Free Energy and Spontaneous Symmetry Breaking in Superconductors — Ginzburg–Landau Theory, BCS Theory, the Higgs Mode and the Nambu–Goldstone Mode, and the Correspondence with “Phonons with Imaginary Frequencies”
— A reading of Chapters 2 and 3 of Naoto Tsuji’s “Physics of Nonequilibrium Superconductivity” (text of the Summer School for Young Condensed-Matter Physicists), with explanations, derivations, and a literature survey for the companion simulator (superconductivity-simulator-en.html)
A superconductor is a state in which electrons form Cooper pairs and condense, so that the complex order parameter ψ = |ψ|eiθ, the “wave function” of the pairs, becomes finite.
|ψ|² is the density of pairs, and the phase θ has a special meaning: the Hamiltonian is invariant under the transformation ψ → eiαψ that rotates θ everywhere at once (a U(1) transformation, the gauge transformation corresponding to charge conservation), yet the superconducting state picks out one particular θ — this is spontaneous symmetry breaking.
Following the equations of Chapter 2 (Ginzburg–Landau theory) and Chapter 3 (BCS theory and time-dependent BCS theory) of the lecture notes one by one, this explanation derives (i) how the “Mexican hat” of the free energy turns into a bowl as the temperature rises, so that the symmetry-broken state becomes unstable, (ii) the two oscillations on the hat — the Higgs mode of the amplitude and the Nambu–Goldstone mode of the phase, (iii) the Anderson–Higgs mechanism, in which the phase mode is absorbed by the electromagnetic field (the Meissner effect), and (iv) the BCS mean-field transition temperature,
and then summarizes (v) how Yoichiro Nambu’s “spontaneous symmetry breaking” arose from the theory of superconductivity, (vi) how the same free energy appears in lattices as the soft mode — the “phonon with imaginary frequency” of DFT — and (vii) a literature survey on whether phonons also have Higgs/Goldstone modes.
Roadmap of this explanation
§1 The order parameter and U(1) symmetry.
§2 The GL free energy: the Mexican hat, a = a₀(T − Tc), the London equation and the Meissner effect, ξ, λ, κ, and Hc.
§3 Determining a and b from measured values (Tc, Hc, λ): a table of 12 elements, and the effect of a uniform vector potential.
§4 Time-dependent GL: the Higgs mode, the Nambu–Goldstone mode, the Anderson–Higgs mechanism, and the dispersion relation ω² = (2Δ)² + vF²q²/3.
§5 Nonlinear response and the third harmonic.
§6 BCS mean field: the gap equation, Tc, the McMillan formula, and the correspondence with Gor’kov’s derivation.
§7 Anderson’s pseudospin and Higgs oscillations (quench and drive).
§8 Nambu and the history of spontaneous symmetry breaking.
§9 Spontaneous symmetry breaking ≡ phonons with imaginary frequencies.
§10 Literature survey: Higgs/Goldstone modes of phonons.
§11 Implementation notes. §12 References.
- The Order Parameter and U(1) Symmetry
- The Ginzburg–Landau Free Energy
- Determining the Coefficients from Measured Values: The Table of Elements
- Time-Dependent GL Theory and Collective Excitations
- Nonlinear Response and the Third Harmonic
- BCS Mean Field and the Transition Temperature
- Anderson’s Pseudospin and Higgs Oscillations
- Nambu and Spontaneous Symmetry Breaking
- Spontaneous Symmetry Breaking ≡ Phonons with Imaginary Frequencies
- Literature Survey: Higgs/Goldstone Modes of Phonons
- Implementation Notes
- References
1. The Order Parameter and U(1) Symmetry
In Landau’s theory of phase transitions, a phase transition is described by an “order parameter” going from 0 to a finite value. The order parameter of a superconductor is the complex field ψ(r) that represents the condensation of Cooper pairs, and
in the language of BCS theory it is proportional to the gap function Δ = (V/N)Σk⟨c−k↓ck↑⟩ (§6). Because this is the expectation value of an operator that annihilates two electrons, something that ought to vanish in a state of fixed electron number has become finite — which means allowing “uncertainty in the electron number,”
that is, the state breaks the U(1) symmetry c → eiαc (ψ → e2iαψ) corresponding to charge conservation. In the language of the lecture notes, the starting point of GL theory is that the free energy is “invariant under the U(1) transformation ψ → eiθψ and under the particle–hole transformation ψ ↔ ψ†.”
Even though the free energy is symmetric, its minimum is not symmetric (a single phase is chosen). This is spontaneous symmetry breaking, and, as we will see below, two kinds of fluctuations always appear around the broken state — one in the direction that restores the symmetry (the phase) and one that changes the magnitude of the order (the amplitude).
2. The Ginzburg–Landau Free Energy — The Mexican Hat Becomes a Bowl
2.1 The free-energy functional (Eq. (1) of the lecture notes)
Near Tc, where the order parameter is small, the free-energy density can be expanded in powers of |ψ|² and in gradients. The GL free-energy density coupled to the electromagnetic field is (the lecture notes set ħ = 1; here we keep ħ)
(1)
m* and e* are the “effective mass” and “effective charge”; from Gor’kov’s microscopic derivation (§6.3), a Cooper pair has m* = 2m and e* = 2e.
The third term is the kinetic energy, written so as to be invariant under the gauge transformation A → A + ∇χ, ψ → eie*χ/ħψ. The last term is the energy of the magnetic field (omitted in the lecture notes).
(2)
For a uniform system (∇ψ = 0) without a magnetic field, f = a|ψ|² + (b/2)|ψ|⁴. We need b > 0 (otherwise it diverges as |ψ| → ∞), and the sign of a determines the phase: if a > 0 the minimum is at ψ = 0 (normal state), and if a < 0
(3)
For a < 0 the minimum is not ψ = 0 but the entire circle of radius √(−a/b) in the complex plane. Plotting |ψ| as the horizontal distance and f as the height gives the “Mexican hat”: the center is raised and the rim is a trough.
Every point on the rim (every phase θ) has the same energy, so the system chooses one of them — this is the spontaneous breaking of U(1) symmetry.
Landau assumed that a changes sign at Tc and, in the simplest way, put a = a₀(T − Tc) (in Gor’kov’s derivation a₀ = N(0)/Tc, §6.3).
As the temperature rises, |a| decreases and the radius of the rim shrinks; the hat flattens, the curvature at the center vanishes at T = Tc, and for T > Tc it is a bowl. “The destabilization of the U(1)-symmetry-broken state with temperature” means that the rim of the hat shrinks toward the center and disappears.
▶ Simulator ①: The Mexican hat and temperature
2.2 The GL equation and the superconducting current
Varying Eq. (1) with respect to ψ* and A gives the GL equation and the current density (Eq. (3) of the lecture notes).
(4)
2.3 The London equation and the Meissner effect
If |ψ| is uniform (|ψ|² = ns) and the phase is constant, only the second term remains in the current of Eq. (4) (Eqs. (4)–(5) of the lecture notes):
(5)
The current is proportional to the vector potential — this is the London equation, and combined with the Maxwell equation ∇×B = μ₀j it gives the Meissner effect, in which the magnetic field decays exponentially from the surface over the penetration depth λ.
Inserting m* = 2m, e* = 2e, and ns = n/2 (the density of pairs) for Cooper pairs gives λ = √(m/μ₀ne²), which is 10–50 nm for elemental metals.
The lecture notes restate this in terms of the optical conductivity: the Drude weight D = πe*²ns/m* of σ(ω) = D[δ(ω) + i/(π(ω + i0))] is finite, and the imaginary part diverges in the dc limit (perfect conductivity); this and the Meissner effect are both consequences of the same Eq. (5).
2.4 Coherence length, penetration depth, κ, and the critical field
Linearizing Eq. (4) near Tc fixes the length scale of the spatial variation of ψ.
(6)
ξ is the shortest length over which the order parameter can vary (the stiffness of the gradient term along the rim of the hat), λ is the penetration depth of the magnetic field, and κ is a ratio that does not depend on temperature: for κ < 1/√2 the superconductor is Type I (it either expels the magnetic field completely or returns to the normal state), and for κ > 1/√2 it is Type II (vortices penetrate).
Hc is the condensation energy −a²/2b measured in units of magnetic-field energy (the thermodynamic critical field), and using the flux quantum Φ₀ = h/2e it can also be written μ₀Hc = Φ₀/(2√2 π λ ξ).
3. Determining the GL Coefficients from Measured Values — The Table of 12 Elements and the Effect of a Vector Potential
3.1 From Hc(0) and λ(0) to a and b
GL theory is a theory for the vicinity of Tc, but for educational purposes we want to know “how deep the hat of a real superconductor is.” We therefore extrapolate the relations of Eqs. (3), (5), and (6) linearly down to T = 0 and fix the coefficients from the measured critical field and penetration depth:
(7)
(8)
Using m* = 2me and e* = 2e gives ns(0) = me/2μ₀e²λ², that is, London’s λ is explained by a density half that of the conduction electrons.
