Spin-polarized quasiparticle tunneling in spin-filter pseudospin-valve devices
aa r X i v : . [ c ond - m a t . m e s - h a ll ] A ug Spin-polarized quasiparticle tunneling in spin-filter pseudospin-valve devices
P. K. Muduli ∗ Department of Materials Science and Metallurgy,University of Cambridge,27 Charles Babbage Road, Cambridge CB3 0FS,United Kingdom (Dated: September 4, 2018)Spin selective nature of spin-filter tunnel junctions can be integrated with conventional metallicferromagnets to regulate spin polarized quasiparticles in superconducting devices. We report fab-rication of pseudo spin-valve device made with a bilayer of nitride spin-filter tunnel barrier (DyNor GdN) and a transition metal ferromagnet (Co and Gd). We show resistance switching in thesedevices corresponding to parallel and antiparallel configuration of their mutual magnetization di-rection. With optimal deposition process partial nitridation of the Co layer can be achieved. Themagnetically dead native CoN x layer at the Co-DyN interface acts the role of the barrier in thesedevices. In pseudo spin-valve with Co, lower resistance was found for antiparallel state comparedto parallel configuration. Reverse resistance switching behavior was observed for the pseudo spin-valves with Gd. Presence of resistance switching in these devices further confirm the spin-filteringnature of DyN and GdN tunnel barrier. Quasiparticle transport at different temperatures in thesedevices was found to be compatible with conventional N-I-S tunnelling model. These devices can befurther engineered to regulate spin polarized supercurrent in superconducting spintronics devices. PACS numbers: 85.30.Mn, 75.76.+j,74.70.Ad,74.50.+r,72.25.Dc,
I. INTRODUCTION
Superconducting spintronics depends on the creationand manipulation of spin polarized current in devicesinvolving superconducting (S) and ferromagnetic (F)materials[1–3]. In particular, Josephson junction of theS-F-S type have got a lot of attention recently due totheir potential application in quantum computing andspintronics [4–6]. The spin singlet Cooper pair ampli-tude undergoes an oscillatory decay inside a ferromag-netic metal. This oscillatory behavior can lead to aphase difference of π between the two superconductorsdepending on the thickness of the ferromagnet in theS-F-S junction[7–9]. The possibility of π junction wasinitially proposed by Bulaevskii et. al. for a Joseph-son junction including magnetic impurities in the tunnelbarrier[10]. A lot of experimental and theoretical inves-tigation has been done on π Josephson junctions sincethen[11–14]. In the case of tunnel junctions the transi-tion from 0 to π -state can be distinguished in the dif-ferential conductance spectra in the quasiparticle tun-nelling regime. In Al-Al O -PdNi-Nb tunnel junctionswhen thickness of the ferromagnetic layer is increasedfeatures in the superconducting density of state (DOS)are reversed with respect to the normal state indicatinga 0 to π state transition[9]. More complicated tunneljunctions of the from SFIFS[15, 16], SIFIS[17], SIFS[18–20] and SIsFS[21–24] has also been proposed theoreti-cally and observed experimentally. However, comparedto metallic S-F-S Josephson junction tunnelling deviceshave been relatively poorly explored experimentally so-far. It has been theoretically predicted that it is possible ∗ Electronic address: [email protected] to generate spin-triplet superconducting correlations ina ferromagnet with magnetic inhomogeneity in contactwith a superconductor[16, 25]. The triplet correlationsdecay over much longer length scale than usual. Someexperimental evidences has also been reported for theirexistence[26–31]. However, a democratic experimentalevidence of long-range odd-frequency spin-triplet pairs isstill missing. As tunnelling process is more spin selec-tive than the diffusive counter part, spin-valves involvingtunnel barriers can provide more definitive evidence forspin-triplet pairsRecently, we have shown that magnetic semiconduc-tors like GdN, DyN, etc., are quite compatible with su-perconducting NbN and are very promising materials forsuperconducting spintronics[32–36]. We have shown thatspin polarized tunnel current can be very effectively gen-erated due to spin filtering through these tunnel barriers.With unique properties of spin-filter tunnel junctions itis possible to create composite structures in combinationwith ferromagnets to design new kind of supercondutingspintronics devices[37–43]. However, integrating normalferromagnets within all nitride device is not trivial andpossible nitridation of the metal layers and nature of in-terfaces have to be carefully considered. In this paper,we report fabrication and electrical characterization ofpsuedospin-valve devices made of a strong ferromagnetand a spin-filter tunnel barrier. We have used GdN andDyN tunnel barrier as spin polarizer and Co and Gd asanalyzer. A comparison of resistance switching behaviourhas been done between Co-DyN and Gd-GdN type de-vices. Devices with different thickness of DyN were fab-ricated to optimize tunnelling regime. The tunneling na-ture of the device has been analyzed through current-voltage (I-V) measurements at different temperatures.Quasiparticle transport through the psuedospin-valve de-vices has been compared to N-I-S tunnelling model. Weshow spin regulation can be achieved in these devices bycontrolling relative magnetization of the two magneticlayers. We provide a qualitatively explanation for the re-sistance switching behavior observed in our psuedo spin-valve device.
