Reversible control of spin-polarized supercurrents in ferromagnetic Josephson junctions

Abstract

Magnetic inhomogeneity at a superconductor (S)-ferromagnet (F) interface converts spin-singlet Cooper pairs into spin-one triplet pairs. These pairs are immune to the pair-breaking exchange field in F and support a long-range proximity effect. Although recent experiments have confirmed the existence of spin-polarized triplet supercurrents in S-F-S Josephson junctions, reversible control of the supercurrent has been impossible because of the robust preconfigured nature of the inhomogeneity. Here, we use a barrier comprising three F layers whose relative magnetic orientation, and hence the interfacial inhomogeneity, can be controlled by small magnetic fields; we show that this enables full control of the triplet supercurrent and, by using finite element micromagnetic simulations, we can directly relate the experimental data to the theoretical models which provide a general framework to understand the role played by magnetic states in long-range supercurrent modulation.

ARTICLE

Received 21 Feb 2014 | Accepted 22 Jul 2014 | Published 26 Aug 2014

Magnetic inhomogeneity at a superconductor (S)ferromagnet (F) interface converts spin-singlet Cooper pairs into spin-one triplet pairs. These pairs are immune to the pair-breaking exchange eld in F and support a long-range proximity effect. Although recent experiments have conrmed the existence of spin-polarized triplet supercurrents in SFS Josephson junctions, reversible control of the supercurrent has been impossible because of the robust precongured nature of the inhomogeneity. Here, we use a barrier comprising three F layers whose relative magnetic orientation, and hence the interfacial inhomogeneity, can be controlled by small magnetic elds; we show that this enables full control of the triplet supercurrent and, by using nite element micromagnetic simulations, we can directly relate the experimental data to the theoretical models which provide a general framework to understand the role played by magnetic states in long-range supercurrent modulation.

DOI: 10.1038/ncomms5771

Reversible control of spin-polarized supercurrents in ferromagnetic Josephson junctions

N. Banerjee1, J.W.A. Robinson1 & M.G. Blamire1

1 Department of Materials Science and Metallurgy, University of Cambridge, 27 Charles Babbage Road, Cambridge CB3 0FS, UK. Correspondence and requests for materials should be addressed to N.B. (email: mailto:[email protected]

Web End [email protected] ).

NATURE COMMUNICATIONS | 5:4771 | DOI: 10.1038/ncomms5771 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 1

& 2014 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5771

The interplay between superconducting and magnetic order parameters constrained by the exclusion principle and fermionic exchange statistics has given rise to rich and

diverse physics and reignited the interest in the problem of coexistence of magnetism and superconductivity1,2. Of particular relevance is the theoretical prediction3 that magnetic inhomogeneity at a SF interface leads to a conversion between singlet and triplet spin pairing states in different quantization bases and produces equal spin Cooper pairs. Recent experimental verication415 of long-ranged supercurrents in ferromagnets has raised the intriguing possibility of taking the next step towards practical implementation as a dissipation-less version of spin electronics (spintronics)16. Two key aspects need to be addressed for a realization of such circuits: efcient generation of spin-polarized supercurrents, and their active control. Efforts1719 in the last few years have been primarily directed towards optimizing the supercurrent; little progress has so far been made in directly controlling it. Optimizing the inhomogeneity in the form of a robust spin-mixer layer that maximizes the singlet to triplet conversion ironically appears to make it difcult to design an externally controllable system.

In this article, we report SF0FF0S Josephson junctions in which the magnetic alignment between thin F0 mixer layers (composed of the soft ferromagnet Ni80Fe20, Permalloy, Py) and a thicker

F layer (Co) can be controlled by the applied magnetic eld (H) and show that the magnitude of the critical current IC is

controlled by the net misalignment of the magnetism in the three layers. In particular, we show that the supercurrent is zero for the parallel aligned case (Fig. 1a,b). This device is the super-conducting analogue of the spin valve, which is the foundation of conventional spintronics20. We analyse our results based on the Houzet and Buzdin model21 of a Josephson junction

incorporating a trilayer magnetic structure which, in combination with nite element analysis, provides a semi-quantitative t to the data.

ResultsTransport measurements of SF0FF0S Josephson junctions.