The results are collected in the table below (the same values as the presets of Simulator ①). Taking Pb as an example, a(0) ≈ −3 µeV, b ≈ 0.3 µeV nm³, and the condensation energy density is 2.6 kJ/m³ = 1.5 µeV per pair — only about 10−7 times the Fermi energy (about 9.5 eV for free electrons), which shows that superconductivity is a “weak” order (10−10–10−6 times depending on the element; a rough guide is Δ²/εF).
| Element | Tc (K) | μ₀Hc(0) (mT) | θD (K) | λ(0) (nm) | ξGL(0) (nm) | κ | ns(0) (10²² cm⁻³) | a(0) (µeV) | b (µeV nm³) | fc (J/m³) | N(0)V | λep | Δ(0) (meV) | 2Δ/kBTc (measured) |
| Al | 1.140 | 10.5 | 428 | 16 | 1385 | 0.012 | 5.5 | −0.00993 | 0.00018 | 44 | 0.165 | 0.43 | 0.173 | 3.40 |
| Ga | 1.091 | 5.1 | 320 | 16† | 2798 | 0.0058 | 5.3 | −0.00243 | 0.000046 | 10 | 0.172 | 0.45 | 0.166 | — |
| In | 3.404 | 29.3 | 108 | 30 | 265 | 0.11 | 1.6 | −0.272 | 0.017 | 340 | 0.279 | 0.80 | 0.517 | 3.60 |
| Tl | 2.390 | 17.1 | 78.5 | 22† | 620 | 0.035 | 2.9 | −0.0495 | 0.0017 | 120 | 0.276 | 0.79 | 0.363 | 3.57 |
| Sn | 3.722 | 30.9 | 200 | 34 | 222 | 0.15 | 1.2 | −0.388 | 0.032 | 380 | 0.243 | 0.66 | 0.566 | 3.50 |
| Pb | 7.193 | 80.3 | 105 | 37 | 78 | 0.47 | 1.0 | −3.11 | 0.30 | 2600 | 0.356 | 1.24 | 1.093 | 4.38 |
| Zn | 0.875 | 5.3 | 327 | 18† | 2507 | 0.0070 | 4.6 | −0.00303 | 0.000066 | 11 | 0.165 | 0.43 | 0.133 | 3.20 |
| Cd | 0.560 | 3.0 | 209 | 110 | 705 | 0.16 | 0.12 | −0.0383 | 0.033 | 3.6 | 0.165 | 0.43 | 0.085 | 3.20 |
| Hg | 4.153 | 41.2 | 72 | 26† | 216 | 0.12 | 2.1 | −0.409 | 0.020 | 680 | 0.336 | 1.10 | 0.631 | 4.60 |
| V | 5.380 | 142.0 | 380 | 50† | 33 | 1.5 | 0.57 | −17.4 | 3.0 | 8000 | 0.228 | 0.61 | 0.818 | 3.40 |
| Nb | 9.500 | 198.0 | 275 | 39 | 30 | 1.3 | 0.93 | −21.0 | 2.3 | 16000 | 0.286 | 0.83 | 1.444 | 3.80 |
| Ta | 4.483 | 83.0 | 240 | 35 | 80 | 0.44 | 1.2 | −2.97 | 0.26 | 2700 | 0.244 | 0.66 | 0.681 | 3.60 |
Data sources and estimated values (marked †)
Tc, Hc(0), and the measured 2Δ/kBTc are from the tables in Kittel’s “Introduction to Solid State Physics”; θD is from Table III of Webb–Marsiglio–Hirsch (2015).
λ(0) is from Kittel (the table of Meservey–Schwartz: the London value of 16 nm for Al, 34 for Sn, 37 for Pb, 110 for Cd, and 39 for Nb); for In it is the direct measurement of weak-field penetration, 29.9 nm (Raychaudhuri et al. 1985), and for Ta it is 33–35 nm (McEvoy et al. 1969, Auer–Ullmaier 1973).
Among the † entries, for Ga, Tl, Zn, and Hg no measured values were available in the literature at hand, so they are estimates obtained by multiplying the free-electron London value √(m/μ₀ne²) (with the n of Ashcroft–Mermin) by the mass enhancement √(1 + λep) due to the electron–phonon interaction; for V, the value was back-calculated with Eq. (8) from κ = 1.5 (Radebaugh–Keesom 1966, analysis of the specific heat in the mixed state) and Hc(0).
ξGL(0) is the GL length calculated with Eq. (8); it is related to the Pippard length of the clean limit, ξ₀ = ħvF/πΔ(0), by ξGL(0) ≈ 0.74 ξ₀ (roughly consistent with the measured values of Al 1600, Sn 230, Pb 83, Cd 760, and Nb 38 nm).
Note that the experimental values themselves scatter by about a factor of 2 (compiled in Table I of Kawamura et al. 2026). N(0)V is from the BCS formula, and λep is the value obtained by inverting the McMillan formula (μ* = 0.13) from Tc (§6.2).
3.2 A uniform vector potential — superfluid velocity and pair breaking
Let us put a uniform A with constant phase into the kinetic term of Eq. (1) of the lecture notes. The gauge-invariant combination is A − (ħ/e*)∇θ, so a uniform A is equivalent to a state with a “phase gradient,” that is, with a superfluid velocity vs:
(9)
(10)
Since the kinetic energy is added to the coefficient of |ψ|², a uniform A (superflow) flattens the hat, and at A = Ac we have aeff = 0 and the superconductivity disappears (the actual depairing current is at the maximum of j, ½m*vs² = |a|/3, where jdp = Φ₀/(3√3 π μ₀λ²ξ)).
In Simulator ① you can vary |A| in units of ħ/2eξ(0). For Pb the orders of magnitude are A ≈ 4 T·nm, vs ≈ 1 km/s, and j ≈ 10⁸ A/cm², in agreement with the first-principles depairing current (Kawamura et al. 2026: 1.5×10⁸ A/cm² for Pb).
For a neutral superfluid with e* = 0, A does not couple, but the same effect from the phase gradient ∇θ (the GL version of Landau’s critical velocity) remains.
4. Time-Dependent GL Theory and Collective Excitations — The Higgs Mode, the Nambu–Goldstone Mode, and the Anderson–Higgs Mechanism
4.1 The Lagrangian of time-dependent GL (lecture notes §2.2)
To treat the dynamics of the order parameter, we consider a Lagrangian obtained by adding time-derivative terms to the free energy (1). Eq. (9) of the lecture notes is (ħ = 1)
(11)
The c₁ term is first order in time (Gross–Pitaevskii type, Schrödinger type), and the c₂ term is second order in time (Klein–Gordon type).
The electromagnetic potentials are included so that the Lagrangian is invariant under the gauge transformation A → A + ∇χ, φ → φ − ∂tχ, ψ → eie*χψ.
As the footnote of the lecture notes emphasizes, time-dependent GL theory can be justified microscopically only in limited cases such as gapless superconductors, and even then the time-derivative term that appears is a damping term (∝ −∂tψ) and cannot describe oscillations. Here we treat it “generously” as a phenomenology (in §7 it is replaced by the microscopic theory).
4.2 Amplitude and phase — the Higgs mode and the Nambu–Goldstone mode
We write the deviation from the ground state ψ₀ = √(−a/b) (taken to be a positive real number) in terms of an amplitude and a phase (Eq. (10) of the lecture notes).
(12)
Substituting this into Eq. (11) and expanding gives (Eq. (11) of the lecture notes)
(13)
- The first term 2aH² is the mass term of the Higgs mode (since a < 0 it is −2|a|H², that is, a restoring force acts on H). The energy gap is ∝ (−2a)1/2 ∝ (Tc − T)1/2. In the picture of the hat, it is the curvature “perpendicular to the rim” at a point on the rim. The second term is the kinetic term of the Higgs mode.
- The phase θ has no mass term (∝ θ²). This is because there is no energy gradient in the direction along the rim of the hat, and this is precisely Goldstone’s theorem: a gapless mode always appears when a continuous symmetry is spontaneously broken. In a charge-neutral (e* = 0) superfluid, a gapless phase mode actually exists (the Anderson–Bogoliubov mode; the first sound of ⁴He, the sound waves of cold atoms).
- The phase and the amplitude mix linearly through −(2c₁/e*) ∂tθ ψ₀H, which comes from the fourth term. For them to be independent modes we need c₁ = 0; since the c₁ term breaks particle–hole symmetry, it is said that “particle–hole symmetry is required to separate the amplitude mode from the phase mode” (Pekker–Varma). In a superconductor, particle–hole symmetry holds approximately at low energies around the Fermi level, so c₁ can be neglected.
4.3 The Anderson–Higgs mechanism — the phase mode is “eaten” by the electromagnetic field
In a charged superconductor (e* ≠ 0), θ appears only in the combinations A − ∇θ/e* or φ + ∂tθ/e*.
