Dirichlet Subgraph: Theory & Applications

• Dirichlet subgraph is a precise construction that restricts Dirichlet forms to an induced subgraph by imposing zero boundary conditions.
• It underpins spectral theory by enabling domain-monotonicity and eigenvalue interlacing, offering practical insights into spectral bracketing.
• The framework extends to fractional and poly-Laplacians, providing sharp eigenvalue bounds and bridging discrete analysis with potential theory.

A Dirichlet subgraph is a precise construction in the theory of Dirichlet forms on graphs, representing the restriction ("part") of a given Dirichlet form to an induced subgraph, with boundary conditions imposed by forcing functions to vanish on the prescribed boundary. This framework unifies a broad spectrum of discrete and metric graph analysis, from spectral theory to potential theory and stochastic processes, and extends naturally to generalizations including fractional and poly-Laplacians (Haeseler, 2017, Wang, 2023, Hua et al., 2024, Carlson, 2011).

1. Dirichlet Forms on Graphs and Subgraph Restriction

Let $G = (V, E)$ be a countable, locally finite graph (no loops, no multiple edges), equipped with a symmetric jump weight $j: V \times V \to [0,\infty)$ and a killing weight $k: V \to [0,\infty)$. Fix a strictly positive vertex measure $m: V \to (0, \infty)$. The Dirichlet form $\mathcal{E}$ is defined on $\ell^2(V, m)$ by

$$\mathcal{E}(f, g) = \frac{1}{2} \sum_{x, y \in V} j(x, y) [f(x) - f(y)][g(x) - g(y)] + \sum_{x \in V} k(x) f(x)g(x).$$

Its natural domain is

$$D(\mathcal{E}) = \{ f \in \ell^{2}(V, m) \mid \mathcal{E}(f, f) < \infty \},$$

forming a closed, Markovian Dirichlet form (Haeseler, 2017).

For an induced subgraph $H \subset G$ with vertex set $W \subset V$, edge set $E(H) = \{\{x, y\} \in E : x, y \in W\}$, define its vertex-boundary as

$$\partial H = \{ x \in W \mid \exists\, y \in V \setminus W \text{ with } j(x, y) > 0 \}.$$

The Dirichlet subgraph (Editor’s term) imposes zero boundary conditions on $\partial H$, i.e., $f(x) = 0$ for all $x \in \partial H$, and the restricted Dirichlet form is

$$\mathcal{E}_H(f, g) = \frac{1}{2}\sum_{\substack{x, y \in W\ \{x, y\} \in E}} j(x, y)[f(x) - f(y)][g(x) - g(y)] + \sum_{x \in W \setminus \partial H} k(x) f(x)g(x),$$

with domain $$D(\mathcal{E}_H) = \{ f \in D(\mathcal{E}) : f(x) = 0 \text{ for all } x \in \partial H \}$$ (Haeseler, 2017).

2. Spectral Theory and Monotonicity on Dirichlet Subgraphs

A fundamental property is spectral domain-monotonicity: if $H_1 \subset H_2$, then for any nonnegative source $f$ supported on $H_1$,

$$(\Delta_{H_1} + \alpha)^{-1}f \leq (\Delta_{H_2} + \alpha)^{-1}f, \quad \alpha > 0,$$

where $\Delta_H$ is the self-adjoint generator of $(\mathcal{E}_H, D(\mathcal{E}_H))$. The sequence of Dirichlet eigenvalues $\{\lambda_k(H)\}$ exhibits spectral bracketing: $\lambda_k(H_2) \leq \lambda_k(H_1)$ for $H_1 \subset H_2$ (Haeseler, 2017).

When subgraphs $H$ exhaust $G$, the Dirichlet forms $\mathcal{E}_H$ converge (in the Mosco sense) to $\mathcal{E}$ on $G$, and the capacities of singleton sets in $H$ tend to their capacities in the full graph. Spectral comparison and trace results under this exhaustion relate discrete spectral theory to potential theory and limit behaviors (Haeseler, 2017).

3. Dirichlet Subgraph Constructions Beyond the Classical Case

For the discrete Laplacian $\Delta$, the Dirichlet subgraph construction underpins extensions to more general operators.

Fractional Laplacians

For $0 < \alpha < 2$, the fractional Laplacian $(-\Delta)^{\alpha/2}$ on a subgraph $\Omega \subset \mathbb{Z}^d$ is defined via zero extension $u^*(x) = u(x)$ for $x \in \Omega$, $0$ for $x \notin \Omega$, and

$$L^\alpha_\Omega u(x) = (-\Delta)^{\alpha/2}u^*(x) \big|_{x \in \Omega}.$$

The Dirichlet eigenvalue problem $L^\alpha_\Omega \phi_j = \lambda_j(\Omega) \phi_j$, with $\phi_j|_{\Omega^c} \equiv 0$, yields a real, positive spectrum. Explicit eigenvalue estimates of Kröger- and Li–Yau-type involve both the volume and a nonlocal boundary term $$|\partial^\alpha \Omega| := \sum_{x \in \Omega, y \notin \Omega} Q_\alpha(x, y)$$ (Wang, 2023).

