06.03.05 · riemann-surfaces / riemann-surfaces

The prescribed-Gaussian-curvature equation on a surface (Kazdan-Warner)

shipped3 tiersLean: none

Anchor (Master): Kazdan-Warner 1974 (Ann. of Math. 99) full; Kazdan-Warner 1974 *Curvature functions for open 2-manifolds* and the S^2 Nirenberg obstruction; Moser 1973 (the Trudinger-Moser inequality); Chang-Yang 1987 (the Nirenberg problem on S^2)

Intuition Beginner

Picture a rubber sheet shaped into a smooth closed surface — a sphere, a doughnut, a pretzel. At each point the surface bends, and a single number records how it bends: positive where the surface domes outward like the top of a ball, zero where it is flat like a cylinder's side, negative where it saddles like a mountain pass. That number is the Gaussian curvature.

Now ask the designer's question in reverse. You are handed a target bending pattern — a desired curvature value at every point — and you may reshape the sheet only by stretching it locally, never tearing or shearing it, so angles between threads stay fixed. Stretching like this is called a conformal change. The puzzle is: which target patterns can you actually achieve by stretching alone?

The surprise is that you cannot achieve every pattern. The overall shape of the surface — whether it is a sphere, a doughnut, or something with more holes — secretly fixes the total amount of bending, and your stretching can only redistribute it.

Visual Beginner

Three closed surfaces sit in a row, each tinted by its curvature: a sphere glowing positive all over, a doughnut with a positive outer ring and a negative inner ring that cancel, and a two-holed surface tinted mostly negative. Below each, a gauge needle points to the same fixed total — large positive for the sphere, exactly zero for the doughnut, large negative for the two-holed surface — no matter how the local tints are rearranged by stretching.

The picture makes the constraint visible. Stretching the rubber moves bright and dark patches around, brightening one region while darkening another, but the gauge needle never moves. The number of holes sets the needle, and the needle decides which curvature patterns are even possible before any detailed reshaping is attempted.

Worked example Beginner

Take the round sphere of radius one. Its Gaussian curvature is the constant value one at every point, and its total area is four times pi. Multiply curvature by area in this constant case and you get four pi: that is the total bending of a sphere, and the number of holes for a sphere is zero.

Now try to reshape this sphere, by stretching only, into a surface whose curvature is negative everywhere. Add up the target: a negative number at every point times a positive area gives a negative total. But stretching cannot change the total away from four pi, a positive value. So the target is unreachable. A sphere refuses to be bent into an everywhere-saddle shape, however cleverly you stretch it.

Compare the doughnut. Its total bending is zero. A target curvature that is positive on the outer rim and negative on the inner throat, balanced so the positive and negative parts cancel, has total zero and matches what the doughnut allows. Such a target stands a chance — and indeed the everyday doughnut already displays exactly this mix. The lesson is the same each time: first check the total against the shape, then ask about the details.

Check your understanding Beginner

Formal definition Intermediate+

Let $M$ be a compact surface without boundary carrying a smooth background Riemannian metric $g_{0}$ with Gaussian curvature $K_{0}$ and Laplace-Beltrami operator $Δ_{0}$ (with the geometers' sign convention, $Δ_{0} = div grad$ , so $Δ_{0}$ is negative-definite on nonconstant functions). A conformal metric is one of the form $g = e^{2 u} g_{0}$ for a smooth function $u : M \to R$ ; the factor $e^{2 u}$ rescales lengths pointwise while preserving angles, and $u$ is the conformal factor.

Definition (the prescribed-curvature equation). Given a target function $K : M \to R$ , the metric $g = e^{2 u} g_{0}$ has Gaussian curvature $K$ exactly when the conformal factor satisfies the semilinear elliptic equation $$ -\Delta_0 u + K_0 = K, e^{2u} , $$ equivalently $Δ_{0} u = K_{0} - K e^{2 u}$ . This is the Gauss-curvature equation (a Liouville-type equation). The unknown is $u$ ; the data are the background curvature $K_{0}$ and the target $K$ .

