Theory of generalized Josephson effects

Aron J. Beekman∗ Department of Physics, and Research and Education Center for Natural Sciences, Keio University, 3-14-1 Hiyoshi, Kohoku-ku, Yokohama 223-8522, Japan (Dated: August 1, 2019) The DC Josephson effect is the flow of supercurrent across a weak link between two superconduc- tors with a different value of their order parameter, the phase. We generalize this notion to any kind of spontaneous continuous symmetry breaking. The quantity that flows between the two systems is identified as the Noether current associated with the broken symmetry. The AC Josephson effect is identified as the oscillations due to the energy difference between the two systems caused by an imposed asymmetric chemical potential. As an example of novel physics, a Josephson effect is pre- dicted between two crystalline solids, potentially measurable as a force periodic in the separation distance.

The prediction by Josephson [1] of a DC current flow- (see Fig. 1). The only assumption is that the order pa- ing between two superconductors separated by a small rameters of the two systems couple to each other in a way distance, with zero voltage bias, revealed two important that favors a uniform configuration. This assumption is concepts. First, the order parameter can rightfully be very plausible if one considers the two systems as parts of regarded as a quantum field with an equation of motion, one large system that therefore prefers to break the sym- that does not vanish abruptly at the edge of a sample metry uniformly. Explicitly, the coupling Hamiltonian but must fall off in a continuous fashion. Second, it set- takes the form: tled the debate around broken symmetry in superconduc- K  † †  tors [2, 3]. It may then be expected that a similar effect HK = − O OR + O OL . (1) 2 L R occurs in any system with spontaneously broken symme- try, since the only prerequisite seems to be a coupling This term must be compensated by HK0 = K|OL||OR| between two systems whose order parameters take dif- so that the energy vanishes when the order parameters ferent values. Indeed, such generalizations have been ex- on left and right are the same. The coupling parameter plored. Of course the phenomenon shows up in helium-4 K > 0 is small with respect to other energy scales. superfluidity because the broken symmetry is essentially Without the coupling Eq. (1), the Hamiltonian H0 is the same as in superconductors, barring the coupling to invariant under a continuous symmetry group G. We gauge fields [4]. Josephson tunneling is also found in can consider the two systems independently, with sym- a a superfluid helium-3, which breaks a non-Abelian symme- metry generators QL and QR, where a runs over the di- try [5]. The Josephson effect has further been generalized mensions of the Lie algebra of G. A symmetry genera- to SO(5)-symmetry proposed for high-Tc superconduc- tor is the volume integral of a Noether charge density: a R a tivity [6, 7]. A spin current flowing between two fer- QL,R = V ∈L,R jt (x, t), and the Noether currents are lo- a a romagnets [8–11] or antiferromagnets [12–14] is another cally conserved ∂tjt (x, t) + ∇ · j (x, t) = 0. We assume manifestation of the Josephson effect. that the symmetry is broken down spontaneously to a Nevertheless, the Josephson effect as an intrinsic prop- subgroup H ⊂ G, due to formation of the order param- erty of spontaneous symmetry breaking has received little eters mL, mR. These order parameters transform under attention, apart from Esposito et al. [15]. They elegantly some representation of G, where the generators Qa are a describe the Josephson effect of a system with inter- represented by Hermitian matrices T ij, so that a general a nal symmetry group O(N) as the appearance of pseudo- transformation generated by QL is parametrized as Goldstone modes due to the explicit but weak breaking iαT a of the doubled symmetry group O(N) × O(N) that gov- mL,i → e ij mL,j, α ∈ R, (2) erns the two uncoupled systems, to the diagonal subgroup and the same for the right system. For convenience we O(N) that leaves the total charge invariant while the rel- have chosen mL, mR to transform under a vector rep- arXiv:1907.13284v1 [cond-mat.other] 31 Jul 2019 ative charge is broken. This method has been applied to resentation, but this is not necessary for anything that spinor Bose–Einstein condensates [16]. We derive the Josephson effect purely in terms of sym- metry transformations of the order parameter operator O. Suppose there are two systems on the left (L) and right (R) in states |ψLi, |ψRi which break some symme- try by the formation of uniform order parameter expec- tation values mL = hψL|OL|ψLi and mR = hψR|OR|ψRi FIG. 1. Typical Josephson setup. Two systems left and right are separated by a small distance w so that their order pa- rameters are coupled. Josephson current I can flow from one ∗ [email protected] system to the other. 