Electron–Lattice Interaction and Cooper Pairs — Retarded Attraction, the Cooper Problem, the BCS–BEC Crossover, and the “Superconducting Bubble”

— A sequel to Lecture 7 (slides 33–34) of Computational Materials Science, “Superconductivity means that fermions behave like bosons”: explanation, derivations, and references for the companion simulator (cooper-pair-simulator-en.html)

In Lecture 7 we explained that in the superconducting state the electrons, which are fermions, pair up two by two (Cooper pairs) and behave like bosons with total spin angular momentum 0, and so condense. The upper half of slide 34 shows the two mechanisms that form pairs—pairing by lattice vibrations (the BCS mechanism, common in alloy systems, with the mechanism fully understood; reaching high temperatures requires ultrahigh pressure) and pairing by antiferromagnetic spin fluctuations (cuprates and others, high temperature at ordinary pressure, but the mechanism is not yet understood)—while the lower half depicts the change of the “spatial region of pair formation,” popularly called the superconducting bubble, as the electron density is moved from the high side to the low side (in the BCS limit the bubbles overlap heavily; in the BEC limit they become small, isolated pairs), and the ease with which BCS pairs are broken by thermal fluctuations (the reason room-temperature superconductivity is difficult).

In this explanation we describe, in order, (i) how a retarded attraction mediated by the lattice acts despite the Coulomb repulsion between the electrons (the difference between heavy and light elements, the isotope effect), (ii) the Cooper problem and the pair size, (iii) the wave function of the BCS ground-state pair seen in real space, and (iv) the BCS–BEC crossover, which connects the BCS limit to the BEC limit in a single theory, together with the quantitative meaning of the “bubble” picture that can be read from it.

Roadmap of this explanation §1 Key points of slides 33–34. §2 The retarded attraction created by the lattice: screening, the slow response of the ions, the Bardeen–Pines effective interaction, the isotope effect, and why the pair is (k↑, −k↓). §3 The Cooper problem: even an arbitrarily weak attraction binds. §4 The BCS ground state and the real-space pair wave function. §5 The BCS–BEC crossover and the “superconducting bubble”: pair size, T* and Tc, thermal fluctuations, and the correspondence with cuprates, cold atoms, and hydrides. §6 Implementation notes. §7 References.
  1. Key points of slides 33–34
  2. The retarded attraction created by the lattice
  3. The Cooper problem
  4. The BCS ground state and the real-space pair wave function
  5. The BCS–BEC crossover and the “superconducting bubble”
  6. Implementation notes
  7. References

1. Key Points of Slides 33–34 — Fermions Behave Like Bosons

In what follows, we back up these four points with equations and numbers.

2. The Retarded Attraction Created by the Lattice — How to Outwit the Coulomb Repulsion

2.1 The repulsion between electrons is screened at 0.1 nm

The Coulomb repulsion between two electrons in a metal is screened, because the surrounding conduction electrons move instantaneously. In the Thomas–Fermi approximation, we have

(1) V(r)=e24πε0re−kTFr, kTF2=3ne22ε0EF ⟹ 1kTF≈0.05–0.1 nm

so the range of the repulsion is no larger than the lattice constant. Therefore, as long as the two electrons are not in the same place at the same time, the repulsion is not a problem. The attraction must act in a way that satisfies this condition.

2.2 The ions are slow — a wake of positive charge remains after the electron has gone

A positive ion is 10⁴–10⁵ times heavier than an electron, and the period of lattice vibration is 2π/ωD ≈ 10⁻¹³ s. An electron, on the other hand, runs at the Fermi velocity vF ≈ 10⁶ m/s, so during a single oscillation of an ion it travels a distance

(2) ℓ=2πvFωD≈100 nm–1 µm ≫ 1kTF

When an electron passes nearby, the ions are attracted to it and start to move, but they bunch up most after the electron has left. The place where they have bunched up is rich in positive charge, so for about 10⁻¹³ s after electron 1 has passed, a “wake of positive charge” remains over several tens to several hundred nm behind electron 1. Electron 2, which arrives later, is drawn toward this wake. It is separated from electron 1 in both time and space, so the Coulomb repulsion does not reach it, and only the attraction transmitted through the “memory” of the lattice acts. This is the retarded attraction. Simulator ① depicts this process with a model in which ions connected by springs (damped harmonic oscillators) are driven by a softened Coulomb attraction:

