Lax pair

A lax pair is a pair of linear operators whose compatibility condition represents a generally nonlinear evolution equation. For operators (L(t)) and (P(t)) acting on a common function space, the defining relation is

[ \frac{dL}{dt}=[P,L]=PL-LP, ]

where ([P,L]) denotes the commutator. This operator equation is called the Lax equation. A nonlinear differential equation admitting such a representation is often associated with an integrable system, although the existence of a formal Lax representation alone does not settle every analytic or Hamiltonian question connected with integrability.

The principal structural consequence of the Lax equation is the preservation of spectral data. When (L\psi=\lambda\psi) and the auxiliary evolution is (\psi_t=P\psi), compatibility implies (\lambda_t=0) under suitable assumptions on operator domains and boundary conditions. The evolution of (L) is therefore isospectral: its coefficients can change with time while its spectrum remains fixed.

Operator formulation

The Lax equation is the compatibility condition for the overdetermined linear system

[ L\psi=\lambda\psi, \qquad \frac{\partial\psi}{\partial t}=P\psi. ]

Differentiation of the spectral equation gives

[ L_t\psi+L\psi_t=\lambda_t\psi+\lambda\psi_t. ]

After substituting (\psi_t=P\psi) and (L\psi=\lambda\psi), this becomes

[ \bigl(L_t-[P,L]\bigr)\psi=\lambda_t\psi. ]

Consequently, the Lax equation makes the spectral parameter independent of time. In finite dimensions, the same statement follows from a similarity transformation. If an invertible operator (G(t)) satisfies

[ G_t=PG, ]

then

[ L(t)=G(t)L(0)G(t)^{-1} ]

solves (L_t=[P,L]). Thus (L(t)) remains in the conjugacy class of its initial value.

For an (N\times N) matrix (L), the coefficients of the characteristic polynomial are conserved. Equivalent conserved quantities arise from

[ I_k=\operatorname{tr}(L^k), ]

because cyclicity of the trace yields

[ \frac{dI_k}{dt} =k\operatorname{tr}\bigl(L^{k-1}[P,L]\bigr)=0. ]

These invariants provide part of the relation between Lax representations and Liouville integrability. A complete Liouville-integrable Hamiltonian system additionally requires sufficiently many functionally independent integrals that commute under the relevant Poisson bracket. The trace invariants of a particular Lax matrix need not automatically satisfy all of those conditions.

For differential operators, the spectral statement depends on the chosen domain and on the behavior imposed at spatial boundaries. Periodic coefficients lead naturally to Floquet theory, whereas rapidly decaying coefficients lead to scattering data on an unbounded domain. The discrete and continuous portions of the spectrum then encode different parts of the associated nonlinear dynamics.

Historical development

The operator formulation was introduced by Peter Lax in 1968 as an abstraction of the spectral structure underlying the Korteweg–De Vries equation. The immediate analytical background was the inverse-scattering solution obtained by Clifford Gardner, John Greene, Martin Kruskal, and Robert Miura. Lax's formulation separated the essential operator identity from the detailed scattering calculation and made the same structure recognizable in other nonlinear systems.

In 1974, Hermann Flaschka and You Watanabe expressed the finite Toda lattice as a matrix Lax equation. Their change of variables converted the exponential interaction between neighboring particles into the evolution of a tridiagonal Jacobi matrix. The eigenvalues of that matrix supplied conserved quantities and connected the mechanical system with the spectral theory of orthogonal polynomials.

A parallel development concerned first-order matrix differential operators. Vladimir Zakharov and Alexei Shabat formulated the spectral problem associated with the nonlinear Schrödinger equation. Their representation placed the nonlinear equation in a zero-curvature form and extended inverse-scattering methods beyond scalar second-order spectral operators.

The Korteweg–De Vries equation

For the Korteweg–De Vries equation in the convention

[ u_t-6uu_x+u_{xxx}=0, ]

a Lax pair is formed by the differential operators

[ L=-\partial_x^2+u(x,t) ]

and

[ P=-4\partial_x^3+6u\partial_x+3u_x. ]

Direct evaluation of ([P,L]) shows that the operator equation (L_t=[P,L]) is equivalent to the nonlinear evolution equation for (u). The operator (L) is a one-dimensional Schrödinger operator, so the Korteweg–De Vries flow preserves its scattering data.

For a rapidly decaying potential, the negative discrete eigenvalues of (L) correspond to soliton components. The continuous spectrum carries radiative information through reflection and transmission coefficients. Under the nonlinear evolution, the eigenvalues remain fixed while the associated norming data acquire explicitly determined time dependence. The inverse scattering transform reconstructs (u(x,t)) from this evolved spectral information.

