Density of States in Superconductor -Normal Metal-Superconductor Junctions arxiv:cond-mat/97756v [cond-mat.mes-hall] 7 Nov 1998 F. Zhou 1,, P. Charlat, B. Spivak 1, B.Pannetier 1 Physics Department, University of Washington, Seattle, USA. Centre de Recherches sur les Très Basses Températures, C.N.R.S., Laboratoire associé à l Université Joseph Fourier, F-384 Grenoble, France Abstract We consider the χ -dependence of the density of states inside the normal metal of a superconductor - normal metal - superconductor (SNS) junction. Here χ is the phase difference of two superconductors of the junction. It is shown that in the absence of electron-electron interaction the energy dependence of the density of states has a gap which decreases as χ increases and closes at χ = π. Both the analytical expressions for the χ dependence of the density of states and the results of numerical simulations are presented. PACS index category: 5.-y, 8.-w I. INTRODUCTION The proximity effect in a normal metal near superconductor-normal metal boundary has been studied for many years. Recently, progress in microfabrication technology has revived the interest in this phenomenon. Due to Andreev reflections at superconductor-normal metal boundaries, the density of states inside the normal metal of a superconductor-normal metal (SN) junction is different from that of the bulk normal metal. This phenomenon has been studied both theoretically and experimentally 1 5. In the superconductor - normal metal - 1
superconductor (SNS) junction it has been shown that at χ = the Andreev reflections at the superconductor-normal metal interfaces cause the density of states to have a gap, which is of the order of the Thouless energy E c = D/L. Here χ is the phase difference between the two superconductors, is the modulus of the order parameter in the superconductors, L is the length of the normal metal of the junction (see Fig.1) and D is the diffusion constant in the metal. In this paper we study the χ dependence of the density of states ν(ǫ, χ ) in the case when the length of the normal metal region is much larger than the electron elastic mean free path (diffusive regime). We study both analytically and numerically the behaviour of the density of states near the gap edge. This quantity can be measured for example in a tunneling experiment. We will show below that it also determines the conductance of the junction G SNS at low temperatures T. The density of states in the normal metal can be expressed in term of the retarded Green s function g R (ǫ, x) = cosθ(ǫ, x) 6 ν(ǫ) = ν L/ cosθ 1 (ǫ, x) cosh θ (ǫ, x)dx. (1) L L/ Here ν = mp F is the density of states in the bulk normal metal and p F is the Fermi wave length; θ = θ 1 + iθ and χ = χ 1 + iχ are the complex variables described by Usadel equations 6, D x θ(ǫ, x) + iǫ sin θ(ǫ, x) D 4 ( xχ(ǫ, x)) sin θ(ǫ, x) = N (x) cos θ, x { x χ(ǫ, x) sin θ(ǫ, x)} =, () N (x) = γ N dǫ cosθ 1 (ǫ, x) sinh θ (ǫ, x) tanh(ǫ/kt). (3) N and γ N > are the modulus of the order parameter and the dimensionless repulsive interaction constant inside the normal metal. The boundary conditions for Eq. are (we consider the case when the transmission coefficient of the normal metal-superconductor boundary t = 1),
θ(ǫ, x = ± L ) = π, χ(ǫ, x = ± L ) = ±χ. (4) Consider the situation when there is no electron-electron interaction in the normal metal and N =. We will see that in this case the energy gap E g (χ ) of the electron spectrum is a decreasing function of χ and is equal to zero only at χ = π (See Fig. ). In two limiting cases, when χ is close to or π we have C (1 C 1 χ ) E g (χ ) = E c C 3 (π χ ) χ π π χ π (5) II. THE ENERGY GAP To get this result we take into account that at ǫ < E g (χ ), θ 1 = π/ following Eq.1. In this case the second integral of Eq. gives the solution for θ (ǫ, x): θ θ (ǫ,x) dθ { ǫ E c (sinh θ sinh θ ) + α (ǫ)(cosh θ cosh θ )} 1 = x L, (6) where θ (ǫ) and α(ǫ) are the functions of energy ǫ and χ determined by the equations θ dθ { ǫ E c (sinh θ sinh θ ) + α (ǫ) θ α(ǫ) dθ cosh θ { ǫ (sinh θ sinh θ ) + α (ǫ) E c (cosh θ cosh θ )} 1 (cosh θ cosh θ )} 1 = 1, χ =. (7) One can show that Eq.7 has solutions only at low energy ǫ. The energy gap E g (χ ) corresponds to the maximum value of ǫ, at which a solution exists. Beyond this value, θ 1 π/ and therefore the density of states ν(ǫ) is non zero. Let us first consider the limiting case corresponding to χ π. We can expand Eq.7 with respect to χ and as a result we have the following equation: 3
