06.03.04 · riemann-surfaces / riemann-surfaces

Uniformization via constant-curvature conformal metrics and Ricci flow on surfaces

shipped3 tiersLean: none

Anchor (Master): Hamilton 1988 *The Ricci flow on surfaces* (Contemp. Math. 71); Chow 1991 *The Ricci flow on the 2-sphere* (J. Diff. Geom. 33); Kazdan-Warner 1974 (Ann. of Math. 99); Chow-Lu-Ni *Hamilton's Ricci Flow* Ch. 5

Intuition Beginner

A Riemann surface comes with a notion of angle but not, on its own, a fixed notion of distance. You are free to choose a ruler — a metric — and the only rule is that it must agree with the angles already there. Picking such a ruler is called choosing a conformal metric: you keep the shapes of tiny figures and only decide how much to stretch or shrink at each point. The stretching factor is allowed to vary from place to place.

Now ask a greedy question. Among all these allowed rulers, is there one that makes the surface look as uniform as possible — bending the same amount everywhere? A surface has a number at each point, its Gaussian curvature, that says how much it bows: positive like a dome, zero like a flat sheet, negative like a saddle. Uniformization is the claim that you can always rescale your ruler so that this bowing becomes one single constant across the whole surface.

Which constant you land on is not your choice. It is fixed by the shape of the surface — by how many handles it has. A sphere is forced to be positively curved, a doughnut perfectly flat, and anything with two or more handles negatively curved. You cannot rescale your way past the topology; the number of holes decides the sign.

There are two ways to find the magic ruler. One solves a single equation directly for the stretching factor. The other lets the surface relax: you watch the ruler flow over time, smoothing its own curvature like heat spreading through a metal plate, until the bowing evens out to a constant. Both arrive at the same uniform surface.

Visual Beginner

Alt text: On the left, three lumpy surfaces with curvature that varies from point to point: a deformed sphere, a bumpy torus, and a warped double-torus. An arrow labelled "Ricci flow" points rightward from each to a smoothed version. On the right, the sphere is perfectly round with curvature plus one, the torus is perfectly flat with curvature zero, and the double-torus has even negative curvature minus one. The picture shows that rescaling the ruler within the same conformal class flattens the curvature to a single constant, whose sign is decided by the number of handles.

Worked example Beginner

Take the round unit sphere, where the curvature is already the constant value $1$ everywhere, and try to rescale it to be flat. Suppose you could find a stretching factor that turns it into a surface of curvature $0$ in the same conformal class. There is a budget you must respect, set by a counting law: when you add up the curvature over the whole surface, weighted by area, you always get the same total, $4 π$ for a sphere, no matter which ruler you chose. This total depends only on the shape, not on the ruler.

If the rescaled curvature were $0$ everywhere, the weighted total would be $0$ times the area, which is $0$ . But the budget demands $4 π$ . Since $0$ is not $4 π$ , no flattening ruler exists. The sphere cannot be made flat while keeping its angles.

Run the same check on a doughnut. Its budget total is $0$ . A constant curvature $c$ would give $c$ times the area, and to hit $0$ you need $c = 0$ . So a doughnut's uniform ruler must be flat — and a flat one does exist, the one you get by rolling up a sheet of paper.

What this tells us: the counting law forces the sign of the constant curvature before you solve anything. The sphere is stuck at positive, the torus at zero, and the budget is the gatekeeper.

Check your understanding Beginner

Formal definition Intermediate+

Let $M$ be a compact oriented surface carrying a background Riemannian metric $g_{0}$ with Gaussian curvature $K_{0}$ and Laplace-Beltrami operator $Δ_{g_{0}}$ . Sign convention: $Δ_{g_{0}}$ is the geometer's (negative-spectrum) Laplacian, so on flat $R^{2}$ it is $\partial_{x}^{2} + \partial_{y}^{2}$ and $Δ_{g_{0}} u \leq 0$ at an interior maximum. The Gaussian curvature $K$ here is the sectional curvature of the tangent plane, equal to half the scalar curvature $R$ , so $R = 2 K$ ; this matches unit 03.02.05. A conformal metric in the class of $g_{0}$ is one of the form $$ g = e^{2u} g_0, \qquad u \in C^\infty(M), $$ where the positive scalar $e^{2 u}$ is the conformal factor. Conformal metrics are precisely those that induce the same angles as $g_{0}$ ; on a Riemann surface every metric compatible with the complex structure is conformal to any other, so the conformal class is the data the complex structure provides.