(3) Mu¨i=−Mω02ui−Mγu˙i+∑e=1,2g(re−Ri)(|re−Ri|2+rc2)3/2

The density of positive charge is colored by the divergence of the displacement, −∇·u (positive where the ions bunch up), and the potential for electron 2 is drawn as the sum of the displacement-induced part of the Coulomb potential of the moved ions and the screened repulsion of Eq. (1) from electron 1. When the mass M is made heavier, ω₀ ∝ M−1/2 decreases and the wake extends far behind the electron. When it is made lighter, the ions follow the electron and return at once, and the wake is short. This is the meaning of the figure at the beginning of Simulator ① (the electron–lattice interaction for heavy and light elements). ▶ Simulator ①: The lattice wake and the retarded attraction

2.3 The Fröhlich and Bardeen–Pines effective interactions and the isotope effect

Written in quantum theory, this picture becomes a second-order perturbation process in which electron 1 emits a phonon (wave vector q, frequency ωq) and electron 2 absorbs it, and the effective interaction, including the screened Coulomb repulsion, is (Bardeen–Pines 1955; Fröhlich 1950)

(4) Veff(q,ω)=e2ε0(q2+kTF2)[1+ωq2ω2−ωq2] =e2ε0(q2+kTF2)ω2ω2−ωq2

ħω is the energy transfer between the two electrons (ξk − ξk′). For |ω| < ωq, Veff < 0, that is, the interaction is attractive: it is attractive only for slow processes, in which the energy transfer is smaller than the phonon energy. This is the momentum-space version of “retardation,” and BCS simplified it to “a constant attraction −V between electrons within ħωD of the Fermi surface.” If the ionic mass M is changed by using isotopes, ωD ∝ M−1/2, and since the BCS Tc = 1.13 θD e−1/N(0)V is proportional to θD, we have

(5) Tc∝M−α, α=12(BCS); measured: Hg 0.50, Pb 0.49, Sn 0.47, Zn 0.45

This isotope effect (Maxwell 1950; Reynolds et al. 1950) became decisive evidence that lattice vibrations are involved in superconductivity. The deviation of α from 1/2 in transition metals is due to the Coulomb pseudopotential μ* (McMillan). The heavier the element, the smaller ωD and the longer the wake persists, but the “window” ħωD in which the attraction acts becomes narrower. Even so, Pb and Hg have high Tc because a softer lattice has a larger electron–lattice coupling λ. Conversely, the lattice of light hydrogen has a very large ωD, which is why hydrides under ultrahigh pressure (H₃S 203 K, LaH₁₀ ≈ 250 K) reach a high Tc by the BCS mechanism. This is the meaning of “high-temperature superconductivity requires ultrahigh pressure” on the slide.

2.4 Why is the pair (k↑, −k↓)?

In the “chase” picture of ①, electron 2 runs in the same direction as electron 1, but an actual Cooper pair consists of momenta k and −k (center of mass at rest). The attraction of Eq. (4) acts between electrons in the shell of thickness ħωD around the Fermi surface, and for center-of-mass momentum K = 0 the number of combinations in which both k and −k lie in the shell is largest (the number of partners to scatter into is largest), so the binding is strongest (§3). The spin is the antisymmetric singlet (↑↓ − ↓↑)/√2, so that the spatial part can be symmetric (s-wave) and make maximal use of the attraction. Please understand that the picture of the lattice wake is the real-space version of a single scattering process, in which “electron 2 receives the lattice vibration (virtual phonon) created by electron 1,” and that the pair itself is defined in momentum space.