Figure 2a shows the IC in a Josephson junction with a Py(1.5)/ Cu(5)/Co(5.5)/Cu(5)/Py(1.5) (thicknesses in nanometres) barrier. The behaviour is distinctly different from the expected dependence of IC on H (IC(H)) in a SFS Josephson junction: an example is shown in Fig. 2b from a junction incorporating Ho mixer layers at the S/F interface but having similar Co layer thickness (6 nm) and comparable dimensions in which, although hysteretic, IC(H)

600

400

200

0

400

200

0

Critical current, I C(A)

Spin mixers

Spin mixers

1,200

900

S F F

F S

S F F

F S

600

300

0

80 60 40 20 20 40 60 80

0

Applied field, 0H (mT)

Figure 1 | SF0FF0S Josephson junction containing a trilayer ferromagnet. (a) At high magnetic elds, the F layers are parallel and the combined

F layer thickness is much greater than the coherence length of the singlet Cooper pairs; no supercurrent ows through the structure.(b) At zero or low magnetic elds, the inhomogeneous or non-collinear F0 layers converts the spin-singlet Cooper pairs in S to equal spin-triplet

Cooper pairs in F thus allowing a nite triplet supercurrent to ow through the structure.

Figure 2 | Dependence of the critical current on applied magnetic eld of a Josephson junction. (a) The junction is composed of a stack of Nb(250 nm)/Cu(5 nm)/Py(1.5 nm)/Cu(5 nm)/Co(5.5 nm)/Cu(5 nm)/ Py(1.5 nm)/Cu(5 nm)/Nb(250 nm). (b) Josephson junction with 6 nm central Co layer but having 4.5-nm thick Ho layers at Nb/Co interface instead of Py to generate spin-polarized supercurrents. It shows a Fraunhofer-like dependence of the junction critical current with prominent side lobes. (c) A Josephson junction having a layer sequence similar toa but with a reduced central Co thickness of 3 nm showing oscillations of the critical current beyond the rst lobe. The approximate dimensions of all the junctions are 600 500 nm. The red dotted lines in ac represent the

shift in the zero critical current line due to the nite non-zero voltage used

to measure the critical current (see Methods for details).

2 NATURE COMMUNICATIONS | 5:4771 | DOI: 10.1038/ncomms5771 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2014 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5771 ARTICLE

clearly follows the expected Fraunhofer-type dependence with distinct second lobes. In Fig. 2a and, Fig. 3a we observe that the overall shape and behaviour of IC(H) is very different: IC goes to zero above a certain eld magnitude instead of showing multiple oscillations with eld (conrmed by the linear currentvoltage characteristic recorded at 40 mT (Fig. 3a inset)); the small rise

seen at high elds is associated with thermal effects arising from the magnet coil. On reducing the central Co layer thickness to 3 nm, which enables a singlet contribution to the supercurrent, although the central peak remains strongly distorted, additional lobes reappear beyond the rst minima (Fig. 2c).

Although unusual IC(H) patterns have been reported before in SFS, SF0FF0S or SIFS Josephson junctions17,2224, these are irreproducible and attributed to stochastic variations of the ux arising from a multi-domain magnetic barrier. This is distinctly

different from what we observe here: a highly reproducible but strongly distorted central peak with zero critical current beyond a certain magnetic eld value.

We start the discussion of these results by setting an upper limit for the singlet IC through such devices. In general, the singlet

IC in SFS Josephson junctions will be oscillatory with multiple 0p transitions with increasing F thickness25,26, but to provide an estimate of the upper limit of the singlet current we just consider the envelope of the IC maximain other words, assuming that the net exchange energy of the barrier is such that singlet pair dephasing is zero and that the supercurrent is just limited by the coherence lengths. The singlet coherence lengths xCo and xPy have been measured to be 3.0 and 1.4 nm, respectively27, meaning that the total F barrier thickness is equivalent to 12 nm of Co for the device shown in Figs 2a and 3a. Taking a typical value of 1.5 mV

Critical current, I C(A)

Normalized I C

100 0

3,000 2,000 1,000

1,000 2,000 3,000

0

600

500 400 300 200

0.08

0.06

0.04

0.02

0.00

0.02

0.04

0.06

0.08

V (mV)

1.2 Maximum critical current

Flux-induced variation, I =1

H (mT)

IC0=sin([p10]1)sin ([p10]2)

I(A)

0.6

80604020 0 20 40 60 80

1.0

iv

I(a.u.)

