Kazdan–Warner Equations in Geometry

• Kazdan–Warner equations are nonlinear elliptic PDEs with exponential nonlinearities used in curvature prescription and variational problems across diverse settings.
• They employ variational methods, sub- and supersolution strategies, and concentration analysis to address critical thresholds and blow‐up phenomena.
• Extensions include formulations on graphs, networks, and foliated manifolds, with results often guided by symmetry identities and topological degree theory.

Searching arXiv for recent and foundational papers on Kazdan–Warner type equations across manifolds, graphs, networks, foliations, and generalized variational formulations. Kazdan–Warner type equations are nonlinear elliptic equations built around exponential source terms and normalization constraints that originate in prescribed curvature problems and extend to mean-field equations, vector-valued moment-map systems, foliated geometry, finite and infinite graphs, networks, and higher-order boundary curvature prescription. In the literature represented here, they include the classical form

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

the critical mean-field equation

$$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$

multi-exponential systems with several weights, and Kazdan–Warner identities that impose symmetry-based obstructions to solvability (Sun et al., 2020, Miyatake, 2020, Sun et al., 2021, Cruz-Blázquez et al., 23 Mar 2026). Their analysis is organized by variational structure, sharp thresholds such as $8\pi$ or $\lambda_1$, concentration and blow-up, sub- and supersolution methods, Brouwer degree, and conformal invariance.

1. Surface variational theory and the critical $8\pi$ regime

On a compact Riemannian surface $(\Sigma,g)$ without boundary, a standard Kazdan–Warner type functional is

$$J_{\alpha,\beta}(u)=\frac12\int_\Sigma \bigl(|\nabla_g u|^2-\alpha u^2\bigr)\,dv_g-\beta\log\int_\Sigma h e^u\,dv_g, \qquad u\in \mathcal H:=\left\{u\in W^{1,2}(\Sigma):\int_\Sigma u\,dv_g=0\right\}.$$

Its critical points satisfy

$$\Delta_g u-\alpha u = \beta\left(\frac{h e^u}{\int_\Sigma h e^u\,dv_g}-\frac{1}{\mathrm{Vol}_g(\Sigma)}\right).$$

When $\alpha<\lambda_1(\Sigma)$, the quadratic part is coercive on $\mathcal H$; for every $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$0, the functional is bounded below and admits a minimizer, while $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$1 or $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$2 forces $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$3 (Yang et al., 2017).

The constant $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$4 is the critical two-dimensional threshold associated with the Trudinger–Moser mechanism. At $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$5, minimizing sequences may fail to converge and instead concentrate at a point. If $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$6 has no minimizer, the infimum can nevertheless be computed exactly: $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$7 where $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$8 is the regular part of the Green function of $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$9. The same framework extends to the large-$8\pi$0 regime by imposing orthogonality to lower eigenspaces and replacing $8\pi$1 by higher spectral thresholds $8\pi$2 (Yang et al., 2017).

A related critical formulation fixes the area of $8\pi$3 to be $8\pi$4 and uses the normalized equation

$8\pi$5

The normalization makes the equation invariant under adding constants to $8\pi$6, and the corresponding Euler–Lagrange functional is

$8\pi$7

This formulation is the critical $8\pi$8 prescribed curvature or Liouville-type problem on a compact surface, with the existence theory again reducing to the behavior of minimizing or critical sequences near the blow-up threshold (Sun et al., 2020).

2. Sign-changing prescribed functions and blow-up asymptotics

A major refinement of the critical theory is the sign-changing case, in which the prescribed function $8\pi$9 is smooth, positive somewhere, and allowed to take negative values. In that setting one loses the maximum principle tools often available when $\lambda_1$0, so minimizing sequences and lower bounds must be analyzed by refined energy estimates. The relevant effective potential is

$\lambda_1$1

where $\lambda_1$2 and $\lambda_1$3 is the regular part of the Green function defined by

$\lambda_1$4

with local expansion

$\lambda_1$5

If, at each maximum point of $\lambda_1$6,

$\lambda_1$7

where $\lambda_1$8 is the Gaussian curvature, then $\lambda_1$9 has a minimizer. The resulting existence theorem generalizes the $8\pi$0 result of Ding–Jost–Li–Wang to prescribed functions that change sign (Sun et al., 2020).