2 follows. This transformation is equivalent to the two systems. This can be made more explicit, by writing the left-hand side of Eq. (6) as h[Qa , O ]i = hT aO i = T a m , (3) L L L ij L,j Z Z a 3 a 3 a a h∂tQLi = d x h∂tjL,ti = − d x h∇ · jLi where by abuse of notation we have also used T as the V V linear transformation acting on the operator O. By Her- I † † † a a a† a = − dS · hjLi = −Iboundary. (8) mitian conjugation [Q , OL] = −OLT = −OLT be- cause Qa and T a are Hermitian. ∂V a The space of values that the order parameters mL and The change of Noether charge hQLi is caused by flow of a mR can take is isomorphic to the quotient space G/H, Noether current hjLi through the boundary ∂V , and the and Eq. (2) can be considered as a rotation in that space. sign of I denoting flow out of the left system (see Fig. 1). We are interested in the situation where mL and mR take This connects with the aforementioned work by Espos- different values in G/H. All the symmetry groups we ito et al. [15] since the broken Noether charge density consider are transitive, which implies that we can always excites Goldstone modes. It also demonstrates that the choose the generators T a in such a way that a single spin Josephson current between two ferromagnets and generator T r connects the right system to the left: that between two antiferromagnets is the same, because they break the same symmetry, even though their or- iαT r mL = mLm0, mR = mRe m0. (4) der parameters are different. Consequently there is no difference in DC Josephson effect between systems with Here m0 is some unit vector, and mL, mR are the mag- type-A or type-B Goldstone modes. nitudes of the order parameters. Most importantly, this result affirms that the Joseph- DC Josephson effect. Turning on the coupling Eq. (1), son effect is general for any type of spontaneous sym- a a the Hamiltonian no longer commutes with QL and QR metry breaking, depending only on ground state order a a separately, although it does commute with QL + QR. We parameters. can now calculate the expectation value with respect to AC Josephson effect. Compared to the DC effect, the the zero-coupling ground state |ψLi ⊗ |ψRi of the time AC Josephson effect is even more ubiquitous. As soon as derivative of the Noether charge of one system, say the one regards the order parameters mL and mR as quan- left, using the Heisenberg equation of motion: tum fields with their own dynamics, it is obvious that there will be an oscillation ∼ eiµt due to an imposed a 1 a h∂tQLi = − h[H,QL]i energy difference ∝ µ between left and right systems. In- i~ deed, the AC Josephson effect does not even require su-   K † a † a perconductivity and has been suggested to occur in nor- = h [OL,QL]OR + OR[OL,QL] i 2i~ mal metals [17]. In the general setting, it follows from K  † a † a  adding an asymmetric chemical potential: = h OLT OR − ORT OL i 2i~ K   H = µ(Qr − Qr ). (9) = m† T am − m† T am . (5) µ R L 2i L R R L ~ This term clearly commutes with the symmetry gener- r r Here we have used the fact that OL and OR commute ators QL, QR separately and does not modify the DC and act on different parts of Hilbert space, so we can effect. The order parameter operators OL, OR in most take expectation values of their products trivially. The cases do not commute with the Hamiltonian H0, which last line immediately shows that there is no current if naively implies a time dependence. This is associated mR/mR = mL/mL. But if the order parameter of the with the Anderson tower of states and leads to an ex- right system is rotated with respect to the left and takes tremely slow oscillation of the order parameter expecta- the form of Eq. (4) we find the general form of the DC tion value with time scale that grows with the volume, Josephson effect: and which vanishes in the thermodynamic limit [18]. We will neglect this trivial time dependence, and calculate: a K †  a −iαT r iαT r a h∂tQ i = mLmRm T e − e T m0. (6) L 2i 0 iµ r ~ ∂tmL,R = h∂tOL,Ri = ∓ h[QL,R, OL,R]i ~ In all cases of interest the right-hand side is only non-zero iµ r when a = r, for which the equation simplifies to = ± T mL,R. (10) ~