The defining computation of the subject is the conformal change law for Gaussian curvature. Under $g = e^{2 u} g_{0}$ on a surface, $$ K_g = e^{-2u}\big(K_0 - \Delta_{g_0} u\big). $$ Prescribing that $g$ have a target curvature $K$ (a given function on $M$ ) is therefore the prescribed-Gaussian-curvature equation, a semilinear elliptic PDE for the conformal factor: $$ \Delta_{g_0} u = K_0 - K, e^{2u}. $$ When the target $K$ is a constant this is the Liouville equation in geometric form. The two normalisations used below are: $K \equiv - 1$ when $χ (M) < 0$ , giving $Δ_{g_{0}} u = K_{0} + e^{2 u}$ ; $K \equiv 0$ when $χ (M) = 0$ ; and $K \equiv + 1$ when $χ (M) > 0$ .

Definition (constant-curvature uniformization). A conformal metric $g = e^{2 u} g_{0}$ on $M$ is a uniformizing metric if $K_{g}$ is constant. The trichotomy asserts that the sign of that constant is forced: by the Gauss-Bonnet theorem, $$ \int_M K_g , dA_g = 2\pi \chi(M), $$ and since $area_{g} (M) > 0$ , a constant $K_{g}$ must share the sign of $χ (M)$ . So $K_{g} > 0$ iff $M ≅ S^{2}$ ( $χ = 2$ ), $K_{g} = 0$ iff $M$ is a torus ( $χ = 0$ ), and $K_{g} < 0$ iff $M$ has genus $\geq 2$ ( $χ < 0$ ). After scaling the metric, the three normalised constants are $+ 1$ , $0$ , $- 1$ .

Definition (normalized Ricci flow on a surface). On a surface the Ricci tensor is $Ric = K g = \frac{1}{2} R g$ , so the unnormalized Ricci flow $\partial_{t} g = - 2 Ric = - R g$ shrinks area. The normalized Ricci flow (equivalently the Yamabe flow in dimension two) adds back the average to preserve area: $$ \frac{\partial g}{\partial t} = (r - R), g, \qquad r := \frac{\int_M R,dA}{\int_M dA} = \frac{4\pi\chi(M)}{\mathrm{area}(M)} = \text{const in } t. $$ Because the flow stays inside the conformal class — writing $g (t) = e^{2 u (t)} g_{0}$ , the equation becomes the scalar parabolic PDE $\partial_{t} u = e^{- 2 u} Δ_{g_{0}} u + r - K_{0} e^{- 2 u}$ — a fixed point of the flow is exactly a metric with $R \equiv r$ , i.e. constant curvature. The non-examples that motivate care: an arbitrary conformal factor need not solve the Liouville equation, and a metric of nonconstant curvature is never a flow fixed point even though it may have the right total curvature.

Counterexamples to common slips

The sign of the constant is not free. One cannot prescribe $K \equiv + 1$ on a genus-2 surface: Gauss-Bonnet would force $\int_{M} e^{2 u} d A_{g_{0}} = 2 π χ < 0$ , impossible for a positive integrand. The target sign must match $χ$ .
Conformal is stronger than "same total curvature". Two metrics with the same $\int K d A$ need not be conformal; conformality is a pointwise angle condition, while Gauss-Bonnet is a single integral constraint.
Ricci flow in 2D is not the unnormalized heat flow of $u$ . The equation $\partial_{t} u = e^{- 2 u} Δ_{g_{0}} u + \dots$ has the curvature-dependent coefficient $e^{- 2 u}$ in front of the Laplacian; treating it as the linear heat equation loses exactly the geometry.

Key theorem with proof Intermediate+

Theorem (uniformization by a constant-curvature conformal metric, $χ < 0$ case). Let $M$ be a compact surface of genus $\geq 2$ with background metric $g_{0}$ . Then there is a unique smooth $u$ for which $g = e^{2 u} g_{0}$ has Gaussian curvature $K_{g} \equiv - 1$ ; equivalently the Liouville equation $$ \Delta_{g_0} u ;=; K_0 + e^{2u} $$ has a unique smooth solution. Consequently every genus- $\geq 2$ Riemann surface admits a hyperbolic (curvature $- 1$ ) metric in its conformal class.