Therefore the gauge transformation A → A − ∇θ/e*, φ → φ + ∂tθ/e* (the unitary gauge) can apparently remove θ. Writing the new potentials again as A and φ, we obtain (Eq. (12) of the lecture notes)
(14)
The phase mode θ has disappeared, and instead a mass term for the electromagnetic field, ∝ e*²ψ₀²A²/2m*, has appeared in the third term. This is the Anderson–Higgs mechanism (Anderson 1958, 1963; Englert–Brout, Higgs, and Guralnik–Hagen–Kibble 1964).
That the electromagnetic field has a mass means that it cannot propagate freely through the superconductor as in vacuum, which is precisely the Meissner effect (the London equation of §2.3 is this mass term differentiated with respect to A).
A massive electromagnetic field has, in addition to the two transverse components, a longitudinal one (a component polarized along the direction of propagation), so the number of modes is
(15)
and the counts match (Eq. (13) of the lecture notes). Physically, the phase mode mixes with the long-range Coulomb interaction and is pushed up to the plasma frequency ωp (ħωp ≈ 10 eV for elemental metals, 10³–10⁴ times 2Δ), and does not remain at low energies.
What remains is only the Higgs mode, which is the only collective excitation of the order parameter that appears at low energies in an ordinary superconductor (the excitation spectrum of Simulator ③).
Near the critical temperature, thermally excited normal electrons screen the electromagnetic field and the phase mode can revive; this is called the Carlson–Goldman mode (1975).
4.4 The dispersion relation of the Higgs mode and the gap 2Δ
Setting c₁ = 0 and forming the Euler–Lagrange equation for H from Eq. (14) gives (Eq. (14) of the lecture notes)
(16)
In the absence of an electromagnetic field, substituting a plane wave H ∼ eiq·r − iωt gives the dispersion of the Higgs mode and its gap (Eqs. (15)–(16) of the lecture notes).
(17)
Time-dependent GL theory does not fix the value of the gap itself, but microscopic calculations (§7) show that ωH = 2Δ (the same as the energy at which Cooper pairs begin to be broken and quasiparticles excited).
When GL theory is derived from the microscopic theory (Gor’kov), the coefficient b satisfies b(ψ₀/Δ)² = 3/4εF (Eq. (18) of the lecture notes), and combining this with ωH = 2Δ fixes c₂ (Eq. (19) of the lecture notes):
(18)
Since these are Cooper pairs, using m* = 2m and εF = ½mvF², the dispersion of the Higgs mode can be written in terms of microscopic quantities alone (Eq. (20) of the lecture notes; it agrees with the result that Littlewood–Varma obtained from a microscopic calculation in 1982).
(19)
The gap 2Δ of the Higgs mode coincides with the lower edge of the quasiparticle continuum, so the Higgs mode “touches” the continuum and easily decays into quasiparticles (the t−1/2 decay of §7.3). This is why the Higgs mode is difficult to observe in superconductors.
▶ Simulator ③: The excitation spectrum and the Anderson–Higgs mechanism
5. Nonlinear Response — Why the Higgs Mode Cannot Be Seen Directly with Light, and the Resonance of the Third Harmonic
Looking at Eq. (14), there is no linear coupling between the Higgs mode H and the electromagnetic field (A, φ) (only terms of the form A²H and φ²H). The Higgs mode carries neither charge nor magnetization, so it cannot be excited in linear response (the ordinary absorption spectrum).
To excite it, we use the nonlinear coupling A²H, that is, two-photon absorption: two photons of frequency Ω create a Higgs mode of frequency 2Ω (Fig. 3 of the lecture notes).
Putting a single-frequency uniform external field A(t) = A₀e−iΩt, φ = 0 into Eq. (16) gives (Eqs. (21)–(22) of the lecture notes)
(20)
The Higgs field oscillates at frequency 2Ω and its amplitude diverges at the resonance 2Ω = ωH = 2Δ. The observable is the current, and from Eq. (14), j = ∂L/∂A = −(2e*²ψ₀/m*)AH + ⋯ (Eq. (23) of the lecture notes).
Substituting Eq. (20) gives (Eq. (24) of the lecture notes)
(21)
The frequency Ω of the external field combines with the 2Ω of the Higgs mode, so the current oscillates at 3Ω — this is third-harmonic generation (Fig. 4 of the lecture notes). Because the intensity of the third harmonic increases divergently at the Higgs-mode resonance,
Matsunaga et al. (2014) applied terahertz light of fixed frequency to an NbN thin film, swept the temperature, and observed the third harmonic increase at the temperature where 2Δ(T) = 2Ω.
In BCS theory, decay into quasiparticles is included, so the power of the divergence is weakened to |2Ω − 2Δ|−1/2 and the phase jumps by π/2 at the resonance (§7.4, Fig. 7 of the lecture notes).
Whether the third harmonic is carried by the Higgs mode or by quasiparticles (Cooper-pair breaking) can be distinguished by its polarization dependence. The discrepancy between the suggestion that quasiparticles dominate in the clean limit (Cea–Castellani–Benfatto 2016) and the polarization independence seen in experiment (Matsunaga et al. 2017) was resolved by including the paramagnetic coupling due to impurities (p·A type) (Tsuji–Nomura 2020: mainly the Higgs mode in NbN).
▶ Simulator ③: Resonance of the third harmonic
6. BCS Mean Field — The Gap Equation, the Transition Temperature, and the Correspondence with GL
6.1 The BCS Hamiltonian and the Nambu spinor (lecture notes §3.1)
BCS theory is the theory in which electrons with momentum k and spin ↑ and with −k and ↓ form Cooper pairs and condense through an attraction mediated by, for example, phonons. Applying the mean-field approximation to the Hamiltonian that retains only the scattering between pairs (Eq. (25) of the lecture notes) and defining the gap function Δ = (V/N)Σk⟨c−k↓ck↑⟩, we obtain (Eqs. (27)–(28) of the lecture notes)
(22)
Δ/V = (1/N)Σ⟨c−k↓ck↑⟩ is the condensate density of Cooper pairs and corresponds to ψ(r) in GL theory. That the Hamiltonian can be written with the two-component Nambu spinor ψk and the 2×2 matrix hk is the starting point for Anderson’s pseudospin (§7) and for the analogy with the Dirac equation (Nambu’s idea, §8).
Diagonalizing hk gives the quasiparticle energy Ek± = ±√(εk² + Δ²): a gap of 2Δ opens above and below the Fermi energy.
6.2 The gap equation and the transition temperature
Determining Δ self-consistently gives (Eq. (32) of the lecture notes)
(23)
The form on the right takes the density of states to be a constant, the value N(0) at the Fermi level (per spin), and cuts off the range in which the attraction acts at the phonon cutoff ħωD (the Debye energy). At T = 0 the integral can be done, giving
Δ = ħωD/sinh(1/N(0)V), and in weak coupling, N(0)V ≪ 1 (Eq. (33) of the lecture notes),
(24)
Tc is the temperature at which Δ(T) = 0, and is obtained by putting Δ = 0 into Eq. (23) and solving (γ = 0.5772 is Euler’s constant).
e−1/x is a non-analytic function all of whose derivatives vanish at x = 0, which means that superconductivity occurs however weak the attraction is, but never appears in perturbation theory.
With ħωD ≈ several tens of meV, reduced by one to two orders of magnitude by the exponential factor, we get Δ ≈ 0.1–1.5 meV (Al 0.17, Pb 1.09, Nb 1.44 meV; 0.05–0.7 THz in terms of 2Δ), which is the scale of elemental superconductors.
Simulator ② solves Eq. (23) numerically by Simpson’s rule and the bisection method to obtain Δ(T) and Tc (the difference from the weak-coupling formula is a few percent even for Pb, with N(0)V ≈ 0.4).
The measured 2Δ(0)/kBTc is 3.4 for Al, 3.5 for Sn, 3.6 for In, and 3.8 for Nb, close to the BCS value, but Pb at 4.38 and Hg at 4.6 deviate greatly. These are strong-coupling superconductors, in which the electron–phonon interaction is strong (λ ≈ 1) and Eliashberg theory, which incorporates the retardation of the phonons, is needed.
The McMillan formula (1968), which summarizes these effects empirically, is
(25)
where λ = 2∫dω α²F(ω)/ω is the electron–phonon coupling constant and μ* ≈ 0.1–0.15 is the pseudopotential of the Coulomb repulsion. McMillan himself solved this formula backward from the measured Tc and θD to obtain λ for each element (about 1.12 for Pb, 1.0 for Hg, 0.82 for Nb, and 0.38 for Al).
The presets of Simulator ② also back-calculate λ by the same procedure (μ* = 0.13). Note that the N(0)V in the table of the BCS paper (reproduced in Kittel) (0.18 for Al, 0.25 for Sn, 0.39 for Pb, etc.) was also obtained by solving Eq. (24) backward from θD and Tc; if θD is chosen differently (375 K versus 428 K for Al), the resulting N(0)V also changes, and the slight difference between 0.18 in the BCS table and 0.165 in the present table changes the predicted Tc by more than 50% — this is the sensitivity of the exponential function (the BCS column of Simulator ①).