3. The Cooper Problem — Even an Arbitrarily Weak Attraction Binds

Cooper (1956) placed two electrons on top of a filled Fermi sea and assumed an attraction −V that acts only within ħωD of the Fermi surface. Writing the relative-motion wave function as Σkgkeik·(r₁−r₂) (k > kF), the Schrödinger equation becomes

(6) (E−2ξk)gk=−V∑k′gk′ ⟹ 1=N(0)V∫0ℏωDdξ2ξ−E ⟹ E≃−2ℏωDexp(−2N(0)V)

A bound state with E < 0 exists however small V is (in contrast to two free particles in three dimensions, for which the attraction has a threshold). The reason is that the Fermi sea blocks the states with k < kF, so the problem becomes effectively two-dimensional, with a constant density of states. The inverse of the momentum uncertainty δk ≈ |E|/ħvF that corresponds to the binding energy |E| is the pair size, and in terms of the BCS Δ it is

(7) ξ0=ℏvFπΔ≈102–103 nm, Nin=n·4π3ξ03=49π(kFξ0)3≈108–1013

Pippard’s coherence length ξ₀ is around 10³ times the electron spacing (0.2–0.3 nm) (from 3×10² times for Nb to 1.5×10⁴ times for Zn), and the 10²–10³ nm in Eq. (7) is an order-of-magnitude guide (about 80 nm for Nb, in the µm range for Al and Zn). Within the extent of a single pair there are 10⁸–10¹³ other electrons (about 10⁸ for Nb, 10¹²–10¹³ for Al and Zn). This is the quantitative meaning of the “overlapping bubbles” on slide 34.

4. The BCS Ground State and the Real-Space Pair Wave Function

In the BCS ground state |ψ⟩ = Πk(uk + vkc†k↑c†−k↓)|0⟩, the probability that (k↑, −k↓) is occupied as a pair is vk², and the pair amplitude (the amplitude for a pair of k to be “present”) is ukvk.

(8) vk2=12(1−ξkEk), φk≡ukvk=Δ2Ek, φ(r)=∑kφkeik·r∝∫0∞k2φksinkrkrdk

Projecting |ψ⟩ onto particle number N gives an antisymmetrized product 𝒜[g(1,2)g(3,4)⋯] in which all the electrons pair up with the same pair function g(r₁ − r₂) (Fourier components gk = vk/uk = Δ/(ξk + Ek)) (Schrieffer’s “condensation of the pair function”). Since gk is largest deep inside the Fermi sea (k → 0) and is not localized at the Fermi surface, to examine the extent of the pair we use the Fourier transform φ(r) (the anomalous average ⟨ψ↓(r)ψ↑(0)⟩) of the pair amplitude φk = ukvk, which is localized around the Fermi surface with width Δ. φ(r) is the amplitude for the relative position of the ↓ partner as seen from an ↑ electron, and on the BCS side only k near the Fermi surface contribute, so φ(r) oscillates with wavelength 2π/kF while extending out to ξ₀. The analytic form of Kadin (2007) is φ(r) ∝ (sin kFr/kFr)·K₀(r/πξ₀) (K₀ is the modified Bessel function), and simulator ③ evaluates the integral in Eq. (8) numerically and draws φ continuously from the BCS side to the BEC side (the molecular e−r/a/r). On the BCS side, where the bubbles overlap, one cannot point in real space to “which two are a pair,” and φ(r) is most accurately read as the amplitude of “how much more likely a ↓ electron is to be found near an ↑ electron than in the uncorrelated case” (pair correlation). ▶ Simulator ③: The pair wave function φ(r)

5. The BCS–BEC Crossover and the “Superconducting Bubble”

5.1 From the BCS limit to the BEC limit in a single theory

As the attraction is strengthened, the BCS state, in which pairs form only in a thin shell around the Fermi surface, changes continuously into a Bose–Einstein condensate of molecules bound in real space (of size a) (Eagles 1969; Leggett 1980; Nozières–Schmitt-Rink 1985). Characterizing the interaction by the scattering length a and using the dimensionless parameter 1/(kFa) (−∞ for weak coupling, +∞ for strong coupling), the T = 0 mean-field (BCS) theory can be written as a pair of simultaneous equations, the gap equation and the number equation (the unit of energy is EF and the unit of wave number is kF; x = k/kF):