Poly-Laplacians

For the discrete poly-Laplace operator $\Delta^{\ell}$, on $\Omega \subset \mathbb{Z}^d$, Dirichlet boundary conditions correspond to extension by zero and restriction: $$\Delta^{\ell, D}_\Omega f := (\Delta^\ell f^*)|_{\Omega}.$$ The Dirichlet eigenvalue problem $$(-1)^\ell \Delta^{\ell, D}_\Omega f(x) = (\lambda_k^{(\ell)}(\Omega))^\ell f(x)$$ produces a positive real spectrum. Sharp eigenvalue-sum bounds and spectral comparison inequalities—such as $$(\lambda_k^{(\ell)}(\Omega))^2 < \lambda_k^{(2\ell)}(\Omega)$$ for subgraphs of $\mathbb{Z}^d$—hold in direct analogy with the classical case (Hua et al., 2024).

4. Boundary Value Problems and Potential Theory

The Dirichlet subgraph framework enables formulation and solution of Dirichlet problems on both finite and infinite graphs, under appropriate metric compactifications and connectivity hypotheses. For an infinite weighted graph $G = (V, E, c, w)$ completed in the minimal-resistance metric, one defines the boundary $\partial \bar{G} = \bar{G} \setminus V$, and the Dirichlet problem seeks $f$ minimizing

$$J(f) = \mathcal{E}(f) - \langle g, f \rangle_{\ell^2(w)},$$

among functions vanishing on the boundary $\partial H$. Existence and uniqueness are guaranteed by a discrete Poincaré inequality under compactness and weak connectivity. This supports the construction of Dirichlet Laplacians with compact resolvent and a purely discrete spectrum, as well as the associated Green’s function and probabilistic semigroups with absorbing (Dirichlet) boundary conditions (Carlson, 2011).

5. Applications: Harmonic Analysis, Heat Kernels, and Capacity

On Dirichlet subgraphs, harmonic functions satisfy $\Delta_H u = 0$ in the interior, with uniqueness and existence following from the Markov property of the Dirichlet form. The form-restriction principle ensures heat kernels on subgraphs are pointwise dominated by those on larger graphs, yielding on- and off-diagonal estimates. Potential theory on graphs leverages restricted capacities: as subgraphs $H$ exhaust $G$, the capacities $\operatorname{cap}_H(A)$ for $A \subset W$ converge monotonically to $\operatorname{cap}_G(A)$, characterizing key global properties such as recurrence and transience (Haeseler, 2017).

6. Canonical Examples and Generalizations

Trees and Lattices

For trees, Dirichlet subgraphs correspond to finite connected subtrees with boundary at the leaves; the resulting Dirichlet Laplacian exhibits interlacing eigenvalues under subtree inclusion. For $\mathbb{Z}^d$, boxes $[-N, N]^d$ serve as Dirichlet subgraphs, with boundaries corresponding to the combinatorial boundary in $\mathbb{Z}^d$, fundamental to discrete Poincaré and Sobolev inequalities (Haeseler, 2017).

Fractional/Poly-Laplacian Eigenvalue Bounds

Explicit Li–Yau- and Kröger-type eigenvalue-sum bounds for both fractional (Wang, 2023) and poly-Laplacians (Hua et al., 2024) are sharp up to boundary error terms depending on nonlocal or higher-order combinatorial stencils. For large periodic subgraphs or boxes, these error terms vanish in the thermodynamic limit, giving Weyl-type asymptotic behaviors. The strict inequality between higher and lower poly-Laplacian spectra on $\mathbb{Z}^d$ links spectral theory and combinatorial geometry.

Operator	Boundary Term	Spectral Property
Laplacian	$|\partial \Omega|$	Spectral domain-monotonicity, interlacing
Fractional Lap.	$|\partial^\alpha \Omega|$	Kröger/Li–Yau eigenvalue bounds
Poly-Laplacian	$|\partial^\ell \Omega|$	$(\lambda^{(\ell)}_k)^2 < \lambda^{(2\ell)}_k$

7. Significance and Theoretical Insights

The Dirichlet subgraph formalism provides a robust, unifying toolkit for discrete analysis, allowing passage from local combinatorial structure to global spectral and potential-theoretic properties. It supports advanced applications in spectral geometry, random walks, heat kernel estimates, and potential theory, and enables precise analysis of spectral convergence, trace forms, and capacity in both finite and infinite graphs. This framework extends naturally to generalizations involving nonlocal operators, higher-order Laplacians, and metric graph compactifications, demonstrating centrality in discrete analysis and mathematical physics (Haeseler, 2017, Wang, 2023, Hua et al., 2024, Carlson, 2011).