Proc. Natl. Acad. Sci. USA Vol. 94, pp. 6013–6018, June 1997

Collective electrodynamics I (superconductor͞quantum coherence͞vector potential͞magnetic interaction)

CARVER A. MEAD

California Institute of Technology, MS 136-93, Pasadena, CA 91125

Contributed by Carver A. Mead, April 2, 1997

ABSTRACT Standard results of electromagnetic theory B, that has a natural connection with the nature of are derived from the direct interaction of macroscopic quan- matter—as highlighted by Aharonov and Bohm (3). tum systems; the only assumptions used are the Einstein– Hamilton’s formulation of classical was—and deBroglie relations, the discrete nature of charge, the Green’s remains—the starting point for the concepts underlying the function for the vector potential, and the continuity of the quantum theory. The would have function. No reference is needed to Maxwell’s equations or to every quantum system approach the behavior of its classical- traditional quantum formalism. Correspondence limits based mechanics counterpart in the limit where the mechanical on are shown to be inappropriate. involved is large compared with Planck’s constant. Although superconductivity was discovered in 1911, the recognition that superconductors manifest quantum phe- But the real glory of is that we can find a way of nomena on a (4) came too late to play a thinking such that the law is evident. – R. P. Feynman role in the formulation of . Through modern experimental methods, however, superconducting Foundations of Physics structures give us direct access to the quantum nature of matter. The superconducting state is a coherent state formed Much has transpired since the first two decades of this century, by the collective interaction of a large fraction of the free when the conceptual foundations for were put in a material. Its properties are dominated by in place. At that , macroscopic mechanical systems were known and controllable interactions within the collective easily accessible and well understood. The nature of electrical ensemble. The dominant interaction is collective because the phenomena was mysterious; were difficult and properties of each depend on the state of the entire their interpretation was murky. Today, quite the reverse is ensemble, and it is electromagnetic because it couples to the true. Electrical experiments of breathtaking clarity can be charges of the electrons. Nowhere in natural phenomena do carried out, even in modestly equipped laboratories. Electronic the basic laws of physics manifest themselves with more apparatus pervade virtually every abode and workplace. Mod- crystalline clarity. ern mechanical experiments rely heavily on electronic instru- This paper is the first in a series in which we start at the mentation. Yet, in spite of this reversal in the range of simplest possible conceptual level, and derive as many experience accessible to the average person, introductory conclusions as possible before moving to the next level of treatments of physics still use classical mechanics as a starting detail. In most cases, understanding the higher level will point. allow us to see why the assumptions of the level below were Ernst Mach wrote (p. 596 in ref. 1), ‘‘The view that makes valid. In this stepwise fashion, we build up an increasingly mechanics the basis of the remaining branches of physics, and comprehensive understanding of the subject, always keeping in view the assumptions required for any given result. We explains all physical phenomena by mechanical ideas, is in our avoid introducing concepts that we must ‘‘unlearn’’ as we judgment a prejudice. . . . The mechanical theory of nature, is, progress. We use as our starting point the magnetic inter- undoubtedly, in a historical view, both intelligible and pardon- action of macroscopic quantum systems through the vector able; and it may also, for a time, have been of much value. But, and scalar potentials Aជ and V, which are the true observable upon the whole, it is an artificial conception.’’ quantities. For clarity, the brief discussion given here is Classical mechanics is indeed inappropriate as a starting limited to situations where the currents and voltages vary point for physics because it is not fundamental; rather, it is the slowly; the four-vector generalization of these relations not limit of an incoherent aggregation of an enormous number of only removes this quasi-static limitation, but gives us elec- quantum elements. To make contact with the fundamental trostatics as well (5, 6). nature of matter, we must in a coherent context where the quantum reality is preserved. Model System R. P. Feynman wrote (p. 15–8 in ref. 2), ‘‘There are many changes in concepts that are important when we go from Our model system is a loop of superconducting wire—the two classical to quantum mechanics. . . . Instead of , we deal ends of the loop being colocated in and either insulated with the way interactions change the of .’’ or shorted, depending on the experimental situation. Experi- Even Maxwell’s equations have their roots in classical me- mentally, the voltage V between the two ends of the loop is chanics. They were conceived as a theory of the ether: They related to the current I flowing through the loop by express relations between the B and the E, which are defined in terms of the classical F ϭ LI ϭ Vdtϭ⌽. [1] q(E ϩ v ϫ B) on a particle of charge q moving with v. ͵ But it is the vector potential A, rather than the magnetic field Two quantities are defined by this relationship: ⌽, called the The publication costs of this article were defrayed in part by page charge *, and L, called the inductance, which depends payment. This article must therefore be hereby marked ‘‘advertisement’’ in on the of the loop. accordance with 18 U.S.C. §1734 solely to indicate this fact. © 1997 by The National Academy of 0027-8424͞97͞946013-6$2.00͞0 *This definition is independent of the shape of the loop, and applies

6013 Downloaded by guest on October 2, 2021 6014 Physics: Mead Proc. Natl. Acad. Sci. USA 94 (1997)

Current is the flow of charge: I ϭ dQ͞dt. Each increment of intuition and common sense, gives us ‘‘a way of thinking such charge dQ carries an increment dW ϭ VdQinto the that the law is evident.’’ Electrons in a superconductor are loop as it enters.† The total energy W stored in the loop is thus described by a that has an amplitude and a phase. The earliest treatment of the wave nature of matter was the 1923 wave mechanics of deBroglie. He applied the 1905 W ϭ ͵VdQϭ͵VI dt Einstein postulate (W ϭប␻) to the energy W of an electron wave, and identified the pជ of an electron with the propagation vector of the wave: p ϭបkជ. Planck’s constant h dI 1 ជ ϭ L IdtϭL IdI ϭ LI2. [2] and its radian equivalent បϭh͞2␲are necessary for merely ͵dt ͵ 2 historical reasons—when our standard units were defined, it was not known that energy and were the same If we reduce the voltage to zero by, for example, connecting the quantity. two ends of the loop to form a closed superconducting path, The Einstein–deBroglie relations apply to the collective the current I will continue to flow indefinitely: a persistent electrons in a superconductor. The of the system can current. If we open the loop and allow it to do work on an be derived from the dispersion relation (10) between ␻ and kជ. external circuit, we can recover all the energy W. Both ␻ and kជ are properties of the phase of the wave function If we examine closely the values of currents under a variety and do not involve the amplitude, which, in collective systems, of conditions, we find the full continuum of values for the is usually determined by some normalization condition. In a quantities I, V, and ⌽, except for persistent currents, where superconductor, the constraint of charge neutrality is such a only certain discrete values occur for any given loop (7, 8). By condition. experimenting with loops of different dimensions, we find the The wave function must be continuous in space; at any given condition that describes the values that occur experimentally: time, we can follow the phase along a path from one end of the loop to the other: the number of radians by which the phase

⌽ ϭ Vdtϭn⌽0. [3] advances as we traverse the path is the phase accumulation ␸ ͵ around the loop. If the phase at one end of the loop changes relative to that at the other end, that change must be reflected Ϫ15 Here, n is any integer, and ⌽0 ϭ 2.06783461 ϫ 10 volt- in the total phase accumulation around the loop. The fre- second is called the flux quantum or fluxoid; its value is quency ␻ of the wave function at any point in space is the rate 9 accurate to a few parts in 10 , independent of the detailed size, at which the phase advances per unit time. If the frequency at shape, or composition of the superconductor forming the loop. one end of the loop (␻1) is the same as that at the other end We also find experimentally that a rather large energy— (␻2), the phase difference between the two ends will remain sufficient to disrupt the superconducting state entirely—is constant, and the phase accumulation will not change with required to change the value of n. time. If the frequency at one end of the loop is higher than that 3 The more we reflect on Eq. , the more remarkable the at the other, the phase accumulation will increase with time, result appears. The quantities involved are the voltage and the and that change must be reflected in the rate at which phase magnetic flux. These