Proof. We use the method of sub- and supersolutions for the operator $F (u) := Δ_{g_{0}} u - K_{0} - e^{2 u}$ , whose zeros are the solutions. The equation is of the form $Δ_{g_{0}} u = f (x, u)$ with $f (x, u) = K_{0} (x) + e^{2 u}$ , and $\partial_{u} f = 2 e^{2 u} > 0$ , so $f$ is increasing in $u$ . This monotonicity drives a comparison principle: if $Δ_{g_{0}} u_{1} - f (\cdot, u_{1}) \geq 0 \geq Δ_{g_{0}} u_{2} - f (\cdot, u_{2})$ and $u_{1} \leq u_{2}$ fails somewhere, then at an interior maximum of $u_{1} - u_{2}$ one has $Δ_{g_{0}} (u_{1} - u_{2}) \leq 0$ while $f (\cdot, u_{1}) - f (\cdot, u_{2}) \geq 0$ , forcing $u_{1} \leq u_{2}$ throughout. Hence solutions are unique, which also gives uniqueness in the statement.

For existence, exhibit a subsolution $u_{-}$ (with $Δ_{g_{0}} u_{-} \geq K_{0} + e^{2 u_{-}}$ ) and a supersolution $u_{+}$ (with $Δ_{g_{0}} u_{+} \leq K_{0} + e^{2 u_{+}}$ ), ordered $u_{-} \leq u_{+}$ . Constants work. For a constant $u \equiv c$ , $Δ_{g_{0}} c = 0$ , so the subsolution condition is $0 \geq K_{0} + e^{2 c}$ , i.e. $e^{2 c} \leq - K_{0}$ , achievable for $c$ very negative provided $K_{0} < 0$ somewhere; and the supersolution condition is $0 \leq K_{0} + e^{2 c}$ , i.e. $e^{2 c} \geq - min K_{0}$ , achievable for $c$ large. To remove the requirement that $K_{0}$ be negative, first replace $g_{0}$ by a conformal metric whose curvature has the correct average sign: since $\int K_{0} d A_{g_{0}} = 2 π χ < 0$ , solving the linear Poisson equation $Δ_{g_{0}} w = K_{0} - \overset{ˉ}{K}_{0}$ (solvable because the right side has zero mean) makes $\tilde{g} = e^{2 w} g_{0}$ have curvature $\tilde{K}_{0} = e^{- 2 w} \overset{ˉ}{K}_{0}$ of constant sign equal to that of $\overset{ˉ}{K}_{0} < 0$ . With this background, $\tilde{K}_{0} < 0$ everywhere and the constant sub/supersolutions above apply.

Given ordered sub/supersolutions, monotone iteration converges. Pick $λ > 0$ with $\partial_{u} f \leq λ$ on the order interval $[u_{-}, u_{+}]$ (compactness of $M$ and the bound $e^{2 u_{+}}$ supply $λ$ ). Define $u_{0} = u_{+}$ and solve the linear elliptic problems $$ (\Delta_{g_0} - \lambda),u_{n+1} = f(\cdot,u_n) - \lambda u_n . $$ Each is uniquely solvable since $Δ_{g_{0}} - λ$ is invertible (its spectrum avoids $0$ ). The map $u_{n} \mapsto u_{n + 1}$ is order-preserving by the choice of $λ$ , and one checks inductively $u_{-} \leq u_{n + 1} \leq u_{n} \leq u_{+}$ , so $u_{n} ↓ u_{\infty}$ monotonically. Elliptic $L^{p}$ and Schauder estimates give uniform $C^{2, α}$ bounds along the sequence, so the limit $u_{\infty}$ is $C^{2, α}$ and satisfies $Δ_{g_{0}} u_{\infty} = f (\cdot, u_{\infty})$ . Elliptic regularity bootstraps $u_{\infty}$ to $C^{\infty}$ . The resulting metric $e^{2 u_{\infty}} \tilde{g}_{0}$ has constant curvature $- 1$ , and pulling the auxiliary factor $e^{2 w}$ back into the conformal factor gives the claimed $u$ relative to the original $g_{0}$ . $■$