6.3 Correspondence with GL — Gor’kov’s derivation
Gor’kov (1959) derived the GL equations from BCS theory near Tc and showed that ψ is proportional to Δ and that e* = 2e and m* = 2m. In the normalization in which ψ is taken to be Δ itself,
(26)
That is, near Tc, Δ(T) ≈ 3.06 kBTc√(1 − T/Tc) (ζ(3) = 1.202). In the readout of Simulator ②, this can be compared with the numerical solution at 0.9Tc.
The same derivation gives, in the clean limit, ξ(T) = 0.74 ξ₀/√(1 − T/Tc) (ξ₀ = ħvF/πΔ(0) is the Pippard–BCS coherence length) and λ(T) = λL(0)/√(2(1 − T/Tc)).
The procedure of §3, “determining a(0) and b from Hc(0) and λ(0),” is a linear extrapolation of these near-Tc relations down to T = 0.
▶ Simulator ②: The gap equation, the McMillan formula, and the table of elements
7. Anderson’s Pseudospin and Oscillations of the Higgs Mode (Lecture Notes §3.2)
7.1 The time-dependent BdG equation and the pseudospin
Time-dependent mean-field theory advances the time evolution of the Nambu spinor, i∂tψk = hkψk (the time-dependent Bogoliubov–de Gennes equation, Eq. (34) of the lecture notes), while determining Δ(t) self-consistently at each time.
If we prepare the equal-time two-point correlation function σkα = ½⟨ψk†ταψk⟩ (α = x, y, z), it has the form of the expectation value of a spin operator and is therefore called Anderson’s pseudospin (1958). The equation of motion takes the form of the Bloch equation (Eqs. (37)–(38) of the lecture notes):
(27)
Since Δ itself enters the effective magnetic field bk, the equation is nonlinear (it is known to be integrable: Barankov–Levitov–Spivak 2004, Yuzbashyan et al. 2005).
The equilibrium state is σkx = (Δ/2Ek)tanh(βEk/2), σy = 0, σz = −(εk/2Ek)tanh(βEk/2).
The z component is the occupation of each k (the momentum distribution), and the x and y components are the Cooper-pair density including the phase. In the normal state the pseudospins form a domain structure, pointing up or down on either side of εk = 0; in the superconducting state they form a structure twisted over a width of ~Δ near the boundary (Fig. 5 of the lecture notes, the lower-left figure of Simulator ②).
7.2 Linearization and Anderson’s condition — the natural frequency 2Δ of the Higgs mode
Linearizing about equilibrium in the deviations δσk and δΔ (taken to be real) gives (for T = 0; Eqs. (40)–(42) of the lecture notes)
(28)
Fourier transforming and combining with the self-consistency condition δΔ = (V/N)Σδσx, the condition for a nontrivial solution is (Eq. (45) of the lecture notes)
(29)
Putting ω = 2Δ into this reproduces the gap equation (23), so a natural oscillation mode of frequency 2Δ exists. This is the oscillation of the amplitude of the gap function = the Higgs mode of GL.
This was pointed out by Anderson (1958), earlier than the paper of Higgs (1964) that predicted the Higgs particle in particle physics.
7.3 Interaction quench — oscillation at 2Δ and decay as t−1/2
If the attraction is suddenly changed from V to V + δV at t = 0, starting from the equilibrium state at T = 0 (a quench), the self-consistency condition becomes δΔ(t) = (Δ/V)δV + (V/N)Σδσx (Eq. (46) of the lecture notes), which can be solved by the Laplace-transform method of Volkov–Kogan (1974). The long-time asymptotic form is (Eq. (51) of the lecture notes)
(30)
The amplitude of the gap oscillates at frequency 2Δ, and the oscillation decays slowly, as a power law t−1/2 (Fig. 6 of the lecture notes). The decay occurs even without a collision term because of a mechanism of the “Landau damping” type: each pseudospin precesses with a different frequency 2Ek for each k, and the phases spread out. However, because of the integrability, the system as a whole does not thermalize.
If the pump light in terahertz pump–probe spectroscopy is regarded as a quench, the gap in NbN has been observed to oscillate at frequency 2Δ while decaying (Matsunaga et al. 2013).
The lower-right figure of Simulator ② integrates the linearized Eq. (28) by RK4 on a 300-point k grid, and one can confirm that it follows the asymptotic form of Eq. (30) (gray dashed line).
7.4 Driving by an oscillating electric field and the third harmonic (BCS version)
When the electric field is introduced through the Peierls phase εk−A(t) (Eqs. (52)–(54) of the lecture notes), the z component of the effective field becomes an even function of A, and the order parameter couples to the square of the electric field (consistent with the two-photon absorption of GL).
The long-time result (Eq. (55) of the lecture notes) has, in addition to the same t−1/2 term as for the quench, a term oscillating at 2Ω, and it resonates at 2Ω = 2Δ; but the power of the divergence is weakened from the GL |2Ω − 2Δ|−1 to |2Ω − 2Δ|−1/2, and the phase jumps by π/2 at the resonance (Fig. 7 of the lecture notes; Tsuji–Aoki 2015).
This is because in BCS the Higgs mode corresponds not to a pole but to a branch point (it is degenerate with the quasiparticle continuum). The third-order component of the current is j(3)(t) = (2/V)ΔδΔ(t)A(t) + (K/6)A(t)³ (Eq. (58) of the lecture notes, for (1,1,1) polarization in a cubic lattice), and as in Eq. (21) of GL, the term proportional to δΔ is responsible for the resonance of the third harmonic.
8. Nambu and “Spontaneous Symmetry Breaking” — From Superconductivity to Particle Physics
Yoichiro Nambu’s 2008 Nobel Prize in Physics was awarded “for the discovery of the mechanism of spontaneous broken symmetry in subatomic physics,” but its starting point was the theory of superconductivity. The history as Nambu himself tells it in his Nobel lecture (Rev. Mod. Phys. 81, 1015 (2009)) is as follows.
- 1956: Schrieffer gave a seminar in Chicago on the theory later to be called BCS theory. Nambu was impressed by the boldness of the state vector but worried that it did not seem to respect gauge invariance.
- Soon afterwards Bogoliubov and Valatin introduced quasiparticles, but a quasiparticle is a superposition of an electron and a hole and has no definite charge. Is it then legitimate to discuss electromagnetic properties such as the Meissner effect with BCS theory? It took him two years to satisfy himself.
- The answer (Nambu 1960, Phys. Rev. 117, 648): what saves gauge invariance (charge conservation) is a massless collective mode, an excitation that tries to restore the lost symmetry (its energy → 0 in the long-wavelength limit). Today it is called the Nambu–Goldstone boson. This mode further mixes with the Coulomb interaction to become the plasmon ωp² = ne²/ε₀m — the prototype of the Anderson–Higgs mechanism of §4.3.
- Because the Bogoliubov–Valatin equation is formally similar to the Dirac equation, Nambu transferred BCS theory to particle physics: the gap Δ corresponds to the nucleon mass M, the electromagnetic current to the axial current; chiral symmetry is broken and the nucleon acquires a mass, and the π meson is its Nambu–Goldstone boson (massless if the symmetry were exact). The concrete model with Jona-Lasinio (the Nambu–Jona-Lasinio model, 1961) has the same form as the BCS model.
- “When a continuous symmetry is spontaneously broken, a gapless mode always appears” became a theorem with Goldstone (1961) and Goldstone–Salam–Weinberg (1962); Anderson (1963) pointed out, for superconductors (plasmons), that in the presence of a gauge field this mode is absorbed by the gauge field and a mass is generated; Englert–Brout, Higgs, and Guralnik–Hagen–Kibble (1964) formulated this relativistically, and it was used in the Weinberg–Salam electroweak theory (1967). In the Nobel lecture he also remarked that the WS theory resembles GL theory and that the NJL model goes over into the Gell-Mann–Lévy model, and that if this analogy is real, the Higgs field may be an effective description of an underlying dynamics.
- At the end of the lecture Nambu remarks that the BCS mechanism creates a mass gap for fermions and produces the Nambu–Goldstone mode and the Higgs mode as low-energy bosons, and that these bosons in turn trigger the next symmetry breaking (tumbling); as an example, he gives the chain “atoms → crystal → phonons → superconductivity”: the Nambu–Goldstone mode at the formation of a crystal is the phonon, and that phonon causes Cooper pairs to form and superconductivity to occur. This is the very subject of §9.
Summary of terms
Nambu–Goldstone mode = oscillation of the phase θ (along the rim, massless). In a superconductor it turns into the plasmon.
Higgs mode = oscillation of the amplitude |ψ| (perpendicular to the rim, gap 2Δ). Pointed out in superconductors by Anderson (1958).
Anderson–Higgs mechanism = a gauge field absorbs the Nambu–Goldstone mode and acquires a mass = the Meissner effect.
The Higgs boson of particle physics is the relativistic version of the Higgs mode, the masses of the W and Z bosons that of the Anderson–Higgs mechanism, and the π meson (and the nucleon mass) that of the Nambu–Goldstone mode (and Δ).