II. EXPERIMENTAL
Multilayer structures NbN-FM-FI-NbN were fabri-cated by DC sputtering in an ultrahigh vacuum (UHV)chamber at room temperature (here FM = Co, Gd andFI= DyN, GdN). Nitride layers were deposited by sput-tering of high purity metal targets in an Ar/N gasmixture with deposition condition as described in thereference[32, 33]. The Co layer was deposited at 1.5 Pain a pure Ar gas with 40 W sputtering power. Ferro-magnetic Gd was also deposited in a similar conditionwith lower 20 W sputtering power. Thickness of differ-ent layers were controlled by regulating the speed of arotating substrate stage. The multilayer structure NbN-FM-FI-NbN was deposited in the order from left to rightinsitu without breaking vacuum. The nitridation of Coand Gd during subsequent deposition of DyN or GdNstrongly depends on time they are exposed to the ni-tride plasma. Therefore, thickness of the DyN or GdNlayer plays a crucial role in controlling the interfacial ni-tridation. In thicker films the Co (Gd) layer is exposedto nitride plasma for a longer time which can lead tohigher nitridation. The CoN x thin film is known to bemagnetic with a perpendicular magnetic anisotropy[59].However, a very thin layer of CoN x might be magneticallydead and prevent magnetic coupling between the FI andFM. In the case of multilayer with Gd, nitrogen deficientGdN x might be formed at the interface. The thicknessof the top and bottom NbN was kept fixed at 50 nm inall the multilayers. While thickness of the Co and DyNlayer was varied in the series. As the top and bottomNbN is common in all the devices, we have used abbre-viation FM( t F M nm)-FI( t F I nm) throughout this paperto represent different kind of devices. Here t F M and t F I represent thickness of the ferromagnetic metal and insu-lator in nanometer, respectively. Devices were fabricatedusing photolithography in a mesa structure. The junc-tion dimension was defined by etching the top NbN inCF plasma for 30 sec and milling 10 min with Ar-ionafter that. Top contact was made with Nb electrode af-ter SiO lift-off. Lateral dimension of the junctions were7 µ m × µ m. Schematic of the mesa device is shown inthe inset of Fig. 1. Differential conductance ( dI/dV ) wasmeasured with a lock-in technique in a custom made dip-stick using liquid helium. Spin valve measurements weredone in a closed-cycle helium refrigerator from Cryogen-ics Lmt.
The measurements were done in a four-probeconfiguration with DC bias current. The magnetizationof the multilayer films deposited at the same time werecarried out with a SQUID magnetometer. In this pa-per measurements done on one representative sample is shown. The reproducibility and behaviour of other de-vices are shown in the supplementary material.
III. RESULTS AND DISCUSSION R ( Ω ) T(K)
NbN NbN SiO DyN Co V I Nb
FIG. 1: (Color online) Temperature dependence of resistanceof a Co(5 nm)-DyN(2 nm) device. The measurement wasdone using a current I = 100 µ A. Inset shows schematic ofthe mesa structure used for devices.
Figure 1 shows temperature dependence of resistance R ( T ) of a Co(5 nm)-DyN(2.5 nm) device measured with abias current I = 100 µ A. The resistance showed semicon-ducting temperature dependence with a small deviationbelow 20 K. Similar R ( T ) was also found for Co-GdNdevices (see supplementary figure SFig. 2). The temper-ature dependence of resistance in these devices is mostlikely determined by the most resistive part i.e., Co-DyNinterface, where formation of disordered CoN x is possi-ble. See supplementary material (SFig. 3) for R ( T ) ofdevices with different thickness of Co and DyN. Super-conducting transition of NbN in the Co(5 nm)-DyN(2.5nm) device can be seen at T C = 10.6 K as a sharp dropin the resistance. The superconducting coherence lengthof NbN in the dirty limit can be determined from theexpression; ξ NbN = (¯ hD NbN / πk B T C ) / . Using diffu-sion constant D NbN = 1.48 × − m s − and T C =10.6K we found ξ NbN = 4.1 nm (See Supplementary Infor-mation for calculation of diffusion constant) [44]. Thesuperconducting coherence length inside Co can be cal-culated from the expression; ξ Co = (¯ hD Co /k B T Curie ) / ; where D Co and T Curie are the diffusion constant andthe Curie temperature of Co, respectively. With D Co =6 × − m s − and T Curie = 1388 K, we found ξ Co =1.8nm [31]. The thickness of the Co in all our Co-DyN de-vices is ∼ ξ Co and ξ NbN .For conventional spin-singlet case the superconductingcorrelations decay over a length ξ Co in the diffusive fer-romagnet. Therefore, our NbN-FM-FI-NbN devices canbe considered as N-I-S type device instead of a S-N-I-Stype device. A. Tunneling behavior -3 -2 -1 0 1 2 3-15-10-5051015 I ( A ) V(mV) G ( V ) / G N -20 -10 0 10 201.01.11.2 (d)(c) (b)