(9) 1kFa=2π∫0∞dx[1−x2E], 23=∫0∞dxx2[1−x2−μ~E], E=(x2−μ~)2+Δ~2

The first equation is the BCS gap equation with its divergence renormalized by the scattering length, and the second is the conservation of electron number (n = Σk2vk²); the key point is that the chemical potential, which BCS theory fixed at μ = EF, is now an unknown. Properties of the solution:

The pair size follows from the momentum-space spread of the pair amplitude φk = ukvk = Δ/2Ek (the quantity defined in Eq. (8), localized around the Fermi surface with width Δ; it is distinct from the pair function vk/uk of the projected state, which is largest deep inside the Fermi sea) (Pistolesi–Strinati 1994):

(10) ξpair2=∫0∞k2|dφk/dk|2dk∫0∞k2φk2dk, kFξpair→{∼EF/Δ(BCS)kFa/2(BEC)

Compared with the electron spacing d = n−1/3 = 3.09/kF, we have ξpair ≫ d on the BCS side (the bubbles overlap) and ξpair ≪ d on the BEC side (isolated molecules). The point where ξpair = d, that is, kFξpair = 3.09, lies slightly on the BCS side of unitarity, and at unitarity (kFξpair ≈ 1.1) the bubble is already smaller than the electron spacing. The number of electrons in one bubble, (4/9π)(kFξpair)³, corresponds to “how many electrons are visible inside a bubble” in the picture on the slide. ▶ Simulator ②: The bubble picture and μ, Δ, ξpair

5.2 The pair-formation temperature T* and the condensation temperature Tc

The transition temperature obtained from the finite-temperature mean-field theory (the condition Δ → 0) is the actual Tc on the BCS side, but on the BEC side it gives only the “temperature at which molecules form,” T* (of the order of the binding energy Eb), and condensation occurs at a much lower temperature:

(11) kBTBEC=2πℏ22m(n/2ζ(3/2))2/3=0.218kBTF

(an ideal Bose gas of mass 2m and density n/2). Therefore Tc rises exponentially along T* on the BCS side, reaches Tc/TF ≈ 0.2 near unitarity (the cold-atom measurement is 0.17, Ku et al. 2012), and saturates on the BEC side at TBEC ∝ n2/3/m. In the region Tc < T < T* on the BEC side there are pairs but no condensation, a “pseudogap” state. The right-hand figure of simulator ② draws the mean-field T*, TBEC, and a guide connecting the two, (1/T*² + 1/TBEC²)−1/2 (a schematic interpolation).

5.3 Thermal fluctuations — why room-temperature superconductivity is hard

The energy needed to break a pair is 2 minkEk, that is, 2Δ for μ > 0 and 2√(μ² + Δ²) ≈ Eb for μ < 0. For BCS elemental superconductors, 2Δ ≈ 3.5 kBTc ≈ 0.3–4 meV, and the pairs are broken at once by the thermal fluctuations at room temperature, kB·300 K = 26 meV—this is the slide’s statement that “this electron pair is easily broken by thermal fluctuations.” If the binding is strengthened by moving toward the BEC side, the pairs no longer break, but the condensation temperature is limited by TBEC of Eq. (11). For a metal with TF ≈ 5×10⁴ K, TBEC ≈ 10⁴ K leaves ample margin, but to get there the coupling would have to be strengthened to 1/(kFa) ≈ 0, whereas for the phonon mechanism λ ≈ 1 is about the limit (the lattice instability in Eliashberg theory).