quantities are integrals of the quantities accumulates with the distance l along the path. The rate at E and B that appear in Maxwell’s equations, and are therefore which phase around the loop accumulates with time is the usually associated with the . Experimen- tally, we know that they can take on a continuum of values— difference in frequency between the two ends. The rate at except under special conditions, when the arrangement of which phase accumulates with distance l is the component of ជ 3 matter in the vicinity causes the flux to take on precisely the propagation vector k in the direction dl along the path. quantized values. In Maxwell’s theory, E and B represented the Thus, the total phase accumulated around the loop is state of strain in a mechanical medium (the ether) induced by . Einstein had a markedly different view (p. 383 3 ␸ ϭ ͑␻ Ϫ ␻ ͒dt ϭ kជ ⅐ dl. [4] in ref. 9): ‘‘I feel that it is a delusion to think of the electrons ͵ 1 2 Ͷ and the fields as two physically different, independent entities. Since neither can exist without the other, there is only one We can understand as an expression of the reality to be described, which happens to have two different single-valued nature of the phase of the wave function. When aspects; and the theory ought to recognize this from the start the two ends of the loop were connected to an external circuit, instead of doing things twice.’’ At the most fundamental level, the two phases could evolve independently. When the ends are the essence of quantum mechanics lies in the wave nature of connected to each other, however, the two phases must match matter. Einstein’s view would suggest that electromagnetic up. But the phase is a quantity that has a cyclic nature— variables are related to the wave properties of the electrons. matching up means being equal modulo 2␲. Thus, for a wave Quantization is a familiar phenomenon in systems where the that is confined to a closed loop, and has a single-valued, boundary conditions give rise to standing waves. The quanti- continuous phase, the integral of Eq. 4 must be n2␲, where n zation of flux (Eq. 3) is a direct manifestation of the wave is an integer. The large energy required to change n is evidence nature of matter, expressed in electromagnetic variables. that the phase constraint is a strong one—as long as the Matter superconducting state stays intact, the wave function remains intact as well. To most nonspecialists, quantum mechanics is a baffling These relations tell us that the magnetic flux and the mixture of waves, , and arbitrary rules, ossified in a propagation vector will be quantized for a given loop; they do matrix of impenetrable formalism. By using a superconductor, not tell us how the frequency ␻ in Eq. 4 is related to the we can avoid the statistics, the rules, and the formalism, and potential V in Eq. 1. To make this connection, we must work directly with the waves. The wave concept, accessible to introduce one additional assumption: The collective electron system represented by the wave function is made up of even to coils with multiple turns. For multiturn coils, what we call the elemental charges of magnitude q0. By the Einstein relation, flux is commonly referred to as the total flux linkage. the energy q0V of an elemental charge corresponds to a † We use this relation to define the voltage V. frequency ␻ ϭ q0V͞ប. Downloaded by guest on October 2, 2021 Physics: Mead Proc. Natl. Acad. Sci. USA 94 (1997) 6015

Electrodynamics (ii) The voltage observed between the two ends of the second loop under open conditions is almost exactly equal to Electrodynamics is the interaction of matter via the electro- that observed across the first loop. magnetic field. We can formulate our first relation between the (iii) When the second loop is shorted, the voltage observed electromagnetic quantities V and ⌽ and the phase accumula- across the first loop is nearly zero, independent of the current. tion ␸ of the wave function by comparing Eq. 1 with Eq. 4: (iv) The current observed in the second loop under shorted conditions is nearly equal to that flowing in the first loop, but q0 q0 is of the opposite sign. ␸ ϭ ␻ dt ϭ Vdtϭ n⌽ ϭn͑2␲͒. [5] ͵ ប ͵ ប 0 Similar measurements performed when the loops are sep- arated allow us to observe how the coupling between the loops depends on their separation and relative orientation. From Eq. 5, we conclude that ⌽ ϭ h͞q . We understand that 0 0 (v) For a given configuration, the voltage observed across the potential V and the frequency ␻ refer to differences in the second loop remains proportional to the voltage across the these quantities between the two ends of the loop. Equiva- first loop. The constant of proportionality, which is nearly