In the corresponding blow-up analysis, one studies critical points $8\pi$1 of $8\pi$2 normalized by

$8\pi$3

If blow-up occurs, then $8\pi$4, the mass $8\pi$5 concentrates into a Dirac mass at a single point $8\pi$6, the negative part vanishes in the limit, and away from $8\pi$7 the normalized sequence converges to $8\pi$8. The peak points $8\pi$9 converge to a critical point of $(\Sigma,g)$0, and for minimizing sequences the limiting point is a maximum point of that function (Sun et al., 2020).

The central asymptotic identity in the blow-up regime is

$(\Sigma,g)$1

This formula couples the small parameter $(\Sigma,g)$2, the blow-up height $(\Sigma,g)$3, and the geometry at the concentration point. It is the sign of $(\Sigma,g)$4 that determines whether the bubble asymptotics are compatible with the minimizing sequence. The associated sharp test-function expansion yields

$(\Sigma,g)$5

and under the positivity condition above, the infimum is attained (Sun et al., 2020).

3. Generalized manifold systems, torus actions, and foliated analogues

Beyond the scalar single-exponential equation, compact manifolds support generalized Kazdan–Warner systems with drift terms or several exponential nonlinearities. One such equation is

$(\Sigma,g)$6

where $(\Sigma,g)$7, $(\Sigma,g)$8, and $(\Sigma,g)$9. Under the sign assumptions

$$J_{\alpha,\beta}(u)=\frac12\int_\Sigma \bigl(|\nabla_g u|^2-\alpha u^2\bigr)\,dv_g-\beta\log\int_\Sigma h e^u\,dv_g, \qquad u\in \mathcal H:=\left\{u\in W^{1,2}(\Sigma):\int_\Sigma u\,dv_g=0\right\}.$$0

there exists a unique smooth solution. The proof can be carried out either by a parabolic flow

$$J_{\alpha,\beta}(u)=\frac12\int_\Sigma \bigl(|\nabla_g u|^2-\alpha u^2\bigr)\,dv_g-\beta\log\int_\Sigma h e^u\,dv_g, \qquad u\in \mathcal H:=\left\{u\in W^{1,2}(\Sigma):\int_\Sigma u\,dv_g=0\right\}.$$1

or by the upper and lower solution method, and the analysis yields a uniform lower bound, a $$J_{\alpha,\beta}(u)=\frac12\int_\Sigma \bigl(|\nabla_g u|^2-\alpha u^2\bigr)\,dv_g-\beta\log\int_\Sigma h e^u\,dv_g, \qquad u\in \mathcal H:=\left\{u\in W^{1,2}(\Sigma):\int_\Sigma u\,dv_g=0\right\}.$$2-to-$$J_{\alpha,\beta}(u)=\frac12\int_\Sigma \bigl(|\nabla_g u|^2-\alpha u^2\bigr)\,dv_g-\beta\log\int_\Sigma h e^u\,dv_g, \qquad u\in \mathcal H:=\left\{u\in W^{1,2}(\Sigma):\int_\Sigma u\,dv_g=0\right\}.$$3 estimate, a uniform $$J_{\alpha,\beta}(u)=\frac12\int_\Sigma \bigl(|\nabla_g u|^2-\alpha u^2\bigr)\,dv_g-\beta\log\int_\Sigma h e^u\,dv_g, \qquad u\in \mathcal H:=\left\{u\in W^{1,2}(\Sigma):\int_\Sigma u\,dv_g=0\right\}.$$4 bound, and, when $$J_{\alpha,\beta}(u)=\frac12\int_\Sigma \bigl(|\nabla_g u|^2-\alpha u^2\bigr)\,dv_g-\beta\log\int_\Sigma h e^u\,dv_g, \qquad u\in \mathcal H:=\left\{u\in W^{1,2}(\Sigma):\int_\Sigma u\,dv_g=0\right\}.$$5, full $$J_{\alpha,\beta}(u)=\frac12\int_\Sigma \bigl(|\nabla_g u|^2-\alpha u^2\bigr)\,dv_g-\beta\log\int_\Sigma h e^u\,dv_g, \qquad u\in \mathcal H:=\left\{u\in W^{1,2}(\Sigma):\int_\Sigma u\,dv_g=0\right\}.$$6 a priori estimates for stationary solutions (Yu, 2023).