I(a.u.)

0.0

0.4

0.2

0.8

0.6

0.8

iii

0.6

0.4

0.4

0.2

0.2

0.0

0.0

80 60 40 20 0 20 40 60 80 100 120

v ii iApplied field, 0H (mT)

v iv iii ii i

Py Co

Py

25 mT

10 mT

0 mT

10 mT 25 mT

Figure 3 | Experimental and simulated critical current variation with in-plane magnetic eld. (a) Critical current versus in-plane magnetic eld of a Nb/Cu(5 nm)/Py(1.5 nm)/Cu(5 nm)/Co(5.5 nm)/Cu(5 nm)/Py(1.5 nm)/Cu(5 nm)/Nb junction measured at 4.2 K. The red dotted line in a representing the shift in the zero critical current line due to the nite non-zero voltage used to measure the critical current (see Methods for details). The inset shows the currentvoltage characteristic of the junction recorded at 40 mT to verify the absence of critical current. (b) Simulated IC(H) pattern (green and

brown curves) showing the combined effect of inhomogeneous magnetic state giving rise to a spin-polarized supercurrent and the effect of the ux taking into account the magnetic inhomogeneity. The green curve takes into account the actual sign of sin f1 sin f2 and thus accounts for the sign of the supercurrent depending on local 0 or p states, whereas the brown curve only takes the modulus of sin f1 sin f2. Inset shows the variation of the maximum supercurrent (IC0) in the junction and the combined effect of ux arising from an inhomogeneous barrier moment and the applied eld on the critical current as a function of an in-plane applied magnetic eld. (c, iv) The plan views of the magnetic states (from OOMMF simulations) for outer Pyand central Co layers are shown with the corresponding magnetic elds as indicated below. The states corresponding to the eld values shown are also marked in b.

NATURE COMMUNICATIONS | 5:4771 | DOI: 10.1038/ncomms5771 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 3

& 2014 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5771

as the characteristic voltage (ICRN) in such junctions (extrapolated for 12 nm Co thickness) previously observed for Nb/Co/Nb26, gives a maximum singlet IC of B 40 mA. This value does not take into account the additional scattering at the multiple interfaces in our structures28. To take account of these interfaces and at least partial cancellation of the dephasing, a more representative number might be obtained by extrapolating from similar sized junctions with much thinner Py(1.6)/Cu(8)/ Co(1) ferromagnetic barriers29, for which the ICRN varied between 0.8 and 2 mV thus, giving a maximum IC of B60 mA when the two F layers were AP. In our devices, the ICRN ranged from 4 to 11 mV with a corresponding IC of B500600 mA. We therefore conclude that the supercurrents cannot originate from singlet pair transport and so must be primarily mediated by spin-one triplet pairs.

The non-collinearity between the adjacent F layers that is required for triplet generation21 arises from the complex magnetic microstructure of the F layers, which itself is due to a competition between the dipolar eld3032, magnetic anisotropies and the external eld. Since this microstructure changes with the applied eld, the maximum critical current, IC0 should depend on H.

For our junctions, the IC(H) modulation is controlled by two factors: rst, the eld-dependent magnetic inhomogeneity determines the maximum triplet supercurrent IC0 and, second, phase variations arising from the applied eld and induced changes to the net barrier moment determine the net IC (which leads to the Fraunhofer IC(H) modulation seen in conventional junctions). Both of these factors depend on the details of the micromagnetic conguration of each magnetic layer and so it is necessary to understand how this depends on H.