lently, we measure each of these quantities at one end of the loop using as a reference the value at the other end of the loop. unity when the loops are superimposed, decreases with the distance between the loops. When we substitute into Eq. 5 the measured value of ⌽0 and the known value of h, we obtain for q a value that is exactly (vi) The constant of proportionality decreases as the axes of 0 the two loops are inclined with respect to each other, goes to twice the charge qe of the free electron. The usual explanation for this somewhat surprising result is that each state in the zero when the two loops are orthogonal, and reverses when one superconductor is occupied by a pair of electrons, rather than loop is flipped with respect to the other. Observation i tells us that the presence of electrons in the by an individual electron, so the elemental charge q0 should be 2q , rather than q . None of the conclusions that we shall reach second loop does not per se affect the operation of the first e e loop. The voltage across a loop is a direct manifestation of the depends on the value of q0. We have established the correspondence between the po- phase accumulation around the loop. Observation ii tells us tential V and the frequency ␻—the time integral of each of that current in a neighboring loop is as effective in producing these equivalent quantities in a closed loop is quantized. The phase accumulation in the wave function as is current in the line integral of the propagation vector kជ around a closed loop same loop. The ability of current in one location to produce also is quantized. We would therefore suspect the existence of phase accumulation in the wave function of electrons in a corresponding electromagnetic quantity, whose line integral another location is called magnetic interaction. Observation vi is the magnetic flux ⌽. That quantity is the well-known vector tells us that the magnetic interaction is vectorial in nature. potential Aជ. The general relations among these quantities, After making these and other similar measurements on many whether or not the loop is closed, are configurations, involving loops of different sizes and shapes, we arrive at the proper generalization of Eqs. 1 and 6: 3 Phase ␸ ϭ ␻ dt ϭ kជ ⅐ dl 3 ͵ Ͷ ប V1 dt ϭ Aជ ⅐ dl1 ϭ ⌽1 ϭ L1I1 ϩ MI2 ⌽ ϭ ␸. [6] ͵ Ͷ 3 q0 Flux ⌽ ϭ Vdtϭ Aជ⅐dl · ͵ Ͷ 3 ជ ͵V2 dt ϭ ͶA ⅐ dl2 ϭ ⌽2 ϭ MI1 ϩ L2I2. [7] Eq. 6 expresses the first set of fundamental relations of collective electrodynamics. 3 3 Here, the line elements dl1 and dl2 are taken along the first and second loops, respectively. The quantity M, which by obser- Coupling vation vi can be positive or negative depending on the config- uration, is called the mutual inductance; it is a measure of how Up to this point, we have tentatively identified the phase effective the current in one loop is at causing phase accumu- accumulation and the magnetic flux as two representations of lation in the other. When L1 ϭ L2 ϭ L, the magnitude of M the same physical entity. We assume that ‘‘winding up’’ the can never exceed L. Observations i–iv were obtained under wave function with a voltage produces a propagation vector in conditions where M Ϸ L. Experiments evaluating the mutual the superconductor related to the of the electrons, and coupling of loops of different sizes, shapes, orientations, and that this motion corresponds to a current because the electrons spacings indicate that each element of wire of dl are charged. This viewpoint will allow us to understand the carrying the current Iជ makes a contribution to Aជ that is interaction between two coupled collective electron systems. proportional to Iជ, and to the inverse of the distance r from the We shall develop these relations in more detail when we study current element to the point at which Aជ is evaluated: the current distribution within the wire itself. Let us consider two identical loops of superconducting wire, ␮0 Iជ ␮0 Jជ the diameter of the wire being much smaller than the loop Aជ ϭ dl f Aជ ϭ dvol. [8] radius. We place an extremely thin insulator between the 4␲ ͵ r 4␲ ͵ r loops, which are superimposed on each other as closely as allowed by the insulator. In this configuration, both loops can The constant ␮0 is called the permeability of free space. The be described, to an excellent approximation, by the same path second form follows from the first if we visualize a distribution in space, despite their being electrically distinct. As we exper- of current as carried by a large number of wires of infinitesimal iment with this configuration, we make the following obser- cross section, and the current Jជas being the number of vations. such wires per unit area normal to the current flow. The 1͞r (i) When the