by cleaningit from surface adatoms and by affecting the magneticorder. The etching does influence the NiO(4.9 nm ∗ ) | YIGsample, showing considerably lower SSE signals thanthe NiO(4.7 nm) | YIG sample. This NiO(4.9 nm ∗ ) | YIGsample is etched only below the Pt strips. Since theNiO(2.6 nm ∗ ) | YIG sample was afterwards exposed to am-bient conditions for lithography purposes the cleaningseems less relevant. The local SSE shows little effect ofthe etch, indicating that the etch is most relevant for aspin current flowing through the NiO layer at low tem-perature.Similar to these NiO | YIG | GGG samples, a low temper-ature upturn in nonlocal SSE signal has been observedfor the paramagnet substrate GGG itself [57]. The signalsize at a distance d = 500 nm is around 5 k V A − m − .This is 3 orders of magnitude smaller than the signals ob-served in the NiO | YIG | GGG geometry. The GGG thusplays a negligible role in the SSE upturn at low temper-atures. For YIG films without the NiO interlayer, peaksin the SSE have been observed at low temperatures [56].An increase in the SSE coefficient of YIG is held respon-sible for the increase in the signal, but the mechanism isnot well understood.The SSE signal can be the result of different sources forthe spin current. In YIG, a decrease of the SSE signal ob-served with increasing magnetic field strength [54] whilean increase is observed with the NiO interlayer. Further-more, the NiO in expected to become less transmissivefor spin currents originating from YIG at low tempera-tures. Therefore, an active role of NiO as a source of spincurrents could be responsible for the upturn of the SSEsignals at low temperatures. The low temperature SSEsignals observed in these NiO | YIG samples indeed showsimilarities to the SSE observed in bulk NiO. The increaseat low temperatures resembles the increase of SSE in bulkNiO [8]. Moreover, the temperature for which the signalarises and the dependence on the magnetic field strengthis similar to bulk NiO, shown to be originating from theimbalance in the population of the magnon modes withopposite spin [8].Initial spin pumping [1] and SSE [2] signals at roomtemperature increase have been observed by insertingup to a few nm of NiO film between the Pt and YIG.This has been attributed to an enhanced spin conduc-tivity [43, 44]. The further exponential decay of SSEsignal strength with increasing NiO thickness might beinfluenced by the increase in N´eel temperature of thickerfilms, shifting the peak in transmissivity towards higher temperatures. At constant temperature, the spin cur-rent transmissivity then decreases for thicker NiO films.Nonetheless, the peak height of the SSE signal also de-creases with increasing NiO thickness [2, 3].However, the low-temperature SSE signal is not relatedto the thickness of the NiO layer, indicating to be unre-lated to the transmissivity of spin currents from YIG.The passive, diffusive role for spin currents in NiO isshown to be minimally depending on the magnetic fieldstrength in case of the electrically injected magnons. Al-though the transmissivity seems unrelated to the sourceof the spin currents, the SSE does increase significantlyat low temperatures. Therefore, we attribute the lowtemperature SSE signal to the NiO itself.
V. CONCLUSION
Spin currents have been injected by electronic andthermal means into NiO | YIG samples with various thick-nesses at a wide range of temperatures and magneticfield strengths. The spin current transmissivity of NiOpeaks at the N´eel temperature and is reduced by loweringthe temperature. At low temperatures, however, thereis a recurring thermally generated spin current whichhas been detected both locally and nonlocally. An in-crease in thermal conductivity could affect thermally gen-erated spin current in this temperature regime. However,the low temperature SSE signals are not affected by theNiO thickness and therefore seem unrelated to possiblechanges in the transmissivity of the NiO layer. On theother hand, the low temperature SSE signals resemblethose observed in bulk NiO: increasing signal strengthswith increasing magnetic field strength and decreasingtemperature. This indicates that, in addition to the pas-sive, diffusive role, the NiO plays an active part in theSSE signals by generating thermal spin currents itself atlow temperatures.
