Kazdan–Warner Equations on Graphs

• Kazdan–Warner-type equations on graphs are nonlinear elliptic equations using discrete Laplacians to extend curvature prescription from smooth manifolds to combinatorial and network settings.
• They employ variational principles, topological degree theory, and heat-flow methods to establish existence and multiplicity of solutions across various parameter regimes.
• Critical thresholds like c₋(h) and the sign-changing behavior of the function h determine uniqueness and bifurcation patterns, paralleling results in the smooth setting.

A Kazdan–Warner-type equation on a graph is a nonlinear elliptic equation of the form

$\Delta u = c - h\,e^u$

or generalizations thereof, where $\Delta$ is an appropriate (possibly fractional or $p$-) Laplacian on the graph, $h$ is a prescribed function on the vertices (or edges in the network case), and $c$ is a real parameter. This framework extends the classical smooth theory of prescribed curvature problems on manifolds to discrete, network, and infinite graph settings, yielding a nonlinear potential theory on combinatorial, metric, and measure-theoretic graphs.

1. Discrete Laplacians, Graph Function Spaces, and Variational Structure

Consider a connected (finite or infinite) graph $G = (V, E)$, with positive vertex measure $\mu: V \to (0, \infty)$ and symmetric edge weights $w: E \to (0, \infty)$, or, more generally, a symmetric conductance $b:X \times X \to [0,\infty)$ for countable $X$ in the canonically compactifiable case (Keller et al., 2017). The discrete Laplacian is

$$\Delta f(x) = \frac{1}{\mu(x)} \sum_{y \sim x} w_{xy}(f(y) - f(x)),$$

and its energy (Dirichlet) form is

$$\mathcal{E}(f) = \frac{1}{2} \sum_{x,y \in V} w_{xy}(f(y) - f(x))^2.$$

Function spaces of relevance include $\ell^p(V, \mu)$, the Dirichlet (energy) domain $$\mathcal{D} = \{ f : |\nabla f| \in \ell^2(V, \mu) \}$$, or the domain $D(Q)$ of a Dirichlet form $Q$ (Keller et al., 2017). On finite graphs, all Sobolev, Poincaré, and Trudinger-Moser-type inequalities become uniform and compact (Grigor'yan et al., 2016, Li et al., 2023). Canonically compactifiable graphs ($D(Q) \subset \ell^\infty(X)$) admit a fully discrete Sobolev embedding with uniform sup-norm control and compactness of embedding, crucial for variational arguments (Keller et al., 2017).

The Kazdan–Warner energy functional on a finite graph is typically

$$J(u) = \frac{1}{2} \int_V |\nabla u|^2\, d\mu - \int_V h(x) e^{u(x)} d\mu + c \int_V u(x)\, d\mu$$

(or its various constrained or sub/super-solution forms). Critical points correspond to strong solutions of the Kazdan–Warner-type equation. For general nonlocal, higher-order, or fractional Laplacians, analogous matrix or spectral definitions are employed (Liu et al., 11 Dec 2025).

2. Existence and Multiplicity: Parameter Regimes and Degree-Theoretic Criteria

The Kazdan–Warner equation

$\Delta u = c - h\,e^u$

on a finite graph admits a sharp existence theory, precisely reflecting continuum obstructions but tailored to the discrete measure-theoretic setting (Grigor'yan et al., 2016, Sun et al., 2021, Keller et al., 2017):

• For $c>0$: solvability if and only if $h(x) > 0$ somewhere on $V$.
• For $c=0$: solvability if and only if $h$ changes sign and $\sum_V h(x)\mu(x) < 0$.
• For $c<0$: existence if and only if $h(x) < 0$ somewhere, and there exists a critical constant $c_-(h)<0$ so that the equation is solvable for every $0>c>c_-(h)$, with $c_-(h) = -\infty$ in the strictly negative $h$ case (Ge, 2016, Sun et al., 2021). At the threshold $c=c_-(h)$, existence holds as well (Ge, 2016).