The Lax representation also generates conservation laws. Formal expansions of spectral quantities at large (\lambda) produce integrals involving (u) and its spatial derivatives. Their conservation reflects the isospectral character of the flow rather than an accidental cancellation specific to a single differential identity.

The finite Toda lattice

The nonperiodic finite Toda lattice has canonical coordinates (q_n) and momenta (p_n), with Hamiltonian

[ H=\frac{1}{2}\sum_n p_n^2+\sum_n e^{q_n-q_{n+1}}. ]

The Flaschka–Watanabe variables are

[ a_n=\frac{1}{2}\exp\left(\frac{q_n-q_{n+1}}{2}\right), \qquad b_n=-\frac{1}{2}p_n. ]

They satisfy

[ \dot a_n=a_n(b_{n+1}-b_n), \qquad \dot b_n=2(a_n^2-a_{n-1}^2). ]

Let (L) be the symmetric tridiagonal matrix whose diagonal entries are (b_n) and whose adjacent off-diagonal entries are (a_n). Let (B) be the skew-symmetric matrix with upper adjacent entries (a_n) and lower adjacent entries (-a_n). The Toda equations then take the form

[ \dot L=[B,L]. ]

The eigenvalues of (L) are constants of motion. Since (L) is a Jacobi matrix, its spectral measure also determines the recurrence coefficients (a_n) and (b_n). The Toda flow can consequently be described as a linear evolution of spectral data followed by reconstruction of the tridiagonal matrix.

The quantities (\operatorname{tr}(L^k)) provide conserved Hamiltonians. Their Poisson-commutation properties follow from the compatible algebraic structure carried by the matrix entries, which can also be formulated through a classical (r)-matrix. This formulation explains why the commutator equation supplies not only spectral invariants but also a hierarchy of mutually compatible flows.

Zero-curvature representation

Many Lax pairs are written using first-order differential systems rather than a single time-dependent spectral operator. Consider

[ \psi_x=U(x,t,\lambda)\psi, \qquad \psi_t=V(x,t,\lambda)\psi, ]

where (U) and (V) are matrices depending on a spectral parameter (\lambda). Equality of mixed derivatives gives

[ U_t-V_x+[U,V]=0. ]

This is the zero-curvature condition. It states that the connection

[ \mathcal A=U,dx+V,dt ]

has vanishing curvature on the two-dimensional space of independent variables. The nonlinear field equation appears when the coefficients of the different powers of (\lambda) are equated.

The operator Lax equation and the zero-curvature equation are closely related. Taking

[ L=\partial_x-U, \qquad P=V ]

converts their compatibility into a commutator relation, subject to the sign convention used for the auxiliary system. The geometric form is particularly natural for integrable hierarchies, because additional evolution variables can be introduced as further compatible components of the same connection.

Nonuniqueness and gauge transformations

A nonlinear equation generally does not determine a unique Lax pair. If (G) is an invertible operator depending on the independent variables, the transformation

[ \psi\mapsto G\psi ]

changes the matrices in a first-order auxiliary system according to

[ U\mapsto GUG^{-1}+G_xG^{-1}, \qquad V\mapsto GVG^{-1}+G_tG^{-1}. ]

The zero-curvature condition is preserved. Such gauge transformations can produce visibly different Lax representations of the same nonlinear equation.

Further nonuniqueness results from changes in the spectral parameter or from enlarging the auxiliary matrix system. Two Lax pairs can therefore encode equivalent dynamics while differing in matrix dimension, differential order, or analytic normalization. Their equivalence is determined by the relation between their spectral problems rather than by superficial agreement of formulas.

Not every formal commutator representation has the same mathematical content. A representation whose spectral operator carries no informative dependence on the dynamical fields can produce conserved quantities that are trivial or redundant. In analytically substantive cases, the spectral problem organizes the initial data, the time evolution acts simply on that data, and an inverse problem reconstructs the nonlinear fields.

Relation to integrable hierarchies

A fixed spectral operator can participate in several commuting Lax equations,

[ \frac{\partial L}{\partial t_n}=[P_n,L]. ]

The associated variables (t_n) define an integrable hierarchy. Compatibility between two such flows requires

[ \frac{\partial P_m}{\partial t_n} -\frac{\partial P_n}{\partial t_m} +[P_m,P_n]=0. ]

For the Korteweg–De Vries hierarchy, the operators (P_n) are constructed from fractional powers of the Schrödinger operator (L), with the differential part extracted from the resulting pseudodifferential operator. This construction generates higher nonlinear evolution equations while preserving the same spectral operator.

The hierarchy viewpoint also connects Lax equations with Hamiltonian mechanics. Compatible Poisson structures can generate the same sequence of flows through different Hamiltonians, producing a bi-Hamiltonian system. The Lax invariants then appear as part of a broader algebraic structure involving recursion operators and commuting Hamiltonian vector fields.