two ends of the second loop are left open, its form of the integrand of Eq. 8 is called the Green’s function; presence has no effect on the operation of the first loop. The it tells us how the vector potential is generated by currents relationship between a current flowing in the first loop and the everywhere in space. It is perhaps more correct to say that the voltage observed between the ends of the first loop follows Eq. vector potential is a bookkeeping device for evaluating the 1, with exactly the same value of L as that observed when the effect at a particular point of all currents everywhere in space. second loop was absent. Ernst Mach wrote (p. 317 in ref. 1), ‘‘We cannot regard it as Downloaded by guest on October 2, 2021 6016 Physics: Mead Proc. Natl. Acad. Sci. USA 94 (1997)

impossible that integral laws . . . will some day take the place length of wire that take part in the motion, integrating Eq. 11 of the . . . differential laws that now make up the science of along the loop gives mechanics. . . . In such an event, the concept of force will have become superfluous.’’ Eqs. 6 and 8 and are the fundamen- ជ 3 tal integral laws for collective electromagnetic interaction. The P ϭ ␩ q0 Ͷ A ⅐ dl ϭ ␩ q0 ⌽ ϭ ␩ q0 LI. [13] 2 .(equivalent differential equation is ٌ Aជ ϭϪ␮0Jជ(5, 6 We can express Eq. 2 in a way that gives us additional insight The current I is carried by the ␩ charges per unit length moving into the energy stored in the coil: at velocity v; therefore, I ϭ ␩ q0v, and Eq. 13 becomes P ϭ L͑␩ q ͒2 v. [14] W ϭ ͵VdQϭ͵VI dt ϭ ͵Id⌽. [9] 0 The momentum is proportional to the velocity, as it should be. It is also proportional to the size of the loop, as reflected by Eq. 9 is valid for any Aជ; it is not limited to the Aជ from the the inductance L. Here we have our second big surprise: current in the coil itself. The integrals in Eq. 9 involve the instead of scaling linearly with the number of charges that take entire coil. From them we can take a conceptual step and, using part in the motion, the momentum of a collective system scales our visualization of the current density, imagine an energy as the square of the number of charges! We can understand this density Jជ ⅐ Aជ ascribed to every point in space: collective behavior as follows. In an arrangement where charges are constrained to move in concert, each charge W ϭ Iជ⅐ Adlជ ϭ Jជ⅐Adជ vol. [10] produces phase accumulation, not only for itself, but for all the ͵ ͵ other charges as well. So the of each charge increases linearly with the number of charges moving in concert. The Electrodynamic Momentum inertia of the ensemble of coupled charges must therefore increase as the square of the number of charges. Feynman commented on the irrelevance of the concept of force in a quantum context. At the fundamental level, we can Forces on Currents understand the behavior of a quantum system using only the wave properties of matter. But we experience forces between In our experiments on coupled loops, we have already seen currents in every encounter with electric motors, relays, and how the current in one loop induces phase accumulation in other electromagnetic actuators. How do these forces arise another loop; the relations involved were captured in Eq. 7.In from the underlying quantum reality? We can make a con- any situation where we change the coupling of collective nection between the classical concept of force and the quan- systems by changing the spatial arrangement, mechanical work tum nature of matter through the concept of momentum. may be involved. Our model system for studying this interac- Using the deBroglie postulate relating the momentum pជ of an tion consists of two identical shorted loops of individual electron to the propagation vector kជ of the wave function, and inductance L0, each carrying a persistent flux ⌽. As long as the identifying the two integrands in Eq. 6, the electrodynamic superconducting state retains its integrity, the cyclic constraint momentum of an elemental charge is on the wave function guarantees that the flux ⌽ in each loop will be constant, independent of the coupling between loops. pជ ϭ បkជ ϭ q0 Aជ. [11] Because M enters symmetrically in Eq. 7, the current I will be the same in both loops. Hence, L0 and ⌽ will remain constant, We shall now investigate the electrodynamic momentum in whereas M and I will be functions of the spatial arrangement one of our loops of superconducting wire. There is an electric of the loops—M will be large and positive when the loops are field E along the loop, the line integral of which is the voltage brought together with their currents flowing in the same V between the ends. From a classical point of view, Newton’s direction, and will be large and negative when the loops are law tells us that the force q0 E on a charge should be equal to brought together with their currents flowing in opposite the time rate of change of momentum. From Eq. 11, directions. From Eq. 7, ⌽ϭ(L0ϩM)I. Substituting ⌽ into Eq. 9, and noting that the total energy of the system is twice that Ѩpជ ѨAជ 3 Ѩ⌽ for a single coil, q Eជ ϭ ϭ q f V ϭ Eជ ⅐ dl ϭ . [12] 0 Ѩt 0 Ѩt Ͷ Ѩt ⌽2 2 W ϭ 2 Id⌽ϭ͑L0ϩM͒I ϭ . [15] Integrating the second form of Eq. 12 with respect to time, we ͵ ͑L0ϩM͒ recover Eq. 6, so the classical idea of inertia is indeed