r K † r r If we take the order parameters to be mL,R = h∂tQ i = − mLmRm T sin(αT )m0. (7) r L 0 iαL,RT ~ mL,Re m0, neglecting higher-order effects by set- ting ∂tm0 = 0, ∂tmL,R = 0, we can derive the equations Eqs. (6) and (7) are the main result of this work. of motion for the phases: We now consider its physical implications. First of all, the current is seen to be the flow of Noether charge µ ∂tαL,R = ∓ . (11) associated with the broken generator Qr that connects ~ 3

Integrating this equation gives the time-dependent phase Heisenberg magnet. The Heisenberg Hamiltonian on a µ 0 0 0 0 difference αL,R(t) = ∓ t+αL,R. Denoting α = αR −αL, bipartite lattice with sites i is the DC and AC effects~ can be combined as:   J X K † r 2µ 0 r H0 = Si · Si+δ. (19) I = mLmRm0T sin ( t + α )T m0. (12) 2 ~ ~ iδ

Josephson energy. There is a potential energy associ- Here Si are the SU(2) spin operators on site i with com- a b c ated with two systems that break the symmetry differ- mutation relations [Si ,Sj ] = iabcSi δij, δ runs over all ently. We can simply calculate elementary lattice vectors and J is the coupling energy. The system prefers a ferromagnetic configuration when hHK0 + HK i = K(mLmR − mL · mR) J < 0 and antiferromagnetic (N´eel)configuration when = KmLmR(1 − cos α), (13) J > 0. This Hamiltonian is invariant under global SU(2) spin rotations with generators Sa = P Sa. The Hamil- where for convenience we have chosen a real vector rep- i i resentation, and α is the angle between the two vectors. tonian is not in canonical form and we cannot perform the Legendre transform to a Lagrangian. Still the lattice It is a remarkable fact that two initially causally dis- a connected systems that form order independently due to Noether current ji,i+δ, which is the current that flows uncontrollable symmetry-breaking dynamics, contain po- through the Josephson junction, can be obtained directly tential energy when brought close together. from the equation of motion: The same result can be found in the standard way, by X X ∂ Sa = −  (SbSc +ScSb ) ≡ − ja . (20) calculating the work needed to increase the phase differ- t i abc i i+δ i i+δ i,i+δ R t δ δ ence from α(t = 0) = 0 to α(t) = α as U = 0 dt Iµ. From Eqs. (8), (9) one immediately sees that this indeed If m(x) is a classical vector corresponding to the mag- has units of energy. Substituting Eqs. (7) and Eq. (11): netic moment at x, then this equation is the discretized, Z t quantum version of (cf. Ref. [11]): † r r U = dt KmLmRm0T sin (α(t)T ) m0∂tα(t) 0 ∂tm = −∂m(m × ∂mm) ≡ −∇ · j. (21) Z t  † r z = −KmLmRm0 dt ∂t cos (α(t)T ) m0 A ferromagnet has magnetization Mi = Si as order 0 parameter, while for an antiferromagnet it is staggered † r i z = KmLmRm0 (1 − cos(αT )) m0. (14) magnetization Mi = (−1) Si , which each break two spin-rotations Sx, Sy while rotations Sz around the We shall now demonstrate these effects at the hand of (staggered) magnetization direction are still symmetries. several examples. The symmetry generators Sa act on (staggered) mag- Superconductor/superfluid. Superconductors and su- netization vectors m = hMi in the vector represen- perfluids are described by a complex scalar order pa- tation (T a) = i . Without loss of generality we rameter operator field ψ, with commutation relation ij aij † 0 0 take the configuration where mL = MLm0 and mR = [ψ(x), ψ (x )] = δ(x−x ) and expectation value hψi = ψ. iαT x MRe m0, with m0 the unit vector (0, 0, 1) and ML, The Lagrangian is invariant under global U(1) rotations M the magnitude of the total (staggered) magnetization ψ → e−iαψ, so that T = −1. The Noether charge is R R 3 † of the left and right systems. The coupling Hamiltonian Q = d x ψ ψ. The coupling term Eq. (1) is: Eq. (1) is: K   H = − ψ† ψ + ψ† ψ , (15) H = −KM · M . (22) K 2 L R R L K L R from which the Feynman equations for the Josephson ef- because the order parameter operators are Hermitian. fect can be derived [19]. Taking ψL = |ψL| and ψR = Then we find the DC spin Josephson current Eq. (6): −iα e |ψR|, Eqs. (7), (8) give: x K † x x K h∂tS i = − MLMRm0T sin(αT )m0 I = |ψL||ψR| sin α, (16) ~ ~ K = − MLMR sin α. (23) which is the standard result for the DC Josephson cur- ~ rent. The AC Josephson effect Eq. (11) is: One can also explicitly verify that for these order param- y z 2µ 2eV eters h∂tS i = h∂tS i = 0. This result agrees with the ∂tα = − = , (17) ~ ~ usual form ∂tm = mL ×mR for ferromagnets [9–11]. For where we have set µ = −eV with e the electron charge the antiferromagnet the equations are usually given in and V the electrostatic potential, to reproduce the stan- terms of the staggered magnetization vector [12–14], but dard AC Josephson equation. Similarly the Josephson here we have shown that is rather the spin current that energy Eq. (14) is: is flowing. The Josephson energy of this configuration is

