# Full text of "Theoretical description of the ferromagnetic $π$-junctions near the critical temperature"

## See other formats

Theoretical description of the ferromagnetic 7r-junctions near the critical temperature A. Buzdin and I. Baladie Condensed Matter Theory Group, CPMOH, UMR 5798, Universite Bordeaux 1, 33405 Talence Cedex, France (02/12/2002) Abstract The theory of ferromagnetic vr-j unction near the critical temperature is pre- sented. It is demonstrated that in the dirty limit the modified Usadel equation adequately describes the proximity effect in ferromagnets. To provide the de- scription of an experimentally relevant situation, oscillations of the Josephson critical current are calculated as a function of ferromagnetic layer thickness for different transparencies of the superconductor-ferromagnet interfaces. I. INTRODUCTION The strong exchange field acting on the electrons in a ferromagnet provokes the damped oscillatory behavior of the superconducting order parameter. This effect is at the origin of the TT-junction realization in superconductor-ferromagnet-superconductor (S/F/S) systems [1,2]. The study of the superconducting vr-state sheds more light on the coexistence of superconductivity and magnetism in general and may be also important for superconducting electronics (see for example [3]). Recently such vr-j unctions have been successfully fabricated and experimentally studied in [4-6]. While the existing theoretical approach [2] provides a good qualitative description 1 of the observed damped oscillations of the critical current as a function of ferromagnetic layer thickness [5,6], a more detailed description is needed to take into account the finite S/F boundary transparency Indeed, in [5] Nb/Al/Al203/PdNi/Nb junctions have been studied. In such junctions, the presence of the AI2O3 barrier can be modeled by a very low transparency of one of the S/F interfaces. In this work we develop a theory of ferromagnetic junctions with different transparencies of the interfaces providing a general description of such junctions near the critical tempera- ture. II. GENERAL FORMALISM We will concentrate on the studies of the properties of an S/F/S junction with a thin F- layer of thickness d and large superconducting electrodes, see Fig. 1. The very convenient set of equations describing an inhomogeneous superconductivity has been elaborated by Eilen- berger [7]. These are transport-hke equations for the energy- integrated Green's functions / and g, assuming that relevant length scales are much larger than atomic length scales. For our geometry all quantities depend only on one coordinate x, and in the presence of the ferromagnetic exchange field h{x) acting on the electron spins in the F-layer, the Eilenberger equations take the form (see for example (u + ih{x) + ^G{x, a;)) / (x, u) + ]^vf cos - {^{x) + a;)) g (x, d, uj) , G{x,uj) ^ j ^g{x,e,uj) , F{x,uj) ^ j {x,e,uj) , (1) f{x,e,u;)r {x,e,u;) + g^ {x,e,u;) = 1, where u = 27rT (n + 1/2) are the Matsubara frequencies and 6 is the angle between the X axis and the direction of the Fermi velocity vp, and r is the elastic scattering time. In the following, we consider the behavior of S/F/S systems close to the critical transition temperature Tc, so we may linearize equations (1) by putting g = sign (a;) . Usually the electron scattering mean free path in S/F/S systems is rather small. In such a dirty limit the angular dependence of the Green's functions is weak, and the Eilenberger equations can be replaced by much simpler Usadel equations [9]. In fact, the conditions of the applicability of Usadel equations are TcT <C 1 and hr <^ 1. The second condition, is much more restrictive due to a large value of the exchange field {h » Tc). Therefore the attempts to retain in the Usadel equations the first correction in the parameter hr have been made in [11-13]. They have resulted in the renormalization of the diffusion constant in F-layer -D/ — > Df{l — 2ihr), where Df — Vpr/3. The critical analysis of this renormalization in [14] has revealed its inaccuracy, but has not provided the right correction. To resolve this controversy we perform below a careful derivation of the Usadel equation for an F-layer retaining the linear correction over parameter hr. In the ferromagnet the Cooper pairing is absent, so A = 0, and considering the case of the strong exchange field uu <^ h, we may write the linearized Eilenberger equation for a function / as where vp and r refer to the Fermi velocity and the