consistent with the quantum behavior of our collective system. The force Fx along some direction x is defined as the rate of Electrodynamic inertia acts exactly as a classical mechanical change of energy with a change in the corresponding coordi- inertia: it relates the integral of a force to a momentum, which nate: is manifest as a current. We note that, for any system of charges that is overall charge neutral, as is our superconductor, the net ѨW ⌽ 2ѨM F ϭ ϭϪ . [16] electromagnetic momentum is zero. For the ϪqAជ of each x Ѩx ͩL ϩMͪѨx 0 electron, we have a canceling ϩqAជ from one of the background positive charges. The electric field that accelerates electrons in The negative sign indicates an attractive force because the one direction exerts an equal force in the opposite direction on mutual inductance M increases as the coils—whose currents the background positive charges. We have, however, just are circulating in the same direction—are moved closer. It is encountered our first big surprise: We recognize the second well known that electric charges of the same sign repel each form of Eq. 12, which came from Newton’s law, as the integral other. We might expect the current, being the spatial analog of form of one of Maxwell’s equations! the charge, to behave in a similar manner. However, Eq. 15 We would expect the total momentum P of the collective indicates that the total energy of the system decreases as M electron system to be the momentum per charge the increases. How does this attractive interaction of currents number of charges in the loop. If there are ␩ charges per unit circulating in the same direction come about? Downloaded by guest on October 2, 2021 Physics: Mead Proc. Natl. Acad. Sci. USA 94 (1997) 6017

The electron velocity is proportional to I.AsMis increased, The velocity vជ ϭ (បkជ Ϫ q0 Aជ)͞m is thus a direct measure of the the electrons in both loops slow down because they have more imbalance between the total momentum បkជ and the electro- inertia due to the coupling with electrons in the other loop. dynamic momentum q0 Aជ. When these two quantities are This effect is evident in Eq. 15, where I ϭ⌽͞(L0ϩM). Thus, matched, the velocity is zero. The current density Jជ is just the there are two competing effects: the decrease in energy due to motion of ᏺ elementary charges per unit : Jជ ϭ q0 ᏺvជ. the lower velocity, and the increase in energy due to the We can thus express Eq. 18 in terms of the wave vector kជ, increase in inertia of each electron. The energy goes as the the vector potential Aជ, and the current density Jជ: square of the velocity, but goes only linearly with the inertia, q0 ᏺ so the velocity wins. The net effect is a decrease in energy as Jជϭ ͑បkជ Ϫ q Aជ ͒. [19] currents in the same direction are coupled, and hence an m 0 attractive force. We can see how the classical force law discovered in 1823 by Ampe`re arises naturally from the Current Distribution collective quantum behavior, which determines not only the magnitude, but also the sign, of the effect. We are now in a position to investigate how current distributes itself inside a superconductor. If Aជ were constant through- Multiturn Coils out the wire, the motion of the electrons would be determined by the common wave vector kជ of the collective electron system, The interaction in a collective system scales as the square of and we would expect the persistent current for a given flux to the number of electrons moving in concert. Thus, we might be proportional to the cross-sectional area of the wire, and thus expect the quantum scaling laws to be most clearly manifest the inductance L of a loop of wire to be inversely related to the in the properties of closely coupled multiturn coils, where the wire cross section. When we perform experiments on loops of number of electrons is proportional to the number of turns. wire that have identical paths in space, however, we find that We can construct an N-turn coil by connecting in series N the inductance is only a weak function of the wire diameter, identical, closely coupled loops. In this arrangement, the indicating that the current is not uniform across the wire, and current through