40 K 30 K 20 K G ( V ) / G ( ) V(mV)(a) -3 -2 -1 0 1 2 3-200-1000100200
Co-DyN DyN Gd-GdN I ( A ) V(mV)
Co-DyN -5 -4 -3 -2 -1 0 1 2 3 4 5-50-2502550 I ( A ) V(mV) G ( V ) / G N G ( V )( a . u ) FIG. 2: (Color online) (a) G ( V ) = dI/dV conductance spec-tra of a Co(5 nm)-DyN(2 nm) tunnel junction measured at20, 30 and 40 K. The conductance spectra is normalized to G (0) (b) The IV and Normalized conductance spectra of thesame Co(5 nm)-DyN(2 nm) tunnel junction measured at 4.2K. The conductance spectra is normalized to normal stateconductance G N . (c) IV and normalized conductance spec-tra of a NbN-DyN-NbN tunnel junction measured at 4.2 K.(d)The IV and normalized conductance spectra of a Gd(5)-GdN(4.5) tunnel junction measured at 4.2 K. Current-voltage (I-V) measurements were done at dif-ferent temperatures to understand the nature of the elec-trical transport in the devices. Figure 2(a) shows theconductance G ( V )(= dI/dV ) normalized to its value at V = 0 for the Co(5 nm)-DyN (2.5 nm) device. Clearparabolic conductance spectra at 20, 30 and 40 K sug-gest tunnelling-type transport in this devices through alltemperature range. The dI/dV spectra of the deviceswere also measured at different temperatures below the T C of NbN. The I-V and normalized conductance spec-tra G ( V ) /G N of the same junction measured at 4.2 Kis shown in Fig. 2(b). Fully developed superconduct-ing gap structure with 2∆ ∼ G ( V ) /G N of a NbN-DyN-NbN tunnel junc-tion without Co. A superconducting gap 4∆ ∼ G ( V ) /G N of a Gd(5nm)-GdN(3 nm) device. The Superconducting gap wasfound to be 2∆ ∼ > > dI/dV spectra below the T C of NbN in our devices can be compared with NIS-type tunnel model. Normalized tunneling conductanceof a NIS junction at a bias voltage V can be written as: G s ( V ) G N ( V ) = dd ( eV ) ∞ Z −∞ N ( E )[ f ( E ) − f ( E + eV )] dE, (1)where f ( E ) is Fermi-Dirac distribution function and N ( E ) is the normalized BCS density of state of thesuperconductor. Here G N ( V ) is the normal state con-ductance of the junction. Following Dynes approach[45]the quasiparticle density of states can be written as, N ( E ) = N (0) (cid:12)(cid:12)(cid:12)(cid:12) Re (cid:18) E/ ∆ − i Γ √ ( E/ ∆ − i Γ) − (cid:19)(cid:12)(cid:12)(cid:12)(cid:12) . Here the smear-ing parameter Γ is included to consider finite lifetime ofquasiparticles. Quasiparticles have finite lifetime at nonzero temperature due to presence of some energy lev-els within the superconducting gap. Besides, impuritiesand pinholes in the tunnel barrier has also shown to con-tribute to Γ[46, 47]. In our case magnetism of the tunnelbarrier (FI) and normal electrode (FN) might further addto Γ.Figure 3(a) shows the temperature evolution of con-ductance spectra of the Co(5 nm)-DyN(2.5 nm) devicemeasured in the range 4.2 to 9.3 K. Fitting of the Eq. (1)(red solid line) to the conductance spectra measured at4.2 and 9.3 K is shown in Fig. 3(b) and (c), respectively.We found acceptable fit of the conductance spectra tothe NIS-tunneling model at all temperatures by adjustingsmearing parameter Γ. Fig. 3(d) shows plot of extractedfitting parameters Γ and ∆ at different temperature. Thered solid line shows a typical BCS type temperaturedependence[48]: ∆( T ) = ∆(0) tanh(1 . p ( T C − T ) /T ),with 2∆(0) = 2.97 meV and T C = 10.94 K. The in-crease in subgap conductance might be due to magnon-assisted Andreev reflection process which can provide ad-ditional channel for subgap transport[49]. Electromag-netic fluctuation in the environment can also lead to simi-lar situations[50, 51]. Moreover, we could not observe anyfeature related to 0 and π transition due to poor couplingbetween the two NbN layers. -3 -2 -1 0 1 2 30.00.51.0 G ( V ) / G N V(mV)4.2 K 9.3 K -4 -2 0 2 40.00.51.0 (d)(c) G ( V ) / G N V(mV)4.2 K -4 -2 0 2 40.81.0 G ( V ) V(mV)9.3 K ( m e V ) T(K) (b) (a)