VI. AKNOWLEDGEMENTS
We acknowledge J. G. Holstein, H. Adema, T. J.Schouten and H. H. de Vries for their technical assistance.In addition, we thank Martin Gottschalk and KarstenRott for support and discussion regarding the TEM ex-periments as well as Andreas H¨utten and G¨unter Reissfor making available the laboratory equipment for samplecharacterization. This work is part of the research pro-gram Magnon Spintronics (MSP) No. 159 financed bythe Nederlandse Organisatie voor Wetenschappelijk On-derzoek (NWO). Further support by the DFG PriorityProgramme 1538 “Spin-Caloric Transport” (KU 3271/1-1) and the Spinoza Prize awarded in 2016 to B. J. vanWees by NWO is gratefully acknowledged. [1] H. Wang, C. Du, P. C. Hammel, and F. Yang, PhysicalReview Letters , 097202 (2014).[2] W. Lin, K. Chen, S. Zhang, and C. L. Chien, PhysicalReview Letters , 186601 (2016).[3] A. Prakash, J. Brangham, F. Yang, and J. P. Heremans,Physical Review B , 014427 (2016).[4] S. M. Wu, W. Zhang, A. Kc, P. Borisov, J. E. Pearson,J. S. Jiang, D. Letterman, A. Hoffmann, and A. Bhat-tacharya, Physical Review Letters , 097204 (2016).[5] R. Lebrun, A. Ross, S. A. Bender, A. Qaiumzadeh,L. Baldrati, J. Cramer, A. Brataas, R. A. Duine, andM. Kl¨aui, Nature , 222 (2018).[6] W. Yuan, Q. Zhu, T. Su, Y. Yao, W. Xing, Y. Chen,Y. Ma, X. Lin, J. Shi, R. Shindou, X. C. Xie, andW. Han, Science Advances , 1098 (2018).[7] W. Xing, L. Qiu, X. Wang, Y. Yao, Y. Ma, R. Cai, S. Jia,X. C. Xie, and W. Han, Physical Review X , 11026(2019).[8] G. R. Hoogeboom and B. J. van Wees, Physical ReviewB , 214415 (2020).[9] T. Wimmer, A. Kamra, J. G¨uckelhorn, M. Opel,S. Gepr¨ags, R. Gross, H. Huebl, and M. Althammer,Phys. Rev. Lett. , 247204 (2020).[10] J. Han, P. Zhang, Z. Bi, Y. Fan, T. S. Safi, J. Xiang,J. Finley, L. Fu, R. Cheng, and L. Liu, Nature Nan-otechnology , 563 (2020).[11] G. R. Hoogeboom, A. Aqeel, T. Kuschel, T. T. M. Pal-stra, and B. J. van Wees, Applied Physics Letters ,052409 (2017).[12] J. Fischer, O. Gomonay, R. Schlitz, K. Ganzhorn, N. Vli-etstra, M. Althammer, H. Huebl, M. Opel, R. Gross,S. T. B. Goennenwein, and S. Gepr¨ags, Physical ReviewB , 014417 (2018).[13] L. Baldrati, A. Ross, T. Niizeki, C. Schneider, R. Ramos,J. Cramer, O. Gomonay, M. Filianina, T. Savchenko,D. Heinze, A. Kleibert, E. Saitoh, J. Sinova, andM. Kl¨aui, Physical Review B , 024422 (2018).[14] F. M. Johnson and A. H. Nethercot, Physical Review , 705 (1959).[15] M. T. Hutchings and E. J. Samuelsen, Physical ReviewB , 3447 (1972).[16] S. Loth, S. Baumann, C. P. Lutz, D. M. Eigler, and A. J.Heinrich, Science , 196 (2012).[17] C. Hahn, G. de Loubens, V. V. Naletov, J. B. Youssef,O. Klein, and M. Viret, European Physics Letters ,571 (2014).[18] Z. Qiu, J. Li, D. Hou, E. Arenholz, A. T. N’Diaye, A. Tan,K. I. Uchida, K. Sato, S. Okamoto, Y. Tserkovnyak, Z. Q.Qiu, and E. Saitoh, Nature Communications , 12670(2016).[19] T. Shang, Q. F. Zhan, H. L. Yang, Z. H. Zuo, Y. L. Xie,L. P. Liu, S. L. Zhang, Y. Zhang, H. H. Li, B. M. Wang,Y. H. Wu, S. Zhang, and R.-W. Li, Applied PhysicsLetters , 032410 (2016).[20] W. Lin and C. L. Chien, Physical Review Letters ,067202 (2017).[21] D. Hou, Z. Qiu, J. Barker, K. Sato, K. Yamamoto,S. V´elez, J. M. Gomez-Perez, L. E. Hueso, F. Casanova,and E. Saitoh, Physical Review Letters , 147202(2017). [22] Y. M. Hung, C. Hahn, H. Chang, M. Wu, H. Ohldag,and A. D. Kent, AIP Advances , 055903 (2017).[23] Z. Z. Luan, F. F. Chang, P. Wang, L. F. Zhou, J. F.Cooper, C. J. Kinane, S. Langridge, J. W. Cai, J. Du,T. Zhu, and D. Wu, Applied Physics Letters , 072406(2018).[24] J. Shan, L. J. Cornelissen, N. Vlietstra, J. B. Youssef,T. Kuschel, R. A. Duine, and B. J. V. Wees, PhysicalReview B , 174437 (2016).[25] L. G. Parratt, Physical Review , 359 (1954).[26] B. L. Henke, E. M. Gullikson, and J. C. Davis, Atomicdata and nuclear data tables , 181 (1993).