ǫ θ χ = dθ (sinh θ sinh θ ) 1 ( E c ) A(θ ), (8) where A is of the order of unity. The right hand side of Eq.8 as a function of θ has a maxima equal to C (1 C 1 χ /). Here C 1, are numerical factors of order of unity. Their values can be obtained from the numerical solution of the Eq. 7 : C 1 =.91 and C = 3.1. At low energies when ǫ E c, Eq.8 has a solution θ ǫ/e c while at large energies ǫ > C E c (1 C 1 χ ), Eq.8 does not have a solution. Therefore in this limit E g = C E c (1 C 1 χ ), as given in Eq.5. When χ π π, the energy gap is small, E g E c and we can expand Eq.7 with respect to the small parameter ǫ/e c. As a result we have ǫ (π χ )E c = π tanh θ {8B 1 + (π χ )πb sinh θ } 1. (9) where B 1, are also constants of the order of unity. Again the right hand side of Eq.9 as a function of θ reaches a maxima when cosh θ = B (π χ ) 1 3 (1) Therefore, at ǫ > C 3 E c (π χ ), Eq.9 has no solution and E g (χ ) is given by Eq.5. From the numerical solution we find B =.1 and C 3 =.43. The insert of Fig. shows the linear dependence of the energy gap near χ = π. III. THE GAP EDGE Let us turn to the calculations of the ǫ-dependence of the density of states ν(ǫ, χ ) at ǫ > E g. In the region ǫ E g E g, the quantities δθ 1 (ǫ, x) = θ 1 (ǫ > E g, x) π, δθ (ǫ, x) = θ (ǫ > E g, x) θ (E g, x), 4
δχ 1 (ǫ, x) = χ 1 (ǫ > E g, x) χ 1 (E g, x), δχ (ǫ, x) = χ (ǫ > E g, x), (11) are small and go to zero as ǫ approaches E g. Here θ (E g, x), χ 1 (E g, x) are the solution of Eq. at ǫ = E g given by Eqs.6,7. When π χ π where θ is given in Eq.1. sinh θ (E g, x) = sinh θ cos πx L, π cosh θ x χ 1 (E g, x) = L cosh θ (E g, x), (1) Expanding Eq. with respect to δθ 1,, δχ 1, we get the following set of equations { D x + E g sinh θ + D ( xχ 1 ) cosh θ }δθ 1 D χ 1 sinh θ x δχ δ 3 θ 1 ( E g 6 sinh θ + 1 3 D( xχ 1 ) cosh θ ) +δθ 1 (E g cosh θ δθ + D( x χ 1 ) sinh θ δθ + D x δχ 1 x χ 1 cosh θ D ( xδχ ) cosh θ ) +δ θ 1 D x δχ x χ 1 sinh θ D x δχ δθ x χ 1 cosh θ D xδχ δχ 1 sinh θ = (ǫ ǫ g ) sinh θ δθ 1 (13) { D x + E g sinh θ + D ( xχ 1 ) cosh θ }δθ + D xχ 1 sinh θ x δχ 1 δ θ 1 ( E g cosh θ + D ( xχ 1 ) sinh θ ) + δθ 1 D x δχ x χ 1 cosh θ D 4 ( xδχ ) sinh θ = (ǫ ǫ g ) cosh θ (14) { x δχ cosh θ δθ 1 x χ 1 sinh θ } = (15) x { x δχ 1 cosh θ + δθ x χ 1 sinh θ δθ 1 x χ 1 cosh θ + δθ 1 x δχ sinh θ } = (16) with the boundary conditions for δθ 1,, δχ 1, : δθ 1 (ǫ, x = ± L ) = δθ (ǫ, x = ± L ) =, δχ 1 (ǫ, x = ± L ) = δχ (ǫ, x = ± L ) =. (17) 5
Consider the linear part of Eq.13-17 at ǫ E g =. Its solution determines the spatial dependence of δθ 1,, δχ 1,. Averaging Eq.13-17 over the sample and comparing the nonlinear terms in Eq.13 with right hand side term proportional to (ǫ E g ) sinh θ δθ 1, we can determine the value of δθ 1 as a function of ǫ E g. As a result we have ǫ Eg (χ ) δθ 1 cosh θ C E 4(χ ) (18) c where C 4 is a function of E g and consequently of χ. Substituting Eqs.1, 18 into Eq.1 we obtain the asymptotic form valid when ǫ E g E g : ν(ǫ, χ ) = ν C 4 (χ ) ǫ E g(χ ) E g (χ ) (19) with C 4 (π χ ) β, where β /3. The value of β is a result of a numerical calculation over the whole energy range 7. The square root energy dependence of the density of states is illustrated in Fig.3 which shows the square of the density of states as function of the energy very close to the energy gap. At higher energy (Fig.3 insert), we find that the