scattering time in the ferromagnet. Let us recall that F is an anomalous Green function averaged over the angles. Then we may present f as f — F{x) + fi{x, 9), where /i ^ F. So for a; > ihf + — + -vp cose*— 2r 2 ox 0. (3) The averaging of this equation over the angles gives (4) Multiplying (3) by cos6', averaging and taking the derivative over x we obtain Then combining (3) and (4) we have 3 Neglecting in this equation the /i term, we obtain the usual Usadel equation with the renormalized diffusion constant Df Df{l — 2i/ir). However the omitted term gives a correction of the same order as that coming from the diffusion constant renormalization [14]. It effectively means that it is needed to keep higher order terms in the expansion of / over cos^ (Legendre polynomials). Multiplying (3) consecutively by cos^^ and cos^^ and taking derivative over x, we obtain after some algebra the following exact equation ihF — 2(2ihr + l) [(2ihr + l) 3 (2i/iT + 1) [y''^ dx^j b dx^ ^ ld''F\ ^ +3a^r°- (7) Here we may already safely neglect the /i term, as it gives a contribution ~ {Jit) . In the result, in the linear approximation over /ir, the Usadel equation is written as Tvl d^F d^F _ 6(4i/iT+l) ^ > Taking in mind that in zero approximation over /ir, the function F satisfies the equation we may finally rewrite (8) in the following form applicable for positive and negative uj ■ ■ / M.r. Df[l-i{2/b)hTsign{uj)]d^F ^ isign {uj) hF ^ = 0. (10) So we conclude that the first correction in the parameter hr to the Usadel equa- tion leads to the somewhat different renormalization of the diffusion constant Df — > Df{l — i{2/5)hTsign{u!)), comparing to what has been suggested before [11-13] {Df Df[l — i2hTsign {u;)]). This result has been also independently obtained by Tagirov [15]. In the superconducting layer, the linearized Usadel equation is , , ^, , Dsd^F{x,u) ., , 1^1 F{x,u;) - -^—^^ = ^(^)' (11) A(x) = |A|7rT^F(x,u;), (12) 4 where A is the BCS couphng constant and Ds is the diffusion coefficient in S layer. The equations (10) and (11) are completed by the general boundaries conditions at the S/F interfaces [16] F,(rf/2) =F;(ci/2)+07B2 . Jd/2 F,(-d/2)^Ff(-d/2)-^nsi[^ , (13) V / —d/2 'dFA cr„ / dF, J ±d/2 \dx J _^^^2 where the notation Fs{Ff) is used for the anomalous Green function in a superconductor (ferromagnet) and (T„ (ag) is the conductivity of the F-layer (S-layer above Tc) , = y ^ is the characteristic length of superconducting correlations decaying in F-layer, and = y ^f' is the superconducting coherence length of the S-layer, the parameter 7^ = ^^y^, where Ri, is the S/F boundary resistance per unit area. Here we introduce two different parameter ■jBi and 7s2, for the left and right S/F interfaces, assuming that their boundary resistances may be different. Note that the parameter 7^ is directly related to the transparency of the S/F interface T = j^p^ [17]. The limit T = (7^ = 00) corresponds to a vanishingly small boundary transparency, and the limit T = 1 (7^ = 0) corresponds to a perfectly transparent interface. In principle, the equations (10) and (11) with boundary conditions (13) give the complete description of the S/F/S junction near the transition temperature. III. CALCULATION OF THE JOSEPHSON CRITICAL CURRENT Let us start with a general solution of the Usadel equation in the ferromagnet for a; > Ff {u; > 0, x) = sinh(A;x) + cosh(A;a:), (14) where A; = [(1 - 0.2/ir) + i(l + 0.2/iT] Similarly for a; < Ff (a; <0,x) = sinh(A;*a;) + cosh{k*x) (15) 5 We see that the complex wave- vector k describes the oscillating exponential damping of the anomalous Green function inside the ferromagnet. In the case of very small scattering time (r — > 0) , the imaginary and real parts of the wave- vector k coincide, and so the char- acteristic lengths for oscillating and damping are exactly the same. The finite scattering length decreases the oscillating period and increases the damping length. This is consistent with the fact that in a pure limit the exponential damping is replaced by a much slower power damping, while the oscillations have a much shorter period comparing to the dirty limit [1,13,18]. To obtain the analytical description of the critical current in S/F/S junctions, it is useful to introduce the functions and F~ [2] F+ {lj>0,x) ^ F{lj>0,x) + F{lj < 0, x) , (16) F- {uj>0,x) ^ F{uj > 0,x) - F{uj < 0, x) . (17) In a superconductor, the equation for is the same as (11) with A — > 2A, while the equation for F~ is much simpler , , ^ , , Dsd'^F-ix,uj) ^ 1^1 F;{^.