The derivation is the conformal transformation law for curvature in two dimensions: under $g = e^{2 u} g_{0}$ the Gaussian curvatures relate by $K_{g} = e^{- 2 u} (K_{0} - Δ_{0} u)$ . Setting $K_{g} = K$ and clearing the exponential yields the equation. The area form transforms by $d A_{g} = e^{2 u} d A_{0}$ , so the curvature integral transforms as $$ \int_M K, dA_g = \int_M K e^{2u}, dA_0 = \int_M (K_0 - \Delta_0 u), dA_0 = \int_M K_0, dA_0 , $$ since the Laplacian of any smooth function integrates to zero on a closed surface. This is the analytic face of the Gauss-Bonnet constraint: the total curvature is invariant under conformal change and equals $2 π χ (M)$ , where $χ (M) = 2 - 2 γ$ is the Euler characteristic and $γ$ the number of holes.

The trichotomy data. The sign of $χ (M)$ partitions the problem: $χ > 0$ for the sphere, $χ = 0$ for the torus, $χ < 0$ for every surface of two or more holes. Each case has its own solvability theory, governed by the constant case where $K \equiv const$ recovers the constant-curvature representative and the uniformization metric of 06.03.04.

Counterexamples to common slips

A target $K$ that is positive everywhere on a torus ( $χ = 0$ ) is unsolvable: $\int_{M} K d A_{g} > 0$ but Gauss-Bonnet pins the integral to $2 π χ = 0$ . The sign of the average is a hard obstruction, not a regularity nuisance.
On the sphere the sign condition $\int K d A > 0$ is necessary but not sufficient. The Kazdan-Warner obstruction (below) rules out further targets, such as $K = 1 + ε x_{1}$ for the height function $x_{1}$ , even though their average is positive. This is the Nirenberg problem.
The equation is genuinely nonlinear: the exponential $e^{2 u}$ means solutions do not superpose, and the linear Dirichlet theory of 06.01.24 solves only the model with $K e^{2 u}$ replaced by a fixed source.

Key theorem with proof Intermediate+

Theorem (Kazdan-Warner, $χ < 0$ case). Let $M$ be a compact surface with $χ (M) < 0$ , normalised so the background metric has constant curvature $K_{0} = - 1$ and total area $- 2 π χ (M)$ . A smooth function $K : M \to R$ is the Gaussian curvature of some conformal metric $e^{2 u} g_{0}$ if and only if $K < 0$ somewhere on $M$ .

Proof. Necessity is Gauss-Bonnet: $\int_{M} K d A_{g} = 2 π χ (M) < 0$ , so $K$ must be negative somewhere, for a function that is nowhere negative cannot have a negative integral. For sufficiency, assume $K < 0$ at some point. The strategy is the method of sub- and super-solutions applied to $$ \Delta_0 u = K_0 - K e^{2u} = -1 - K e^{2u}. $$ A super-solution $\overline{u}$ satisfies $Δ_{0} \overline{u} \leq - 1 - K e^{2 \overline{u}}$ and a sub-solution $\underline{u}$ satisfies $Δ_{0} \underline{u} \geq - 1 - K e^{2 \underline{u}}$ , with $\underline{u} \leq \overline{u}$ ; the monotone-iteration scheme then produces a genuine solution trapped between them.

Build the super-solution as a large constant. For a constant $u = c$ , the left side is $0$ and the requirement reads $0 \leq - 1 - K e^{2 c}$ , i.e. $K e^{2 c} \leq - 1$ . Where $K < 0$ this holds for $c$ large; where $K \geq 0$ it fails. So a constant alone will not serve everywhere, and one instead solves an auxiliary linear equation. Let $\overset{ˉ}{K}$ be the average of $K$ and pick $v$ solving $Δ_{0} v = K - \overset{ˉ}{K}$ (solvable since the right side has mean zero, by the linear theory descending from 06.01.24 and 06.01.11). Then a super-solution of the form $\overline{u} = v + c$ for suitable large $c$ , and a sub-solution $\underline{u} = v + c^{'}$ for suitable very negative $c^{'}$ , can be checked directly: as $c^{'} \to - \infty$ the exponential $e^{2 \underline{u}} \to 0$ uniformly, so $Δ_{0} \underline{u} = K - \overset{ˉ}{K} \geq - 1 - K e^{2 \underline{u}}$ holds once $e^{2 \underline{u}}$ is small enough and $\overset{ˉ}{K} = 2 π χ / Area < 0$ . With $\underline{u} \leq \overline{u}$ arranged by taking $c$ large and $c^{'}$ small, monotone iteration starting from $\overline{u}$ converges in $C^{2, α}$ to a solution $u$ . Elliptic regularity bootstraps $u$ to smoothness. $□$