Bridge. This existence proof builds toward the analytic reading of all of surface theory: the same Liouville operator $Δ_{g_{0}} u - K_{0} + K e^{2 u}$ governs the prescribed-curvature problem in full, and the monotone-iteration scheme is exactly the discrete shadow of the parabolic flow developed in the Master tier — the bridge is that running the flow is solving the elliptic equation by relaxation in time. This is the foundational reason the two proofs of uniformization coincide: the elliptic fixed point and the parabolic limit are the same metric. The potential-theoretic route of 06.03.03 is dual to this one — there the universal cover is uniformized by a Green's function and the metric is pulled back, here the metric is found directly on $M$ by curvature — and putting these together, the constant-curvature representative appears again in 03.02.06 as the model space $S^{2}$ , $C /Λ$ , or $D$ whose isometry group acts as the deck group. The central insight is that topology fixes the sign through Gauss-Bonnet and analysis then supplies the metric.

Exercises Intermediate+

Exercise (hard, short-answer). On a surface, the curvature evolves under normalized Ricci flow by $\partial_t R = \Delta R + R(R - r)$ (with $\Delta$ the time-dependent Laplacian). Use the maximum principle to show that if $R \ge r$ somewhere is replaced by $R_{\min}(0) > 0$, then $R$ stays positive and $R_{\min}(t)$ is nondecreasing while it is below $r$.

Hint

At a spatial minimum $Δ R \geq 0$ . Compare $R_{m i n} (t)$ to the ODE $\overset{ρ}{˙} = ρ (ρ - r)$ .

Answer

Rubric. By the scalar maximum principle, $R_{m i n} (t)$ satisfies the differential inequality $\frac{d}{d t} R_{m i n} \geq R_{m i n} (R_{m i n} - r)$ , since at the minimizing point $Δ R \geq 0$ . The comparison ODE $\overset{ρ}{˙} = ρ (ρ - r)$ with $ρ (0) = R_{m i n} (0) > 0$ keeps $ρ > 0$ ; if $0 < ρ < r$ then $ρ (ρ - r) < 0$ — so the ODE pushes $ρ$ down, but bounded below by $0$ , and the precise Hamilton statement is that $R_{m i n}$ increases toward $r$ from below once the lower bound is improved by the entropy estimate. Full credit invokes the minimum principle and the sign of $ρ (ρ - r)$ , and notes $R$ cannot cross $0$ from above because $R \equiv 0$ is a barrier.

Lean formalization Intermediate+

Mathlib provides the Laplace-Beltrami operator and Gaussian curvature of a surface but has no conformal-change-of-metric API, no semilinear elliptic existence theory of the Liouville type, and no geometric flow, so the central results are not formalisable against current Mathlib (lean_status: none). The shape one would target is the conformal-rescaling law and the prescribed-curvature equation as predicates:

-- Statement target (NOT compiling against current Mathlib):
variable {M : Type*} [Surface M] (g0 : RiemannianMetric M)

-- Conformal rescaling of a metric by e^{2u}.
def conformalScale (u : M → ℝ) (g0 : RiemannianMetric M) : RiemannianMetric M :=
  fun x => Real.exp (2 * u x) • g0 x

-- The conformal change law for Gaussian curvature.
theorem gauss_curv_conformal (u : M → ℝ) (x : M) :
    gaussCurv (conformalScale u g0) x
      = Real.exp (-2 * u x) * (gaussCurv g0 x - laplaceBeltrami g0 u x) := sorry

-- Uniformization (χ < 0): a curvature −1 conformal metric exists.
theorem uniformize_hyperbolic (h : eulerChar M < 0) :
    ∃ u : M → ℝ, ∀ x, gaussCurv (conformalScale u g0) x = -1 := sorry

the Mathlib gap analysis enumerates the missing primitives: conformal class and rescaling, the Liouville/prescribed-curvature equation and its sub/supersolution solvability, and the normalized Ricci/Yamabe flow with its Harnack and entropy estimates.