density of states exhibits a smooth bump above the gap when phase difference is small. This smooth maximum turns to a sharper and sharper peak as the phase difference approaches π, i.e. as the gap closes. It should be noted at this point that the asymptotic density of states at high energy or at χ = π is smaller than the normal state value. The region near the superconductor provides a small contribution to the (spatially averaged) density of states. This apparent deficit of states is balanced by the excess density of states above the superconducting gap which according to our assumption is far above the energy range of interest. This deficit does not exist at the center of the normal metal or in a geometry where the relative area of the boundary between N and S goes to zero. The latter case is met in the billard geometry of Ref 4. IV. DISCUSSION 6
Although the above results were derived for the junction geometry shown in Fig.1, the χ and ǫ dependences of the density of states in Eqs.5,19 as well as the statement that the gap closes at χ = π are general 9. We believe that they are independent of the transmission coefficient of the superconductor-normal metal boundaries and the geometry of the normal region. The values of E c and C i, however, depend on these parameters. Similar conclusions were also reached in 4. So far we assumed the electrons inside the normal metal do not interact with each other. In the presence of electron-electron interactions, N is not zero and one can study the interaction effects on the density of state by taking into account N in Eq.7 for E g. Since γ N 1, one can carry out the perturbative calculation with respect to γ N. As a result, at χ π, the gap turns out to be smaller than that given in Eq.5 for the noninteracting case, i.e. E g () E g (γ N ) γ N E c. Furthermore, the gap closes at χ smaller than π π χ γ N () Quantum fluctuations of the phase of the order parameter can also change the results derived above. At finite temperature, in principle one also has to take into account the electron level broadening due to inelastic scattering. Such an inelastic process will introduce a temperature dependent density of state at the Fermi surface, i.e. ν(ǫ = ) 1/E c τ ǫ. However at T E g (χ ), the inelastic scattering rate τ 1 ǫ due to electron-phonon interaction is exponentially small. Finally, let us calculate the conductance of the junction at T E c. The conductance G SNS is the proportionality coefficient between the applied voltage and the disspative current averaged over the period of Josephson oscillations. At low temperature, the main contribution to G SNS comes from the Debye relaxation mechanism 8. The qualitative picture is the following. When voltage V is applied across the junction, the time dependence of χ is determined by the Josephson relation dχ dt = ev (1) 7
At small V the time dependence of ν is determined by the corresponding time dependence of χ (t). In other words the quasiparticle energy levels move adiabatically with frequency ev. The electron population of the energy levels follows the motion of the levels and as a result the electron distribution becomes nonequilibrium. The relaxation of the nonequilibrium distribution due to inelastic processes leads to the entropy production and therefore contributes to the conductance. As a result, we have 8, G SNS = e v h ν dǫ ǫ tanh ǫ kt π ǫ dχ τ ǫ (χ ){ dǫ dν(ǫ, χ ) } () dχ where τ ǫ (χ ) is the energy relaxation time of quasi particle of energy ǫ. In this case 1, 1 τ ǫ (χ ) ǫ3 exp( E Ω g (χ )/kt) (3) D is exponentially small in the time interval when E g (χ (t)) T. Ω D is the Debye frequency. On the other hand, the concentration of quasiparticles in this case is also exponentially small. These two exponential factors cancel each other and the main contribution to Eq. comes from the time interval when E g (χ (t)) T. Since χ changes linearly with time and E g (χ ) vanishes