FIG. 3: (Color online)(a) Normalized conductance spectra G ( V ) /G N of the Co(5 nm)-DyN(2 nm) tunnel device mea-sured in the temperature range 4.2 to 9.3 K. (b)NIS-tunnelingmodel fitting of the conductance spectra measured at 4.2K. (c)NIS-tunneling model fitting of the conductance spec-tra measured at 9.3 K (d)Temperature dependence of super-conducting gap ∆ and smearing parameter Γ obtained fromfitting. The red solid line shows fitting to the BCS tempera-ture dependence of ∆( T ). B. Spin-valve behavior
Figure 4(a) shows magnetic field dependence of resis-tance of the Co(5 nm)-DyN(2 nm) device measured at2 K. The measurement was done with a bias current I = 500 µ A. The resistance was found to increase withmagnetic field up to H p ∼ H p was found to vary from device to deviceand a complete decreasing trend was observed for de-vices with thicker DyN (see Supplementary figure SFig.6). Fig. 4(b) shows field dependence of magnetizationmeasured at 12 and 100 K of a NbN-Co(5 nm)-DyN(4.5nm)-NbN film deposited at the same time. As 4.5 nmthick DyN have very small magnetic moment comparedto Co the M-H loop is mostly dominated by Co. Fig. 4(c)shows field dependence of resistance in the field range ±
30 mT. One can clearly see resistance switching at ± ±
25 mT. Although, coercive field ofthe NbN-Co(5 nm)-DyN(4.5 nm)-NbN film is 7 mT, re-sistance switching at 25 mT in the Co(5 nm)-DyN(2 nm)device is due to the reduced dimension of the junction i.e.,7 µ m × µ m. We found slight variation in the switchingfield from device to device depending on the thickness ofDyN (See supplementary figure SFig. 5). This might bedue to the different nitridation of Co which makes effec-tive Co layer thinner. One striking feature to notice is alow resistance state for antiparallel configuration in com-parison to parallel configuration. Fig. 4(d) shows sim-ilar measurements on Gd(5 nm)-GdN(3 nm) device. Ahigher resistance for antiparallel state can be seen in thiscase. In Gd-GdN devices resistance switching was found only at low temperature below 13 K (see supplementaryfigure SFig.7). This might be due to the absence of awell defined magnetic decoupling layer at the interfacebetween Gd and GdN. In this case a gradient of nitrogendeficiency might separate Gd and GdN layer from eachother providing poor magnetic isolator. -0.50 -0.25 0.00 0.25 0.5014.014.114.214.314.414.514.614.7 Co-DyNGd-GdNCo-DyN (d)(c) (b) R () H(T)(a)
Co-DyNH p -50 -25 0 25 50-100-50050100
12K 100 K M ( e m u ) H(mT) H C ~ 7 mT-30 -20 -10 0 10 20 3014.8014.8514.9014.95 R () H(mT) -50 -25 0 25 50144148152 R () H(mT)
FIG. 4: (Color online)(a) Magnetic field dependence of resis-tance of a Co(5 nm)-DyN(2 nm) device measured at 2 K usinga current I = 500 µ A. (b) Magnetic field dependence of mag-netization at 12, 20, and 100 K for a NbN-Co(5 nm)-DyN(4.5nm)-NbN film deposited at the same time. (c)R-H loop of theCo(5 nm)-DyN(2 nm) device in the low-field range. (d)Fielddependence of resistance of a Gd(5 nm)-GdN(4.5 nm) devicemeasured at 2 K.
Figure 5 shows R-H loops of the Co(5 nm)-DyN(2 nm)device measured at different temperatures. The resis-tance switching was found to disappear as temperaturewas increased to 30 K. This is expected as DyN gets intoparamagnetic phase above T Curie ∼