Advanced results Master

The prescribed-Gaussian-curvature trichotomy (Kazdan-Warner). The Liouville operator governs not only constant targets but the full question: which functions $K$ on a compact surface arise as the Gaussian curvature of some conformal metric $e^{2 u} g_{0}$ ? The answer splits by the sign of $χ$ . For $χ < 0$ , the equation $Δ_{g_{0}} u = K_{0} - K e^{2 u}$ is solvable whenever $K < 0$ somewhere and the sign-compatible average condition holds; the negative case is carried in full by the sub/supersolution argument above, generalised to nonconstant $K$ by replacing the constant barriers with $K$ -adapted ones ^{[Kazdan-Warner 1974]}. For $χ = 0$ , solvability requires $K$ to change sign (unless $K \equiv 0$ ) together with the linear constraint $\int K e^{2 u} d A_{g_{0}} = 0$ read through the equation. For $χ > 0$ — the sphere — the Nirenberg problem appears: $K = K_{S^{2}} + const$ is obstructed by the Kazdan-Warner identity $\int_{S^{2}} ⟨ \nabla K, \nabla x_{j} ⟩ e^{2 u} d A = 0$ for the coordinate functions $x_{j}$ , which rules out monotone $K$ and shows not every positive $K$ is admissible. The Moser-Trudinger inequality supplies the critical Sobolev embedding controlling the variational functional in the $S^{2}$ case.

The normalized Ricci flow reproves uniformization (Hamilton-Chow). Hamilton's 1988 analysis of $\partial_{t} g = (r - R) g$ on a surface shows the flow exists for all time and converges, exponentially, to a constant-curvature metric in the conformal class — an entirely parabolic proof of uniformization ^{[Hamilton 1988]}. The mechanism differs by sign of $χ$ . When $χ < 0$ ( $r < 0$ ), the maximum principle on $\partial_{t} R = Δ R + R (R - r)$ squeezes $R$ toward $r$ from both sides, and a gradient estimate yields exponential convergence to the hyperbolic metric. When $χ = 0$ ( $r = 0$ ), the flow is $\partial_{t} R = Δ R + R^{2}$ and an entropy bound forces $R \to 0$ , giving the flat metric. The hard case is $χ > 0$ ( $r > 0$ , the sphere): Hamilton settled it under the assumption $R > 0$ using a Lyapunov entropy $N = \int_{M} R lo g R d A$ whose monotonicity, combined with a differential Harnack inequality, controls the geometry; Chow 1991 removed the curvature-positivity assumption, proving that on $S^{2}$ any initial metric flows to the round metric, completing the flow proof for all genera ^{[Chow 1991]}. The fixed point $R \equiv r$ is the constant-curvature representative, so the long-time limit of the flow is the uniformizing metric.

Recovering the three model geometries. The constant-curvature representative realises $M$ as a quotient of one of the three simply connected model surfaces of 03.02.06. Genus $0$ : curvature $+ 1$ , universal cover the round sphere $S^{2}$ , no quotient (the sphere is its own cover). Genus $1$ : curvature $0$ , universal cover the Euclidean plane, $M = C /Λ$ a flat torus. Genus $\geq 2$ : curvature $- 1$ , universal cover the hyperbolic disk $D$ (Poincaré metric), $M = D /Γ$ for a discrete torsion-free cocompact group $Γ$ of Möbius isometries. The flow or the Liouville solution constructs the constant-curvature metric on $M$ ; passing to the universal cover and using the classification of complete simply connected constant-curvature surfaces (Killing-Hopf, 03.02.06) identifies the cover with the model, and the deck group with a discrete isometry group. This is the differential-geometric proof of uniformization, complementary to the function-theoretic proof of 06.03.03.

Solitons and the entropy formalism. The two-dimensional flow has a rich fixed-point structure beyond constant curvature: the cigar soliton $g = (d x^{2} + d y^{2}) / (1 + x^{2} + y^{2})$ is a complete steady gradient Ricci soliton on $R^{2}$ with $R = 2/ (1 + ∣ z ∣^{2}) > 0$ , the model singularity for the unnormalized flow and the prototype for Perelman's later entropy functionals. On compact surfaces no nontrivial solitons exist (the only ones are constant-curvature), which is why convergence is unobstructed; the soliton enters as the blow-up limit ruled out by Hamilton's Harnack estimate.

Synthesis. Putting these together, uniformization is the central insight that one topological invariant, $χ (M)$ , controls three layers at once: it fixes the sign of the curvature constant through Gauss-Bonnet, the solvability of the Liouville equation through the average constraint, and the limiting curvature $r$ of the Ricci flow. The elliptic and parabolic routes are dual — the Liouville fixed point of 06.03.03's complement is exactly the long-time limit of the flow, so solving the equation and running the flow are the same statement read statically versus dynamically. This generalises the linear potential theory of 06.03.03 to a nonlinear curvature equation, and it is the foundational reason the model geometries of 03.02.06 are forced: the constant-curvature metric whose existence the flow guarantees pulls back to the universal cover as one of $S^{2}$ , $C$ , $D$ , with the deck group acting by isometries. The bridge is that curvature, not function theory, supplies the metric directly on $M$ , and the higher-dimensional shadow of the same circle of ideas — the Yamabe problem and Perelman's entropy — appears again in the analysis of the Ricci flow on $3$ -manifolds, where the surface case is the solved prototype.

