What Is the Global Dynamics of the 1D Quartic Klein-Gordon Equation With an Internal Mode?

An internal mode creates linear bound states

The model is ∂²ₜu − ∂²ₓu + u + V(x)u = β(x)u² + u⁴ on the line, with a fast-decaying potential V and a localized quadratic coefficient β. The linearized operator H + 1 = −∂²ₓ + 1 + V(x) is assumed to have purely absolutely continuous spectrum [1, ∞) plus exactly one positive discrete eigenvalue λ², with 0 < λ² < 1, and a normalized exponentially localized eigenfunction φ.

That single eigenvalue is what the paper calls an internal mode. It produces spatially localized, time-periodic solutions of the linear equation, φ_{A,θ}(t,x) = A cos(λt + θ) φ(x), known as bound states. The question the paper answers is whether these small bound states survive once the quadratic and quartic nonlinear terms are switched on.

The paper also assumes V is generic: no resonance at the edge of the continuous spectrum, meaning there is no nonzero bounded ζ with Hζ = 0. This is a spectral non-degeneracy condition, not a statement about the size of the nonlinearity.

The Fermi Golden Rule drives radiation damping

The mechanism that decides the fate of the internal mode is a nonlinear resonance condition, the Fermi Golden Rule. In the paper's decomposition, the amplitude A(t) of the discrete component obeys an evolution equation whose leading nonlinear correction is (1/2iλ)(Λ − iΓ)|A|²A, where Λ and Γ are real constants built from θ = P_c(βφ²).

The constant Γ is the damping coefficient: Γ = (π/2λ)(θ, δ(L − 2λ)θ), a spectral measure evaluated at the resonant frequency 2λ. When this term is present, the internal mode loses energy to the continuous spectrum — the radiation field — and its amplitude decays. The paper states that the Fermi Golden Rule gives the nonlinear damping term −(Γ/2λ)|A|²A.

This is the paper's central causal chain: the discrete eigenvalue couples to the continuous spectrum at frequency 2λ, the coupling is non-degenerate, and the result is radiation damping rather than a persistent oscillation. The assumption is stated explicitly rather than derived, so the decay conclusion is conditional on it.

Amplitude decays while radiation scatters

Under the spectral assumptions and the Fermi Golden Rule, the main theorem gives a global description for sufficiently small real initial data. The internal-mode amplitude satisfies |A(t)| = (1/2λ)√(ȧ²(t) + λ²a²(t)) ≈ ε₀(1 + ε₀²|t|)^(-1/2), so the discrete component decays slowly, at the rate t^(-1/2) for large time.

The radiation component v is controlled in L∞ by ⟨t⟩^(-1/2+δ)‖v(t)‖_{L∞} ≲ ε₀^α, with a sharper logarithmic bound ‖v(t)‖_{L∞} ≲ log(2 + ε₀²⟨t⟩)/⟨t⟩^(1/2) for t ≳ ε₀^(-2). Its Sobolev norm stays bounded: ‖v(t)‖_{H^N} ≲ ε₀.

The radiation also scatters: there exists Φ₊ in P_c H^{N-1} such that ‖(∂ₜ + iL)v − e^{itL}Φ₊‖_{H^{N-1}} ≲ ε₀(1 + ε₀²t)^(-σ) for some fixed σ > 0, with the analogous statement as t → −∞ by time reversal. Scattering here means the radiation asymptotically follows the linear Klein-Gordon flow with the potential, not that it vanishes.

The amplitude follows a logarithmic phase law

Beyond the decay rate, the paper gives the asymptotic form of the internal-mode amplitude. There is a real phase ψ∞ depending on the initial data and constants c₂, β₁ depending only on the equation such that A(t) = exp[i(ψ∞ − (c₂ − β₁)(λ/Γ) log(1 + (Γ/λ)ρ₀t))] √(λ/Γ)t + O(t^(-1/2-γ)) for t ≫ ε₀^(-2), for some small γ > 0, where ρ₀ = |A(0)|².

