Solitary waves, solitons & integrability

A localized wave that travels without changing shape, a pair of waves that scatters elastically, and an integrable evolution equation are three different mathematical claims. The decaying KdV equation supplies examples of all three, but verifying the first does not prove the others. This reference fixes the meanings used in the site’s KdV sequence and explains which calculation supports each claim. It does not impose one definition of integrability on every branch of the subject.

Required background. Recognize a traveling profile $U(x-vt)$ and distinguish an equation from one of its solutions; the ODE bridge develops that distinction. Helpful background. The KdV introduction and collision lesson give explicit examples.

Localized traveling waves

In the zero-background setting on $x\in\mathbb R$, call a nonzero profile a solitary wave when

$$u(x,t)=U(x-vt-a), \qquad U(\xi)\longrightarrow0\quad(\xi\to\pm\infty),$$

and this expression solves the stated evolution equation. Here $v$ is its constant speed and $a$ its position parameter. Required decay and regularity depend on the problem; the KdV profiles below are smooth and exponentially decaying. Periodic waves and waves joining different nonzero limits do not meet this particular zero-background definition. Check the regime before transferring terminology between models.

For our KdV sign convention,

$$u_t+6uu_x+u_{xxx}=0,$$

the solitary profile is

$$u(x,t)=2\kappa^2\operatorname{sech}^2 \!\left[\kappa(x-4\kappa^2t-a)\right], \qquad\kappa\gt0.$$

Its amplitude is $2\kappa^2$, speed is $4\kappa^2$, and width scale is $1/\kappa$. Substituting a traveling wave reduces the PDE to an ODE; the identity $U''=4\kappa^2U-3U^2$ verifies the result after differentiation. The traveling-wave derivation shows how the decay conditions fix the integration constants.

Shape preservation alone is a weaker property than nonlinear soliton scattering. For example, every differentiable localized profile $F$ gives an exact solution $F(x-vt)$ of the linear transport equation $u_t+vu_x=0$. Direct differentiation verifies this for arbitrary $F$. Merely observing an unchanging pulse therefore does not identify a nonlinear interaction law or an inverse-scattering structure.

What makes the KdV pulse a soliton?

Within the decaying KdV scattering problem, soliton has a precise spectral meaning. The field determines the Schrödinger operator $L=-\partial_x^2-u$. Reflectionless data with one negative bound eigenvalue $-\kappa^2$ reconstruct the displayed pulse. Its eigenvalue fixes scale and speed; its positive norming coefficient fixes position. A pure multisoliton solution comes from finitely many such bound levels and zero reflection coefficient. This is the KdV usage in Aktosun 2009, § I and § IX, equations (9.1)–(9.3); his field is $-u$ in the convention used here.

The one-pole reconstruction checks the bound state, absence of reflection, and pulse directly. Thus calling that solution a soliton does not require waiting for a numerical collision experiment. Its membership in the exact scattering family supplies the extra structure.

For two distinct parameters $\kappa_1\gt\kappa_2\gt0$, the exact collision recovers the two amplitudes and speeds as $t\to\pm\infty$. Their asymptotic center lines are

$$x=4\kappa_j^2t+b_j^\pm.$$

The changes in intercept are

$$\begin{aligned} \Delta x_1&=\frac1{\kappa_1} \log\frac{\kappa_1+\kappa_2}{\kappa_1-\kappa_2}\gt0,\\ \Delta x_2&=-\frac1{\kappa_2} \log\frac{\kappa_1+\kappa_2}{\kappa_1-\kappa_2}\lt0. \end{aligned}$$

Here $\Delta x_j=b_j^+-b_j^-$. The two-soliton derivation obtains these limits in the two moving frames. The preserved asymptotic shapes and speeds, together with the shifts, are the elastic-scattering statement. The interaction still changes the trajectories.

A “phase shift” in this real KdV example means a spatial displacement, not multiplication of a complex field by $e^{i\theta}$. The pulses are labeled by their distinct spectral parameters. Two isolated maxima need not remain identifiable while the waves overlap, so following whichever peak is tallest can misidentify a soliton.

Claims that require different evidence

Claim	What must be checked	What this alone does not establish
A solitary wave exists	A localized traveling profile satisfies the equation and its stated boundary conditions.	Collision behavior, stability, or a solution method for general data.
Two KdV solitons scatter elastically	One interacting solution has the stated incoming and outgoing profiles, with their shifts.	Stability against arbitrary perturbations or behavior of every initial profile.
A numerical collision agrees with a formula	The same initial data, equation, domain and norms agree under independent refinements.	An exact continuum identity or a completeness theorem.
A pulse is stable in a stated sense	Small perturbations remain controlled in a specified norm, with any allowed translations stated.	A general inverse-scattering solution of the equation.
A model is integrable in a stated sense	The model-specific structure and its hypotheses are established.	Every stronger meaning of “solvable.”

In particular, two isolated KdV pulses cannot simply be added to obtain an exact collision. If each $u_j$ solves KdV, their traveling sum has residual

$$6\partial_x(u_1u_2).$$

This is not identically zero at finite separation, although it can vanish at particular points. Starting a numerical evolution from that sum specifies different initial data from the exact two-soliton tau profile. Small differences at large separation can be useful approximations, but they must not be renamed an exact benchmark.

Likewise, an exact formula for one or two special solutions is not a stability theorem. Stability quantifies a neighborhood of perturbed initial data, not just the chosen family. A long-time decomposition of general data into solitons and dispersive contributions needs further hypotheses and estimates. Neither claim follows from the two-soliton calculation on this site.

Integrability belongs to a model and a regime

For the decaying KdV initial-value problem, the relevant structure is inverse scattering: construct spectral data from the initial field, evolve those data, and reconstruct the field. The finite-rank examples carry out this program for reflectionless data. A general theorem must also handle the continuous scattering contribution and the analytic conditions that make each step valid; Aktosun, § III and §§ VI–VIII describes that framework. Changing to periodic or nonzero-background data changes the spectral problem.

Finite-dimensional Hamiltonian integrability asks a different question. On a $2n$-dimensional symplectic phase space, Liouville integrability requires $n$ independent conserved functions in involution on the stated regular set. The harmonic oscillator supplies a simple example with one degree of freedom and a nonconstant conserved Hamiltonian away from its equilibrium. Its phase-space motion is integrable without being a localized field on a spatial line. The Liouville–Arnold reference states the additional conditions needed for the torus and action–angle conclusions.

For the finite XXX chain, the transfer-matrix proof instead establishes a commuting quantum operator family containing the Hamiltonian. That proof distinguishes commutativity from independence and Bethe-state completeness. A quantum spin chain need not first exhibit a classical traveling pulse to possess that operator structure.

These examples explain why the words should stay attached to their objects: a solitary wave describes a solution’s spatial and temporal form; a KdV soliton belongs to a particular nonlinear scattering construction; and an integrability claim identifies a structure for an equation or model with specified boundaries and hypotheses.

References

• Aktosun, Tuncay. “Inverse Scattering Transform and the Theory of Solitons.” In Robert A. Meyers (ed.), Encyclopedia of Complexity and Systems Science. Springer, 2009, 4960–4971. DOI. Author version arXiv:0905.4746v1 [nlin.SI], 28 May 2009. Open HTML. Sections I, III and VI–IX support the stated decaying-wave and inverse-scattering usage; they do not supply a universal definition for the other forms of integrability discussed here.