Full proof set Master

Proposition 1 (conformal change law). On a surface, if $g = e^{2 u} g_{0}$ then $K_{g} = e^{- 2 u} (K_{0} - Δ_{g_{0}} u)$ , where $Δ_{g_{0}}$ is the geometer's Laplace-Beltrami operator of $g_{0}$ .

Proof. Work in $g_{0}$ -isothermal coordinates, so $g_{0} = e^{2 φ} (d x^{2} + d y^{2})$ for a local potential $φ$ , and $g = e^{2 (u + φ)} (d x^{2} + d y^{2})$ . For any conformally flat metric $e^{2 ψ} (d x^{2} + d y^{2})$ the Gaussian curvature is $- e^{- 2 ψ} (ψ_{xx} + ψ_{y y})$ , a direct Brioschi/Liouville computation. Apply this with $ψ = u + φ$ : $$ K_g = -e^{-2(u+\varphi)}\big((u+\varphi){xx}+(u+\varphi){yy}\big) = e^{-2u}\Big(-e^{-2\varphi}(\varphi_{xx}+\varphi_{yy}) - e^{-2\varphi}(u_{xx}+u_{yy})\Big). $$ The first bracket is $K_{0}$ and the second is $Δ_{g_{0}} u$ because $Δ_{g_{0}} = e^{- 2 φ} (\partial_{x}^{2} + \partial_{y}^{2})$ in these coordinates. Hence $K_{g} = e^{- 2 u} (K_{0} - Δ_{g_{0}} u)$ , an identity independent of the coordinate chart. $■$

Proposition 2 (Gauss-Bonnet forces the sign of the uniformizing constant). If $g = e^{2 u} g_{0}$ has constant curvature $c$ , then $sgn (c) = sgn (χ (M))$ .

Proof. By Gauss-Bonnet, $\int_{M} K_{g} d A_{g} = 2 π χ (M)$ . With $K_{g} \equiv c$ the left side is $c \cdot area_{g} (M)$ , and $area_{g} (M) > 0$ for any metric, so $c \cdot area_{g} = 2 π χ$ has $c$ and $χ$ of the same sign (and $c = 0$ iff $χ = 0$ ). $■$

Proposition 3 (comparison principle and uniqueness for the Liouville equation). For $K \equiv - 1$ the solution of $Δ_{g_{0}} u = K_{0} + e^{2 u}$ on a compact surface is unique.

Proof. Suppose $u_{1}, u_{2}$ both solve it and set $v = u_{1} - u_{2}$ . Subtracting, $$ \Delta_{g_0}v = e^{2u_1} - e^{2u_2} = \Big(\int_0^1 2e^{2(u_2 + s v)},ds\Big) v = a(x), v, $$ with $a (x) > 0$ . At an interior maximum point $x_{0}$ of $v$ (compactness gives one), $Δ_{g_{0}} v (x_{0}) \leq 0$ , while $a (x_{0}) v (x_{0}) = Δ_{g_{0}} v (x_{0}) \leq 0$ forces $v (x_{0}) \leq 0$ , so $max v \leq 0$ . Symmetrically $min v \geq 0$ , hence $v \equiv 0$ . $■$

Proposition 4 (constant in $t$ of the area, hence of $r$ , under normalized flow). Along $\partial_{t} g = (r - R) g$ on a compact surface, $area (M, g (t))$ is constant, and therefore $r = 4 π χ / area$ is constant.

Proof. For a $2$ -dimensional metric, $\partial_{t} d A_{g} = \frac{1}{2} tr_{g} (\partial_{t} g) d A_{g} = \frac{1}{2} tr_{g} ((r - R) g) d A_{g} = (r - R) d A_{g}$ . Integrating, $\frac{d}{d t} area = \int_{M} (r - R) d A_{g} = r area - \int_{M} R d A_{g} = r area - 4 π χ$ . By the definition $r = 4 π χ / area$ , the right side vanishes, so the area is constant; $χ$ is a topological invariant, so $r$ is constant in $t$ . $■$

Proposition 5 (the curvature evolution equation). Under the normalized flow on a surface, $\partial_{t} R = Δ_{g (t)} R + R (R - r)$ .