35 K. Temperaturedependence of switching strongly suggest that the switch-ing arises due to the relative mutual magnetization of Coand DyN. We could not find any switching in devices withthicker DyN ( > x is mostlikely very thick which causes loss of spin polarizationof electrons after filtering through DyN. The high fieldmagnetoresistance seen in Figure 4(a,d) seems likely tooriginate from a field-enhanced magnetisation [33] andhence an exchange splitting which increases with field.The resistance switching in our devices can be under-stood considering spin dependent density of states (DOS)of different layers as shown in Fig. 6. When magnetiza-tion of Co and DyN are parallel to each other, the up-spinelectrons tunneling through DyN experience a lower bar-rier height compared to down-spin electrons. During tun-neling process spin orientation of electrons are conserved,therefore, up and down-spin electrons can tunnel onlyinto spin-up and spin-down states of Co, respectively[52].Similarly when magnetization of Co and DyN are an-tiparallel to each other down-spin electrons are filtered -100 -50 0 50 10021.7521.7821.81 (d)(c) (b)(a) R () H(mT)13 K -100 -50 0 50 10021.6821.7021.72 R () H(mT)20 K -100 -50 0 50 10020.9020.9120.9220.9320.94 R () H(mT)30 K -100 -50 0 50 10021.8121.8421.87 R () H(mT)15 K
FIG. 5: (Color online)Field dependence of resistance of theNbN-Co(5 nm)-DyN(2 nm)-NbN device measured at (a) 13K (b) 15 K, (c)20 K and (d) 30 K. Measurements were donewith a current I = 500 µA . E F E F (a) (b) eV DyN Co NbN eV GdN NbN Gd FIG. 6: (Color online) Schematic of the spin-resolved den-sity of state for the (a)Co-DyN and (b)Gd-GdN device. Thedensity of state of Co and Gd is shifted by eV along the en-ergy axis when a bias voltage V is applied. All spin-valvemeasurements were done with bias voltage eV > ∆. through. In the case of Co the DOS at the Fermi level E F for up-spin electrons is lower than DOS for down-spinelectrons[53]. Therefore the transparency through theDyN-Co bilayer will be higher for antiparallel configura-tion compared to parallel configuration. This will cause alow-resistance state for situation when magnetization ofCo and DyN are antiparallel to each other. This conceptcan be further verified in Gd-GdN devices where Co isreplaced by Gd. Schematic of spin dependent DOS of Gdis shown in Fig. 6(b). In case of Gd the up-spin DOS at E F is slightly higher than down-spin DOS[54]. Moreover,unlike Co in the case of Gd tunneling spin polarization is usually found positive experimentally[55]. Therefore, oneexpect a higher resistance for antiparallel configurationcompared to parallel configuration as observed in normalspin-valves[52]. The resistance switching behavior shownin Fig. 4 confirms our interpretation. The sign of thetunnel spin polarization is known to depend strongly onthe nature of the interface bonding which is likely dif-ferent between Co-DyN and Gd-GdN. A detailed bandstructure calculation is needed to understand differentsign of spin polarization in these interfaces. IV. CONCLUSIONS
In conclusion, we have fabricated NbN-FM-FI-NbNpsuedospin-valve devices (with FM = Co, Gd and FI=DyN, GdN) and made an extensive study of electri-cal transport measurements. Tunnelling regime wasachieved in these devices by optimizing thickness of theDyN layer. In the tunnelling regime quasiparticle tun-nelling spectra through the spin-valve were compared totunneling spectra of NIS tunnel model with a nonmag-netic barrier. We also measured R-H loop of the de-vices at different temperatures. Clear resistance switch-ing was observed in these devices corresponding to theirmutual magnetization direction. The resistance switch-ing was found to be sensitive to density of state (DOS)at the FM-FI interface. In the case of Co a low resis-tance state was found for antiparallel configuration dueto lower spin-up DOS at E F . A reverse behaviour withhigh resistance for antiparallel configuration was foundfor Gd devices. The resistance switching in these spinvalve measurements confirm the spin filtering nature ofDyN tunnel barrier. Optimization of such devices withideal deposition condition and interface can lead to hugeMR. Thickness of Co (Gd) and barrier transparency ofDyN (GdN) can be further tuned to create stronger cou-pling between the two NbN leading to a S-I-S type de-vice with a spin-valve sandwiched in between.This kindof devices can be tuned between 0 and π state throughmagnetization configuration of the spin valve[38, 58]. Acknowledgments