Finite element analysis of IC dependence on magnetic eld. Experimentally, it is hard to directly visualize these states in sub-micron devices and, although previous SFS experiments have used indirect information from magnetic measurements of unpatterned lms, the much weaker role of dipolar elds in continuous lms means that it is impossible to directly relate the details of the micromagnetic structure of nanopillar devices from such measurements. Instead, we have used nite element micromagnetic simulations33 using Object Oriented Micromagnetic Framework (OOMMF), which allows simulation of the magnetic state up to a resolution of few nanometres and make semi-quantitative predictions relating the magnetic structure to the spin-polarized supercurrent owing through the device. The saturation magnetization, exchange coefcient and uniaxial anisotropy for Co were set to 1,400 103 Am 1,

3 10 11 Jm 1 and 208 103 Jm 3, respectively, while for

Py these values were 860 103 Am 1, 1.3 10 11 Jm 1 and

150 Jm 3, respectively. The saturation magnetization and exchange coefcient values were taken from the OOMMF database (used in the literature); the saturation magnetization agrees closely with the values we have calculated from bulk lms deposited on SiO2 substrates sandwiched between 100-nm thick

Cu. The uniaxial anisotropy for Py was calculated from the difference in area of the hysteresis loops measured along the hard and easy axis and the direction, originally set by the growth eld, was orthogonal to the applied eld H. However, it is seen that the dipolar energy term in this case is much larger than the Py anisotropy energy and alone dictates the ground state conguration. To determine the value of Co anisotropy, we have simulated a spin valve structure consisting of Co (1.5)/Cu(7.5)/Py (1.5) similar to the one used in ref. 29, which was grown under similar conditions and modied the Co anisotropy value to match the switching eld obtained from magnetoresistance

measurements on these structures. The value obtained from these simulations is 40% of the reported value in OOMMF database; this is not unexpected since the anisotropy strongly depends on the growth conditions, the substrate used and the lm thickness34. The Co anisotropy was in the plane of the layer and the direction was chosen from a random vector eld, which reects the polycrystalline nature of the sputtered lms. The damping coefcient was set to 0.5, which allowed for rapid convergence.

Figure 3a shows one branch (positive to negative eld sweep) of IC(H) for a device. Micromagnetic simulations for this device have been performed at 5-mT intervals for an equivalent eld sweep: Fig. 3c shows plan views of the magnetic structure of each layer at representative elds. The colour scheme adopted to represent magnetization direction is red-white-blue with red (blue) pixels representing magnetization aligned along the positive (negative) external eld direction. White pixels represent magnetic moments orthogonal to the applied eld direction. At the highest eld magnitudes, the three F layers are parallel. Around 10 mT, the Py layers start inhomogeneously reversing under the dipolar magnetostatic interaction from the Co layer and are fully reversed at zero eld. As the eld increases in the negative direction the Co layer eventually reverses beyond

10 mT. It is clear from the images that signicant noncollinearity exists within and between all layers during the reversal process.

This observation is important in its own right as there have been speculations about the specic origin of spin-polarized supercurrents in SF0FF0S devices before. Although it was concluded by Khasawneh et al.23 that non-collinearity between F0 and F layers most likely gives rise to the spin-polarized supercurrents rather than inhomogeneity in F0 layers, our simulations indicate a more subtle effect at play. Intuitively one might be inclined to believe that there is little inhomogeneity in nano-pillar devices, but it is evident here that inhomogeneity does exist and if engineered properly using F layer with difference in coercivities, this can be translated to a local non-collinearity between Py and Co layers, which is critical for spin-polarized supercurrent generation.

To proceed further, a quantitative estimate the magnetic inhomogeneity as a function of H is required to estimate the IC0

through the junction. According to the HouzetBuzdin model, IC0 for a Josephson junction at a xed temperature with a F10FF20 barrier is proportional to the product of the sines of the angles between adjacent magnetic layers (f1 and f2), that is,

IC0 / sin f1 sin f2: 1 Since the F layers cannot be approximated by a macrospin, it implies that we have to apply the model by calculating the product of the sine of the angle between the cells of two adjacent magnetic layers for each vertical cell stack used in the simulations within which a continuum approximation implies a uniform magnetization. The components of the magnetization in each cell are known from the OOMMF simulation and sin f1 is obtained from the inner product of the magnetization in the ith cell of the top Py with the corresponding cell in Co. The same procedure is repeated for the ith cell of the bottom Py and Co to obtain sin f2.