A more structural generalization is the vector-valued equation associated with a linear action of a torus on $$J_{\alpha,\beta}(u)=\frac12\int_\Sigma \bigl(|\nabla_g u|^2-\alpha u^2\bigr)\,dv_g-\beta\log\int_\Sigma h e^u\,dv_g, \qquad u\in \mathcal H:=\left\{u\in W^{1,2}(\Sigma):\int_\Sigma u\,dv_g=0\right\}.$$7: $$J_{\alpha,\beta}(u)=\frac12\int_\Sigma \bigl(|\nabla_g u|^2-\alpha u^2\bigr)\,dv_g-\beta\log\int_\Sigma h e^u\,dv_g, \qquad u\in \mathcal H:=\left\{u\in W^{1,2}(\Sigma):\int_\Sigma u\,dv_g=0\right\}.$$8 where $$J_{\alpha,\beta}(u)=\frac12\int_\Sigma \bigl(|\nabla_g u|^2-\alpha u^2\bigr)\,dv_g-\beta\log\int_\Sigma h e^u\,dv_g, \qquad u\in \mathcal H:=\left\{u\in W^{1,2}(\Sigma):\int_\Sigma u\,dv_g=0\right\}.$$9, the $$\Delta_g u-\alpha u = \beta\left(\frac{h e^u}{\int_\Sigma h e^u\,dv_g}-\frac{1}{\mathrm{Vol}_g(\Sigma)}\right).$$0 are nonnegative functions, and the weights $$\Delta_g u-\alpha u = \beta\left(\frac{h e^u}{\int_\Sigma h e^u\,dv_g}-\frac{1}{\mathrm{Vol}_g(\Sigma)}\right).$$1 come from the torus action. Existence is equivalent to the cone condition

$$\Delta_g u-\alpha u = \beta\left(\frac{h e^u}{\int_\Sigma h e^u\,dv_g}-\frac{1}{\mathrm{Vol}_g(\Sigma)}\right).$$2

and solutions are unique modulo constant elements in the orthogonal complement of the span of the active weights. The associated energy is convex, and the solvability condition has an explicit moment-map and GIT interpretation. In a special case this system reduces to the periodic Toda equation, and it gives a new proof that cyclic Higgs bundles produce Toda solutions (Miyatake, 2020).

On compact foliated manifolds, the same generalized formalism persists in the basic subcomplex. If the coefficients $$\Delta_g u-\alpha u = \beta\left(\frac{h e^u}{\int_\Sigma h e^u\,dv_g}-\frac{1}{\mathrm{Vol}_g(\Sigma)}\right).$$3 are basic and the ambient Laplacian preserves basic functions, then solvability of

$$\Delta_g u-\alpha u = \beta\left(\frac{h e^u}{\int_\Sigma h e^u\,dv_g}-\frac{1}{\mathrm{Vol}_g(\Sigma)}\right).$$4

is equivalent to solvability of the reduced basic equation

$$\Delta_g u-\alpha u = \beta\left(\frac{h e^u}{\int_\Sigma h e^u\,dv_g}-\frac{1}{\mathrm{Vol}_g(\Sigma)}\right).$$5

The existence and uniqueness criterion is unchanged from the non-foliated theorem, and any solution can be chosen basic. The principal example is the transverse Hitchin equation for a diagonal harmonic metric on a basic cyclic Higgs bundle, where the PDE becomes a $$\Delta_g u-\alpha u = \beta\left(\frac{h e^u}{\int_\Sigma h e^u\,dv_g}-\frac{1}{\mathrm{Vol}_g(\Sigma)}\right).$$6-valued generalized Kazdan–Warner system on a foliated manifold (Miyatake, 2022).