This work was done when PKM was supported by theERC Advanced Investigator Grant SUPERSPIN duringthe period June 2012-April 2015. PKM acknowledgesDr X. L. Wang for SQUID measurements and DeviceMaterials Group (DMG) for experimental facility. [1] J. Linder and J. W. A. Robinson, Nature Phys. , 307(2015).[2] M. G. Blamire and J. W. A. Robinson, Journal ofPhysics: Condensed Matter , 453201 (2014).[3] T. Wakamura, H. Akaike, Y. Omori, Y. Niimi, S. Taka-hashi, A. Fujimaki, S. Maekawa, and Y. Otani, Nature Materials , 675(2015).[4] Y. Makhlin, G. Sch¨oen, and A. Shnirman, Rev. Mod.Phys. , 357 (2001).[5] John Clarke and Frank K. Wilhelm, Nature , 1031(2008).[6] T. Yamashita, K. Tanikawa, S. Takahashi, and S. Maekawa, Phys. Rev. Lett. , 097001 (2005).[7] A. I. Buzdin, Rev. Mod. Phys. , 935 (2005).[8] V. V. Ryazanov, V. A. Oboznov, A. Yu Rusanov, A. V.Veretennikov, A. A. Golubov, and J. Aarts, Phys. Rev.Lett. , 2427 (2001).[9] T. Kontos, M. Aprili, J. Lesueur, and X. Grison, Phys.Rev. Lett. , 304 (2001).[10] L. N. Bulaevskii, V. V. Kuzii and A. A. Sobyanin, Pis’maZh. Eksp. Teor. Fiz. (1977) 314 [JETP Lett. (1977)290].[11] C. Bell, G. Burnell, C. W. Leung, E. J. Tarte, D.-J. Kangand M. G. Blamire, Appl. Phys. Lett. , 1153 (2004).[12] J. W. A. Robinson, G. B. Hal´asz, A. I. Buzdin, and M.G. Blamire, Phys. Rev. Lett. , 207001 (2010)[13] H. Sellier, C. Baraduc, F. Lefloch, and R. Calemczuk,Phys. Rev. Lett. , 257005 (2004).[14] Mohammad Alidoust and Klaus Halterman, Phys. Rev.B , 195111 (2014).[15] V. N. Krivoruchko and E. A. Koshina, Phys. Rev. B ,172511 (2001).[16] F. S. Bergeret, A. F. Volkov, and K. B. Efetov, Phys.Rev. Lett. , 3140 (2001); F. S. Bergeret,A. F. Volkovand K. B. Efetov, Rev. Mod. Phys. , 1321-1373 (2005).[17] Ivana Petkovic, Nikolai M. Chtchelkatchev, and ZoranRadovic,Phys. Rev. B , 184510 (2006).[18] M. Weides, H. Kohlstedt, R. Waser, M. Kemmler, J.Pfeiffer, D. Koelle, R. Kleiner, E. Goldobin, Appl.Phys. A , 613617 (2007); M. Weides,M. Kemmler,E. Goldobin, D. Koelle, R. Kleiner,H. Kohlsted, and A.Buzdin, Appl. Phys. Lett. , 122511 (2006).[19] T. I. Larkin, V. V. Bolginov, V. S. Stolyarov, V. V.Ryazanov, I. V. Vernik, S. K. Tolpygo, and O. A.Mukhanov, Appl. Phys. Lett. , 222601 (2012).[20] G. Wild, C. Probst, A. Marx, and R. Gross, Eur. Phys.J. B , 509523 (2010).[21] S. V. Bakurskiy, N. V. Klenov, I. I. Soloviev, V. V. Bolgi-nov, V. V. Ryazanov, I. V. Vernik, O. A. Mukhanov, M.Yu. Kupriyanov, and A. A. Golubov, Appl. Phys. Lett. , 192603 (2013).[22] S. V. Bakurskiy, N. V. Klenov, I. I. Soloviev, M. Yu.Kupriyanov, and A. A. Golubov, Phys. Rev. B ,144519 (2013).[23] I. V. Vernik, V. V. Bolginov, S. V. Bakurskiy, A. A.Golubov, M. Y. Kupriyanov, V. V. Ryazanov, and O. A.Mukhanov, IEEE Trans. Appl. Supercond. , 1701208(2013).[24] V.V. Ryazanov,V.V. Bolginov, D. S. Sobanin, I. V.Vernik, S. K. Tolpygo, A. M. Kadin, and O. A.Mukhanov, Physics Procedia , 35 (2012).[25] M. Eschrig, Phys. Today , 43 (2011).[26] I. Sosnin, H. Cho, V. T. Petrashov, and A. F. Volkov,Phys. Rev. Lett. , 157002 (2006).[27] J. W. A. Robinson, J. D. S. Witt, and M. G. Blamire,Science , 59 (2010).[28] T. S. Khaire, M. A. Khasawneh, W. P. Pratt, and N. O.Birge, Phys. Rev. Lett. , 137002 (2010).[29] R. S. Keizer, S. T. B. Goennenwein, T. M. Klapwijk, G.Miao, G. Xiao, and A. Gupta, Nature (London) , 825(2006).[30] M. S. Anwar, F. Czeschka, M. Hesselberth, M. Porcu,and J. Aarts, Phys. Rev. B , 100501 (2010).[31] J. Wang, M. Singh, M. Tian, N. Kumar, B. Liu, C. Shi,J. K. Jain, N. Samarth, T. E. Mallouk, and M. H. W.Chan, Nat. Phys. 6, 389 (2010). [32] P. K. Muduli, A. Pal, M. G. Blamire, Phys. Rev. B ,094414 (2014).[33] P. K. Muduli, X. L. Wang, J. H. Zhao, M. G.Blamire, arXiv:1410.6741 [34] A. Pal, K. Senapati, Z. H. Barber, M. G. Blamire, Adv.Mater. , 5581 (2013).[35] M. G. Blamire, A. Pal, Z. H. Barber, and K. Senapati,Proc. SPIE , 84610J (2012).[36] K. Senapati, M. G. Blamire, and Z. H. Barber, NatureMaterials , 849 (2011).[37] T. Nagahama, T. S. Santos, and J. S. Moodera,Phys.Rev. Lett. , 016602 (2007).[38] Burm Baek, William H. Rippard, Matthew R. Pufall,Samuel P. Benz, Stephen E. Russek, Horst Rogalla,and Paul D. Dresselhaus, Phys. Rev. Applied , 011001(2015).[39] P. Leclair, J. K. Ha, H. J. M. Swagten, J. T. Kohlhepp,C. H. van de Vin, W. J. M. de Jonge, Appl. Phys. Lett. , 625 (2002).