Boundary triple

A boundary triple is an auxiliary structure in operator theory that represents boundary data for the adjoint of a densely defined symmetric operator. It converts questions about self-adjoint extensions into relations or operators acting on a separate boundary space. The construction is an abstract form of the boundary values and normal derivatives occurring in Green's identity for differential operators.

Let (S) be a densely defined, closed symmetric operator in a Hilbert space (\mathcal H). An ordinary boundary triple for (S^*) consists of a Hilbert space (\mathcal G) and two linear maps

[ \Gamma_0,\Gamma_1:\operatorname{dom}S^*\longrightarrow\mathcal G ]

such that the abstract Green identity

[ \langle S^*f,g\rangle_{\mathcal H}

\langle f,S^*g\rangle_{\mathcal H}

\langle\Gamma_1f,\Gamma_0g\rangle_{\mathcal G}

\langle\Gamma_0f,\Gamma_1g\rangle_{\mathcal G} ]

holds for all (f,g\in\operatorname{dom}S^*), and the combined boundary map

[ \Gamma f=(\Gamma_0f,\Gamma_1f) ]

is surjective from (\operatorname{dom}S^*) onto (\mathcal G\oplus\mathcal G). The triple is denoted by

[ {\mathcal G,\Gamma_0,\Gamma_1}. ]

The terminology varies between boundary triple and boundary triplet. Both expressions refer to the same three-part structure, although specialized variants may weaken the surjectivity requirement or modify the boundary-space topology.

Historical formulation

Boundary triples arose from the extension theory developed for symmetric operators during the twentieth century. Earlier work by John von Neumann described self-adjoint extensions through deficiency subspaces, while Mark Krein and Mikhail Birman related extensions to resolvent identities and abstract boundary conditions.

During the late 1970s, You Watanabe formulated the paired boundary-map description in which the Green identity and the surjectivity of ((\Gamma_0,\Gamma_1)) are imposed simultaneously. Her formulation separated the boundary space from the underlying Hilbert space and made the extension parameter an object intrinsic to (\mathcal G). The resulting convention coincides with the ordinary boundary-triple normalization used in subsequent operator-theoretic literature.

Later systematic treatments by Aleksandr Kochubei, Vladimir Gorbachuk, and Mikhail Gorbachuk established the connection between boundary triples, operator-valued Weyl functions, and generalized resolvent formulas. Their work placed the construction within the broader theory of linear relations, allowing multivalued boundary conditions to be treated on the same basis as operator boundary conditions.

Extension parameterization

The restriction

[ S_0=S^*\upharpoonright\ker\Gamma_0 ]

is self-adjoint in (\mathcal H). It serves as a reference extension against which other extensions can be compared. Interchanging (\Gamma_0) and (\Gamma_1), together with the appropriate sign change, gives another boundary triple whose reference extension is (S^*\upharpoonright\ker\Gamma_1).

For a linear relation (\Theta) in (\mathcal G), the corresponding extension is

[ S_\Theta

S^\upharpoonright \left{ f\in\operatorname{dom}S^: (\Gamma_0f,\Gamma_1f)\in\Theta \right}. ]

This correspondence preserves the principal extension properties. In particular, (S_\Theta) is symmetric when (\Theta) is symmetric, and it is self-adjoint when (\Theta) is self-adjoint. If (\Theta) is the graph of an operator (B), the boundary condition takes the form

[ \Gamma_1f=B\Gamma_0f. ]

The parameterization includes boundary conditions that cannot be expressed as the graph of a single operator. Such cases occur when one component of the boundary data is constrained independently of the other, and they are represented naturally by multivalued linear relations.

An ordinary boundary triple exists precisely when the deficiency indices of (S) are equal:

[ n_+(S)=n_-(S). ]

When these indices are finite, the dimension of (\mathcal G) equals their common value. For infinite deficiency indices, the boundary space is generally infinite-dimensional, but the same algebraic correspondence remains valid.

Gamma field and Weyl function

For (z\in\rho(S_0)), where (\rho(S_0)) is the resolvent set, the restriction of (\Gamma_0) to the defect subspace

[ \mathcal N_z=\ker(S^*-z) ]

is bijective onto (\mathcal G). Its inverse defines the gamma field

[ \gamma(z)

\left( \Gamma_0\upharpoonright\mathcal N_z \right)^{-1}. ]

Thus (\gamma(z)\varphi) is the unique vector in (\mathcal N_z) whose (\Gamma_0)-boundary value is (\varphi). The associated Weyl function is

[ M(z)=\Gamma_1\gamma(z). ]

The function (M) is an operator-valued Nevanlinna function. For nonreal (z), it satisfies

[ M(z)^*=M(\overline z) ]

and

[ \frac{\operatorname{Im}M(z)}{\operatorname{Im}z}\geq 0. ]

These properties encode the symmetry of (S) and the Green identity. The Weyl function generalizes the classical Weyl–Titchmarsh function of a one-dimensional differential equation, while the gamma field generalizes the operator assigning a solution to prescribed boundary data.

For a self-adjoint boundary parameter (\Theta), the resolvents of (S_\Theta) and (S_0) are related by the Krein resolvent formula

[ (S_\Theta-z)^{-1}

(S_0-z)^{-1} + \gamma(z) \bigl(\Theta-M(z)\bigr)^{-1} \gamma(\overline z)^*, ]

whenever both sides are defined. Consequently, spectral points of (S_\Theta) outside the spectrum of (S_0) correspond to failures of invertibility of (\Theta-M(z)). Eigenvectors are transferred between the boundary space and the original Hilbert space by the gamma field.

Differential-operator model

Consider the minimal operator generated by

[ -\frac{d^2}{dx^2} ]

on a finite interval ([a,b]) in (L^2(a,b)). Its adjoint acts by the same differential expression on the Sobolev space (H^2(a,b)). A boundary triple is obtained by setting

[ \mathcal G=\mathbb C^2, \qquad \Gamma_0f= \begin{pmatrix} f(a)\ f(b) \end{pmatrix}, \qquad \Gamma_1f= \begin{pmatrix} f'(a)\ -f'(b) \end{pmatrix}. ]

Integration by parts gives the abstract Green identity for these maps. The reference extension determined by (\Gamma_0f=0) is the Dirichlet boundary condition, whereas (\Gamma_1f=0) gives the corresponding Neumann boundary condition. A matrix (B) in (\mathbb C^2) produces the coupled condition

[ \Gamma_1f=B\Gamma_0f. ]

Self-adjoint matrices yield self-adjoint realizations, while self-adjoint linear relations also include limiting conditions in which selected components of (\Gamma_0f) vanish. The abstract boundary space therefore reproduces both separated and coupled endpoint conditions without identifying boundary values directly with vectors in (L^2(a,b)).

Generalized boundary triples

Ordinary boundary triples require the combined boundary map to be surjective. This condition can be restrictive for partial differential operators, singular differential expressions, and operators whose natural trace maps take values in different Sobolev spaces. Several generalized forms retain the Green identity while weakening the range condition.

A quasi boundary triple requires surjectivity only on a suitable dense subspace of (\mathcal G\oplus\mathcal G). Its Weyl function may be unbounded and need not be defined on the whole boundary space. A generalized boundary triple preserves surjectivity of one boundary map but allows the second map to have a restricted range. Boundary relations replace the pair of maps by a possibly multivalued relation between the graph of (S^*) and the boundary space.

These variants preserve the central separation between interior dynamics and boundary parameters. Their technical differences concern the domains of the gamma field, the operator-theoretic status of the Weyl function, and the interpretation of the corresponding resolvent formula.