linearly as a function of χ π at χ close to π, following Eq. we have Here G N = e Dν S L and τ 1 in G SNS G N C 4( T E c )τ in T (4) T 3 /Ω D. C 4 is given in Eq.19. At E c T, Eq.4 matches the result of the conductance, G N τ in E c/t, which was obtained in 8 at the high temperature limit T E c. Following the above arguments, the time dependence of the conductance has the form of narrow peaks with the amplitude of order of G N C 4 (T/E c)τ in E c and duration of the order of (ev ) 1 TE c. This phenomenon can be connected with the well known cos χ problem which has been investigated both experimentally and theoretically 11,1. V. CONCLUSION 8
In conclusion we have shown that the density of states in the normal part of a SNS junction has a gap which closes when the superconducting phase difference is χ = π. The energy dependence of the density of states near the gap edge has been calculated. We should mention that nonequilibrium effects in superconductors which are connected with time dependence of quasiparticle spectrum have been considered in 13,14. In these papers the enhancement of the critical current due to the nonequilibrium effects was considered. The effect considered there is proportional to (ev τ ǫ ) while the DC current calculated above is linear in (ev τ ǫ ). We would like to acknowledge discussions with H. Courtois. This work was partially supported by the NATO CRG 96597. 9
REFERENCES 1 A. A. Golubov, M. Y. Kupriyanov, Pisma. Zh. Eksp. Teor. Fiz. 61, 83 (1995) [ JETP Lett. 61, 851 (1995)]. A. Altland, and M. R. Zirnbauer, Phys. Rev. Lett. 76, 34 (1996). 3 K. M. Frahm, P. W. Brouwer, J. A. Melsen, C. W. J. Beenakker, Phys. Rev. Lett. 76, 981 (1996). 4 J. A. Melsen, P. W. Brouwer, K. M. Frahm, C. W. J. Beenakker, Physica Scripta T69 (1997) 3-5. 5 S. Guéron, H. Pothier, N. O. Birge, D..Estève, M. Devoret, Phys. Rev. Lett. 77, 35 (1996). 6 F. Zhou, B. Spivak, A. Zyuzin, Phys. Rev. B5, 4467 (1995) and the references therein. 7 P.Charlat, PhD thesis, University J. Fourier Grenoble, 1997, unpublished. 8 F. Zhou, B. Spivak, Pisma. Zh. Eksp. Teor. Fiz. 65, 347 (1997) [JETP Lett. 65, 369 (1997)]. 9 The approach in this paper (Usadel Equations) assumes that the electron spectrum is continuous. The statement that the gap closes at χ = π is valid when the single particle level spacing is taken to be zero. A minigap of order of one level spacing was obtained using the Random Matrix Theory by A.Altland and M.R. Zirnbauer, Phys. Rev. B55, 114, (1997). Its discussion is beyond our approximation. 1 A. G. Aronov, Y. M. Galperin, V. L. Gurevich and V. I. Kozub, Adv. Phys. 3, 539 (1981). 11 N. F. Pedersen, T. F. Finnegan, D. N. Langenberg, Phys. Rev. B6, 4151(197). 1 A. E. Zorin, I. O. Kulik, K. K. Likharev, J. R. Schrieffer, Fiz. Nizk. Temp. 5, 1138 (1979) [Sov. Jour. Low. Temp. Phys. 5(1), 537(1979)]. 1
13 L.G.Aslamasov, A.I.Larkin, Zh. Eksp. Teor. Fiz. 71, 134 (1976) [Sov. Phys. JETP. 43, 698 (1976)]. 14 A. Schmid, G. Schon, M. Tinkham, Phys. Rev. B1, 576 (198). 11
FIGURES S L S - χ N χ x FIG. 1. The S-N-S system: the origin of coordinate x is the center of the normal metal (length L). The phases of the superconducting order parameters in the right and left superconducting electrodes are respectively χ / and +χ /. E g /E c 3..95π π 1.5π 1 π χ π FIG.. Energy gap E g in units of the Thouless energy E c vs the phase difference χ between superconducting contacts. The insert shows the linear dependence of E g near χ = π. 1
.4 (ν/ν ). ν/ν 6 5 4 3 1 1 7 3 6 ε/e c 6 5 4 3.99 1 1.1 ε/e g FIG. 3. The square power of the density of states vs reduced energy ǫ/e g (χ ) near the gap edge. Curves labeled, 1,, 3, 4, 5, 6 and 7 are for χ =, π/, 3π/4, 7π/8, 15π/16, 31π/3, 63π/64 and π. The insert shows the full density of states curves (here E g ). C 4 1 1.1 1 π χ -/3 FIG. 4. Divergence of the C 4 numerical coefficient near χ = π. The dashed line is an asymptotic law (π χ ) β with β = /3. 13