[40] J. S. Moodera, T. S. Santos, and T. Nagahama, J. Phys.:Condens. Matter. , 165202 (2007).[41] Shiro Kawabata, Andrey S. Vasenko, Asier Ozaeta, Se-bastian F. Bergeret and Frank W. J. Hekkingb,Blythe,J. Magn. Magn. Mater. ,157 (2015).[42] F. S. Bergeret and F. Giazotto,Phys. Rev. B , 014515(2013).[43] Shiro Kawabata, Satoshi Kashiwaya, Yasuhiro Asano,Yukio Tanaka, and Alexander A. Golubov,Phys. Rev. B , 180502(R) (2006).[44] Kartik Senapati, Mark G. Blamire and Zoe H. Bar-ber,Appl. Phys. Lett. 103, 132406 (2013).[45] R. C. Dynes, V. Narayanamurti, and J. P. Garno, Phys.Rev. Lett. , 1509 (1978).[46] T. M. Klapwijk, G. E. Blonder, and M. Tinkham, Physica(Amsterdam) 109110B+C, 1657 (1982).[47] J. R. Schrieffer and J.W. Wilkins, Phys. Rev. Lett. 10,17 (1963).[48] H. Takayanagi and J. Nitta in Towards the Control-lable Quantum States: Mesoscopic Superconductivity andSpintronics , 2002, Page-169[49] G. Tkachov, E. McCann, V. I. Falko, Phys. Rev. B ,024519 (2001).[50] J. P. Pekola, V. Maisi, S. Kafanov, N. Chekurov, A.Kemppinen, Yu. Pashkin, O.-P. Saira, M. Mottonen, andJ. Tsai, Phys. Rev. Lett. , 026803 (2010).[51] A. Di Marco, V. F. Maisi, J. P. Pekola, F. W. J. Hekking,Phys. Rev. B 88, 174507 (2013).[52] M. Julliere,Phys. Lett. A , 225 (1975).[53] J.M.D. Coey, Magnetism and magnetic materials , Cam-bridge University press, Page-149.[54] B. N. Harmon and A. J. Freeman,Phys. Rev. B , 1979(1974).[55] P. M. Tedrow and R. Meservey,Phys. Rev. B , 318(1973).[56] E. Y. Tsymbal, O. N. Mryasov, P. R. Leclar, Journal ofPhysics: Condensed Matter 15, R109 (2003).[57] B. Baek, W. H. Rippard, S. P. Benz, S. E. Russek, P. D.Dresselhaus, Nature Communications , 3888 (2014).[58] E. C. Gingrich, Bethany M. Niedzielski, Joseph A. Glick,Yixing Wang, D.L. Miller, Reza Loloee, W. P. Pratt, Jr.,and Norman O. Birge, Nature Physics , 564 (2016).[59] M. Matsuoka, K. Ono, T. Inukai,Appl. Phys. Lett. ,977 (1986);K. Oda, T. Yoshio, K. Oda, J. Mater. Sci. , 2729 (1987). Supplementary Information
Spin-polarized quasiparticle tunneling in spin-filter pseudospin-valve devices
Pseudospin-valve devices were fabricated from NbN-FM-FI-NbN multilayered thin films deposited by DC sputteringmethod. Here FM = Co and Gd; FI = DyN and GdN. A series of devices were fabricated from Co(5 nm)-DyN( t )multilayer with different thickness of DyN. Few representative devices were fabricated in the configuration Co-GdNand Gd-GdN for comparison. More detailed experiment was done in the Co-DyN series compared to others. In thismanuscript a comparison has been made between devices with GdN and DyN for spin-valve measurements. Thecomparison is reasonable as GdN and DyN are similar ferromagnetic semiconductors with only different spin-filteringefficiency. Measurements done on different devices are summarized in the figures below. Calculation of Diffusion constant of NbN : R , () T(K) T C =10.8 K R , () T(K)
NbN (50 nm)
SFig. 1: (Color online)Temperature dependence of resistance of a 50 nm thick NbN film deposited under similar condition asthe devices. Inset shows measurement done in a different Van der Pauw configuration. Here residual resistivity RR( R RT /R K ) ≈ ρ (15 K ) ≈ µ Ω − cm . The diffusion constant can be calculated from Einstein relation; D = 1 e D ( E f ) ρ (2)Where D ( E f ) is the density of state at Fermi energy and ρ is the resistivity. The diffusion constant can also bewritten as; D = v F l e , with v F is Fermi velocity and l e is the average electron mean free path. For NbN with D ( E f ) = 1 . × eV − m − [Chockalingam et. al. Phys. Rev. B , 77, 214503.] and ρ = 2 . µ Ω − m , wefound D = 1 . × − m s − . R () T(K)NbN--Co(2.5 nm)-GdN(2 nm)-NbNI =100 A
SFig. 2: (Color online)Temperature dependence of resistance of a Co(2.5 nm)-GdN(2 nm) device measured with a current I=100 µ A. The R ( T ) in these devices may not be determined by GdN but more accurately the temperature dependence isdecided by the composite of Co-GdN. For devices with thinner Co, complete nitridation of Co cannot be ruled out. Therefore, R ( T ) may vary from device to device. R () T(K)
