WorksPhysics and Astronomy
Supercurrent decay in ballistic magnetic Josephson junctions
Abstract
We investigate transport properties of ballistic magnetic Josephson junctions and establish that suppression of supercurrent is an intrinsic property of the junctions, even in absence of disorder. By studying the role of ferromagnet thickness, magnetization, and crystal orientation we show how the supercurrent decays exponentially with thickness and identify two mechanisms responsible for the effect: (i) large exchange splitting may gap out minority or majority carriers leading to the suppression of Andreev reflection in the junction, (ii) loss of synchronization between different modes due to the significant dispersion of the quasiparticle velocity with the transverse momentum. Our results for Nb/Ni/Nb junctions are in good agreement with recent experimental studies. Our approach combines density functional theory and Bogoliubov-de Gennes model and opens a path for material composition optimization in magnetic Josephson junctions and superconducting magnetic spin valves.
Introduction
Coherent quantum tunneling of Cooper pairs through a thin barrier is one of the first examples of macroscopic quantum coherent phenomena. Predicted by Josephson more than 50 years ago [1], it has important applications in quantum circuits used in metrology, quantum sensing and quantum information processing [2].
Most of the previous studies focused on conventional Josephson junctions (JJs) consisting of two s-wave superconductors (S) that are connected by an insulating (I) or a normal (N) region [3, 4]. The flow of supercurrent through a JJ depends on the superconducting phase difference between two superconductors and, in general, is characterized by the current-phase relationship (CPR). In conventional JJs CPR should be periodic with , which follows from the BCS theory [5]. This result is a manifestation of a charge of Cooper pairs, and is used in metrology to measure electron charge. Time-reversal symmetry requires that which imposes a constraint that the supercurrent should be zero for where is integer. In general, CPR can be expanded in Fourier harmonics, .
In many cases, however, CPR is well approximated by the first harmonic with being the maximum supercurrent that can flow through the junction, i.e. the critical current. At a microscopic level, the supercurrent through a short SNS junction is determined by bound states forming in the constriction due to Andreev reflection at the NS interfaces. In the Andreev reflection process an incident electron-like quasiparticle with spin gets reflected at the NS interface as a hole-like quasiparticle with spin and a Cooper pair is emitted into the condensate. When time-reversal symmetry is not broken, electrons and holes propagate with the same velocity in the normal region. In this case no phase shift accumulates between this pair of quasiparticles along their trajectories in the N region, and the sign of is fixed. When we refer to this case as 0-junction.
In a magnetic Josephson junction (MJJ), exchange splitting breaks time-reversal symmetry and leads to an interesting interplay of superconductivity and magnetism [4, 7, 8, 9, 10]. In superconductor-ferromagnet-superconductor (SFS) junctions [Figure 1(a)] the correlated quasi-particles and quasi-holes forming Andreev bound states propagate through the junction under the exchange field of the ferromagnet (F). In many ferromagnets, such as Fe or Ni, the exchange splitting is large (of the order of eV) and significantly perturbs the band structure of a metal and, consequently, significantly modifies Fermi velocities of minority (spin ) and majority (spin ) carriers. Strong time-reversal symmetry breaking leads to the appearance of characteristic superconducting correlations with oscillatory dependence determined by the difference in wave numbers, [11, 12]. This effect opens a possibility for the supercurrent reversal as a function of the thickness of the ferromagnetic region, the so-called Josephson -junction [13, 14]. The correlation between the phase shift of the supercurrent and the magnetization provides a possibility for realizing magnetic spin valves, see Figure 1(b), which may have promising novel applications for cryogenic superconducting digital technologies [6, 15, 16]. Understanding the microscopic physics of MJJs is of great scientific interest as well as technological importance. 0- transitions in SFS junctions have been extensively studied experimentally since the early 2000’s and have been observed in different material systems [14, 17, 18, 19, 20, 21, 22, 23, 24, 25]. While qualitatively these observations are consistent with the previous phenomenological theories [26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42] the roles of the microscopic band structure arising from the atomic lattice on the supercurrent suppression with junction thickness remain unclear. Our primary goal is to address these essential points. The supercurrent suppression that is exponential in the junction length is often associated with presence of disorder in
Figure 1. (a) Schematic view of the SFS junction. (b) SFNFS junction. Arrows indicate possible magnetization of ferromagnets. The supercurrent through the spin-valve JJ depends on the relative magnetization of the ferromagnets, which governs the properties of the Josephson magnetic random-access memory (JMRAM) [6]. (c) Ball-and-stick representation of the Nb(110)/Ni(111)/Nb(110) junction with 5 layers of Ni. Nb atoms are light grey, Ni atoms are blue. The top and bottom atomic planes of Nb(110) are repeated periodically in the -direction to create the semi-infinite Nb leads of the junction through which the current flows. Periodic boundary conditions are used in the -plane. The corresponding reciprocal space defines two-dimensional vectors, i.e. the transverse modes, used in the calculations.
ends_mid=1 the ferromagnetic region [7, 43]. However, significant supercurrent suppression can also appear in relatively clean metals (e.g., Ni) whose mean free path is larger than the junction thickness [22, 23, 44]. Here we study the suppression of the critical current in the MJJs shown in Figure 1(a) and identify two microscopic mechanisms for its suppression, both a consequence of the band structure asymmetry of the majority and minority carriers in the F region. First, there is an asymmetry in the structure of the Fermi surface, see Figure 2(a). For certain bands and momenta, the Fermi surface present in one spin channel may be absent in the other. This is typical in ferromagnetic materials like Fe, Co, and Ni because the bandwidth for -electrons is relatively small and is often comparable to the exchange splitting. As a result, the wave number of one of the constituent quasiparticles forming Andreev bound states in MJJ becomes imaginary and the supercurrent becomes suppressed. We label this scenario as mechanism (i). In a second mechanism (ii), we show there is a dephasing of a harmonic signal originating from different Fourier components of the supercurrent due the Fermi velocity dispersion, see Figure 3(a). Both these mechanisms lead to an exponential suppression of the supercurrent, which was previously believed to occur due to the presence of disorder in the magnetic region. We show that band structure effects are important and may be even dominant in many cases.
Figure 2. (a) Simplified band structure of a ferromagnet, with majority and minority bands split by . and are the Fermi momenta for majority and minority carriers, respectively. Fermi level corresponds to large , where the minority band is pushed above . corresponds to small . Thus, the propagation of minority quasiparticles is characterized by an imaginary momentum and is suppressed. (b) Supercurrent in SFS junction is carried by Andreev bound states localized in the junction. Solid red line represents quasi-classical trajectory corresponding to an Andreev bound state. The spectrum of Andreev states depends on the relative superconducting phase difference across the junction as well as the phase, , accumulated due to the difference of Fermi momenta for majority and minority carriers. Note that in scenario (i) the propagation of minority carriers is suppressed leading to an overall exponential decay of the supercurrent with . This is to be contrasted with the normal transport through the junction.
Figure 3. (a) First Fourier harmonic of the supercurrent density Eq. (3), black circles and its fit Eq. (14), solid black line as a function of Ni layer thickness, , calculated for the Nb(110)/Ni(111)/Nb(110) junctions shown in Figure 1(c). Green semitransparent curves correspond to for individual . (b) Normal state conductance per unit of area as a function of for majority () and minority () spins Eq. (10) as well as . does not depend on and is approximated by the single value . (c)–(e) Supercurrent as a function of phase difference for (c) 4 atomic layers of Ni (strong 0-junction regime), (d) 8 layers (intermediate regime), and (e) 13 layers (strong -junction regime). In the 0- and -junction regimes, the component dominates. In the intermediate regime higher-order terms may prevail.
In order to capture realistic band structure, we develop a microscopic theory for the supercurrent in realistic MJJs. We use a combination of density functional theory (DFT) and Bogoliubov-de Gennes (BdG) model to investigate the 0- transition in realistic material stacks of Nb/Ni/Nb junctions in the clean limit. This method allows one to predict and explain key properties of MJJs such as the period and decay of the critical current oscillations with the ferromagnet thickness.
The paper is organized as follows. We describe our findings and summarize our main results in § 1.1. In § 2 we discuss the numerical method we developed to perform first principles supercurrent calculations. Detailed discussion of the main results is presented in § 2.1. We conclude with § 2.2. Some technical points are relegated to Appendices A–§ 2.8.