Bridge. This existence theorem builds toward the curvature route to uniformization in 06.03.04: take the target $K$ to be the constant $- 1$ and the theorem hands a conformal metric of constant negative curvature on every higher-genus surface, which is exactly the hyperbolic representative the uniformization theorem promises. The foundational reason the negative case is the friendly one is that the nonlinearity $- K e^{2 u}$ is monotone in the right direction when $K < 0$ — increasing $u$ makes the right side more negative, matching the maximum principle — so the sub/super-solution sandwich closes; this is exactly the structural advantage that fails when $χ \geq 0$ . The construction generalises the linear Dirichlet solvability of 06.01.24 to a semilinear source, and it is dual to the parabolic approach of 06.03.04, where the normalized Ricci flow reaches the same constant-curvature metric as a long-time limit rather than a variational minimum. Putting these together, the bridge is that the sign of the Euler characteristic, read through Gauss-Bonnet, decides both whether a constant-curvature metric exists and which analytic machine — elliptic sandwich here, parabolic flow there — delivers it, and the central insight that the average of $K$ must match $2 π χ$ reappears again in 03.02.05 as the global obstruction that no amount of local curvature engineering can evade.

Exercises Intermediate+

Exercise 1 (easy, short-answer).

Derive the Gauss-Bonnet constraint $\int_{M} K d A_{g} = \int_{M} K_{0} d A_{0}$ directly from the prescribed-curvature equation by integrating it against the background area form.

Hint

Integrate $- Δ_{0} u + K_{0} = K e^{2 u}$ over $M$ and use that the integral of a Laplacian over a closed surface vanishes. Recall $d A_{g} = e^{2 u} d A_{0}$ .

Answer

Integrating the equation: $\int_{M} (- Δ_{0} u) d A_{0} + \int_{M} K_{0} d A_{0} = \int_{M} K e^{2 u} d A_{0}$ . The first integral is zero, since the Laplacian of a smooth function on a closed surface integrates to zero (it is a divergence). The right side is $\int_{M} K d A_{g}$ because $d A_{g} = e^{2 u} d A_{0}$ . Hence $\int_{M} K_{0} d A_{0} = \int_{M} K d A_{g}$ . Both equal $2 π χ (M)$ by Gauss-Bonnet for the respective metrics. Rubric: full credit for the vanishing-Laplacian step and the area-form substitution.

Exercise 3 (medium, symbolic).

On the flat torus $C /Λ$ with $K_{0} = 0$ , write down the prescribed-curvature equation and show that a smooth target $K$ admits a solution only if $\int_{M} K e^{2 u} d A_{0} = 0$ for the solution, and deduce that $K$ must change sign (or vanish) unless $K \equiv 0$ .

Hint

With $K_{0} = 0$ the equation is $Δ_{0} u = - K e^{2 u}$ . Integrate over the torus.

Answer

The equation is $- Δ_{0} u = K e^{2 u}$ , i.e. $Δ_{0} u = - K e^{2 u}$ . Integrating over the closed torus, the left side gives zero, so $\int_{M} K e^{2 u} d A_{0} = 0$ , that is $\int_{M} K d A_{g} = 0$ . Since $e^{2 u} > 0$ is a positive weight, an integral of $K$ against a positive weight vanishing forces $K$ to take both signs or be identically zero. Hence on the torus a nonzero target must change sign. The constant target $K \equiv 0$ is the flat metric itself. Rubric: full credit for the zero-mean weighted integral and the sign-change deduction.

Exercise 4 (medium, short-answer).

In the $χ < 0$ proof, explain why a single large constant fails as a super-solution and why the auxiliary function $v$ solving $Δ_{0} v = K - \overset{ˉ}{K}$ repairs the construction.

Hint

A constant has zero Laplacian, so the super-solution inequality becomes a pointwise condition on $K e^{2 c}$ that can fail where $K \geq 0$ . The function $v$ absorbs the non-constant part of $K$ .