Generalized Kazdan–Warner equations also appear as the analytic core of adiabatic limits for vortex-type equations. After passing to complex gauge, one obtains families of the form

$$\Delta_g u-\alpha u = \beta\left(\frac{h e^u}{\int_\Sigma h e^u\,dv_g}-\frac{1}{\mathrm{Vol}_g(\Sigma)}\right).$$7

and uniform interior $$\Delta_g u-\alpha u = \beta\left(\frac{h e^u}{\int_\Sigma h e^u\,dv_g}-\frac{1}{\mathrm{Vol}_g(\Sigma)}\right).$$8 bounds for these equations, independent of $$\Delta_g u-\alpha u = \beta\left(\frac{h e^u}{\int_\Sigma h e^u\,dv_g}-\frac{1}{\mathrm{Vol}_g(\Sigma)}\right).$$9, are the key input for smooth convergence away from finitely many points in generalized vortex and multiple-spinor Seiberg–Witten problems (Doan, 2017).

4. Finite graph equations, thresholds, and degree theory

On a connected finite graph, Kazdan–Warner type equations become finite-dimensional nonlinear equations with a graph Laplacian in place of the Laplace–Beltrami operator. A central model is

$\alpha<\lambda_1(\Sigma)$0

with weighted graph Laplacian

$\alpha<\lambda_1(\Sigma)$1

Because the function space is finite dimensional, compactness is automatic, and variational, monotone, and topological-degree methods become განსაკუთრებით sharp (Ge, 2016, Sun et al., 2021).

In the negative case $\alpha<\lambda_1(\Sigma)$2, the solvability picture is governed by a threshold $\alpha<\lambda_1(\Sigma)$3. Earlier results showed that solvability implies $\alpha<\lambda_1(\Sigma)$4, and that if $\alpha<\lambda_1(\Sigma)$5 then there exists $\alpha<\lambda_1(\Sigma)$6 such that the equation is solvable for $\alpha<\lambda_1(\Sigma)$7 and not solvable for $\alpha<\lambda_1(\Sigma)$8. The borderline question is whether the equation is solvable at $\alpha<\lambda_1(\Sigma)$9. The answer is affirmative: if $\mathcal H$0, then there exists at least one solution to

$\mathcal H$1

Moreover,

$\mathcal H$2

in the negative regime (Ge, 2016).

Brouwer degree gives a complementary global description. For the finite graph equation

$\mathcal H$3

all solutions are uniformly bounded under the standard solvability hypotheses, so the degree is well defined on the mean-zero subspace. The degree values are

$\mathcal H$4

As consequences, one recovers existence for $\mathcal H$5 when $\mathcal H$6, existence for $\mathcal H$7 when $\mathcal H$8 changes sign and $\mathcal H$9, uniqueness for $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$00 when $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$01, and a threshold-and-multiplicity picture when $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$02, $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$03, and $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$04: for $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$05 there are at least two distinct solutions, for $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$06 at least one stable solution, and for $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$07 no solution (Sun et al., 2021).

A spectral variational variant considers

$$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$08

and

$$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$09

on the mean-zero space $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$10. For any $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$11, $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$12 has a minimizer. If $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$13, then $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$14 has a minimizer for all $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$15; if $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$16, then $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$17; and at $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$18, solvability depends sharply on the sign of $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$19, with $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$20 requiring minimization on the orthogonal complement of the first eigenspace. The same pattern extends to higher eigenvalues (Li et al., 2023).

Further graph variants enlarge both the operator and the nonlinearity. For the discrete $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$21-Laplacian,

$$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$22

the operator $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$23 with $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$24 is one-to-one, onto, and order preserving after sign reversal, which supports a full sub- and supersolution theory and extends the $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$25 graph results to all $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$26 (Ge, 2016). For the nonstandard nonlinearity

$$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$27

Brouwer degree can still be computed because the reduced scalar problem has exactly three constant solutions, a fact proved by a connectivity argument specific to graphs (Yu, 2024). A different negative-curvature graph analogue,

$$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$28

exhibits a precise threshold $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$29: unique solvability for $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$30, at least two solutions for $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$31, at least one solution for $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$32, and no solution for $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$33 (Liu et al., 2020).