Qualitative discussion and main results
In this section we describe basic concepts for the supercurrent flow in MJJs and summarize our results. Our main qualitative conclusions are supported by microscopic calculations for Nb/Ni/Nb MJJs. Ni appears to have fairly long mean free path Å (see estimations in Appendix A), which is comparable or larger than the typical thickness of the ferromagnet used in recent experiments [22, 23]. Therefore, the motion of quasiparticles in Nb/Ni/Nb junction is quasi-ballistic, and our method is applicable to this system. Most of this paper focuses on clean Nb/Ni/Nb junctions.
First it is illuminating to consider a toy model, a one-dimensional SFS junction, and calculate the supercurrent in such a system for different Fermi energies, see Figure 2(a). Using the results of Ref. 45, one finds that both majority and minority spin bands are both occupied in the limit [i.e. scenario (ii) in Figure 2(a)]; the supercurrent does not decay with ferromagnet thickness at zero temperature,
Here perfect interface transparency is assumed, . The phase offset originates from the Fermi momentum difference of a quasi-particle and a quasi-hole forming Andreev bound state in the junction, see Figure 2(b), and is given by . The 0- and -junction regimes are clearly identified at and , respectively. At the intermediate values , the CPR is anharmonic which is a generic feature at 0- transition as shown below. In the low transparency regime, , one would expect qualitatively similar results with the maximal critical current being suppressed but still independent of the ferromagnet thickness, .
In the case of large exchange splitting [scenario (i) in Figure 2(a)] the minority band may become unoccupied. Given that minority carriers are gapped out and their propagation through the junction is suppressed, the supercurrent decays exponentially with [45],
Here and with being effective electron mass. One may notice the drastic difference between normal-state and superconducting transport in this case — the former is weakly affected (because majority and minority contributions are additive) whereas the supercurrent is strongly suppressed. In this case, the measurement of normal-state junction resistance does not necessarily predict the magnitude of the supercurrent through the junction.
We now generalize above results for the realistic three-dimensional (3D) geometry and material composition of the MJJ. In the clean limit, the supercurrent in the short-junction limit ( much smaller than the coherence length of the superconductor) is obtained from the spectrum of the Andreev bound states localized in the junction [3] which now also depends on the parallel momentum . The supercurrent density per junction area is . For the junction with periodic atomic structure in -plane the supercurrent density at zero temperature is given by
where the integration is performed over the Brillouin zone (BZ) of the corresponding surface supercell of area, , and the sum is carried over positive quasiparticle energies, . Note that we use spin-resolved and therefore omit spin prefactor 2 in Eq. (3). The derivative is periodic in and can be represented as a Fourier series
so that Eq. (3) can be written as
As we will show below, away from 0- transition the supercurrent is dominated by the first () harmonic. Therefore, we focus henceforth on the first harmonic contribution and drop index in the following discussion. Next, one may notice that the supercurrent amplitudes and phase offsets depend on the parallel momentum . One may include this dependence and define an effective Fermi energy that counts the energy corresponding to in each band of the ferromagnet from the bottom of the band. Depending on and , either scenario (i) or (ii) of Figure 2(a) may be realized.
Thus, it is important to compare the exchange splitting with the bandwidth of -character states in transition metals to make sure that a perturbation theory in is justified. Assuming it is the case, one can estimate the phase offset by expanding in exchange splitting to find
In general, the dependence of on is complicated, especially in -transition metals. The combination of complicated amplitude and phase offset dependence on leads to a non-trivial supercurrent dependence on the ferromagnet thickness . As shown in Figure 3(a), the critical current decays with for the Nb(110)/Ni(111)/Nb(110) junctions. We analyze the details of the decay and perform an exponential fit in § 2.1. At small thicknesses, below 30 Å, this decay originates from the evanescent modes corresponding to gapped out minority or majority carriers which cannot propagate through the junction, see Table I. This is the mechanism (i) discussed above. At larger thicknesses (i.e. Å) a loss of synchronization between different modes due to the dispersion of becomes important. This second mechanism (ii) has been previously discussed in the literature [7, 13] under assumptions of a single spherical Fermi surface and a small uniform exchange splitting in the magnetic region. Within these assumptions, one finds that critical current should decay algebraically with the thickness of a magnetic layer [13]. However, as we show below most of these assumptions do not apply to realistic SFS junctions involving transition metals. Thus, in order to understand CPR in realistic MJJ's, one needs to use accurate methods, which capture the physical effects described above.
| Band | \multicolumn{8}{c}{Majority ()} | \multicolumn{5}{c}{Minority ()} | |||||||||||
| [] | [] | [] | [] | [] | [eV] | [\text{\AA}] | [] | [] | [] | [] | [] | ||
| 2 | 0.68 | 12 | |||||||||||
| 3 | 3.90 | 13 | 2.7 | 3.6 | 3.3 | 3.6 | 0.004 | ||||||
| 4 | 0.61 | 47 | 0.8 | 1.8 | 1.1 | 1.2 | 0.015 | ||||||
| 5 | 2.95 | 22 | 0.6 | 3.0 | 1.5 | 1.8 | 0.173 | ||||||
| 6 | 3.5 | 6.0 | 4.6 | 5.2 | 0.029 | 0.6 | 3.0 | 2.3 | 2.6 | 0.045 |
Table I. Minimum, maximum, average, and root mean square of the Fermi velocities for the majority and minority Fermi surfaces in bulk Ni. is the density of states at the Fermi level, is the energy of the valence band maximum relative to for the majority bands not crossing , () is the effective (bare) electron mass, and estimates the decay exponent of the evanescent mode for a given band at .
To make a connection between the decay seen in Figure 3 and the mechanisms responsible for it, Figure 4(a) presents the energy band structure of bulk Ni. It is probably the highest fidelity band structure available: it very closely reproduces ARPES data in both majority and minority spin bands, with exchange splitting eV [46], and should be an excellent predictor of the real Fermi surface and velocities in Ni.
Figure 4. (a) Electronic band structure of bulk Ni (solid lines) calculated from first principles including many-body effects, see Ref. 46. It is the highest fidelity available and is very close to ARPES data (diamonds) in both majority (red) and minority (green) spin bands, with exchange splitting eV. (b) Majority (solid line) and minority (dashed line) Fermi surfaces of bulk Ni in the plane (with ) corresponding to the (111) plane direction used for the stacking of the Nb/Ni/Nb junctions, as discussed in the text. Axes correspond to two -L lines: the Fermi surface has a three-fold symmetry in the entire plane. Majority band is depicted as solid red-blue line with color interpolating between red and blue depending on the Fermi velocity , which ranges between m/s (red) and m/s (blue). Minority bands are shown by dashed lines: (blue), (cyan), (green), and (red).
Here we use this potential to analyze the bulk Ni Fermi surface (Figure 4(b)) and Fermi velocities (Table I). Counting from the bottom -band, bands of Ni character are bands 2 to 6. These bands are nearly full: only majority band and minority bands – cross the Fermi level . Only band 6 has both majority and minority carriers at the Fermi surface. Bands and have roughly the same shape and the energy splitting is approximately constant and equal to (see the right panel of Figure 4 in Ref. [46]). Beyond this, however, the correspondence between the Ni band structure and a simple parabolic band structure deviate substantially, in two ways that critically affect the analysis. First, the Fermi velocity, , is not fixed even for a single pocket: it varies in band by a factor of 2 [Figure 4 and Table I]. Accordingly the splitting between bands and varies by factor of two as expected from the twofold variation in . Second, bands – have no majority counterpart at , indicating that the wave number of bands – is complex. This is the origin for the exponential decay in scenario (i) in Figure 2(a) as noted above: a large portion of Andreev levels are carried by Cooper pairs made of single-particle wave functions with one or both of and having an imaginary component. The magnitude of depends on the particular mode and leading to a distribution of decay exponents. The slowest decay in each of these evanescent modes can be estimated from the distance of the closest approach to and the effective mass , using and decay , see Table I. ( is found to be highly anisotropic, so only the effective transport mass is shown.) is only a rough measure of the evanescent mode decay for a given band. We discuss in detail the distribution of decay exponents and phase offsets in § 2.1.