9. Spontaneous Symmetry Breaking Is Essentially the Same Thing as a “Phonon with Imaginary Frequency”
9.1 Landau theory of structural phase transitions and the soft mode
The order parameter of a structural phase transition is a pattern of atomic displacements Q (the amplitude of the one mode, among the lattice vibrations of the high-symmetry structure, that freezes at the transition). Landau’s free energy is
(31)
which has the same form as Eq. (1) (a ↔ ½A(T − T₀), b/2 ↔ ¼B). For T > T₀ the high-symmetry structure with Q = 0 is stable, and the oscillation of Q is given, using the effective mass M of the mode, by
(32)
so it softens toward the transition point (the soft mode of Cochran 1960; see the review by Scott 1974). If we make the harmonic approximation about the high-symmetry structure for T < T₀, then Mω² = A(T − T₀) < 0, that is, the frequency is imaginary.
Why ω² is proportional to temperature, and why the proportionality breaks down near 0 K
The linear law Mω² = A(T − T₀) arises because the quartic anharmonic term stiffens the soft mode through the mean-square thermal atomic displacement: ω²(T) = ωharm² + g⟨u²⟩. In the classical regime, equipartition gives ⟨u²⟩ ∝ kBT, so ω² is proportional to T, and ωharm² < 0 (the negative curvature given by DFT at T = 0) is cancelled at T₀. Quantum mechanically, ⟨u²⟩ = (ħ/2MΩ)coth(ħΩ/2kBT), which stays at the zero-point value as 0 K is approached, so the temperature dependence of ω² flattens out (Barrett’s formula ω² ∝ (T₁/2)coth(T₁/2T) − T₀, Barrett 1952). If ω² remains positive where it flattens out, the transition does not occur, and one has the quantum paraelectricity of SrTiO₃ and KTaO₃ (Müller–Burkard 1979). As a real example of passing to a quantum paraelectric through composition, in Li₂Sr(Nb1−xTax)₂O₇ the temperature of the weak ferroelectric transition falls from 220 K (x = 0) to 40 K (x = 0.3) as the Ta content x is increased, and at x = 0.4 the transition disappears and the material becomes an incipient ferroelectric (Nagai, Mochizuki et al. 2020: analyzed with Barrett’s formula). In Simulator ④, the slider “Quantum fluctuations T₁/T₀” lets you introduce Barrett’s saturation: with the effective ã = (θ/2)coth(θ/2τ) − 1 (θ = T₁/T₀, τ = T/T₀), the classical reference line and ω² including the quantum saturation are overlaid, the transition temperature is Tc/T₀ = θ/ln((2 + θ)/(2 − θ)), and for θ ≥ 2 no transition occurs and the system remains a single well (the SrTiO₃-like preset).
9.2 The “phonon with imaginary frequency” of DFT = negative curvature at the center of the hat
In first-principles calculations of the phonons of a high-symmetry structure (DFPT or the finite-displacement method of phonopy), branches may appear for which the eigenvalue ω² of the dynamical matrix is negative. By convention these are displayed as negative frequencies, −√|ω²|; they are the so-called phonons with imaginary frequencies.
This is precisely the value Mω² = −AT₀ of Eq. (32) at T = 0, and means that the energy surface is not convex at the center of the hat (the symmetric structure), that is, the symmetric structure is unstable and the symmetry is spontaneously broken.
In an actual calculation, when atoms are moved along the eigenvector of the imaginary mode to lower the energy, the structure falls to the rim of the hat (the low-symmetry structure at Q = ±Q₀ = ±√(AT₀/B)); recalculating the phonons there, the imaginary mode has disappeared, and
(33)
an amplitude mode with a real frequency appears. This is exactly the same calculation as the mass term 2aH² of the Higgs mode in §4.2 (the curvature of the potential, ∂²/∂H², is −4a): inserting a ↔ ½A(T − T₀) gives −4a = 2A(T₀ − T), which agrees even in the coefficients. The correspondence is summarized in the following table.
| Superconductivity (GL) | Lattice (Landau, soft mode) |
| Order parameter ψ = |ψ|eiθ (complex, U(1)) | Amplitude Q of the soft mode (real: Z₂) or two components (Q₁, Q₂) (U(1)-like) |
| a = a₀(T − Tc) | ½A(T − T₀) = ½Mωsoft² (harmonic approximation of the high-symmetry structure) |
| a < 0: ψ = 0 is unstable (the center of the hat) | ω² < 0: phonon with imaginary frequency (negative frequency in DFT) |
| |ψ|₀ = √(−a/b) ∝ (Tc − T)1/2 | Q₀ = √(A(T₀ − T)/B): distorted low-symmetry structure |
| Condensation energy −a²/2b = −μ₀Hc²/2 | Energy gain from the distortion −A²(T₀ − T)²/4B (ΔE in DFT) |
| Higgs mode ωH² ∝ −2a (2Δ microscopically) | Amplitude mode MωA² = 2A(T₀ − T): fully symmetric (A₁) phonon of the low-symmetry phase, Raman active |
| Nambu–Goldstone mode (phase θ, ω = 0) | Z₂: absent (the minimum is at two points) / U(1): phason, Goldstone phonon (pseudo-Goldstone because of anisotropy) |
| Anderson–Higgs mechanism (the phase is absorbed by the electromagnetic field and becomes ωp) | There is no corresponding gauge field. However, a phason acquires a mass through impurity pinning and commensurability |
| The Higgs mode is not seen in linear response (A²H coupling) | The A₁ mode is Raman active (E²Q coupling = the same second-order coupling) and is seen directly |
| |ψ|² carries a condensation energy of neV–µeV per pair | The distortion energy is meV–100 meV per atom |
In this sense, the spontaneous symmetry breaking of superconductivity and the “phonon with imaginary frequency” are essentially the same phenomenon: in both, the curvature of the energy surface becomes negative in the symmetric state (Landau’s a < 0), and the system falls into a state with a finite order parameter. The Higgs mode is the oscillation “perpendicular to the rim” about the new equilibrium point: in a superconductor it is the oscillation of the Cooper-pair amplitude at 2Δ, and in the lattice it is the A₁ phonon of the low-symmetry phase.
The chain “atoms → crystal → phonons → superconductivity” that Nambu cited in his lecture expresses this correspondence, including the fact that the Goldstone mode of crystallization (breaking of translational symmetry) is the acoustic phonon.
9.3 Where they differ — Z₂ or U(1), and the gauge field
- Type of symmetry: The superconducting ψ is a complex number, so the symmetry that is broken is the continuous U(1), and a Nambu–Goldstone mode of the phase necessarily appears. On the other hand, a structural phase transition driven by a single real order parameter Q (for example, octahedral rotation in a perovskite, ferroelectric distortion, or Peierls dimerization) is the breaking of the discrete symmetry Z₂ (or a larger discrete group) of Q → −Q, and the minimum is not the rim of the hat but isolated points.
Therefore there is an amplitude mode (the counterpart of the Higgs mode) but no Goldstone mode. A Goldstone phonon appears only in the special case where the order parameter belongs to an irreducible representation of dimension two or more and the low-order Landau expansion is rotationally symmetric (a U(1)-like Mexican hat) (§10).
- Gauge field: The lattice has no gauge field corresponding to the electromagnetic field, so there is no Anderson–Higgs mechanism, and in the U(1)-like case the Goldstone phonon is truly gapless (apart from the anisotropy due to the discreteness of the crystal). The phason of a charge density wave acquires a mass by being pinned by impurities, but this is due not to a gauge field but to the explicit breaking of translational symmetry.
- Weighting by mass: The Higgs/Goldstone modes are eigenvectors of the curvature of the free energy (the force-constant matrix), whereas phonons are eigenvectors of the dynamical matrix Φij/√(MiMj). Unless all the atomic masses are the same, the two do not coincide, and the Higgs/Goldstone modes are distributed over groups of phonons of the same irreducible representation (Meier et al. 2020).
- Microscopic nature: The gap 2Δ of the Higgs mode of a superconductor coincides with the lower edge of the quasiparticle continuum, and it decays into quasiparticles (§4.4, §7.3). The amplitude mode of a lattice is not necessarily buried in another continuum, and can often be observed as an isolated, sharp Raman peak. Conversely, in a lattice the quartic term B and the sixth-order anisotropy are large, and at high temperature the transition may be not of the displacive type but of the order–disorder type (the Q at each site has already fallen to ±Q₀ and the domains fluctuate), in which case the soft mode is not seen.
Procedure for treating phonons with imaginary frequencies in DFT (the correspondence in practice)
(1) Calculate the phonons of the high-symmetry structure and identify the imaginary modes and their irreducible representations (symmetry) (the curvature at the center of the hat, and the symmetry of the order parameter).
(2) Distort the structure along the eigenvector of the imaginary mode, calculate the energy E(Q), and fit it to a double well (or a Mexican hat) to obtain the Landau coefficients A, B (and c′) (Meier et al. obtained a = −0.82 eV Å−2 and b = 1.13 eV Å−4 for InMnO₃, and −3.83 and 6.20 for ErMnO₃).