Answer

A constant $u = c$ has $Δ_{0} u = 0$ , so the super-solution inequality $0 \leq - 1 - K e^{2 c}$ requires $K e^{2 c} \leq - 1$ pointwise, which fails wherever $K \geq 0$ . The torus-style obstruction is that $K$ need not be sign-definite. Solving $Δ_{0} v = K - \overset{ˉ}{K}$ (possible because $K - \overset{ˉ}{K}$ has mean zero) lets one write $u = v + const$ , so the left side $Δ_{0} u = K - \overset{ˉ}{K}$ now carries the variable part of $K$ ; the remaining inequality involves only the average $\overset{ˉ}{K} = 2 π χ / Area < 0$ , which has the favorable sign. The sub-solution then comes from sending the additive constant to $- \infty$ , killing the exponential. Rubric: full credit for the zero-Laplacian failure and the mean-zero solvability of the $v$ -equation.

Exercise 5 (hard, short-answer).

State the Kazdan-Warner obstruction on $S^{2}$ and use it to exhibit a positive function $K$ on the sphere that cannot be a conformal curvature despite having positive integral.

Hint

The obstruction says $\int_{S^{2}} ⟨ \nabla K, \nabla x_{j} ⟩ e^{2 u} d A_{0} = 0$ for each ambient coordinate $x_{j}$ restricted to the sphere (a first spherical harmonic). Choose $K$ so this fails for every conformal factor's sign.

Answer

The Kazdan-Warner identity states that if $e^{2 u} g_{0}$ on $S^{2}$ (with $g_{0}$ round) has curvature $K$ , then for each restriction $x_{j}$ of an ambient linear coordinate — a first eigenfunction of the round Laplacian — one has $\int_{S^{2}} ⟨ \nabla K, \nabla x_{j} ⟩_{g_{0}} e^{2 u} d A_{0} = 0$ . Take $K = 1 + ε x_{1}$ with small $ε > 0$ , which is positive on the sphere and has positive average. Then $\nabla K = ε \nabla x_{1}$ , so the integrand becomes $ε ∣\nabla x_{1} ∣^{2} e^{2 u} > 0$ wherever $\nabla x_{1} \neq = 0$ , forcing the integral to be strictly positive rather than zero. The obstruction is violated, so no conformal factor $u$ can realise this $K$ . Rubric: full credit for stating the identity and producing a monotone $K = 1 + ε x_{1}$ whose integral against $\nabla x_{1}$ cannot vanish.

Exercise 6 (hard, short-answer).

Set up the variational functional whose critical points solve the $χ < 0$ equation and identify the role of the Trudinger-Moser inequality in the sphere case.

Hint

Consider $E (u) = \frac{1}{2} \int ∣\nabla u ∣^{2} + \int K_{0} u - \frac{1}{2} \int K e^{2 u}$ (signs adapted to the convention). Coercivity needs control of $\int e^{2 u}$ by the Dirichlet energy.

Answer

A solution of $- Δ_{0} u + K_{0} = K e^{2 u}$ is a critical point of $$ E(u) = \tfrac12 \int_M |\nabla u|^2, dA_0 + \int_M K_0 u, dA_0 - \tfrac12 \int_M K, e^{2u}, dA_0, $$ since the Euler-Lagrange equation reads $- Δ_{0} u + K_{0} - K e^{2 u} = 0$ . For $χ < 0$ with $K < 0$ , the term $- \frac{1}{2} \int K e^{2 u} > 0$ helps coercivity and a minimizer exists in $H^{1}$ . In the borderline $S^{2}$ case the energy sits at the critical exponent for the exponential nonlinearity, and the Trudinger-Moser inequality $\int e^{2 u} \leq C exp (\frac{1}{2 π} \int ∣\nabla u ∣^{2} + 2 \overset{u}{ˉ})$ is exactly the sharp embedding that controls $\int e^{2 u}$ by the Dirichlet energy with the best constant; loss of compactness at that constant is the analytic reason the sphere case is delicate and the source of the Kazdan-Warner obstruction. Rubric: full credit for the correct functional and for naming Trudinger-Moser as the borderline-compactness tool.