Let us now focus on the mechanism (ii) for the supercurrent decay, i.e. loss of synchronization between different transverse modes. This mechanism is well-known in diffusive systems where quasiparticle trajectory is random and thus the phase offset accumulated along such a trajectory also gets randomized. Thus, upon averaging Eq. (3) over different disorder realizations, one ends up with exponentially decaying critical current [7]. Previously, such a suppression of the supercurrent with junction thickness, , was often associated with impurity scattering in the ferromagnet. Here we demonstrate that this dephasing mechanism can also appear in clean systems (where quasiparticle trajectory is well defined) due to the dispersion of the velocity with in-plane momentum . Specifically, we find that, in the Nb/Ni/Nb junctions, this mechanism becomes relevant for junctions thicker than 50 Å, see Figure 3 and § 2.1. In SFS junctions the combination of disorder in the ferromagnet, interface scattering as well band-structure-induced dephasing ultimately determines the magnitude of the supercurrent. However, we believe that band structure effects provide an upper bound on the magnitude of the critical current as a function of .
In Sec. 4, we present numerical results which support the qualitative discussion presented above.
Method
We develop a numerical method to perform realistic simulations of MJJs using a combination of first-principles DFT and BdG calculations. The former is used to obtain the normal-state properties (e.g., band structure, Fermi velocities, magnetization) and to calculate the normal scattering matrices through the inhomogeneous 3D realistic junctions. As a next step we take superconductivity into account and calculate supercurrent through the stack assuming the short junction limit.
Normal transport: ab initio description
To calculate the normal scattering matrix we use the Questaal package for electronic structure calculations based on the linear muffin-tin orbital (LMTO) method [47]. It calculates the full non-linear, i.e. non-equilibrium, transport properties of an infinite system describing a central (C) region cladded by two semi-infinite left (L) and right (R) leads [48, 49], as represented below:
The LCR system [Fig. 1(c)] can be partitioned into an infinite stack of principal layers (PLs) which interact only with their nearest-neighbors. This is possible because the screened LMTO structure constants are short-ranged [50]. In the present case the C region consists of the ferromagnet, plus two layers of Nb at the LC and CR interfaces respectively. This is the range over which the perturbation from C significantly modifies the potential in the L or R region. To construct the Nb/Ni/Nb stack, coincident site lattices for Ni and Nb must be found (details of how this was accomplished are given in Appendix B). Planes of coincident site lattices are stacked to form the Nb/Ni/Nb structures. Figure 5(a) shows the Nb and Ni planes we used, which are denoted here as ‘surface supercells.’
Figure 5. (a) Top view of the Nb(110) and Ni(111) surface supercells used to build the Nb/Ni interfaces and the Nb(110)/Ni(111)/Nb(110) stacks shown in Figure 1(c). The surface supercell is defined from two 2D vectors Å and Å with periodic boundary condition in the 2D plane. The corresponding reciprocal space defines the 2D vectors used in the calculations. Each atomic plane of Ni (Nb) contains 14 (10) atoms of Ni (Nb). (b) Magnetic moment profile of Nb(110)/Ni(111)/Nb(110) junctions for different thickness of Ni (from 3 to 9 layers). The value of the moment is an averaged over the moments of the 14 Ni atoms in each atomic plane. Note the magnetic dead layer at the Nb/Ni interfaces and that all moments vanish for the shortest junction made of 3 layers of Ni.
The electronic current flows along the direction, perpendicular to the PLs lying in the -plane (transverse direction), see Figure 1(c). Periodic boundary conditions are used within each PL. The corresponding reciprocal space defines the two-dimensional (2D) vectors, i.e. the transverse modes, used in the calculations. The mesh is discretized and integrals over are performed numerically.
The electronic structure of the C region can be separated from L and R regions through self-energies, and , that modify the Hamiltonian of C region. They are most easily calculated if the potential of each PL in the L or R region are identical all through the bulk region. This is the reason for adding a few Nb layers folded into the C region. Thus the periodically repeating unit cells in the L and R regions can be safely assumed to have the potential of the bulk crystal. To construct the self-energies, the potentials of the PL in an infinite stack are needed. These potentials are functions only of the PL in their own region, and may be calculated in several ways. and are obtained from ‘surface’ Green’s function (a fictitious system which consists of a semi-infinite stack of PL, each with the same potential). Note that the potential of the C region is calculated self-consistently. This is important, as the local moments of Ni are small at the boundary layers, and build up gradually, see Figure 5(b).
With the potential in hand, the normal-state transport can be calculated using scattering formalism [49, 51]. For this, knowing the retarded Green’s function, , of the junction is sufficient. However, this is not the case for the Josephson current: the individual eigenfunctions are required. Within a Green’s function framework, must be organized by normal modes which correspond to the eigenstates of the L and R leads for a prescribed energy . In the PL representation, the Hamiltonian has been discretized into the linear combinations of the LMTO basis functions, and the normal modes are represented as eigenvectors of these basis functions. The Schrödinger equation becomes a difference equation in the PL basis functions [52]. Eigenvectors are calculated by solving a quadratic eigenvalue problem [52, 53], whose eigenvalues correspond to , where is the thickness of the PL. The wave number of the normal mode can be complex, but to correspond to a propagating mode must be real. By solving the equation as a function of the energy , one gets all the eigenvalues and eigenvectors which provide the information needed to construct the self-energies and (for each and each spin ). Note that, in the mode basis, the imaginary part of the self-energies is proportional to the (band) velocity of the modes, and is diagonal for non-degenerate modes [53, 54].
The retarded Green’s function of the C region (connected to the L and R leads) can be written as a matrix in the normal mode basis,
where is the Green’s function of the isolated C region. In this basis, is decomposed into four blocks,
upon projecting onto the propagating modes of the L and R leads. These four quantities and the mode velocities completely determine the normal state transport properties of the junctions. The transmission matrices are defined by the off-diagonal parts of Eq. (6). More specifically, the transmission coefficients , connecting L and R regions, are given by [51]
where and are the velocity matrix elements for propagating modes and in the L and R leads, respectively. The transmission matrix can be obtained from Eq. (7) by replacing L R. The reflection coefficients are given by the diagonal blocks of Eq. (6). For instance, on the L side
The reflection matrix on the R side is obtained from Eq. (8) by replacing L R. We define the normal scattering matrix as
where we omit - and -dependence for brevity. The linear-response normal conductance, , per spin is given by
It is calculated at the Fermi energy, . The total conductance is given by . Figure 3(b) shows that , , and are almost independent of the junction thickness, .
Superconducting transport: scattering matrix approach
Equation (9) is the normal-state scattering matrix for metal-ferromagnet-metal (NFN) structure taking into account reflection at both NF interfaces. To account for superconductivity, we introduce a step-like superconducting pairing potential meV and use the Andreev approximation to account for electron-hole scattering processes [3, 10]. This approach combines the details of the atomic structure of Nb/Ni/Nb and superconductivity within the mean-field approximation. The direct contact between S and F layers, as shown in Figure 1(a), would lead to an interaction between them. The back-action of the ferromagnet on the superconductor (i.e. the inverse proximity effect) results in a spatial dependence of the pairing potential near the SF interfaces [55, 56, 57]. In the case of a clean SFS junction model,
the self-consistent BdG calculations have been discussed in Refs. 58–62. In typical experimental systems the superconductor is disordered, so disorder effect on pairing potential needs to be considered, see, e.g., Ref. 63. Furthermore, in recent experiments S and F layers are separated by an intermediate spacer layer, which significantly reduces this inverse proximity effect. Thus, to understand inverse proximity effect in realistic SFS devices one would need to include both of the abovementioned ingredients in the model as well as to take into account the inhomogeneous magnetization in the ferromagnet (discussed in the next section) which is outside the scope of this paper. For the sake of clarity, we focus here on a realistic band structure in the ferromagnet and its effect on the supercurrent in the SFS structures. Henceforth, we also neglect the orbital effects of the fringe magnetic field, created by the ferromagnet.
Since , spin-resolved Andreev reflection at SN and NS interfaces is described by
where is the phase difference between left and right superconducting leads and is the identity matrix. In the short junction limit, the main contribution to the supercurrent comes from Andreev bound states localized in the junction having energy . The energy spectrum of Andreev states can be obtained using the following equation [64],
where . The vector corresponds to the electron- and hole-like (e/h) waves in N and N regions incident on the F region from the left () and from the right ().