In negative and sign-changing cases, multiplicity may arise. There exists a regime where at least two distinct solutions appear, characterized by changes in topological degree or bifurcation phenomena analogous to the smooth Ding-Liu–Yang-Zhu continuum results (Liu et al., 2020, Liu et al., 11 Dec 2025).

In the negative fractional setting on finite graphs,

$$(-\Delta)^s u = h_\lambda e^{2u} - c,\quad s \in (0,1),\; c < 0,$$

a unique solution exists for $\lambda \leq 0$, while there is a critical parameter $\Lambda^*_s$ for which the problem admits at least two solutions for $0 < \lambda < \Lambda^*_s$ and none for $\lambda > \Lambda^*_s$. This matches the classical critical parameter windows for the negative-curvature case (Liu et al., 11 Dec 2025).

The use of topological (Brouwer) degree, with explicit calculation based on a priori bounds and graph reduction to two vertices, allows for a complete existence theory even under generalized exponential nonlinearities (Hua et al., 20 May 2025, Yu, 2024).

3. Analytical and Topological Methods: Variational, Degree, and Heat Flow Approaches

The solvability theory on graphs is constructed through an overview of discrete variational, monotone iterative, and topological degree-theoretic arguments:

• Variational methods: On finite graphs or canonically compactifiable graphs, constrained minimization and Lagrange multipliers produce solutions in all sign regimes (Grigor'yan et al., 2016, Keller et al., 2017, Li et al., 2023), using coercivity from Poincaré and compactness in the discrete Sobolev space.
• Sub/super-solution constructions and monotone iteration: Upper/lower solutions bracket a true solution within order intervals, exploiting the bijectivity and order-preservation of the (possibly nonlinear) discrete Laplacian (Ge, 2016, Ge, 2016, Liu et al., 2020, Hua et al., 20 May 2025).
• Degree theory: The Brouwer degree is defined for the operator $F(u) = -\Delta u + h(x)e^{u(x)} - c$ in $\mathbb{R}^N$, with explicit computation based on limiting and blow-up scenarios (Sun et al., 2021, Yu, 2024, Hua et al., 20 May 2025). Topological invariance under homotopy tracks bifurcations and multiplicity.
• Heat-flow methods: On infinite graphs, the lack of compactness invalidates straightforward variational arguments. Instead, existence is established by parabolic flow (discrete heat equation with exponential source), uniform a priori energy bounds via Lyapunov functionals, and passage to time and domain limits (Ge et al., 2017).

For nonlocal problems involving the spectral fractional Laplacian or the $p$-Laplacian, the theory is extended with uniform bounds, maximum principles, compactness, and careful spectral or monotonicity arguments (Liu et al., 11 Dec 2025, Ge, 2016).

4. Infinite, Canonically Compactifiable, and Network Graph Extensions

The discrete Kazdan–Warner paradigm generalizes across various classes of graphs:

• Infinite graphs: Under $h \leq 0$ and appropriate integrability/geometry (e.g., Cheeger constant), the infinite graph equation

$\Delta f = g - h\,e^f$

admits global solutions, with Cheeger inequality imitating Sobolev embedding and ensuring uniform a priori bounds on large finite exhaustions (Ge et al., 2017).

• Canonically compactifiable graphs: Here, the Dirichlet form domain $D(Q)$ is bounded in $\ell^\infty$; solvability closely matches the smooth precompact manifold regime. The existence theory is variational, with Moser–Trudinger and Poincaré inequalities, and follows the same sign/integral obstructions as the smooth case (Keller et al., 2017).
• Metric graphs and networks: On finite metric graphs ("networks") with edgewise equations and Kirchhoff–Neumann vertex conditions, the entire Kazdan–Warner existence dichotomy persists: solution existence is controlled by the sign of integrals and sign-changing properties of $h$, and the only role of network topology is in Poincaré and embedding constants (Camilli et al., 2019).