all loops is equal to the current I through the therefore that Aជ is far from constant. If we make a loop of coil, and the voltage V across the coil is equal to the sum of superconducting tubing, instead of wire, we find that it has the individual voltages across the loops. If A0 is the vector exactly the same inductance as does a loop made with wire of potential from the current in one loop, we expect the vector the same diameter, indicating that current is flowing at the potential from N loops to be NA0, because the current in surface of the loop, but is not flowing throughout the bulk. each loop contributes. The flux integral is taken around N Before taking on the distribution of current in a wire, we can turns, so the path is N times the length l0 of a single turn. The examine a simpler example. In a simply connected bulk total flux integral is thus superconductor, the single-valued nature of the wave function can be satisfied only if the phase is everywhere the same: kជ ϭ Nl0 0. Any phase accumulation induced through the Aជ vector 2 ⌽ ϭ ͵Vdtϭ͵ NA0 ⅐ dl ϭ N L0 I. [17] created by an external current will be canceled by a screening 0 current density Jជ in the opposite direction, as we saw in observations iii and iv. To make the problem tractable, we From Eq. 17 we conclude that an N-turn closely coupled coil consider a situation where a vector potential A0 at the surface 2 has an inductance L ϭ N L0. Once again, we see the of a bulk superconducting slab is created by distant currents collective interaction scaling as the square of the number of parallel to the surface of the slab. The current distribution interacting charges. We remarked that collective quantum perpendicular to the surface is a highly localized phenomenon, systems have a correspondence limit markedly different from so it is most convenient to use the differential formulation of that of classical mechanical systems. When two classical Eq. 8. We suppose that conditions are the same at all points massive bodies, each body having a separate inertia, are on the surface, and therefore that A changes in only the x bolted together, the inertia of the resulting composite body direction, perpendicular to the surface, implying that ٌ2A ϭ is simply the sum of the two individual . The inertia Ѩ2A͞Ѩx2. of a collective system, however, is a manifestation of the 2 2 interaction, and cannot be assigned to the elements sepa- Ѩ A ␮0 q0 ᏺ [2A ϭ ϭ Ϫ␮ Jជϭ A. [20ٌ rately. This difference between classical and quantum sys- Ѩx2 0 m tems has nothing to do with the size scale of the system. Eq. 17 is valid for large as well as for small systems; it is valid The solution to Eq. 20 is where the total phase accumulation is an arbitrary number m of cycles—where the granularity of the flux due to ប is as Ϫx͞␭ 2 A ϭ A0 e ␭ ϭ 2 . [21] small as might be required by any correspondence procedure. ␮0 q0 ᏺ Thus, it is clear that collective quantum systems do not have a classical correspondence limit. The particular form of Eq. 21 depends on the , but the qualitative result is always the same, and can be understood Total Momentum as follows: The current is the imbalance between the wave vector and the vector potential. When an imbalance exists, a To see why our simplistic approach has taken us so far, we must current proportional to that imbalance will flow such that it understand the current distribution within the superconductor cancels out the imbalance. The resulting screening current dies itself. We saw that the vector potential made a contribution to out exponentially with distance from the source of imbalance. the momentum of each electron, which we called the electro- The distance scale at which the decay occurs is given by ␭, the dynamic momentum: pជel ϭ qAជ. The m of an electron screening distance, penetration depth, or skin depth. For a moving with velocity vជ also contributes to the electron’s typical superconductor, ᏺ is of the order of 1028͞m3,so␭ momentum: pជmv ϭ mvជ. The total momentum is the sum of should be a few tens of nanometers. Experimentally, simple these two contributions: superconductors have ␭ Ϸ 50 nanometers—many orders of magnitude smaller than the macroscopic wire thickness that បkជ ϭ pជ ϭ pជ el ϩ pជmv ϭ q0 Aជ ϩ mvជ. [18] we are using. Downloaded by guest on October 2, 2021 6018 Physics: Mead Proc. Natl. Acad. Sci. USA 94 (1997)

Current in a Wire Conclusion