Simulations show that is weakly-dependent on in the range . Therefore, we expand in and keep only the leading term, i.e. . Using this approximation, one can simplify the quantization condition (12) and reduce it to the matrix eigenvalue problem (see details in Ref. 65). This approach allows one to reliably calculate the Andreev bound states spectrum, . The zero-temperature supercurrent, , through the junction is given by Eq. (3). Figure 3(c)–Figure 3(e) show as a function of a phase difference, . Figure 3(a) shows the first Fourier harmonic of the supercurrent as a function of the junction thickness, .
Results
It is illuminating to compare our numerical simulations for the supercurrent with the experimental measurements involving quasi-ballistic MJJs. As previously discussed, we believe that the Nb/Ni/Nb junctions represent a good model system for which experimental data is readily available [16, 22, 23]. The best-performing stacks consist of Nb(110)/Cu/Ni(111)/Cu/Nb(110). Cu spacer layers seem to be essential to get strong supercurrent, likely because it prevents intermixing of the Ni and Nb. Our model junction simulates this geometry, via supercells in the plane normal to the stack to account for the lattice mismatch (see Figure 1 and Appendix B), though we do not include the Cu layers. We anticipate that Cu spacers will mainly affect transmission matrix elements rather than the dependence of the supercurrent on ferromagnet thickness, which is the main focus of this work. Furthermore, as discussed before, the Cu spacers will suppress the direct interaction between the ferromagnet and the superconductor and reduce inverse proximity effect justifying Andreev approximation for the boundary conditions, see Eq. (11). Therefore, we consider only the simplified Nb/Ni/Nb stack and vary the number of layers (atomic planes) of Ni. Additionally, we also consider effect of different crystallographic orientation of the Ni planes and investigate Nb/Ni(110)/Nb junctions in Appendix § 2.6.
To make a Nb/Ni superlattice, the unit cells of the Nb and Ni regions in the plane normal to the interface must be coincident. This is complicated by the severe lattice constant mismatch, and also the incompatibility of the (110) and (111) atomic planes. It is necessary to construct superlattices with Nb(110) and Ni(111) both rotated to the axis, and with lattice vectors in the plane coincident. A supercell with nearly coincident vectors was found (see Appendix B for details). By applying a small shear strain to the Ni, the lattice vectors are made exactly coincident. Figure 5(a) shows the Nb(110) surface supercell and the Ni(111) surface supercell with equal lattice vectors used to match the Nb/Ni interfaces. Each atomic plane of Ni(111) contains 14 atoms and each atomic plane of Nb(110) consists of 10 atoms. The atomic structure of the Nb(110)/Ni(111)/Nb(110) for 5 layers (atomic planes) of Ni is shown in Figure 1(c).
Next, we performed self-consistent DFT calculations within the local density approximation (LDA) in order to obtain the relaxed structure and corresponding electronic structure. For the smallest structures we performed a constrained optimization. Only the atoms in the planes closest to the Nb/Ni interfaces are allowed to relax to facilitate stacking of arbitrarily large cells. The Nb/Ni interplanar spacing has also been optimized to minimize the total energy, see Appendix § 2.8 for more details.
Once the structure is determined, one can determine the normal-state thermodynamic and transport properties of the junction, e.g., calculate the magnetization profile and spin-resolved conductance through the junction
as a function of Ni thickness. For transport calculations, we use a layer transport technique [48] which employs the atomic spheres approximation (ASA). Careful checks were made of ASA band structures of elemental Nb and Ni, and also superlattices, to confirm that they are very similar to the full potential LMTO DFT-LDA ones.
We find that the magnetic properties of Ni are sensitive to their local environment, indicative of the itinerant ferromagnetism. As shown in Figure 5(b), the magnetization profile is non-uniform in the junction with averaged magnetic moments per atom being suppressed near the Nb interface. For thickness larger than 4 layers, one recovers the bulk value of in the middle layers, away from the Nb/Ni interfaces. The averaged moment drops down towards the edges and becomes considerably reduced down to at the interface with Nb. The strong reduction of magnetism is exemplified for the short junction with 3 layers of Ni where the moments on the Ni atoms have completely vanished. Such a non-uniform magnetic moment dependence in Nb/Ni/Nb junctions affect superconducting properties of the SFS junctions in a non-trivial way. For example, Nb/Ni/Nb junctions thinner than 4 layers of Ni behave as essentially SNS junctions.
It is well known that the LDA tends to overestimate local moments in itinerant magnets [66] because spin fluctuations reduce the average moment [67], and underestimate when local moments are very large [46]. For Ni, LDA yields in good agreement with the experiment, but this is likely an artifact of an accidental cancellation of errors. Most important for transport is the exchange splitting , which the LDA predicts to be 0.6 eV, about twice larger than the experimental value of 0.3 eV [68]. It is possible to reproduce both and at the same time, but a high-level theory, potentially including spin-orbit coupling, is needed to surmount both kinds of errors inherent the LDA [46, 69]. The high cost and poor scaling of such a theory is not practical for these junctions, so we elect to stay within the LDA and scale the self-consistently calculated . This was the approach Karlsson and Aryasetiawan used to calculate the spin wave spectra in Ni [70]. Scaling of can be accomplished using different approaches, e.g., by adding some effective magnetic field to simulate the effect of spin fluctuations. Since Ni is a simple case with a nearly linear relation between and , the band structure hardly depends on the details in which the LDA potential is modified. Here we first perform fully self-consistent calculations. Then, to construct the potential for transport properties, we rescale the spin component of the density by a constant factor, which we denote as . This enables parametric studies of transport as a function of . yields the observed , and we use this scaling unless stated otherwise.
The conductance per unit of area, , is shown in Figure 3(b). It is weakly dependent on the thickness, , of the magnetic layer, as expected for a metallic system in the absence of disorder.
We now focus on superconducting properties. The dependence of the supercurrent on the phase difference for 5, 8, and 11 Ni layers is shown in Figure 3(c)–Figure 3(e). One can present current-phase relation, in a form of a Fourier series,
In the 0-junction mode, the first term in the Fourier series dominates with [see solid black line in Figure 3(c)]. In -junction case [Figure 3(e)], the supercurrent is also mostly defined by the first harmonic but with . Close to the 0- transition dies out, so that the behavior is governed by higher Fourier harmonics [71], e.g. for 8 Ni layers supercurrent has mostly second harmonic, , shown by dashed black line in Figure 3(d).
Figure 6 shows the critical current density, , normalized by normal-state conductance, , as a function of (blue squares). Far from the 0- transitions, the critical current coincides with the absolute value of the first Fourier harmonic, (red circles). Since contains a sign of the current and has better numerical stability than , we use this quantity for the analysis. We exclude very thin junctions (3 Ni layers or less) from the analysis since the magnetic properties are suppressed there.1
Figure 6. Comparison of critical current density, (blue squares), absolute value of first Fourier component for supercurrent, [see Eq. (13), red circles], and its fitting [Eq. (14), black curve]. All these quantities are ‘normalized’ by normal-state conductance, .
The dependence on can be fit by the following expression,
where , , , and are fitting parameters. We interpret as a decay length, as the ‘half-period’ of the oscillation in as a function of , and as a measure of the suppressed magnetization in Ni layers near the Ni/Nb boundaries. accurately fits the discrete points as shown in Figure 3(a), Figure 6, Figure 7, and Figure 10 by black solid line and black circles, accordingly.

Figure 7. First Fourier harmonic, , of the supercurrent as a function of junction thickness, . Full black circles correspond to calculated with 290 -points, empty circles (mostly superposed onto the full black circles) correspond to 4142 -points. Solid black line is [ fit given by Eq. (14)]. Green semitransparent dashes show contributions [Eq. (15b)] for the individual -points. The vertical ‘errorbars’ correspond to the standard deviation of with respect to . The standard deviation is the same for both sets of -points indicating that these results are independent of the chosen discretization.
Figure 10. for different rescaling of the magnetic moments as a function of (a) the thickness, , and (b) the rescaled thickness, , where is the fitting parameter in Eq. (14) and Å. values are shown by circles, corresponding fittings are shown by lines of the same color. (Inset) Half-period, , fitted using Eq. (14) versus .