^) - -y ^ = 0, (18) and the self consistency equation reads as I^^\\\nTY.Ft{x,uj). (19) a;>0 The boundary conditions (13) will be the same for F+ and F~ functions too. Using the solution of (18) for F~{x,io), it may be demonstrated (see also [2]) that in the case of a rather large resistivity of the F layer ^ l) or in the case low transparency of the interfaces, when the condition ^j^^ <^ 1 is satisfied, the boundary conditions for F^ functions are essentially simplified ^ d/2 (dF~ \ ^ =0- (21) ^ -d/2 Further on, introducing the functions :F+ = |A|7rT^F+(u; > 0,x), (22) a;>0 JP- = |A|7rT^F-(a; > 0,a;), (23) u)>0 we obtain the following system of equations for the order parameter and its derivatives on the left and right sides of the junction ^ =0, (24) ^ =0' (25) / —d/2 A (x = d/2) = A+ = {d/2) + e/7s2 [-^\ , A (:, = -d/2) = A- = (-d/2) - e/7Bi (^^J (26) (27) (28) (29) This system of six equations, after the elimination of four coefficients of the type ^|A| ttT ma-y be reduced to a standard linear form of (^-) , (^^)) ^~ A+ = Mn A- + Mi2 j , (30) which was used by De Gennes in his general description of the Josephson junction close to Tc [19]. Following the approach [19], the critical current of the SFS junction is given directly by the expression ^ -ieT:Ns{Q)Ds (^A*-A+ - c.c) , (32) where iVs(O) is the electron density of states in S electrodes when they are in the normal state. The calculation of the coefficient M12 is straightforward but rather cumbersome for a general case. The corresponding formula is to intricate to be presented here. So, to discuss the essential physics, we consider below only the most interesting limits. In the case of highly transparent interfaces, it is possible to obtain an analytical expres- sion of the critical current taking into account the first order correction over hr in the Usadel equation (10) ttA^ an 8eT, k + c.c. (33) sinh cosh where the complex wave vector k — [{1 — 0.2/it) + i(l + 0.2/it] -sJ^- The dependences Ic{d) for different parameter hr are presented in Fig. 2. When the critical current goes through zero, the transition from to vr-shift Josephson contact takes place. As expected, with the increase of the scattering time the period of oscillations becomes somewhat shorter, while its amplitude increases. This is consistent with a power low decrease of the critical current with F-layer thickness in the clean limit [1] . Further we concentrate on the influence of the S/F boundary transparency on the Ic oscillations, and then we neglect the hr correction in the diffusion coefficient. Starting with the case of completely transparent interfaces 7^1,2 0, we retrieve the corresponding expression previously obtained in [2] sin(a;) cosh(,x) + cos(a;) sinh(a:) (34) cos(2x') — cosh(2a;) where x — d/^f. In the case of low transparency of both interfaces 7bi,2 ^ 1, the critical current is 1 (Tr. vrA^ cos(a;) sinh(x) — sin(a;) cosh(a;) cos(2x) — cosh(2a;) (35) If one interface (the left S/F interface) is transparent, 7^1 — > 0, the critical current is o^ttA^ cosh(a;) sin(a;) + cos(a;) sinh(x) + 27^2 cos(x) cosh(a;) [1 - 27I2) cos(2a;) - (1 + 27^2) cosh(2a;) - 27^2 [sin(2a;) + sinh(2a;) (36) 8 If one interface (the first S/F interface) has a low transparency, 7^1 » 1, the critical current is cos{x) cosh(,T) — [cosh(a;) sin(x) — cos(a;) sinh(x)] ;i - 27I2) cos(2x) + (1 + 27I2) cosh(2x) - 27B2 [sin(2x) - sinh(2x) (37) The evolution of the Ic{d) dependence with the decrease of the transparency of the second interface is presented in Fig. 3. We observe that with the increase of 7^2 the amplitude of oscillations decreases, as well as the thickness of the ferromagnetic layer corresponding to the first zero of the critical current. In the experiment [5], the presence of AI2O3 barrier at one S/F boundary can be modeled by a low transparency interface (7^1 ^ 1) while the other boundary is quite transparent 7s2 — 0. The resistance i?„ per unit area of the S/F/S junction in this experiment is dominated by the resistance of the tunnel barrier and thus i?