Advanced results Master

The full Kazdan-Warner theorem partitions solvability by the sign of $χ (M)$ into a clean trichotomy, and the analytic character of the equation changes completely across the three cases. For $χ (M) < 0$ the equation $Δ_{0} u = K_{0} - K e^{2 u}$ has a coercive variational structure, the nonlinearity pushes in the direction the maximum principle wants, and the sub/super-solution method of the Key theorem settles existence with the single sharp hypothesis $K < 0$ somewhere; the constant-curvature representative $K \equiv - 1$ is the hyperbolic uniformization metric of 06.03.04, and the solution is in fact unique. This is the analytic heart of the geometric-analysis reading of surface theory: every higher-genus surface is uniformized by solving one semilinear elliptic equation.

For $χ (M) = 0$ the torus carries $K_{0} = 0$ and the equation is $Δ_{0} u = - K e^{2 u}$ . Kazdan-Warner prove that a nonconstant $K$ is solvable if and only if $K$ changes sign and $\int_{M} K d A_{0} < 0$ — a mean condition on the background area form, not merely a pointwise sign change. The asymmetry (the strict negativity of the unweighted mean, even though the weighted mean $\int K e^{2 u}$ must vanish) comes from a delicate use of the maximum principle that distinguishes $K e^{2 u}$ from $K$ ; the constant solution $K \equiv 0$ is the flat metric. The borderline nature of $χ = 0$ — neither the favorable coercivity of $χ < 0$ nor the conformal-group degeneracy of $χ > 0$ — makes it the most arithmetically rigid of the three.

For $χ (M) > 0$ the sphere is governed by the Nirenberg problem, and here the sign condition $\int K d A > 0$ is necessary but far from sufficient. Two obstructions intervene. First, the conformal group of $S^{2}$ is noncompact (the Mobius transformations), so the variational functional loses compactness exactly at the Trudinger-Moser critical constant, and minimizing sequences can concentrate at a point. Second, there is the explicit Kazdan-Warner identity: if $K$ is the curvature of $e^{2 u} g_{0}$ on the round sphere then $\int_{S^{2}} ⟨ \nabla K, \nabla x_{j} ⟩ e^{2 u} d A_{0} = 0$ for each first spherical harmonic $x_{j}$ . This rules out, for example, every $K$ that is a strictly monotone function of a single linear height, even with positive average. Positive resolution requires further hypotheses — Moser's antipodal-symmetry condition, or the Chang-Yang degree-theoretic count of solutions — and the problem remains the richest of the three.

A unifying perspective is the total-curvature flow picture of 06.03.04: the normalized Ricci flow $\partial_{t} g = (r - R) g$ on a surface (with $R$ the scalar curvature, twice the Gaussian, and $r$ its mean) is the gradient flow of a curvature functional, and its long-time limit is the constant-curvature metric. Where the elliptic Kazdan-Warner theory produces the prescribed-curvature metric as a critical point, the parabolic flow produces the constant-curvature special case as a limit; the two are the stationary and evolutionary faces of the same conformal-geometry problem, and the Euler-characteristic trichotomy governs the convergence behavior of the flow exactly as it governs the solvability of the equation.

Synthesis. The prescribed-Gaussian-curvature equation is the foundational reason that conformal geometry on a surface reduces to a single semilinear elliptic PDE, and the sign of the Euler characteristic is the central insight that organizes its entire solvability theory. The Gauss-Bonnet constraint $\int K d A = 2 π χ$ is exactly the necessary condition in all three regimes, and it generalises the elementary observation that total curvature is a topological invariant into a sharp algebraic obstruction on which targets are reachable. Putting these together, the trichotomy is the analytic shadow of the geometric uniformization trichotomy: $χ < 0$ gives coercive existence and the hyperbolic metric, $χ = 0$ gives a rigid mean condition and the flat metric, $χ > 0$ gives the obstructed Nirenberg problem and the spherical metric — and this is exactly the same three-way split that 06.03.04 reaches by the normalized Ricci flow and that 06.03.03 reaches by potential theory. The bridge is that all three routes — elliptic sandwich, parabolic flow, Green's-function potential theory — must respect the one invariant $2 π χ$ , so the equation studied here is dual to the flow of 06.03.04 and is the analytic engine beneath the geometric statements of 06.03.03, with the Kazdan-Warner identity standing as the sharpest measure of how much the sphere's conformal symmetry obstructs naive curvature prescription.