In order to gain insight into the evolution of with , let us resolve contributions from different . For this, we rewrite Eq. (3) as
Here is the area of the surface supercell shown in Figure 5(a). Similar to Eq. (13), we denote the first -harmonic of as . The evolution of the first Fourier harmonic of the supercurrent and as a function of are shown in Figure 7. Calculations were performed for two different sets of with 290 discrete -points (full black circles) and 4142 -points (empty black circles). One can see that both sets give the same result for , establishing that the integration is well converged. In Figure 7, ‘errorbars’ denote the standard deviation in with respect to the -summed average, . Individual are shown by semi-transparent horizontal dashes. The important observation is that while decays with , the dispersion in does not change significantly. Figure 8 shows colorplots of corresponding to local extrema of (4, 13, 25, 36, 48, and 60 layers labeled in Figure 7). For small , one can see that most of all the contributions to have the same sign, i.e. positive in 0-junction regime and negative in -junction regime. In this regime the decay is predominantly due to evanescent modes decaying into the junction. For larger , the dephasing mechanism becomes important since the phase offset spread grows with . One can observe the apparition of contributions of the opposite sign for Å. This dephasing mechanism is mainly due to the variation of the Fermi velocity with , and becomes more important with increasing .
Figure 8. Colorplot of with . Each panel corresponds to the local extrema of shown in Figure 7. For 4 and 13 layers all the channels contribute with the same sign. For 48 layers and larger, different channels lose synchronization and contribute to the total supercurrent with different signs.
In order to study the distribution of the phase offsets and decay exponents for different modes, we fit the individual using an expression analogous to Eq. (14). The set of the resulting fitted curves for 4142 -points are shown by the green semitransparent curves in Figure 3(a). Here to minimize the numerical ‘noise,’ curves are smoothed over the 2D space using Gaussian filter with Å which is of the order of the Fermi wave vector in Nb. Thus, each data point approximately corresponds to a transverse conducting channel. This fitting procedure works reasonably well, e.g., the relationship in Eq. (15a) holds if one replaces by and by for Å.
The distribution of the fitting parameters for is shown in Figure 9. Histograms for decay lengths and half-periods reveal a complicated picture describing different contributions to the supercurrent in real materials. First of all, in Figure 9(a) one can see the distribution of the half-periods, , which is similar to a Gaussian distribution with a mean value Å and standard deviation 2.8 Å. The mean value is very close to [see text after Eq. (14)] while the spread in leads to dephasing and is responsible for the exponential decay of at large . Indeed, it is well-known that the average of an oscillatory function with respect to a random fluctuating phase (described by a Gaussian distribution) results in an exponentially decaying function.
Figure 9. Analysis of the fitting parameters of for individual , shown in Figure 3(a) by semitransparent green lines. Distribution of (a) half-periods, , and (b) decay lengths, .
In addition to the dephasing mechanism, the decay of the supercurrent originates from the evanescent modes. The histogram for decay lengths, is shown in Figure 9(b). Here small corresponds to fast-decaying , large is responsible for non-decaying modes (i.e. modes with the decay exponents larger than the junction thickness). The right-skewed distribution of the decay exponents has a mean value of Å which is much larger than in Eq. (14). The shoulder at small presumably corresponds to the evanescent mode decay comprising of bands 2 and 3, see Table I, whereas the tail at large originates predominantly from the band 6. Overall, one can see that a fit with a single decay exponent, discussed in Eq. (14), is quite oversimplified for a Nb/Ni/Nb junction considered here.
We now turn to the discussion of the effect of exchange splitting energy on the supercurrent in MJJs. So far we have used , which yields the experimentally observed eV. It is interesting to investigate how a ferromagnet with a different (but otherwise the same band structure as Ni) would affect the -dependence of . In Figure 10 we show the results for parametric variations in . One can see in Figure 10(a) that the half-period, , and the decay length, , strongly depend on . Here black points correspond to , and the self-consistent calculations with no rescaling correspond to . In order to understand how and depend on , we perform the fitting procedure Eq. (14) for different magnetic moments and plot, in Figure 10(b), the supercurrent density as a function of the rescaled thickness, with Å. Remarkably , as a function of the rescaled thickness, collapses to the same universal
curve. The inset demonstrates that the fitted half-period, , is proportional to showing that the oscillation period scales linearly with the inverse of in this parameter range.2
We now discuss the difference in crystal orientation in Nb/Ni/Nb junctions. We have performed calculations for Nb/Ni/Nb junctions built from stacking the Ni atomic planes in the (110) orientation instead of the (111) orientation. The results are given in Appendix C. Qualitatively, the same physics hold for both stacks built from (111) and (110) Ni planes. However, our calculations show that the actual value for the period of oscillation and for the current decay depend crucially on the details of the electronic structure of the junctions, such as the relative crystal orientation.
Finally, we considered effect of spin-orbit coupling in SFS junctions. Spin-orbit coupling leads to mixing of the minority and majority channels and may change current-phase relationship. The interplay between Zeeman splitting and spin-orbit coupling have been discussed in Ref. 45; the regime of interest is Zeeman-field dominated regime considered there. Indeed, we find that SOC in Ni is much smaller than the exchange splitting because of low atomic number of Ni. As we show in Appendix D, the SOC in Ni-based MJJ considered here does not change qualitative picture described above but rather leads to small quantitative changes to the Josephson current.
Conclusion
In this paper we identified two generic mechanisms for the decay of the supercurrent with junction thickness: (i) exchange-splitting induced gap opening for minority or majority carriers and (ii) dephasing between different modes due to the significant quasiparticle velocity dispersion with the transverse momentum. It was previously believed that disorder in the ferromagnet is mainly responsible for the supercurrent decay in SFS junctions. In the present work we have shown that band structure effects also contribute to the critical current suppression and thus provide an upper bound for the supercurrent in ideal (i.e. disorder-free) structure.
We found that the Nb/Ni/Nb junction is a suitable system for comparison with the simulations because of the long mean free path in Ni relative to the junction thickness and the quasi-ballistic nature of quasiparticle propagation in the ferromagnet. We have found good agreement with published experimental data for the half-period of the critical current oscillations: Å [see Eq. (14) and text after it] versus Å in experiment, Ref. 22. We have also found that the critical current decays exponentially with the ferromagnet thickness . This is to be contrasted with previously assumed algebraic decay based on results for the clean SFS junctions using simple parabolic-like band structure. We believe that in measured Nb/Ni/Nb junctions with Å the mechanism (i) is likely to be responsible for the supercurrent decay. This finding is crucial for material and geometry optimization of MJJs and superconducting magnetic spin valves.
Understanding the interplay of band structure effects and disorder in MJJs is an interesting open problem. We believe that interfacial disorder due to, for example, surface roughness will mix different modes and will lead to a larger spread of half-periods. This, in turn, will further enhance the dephasing mechanism (ii) of the supercurrent decay discussed here. Strong disorder in the bulk (i.e. mean free path much smaller than junction thickness ) would lead to the diffusive motion of quasiparticles in the ferromagnet which is a significant departure from the quasi-ballistic junction limit considered here. We think that bulk disorder would induce even more dephasing between different modes because phase offsets in this case will depend on different random trajectories of minority and majority carriers. We, therefore, believe that bulk disorder will lead to even stronger decay of the supercurrent with junction thickness, .
Acknowledgments
HN is much indebted to Dimitar Pashov for stimulating discussions about developing the Questaal package. The authors express their gratitude to Mason Thomas for the organizational help and discussions at the early stages of the project. The authors acknowledge stimulating discussions with Norman Birge, Anna Herr, Tom Ambrose, Nick Rizzo, and Don Miller.
This work is based on support by the U.S. Department of Energy, Office of Science through the Quantum Science Center (QSC), a National Quantum Information Science Research Center. HN and MvS acknowledge financial support from Microsoft Station Q via a sponsor agreement between KCL and Microsoft Quantum. In the late stages of this work MvS was supported by the U.S. Department of Energy, Office of Science, Basic Energy Sciences under Award # FWP FRW7906.
A Mean free path estimate
In this section we provide an estimate for the mean free path in the Ni ferromagnet. Our approach is similar to Ref. 44. We use the Kubo formula for conductivity
where index labels the Ni bands, is the spin projection, and is the scattering time in each band assumed to be momentum independent. Mean square velocity, , and density of states at the Fermi level, , are obtained from our ab initio model calculations and are provided in Table I.