[27] Y. M. Lu, W. S. Hwang, and J. S. Yang, Surface andCoatings Technology , 231 (2002).[28] J. M. Gomez-Perez, K. Oyanagi, R. Yahiro, R. Ramos,L. E. Hueso, E. Saitoh, and F. Casanova, AppliedPhysics Letters , 032401 (2020).[29] N. Vlietstra, J. Shan, B. J. van Wees, M. Isasa,F. Casanova, and J. B. Youssef, Physical Review B ,174436 (2014).[30] S. M. Rezende, A. Azevedo, and R. L. Rodr´ıguez-Su´arez,Journal of Applied Physics , 151101 (2019).[31] G. A. Gurevich, A. G. ; Melkov, CRC Press , Vol. 13(1996) pp. 74–76.[32] J. Xu, C. Zhou, M. Jia, D. Shi, C. Liu, H. Chen, G. Chen,G. Zhang, Y. Liang, J. Li, W. Zhang, and Y. Wu, Phys-ical Review B , 134413 (2019).[33] J. M. Logan, H. C. Kim, D. Rosenmann, Z. Cai, R. Divan,O. G. Shpyrko, and E. D. Isaacs, Applied Physics Letters , 192405 (2012).[34] E. G. Tveten, A. Qaiumzadeh, O. A. Tretiakov, andA. Brataas, Physical Review Letters , 127208 (2013).[35] R. P. Michel, N. E. Israeloff, M. B. Weissman, J. A. Dura,and C. P. Flynn, Physical Review B , 7413 (1991).[36] S. V´elez, A. Bedoya-Pinto, W. Yan, L. E. Hueso, andF. Casanova, Physical Review B , 174405 (2016).[37] S. M. Rezende, R. L. Rodr´ıguez-Su´arez, and A. Azevedo,Physical Review B , 054412 (2016).[38] H. Wang, C. Du, P. C. Hammel, and F. Yang, PhysicalReview B , 220410 (2015).[39] R. Iguchi, K. I. Uchida, S. Daimon, and E. Saitoh, Phys-ical Review B , 1 (2017).[40] Y. Chen, E. Cogulu, D. Roy, J. Ding, J. B. Mohammadi,P. G. Kotula, N. A. Missert, M. Wu, and A. D. Kent,AIP Advances , 105319 (2019).[41] Z. Qiu, D. Hou, J. Barker, K. Yamamoto, O. Gomonay,and E. Saitoh, Nature Materials , 577 (2018).[42] J. Milano and M. Grimsditch, Physical Review B ,094415 (2010).[43] Q. Li, M. Yang, C. Klewe, P. Shafer, A. T. N’Diaye,D. Hou, T. Y. Wang, N. Gao, E. Saitoh, C. Hwang, R. J.Hicken, J. Li, E. Arenholz, and Z. Q. Qiu, Nature Com-munications , 5265 (2019).[44] M. D¸abrowski, T. Nakano, D. M. Burn, A. Frisk, D. G.Newman, C. Klewe, Q. Li, M. Yang, P. Shafer, E. Aren-holz, T. Hesjedal, G. Van Der Laan, Z. Q. Qiu, and R. J.Hicken, Physical Review Letters , 217201 (2020).[45] P. R. T. Ribeiro, F. L. A. Machado, M. Gamino,A. Azevedo, and S. M. Rezende, Physical Review B ,094432 (2019). [46] S. M. Rezende, A. Azevedo, and R. L. Rodriguez-Su´arez,Journal of Physics D: Applied Physics , 174004 (2018).[47] R. Cheng, J. Xiao, Q. Niu, and A. Brataas, PhysicalReview Letters , 057601 (2014).[48] A. Ross, R. Lebrun, O. Gomonay, D. A. Grave, A. Kay,L. Baldrati, S. Becker, A. Qaiumzadeh, C. Ulloa,G. Jakob, F. Kronast, J. Sinova, R. Duine, A. Brataas,A. Rothschild, and M. Kl¨aui, Nano Letters , 306(2020).[49] M. I. Dyakonov, Physical Review Letters , 126601(2007).[50] S. V´elez, V. N. Golovach, A. Bedoya-Pinto, M. Isasa,E. Sagasta, M. Abadia, C. Rogero, L. E. Hueso, F. S.Bergeret, and F. Casanova, Physical Review Letters ,016603 (2016).[51] J. Shan, P. Bougiatioti, L. Liang, G. Reiss, T. Kuschel,and B. J. Van Wees, Applied Physics Letters , 132406 (2017).[52] S. Gepr¨ags, A. Kehlberger, F. Coletta, Z. Qiu, E. J.Guo, T. Schulz, C. Mix, S. Meyer, A. Kamra, M. Al-thammer, H. Huebl, G. Jakob, Y. Ohnuma, H. Adachi,J. Barker, S. Maekawa, G. E. W. Bauer, E. Saitoh,R. Gross, S. T. B. Goennenwein, and M. Kl¨aui, NatureCommunications , 10452 (2016).[53] L. J. Cornelissen, J. Liu, R. A. Duine, J. b. Youssef, andB. J. Van Wees, Nature Physics , 1022 (2015).[54] L. J. Cornelissen and B. J. Van Wees, Physical ReviewB , 020403 (2016).[55] A. Kamra, T. Wimmer, H. Huebl, and M. Althammer,Physical Review B , 174445 (2020).[56] L. J. Cornelissen, J. Shan, and B. J. Van Wees, PhysicalReview B , 180402(R) (2016).[57] K. Oyanagi, S. Takahashi, L. J. Cornelissen, J. Shan,S. Daimon, T. Kikkawa, G. Bauer, B. J. van Wees, andE. Saitoh, Nature Communications10