Full proof set Master

The $χ < 0$ existence theorem is proved in full in the Key theorem section. The remaining structural claims are recorded here.

Proposition (conformal transformation of Gaussian curvature in two dimensions). If $g = e^{2 u} g_{0}$ on a surface, then the Gaussian curvatures relate by $K_{g} = e^{- 2 u} (K_{0} - Δ_{0} u)$ , where $Δ_{0}$ is the Laplace-Beltrami operator of $g_{0}$ .

Proof. Work in an isothermal coordinate chart for $g_{0}$ , in which $g_{0} = e^{2 φ} (d x^{2} + d y^{2})$ for some local function $φ$ . In two dimensions the Gaussian curvature of a conformally flat metric $e^{2 ψ} (d x^{2} + d y^{2})$ is $K = - e^{- 2 ψ} Δ_{eucl} ψ$ , where $Δ_{eucl} = \partial_{x}^{2} + \partial_{y}^{2}$ is the flat Laplacian. Applying this to $g_{0}$ (so $ψ = φ$ ) gives $K_{0} = - e^{- 2 φ} Δ_{eucl} φ$ , and to $g = e^{2 (u + φ)} (d x^{2} + d y^{2})$ (so $ψ = u + φ$ ) gives $$ K_g = -e^{-2(u+\varphi)}\Delta_{\mathrm{eucl}}(u + \varphi) = e^{-2u}\big( -e^{-2\varphi}\Delta_{\mathrm{eucl}} u - e^{-2\varphi}\Delta_{\mathrm{eucl}}\varphi\big). $$ The Laplace-Beltrami operator of $g_{0}$ in this chart is $Δ_{0} = e^{- 2 φ} Δ_{eucl}$ , so the first term is $- e^{- 2 u} Δ_{0} u$ and the second is $e^{- 2 u} K_{0}$ . Hence $K_{g} = e^{- 2 u} (K_{0} - Δ_{0} u)$ . The expression is coordinate-independent because both sides are scalar functions. $□$

Proposition (Gauss-Bonnet invariance of total curvature under conformal change). For any conformal metric $g = e^{2 u} g_{0}$ on a closed surface $M$ , $\int_{M} K_{g} d A_{g} = \int_{M} K_{0} d A_{0} = 2 π χ (M)$ .

Proof. By the previous proposition $K_{g} e^{2 u} = K_{0} - Δ_{0} u$ , and $d A_{g} = e^{2 u} d A_{0}$ , so $K_{g} d A_{g} = (K_{0} - Δ_{0} u) d A_{0}$ . Integrate over $M$ : $\int_{M} K_{g} d A_{g} = \int_{M} K_{0} d A_{0} - \int_{M} Δ_{0} u d A_{0}$ . The last integral is the integral of a divergence over a closed manifold, hence zero by the divergence theorem with no boundary. So $\int_{M} K_{g} d A_{g} = \int_{M} K_{0} d A_{0}$ . That this common value equals $2 π χ (M)$ is the Gauss-Bonnet theorem for $g_{0}$ . $□$

Proposition (Kazdan-Warner identity / first-harmonic obstruction on $S^{2}$ ). Let $g_{0}$ be the round metric on $S^{2}$ and $x_{j}$ ( $j = 1, 2, 3$ ) the restrictions of the ambient linear coordinates, which are first eigenfunctions of $Δ_{0}$ with $Δ_{0} x_{j} = - 2 x_{j}$ . If $e^{2 u} g_{0}$ has Gaussian curvature $K$ , then $\int_{S^{2}} ⟨ \nabla K, \nabla x_{j} ⟩_{g_{0}} e^{2 u} d A_{0} = 0$ .