We use the available experimental data [74, 75] for thick Ni samples Å: the low temperature linear resistance nm, and bulk spin scattering asymmetry, . Only a single band of the majority spin is occupied. Thus, one can readily estimate mean free path for majority carriers Å. For the minority channel we will assume that the conductance at large thicknesses is dominated by the most mobile band, i.e. [76]. This gives an estimate for the mean free path Å, similar to its exchange-split partner . These estimates are consistent with the available ARPES [77] and computational data [44]. As mentioned in the main text, the mean free path for and carriers exceeds the typical junction thicknesses measured experimentally.
B Nb(110)/Ni(111)/Nb(110) junctions
Here we describe how coincident site lattices are constructed for the Nb(110)/Ni(111) interface. Supercells of both Ni and Nb are separately constructed in the following way, to make a coincident site lattice. The primitive unit cells (fcc in the Ni case, with a 0 K lattice constant 3.515 Å, and bcc in the Nb case with a 0 K lattice constant 3.295 Å) are rotated. For Ni, the rotation is compactly described in terms of three Euler angles: rotation about by , rotation about by , rotation about by . Axes are shown in Figure 1(c); and are axes after first and second rotation, respectively. From the first two rotations the axis becomes the new axis. The last rotation is needed to make it approximately coincident with Nb. The Nb is rotated about by , about by , and about again by . After rotations both Ni and Nb planes are normal to , which is the propagation direction.
Next, superlattices must be constructed. They are generated by scaling the primitive lattice vectors by the following integer multiples, for Ni and Nb respectively:
The Ni supercell contains 14 atoms, all in a single plane perpendicular to ; the Nb supercell consists of 2 planes with 10 atoms per plane [Figure 5(a)]. The lattice vectors
transverse to are nearly coincident, but not identically so. To render them coincident, we opt to shear the Ni by the following linear transformation,
This shear is a measure of the remaining mismatch of the undistorted Ni and Nb lattices. It is close enough to unity to have a minor effect on the Ni band structure.
Along this axis, Nb planes form an ABAB... stacking pattern; the Ni are stacked ABCABC... Ni/Nb interfaces are formed by stacking varying numbers of Ni planes on a Nb substrate, and relaxing the supercells to minimize the total energy. This was done for a few small superlattices with 3, 4, and 5 Ni planes, restricting the relaxation to the two Ni and two Nb frontier planes. In this way we can build up structures of arbitrarily many Ni planes, sandwiching unrelaxed planes between frontier planes. Thus the entire set of structures has three families of interfaces: those with integer numbers , or of Ni planes.
Figure 3(b) shows that normal conductance, , experiences a deviations from its mean following the pattern of these three families. Overall, both normal conductance and supercurrent (see Figure 7) are reasonably smooth functions of the thickness , which suggests that the discreteness of the lattice and the details of lattice relaxation plays a minor role.
C Nb(110)/Ni(110)/Nb(110) junctions
In this section, we present results for the Nb/Ni/Nb junctions built from stacking the Ni atomic planes with the (110) orientation instead of the (111) orientation. Although the (110) plane orientation is not the most stable configuration for the Nb/Ni interface and most experimental studies focus on (111) orientation, it is interesting to ascertain whether there is an effect of crystallographic orientation on the supercurrent. Indeed, the supercurrent decay and half-period of oscillations with do depend on crystal orientation.
We study the Nb(110)/Ni(110)/Nb(110) trilayer shown in Figure 11(a). The inter-plane distance between the Nb and Ni planes at the interfaces is taken to be the average of the Nb and Ni inter-plane distances. For the sake of simplicity, we do not perform atomic relaxations for this system. The lattice parameter of Nb (3.295 Å) has been increased to match the lattice parameter of Ni (3.515 Å) to simplify the construction of periodic supercell. This has a slight effect on the Nb band structure.
Figure 11. (a) Top view of the Nb(110)/Ni(110) interface supercell. Ni (Nb) atoms are shown in light blue (grey). A repetition (3×3) of the surface supercell is shown. The corresponding Nb(110) and Ni(110) cells contain 2 atoms each. (b) Transverse view of the Nb(110)/Ni(110)/Nb(110) junction with 5 layers of Ni. (c) component for the Nb(110)/Ni(110)/Nb(110) and the Nb(110)/Ni(111)/Nb(110) junctions versus the thickness . Both currents have a decaying oscillatory behavior, with a half period of oscillation of Å for the Ni(110) case and of Å for the Ni(111) case.
Calculations of the trilayers were performed self-consistently, and for transport we reconstruct the potential by rescaling the magnetic moments by 0.5, as in the Ni(111) case. The - and - transitions occur around Å and 41 Å [shown in red in Figure 11(c)], yielding a half-period for the oscillations of Å. This period is significantly larger than Å for the Nb(110)/Ni(111)/Nb(110) case [shown in black in Figure 11(c)]. The difference of Å between the two half periods represents roughly 5 inter-plane distances for Ni(110) [3 inter-plane distances for Ni(111)], and is not induced by the unrelaxed atomic structures at the Nb(110)/Ni(110) interfaces. It is noteworthy that the calculated period for the Ni(111) trilayers better coincides with available experimental data [22] than for the Ni(110) case.
Effect of spin-orbit interaction
In this section, we discuss the effect of spin-orbit coupling (SOC) on Josephson current in quasi-ballistic SFS junctions. The effect of spin-orbit interaction is two-fold: (i) SOC changes the band structure and, therefore, may modify the dependence of the quasiparticle velocities on transverse momenta and (ii) SOC couples spin and orbital motion and leads to spin precession along the junction. The combination of spin precession and scattering on non-magnetic impurities may introduce random spin-flip processes which suppress the phase difference between minority and majority quasiparticles. We analyze both mechanisms below and show that they represent weak perturbations to our main results and do not change qualitative predictions for quasi-ballistic Ni-based MJJs.
Detailed theoretical analysis of the bulk Ni band structure with SOC and different orientations of the magnetic moment has been discussed in Ref. 69. In Figure 12, we compare Fermi surface of Ni with and without SOC using the same high-level theory [46] of Figure 4(a). The Questaal code can treat SOC effects, see e.g. tutorial ‘Spin and spin orbit coupling.’ The SOC implementation is conventional for the energy-dependent local basis set and band structure calculations (see Sec. 2.8.2 in Ref. 47), and has also been developed for the Green’s functions (see Supplementary Materials in Ref. 78 and Sec. 2.16 in Ref. 47).
Figure 12. GW Fermi surfaces of bulk Ni in the plane corresponding to Ni(111). (a) No SOC [same as Figure 4(b)]. Shown are majority band (red) and minority bands (blue), (dark blue), (cyan), and (green). (b) With SOC. SOC opens gaps around the small regions close to the band crossings (circled).
One may notice that the spin-orbit coupling only weakly modifies the Fermi surface shown in Figure 12. Indeed, apart from several pockets (close to the degeneracy points) across the Brillouin zone, the SOC only weakly perturbs the band structure. In fcc crystals, the -orbitals are split into degenerate and levels, at the point, according to the symmetry. The SOC splits further the degenerate levels. From these level shifts (taken at the -point), we estimate local SOC coupling to be of the order of meV which is much smaller than the exchange splitting in Ni. In Ref. 69, the corresponding SOC strength was found to be 68meV which further corroborates our conclusions.
One can now evaluate effect of SOC coupling on quasi-particle propagation through the junction. Given that bulk band structure of Ni is centro-symmetric, Dresselhaus spin orbit coupling is forbidden by symmetry. The Rashba SOC may appear due to the inversion symmetry-breaking along the direction of the junction. However, the rapid screening of the interface potential in the ferromagnet would limit Rashba SOC to a few atomic layers in the junction. Finally, we believe that effect of local spin-flip processes due to scattering on impurities is weak in the regime of interest. Indeed, here we are considering quasi-ballistic magnetic Josephson junctions (i.e. the thickness of the junction smaller than the mean free path in Ni). Thus, we expect impurity-induced spin-relaxation rate to be suppressed as well.
Normal scattering matrix calculation
The Questaal code is based on the LMTO technique which uses a set of electron wave-function and its energy-derivative as a basis set. The wave-functions are solutions of the spherical Schrödinger equation in a sphere around a given atom at position with angular quantum numbers . The LMTO technique is an all-electron approach, with does not rely on the use of pseudo-potential. The basis set contains core electrons as well as ‘valence’ (non-core) electrons.