Two features matter for reading this correctly. First, the decay is algebraic, not exponential: the internal mode is metastable, losing energy gradually through the resonant coupling. Second, the phase carries a logarithmic correction whose coefficient involves λ/Γ, so the damping constant Γ controls not only how fast the amplitude shrinks but also how the oscillation phase drifts.

The remainder is understood as t → ∞ for fixed data, and the constants c₂ and β₁ are given exactly in the paper's Theorem 9.3 rather than in the summary statement.

Distorted Fourier transform handles the low-degree term

The technical obstacle is that the linear solutions decay only slowly, at the pointwise rate t^(-1/2), and the coupling to the discrete spectrum acts as a singular forcing. The paper's approach is built on the distorted Fourier transform together with refined dispersive decay and smoothing estimates, including local decay and smoothing-type bounds used to close the bootstrap.

The paper states that a quartic, non-localized interaction is essentially sharp for this analysis: closing the estimates with such a low-degree nonlinear term requires careful and delicate treatment. The authors note that adding a localized cubic term β₂(x)u³, or quintic and higher terms u^p with p ≥ 5, does not bring further complications, which is why the model is written with a localized quadratic term and a quartic term for concreteness.

The related cubic problem — global-in-space dispersive asymptotics for perturbations of kink solutions in the classical φ⁴ model — is described as remaining a challenging open question. The quartic model is presented as an intermediate step that illustrates the difficulties of combining an internal mode with a low pure power nonlinearity.

Kink asymptotic stability follows as an application

As a direct application, the paper proves asymptotic stability of kinks for classes of scalar field models that can be seen as perturbations of the φ⁴ model. In the application, the wave equation has an odd kink solution K(x) with a double-well potential, and the linearized operator satisfies the same spectral conditions.

For odd perturbations φ(t,x) = K(x) + u(t,x), with initial data satisfying |(φ_ω, u₀)| + |(φ_ω, u₁)| = ε₀ and the stated weighted Sobolev bounds, the equation admits a unique global solution. Its discrete component satisfies √(ȧ²(t) + ω²a²(t)) ≈ ε₀(1 + ε₀²t)^(-1/2), its radiation obeys ‖v(t)‖_{L∞} ≲ log(2 + ε₀²t)/⟨t⟩^(1/2) for t ≥ ε₀^(-2), and the field w = ∂ₜv + iLv scatters to a linear Klein-Gordon wave in H^{N-1}.

The conclusion is that the kink is asymptotically stable under odd perturbations on the whole line. The oddness restriction is part of the application's setup: it avoids the zero eigenvalue and the resonance in the spectrum of L, which the paper identifies as substantial additional challenges.

Where the result applies and where it stops

The theorem is a small-data statement: it requires initial data small in a weighted Sobolev space, with |(φ, u₀)| + |(φ, u₁)| = ε₀ and the H^{N-1} and weighted H² bounds of condition (2.22) both at most ε₀, for ε₀ below a sufficiently small threshold. The bootstrap parameters α = 10^(-11), δ = 10^(-10), δ₂ = 10^(-9), δ₁ = 10^(-2) and N = 10⁵ + 1 are explicitly not optimized, so the thresholds are not sharp numerical constants.

The result is also conditional on the spectral assumptions: a single discrete eigenvalue λ² in (0,1), a generic potential with no resonance at the continuous-spectrum edge, and the Fermi Golden Rule. The paper assumes the Fermi Golden Rule rather than proving it for a given potential, so the decay and scattering conclusions hold for the class of models where that non-degeneracy condition is satisfied.

The most decision-relevant distinction is between the two components. The internal-mode amplitude decays only at the slow rate (1 + ε₀²|t|)^(-1/2) and is metastable, while the radiation scatters to the linear flow with an H^{N-1} error decaying like (1 + ε₀²t)^(-σ). Reading the result as fast damping of the discrete mode, or as the radiation disappearing, misstates what the theorem establishes: the bound state does not persist, but the internal mode fades slowly and the radiation survives as a linear scattering state.

Sources used in this answer