„ ~ Qq j^g^y expect that the experimental situation [5] must be described by the expression (37) with 7^2 = i.e. IrR cos(x) cosh(a;) (38) cos(2a;) + cosh(2a;) Note that this expression is quite different from that used in [5] to fit the experimental data. Unfortunately in [5] there is no information which could shed light not only on the reasons of the discrepancy between our formula but even at the origin of the used theoretical expression. The experimental results on the Idd) oscillations presented in [5] have been obtained at low temperature. Nevertheless we think that the overall shape of the Ic{d) curve is not very sensitive to the temperature and we compare the experimental data [5] with our theoretical expression (38) in Fig. 4. We use as the fitting parameter ~ 30 A, while the experiment [20] provided a value of around 35 A in our notation. The value of the other fitting parameter is 110 fj,V; this parameter is quite difficult to estimate because of the uncertainty on the value of A at the S/F interface in the geometry of experiment [5]. Also following the analysis of the authors [5] presented in their previous publication [20], 9 we have taken into account that the actual ferromagnetic thickness of PdNi layer is reduced by 15 A due to some interdiffusion at the S/F interface. The obtained description of the experimental data [5] is quite satisfactory. In conclusion, we have presented the general theoretical description of the ferromagnetic TT-junctions near superconducting transition temperature and proposed a simple way to take into account the finite elastic scattering time in Usadel equation. The obtained analytical expressions may be useful for the analysis of different experimental realizations of such junc- tions. Our analysis may be also easily generalized to the situation when the superconducting electrodes are fabricated from different materials. We thank M. Aprili and V. Ryazanov for stimulating discussions. This work was sup- ported by the ESF "vortex" Programme and the ACT "supra-nanometrique" . 10 REFERENCES [1] A. I. Buzdin, L. N. Bulaevskii, and S. V. Panyukov, Pis'ma Zh. Eksp. Teor. Fiz. 35, 147 (1982) [JETP Lett. 35, 178 (1982)]. [2] A. I. Buzdin and M. Y. Kuprianov, Pis'ma Zh. Eksp. Teor. Fiz. 53, 308 (1991) [JETP Lett. 53, 321 (1991)]. [3] L. B. lofFe, V. B. Geshkenbein, M. V. Feigelman, A. L. Fauchere, and G. Blatter, Nature (London) 398, 679 (1999). [4] V. V. Ryazanov, V. A. Oboznov, A. Yu. Rusanov, A. V. Veretennikov, A. A. Golubov, and J. Aarts, Phys. Rev. Lett. 86, 2427 (2001). [5] T. Kontos, M. Aprili, J. Lesueur, F. Genet, B. Stephanidis, and R. Boursier, Phys. Rev. Lett. 89, 137007 (2002). [6] Y. Blum, A. Tsukernik, M. Karpovski, and A. Palevski, Phys. Rev. Lett. 89, 187004 (2002). [7] G. Eilenberger, Z. Phys. 214, 195 (1968). [8] L. N. Bulaevskii, A. I. Buzdin, M. L. Kuhc, and S. V. Panyukov, Adv. Phys. 34, 175 (1985). [9] L. Usadel, Phys. Rev. Lett. 95, 507 (1970). [10] L. R. Tagirov, Physica C 307, 145 (1998) [11] Y. N. Proshin and M. G. Khusainov, Zh. Eksp. Theor. Fiz. 113, 1708 (1998) [Sov. Phys. JETP 86, 930 (1998)]; 116, 1887 (1999) [Sov. Phys. JETP 89, 1021 (1999)]. [12] Y. A. Izyumov, Y. N. Proshin, M. G. Khusainov, Phys. Uspekhi. 45, 109 (2002). [13] I. Baladie and A. Buzdin, Phys. Rev. B 64, 224514 (2001). [14] Ya. V. Fominov, N. M. Chtchelkatchev, and A. A. Golubov, Phys. Rev. B 66, 14507 11 (2002). [15] L. Tagirov, private communication. [16] M. Y. Kuprianov and V. F. Lukichev, Zh. Eksp. Theor. Fiz. 94, 139 (1988) [Sov. Phys. JETP 67, 1163 (1988)]. [17] J. Aarts, J. M. E. Geers, E. Briick, A. A. Golubov, and R. Coehoorn, Phys. Rev. B 56, 2779 (1997). [18] F. S. Bergeret, A. F. Volkov, and K. B. Efetov, Phys. Rev. B 65, 134505 (2002) [19] P. G. De Gennes, Superconductivity of Metals and Alloys, W. A. Benjamin, New York (1966). [20] T. Kontos, M. Aprih, J. Lesueur, and X. Grison, Phys. Rev. Lett. 86, 304 (2001). Figure captions FIG. 1 Geometry of the considered S/F/S system. The thickness of the ferromagnetic layer is d. The transparency of the left S/F interface is characterized by the coefficient 7^1 and the transparency of the right F/S interface is characterized by 752- FIG. 2 Critical current of the S/F/S junction as a function of the thickness of the ferromagnetic layer normalized by C,f for different scattering time. Both S/F interfaces are completely transparent. The parameter hr is equal to 0, 0.25 and 0.5. FIG. 3 Critical current of the S/F/S junction as a function of the thickness of the ferro- magnetic layer normalized by The first S/F interface has a low transparency (7^1 ^ 1). The parameter 7^2 characterizing the transparency of the second interface is chosen as 0.2, 1.5, 3.0, and 10. FIG. 4 The experimental points correspond to the measurement of the critical current, done by Kontos et al [5], vs the PdNi layer thickness. The theoretical curve is the best fit obtained by using formula (38). The fitting parameters are ~ 30 A and ~ 110 iJ,V. 12 Figure 2