U = K|ψL||ψR| (1 − cos α) . (18) U = KMLMR (1 − cos α) . (24) 4

a R is Q = Πa = V πa, and πi(x) the canonical momentum 0 0 conjugate to φi(x) with [φi(x), πj(x )] = iδijδ(x − x ). For a homogeneous medium we have hφi(x)i = xi, and this suffices to break the symmetry [21]. For a solid, we need to preserve lattice translations, and the order pa- rameter should take values in G/H ' U(1)⊗d. We write d the order parameter as O(x) = ⊗i=1Oi(x), with

2πiφi(x)/ci Oi(x) = e ρ(x), (26)

FIG. 2. The crystalline DC Josephson effect occurs when left where ci is the lattice constant in direction i and ρ(x) the (blue) and right (red) crystalline solids with identical lattice density. This quantity is invariant under lattice trans- are displaced by a vector ~w that is is not a lattice vector. The a lations φi → φi + nci, n ∈ Z. We have [Q , Oi] = empty blue dots indicate the order that is preferred by the 2π a 2π δaiOi, represented by T = δaiδij. For a lat- ci ij ca left crystal. tice with primitive unit cells, the density is ρ(x) = Qd P δ (xi − (ni + ϕi)ci). Here ϕi ∈ [0, 1) is the i=1 ni∈Z The AC Josephson effect follows from the Hamiltonian position of the origin of the unit cell with respect to the coordinate system x, in units of ci, modulo lattice trans- Hµ = µ · (SL − SR) where µ is proportional to an ex- ternal magnetic field imposed in opposite directions for lations. The order parameter expectation value is, aver- i2πϕi left and right systems. While Eq. (10) is valid, it is more aging over one unit cell: hOi(x)i = e , insightful to simply derive: For the DC Josephson effect, we regard two crystals with perfect surface along a crystal direction, separated