How about actual high-temperature superconductors? In the Uemura plot (Tc versus TF), the cuprates reach Tc/TF ≈ 0.02–0.05, which is 1/4 to 1/10 of the upper limit on the BEC side (≈ 0.2) (and higher by 2 to 4 orders of magnitude than the roughly 10⁻⁵–10⁻⁴ of conventional superconductors); the coherence length is 1–2 nm and there are a few to a dozen or so electrons inside a bubble, a state close to the “crossover” picture on the slide. On the underdoped side there is a pseudogap at temperatures above Tc, which Emery and Kivelson (1995) interpreted as a state in which “pairs have formed but their phases are not aligned.” Scanning tunneling microscopy of Bi₂Sr₂CaCu₂O8+δ shows the pairing gap appearing above Tc as nanometer-scale patches, which grow on cooling and cover the whole sample (Gomes et al. 2007)—an example in which “superconducting bubbles” are literally visible as individual objects. In disordered thin-film superconductors as well, strong disorder makes the superconductivity island-like (emergent granularity), and theory (Ghosal–Randeria–Trivedi 2001; Dubi–Meir–Avishai 2007) and STM (Sacépé et al. 2011) have shown that, although pairs exist inside the islands, the phases of different islands are not aligned and the film becomes an insulator.

Summary: how Cooper pairs form (1) When an electron passes, the heavy ions are pulled in after a delay, leaving a wake of positive charge several tens to several hundred nm behind the electron. (2) Another electron is attracted to the wake—since the two electrons are never at the same place at the same time, they avoid the Coulomb repulsion (screened at 0.1 nm). In momentum space, the exchange of phonons with ħω < ħωD is the attraction. (3) On top of a Fermi sea, even this weak attraction always produces a bound state (Cooper), and the pair size is ξ₀ = ħvF/πΔ ≈ 10²–10³ nm (about 80 nm for Nb, in the µm range for Al and Zn), with 10⁸–10¹³ electrons inside a bubble (BCS limit). (4) As the coupling is strengthened (the density lowered), the bubble shrinks, becomes comparable to the electron spacing near unitarity, and in the BEC limit becomes an isolated molecule. (5) BCS pairs, with 2Δ of the order of meV, are broken by thermal fluctuations at room temperature, and on the BEC side Tc levels off, with Tc ≤ 0.2 TF—this is why room-temperature superconductivity is difficult.