The electron ground state of the systems is obtained with the DFT-LDA-ASA framework [47]. The (energy-dependent) basis set consists of partial waves of , , character on each of the atomic sites. Core levels are integrated separately from the valence partial waves to obtain the all-electron charge density. However, they are not included in the secular matrix.
The calculations are converged when variations in the electron density and total energy between the last iterations, are below () respectively. To obtain the self-consistent charge density, calculations were performed with a mesh of . As is well known, for purposes of determining the density, a finer mesh is not needed since the output is a -independent potential for a subsequent transport calculation.
Atomic relaxations are performed on relatively small cells with a full-potential method [47], with periodic boundary conditions. Once the ground-state is reached, the atomic positions of the frontier atoms at the Nb/Ni and Ni/Nb interfaces are allowed to relax to minimize the total energy. Convergence is achieved when all the forces on the relaxed atoms are below mRy/bohr. Atomic relaxations are constrained to the first and second frontier layers; these shifts are then added to the ideal geometry of the (larger) stacked cells in the subsequent ASA calculations of transport. More details of the calculations can be found in the Questaal tutorial ‘Nb/Ni superlattice’.
A denser mesh is needed as the transmission (reflection) probability is a quantity very sensitive to the -point sampling. As shown in Figure 3 of the main text, we used a mesh of and a mesh of points, giving rise to 290 and 4142 irreducible points respectively. This establishes that our transport calculations are well converged in .
Note that the principal layer technique (used for the transport) does not rely on periodic boundary conditions in the (transport) direction perpendicular to the layers (in our case the -axis). Hence there is not point sampling, the wave number should be understood as a continuous complex function of .
Further details for the constructions of the family of the , or of Ni planes of the Nb(110)/Ni(111)/Nb(110) junctions can be found in the Questaal tutorial ‘Nb(110)/Ni/Nb(110) metallic trilayers.’
References
References
- [1]B. Josephson, Possible new effects in superconductive tunnelling, Phys. Lett. 1, 251 (1962).DOI
- [2]P. A. Warburton, The josephson effect: 50 years of science and technology, Phys. Ed. 46, 669 (2011).DOI
- [3]C. W. J. Beenakker, Three “universal” mesoscopic Josephson effects, in Transport phenomena in mesoscopic systems, Vol. 109, edited by H. Fukuyama and T. Ando (Springer, Berlin, Heidelberg, 1992) pp. 235–253.
- [4]A. A. Golubov, M. Y. Kupriyanov, and E. Il’ichev, The current-phase relation in Josephson junctions, Rev. Mod. Phys. 76, 411 (2004).DOI
- [5]J. Bardeen, L. N. Cooper, and J. R. Schrieffer, Microscopic theory of superconductivity, Phys. Rev. 106, 162 (1957).
- [6]I. M. Dayton, T. Sage, E. C. Gingrich, M. G. Loving, T. F. Ambrose, N. P. Sivak, S. Keebaugh, C. Kirby, D. L. Miller, A. Y. Herr, Q. P. Herr, and O. Naaman, Experimental demonstration of a Josephson magnetic memory cell with a programmable π-junction, IEEE Magn. Lett. 9, 1 (2018).DOI
- [7]A. I. Buzdin, Proximity effects in superconductor-ferromagnet heterostructures, Rev. Mod. Phys. 77, 935 (2005).DOI
- [8]F. S. Bergeret, A. F. Volkov, and K. B. Efetov, Odd triplet superconductivity and related phenomena in superconductor-ferromagnet structures, Rev. Mod. Phys. 77, 1321 (2005).DOI
- [9]M. G. Blamire and J. W. A. Robinson, The interface between superconductivity and magnetism: understanding and device prospects, J. Condens. Matter Phys. 26, 453201 (2014).DOI
- [10]M. Eschrig, Spin-polarized supercurrents for spintronics: a review of current progress, Rep. Prog. Phys 78, 104501 (2015).DOI
- [11]P. Fulde and R. A. Ferrell, Superconductivity in a strong spin-exchange field, Phys. Rev. 135, A550 (1964).DOI
- [12]A. I. Larkin and Y. N. Ovchinnikov, Inhomogeneous state of superconductors, Sov. Phys. JETP 20, 762 (1965).
- [13]A. L. Buzdin, L. N. Bulaevskii, and S. V. Panyukov, Critical-current oscillations as a function of the exchange field and thickness of the ferromagnetic metal (F) in an S-F-S Josephson junction, JETP Lett. 35, 178 (1982).
- [14]V. V. Ryazanov, V. A. Oboznov, A. Y. Rusanov, A. V. Veretennikov, A. A. Golubov, and J. Aarts, Coupling of two superconductors through a ferromagnet: Evidence for a π junction, Phys. Rev. Lett. 86, 2427 (2001).pubmed.ncbi.nlm.nih.gov/11289946
- [15]C. Bell, G. Burnell, C. W. Leung, E. J. Tarte, D.-J. Kang, and M. G. Blamire, Controllable Josephson current through a pseudospin-valve structure, Appl. Phys. Lett 84, 1153 (2004).DOI
- [16]E. C. Gingrich, B. M. Niedzielski, J. A. Glick, Y. Wang, D. L. Miller, R. Loloee, W. P. P. Jr, and N. O. Birge, Controllable 0-π Josephson junctions containing a ferromagnetic spin valve, Nat. Phys. 12, 564 (2016).DOI
- [17]V. V. Ryazanov, V. A. Oboznov, A. V. Veretennikov, and A. Y. Rusanov, Intrinsically frustrated superconducting array of superconductor-ferromagnet-superconductor π junctions, Phys. Rev. B 65, 020501 (2001).DOI
- [18]T. Kontos, M. Aprili, J. Lesueur, F. Genêt, B. Stephanidis, and R. Boursier, Josephson junction through a thin ferromagnetic layer: Negative coupling, Phys. Rev. Lett. 89, 137007 (2002).pubmed.ncbi.nlm.nih.gov/12225057
- [19]H. Sellier, C. Baraduc, F. m. c. Lefloch, and R. Calemczuk, Temperature-induced crossover between 0 and π states in S/F/S junctions, Phys. Rev. B 68, 054531 (2003).DOI
- [20]J. W. A. Robinson, S. Piano, G. Burnell, C. Bell, and M. G. Blamire, Critical current oscillations in strong ferromagnetic π junctions, Phys. Rev. Lett. 97, 177003 (2006).pubmed.ncbi.nlm.nih.gov/17155498
- [21]T. S. Khaire, W. P. Pratt, and N. O. Birge, Critical current behavior in Josephson junctions with the weak ferromagnet PdNi, Phys. Rev. B 79, 094523 (2009).DOI
- [22]B. Baek, M. L. Schneider, M. R. Pufall, and W. H. Rippard, Phase offsets in the critical-current oscillations of Josephson junctions based on Ni and Ni-(Ni₁₈Fe₁₉)ₓNbᵧ barriers, Phys. Rev. Applied 7, 064013 (2017).
- [23]B. Baek, M. L. Schneider, M. R. Pufall, and W. H. Rippard, Anomalous supercurrent modulation in Josephson junctions with Ni-based barriers, IEEE Trans. Appl. Supercond 28, 1 (2018).DOI
- [24]V. Aguilar, D. Korucu, J. A. Glick, R. Loloee, W. P. Pratt, and N. O. Birge, Spin-polarized triplet supercurrent in Josephson junctions with perpendicular ferromagnetic layers, Phys. Rev. B 102, 024518 (2020).DOI
- [25]S. S. Mishra, R. Loloee, and N. O. Birge, Supercurrent transmission through Ni/Ru/Ni synthetic antiferromagnets, Appl. Phys. Lett. 119, 172603 (2021).DOI
- [26]W. L. McMillan, Theory of superconductor—normal-metal interfaces, Phys. Rev. 175, 559 (1968).DOI
- [27]T. Wolfram, Tomasch oscillations in the density of states of superconducting films, Phys. Rev. 170, 481 (1968).DOI
- [28]I. O. Kulik, Macroscopic quantization and the proximity effect in S-N-S junctions, J. Exp. Theor. Phys. 30, 944 (1970).
- [29]J. Demers and A. Griffin, Scattering and tunneling of electronic excitations in the intermediate state of superconductors, Can. J. Phys. 49, 285 (1971).DOI
- [30]A. Griffin and J. Demers, Tunneling in the normal-metal-insulator-superconductor geometry using the Bogoliubov equations of motion, Phys. Rev. B 4, 2202 (1971).DOI
- [31]O. Entin-Wolhman, Effect of a barrier at the superconducting-normal metal interface, J. Low Temp. Phys 27, 777 (1977).