5. Infinite graphs, canonically compactifiable graphs, and networks

Infinite graphs require additional compactness or integrability input. On canonically compactifiable graphs, defined by

$$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$34

finite-energy functions are automatically bounded, the embedding $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$35 is compact, the Neumann Laplacian has discrete spectrum, and $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$36. In this setting the equation

$$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$37

admits a trichotomy closely paralleling the compact-manifold theory: for $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$38, solvability is equivalent to $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$39 and sign change of $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$40; for $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$41, solvability is equivalent to $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$42 being positive somewhere; for $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$43, $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$44 is necessary and one gets a threshold $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$45, with $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$46 if $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$47, $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$48. The proofs combine variational arguments, a graph Trudinger–Moser inequality, and monotone iteration (Keller et al., 2017).

For general connected infinite locally finite graphs, a heat-flow method replaces finite-dimensional compactness. The equation is written as

$$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$49

and solvability is proved under either of two hypotheses: $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$50, $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$51, and $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$52; or $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$53 is a Cheeger graph, $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$54, $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$55, and $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$56. The argument uses an exhaustion by finite full subgraphs, a parabolic flow on each finite piece, energy monotonicity, and uniform $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$57 bounds. Corollaries include global solvability of the Poisson equation on Cheeger graphs and existence for

$$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$58

when $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$59 and $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$60 (Ge et al., 2017).

Networks occupy an intermediate position between manifolds and discrete graphs. On a finite connected network $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$61, the Kazdan–Warner equation is imposed edgewise,

$$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$62

together with continuity at vertices and Kirchhoff conditions

$$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$63

The classical trichotomy persists: for $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$64, solvability is equivalent to sign change of $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$65 and $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$66; for $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$67, solvability is equivalent to positivity of $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$68 somewhere; for $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$69, there is a threshold $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$70, and if $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$71 then the critical endpoint $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$72 is also solvable. The proofs use constrained minimization for $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$73 and upper/lower solutions plus network maximum principles for $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$74 (Camilli et al., 2019).

6. Kazdan–Warner identities, conformal symmetry, and higher-order obstructions

Kazdan–Warner type equations are accompanied by integral identities that encode symmetry obstructions to solvability. In a broad variational framework, any naturally conformally variational scalar invariant $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$75 on a closed conformal manifold satisfies

$$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$76

for every conformal vector field $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$77. More generally, if $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$78 is a locally conserved symmetric $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$79-tensor and $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$80, then on a manifold with boundary one has a Pohozaev–Schoen type identity

$$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$81

This framework subsumes the classical Kazdan–Warner identity for scalar curvature, Schoen’s unification of Kazdan–Warner and Pohozaev identities, and further examples involving $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$82-curvature, renormalized volume coefficients, Gauss–Bonnet curvatures, and mean curvature of conformal immersions (Gover et al., 2010).

A recent higher-order boundary analogue appears in the conformal prescription of interior $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$83-curvature and boundary $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$84-curvature on the upper hemisphere $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$85. The model problem is

$$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$86

For every boundary-preserving conformal vector field $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$87, any solution satisfies the Kazdan–Warner type identity

$$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$88

If there exists such an $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$89 with $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$90 in $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$91, $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$92 on $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$93, and at least one inequality strict somewhere, then the boundary problem has no solution. In this sense, the Kazdan–Warner philosophy extends from prescribed Gaussian or scalar curvature to coupled interior–boundary prescription of $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$94- and $$-\Delta u = 8\pi\left(\frac{h e^u}{\int_\Sigma h e^u}-1\right),$$95-curvatures (Cruz-Blázquez et al., 23 Mar 2026).

Across these settings, Kazdan–Warner type equations are characterized less by a single formula than by a recurrent analytic and geometric pattern: exponential nonlinearity coupled to a normalization or curvature prescription, threshold behavior at critical parameters, concentration phenomena at loss of compactness, and conformal or discrete symmetry identities that sharply delimit the solvable regime.