6. Implementation Notes

7. References

  1. Computational Materials Science, Lecture 7 (lecture slides, PDF, in Japanese), slides 32–34: fermions and bosons, electron-pair formation by lattice vibrations or antiferromagnetic spin fluctuations, and the “superconducting bubble” from the BCS limit to the BEC limit.
  2. H. Fröhlich, Theory of the Superconducting State. I, Phys. Rev. 79, 845 (1950).
  3. J. Bardeen and D. Pines, Electron-Phonon Interaction in Metals, Phys. Rev. 99, 1140 (1955).
  4. L. N. Cooper, Bound Electron Pairs in a Degenerate Fermi Gas, Phys. Rev. 104, 1189 (1956).
  5. J. Bardeen, L. N. Cooper, and J. R. Schrieffer, Theory of Superconductivity, Phys. Rev. 108, 1175 (1957).
  6. E. Maxwell, Phys. Rev. 78, 477 (1950); C. A. Reynolds, B. Serin, W. H. Wright, and L. B. Nesbitt, Phys. Rev. 78, 487 (1950). — Isotope effect.
  7. J. R. Schrieffer, Theory of Superconductivity (Benjamin, 1964); M. Tinkham, Introduction to Superconductivity, 2nd ed. (McGraw-Hill, 1996); P. G. de Gennes, Superconductivity of Metals and Alloys (Benjamin, 1966).
  8. A. M. Kadin, Spatial Structure of the Cooper Pair, J. Supercond. Nov. Magn. 20, 285 (2007), arXiv:cond-mat/0510279. — Real-space pair wave function.
  9. W. L. McMillan, Phys. Rev. 167, 331 (1968); G. M. Eliashberg, Sov. Phys. JETP 11, 696 (1960); A. B. Migdal, Sov. Phys. JETP 7, 996 (1958).
  10. D. M. Eagles, Possible Pairing without Superconductivity at Low Carrier Concentrations in Bulk and Thin-Film Superconducting Semiconductors, Phys. Rev. 186, 456 (1969).
  11. A. J. Leggett, Diatomic molecules and Cooper pairs, in Modern Trends in the Theory of Condensed Matter, edited by A. Pękalski and J. Przystawa (Springer, 1980), p. 13.
  12. P. Nozières and S. Schmitt-Rink, Bose condensation in an attractive fermion gas: From weak to strong coupling superconductivity, J. Low Temp. Phys. 59, 195 (1985).
  13. F. Pistolesi and G. C. Strinati, Evolution from BCS superconductivity to Bose condensation: Role of the parameter kFξ, Phys. Rev. B 49, 6356 (1994); M. Marini, F. Pistolesi, and G. C. Strinati, Eur. Phys. J. B 1, 151 (1998).
  14. Q. Chen, J. Stajic, S. Tan, and K. Levin, BCS–BEC crossover: From high temperature superconductors to ultracold superfluids, Phys. Rep. 412, 1 (2005); M. Randeria and E. Taylor, Annu. Rev. Condens. Matter Phys. 5, 209 (2014); G. C. Strinati, P. Pieri, G. Röpke, P. Schuck, and M. Urban, Phys. Rep. 738, 1 (2018).
  15. M. J. H. Ku, A. T. Sommer, L. W. Cheuk, and M. W. Zwierlein, Revealing the Superfluid Lambda Transition in the Universal Thermodynamics of a Unitary Fermi Gas, Science 335, 563 (2012). — Tc/TF = 0.167 for the unitary gas.
  16. Y. J. Uemura et al., Universal Correlations between Tc and ns/m* in High-Tc Cuprate Superconductors, Phys. Rev. Lett. 62, 2317 (1989); Y. J. Uemura et al., Phys. Rev. Lett. 66, 2665 (1991). — The Uemura plot.
  17. V. J. Emery and S. A. Kivelson, Importance of phase fluctuations in superconductors with small superfluid density, Nature 374, 434 (1995).
  18. K. K. Gomes, A. N. Pasupathy, A. Pushp, S. Ono, Y. Ando, and A. Yazdani, Visualizing pair formation on the atomic scale in the high-Tc superconductor Bi₂Sr₂CaCu₂O8+δ, Nature 447, 569 (2007).
  19. A. Ghosal, M. Randeria, and N. Trivedi, Phys. Rev. B 65, 014501 (2001); Y. Dubi, Y. Meir, and Y. Avishai, Nature 449, 876 (2007); B. Sacépé et al., Nat. Phys. 7, 239 (2011). — Island formation in disordered superconductors.
  20. A. P. Drozdov, M. I. Eremets, I. A. Troyan, V. Ksenofontov, and S. I. Shylin, Nature 525, 73 (2015) (H₃S); A. P. Drozdov et al., Nature 569, 528 (2019) (LaH₁₀). — Hydrides under ultrahigh pressure.
  21. T. Moriya and K. Ueda, Spin fluctuations and high temperature superconductivity, Adv. Phys. 49, 555 (2000); D. J. Scalapino, A common thread: The pairing interaction for unconventional superconductors, Rev. Mod. Phys. 84, 1383 (2012). — The spin-fluctuation mechanism.
  22. N. W. Ashcroft and N. D. Mermin, Solid State Physics (Saunders, 1976); C. Kittel, Introduction to Solid State Physics (table of isotope effects, Thomas–Fermi screening); G. A. Webb, F. Marsiglio, and J. E. Hirsch, Physica C 514, 17 (2015) (θD); M. Kawamura et al., arXiv:2603.05123 (2026) (vF).

Mochizuki Research Group, Tokyo University of Science. This page quantifies the “fermions behave like bosons” picture and the “superconducting bubble” of Lecture 7 (slides 33–34) of Computational Materials Science in terms of the electron–lattice interaction, the Cooper problem, and the BCS–BEC crossover. For numerical experiments, go to the simulator (cooper-pair-simulator-en.html). Related: the free energy of superconductors and spontaneous symmetry breaking and relativistic quantum mechanics. The equations are typeset in the browser’s native MathML (there are no external libraries or network communication at all).