(3) Recalculating the phonons at the minimum structure, the amplitude mode (A₁) appears with a real frequency (Higgs). For a two-component order parameter, the phase mode is also identified (Goldstone).
(4) Determining the space group of the low-symmetry phase from the irreducible representation (the isotropy subgroup) is precisely the group-theoretical problem that CrystOD deals with.
▶ Simulator ④: Double well, Mexican hat, and the temperature dependence of the squared frequency; amplitudon/phason of a one-dimensional chain
10. Literature Survey — Do Phonons Also Have a Higgs Mode and a Nambu–Goldstone Mode?
To state the conclusion first: yes, they do. However, a Goldstone phonon appears only when the order parameter has a continuous symmetry (in practice, an approximate continuous symmetry within a discrete group), and a Higgs phonon is defined as an eigenmode of the curvature of the free energy, so it does not correspond one-to-one to a single phonon. Below, the main literature is organized chronologically.
10.1 The amplitudon and phason of charge density waves (1974–)
The oldest and best established example is the charge density wave (CDW). From the Fröhlich Hamiltonian, Lee–Rice–Anderson (1974) derived, for the order parameter Δeiφ of an incommensurate CDW, an oscillation of the amplitude, the amplitudon (gap ωA ≈ √λ ω2kF, smaller than the CDW gap 2Δ), and a sliding of the phase, the phason (gapless, with sound velocity ∝ vF√(m/m*)). These are systematically summarized in Grüner’s review (Rev. Mod. Phys. 60, 1129 (1988)).
A phason is pinned by impurities or commensurability and acquires a finite mass, and the CDW slides (nonlinear conduction) in an electric field that overcomes that mass.
Littlewood and Varma (1981, 1982) explained a new peak that appeared together with the superconducting transition in the Raman scattering of 2H-NbSe₂ (Sooryakumar–Klein 1980) as the coupling of the amplitudon of the CDW to the Higgs mode of the superconductivity — this is regarded as the first observation of a Higgs mode in condensed matter (review by Pekker–Varma 2015).
As a recent theory, Yang–Chen (arXiv:2505.05025, 2025) constructed a self-consistent phase-transition theory that incorporates phase fluctuations of the CDW, and showed that in (TaSe₄)₂I the phason softens with increasing temperature and is depinned (becomes gapless) at Td ≈ 160 K, that a first-order transition occurs at Tc ≈ 268 K, that 2Δ/kBTc ≈ 17 deviates greatly from the BCS value 3.53, and that the damping of the amplitudon (0.23 THz) changes from light to heavy at Td.
Unlike in superconductors, there is no particle–hole symmetry, so the amplitudon and the phason couple (corresponding to the c₁ term of §4.2), and this is the origin of the damping.
In the rare-earth tritellurides RTe₃, the CDW amplitudon shows twofold symmetry in polarized Raman scattering and is called the “axial Higgs mode” (Wang et al., Nature 606, 896 (2022)); Wulferding et al. (arXiv:2411.08331; with Takami Tohyama of Tokyo University of Science as a coauthor) showed in GdTe₃ and LaTe₃ that a magnetic field perpendicular to the plane increases the axial component of the Higgs mode (the antisymmetric off-diagonal component of the Raman tensor) in proportion to the field and that the phase changes by 90° when the field is reversed, and concluded that the origin of the axiality is not spin but orbital angular momentum (the px ± ipy hybridization of Te).
10.2 Goldstone phonons in structural phase transitions (2005–)
- Cd₂Re₂O₇ (pyrochlore; Kendziora et al., PRL 95, 125503 (2005)): The cubic-to-tetragonal transition at 200 K is caused by an instability of the doubly degenerate Γ₃⁻ (Eu) phonon (Sergienko–Curnoe 2003), and since the anisotropy terms of the Landau expansion appear only at sixth order, it becomes a Mexican hat. In polarized Raman scattering a zero-frequency peak of B₁ symmetry appears below the transition with divergent intensity, and it was taken to be the first Goldstone phonon in a crystal. Later a reinterpretation as an improper displacive transition was proposed from time-resolved coherent-phonon spectroscopy (Harter et al. 2018), and the existence of a Goldstone mode has been supported by machine-learning analysis of diffuse scattering (Venderley et al. 2022).
- PbSr₂Ti₂O₇ (Ruddlesden–Popper; Nakhmanson–Naumov, PRL 104, 097601 (2010)): A first-principles calculation predicted a Goldstone-like state in which the polarization direction is degenerate within the plane.
- (111)-strained SrMnO₃ (Marthinsen et al., Phys. Rev. Materials 2, 014404 (2018)): It was predicted by first-principles calculations that under compressive strain two in-plane rotational instabilities couple to form a Mexican-hat-shaped energy surface, so that a Goldstone-like phonon mode appears. The authors point out that Goldstone modes are known in particle physics, superfluid He, superconductors, and magnons, but that structural Goldstone modes are rare.
- Others that have been reported include the extremely low thermal conductivity due to the collective Goldstone mode of the filled skutterudite SnFe₄Sb₁₂ (Fu et al. 2018), Goldstone-like and Higgs-like structural modes of the relaxor ferroelectric PbMg1/3Nb2/3O₃ (Prosandeev et al. 2020), and hexagonal BaMnO₃ (Zhang–Ye–Li 2021).
10.3 Hexagonal manganites — a “textbook” structural Higgs mode (Meier et al. 2020)
Meier, Stucky, Teyssier, Griffin, van der Marel, and Spaldin, Phys. Rev. B 102, 014102 (2020), is the paper that answers the question of this survey most directly.
- System: The P6₃/mmc → P6₃cm transition of the hexagonal manganites RMnO₃ is caused by a trimerizing tilt of the MnO₅ trigonal bipyramids (a two-dimensional order parameter: the magnitude Q and the angle θ of the tilt), and for small Q the energy does not depend on θ — unusually for a crystal, this is a U(1)-symmetric Mexican hat (Fennie–Rabe 2005; Artyukhin et al. 2014). The discreteness creates six dips on the rim (Z₆) through a sixth-order c′cos 6θ term and a coupling −gQ³Pcos 3θ to the secondary order parameter (the polarization P).
- Theory: After showing within Landau theory that the curvature of the Higgs mode is −2a(T) (the same as Eq. (33) of this explanation) and that of the Goldstone mode is 0, the authors point out that in a multi-element crystal the Higgs/Goldstone modes are eigenvectors of the force-constant matrix and do not coincide with the eigenvectors (phonons) of the dynamical matrix divided by the masses. Only the Goldstone mode (ω = 0) corresponds to a single phonon independently of the masses, whereas the Higgs mode is distributed over a group of phonons of the same irreducible representation (A₁). From the DFT Landau coefficients (InMnO₃: a = −0.82 eV Å−2, b = 1.13 eV Å−4, c′ = 0.04 eV Å−6; ErMnO₃: −3.83, 6.20, 0.06), the height of the hat is ~500 meV for ErMnO₃ and ~170 meV for InMnO₃, and the dips on the rim are ~200 meV and ~50 meV, respectively.
- Identification of the modes: The lowest A₁ phonon (InMnO₃ 128 cm−1, ErMnO₃ 130 cm−1; 96 and 106 in Landau theory) has the strongest Higgs character, and the lowest B₁ phonon (InMnO₃ 65 cm−1, ErMnO₃ 107 cm−1; 21 and 82 in Landau theory) has the strongest Goldstone character. That the Goldstone-like mode is low in InMnO₃, where the dips on the rim are small, is consistent with the gap ∝ √c′ of the pseudo-Goldstone mode in Simulator ④.
- Experiment: In temperature-dependent Raman scattering (4–300 K), the lowest A₁ mode of InMnO₃ (TC ≈ 500 K) softens by ~20 cm−1 by 300 K (Landau theory predicts ~30 cm−1), which cannot be explained by the effect of thermal expansion calculated in the quasi-harmonic approximation; the authors therefore concluded that this is the first experimental identification of the Higgs mode of a structural phase transition. In ErMnO₃ (TC ≈ 1200 K) the effect of thermal expansion dominates up to 300 K.
As a sequel to this system, Juraschek–Meier–Narang (PRL 124, 117401 (2020)) proposed a method of indirectly exciting an optically silent Goldstone-like phonon, through the nonlinear coupling, by resonantly driving an infrared-active Higgs-like phonon with a terahertz pulse, and Rangwala–Ganesan (arXiv:2602.07462, 2026) numerically investigated, with their nonlinear phononics model (Higgs 4.0 THz, Goldstone 2.4 THz), the conditions for generating a phonon frequency comb through the Higgs–Goldstone coupling.
Meier et al. (arXiv:2107.11332, 2021) further showed by Landau theory, for a structural phase transition involving several irreducible representations (the pyrochlore Cd₂Nb₂O₇), the existence of an “anti-phase Goldstone mode (anti-phason)” corresponding to the Leggett mode of multiband superconductors (an oscillation of the relative phase of two order parameters), and of a new “anti-phase Higgs mode.”