∂tmL = −µ × mL, ∂tmR = µ × mR. (25) by a short distance wk, and relatively displaced along the surface by w⊥, see Fig. 2. The Josephson effect occurs for which mutatis mutandis agrees with Ref. [12]. each lattice direction a when wa is not an exact multiple Helium-3. A helium-3 superfluid has triplet pairing of the lattice constant ca. (For the longitudinal effect, and an anisotropic order parameter transforming under it is necessary that the medium in between can support rotations (L), spin rotations (S) and global phase ro- the field φi(x) over the distance wk.) This leads to a tations, with a total symmetry group G = SO(3)L × displacement (phonon) current from the left to the right SO(3)S ×U(1), with seven generators spanning the Lie al- system: gebra. There are many symmetry breaking patterns [20], a K 2π wa but we will here look at the B-phase where the spin is I = sin ∆a, ∆a = 2π . (27) locked to angular momentum, with residual symmetry ~ ca ca group H = SO(3)L+S. There are four broken genera- P The Josephson energy is U = a K(1 − cos ∆a). In con- tors; the order parameter space is isomorphic to G/H ' trast with Josephson effects related to broken internal SO(3) × U(1). The U(1) part follows the pattern of the symmetries, the energy depends periodically on displace- ordinary superfluid above, so we focus on the SO(3) sub- ment w between the two crystals (on top of a possible space. The order parameter is a fixed orthogonal 3 × 3 wk-dependence of the coupling K). Then there is an as- matrix Aµj, where µ transforms under S and j under L, 2π sociated force Fa = −∂w U = K sin ∆a (neglecting denoting the rotation from the position to the spin coor- a ca ∂ K). This Josephson force is purely due to the sym- dinate frame [20]. The three broken generators La − Sa wa a metry breaking and adds to or competes with for instance are represented by the matrices T ij = iaij acting on r Casimir/Van der Waals forces. A. Suppose that m = m A and m = m eiαT A. The L L R R Outlook. We have shown that the Josephson effect is DC Josephson current Eq. (7) is: intrinsic to the phenomenon of spontaneous symmetry breaking. Other generalized phenomena similarly ap- r K T r r
K I = mLmRTr A T sin(αT )A = mLmR2 sin α, pear, such as the Shapiro effect when the applied current ~ ~ is periodic. This paves the way to explore hitherto un- where we used the cyclic property of the matrix trace Tr addressed Josephson currents. It should be kept in mind and AAT = 1. The current Ia for a 6= r vanishes. that it is in general necessary to apply an external current Crystalline solid. The most interesting prediction in order to detect the Josephson current, and this may that follows from Eq. (6) is a Josephson effect for d- not be straightforward in some settings. Atomic conden- dimensional solids. It is necessary to carefully deter- sates may form a source of inspiration [22]. For the effect mine the order parameter capturing the translational in crystalline solids, measurement of the force F = −∇U symmetry breaking Rd → Zd to discrete lattice trans- looks most promising. Its magnitude is difficult to esti- lations. A medium is described by a field φi(x) de- mate in full generality. The tunneling of phonons should noting the comoving coordinates of the the volume el- fall off exponentially with wk, with a length scale ξ that ement at x. The action is invariant under internal shifts is the healing length for crystalline order. This should be a φi(x) → φi(x)+α δia, for which the symmetry generator related to the core size of dislocations [23]. 5

Acknowledgments. I thank Jasper van Wezel and Louk MEXT-Supported Program for the Strategic Research Rademaker for collaboration on a related project that Foundation at Private Universities “Topological Science” inspired this work and for many useful discussions, and (Grant No. S1511006) and by JSPS Grant-in-Aid for Naoki Yamamoto and Antonino Flachi for careful read- Early-Career Scientists (Grant No. 18K13502). ing of the manuscript. This work is supported by the