Proof. Writing $g = e^{2 u} g_{0}$ and using Proposition 1, $R = 2 K_{g} = 2 e^{- 2 u} (K_{0} - Δ_{g_{0}} u)$ , while the flow gives $\overset{u}{˙} = \frac{1}{2} (r - R)$ (from $\partial_{t} g = 2 \overset{u}{˙} g = (r - R) g$ ). Differentiating $R$ in $t$ and using $\partial_{t} Δ_{g (t)} = - 2 \overset{u}{˙} Δ_{g (t)}$ for the conformal Laplacian, together with $Δ_{g (t)} = e^{- 2 u} Δ_{g_{0}}$ , a direct computation collapses to $\partial_{t} R = Δ_{g (t)} R + R (R - r)$ ; the reaction term $R (R - r)$ comes from differentiating the curvature factor $e^{- 2 u}$ against $\overset{u}{˙} = \frac{1}{2} (r - R)$ and the diffusion term from the Laplacian acting on the evolving conformal factor. $■$

The Hamilton-Chow long-time existence and exponential convergence, the entropy monotonicity of $N = \int R lo g R d A$ , and the differential Harnack inequality on $S^{2}$ are stated in the Advanced results without proof here — see Hamilton ^{[Hamilton 1988]} for the $χ \leq 0$ and the $R > 0$ sphere cases, and Chow ^{[Chow 1991]} for the removal of the positivity hypothesis on $S^{2}$ .

Connections Master

Uniformization theorem, potential-theoretic 06.03.03. That unit proves uniformization by solving the Dirichlet problem / constructing a Green's function on the universal cover and pulling the metric back. This unit gives the dual differential-geometric proof: it finds the constant-curvature metric directly on $M$ by solving the Liouville equation or running the Ricci flow, and the two limits coincide. The potential theory there is the linear core that this nonlinear curvature equation generalises.
Constant-curvature spaces and Killing-Hopf 03.02.06. The output of this unit is a metric of constant curvature $+ 1$ , $0$ , or $- 1$ on $M$ ; that unit classifies the complete simply connected models $S^{2}$ , $E^{2}$ , $H^{2}$ and identifies $M$ as a quotient by a discrete isometry group. The uniformizing metric constructed here is exactly the metric whose universal cover Killing-Hopf classifies, so the two units compose into the full geometrization of surfaces.
Sectional, Ricci, and scalar curvature 03.02.05. The Gaussian curvature $K$ , the scalar curvature $R = 2 K$ , and the Ricci tensor $Ric = K g$ used throughout are defined there; the normalized Ricci flow is the surface case of the general Ricci flow $\partial_{t} g = - 2 Ric$ , and the conformal change law is the two-dimensional specialisation of the general conformal transformation of scalar curvature.
Riemann mapping theorem 06.01.06. The flat and hyperbolic uniformizing metrics on a simply connected planar domain reproduce, at the level of the universal cover, the conformal map to the disk or plane; uniformization is the closed-surface and higher-genus extension of the Riemann mapping theorem, and the constant-curvature metric is the geometric content of the conformal coordinate it provides.

Historical & philosophical context Master

Poincaré and Koebe stated and proved the uniformization theorem in 1907 by function-theoretic means, but the idea that a conformal metric of constant curvature should exist in each conformal class goes back to Poincaré's 1898 study of Fuchsian functions, where the hyperbolic metric on a genus- $\geq 2$ surface appears as the pulled-back Poincaré metric of the disk. The reformulation as a single nonlinear PDE — the Liouville equation $Δ u = K_{0} - K e^{2 u}$ — and the systematic prescribed-curvature theory were carried out by Berger (1971) and decisively by Kazdan and Warner, whose 1974 paper Curvature functions for compact 2-manifolds (Annals of Mathematics 99, 14–47) settled the solvability trichotomy and exhibited the obstruction now named for them on $S^{2}$ ^{[Kazdan-Warner 1974]}. Moser's 1973 sharp form of the Trudinger inequality supplied the critical Sobolev tool for the sphere case.