Proof sketch. Multiply the equation $- Δ_{0} u + 1 = K e^{2 u}$ (with $K_{0} = 1$ on the unit sphere) by $⟨ \nabla x_{j}, \nabla (\cdot)⟩$ in the appropriate weighted sense and integrate by parts, using the conformal Killing structure of $\nabla x_{j}$ on $S^{2}$ : the gradient of a first eigenfunction generates a conformal vector field, so $\nabla x_{j}$ is a conformal Killing field. Pairing the equation against this field and integrating, the contributions from $Δ_{0} u$ and the constant term cancel by the conformal-Killing identity, leaving precisely $\int_{S^{2}} ⟨ \nabla K, \nabla x_{j} ⟩ e^{2 u} d A_{0} = 0$ . The mechanism is that the first eigenspace is the kernel of the linearization, so the nonlinear equation projects to zero against it. (Full computation: Kazdan-Warner 1974, §6.) $□$

Connections Master

Sectional, Ricci, and scalar curvature 03.02.05 supplies the very notion of Gaussian curvature that this unit prescribes, and it carries the Gauss-Bonnet theorem in the form $\int_{M} K d A = 2 π χ (M)$ that is the single global obstruction governing every case of the trichotomy. The conformal transformation law $K_{g} = e^{- 2 u} (K_{0} - Δ_{0} u)$ proved here is the two-dimensional specialization of the general curvature-under-conformal-change formulas of that unit, and the constraint that no local curvature engineering can move the total away from $2 π χ$ is the recurring theme that ties this PDE back to the topology fixed there.

Uniformization via constant-curvature conformal metrics and the Ricci flow 06.03.04 is the immediate geometric consumer of this analytic engine. Setting the target $K$ to the constant $+ 1$ , $0$ , or $- 1$ in the prescribed-curvature equation is exactly the problem of finding the uniformizing metric, and the $χ < 0$ existence theorem of this unit is the elliptic proof of hyperbolic uniformization; the normalized Ricci flow of that unit reaches the same constant-curvature metric as a parabolic limit, so the two units are the stationary and evolutionary faces of one conformal-geometry problem, split by the same Euler-characteristic trichotomy.

The uniformization theorem 06.03.03 states the same three-model conclusion — sphere, plane, disc — but proves it by potential theory through the Green's function and the Dirichlet problem rather than by the semilinear curvature equation. This unit is the differential-geometric complement: it reaches the constant-curvature representatives by solving a nonlinear elliptic PDE, and it makes explicit, through Gauss-Bonnet, why the sign of the curvature is forced by the topology — a fact that the potential-theoretic route encodes more implicitly in the type of the universal cover.

The Dirichlet problem and the Perron method 06.01.24 provide the linear elliptic solvability that the sub/super-solution scheme bootstraps. The auxiliary equation $Δ_{0} v = K - \overset{ˉ}{K}$ in the $χ < 0$ proof is solved by exactly the linear theory developed there, and the monotone-iteration scheme that closes the semilinear sandwich is the nonlinear extension of the linear maximum principle and Perron's method of subharmonic functions.

Harmonic functions in the plane and the Laplacian 06.01.11 is the analytic bedrock: the operator $Δ_{0}$ at the center of the prescribed-curvature equation is the Laplace-Beltrami operator whose flat-plane prototype, with its mean-value property and maximum principle, is established there. The vanishing of $\int_{M} Δ_{0} u d A_{0}$ on a closed surface — the step that makes Gauss-Bonnet conformally invariant — is the global counterpart of the local mean-value identity for harmonic functions.

Historical & philosophical context Master

The question of which functions arise as Gaussian curvatures of a metric in a fixed conformal class was settled, for compact surfaces, by Jerry Kazdan and Frank Warner in their 1974 paper Curvature functions for compact 2-manifolds (Annals of Mathematics 99, 14–47) ^{[Kazdan-Warner 1974]}, which isolated the trichotomy by the sign of the Euler characteristic and discovered the obstruction identity that now bears their names. The $χ < 0$ existence result had been obtained by Melvyn Berger in 1971 (Riemannian structures of prescribed Gaussian curvature for compact 2-manifolds, J. Differential Geometry 5, 325–332) ^{[Berger 1971]} using the variational method, and the sharp analytic tool for the borderline sphere case is Jürgen Moser's 1973 form of the Trudinger inequality ^{[Moser 1973]}, whose best constant is exactly where compactness is lost on $S^{2}$ . The lineage runs back to Poincaré and Koebe's uniformization and forward to the Yamabe problem in higher dimensions, of which the surface case is the conformally critical prototype.