- [32]G. E. Blonder, M. Tinkham, and T. M. Klapwijk, Transition from metallic to tunneling regimes in superconducting microconstrictions: Excess current, charge imbalance, and supercurrent conversion, Phys. Rev. B 25, 4515 (1982).DOI
- [33]A. Furusaki and M. Tsukada, Dc Josephson effect and Andreev reflection, Solid State Commun 78, 299 (1991).DOI
- [34]A. Furusaki, H. Takayanagi, and M. Tsukada, Josephson effect of the superconducting quantum point contact, Phys. Rev. B 45, 10563 (1992).pubmed.ncbi.nlm.nih.gov/10000963
- [35]A. Furusaki, DC Josephson effect in dirty SNS junctions: Numerical study, Physica B 203, 214 (1994).DOI
- [36]M. J. M. de Jong and C. W. J. Beenakker, Andreev reflection in ferromagnet-superconductor junctions, Phys. Rev. Lett. 74, 1657 (1995).
- [37]Y. Tanaka and S. Kashiwaya, Theory of Josephson effect in superconductor-ferromagnet-insulator-superconductor junction, Physica C 274, 357 (1997).DOI
- [38]I. Žutić and O. T. Valls, Spin-polarized tunneling in ferromagnet/unconventional superconductor junctions, Phys. Rev. B 60, 6320 (1999).
- [39]Z. Radović, N. Lazarides, and N. Flytzanis, Josephson effect in double-barrier superconductor-ferromagnet junctions, Phys. Rev. B 68, 014501 (2003).DOI
- [40]J. Cayssol and G. Montambaux, Incomplete andreev reflection in a clean superconductor-ferromagnet-superconductor junction, Phys. Rev. B 71, 012507 (2005).DOI
- [41]F. Konschelle, J. Cayssol, and A. I. Buzdin, Nonsinusoidal current-phase relation in strongly ferromagnetic and moderately disordered sfs junctions, Phys. Rev. B 78, 134505 (2008).DOI
- [42]A. F. Tzortzakakis and N. Flytzanis, Josephson junctions with spin-orbit and spin-flip interactions, Master’s thesis, University of Crete (2019).arxiv.org/abs/1909.10617
- [43]E. A. Demler, G. B. Arnold, and M. R. Beasley, Superconducting proximity effects in magnetic metals, Phys. Rev. B 55, 15174 (1997).DOI
- [44]D. Gall, Electron mean free path in elemental metals, J. Appl. Phys. 119, 085101 (2016).DOI
- [46]L. Sponza, P. Pisanti, A. Vishina, D. Pashov, C. Weber, M. van Schilfgaarde, S. Acharya, J. Vidal, and G. Kotliar, Self-energies in itinerant magnets: A focus on Fe and Ni, Phys. Rev. B 95, 041112 (2017).DOI
- [47]D. Pashov, S. Acharya, W. R. Lambrecht, J. Jackson, K. D. Belashchenko, A. Chantis, F. Jamet, and M. van Schilfgaarde, Quastaal: A package of electronic structure methods based on the linear muffin-tin orbital technique, Comput. Phys. Commun 249, 107065 (2020).
- [48]S. V. Faleev, F. Léonard, D. A. Stewart, and M. van Schilfgaarde, Ab initio tight-binding LMTO method for nonequilibrium electron transport in nanosystems, Phys. Rev. B 71, 195422 (2005).DOI
- [49]Y. Meir and N. S. Wingreen, Landauer formula for current through an interacting electron region, Phys. Rev. Lett. 68, 2512 (1992).pubmed.ncbi.nlm.nih.gov/10045416
- [50]O. K. Andersen and O. Jepsen, Explicit, first-principles tight-binding theory, Phys. Rev. Lett. 53, 2571 (1984).DOI
- [51]D. S. Fisher and P. A. Lee, Relation between conductivity and transmission matrix, Phys. Rev. B 23, 6851 (1981).DOI
- [52]A.-B. Chen, Y.-M. Lai-Hsu, and W. Chen, Difference-equation approach to the electronic structures of surfaces, interfaces, and superlattices, Phys. Rev. B 39, 923 (1989).pubmed.ncbi.nlm.nih.gov/9948270
- [53]Y. Fujimoto and K. Hirose, First-principles treatments of electron transport properties for nanoscale junctions, Phys. Rev. B 67, 195315 (2003).DOI
- [54]M. Wimmer, Quantum transport in nanostructures: From computational concepts to spintronics in graphene and magnetic tunnel junctions, Ph.D. thesis, Universität Regensburg (2008).DOI
- [55]K. Halterman and O. T. Valls, Proximity effects at ferromagnet-superconductor interfaces, Phys. Rev. B 65, 014509 (2001).
- [56]K. Halterman and O. T. Valls, Proximity effects and characteristic lengths in ferromagnet-superconductor structures, Phys. Rev. B 66, 224516 (2002).DOI
- [57]G. Csire, A. Deák, B. Nyári, H. Ebert, J. F. Annett, and B. Újfalussy, Relativistic spin-polarized KKR theory for superconducting heterostructures: Oscillating order parameter in the Au layer of Nb/Au/Fe trilayers, Phys. Rev. B 97, 024514 (2018).DOI
- [64]C. W. J. Beenakker, Universal limit of critical-current fluctuations in mesoscopic josephson junctions, Phys. Rev. Lett. 67, 3836 (1991).pubmed.ncbi.nlm.nih.gov/10044838
- [66]A. Aguayo, I. I. Mazin, and D. J. Singh, Why Ni₃Al is an itinerant ferromagnet but Ni₃Ga is not, Phys. Rev. Lett. 92, 147201 (2004).pubmed.ncbi.nlm.nih.gov/15089568
- [67]T. Moriya, Spin fluctuations in itinerant electron magnetism (Springer-Verlag, Berlin, 1985).
- [68]F. J. Himpsel, J. A. Knapp, and D. E. Eastman, Experimental energy-band dispersions and exchange splitting for ni, Phys. Rev. B 19, 2919 (1979).DOI
- [69]J. Bünemann, F. Gebhard, T. Ohm, S. Weiser, and W. Weber, Spin-orbit coupling in ferromagnetic nickel, Phys. Rev. Lett. 101, 236404 (2008).DOI
- [70]K. Karlsson and F. Aryasetiawan, A many-body approach to spin-wave excitations in itinerant magnetic systems, J. Phys. Condens. Matter 12, 7617 (2000).DOI
- [71]M. J. A. Stoutimore, A. N. Rossolenko, V. V. Bolginov, V. A. Obnoznov, A. Y. Rusanov, D. S. Baranov, N. Pugach, S. M. Frolov, V. V. Ryazanov, and D. J. Van Harlingen, Second-harmonic current-phase relation in Josephson junctions with ferromagnetic barriers, Phys.DOI
- [72]O. N. Dorokhov, On the coexistence of localized and extended electronic states in the metallic phase, Solid State Commun 51, 381 (1984).
- [73]P. A. Mello, P. Pereyra, and N. Kumar, Macroscopic approach to multichannel disordered conductors, Ann Phys 181, 290 (1988).DOI
- [74]C. E. Moreau, I. C. Moraru, N. O. Birge, and W. P. Pratt, Measurement of spin diffusion length in sputtered Ni films using a special exchange-biased spin valve geometry, Appl. Phys. Lett 90, 012101 (2007).DOI
- [75]J. Bass, Giant magnetoresistance: Experiment, in Handbook of spin transport and magnetism, edited by I. Zutic and E. Y. Tsymbal (CRC Press, 2011) pp. 69–94.DOI
- [76]E. Y. Tsymbal, D. Pettifor, and S. Maekawa, Giant magnetoresistance: Theory, in Handbook of spin transport and magnetism, edited by I. Zutic and E. Y. Tsymbal (CRC Press, 2011) pp. 95–114.DOI
- [77]D. Y. Petrovykh, K. N. Altmann, H. Höchst, M. Laubscher, S. Maat, G. J. Mankey, and F. J. Himpsel, Spin-dependent band structure, Fermi surface, and carrier lifetime of permalloy, Appl. Phys. Lett 73, 3459 (1998).