The parabolic proof is due to Richard Hamilton, who introduced the Ricci flow in 1982 and analysed its two-dimensional normalized form in The Ricci flow on surfaces (Contemporary Mathematics 71, 1988, pp. 237–262), proving convergence to a constant-curvature metric for $χ \leq 0$ and for $χ > 0$ under the assumption of positive curvature, using an entropy functional and a differential Harnack estimate ^{[Hamilton 1988]}. Bennett Chow removed the positivity hypothesis in The Ricci flow on the 2-sphere (Journal of Differential Geometry 33, 1991, pp. 325–334), completing the flow proof of uniformization on all closed surfaces ^{[Chow 1991]}. The surface case became the tested prototype for Hamilton's program on three-manifolds and for Perelman's entropy and no-local-collapsing methods two decades later.

Bibliography Master

@incollection{Hamilton1988RicciSurfaces,
  author    = {Hamilton, Richard S.},
  title     = {The {R}icci flow on surfaces},
  booktitle = {Mathematics and General Relativity},
  series    = {Contemporary Mathematics},
  volume    = {71},
  pages     = {237--262},
  publisher = {American Mathematical Society},
  year      = {1988}
}

@article{Chow1991Sphere,
  author  = {Chow, Bennett},
  title   = {The {R}icci flow on the 2-sphere},
  journal = {Journal of Differential Geometry},
  volume  = {33},
  number  = {2},
  pages   = {325--334},
  year    = {1991}
}

@article{KazdanWarner1974,
  author  = {Kazdan, Jerry L. and Warner, F. W.},
  title   = {Curvature functions for compact 2-manifolds},
  journal = {Annals of Mathematics},
  volume  = {99},
  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},
  pages   = {325--332},
  year    = {1971}
}

@article{Moser1973,
  author  = {Moser, J{\"u}rgen},
  title   = {On a nonlinear problem in differential geometry},
  journal = {Dynamical Systems (Academic Press)},
  pages   = {273--280},
  year    = {1973}
}

@book{ChowLuNi2006,
  author    = {Chow, Bennett and Lu, Peng and Ni, Lei},
  title     = {Hamilton's Ricci Flow},
  series    = {Graduate Studies in Mathematics},
  volume    = {77},
  publisher = {American Mathematical Society},
  year      = {2006}
}

@book{Jost2006,
  author    = {Jost, J{\"u}rgen},
  title     = {Compact Riemann Surfaces},
  edition   = {3rd},
  publisher = {Springer},
  year      = {2006}
}

@incollection{Poincare1898,
  author    = {Poincar{\'e}, Henri},
  title     = {Les fonctions fuchsiennes et l'{\'e}quation $\Delta u = e^u$},
  booktitle = {Journal de Math{\'e}matiques Pures et Appliqu{\'e}es},
  volume    = {4},
  pages     = {137--230},
  year      = {1898}
}

Prerequisites

06.03.03
03.02.05
03.02.06

Tier anchors

beginner: Needham *Visual Differential Geometry and Forms* (Gaussian curvature, conformal stretching); Tristan Needham on metrics that bend a sphere flat
intermediate: Jost *Compact Riemann Surfaces* §2.3, §4.4 (conformal metrics, the Liouville equation, uniformization by curvature); do Carmo *Differential Geometry of Curves and Surfaces* §4 (Gauss-Bonnet)
master: Hamilton 1988 *The Ricci flow on surfaces* (Contemp. Math. 71); Chow 1991 *The Ricci flow on the 2-sphere* (J. Diff. Geom. 33); Kazdan-Warner 1974 (Ann. of Math. 99); Chow-Lu-Ni *Hamilton's Ricci Flow* Ch. 5

References

Hamilton — The Ricci flow on surfaces (1988) · Contemporary Mathematics 71, pp. 237--262
Chow — The Ricci flow on the 2-sphere (1991) · Journal of Differential Geometry 33, pp. 325--334
Kazdan-Warner — Curvature functions for compact 2-manifolds (1974) · Annals of Mathematics 99, pp. 14--47
Jost — Compact Riemann Surfaces (3rd ed., 2006) · §2.3 (conformal metrics), §4.4 (uniformization)
03-modern-geometry/Bott-LecturesOnMorseTheory.pdf · background on the second-variation / energy-functional viewpoint used in passing

Estimated time

beginner: 18m
intermediate: 45m
master: 82m