The product sin f1 sin f2 indicates the combined inhomo

geneity arising from the three F layers (outer Py layers and the central Co layer). This procedure is repeated for each cell in the entire layer and an average value of sin f1 sin f2 is obtained by

summing the product for all the cells and dividing by the total number of cells. We have taken into account the actual sign of the product sin f1 sin f2, since according to the HouzetBuzdin

model, the junction can be in a 0 (p) state depending on the anti-parallel (parallel) orientation of the magnetization of the

4 NATURE COMMUNICATIONS | 5:4771 | DOI: 10.1038/ncomms5771 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2014 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5771 ARTICLE

outer layers. This is clear from the micromagnetic simulations shown in Fig. 3c; at low elds, local regions of the junctions are in a 0 or p state thereby reducing the total critical current through the junction. Also, the dependence of the critical current on the relative angle between two F layers reects the fact that noncollinearity induced by inhomogeneity between two F layers is more important than inhomogeneity in a single F layer where it occurs at the scale of the magnetic exchange length, which far exceeds the coherence length of a Cooper pair in the F layer.

Figure 3b (inset) shows the dependence of IC0 on H: there are

two distinct peaks (indicating maximum inhomogeneity), with the rst peak at a positive eld (B10 mT) related primarily to Py reversal, while the second (B 10 mT) is due to the Co layer

reversal.

To calculate the phase variation owing to the local ux density B arising from a combination of the inhomogeneous barrier magnetization and H, we integrate the variation of the phase difference of the superconducting order parameter (f) over the junction area

ICIC0 ZS

sin f0

where

DH 2 l dFM

bm0MdFM; 6

b is the effective coupling of the ux originating from the saturation magnetization of the ferromagnet M, and dFM is the

thickness of the ferromagnet. We can estimate b from Nb/Ho/Co/ Ho/Nb junctions with a similar size and shape: the inset to Fig. 3b in ref. 4 shows the eld offset (DH) versus dCo and a linear t to (DH versus dCo/(2*l dCo) (Fig. 3b inset, ref. 4) gives b 0.2.

This implies a signicant partial cancellation of the magnetization ux arising from fringing elds producing a ux in the opposite direction in the region within the penetration depth of the superconductor. Using b 0.2, IC calculated from equation (5)

with IC0 1 is shown in Fig. 3b (inset); the distortion of an ideal

Fraunhofer pattern arises due to eld-dependent inhomogeneous magnetism of the barrier. Figure 3b (green curve) shows the full solution of equation (5) by including IC0(H) shown in the other inset and thus shows the combined effect of the dependence of the triplet supercurrent on the magnetic structure and the ux arising from the inhomogeneous magnetic barrier.

DiscussionInspection of Fig. 3b shows that several features of the experimental curve are well reproduced. These include the rapid decay of IC above a critical eld and the severely suppressed higher order lobes as a result of a more homogenous magnetic structure where IC0-0. The small remnant oscillations in the simulated curve arises from a residual inhomogeneity at the edges arising from dipolar elds between Py and Co layers, which always remain in the simulation but, experimentally may not contribute because of surface oxidation and intermixing arising from the ion-milling during fabrication. The dip near zero eld (less prominent in the experimental curve) is quite sensitive to the magnetic conguration of each layer. At low elds, such congurations are quite prone to stochastic variations induced by factors like the lm microstructure, exact device dimensions and magnetic history, and direct comparison with experiments are difcult to make in that eld region. Ideally, the low and zero eld conguration is expected to be symmetric with respect to the magnetic state of the outer layers and noncolinear to the central Co layer; this implies that globally the junction is in a p state and the sign of the product of sin f1 sin f2 is irrelevant. Taking this fact into consideration, we have simulated the same junction (Fig. 3b, brown curve). The two simulations differ only at or near zero eld. This brown curve, therefore, provides an upper limit to the critical current close to zero eld for a junction with homogeneously symmetric (or antisymmetric) outer Py layers.

In view of the above, the behaviour of devices with thinner Co (Fig. 2c) becomes clear: on reducing the central Co layer a background singlet current ows whose maximum value is insensitive to the magnetic state in the device and is thus visible as phase-controlled IC oscillations beyond the central lobe.

From the point of view of applications, the key aspect of this result is the experimental proof that the triplet supercurrent amplitude can be reversibly controlled by changing the magnetic inhomogeneity within the barrier. This provides direct control over the spin-polarized supercurrent, which is of fundamental importance towards the realization of practical superconducting spintronic circuits. Perhaps equally as importantly, we demonstrate that signicant inhomogeneity can be generated even in nanoscale junctions and appropriate engineering of the micro-magnetic structure offers the potential to optimize the response of the system to very small eld changes or spin transfer torques3638.