10.4 A field-theoretic summary, and the conclusions of this survey
Vallone (Phys. Status Solidi B 257, 1900443 (2020)) treats the Higgs/Goldstone phonons of crystals as a field theory, and Pekker–Varma (Annu. Rev. Condens. Matter Phys. 6, 269 (2015)) is a review of amplitude (Higgs) modes across superconductors, CDWs, magnets, and cold atoms.
To summarize:
- The GL free energy of a superconductor and the Landau free energy of a structural phase transition are of the same φ⁴ type, and the phonon with imaginary frequency of DFT is precisely the Landau coefficient a < 0 (the negative curvature at the center of the hat) (§9).
- The amplitude mode (A₁ phonon) of the low-symmetry phase is the lattice version of the Higgs mode, and since it is Raman active it can be seen directly. It was identified by temperature-dependent Raman scattering in hexagonal InMnO₃ (Meier et al. 2020).
- A Goldstone phonon appears only when the order parameter has two or more components and the low-order Landau expansion is rotationally symmetric: the CDW phason (1974) is the prototype, Cd₂Re₂O₇ (2005) was the first in crystal structures, and it has spread to hexagonal manganites, strained SrMnO₃, pyrochlore Cd₂Nb₂O₇, and others. It has a small gap (pseudo-Goldstone) because of the sixth-order anisotropy due to the discreteness of the crystal.
- Since a lattice has no gauge field, there is no Anderson–Higgs mechanism (conversion to a plasmon). Instead, a phason acquires a mass through impurity pinning.
- In multi-element crystals, the Higgs/Goldstone modes (eigenmodes of the force-constant matrix) and single phonons (eigenmodes of the dynamical matrix) do not coincide and are distributed over groups of phonons of the same irreducible representation — a “Higgs phonon” is, strictly speaking, a “phonon with a strong Higgs character.”
11. Implementation Notes
- Units and constants: SI (CODATA 2022). The default is m* = 2me and e* = 2e for Cooper pairs, and the sliders can vary m*/me and e*/e independently (e* = 0 gives a neutral superfluid). The GL coefficients are determined from measured values by Eqs. (7)–(8) of §3.1, and a(T) is linear down to T = 0. The empirical formula Hc(T) = Hc(0)(1 − t²) has a slope near Tc twice that of the linear model, so it is overlaid for reference in the right-hand figure.
- Drawing the Mexican hat: f̃ = 2ãr̃² + (b/b₀)r̃⁴ (ã = aeff/|a(0)|, r̃ = |ψ|/ψ₀(0), unit of height fc = a(0)²/2b) is drawn on a polar mesh (20 × 40), with hidden surfaces removed by the painter’s algorithm, painting from back to front. The U(1) hat of ④ has the anisotropy c′r̃⁶(1 − cos 6θ)/2 added. The apparent frequency of the Higgs oscillation of the ball is made proportional to √(−2ã), expressing critical slowing down.
- BCS gap equation: Eq. (23) is put in the dimensionless form with ħωD = 1, namely 1/N(0)V = ∫₀¹ dx tanh(√(x² + d²)/2τ)/√(x² + d²), integrated by Simpson’s rule (400 subdivisions), and d = Δ/ħωD and τc = kBTc/ħωD are solved by the bisection method. The interpolation formula Δ(0)tanh(1.74√(Tc/T − 1)) for Δ(T) is used in ① and ③.
- Pseudospin quench: The linearized Eq. (28) is discretized on a nonuniform grid ε = Δ sinh u (301 points, |ε| ≤ ħωD), and, using N(0)V = 1/Σwi/2Ei, which satisfies the gap equation exactly as a discrete sum, it is integrated by RK4 (dt = 0.02/Δ, Δt ≤ 40) together with the self-consistency condition δΔ = ΔδV/V + N(0)VΣwiδσxi. The asymptotic form (30) is displayed overlaid.
- Inverting the McMillan formula: λ = (1.04 + Lμ*)/(L(1 − 0.62μ*) − 1.04), L = ln(θD/1.45Tc). The BCS inversion is N(0)V = 1/ln(1.134 θD/Tc).
- The excitation spectrum of ③: In units of the wave number q₀ = 2√3Δ(0)/ħvF, the Higgs dispersion (19) is ω/2Δ(0) = √(Δ̃² + q̃²) and the neutral Nambu–Goldstone mode is ω = vFq/√3. vF is either the Fermi-surface average of Kawamura et al. (2026) (Al, In, Sn, Pb, Nb, Ta) or the free-electron value of Ashcroft–Mermin, and ωp is from the free-electron density (a typical value of 10 eV for V and Ta). The resonance curves of the third harmonic are schematic plots of the GL |2Ω − ωH|−1 and the BCS |2Ω − 2Δ|−1/2.
- The lattice model of ④: Ẽ = 2ãq² + q⁴, ω̃² = Ẽ″/4 (in units of ω₀² = AT₀/M). Classically ã = T/T₀ − 1, and including the quantum fluctuations T₁/T₀ = θ gives Barrett’s formula ã = (θ/2)coth(θ/2τ) − 1 (which returns to the classical case as θ → 0). The harmonic approximation of the high-symmetry structure is ω̃² = ã (negative for T < Tc), the amplitude mode is ω̃A² = −2ã, and the pseudo-Goldstone mode is ω̃G² = 4.5c′(−ã)² (Q₀² ∝ −ã). Since the vertical axis of the right-hand figure is ω²/ω₀², the soft mode and the harmonic approximation of the high-symmetry structure appear as a single straight line passing through the transition point (Cochran’s ω² ∝ T − T₀), and there is no need to draw an unstable mode on the imaginary axis. The transition temperature is Tc/T₀ = θ/ln((2 + θ)/(2 − θ)), with no transition for θ ≥ 2. The one-dimensional chain has dimerization (−1)nu for Z₂ and an incommensurate modulation ucos(Qna + θ) for U(1) (Qa = 2π/4.3).
- Conventions of the page: Equations are set in the browser’s standard MathML, with no external libraries or network communication.
?embed=1 embeds the page, ?only=gl|bcs|modes|phonon shows a single tab, #bcs and so on select a tab, and the height is posted to the parent page by postMessage.
12. References
- Naoto Tsuji, Physics of Nonequilibrium Superconductivity (in Japanese), lecture text of the Summer School for Young Condensed-Matter Physicists (Bussei Wakate Natsu no Gakko), PDF/5_69th_lecture_E.pdf. §2 and §4–§7 of this explanation follow its Chapters 2 and 3.
- V. L. Ginzburg and L. D. Landau, Zh. Eksp. Teor. Fiz. 20, 1064 (1950).
- J. Bardeen, L. N. Cooper, and J. R. Schrieffer, Theory of Superconductivity, Phys. Rev. 108, 1175 (1957).
- L. P. Gor’kov, Microscopic Derivation of the Ginzburg–Landau Equations in the Theory of Superconductivity, Sov. Phys. JETP 9, 1364 (1959).
- P. W. Anderson, Random-Phase Approximation in the Theory of Superconductivity, Phys. Rev. 112, 1900 (1958). — Higgs mode (collective oscillation of the amplitude, 2Δ) and pseudospin.
- P. W. Anderson, Plasmons, Gauge Invariance, and Mass, Phys. Rev. 130, 439 (1963). — The Anderson–Higgs mechanism.
- Y. Nambu, Quasi-Particles and Gauge Invariance in the Theory of Superconductivity, Phys. Rev. 117, 648 (1960).
- Y. Nambu and G. Jona-Lasinio, Dynamical Model of Elementary Particles Based on an Analogy with Superconductivity. I, Phys. Rev. 122, 345 (1961).
- Y. Nambu, Nobel Lecture: Spontaneous symmetry breaking in particle physics: A case of cross fertilization, Rev. Mod. Phys. 81, 1015 (2009). — The history in §8 is based on this lecture.
- J. Goldstone, Field theories with «Superconductor» solutions, Il Nuovo Cimento 19, 154 (1961); J. Goldstone, A. Salam, and S. Weinberg, Broken Symmetries, Phys. Rev. 127, 965 (1962).
- P. W. Higgs, Phys. Rev. Lett. 13, 508 (1964); F. Englert and R. Brout, Phys. Rev. Lett. 13, 321 (1964); G. S. Guralnik, C. R. Hagen, and T. W. B. Kibble, Phys. Rev. Lett. 13, 585 (1964).
- N. N. Bogoliubov, V. V. Tolmachev, and D. V. Shirkov, A New Method in the Theory of Superconductivity (Academy of Sciences of USSR, 1958).
- R. V. Carlson and A. M. Goldman, Phys. Rev. Lett. 34, 11 (1975). — The Carlson–Goldman mode.
- P. B. Littlewood and C. M. Varma, Gauge-Invariant Theory of the Dynamical Interaction of Charge Density Waves and Superconductivity, Phys. Rev. Lett. 47, 811 (1981); Amplitude collective modes in superconductors and their coupling to charge-density waves, Phys. Rev. B 26, 4883 (1982).