5. Generalizations: Higher-Order, Fractional, and Fully Nonlinear Elliptic Graph Problems

Generalizations encompass:

• $p$-Laplacian and higher-order operators: The theory extends—using similar variational and maximum-principle techniques—to nonlinear $p$-Laplacians and polyharmonic (higher-order) graph Laplacians (Ge, 2016, Grigor'yan et al., 2016, Pinamonti et al., 2021). The existence/threshold structure for solutions often parallels the classical case when appropriate discrete Sobolev and maximum principles are available.
• Fractional graph Laplacian: Spectral or kernel forms of the graph fractional Laplacian $(-\Delta)^s$ admit an existence and multiplicity theory for negative exponentials, with threshold and uniqueness properties matching the local case (Liu et al., 11 Dec 2025).
• General exponential nonlinearities: Equations of the form $-\Delta u = F_n(x, u(x))$ with $F_n(x, y) = \sum_{i=1}^n f_i(x) e^{i y} + c$ are handled via single-vertex a priori bounds, reduction to small graphs (Schur complement), and explicit degree calculation; solutions exist when the computed degree is nonzero, and can sometimes be constructed via sub/super-solution when the degree vanishes (Hua et al., 20 May 2025).
• Kazdan–Warner-type equations with nonstandard nonlinearity: For nonlinearities such as $f(u) = (1-1/(1+u^{2n})) e^u$, careful classification of roots and connectivity arguments yield sharp existence and multiplicity theorems, with the number of constant solutions controlling the Brouwer degree (Yu, 2024).

6. Analytical Thresholds, Critical Parameters, and Uniqueness Issues

The location of critical thresholds, parameter dependence, and uniqueness are central:

• Thresholds $c_-(h)$: On finite graphs, for $h < 0$ the critical constant

$$c_-(h) = \inf\{ c < 0 : \text{Kazdan--Warner equation is solvable} \}$$

is negative or $-\infty$, corresponding to the sharp curvature lower bound in the smooth case. At $c = c_-(h)$, solvability still holds (Ge, 2016).

• Multiplicity: For negative total curvature (negative $c$), and sign-changing $h$, there is a critical parameter range exhibiting bifurcation, in which at least two solutions exist; this mirrors the Yang–Zhu, Ding–Liu multiplicity theorems on surfaces (Liu et al., 2020, Liu et al., 11 Dec 2025).
• Uniqueness: In strictly negative or monotone regimes (finite graphs, $h \leq 0$), uniqueness can be proved via strong maximum principles and monotonicity of the nonlinearity or the operator (Ge, 2016, Pinamonti et al., 2021). On infinite or more general graphs, uniqueness may fail or remains open without further structural assumptions (Ge et al., 2017).

These analytical results rigorously parallel the integral geometric obstructions and phenomena in the classical Kazdan–Warner theory on compact and punctured Riemann surfaces.

7. Relation to the Smooth and Networks Theories; Further Developments

Kazdan–Warner equations on graphs realize discrete analogues of the problem of prescribing Gaussian curvature (or higher Q-curvature) under conformal change on smooth manifolds:

• The discrete Laplacian mimics the geometric Laplacian on the manifold.
• Solution criteria—sign of $h$, total curvature, sign-changing properties—exactly parallel the role of integral obstructions and average curvature in the continuum Kazdan–Warner classification (Keller et al., 2017, Grigor'yan et al., 2016).
• On networks/metric graphs and infinite graphs, the theory extends with adjustment for boundary/vertex conditions, Cheeger-type inequalities, or noncompactness.
• Recent advances cover fractional Laplacians, general exponential nonlinearities, and problems with "fractional curvature" or "multi-exponential" structure, further connecting the graph case with developments in geometric analysis and statistical mechanics (Liu et al., 11 Dec 2025, Hua et al., 20 May 2025).

The extension of Kazdan–Warner-type equations to graphs continues to inform both discrete geometric analysis and applications to network models with prescribed structural or curvature-like effects. Open problems include infinite-graph regularity, blow-up phenomena, classification of solution branches in multi-parameter regimes, and connections to discrete conformal geometry.