2

64

Za

8 >

<

>

:

2e

0

B

@

l d2

l d2

Bdz

Z

1

C

A

bz

9 >

=

>

;

dl

3

75dS;

2

where l is the penetration depth of the superconductor,

bz is the direction normal to the plane of the junction and e is the charge of an electron. Here, the line integral is carried out for all the points a dening the junction by starting from the origin where f0 is dened. The critical current is nally obtained by maximizing with respect to f0 of the surface integral dening the junction over the points a. The effective value of l for our materials and geometry is estimated to be 90 nm (by measuring the eld corresponding to one ux quantum) from devices with similar dimensions and Nb thicknesses but with Ho as the triplet generators instead of Py (Fig. 2b). Given the complex magnetization distribution in our junctions, a simple analytical solution to equation (2) is not possible and so we apply a numerical technique35 as outlined below:

The local B elds obtained from the micromagnetic simulations when integrated vertically normal to the plane of the layers for the whole barrier thickness including the London penetration depth (l) of the Nb electrodes, gives a linear ux density matrix (Cij) according to

Cij Z

l d2

l d2

Bdz: 3

Here z is the direction normal to the plane of the lms. The Cij

matrix is then converted to an equivalent matrix of phase gradients (Fij0) according to

F0ijCij

bz: 4 The critical current is then obtained by performing the summations

ICIC0

Xx Xysin f0 2e

Xxi1F0i1 dx 2e

!

:

5

Xyj1F0xj dy

and maximizing with respect to f0, where f0 is set at (1,1).

To compare our simulations with experimental data, it is necessary to know the effective coupling of the ux originating from the barrier magnetization into the junction. In an SFS junction with a single, homogeneously magnetized ferromagnetic barrier the maximum critical current is achieved when H DH

NATURE COMMUNICATIONS | 5:4771 | DOI: 10.1038/ncomms5771 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 5

& 2014 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5771

Methods

Film growth. Nb(250)/Cu(5)/Py(y)/Cu(5)/Co(x)/Cu(5)/Py(y)/Cu(5)/Nb(250) (thicknesses in nanometres) samples were grown on unheated (001) Si substrates with a 250-nm thick SiO2 coating by dc magnetron sputtering in ultra-high vacuum chamber. The base pressure was maintained below 10 8 Pa while the chamber was cooled via a liquid nitrogen jacket. The targets were pre-sputtered for 1520 min to clean the surfaces and the lms were grown in 50 mT (approximately) magnetic eld by placing the substrates between two bar magnets. This induces an easy axis for the Py lms along the growth-eld direction. The Cu layer between the base Nb and Py was inserted to improve the magnetic properties of Py. The Co thickness (x) was varied between 3 and 9 nm.

Device fabrication. Devices were prepared with either 1.5 or 2.5 nm Py layers (y); in general, these showed similar results. Standard optical lithography and Ar-ion milling were used to dene 4-mm-wide tracks, which were narrowed down by focused-ion-beam milling to make current-perpendicular-to-plane devices: details of the process are described elsewhere39. The average device dimensions were in the range of 600 500 nm.

Transport measurements. A custom-built liquid He dip probe was used to cool the devices down to 4.2 K by dipping it in a liquid He dewar. Currentvoltage characteristics were measured by a four-point technique using a current-biased circuit attached to a lock-in amplier. The Josephson effect in the devices was measured by applying an in-plane magnetic eld and measuring the critical current IC as a function of the applied eld (H) (Fig. 2a). The critical current was determined using a voltage criterion and hence a nite value is recorded even in the absence of a supercurrent. To subtract this background contribution, we have divided this criterion voltage by the normal state resistance of the junction, which shifts the effective zero critical current line to the values shown by the red dotted line in each gure. The eld was applied perpendicular to the Py easy axis which gives a weak tendency of the Py to align itself perpendicular to the Co layer at low or zero external elds.

References

1. Buzdin, A. I. Proximity effects in superconductor-ferromagnet heterostructures. Rev. Mod. Phys. 77, 935976 (2005).

2. Bergeret, F. S., Volkov, A. & Efetov, K. Odd triplet superconductivity and related phenomena in superconductor-ferromagnet structures. Rev. Mod. Phys. 77, 13211373 (2005).