- R. Sooryakumar and M. V. Klein, Raman Scattering by Superconducting-Gap Excitations and Their Coupling to Charge-Density Waves, Phys. Rev. Lett. 45, 660 (1980).
- D. Pekker and C. M. Varma, Amplitude/Higgs Modes in Condensed Matter Physics, Annu. Rev. Condens. Matter Phys. 6, 269 (2015).
- R. Shimano and N. Tsuji, Higgs Mode in Superconductors, Annu. Rev. Condens. Matter Phys. 11, 103 (2020).
- R. Matsunaga, Y. I. Hamada, K. Makise, Y. Uzawa, H. Terai, Z. Wang, and R. Shimano, Higgs Amplitude Mode in the BCS Superconductors Nb1−xTixN Induced by Terahertz Pulse Excitation, Phys. Rev. Lett. 111, 057002 (2013).
- R. Matsunaga, N. Tsuji, H. Fujita, A. Sugioka, K. Makise, Y. Uzawa, H. Terai, Z. Wang, H. Aoki, and R. Shimano, Light-induced collective pseudospin precession resonating with Higgs mode in a superconductor, Science 345, 1145 (2014).
- N. Tsuji and H. Aoki, Theory of Anderson pseudospin resonance with Higgs mode in superconductors, Phys. Rev. B 92, 064508 (2015).
- T. Cea, C. Castellani, and L. Benfatto, Nonlinear optical effects and third-harmonic generation in superconductors: Cooper pairs versus Higgs mode contribution, Phys. Rev. B 93, 180507 (2016); R. Matsunaga, N. Tsuji, K. Makise, H. Terai, H. Aoki, and R. Shimano, Phys. Rev. B 96, 020505 (2017); N. Tsuji and Y. Nomura, Higgs-mode resonance in third harmonic generation in NbN superconductors: Multiband electron-phonon coupling, impurity scattering, and polarization-angle dependence, Phys. Rev. Res. 2, 043029 (2020).
- A. F. Volkov and S. M. Kogan, Collisionless relaxation of the energy gap in superconductors, Sov. Phys. JETP 38, 1018 (1974); R. A. Barankov, L. S. Levitov, and B. Z. Spivak, Phys. Rev. Lett. 93, 160401 (2004); E. A. Yuzbashyan, B. L. Altshuler, V. B. Kuznetsov, and V. Z. Enolskii, Phys. Rev. B 72, 220503 (2005).
- W. L. McMillan, Transition Temperature of Strong-Coupled Superconductors, Phys. Rev. 167, 331 (1968); P. B. Allen and R. C. Dynes, Phys. Rev. B 12, 905 (1975).
- C. Kittel, Introduction to Solid State Physics, 7th/8th ed. (Wiley), tables in Chap. 10/12 (Tc, Hc(0), N(0)V, 2Δ/kBTc, ξ₀ and λL (Meservey–Schwartz)).
- M. Tinkham, Introduction to Superconductivity, 2nd ed. (McGraw-Hill, 1996). — The standard textbook on GL theory.
- G. A. Webb, F. Marsiglio, and J. E. Hirsch, Superconductivity in the elements, alloys and simple compounds, Physica C 514, 17 (2015), arXiv:1502.04724. — Table III (Tc, θD).
- M. Kawamura, T. Nomoto, N. Witt, and R. Arita, First-principles calculation of coherence length and penetration depth based on density functional theory for superconductors, arXiv:2603.05123 (2026). — Tables I and S2 (compilation of the measured values of ξ₀ and λL, vF, λep, and the depairing current).
- A. K. Raychaudhuri, C. Egloff, and L. Rinderer, J. Low Temp. Phys. 59, 413 (1985) (λ of In); J. Auer and H. Ullmaier, Phys. Rev. B 7, 136 (1973), J. P. McEvoy, D. P. Jones, and J. G. Park, Phys. Rev. Lett. 22, 229 (1969) (Ta); R. Radebaugh and P. H. Keesom, Phys. Rev. 149, 217 (1966) (κ of V).
- N. W. Ashcroft and N. D. Mermin, Solid State Physics (Saunders, 1976), Tables 1.1, 2.1 (n, vF).
- W. Cochran, Crystal stability and the theory of ferroelectricity, Adv. Phys. 9, 387 (1960); J. F. Scott, Soft-mode spectroscopy, Rev. Mod. Phys. 46, 83 (1974); R. A. Cowley, Structural phase transitions I. Landau theory, Adv. Phys. 29, 1 (1980); M. T. Dove, Introduction to Lattice Dynamics (Cambridge, 1993); J. H. Barrett, Dielectric Constant in Perovskite Type Crystals, Phys. Rev. 86, 118 (1952); K. A. Müller and H. Burkard, SrTiO₃: An intrinsic quantum paraelectric below 4 K, Phys. Rev. B 19, 3593 (1979); T. Nagai, Y. Mochizuki, H. Shirakuni, A. Nakano, F. Oba, I. Terasaki, and H. Taniguchi, Phase Transition from Weak Ferroelectricity to Incipient Ferroelectricity in Li₂Sr(Nb1−xTax)₂O₇, Chem. Mater. 32, 744 (2020).
- A. Togo and I. Tanaka, First principles phonon calculations in materials science, Scr. Mater. 108, 1 (2015) (phonopy); S. Baroni, S. de Gironcoli, A. Dal Corso, and P. Giannozzi, Rev. Mod. Phys. 73, 515 (2001) (DFPT); H. T. Stokes and D. M. Hatch, ISOTROPY software suite; B. J. Campbell et al., J. Appl. Cryst. 39, 607 (2006) (ISODISPLACE).
- P. A. Lee, T. M. Rice, and P. W. Anderson, Conductivity from charge or spin density waves, Solid State Commun. 14, 703 (1974); G. Grüner, The dynamics of charge-density waves, Rev. Mod. Phys. 60, 1129 (1988).
- C. A. Kendziora, I. A. Sergienko, R. Jin, J. He, V. Keppens, B. C. Sales, and D. Mandrus, Goldstone-Mode Phonon Dynamics in the Pyrochlore Cd₂Re₂O₇, Phys. Rev. Lett. 95, 125503 (2005); I. A. Sergienko and S. H. Curnoe, J. Phys. Soc. Jpn. 72, 1607 (2003); J. W. Harter et al., Phys. Rev. Lett. 120, 047601 (2018).
- S. M. Nakhmanson and I. Naumov, Goldstone-like States in a Layered Perovskite with Frustrated Polarization: A First-Principles Investigation of PbSr₂Ti₂O₇, Phys. Rev. Lett. 104, 097601 (2010).
- A. Marthinsen, S. M. Griffin, M. Moreau, T. Grande, T. Tybell, and S. M. Selbach, Goldstone-like phonon modes in a (111)-strained perovskite, Phys. Rev. Materials 2, 014404 (2018).
- Q. N. Meier, A. Stucky, J. Teyssier, S. M. Griffin, D. van der Marel, and N. A. Spaldin, Manifestation of structural Higgs and Goldstone modes in the hexagonal manganites, Phys. Rev. B 102, 014102 (2020), arXiv:1912.04942.
- C. J. Fennie and K. M. Rabe, Phys. Rev. B 72, 100103(R) (2005); S. Artyukhin, K. T. Delaney, N. A. Spaldin, and M. Mostovoy, Nat. Mater. 13, 42 (2014). — Landau theory of the hexagonal manganites.
- D. M. Juraschek, Q. N. Meier, and P. Narang, Parametric Excitation of an Optically Silent Goldstone-Like Phonon Mode, Phys. Rev. Lett. 124, 117401 (2020).
- Q. N. Meier, D. Hickox-Young, G. Laurita, N. A. Spaldin, J. M. Rondinelli, and M. R. Norman, Leggett Modes Accompanying Crystallographic Phase Transitions, arXiv:2107.11332 (2021/2022).
- M. Vallone, Higgs and Goldstone modes in crystalline solids, Phys. Status Solidi B 257, 1900443 (2020).
- Y. Fu, X. He, L. Zhang, and D. J. Singh, Phys. Rev. B 97, 024301 (2018); S. Prosandeev et al., Phys. Rev. B 102, 104110 (2020); X. Zhang, Q.-J. Ye, and X.-Z. Li, Phys. Rev. B 103, 024101 (2021).
- Y. Wang et al., Axial Higgs mode detected by quantum pathway interference in RTe₃, Nature 606, 896 (2022); D. Wulferding, J. Park, T. Tohyama, S. R. Park, and C. Kim, Magnetic field control over the axialness of Higgs modes in charge-density wave compounds, arXiv:2411.08331 (2024), Nature Communications (2025).
- F. Yang and L. Q. Chen, Microscopic phase-transition theory of charge density waves: revealing hidden transitions of phason and amplitudon, arXiv:2505.05025 (2025).
- M. Rangwala and A. Ganesan, Spontaneous Symmetry Breaking and Collective Higgs–Goldstone Dynamics in Solid-State Phononic Frequency Combs, arXiv:2602.07462 (2026).