At long last, we can visualize the current distribution within the We took to heart Einstein’s belief that the electrons and the superconducting wire itself. Because the skin depth is so small, fields were two aspects of the same reality, and were able to the surface of the wire appears flat on that scale, and we can treat the macroscopic quantum system and the electromag- use the solution for a flat surface. The current will be a netic field as elements of a unified subject. We heeded Mach’s maximum at the surface of the wire, and will die off expo- advice that classical mechanics was not the place to start, nentially with distance into the interior of the wire. We can followed Feynman’s directive that interactions change the appreciate the relations involved by examining a simple ex- wavelengths of waves, and saw that there is a correspondence ample. A 10-cm-diameter loop of 0.1-mm-diameter wire has an limit more appropriate than the classical-mechanics version inductance of 4.4 ϫ 10Ϫ7 Henry (p. 193 in ref. 11): A persistent used in traditional introductions to quantum theory. We found current of 1 Ampere in this loop produces a flux of 4.4 ϫ 10Ϫ7 Newton’s law masquerading as one of Maxwell’s equations. We volt-second, which is 2.1 ϫ 108 flux quanta. The electron wave were able to derive a number of important results using only function thus has a total phase accumulation of 2.1 ϫ 108 cycles the simplest properties of waves, the Einstein postulate relating along the length of the wire, corresponding to a wave vector frequency to energy, the deBroglie postulate relating momen- k ϭ 4.25 ϫ 109 mϪ1. Due to the cyclic constraint on the wave tum to wave vector, and the discrete charge of the electron. It function, this phase accumulation is shared by all electrons in thus appears possible to formulate a unified, conceptually the wire, whether or not they are carrying current. correct introduction to both the quantum nature of matter and In the region where current is flowing, the moving mass of the fundamental laws of electromagnetic interaction without the electrons contributes to the total phase accumulation. The using either Maxwell’s equations or standard quantum formal- 1-Ampere of current results from a current density of 6.4 ϫ ism. 1010 Amperes per square meter flowing in a thin ‘‘skin’’ Ϸ␭ thick, just inside the surface. This current density is the result I am indebted to Richard F. Lyon, Sanjoy Mahajan, William B. of the 1028 electrons per cubic meter moving with a velocity of Bridges, Rahul Sarpeshkar, Richard Neville, and Lyn Dupre´for v Ϸ 20 meters per second. The mass of the electron moving at helpful discussion and critique of the material, and to Calvin Jackson this velocity contributes mv͞បϭ1.7 ϫ 105 mϪ1 to the total for his help in preparing the manuscript. The work was supported by wave vector of the wave function, which is less than one part the Arnold and Mabel Beckman Foundation, and by Gordon and Betty in 104 of that contributed by the vector potential. That small Moore. difference, existing in about 1 part in 106 of the cross-sectional area, is enough to bring kជ and Aជ into balance in the interior of 1. Mach, E. (1960) The Science of Mechanics (Open Court, La Salle, the wire. IL). 2. Feynman, R. P., Leighton, R. B. & Sands, M. (1964) The Feyn- In the interior of the wire, the propagation vector of the man Lectures on Physics (Addison–Wesley, Reading, MA), Vol. wave function is matched to the vector potential, and the 2. current is therefore zero. As we approach the surface, A 3. Aharonov, Y. & Bohm, D. (1959) Phys. Rev. 115, 485–491. decreases slightly, and the difference between k and Aq0͞ប is 4. London, F. (1950) Superfluids (Wiley, New York). manifest as a current. At the surface, the value and radial slope 5. Sommerfeld, A. (1952) Electrodynamics (Academic, New York). of A inside and outside the wire match, and the value of A is 6. Morse, P. M. & Feshbach, H. (1953) Methods of Theoretical still within one part in 104 of that in the center of the wire. So Physics I (McGraw–Hill, New York). our simplistic view—that the vector potential and the wave 7. Deaver, B. S. & Fairbank, W. M. (1961) Phys. Rev. Lett. 7, 43–46. vector were two representations of the same quantity—is 8. Doll, R. & Nabauer, M. (1961) Phys. Rev. Lett. 7, 51–52. precisely true in the center of the wire, and is nearly true even 9. Jaynes, E. T. (1990) in Complexity, , and the Physics of Information, ed. Zurek, W. H. (Addison–Wesley, Reading, MA), at the surface. The current ជI is not the propagation vector kជ of ជ ជ pp. 381–403. the wave, but, for a fixed configuration, I is proportional to k 10. Feynman, R. P. & Hibbs, A. R. (1965) Quantum Mechanics and by Eqs. 8 and 19. For that reason, we were able to deduce the Path Integrals (McGraw–Hill, New York). electromagnetic laws relating current and voltage from the 11. Ramo, S., Whinnery, J. R. & van Duzer, T. (1994) Fields and quantum relations between wave vector and frequency. Waves in Communication Electronics (Wiley, New York). Downloaded by guest on October 2, 2021