3. Bergeret, F. S., Volkov, A. & Efetov, K. Long-range proximity Effects in superconductor-ferromagnet structures. Phys. Rev. Lett. 86, 40964099 (2001).

4. Robinson, J. W. A., Witt, J. D. S. & Blamire, M. G. Controlled injectionof spin-triplet supercurrents into a strong ferromagnet. Science 329, 5961 (2010).

5. Khaire, T. S., Khasawneh, M. A., Pratt, W. P. & Birge, N. O. Observation of spin-triplet superconductivity in Co-based Josephson junctions. Phys. Rev. Lett. 104, 137002 (2010).

6. Keizer, R. S. et al. A spin triplet supercurrent through the half-metallic ferromagnet CrO2. Nature 439, 825827 (2006).

7. Anwar, M. S., Czeschka, F., Hesselberth, M., Porcu, M. & Aarts, J. Long-range supercurrents through half-metallic ferromagnetic CrO2. Phys. Rev. B 82, 100501 (2010).

8. Banerjee, N. et al. Evidence for spin selectivity of triplet pairs in superconducting spin valves. Nat. Commun. 5, 3048 (2014).

9. Sprungmann, D., Westerholt, K., Zabel, H., Weides, M. & Kohlstedt, H. Evidence for triplet superconductivity in Josephson junctions with barriers of the ferromagnetic Heusler alloy Cu2MnAl. Phys. Rev. B 82, 060505 (2010).

10. Kalcheim, Y., Millo, O., Egilmez, M., Robinson, J. W. A. & Blamire, M. G. Evidence for anisotropic triplet superconductor order parameter in half-metallic ferromagnetic La0.7Ca0.3MnO3 proximity coupled to superconducting

Pr1.85Ce0.15CuO4. Phys. Rev. B 85, 104504 (2012).

11. Visani, C. et al. Equal-spin Andreev reection and long-range coherent transport in high-temperature superconductor/half-metallic ferromagnet junctions. Nat. Phys. 8, 539543 (2012).

12. Kalcheim, Y., Kirzhner, T., Koren, G. & Millo, O. Long-range proximity effect in La2/3Ca1/3MnO3/(100)YBa2Cu3O7-d ferromagnet/superconductor bilayers: evidence for induced triplet superconductivity in the ferromagnet. Phys. Rev. B 83, 064510 (2011).

13. Anwar, M. S. & Aarts, J. Inducing supercurrents in thin lms of ferromagnetic CrO2. Supercond. Sci. Technol. 24, 024016 (2011).

14. Sosnin, I., Cho, H., Petrashov, V. & Volkov, A. Superconducting phase coherent electron transport in proximity conical ferromagnets. Phys. Rev. Lett. 96, 157002 (2006).

15. Usman, I. T. M. et al. Evidence for spin mixing in holmium thin lm and crystal samples. Phys. Rev. B. 83, 144518 (2011).

16. Eschrig, M. Spin-polarized supercurrents for spintronics. Phys. Today 64, 43 (2011).

17. Gingrich, E. C. et al. Spin-triplet supercurrent in Co/Ni multilayer Josephson junctions with perpendicular anisotropy. Phys. Rev. B 86, 224506 (2012).

18. Klose, C. et al. Optimization of spin-triplet supercurrent in Ferromagnetic Josephson junctions. Phys. Rev. Lett. 108, 127002 (2012).

19. Robinson, J. W. A., Chiodi, F., Egilmez, M., Halsz, G. B. & Blamire, M. G. Supercurrent enhancement in Bloch domain walls. Sci. Rep. 2, 699 (2012).20. Chappert, C., Fert, A. & Van Dau, F. N. The emergence of spin electronics in data storage. Nat. Mater. 6, 813823 (2007).

21. Houzet, M. & Buzdin, A. I. Long range triplet Josephson effect through a ferromagnetic trilayer. Phys. Rev. B 76, 060504 (2007).

22. Khaire, T., Pratt, W. & Birge, N. O. Critical current behavior in Josephson junctions with the weak ferromagnet PdNi. Phys. Rev. B 79, 094523 (2009).