The philosophical content is a clean instance of how topology constrains analysis. The Euler characteristic is a discrete, combinatorial invariant; the curvature is a continuous, infinitely-variable field; and the Gauss-Bonnet theorem welds them so tightly that the discrete number dictates the global sign the continuous field must carry. Kazdan and Warner's discovery that the sign condition is sufficient for $χ \leq 0$ but insufficient for $χ > 0$ — that the sphere hides a further obstruction invisible to the integral — taught geometers that the symmetry group of the model space is itself an analytic obstacle: the noncompactness of the Möbius group on $S^{2}$ is the precise reason the variational problem loses compactness, and the obstruction identity is the algebraic fingerprint of that lost symmetry. The surface case thus became the laboratory in which the interplay of conformal invariance, critical Sobolev exponents, and concentration-compactness was first understood, before those ideas reshaped geometric analysis in every dimension.

Bibliography Master

@article{kazdanwarner1974,
  author  = {Kazdan, Jerry L. and Warner, Frank W.},
  title   = {Curvature functions for compact 2-manifolds},
  journal = {Annals of Mathematics},
  volume  = {99},
  number  = {1},
  pages   = {14--47},
  year    = {1974}
}

@article{berger1971,
  author  = {Berger, Melvyn S.},
  title   = {Riemannian structures of prescribed {G}aussian curvature for compact 2-manifolds},
  journal = {Journal of Differential Geometry},
  volume  = {5},
  number  = {3-4},
  pages   = {325--332},
  year    = {1971}
}

@incollection{moser1973,
  author    = {Moser, J\"urgen},
  title     = {On a nonlinear problem in differential geometry},
  booktitle = {Dynamical Systems},
  editor    = {Peixoto, M. M.},
  publisher = {Academic Press},
  pages     = {273--280},
  year      = {1973}
}

@book{kazdan1985,
  author    = {Kazdan, Jerry L.},
  title     = {Prescribing the Curvature of a Riemannian Manifold},
  series    = {CBMS Regional Conference Series in Mathematics},
  volume    = {57},
  publisher = {American Mathematical Society},
  year      = {1985}
}

@book{aubin1998,
  author    = {Aubin, Thierry},
  title     = {Some Nonlinear Problems in Riemannian Geometry},
  series    = {Springer Monographs in Mathematics},
  publisher = {Springer-Verlag, Berlin},
  year      = {1998}
}

@article{changyang1987,
  author  = {Chang, Sun-Yung A. and Yang, Paul C.},
  title   = {Prescribing {G}aussian curvature on {$S^2$}},
  journal = {Acta Mathematica},
  volume  = {159},
  pages   = {215--259},
  year    = {1987}
}

Prerequisites

06.01.24
06.01.11
03.02.05

Tier anchors

beginner: Kazdan 1985 *Prescribing the Curvature of a Riemannian Manifold* (CBMS 57) Ch. 1-2 (the conformal-factor picture, informal); do Carmo 1976 *Differential Geometry of Curves and Surfaces* §4-5 (Gauss curvature, Gauss-Bonnet)
intermediate: Kazdan-Warner 1974 *Curvature functions for compact 2-manifolds* (Ann. of Math. 99) §§1-3 (the equation, the trichotomy by sign of chi); Aubin 1998 *Some Nonlinear Problems in Riemannian Geometry* Ch. 5
master: Kazdan-Warner 1974 (Ann. of Math. 99) full; Kazdan-Warner 1974 *Curvature functions for open 2-manifolds* and the S^2 Nirenberg obstruction; Moser 1973 (the Trudinger-Moser inequality); Chang-Yang 1987 (the Nirenberg problem on S^2)

References

Kazdan, Warner — Curvature functions for compact 2-manifolds (Annals of Mathematics 99, 1974) · pp. 14--47; the conformal Gauss-curvature equation, the trichotomy by sign of the Euler characteristic, and the chi=0 / chi>0 obstruction identity
Berger — Riemannian structures of prescribed Gaussian curvature for compact 2-manifolds (J. Diff. Geom. 5, 1971) · pp. 325--332; the chi<0 existence theorem and the variational method
Moser — On a nonlinear problem in differential geometry (1973) · in Dynamical Systems (ed. Peixoto), pp. 273--280; the sharp Trudinger-Moser inequality used for the S^2 case

Estimated time

beginner: 18m
intermediate: 45m
master: 85m