[1] B. D. Josephson, Possible new effects in superconductive romagnetic d-wave superconductors, Phys. Rev. B 81, tunnelling, Phys. Lett 1, 251 (1962). 115102 (2010). [2] P. W. Anderson, Special effects in superconductivity, in [13] A. Moor, A. F. Volkov, and K. B. Efetov, Josephson-like Lectures on the Many-Body Problem, Vol. 2, edited by spin current in junctions composed of antiferromagnets E. R. Caianello (Academic, New York, 1964) pp. 113– and ferromagnets, Phys. Rev. B 85, 014523 (2012). 135. [14] Y. Liu, G. Yin, J. Zang, R. K. Lake, and Y. Barlas, Spin- [3] B. D. Josephson, The discovery of tunnelling supercur- Josephson effects in exchange coupled antiferromagnetic rents, Rev. Mod. Phys. 46, 251 (1974). insulators, Phys. Rev. B 94, 094434 (2016). [4] P. W. Anderson, Considerations on the flow of superfluid [15] F. P. Esposito, L.-P. Guay, R. B. MacKenzie, M. B. helium, Rev. Mod. Phys. 38, 298 (1966). Paranjape, and L. C. R. Wijewardhana, Field theoretic [5] V. Ambegaokar, P. G. de Gennes, and D. Rainer, description of the Abelian and non-Abelian Josephson Landau-Ginsburg equations for an anisotropic superfluid, effect, Phys. Rev. Lett. 98, 241602 (2007). Phys. Rev. A 9, 2676 (1974). [16] R. Qi, X.-L. Yu, Z. B. Li, and W. M. Liu, Non-Abelian [6] S.-C. Zhang, A unified theory based on SO(5) symme- Josephson effect between two F = 2 spinor Bose-Einstein try of superconductivity and antiferromagnetism, Science condensates in double optical traps, Phys. Rev. Lett. 275, 1089 (1997). 102, 185301 (2009). [7] E. Demler, A. J. Berlinsky, C. Kallin, G. B. Arnold, and [17] B. Gaury, J. Weston, and X. Waintal, The a.c. Josephson M. R. Beasley, Proximity effect and Josephson coupling effect without superconductivity, Nat. Comm. 6, 6524 in the SO(5) theory of high-Tc superconductivity, Phys. (2015). Rev. Lett. 80, 2917 (1998). [18] P. W. Anderson, An approximate quantum theory of [8] F. S. Nogueira and K.-H. Bennemann, Spin Josephson the antiferromagnetic ground state, Phys. Rev. 86, 694 effect in ferromagnet/ferromagnet tunnel junctions, EPL (1952). 67, 620 (2004). [19] R. P. Feynman, R. B. Leighton, and M. Sands, The Feyn- [9] Y.-L. Lee and Y.-W. Lee, Quantum dynamics of tun- man Lectures on Physics, Vol. 3 (Addison-Wesley, 1965). neling between ferromagnets, Phys. Rev. B 68, 184413 [20] D. Vollhardt and P. W¨olfle, The Superfluid Phases of (2003). Helium 3 (Taylor & Francis, 1990). [10] W. Chen, P. Horsch, and D. Manske, Dissipationless spin [21] A. Nicolis, R. Penco, F. Piazza, and R. Rattazzi, Zool- current between two coupled ferromagnets, Phys. Rev. B ogy of condensed matter: Framids, ordinary stuff, extra- 89, 064427 (2014). ordinary stuff, JHEP 06, 155. [11] A. R¨uckriegel and P. Kopietz, Spin currents, spin torques, [22] K. Sukhatme, Y. Mukharsky, T. Chui, and D. Pearson, and the concept of spin superfluidity, Phys. Rev. B 95, Observation of the ideal Josephson effect in superfluid 104436 (2017). 4He, Nature 411, 280 (2001). [12] D. Chass´eand A.-M. S. Tremblay, Generalized dc and [23] R. Peierls, The size of a dislocation, Proc. Phys. Soc. 52, ac Josephson effects in antiferromagnets and in antifer- 34 (1940).