23. Khasawneh, M. A., Khaire, T. S., Klose, C., Pratt, Jr W. P. & Birge, N. O. Spin-triplet supercurrent in Co-based Josephson junctions. Supercond. Sci. Technol. 24, 024005 (2011).

24. Weides, M. Magnetic anisotropy in ferromagnetic Josephson junctions. Appl. Phys. Lett. 93, 052502 (2008).

25. Kontos, T. et al. Josephson junction through a Thin ferromagnetic layer: negative coupling. Phys. Rev. Lett. 89, 137007 (2002).

26. Robinson, J. W. A., Piano, S., Burnell, G., Bell, C. & Blamire, M. G. Critical current oscillations in strong ferromagnetic p junctions. Phys. Rev. Lett. 76, 177003 (2006).

27. Robinson, J. W. A., Piano, S., Burnell, G., Bell, C. & Blamire, M. G. Zero to p transition in superconductor-ferromagnet-superconductor junctions. Phys. Rev. B 76, 094522 (2007).

28. Robinson, J. W. A., Halsz, G. B., Buzdin, A. I. & Blamire, M. G. Enhanced supercurrents in Josephson junctions containing nonparallel ferromagnetic domains. Phys. Rev. Lett. 104, 207001 (2010).

29. Bell, C. et al. Controllable Josephson current through a pseudospin-valve structure. Appl. Phys. Lett. 84, 1153 (2004).

30. Aziz, A. et al. Nonlinear giant magnetoresistance in dual spin valves. Phys. Rev. Lett. 103, 237203 (2009).

31. Banerjee, N. et al. Thickness dependence and the role of spin transfer torque in nonlinear giant magnetoresistance of permalloy dual spin valves. Phys. Rev. B 82, 224402 (2010).

32. Cowburn, R. P. & Welland, M. E. Phase transitions in planar magnetic nanostructures. Appl. Phys. Lett. 72, 2041 (1998).

33. Donahue, M. J. & Porter, D. G. OOMMF Users guide, Version 1.0, Interagency Report NIST IR 6376 (Gaithersburg, MD (1999).

34. Yang, W., Lambeth, D. N. & Laughlin, D. E. Dependence of Co anisotropy constants on temperature, processing, and underlayer. J. Appl. Phys. 87, 68846886 (2000).

35. Blamire, M. G., Smiet, C. B., Banerjee, N. & Robinson, J. W. A. Field modulation of the critical current in magnetic Josephson junctions. Supercond. Sci. Technol. 26, 055017 (2013).

36. Linder, J., Yokoyama, T. & Sudb, A. Spin-transfer torque and magnetoresistance in superconducting spin valves. Phys. Rev. B 79, 224504 (2009).

37. Linder, J. & Halterman, K. Superconducting spintronics with magnetic domain walls. Preprint at http://arxiv.org/abs/1401.5806

Web End =http://arxiv.org/abs/1401.5806 (2014).

38. Braude, V. & Blanter, Y. Triplet Josephson effect with magnetic feedback in a superconductor-ferromagnet heterostructure. Phys. Rev. Lett. 100, 207001 (2008).

39. Bell, C. et al. Fabrication of nanoscale heterostructure devices with a focused ion beam microscope. Nanotechnology 14, 630 (2003).

Acknowledgements

M.G.B. acknowledges funding from the UK EPSRC (EP/I038047/1) and the European

Commission through an ERC Advanced Investigator Grant Superspin. J.W.A.R.

acknowledges funding from the Royal Society.

Author contributions

N.B. prepared the samples, fabricated the devices and carried out the measurements. N.B.

performed the micromagnetic simulations and together with M.G.B. carried out the nite

element simulations. N.B., M.G.B. and J.W.A.R. wrote the manuscript.

Additional information

Competing nancial interests: The authors declare no competing nancial interests.

Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

Web End =http://npg.nature.com/

http://npg.nature.com/reprintsandpermissions/

Web End =reprintsandpermissions/

How to cite this article: Banerjee, N. et al. Reversible control of spin-polarized

supercurrents in ferromagnetic Josephson junctions. Nat. Commun. 5:4771

doi: 10.1038/ncomms5771 (2014).

6 NATURE COMMUNICATIONS | 5:4771 | DOI: 10.1038/ncomms5771 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2014 Macmillan Publishers Limited. All rights reserved.