SFig. 3: (Color online)Temperature dependence of resistance of Co(2.5 nm)-DyN( t nm) devices with different thicknesses ofDyN. R () T(K)
I=10 ANbN-Gd(5 nm)-GdN(3 nm)-NbN
SFig. 4: (Color online)Temperature dependence of resistance of a Gd(5 nm)-GdN(3 nm) device measured with a bias currentI = 10 µ A. -30 -20 -10 0 10 20 3014.8014.8514.9014.95 R () H(mT) I = 500 AT= 2 K(a) Co(5 nm)-DyN(2 nm) -30 -20 -10 0 10 20 3014.8514.9014.9515.00 R () H(mT) T= 5 KI = 500 A(b) Co(5 nm)-DyN(2 nm) -30 -20 -10 0 10 20 3026.4326.4626.4926.52 R () H(mT) T = 10 K(c) Co(5 nm)-DyN(2 nm)I = 500 A -0.10 -0.05 0.00 0.05 0.1093949596 R () H(mT) I =100 AT= 5 K(d) Co(5 nm)-DyN(2.5 nm) -0.10 -0.05 0.00 0.05 0.10108.6109.2109.8 (e) Co(5 nm)-DyN(2.5 nm) R () H(mT) I =100 AT = 10 K -100 -50 0 50 10096.096.396.696.9 R () H(mT)T= 15 KI =100 A(f) Co(5 nm)-DyN(2.5 nm) -100 -50 0 50 1002300240025002600 (g) Co(5 nm)-DyN(3.5 nm) R () H(mT) T = 2 KI = 1 A -100 -50 0 50 1002500260027002800 R () H(mT) I =1 AT = 5 K(h) Co(5nm)-DyN(3.5 nm) -100 -50 0 50 100270028002900 (i) Co(5nm)-DyN(3.5 nm) R () H(mT) I = 1 AT= 10 K -100 -50 0 50 1002670270027302760 (j) Co(5nm)-DyN(3.5 nm)I = 1 A R () H(mT) T = 15 K
SFig. 5: The RH loop of the (a)-(c)Co(5 nm)-DyN(2 nm),(d)-(f)Co(5 nm)-DyN(2.5 nm) and (g)-(j) Co(5 nm)-DyN(3.5 nm)devices measured at different temperature. The resistance switching field can be seen between ± -1.0 -0.5 0.0 0.5 1.08.08.18.2 (a) Co(5 nm)/DyN(1 nm) R () H(T) T = 1.8 KI= 500 A -4 -3 -2 -1 0 1 2 3 44550556065707580859095 R () H(T) I =100 AT = 1.6 K(b)Co(5 nm)-DyN(2.5 nm) -4 -3 -2 -1 0 1 2 3 4102030 R ( k ) H(T) I = 1 AT= 5 K (c) Co( 5nm)-DyN(4.5 nm)
SFig. 6: The RH loop of the Co(5 nm)-DyN( t nm) device measured up to high field µ H = 4 T. -50 -40 -30 -20 -10 0 10 20 30 40 5029.430.130.8 R () H(mT) T= 5 KI= 500 A (a) Gd(8 nm)-GdN(3 nm) -50 -40 -30 -20 -10 0 10 20 30 40 5037.538.038.539.0 (b) Gd(8 nm)-GdN(3 nm)
T= 7 KI= 500 A R () H(mT) -50 -40 -30 -20 -10 0 10 20 30 40 5028.028.829.6 (c) Gd(8 nm)-GdN(3 nm)
T= 13 KI= 500 A R () H(mT)
SFig. 7: R-H loop of the Gd(8 nm)-GdN(3 nm) device measured at different temperature with bias current I = 500 µ A. -3 -2 -1 0 1 2 3-0.20-0.15-0.10-0.050.000.050.100.150.20 I ( A ) V(mV) T = 4.2 K (a) Co-(5 nm)-DyN( 4 nm) G ( V ) / G N -4 -3 -2 -1 0 1 2 3 4-0.50.00.5 I ( A ) V(mV) T = 4.2 K G ( V ) / G N (b) Co(5 nm)-DyN(3.5 nm) -4 -3 -2 -1 0 1 2 3 4-3-2-10123 G ( V ) / G N I ( A ) V(mV) (c) Co(5 nm)-DyN(3 nm)
T = 4.2 K -4 -3 -2 -1 0 1 2 3 4-40-30-20-10010203040 I ( A ) V(mV) G ( V ) / G N (d) Co(5 nm)-DyN(2.5 nm) T= 4.2 K -4 -3 -2 -1 0 1 2 3 4-300-200-1000100200300 G ( V ) / G N I ( A ) V(mV) (e) Co(5 nm)-DyN(2 nm)
T = 4.2 K -3 -2 -1 0 1 2 3-400-300-200-1000100200300400 G ( V ) / G N I ( A ) V(mV) T = 4.2 K (f) Co(5 nm)-DyN(1 nm)
SFig. 8: I-V and normalized conductance spectra ( G ( V ) /G N ) of Co-DyN devices with different thickness of DyN. -4 -3 -2 -1 0 1 2 3 4-40-2002040 I ( A ) V(mV) G ( V ) / G N (a) Co(5 nm)-DyN(2.5 nm) T = 4.2 KJn5 -4 -3 -2 -1 0 1 2 3 4-40-2002040 G ( V ) / G N (b) Co(5 nm)-DyN(2.5 nm) T = 4.2 K I ( A ) V(mV) Jn6 G ( V ) / G N T = 4.2 K(c) Co(5 nm)-DyN(2.5 nm) I ( A ) V(mV) Jn7 G ( V ) / G N (d) Co(5 nm)-DyN(2.5 nm)T = 4.2 K I ( A ) V(mV) Jn8
SFig. 9: I-V and normalized conductance spectra ( G ( V ) /G N ) of different Co-DyN device on the same chip. Each chip contain8 identical junctions. This shows the reproducibility from device to device is quite good. -4 -3 -2 -1 0 1 2 3 4-500-400-300-200-1000100200300400500 I ( A ) V(mV) T= 4.3 K G ( V ) / G N (a) Co(2.5 nm)-GdN( 2 nm) -4 -3 -2 -1 0 1 2 3 40.00.51.0 T = 4.2 K G ( V ) / G N V(mV)4.2 K 9.5 K(b) Gd(5 nm)-GdN(3 nm)
SFig. 10: I-V and normalized conductance spectra G ( V ) /G N of (a) Co-GdN and (b) Gd-GdN device device (The conductancespectra at different temperature is shifted lightly for clarity).
100 -50 0 50 100180184188 R () H(mT) I=1 A(a) Co(5nm)-DyN(2.5 nm)T = 1.6 K
100 -50 0 50 100919293 (b) Co( 5 nm)-DyN(2.5 nm) R () H(mT) I = 100 AT= 1.6 K
100 -50 0 50 10071.4571.5071.5571.6071.6571.7071.75 (c) Co( 5 nm)/DyN(2.5 nm)I = 500 A T= 1.6 K R () H(mT)
50 100 150 200 250 3000.210.280.35 R , () T(K) R , ()()