Finding Einstein Metrics That “Optimize” Ricci Curvature
In Riemannian geometry, one of the most profound questions is: what is the “best” metric on a manifold? The Ricci flow, introduced by Richard Hamilton in 1982, answers this by evolving a metric toward a canonical shape. Its fixed points — metrics that don’t change under the flow — are called Einstein metrics, and they represent a kind of geometric optimum.
The Big Picture: What Is the Ricci Flow?
The Ricci flow is a geometric PDE that deforms a Riemannian metric $g_{ij}$ in the direction of its Ricci curvature:
$$\frac{\partial g_{ij}}{\partial t} = -2 R_{ij}$$
where $R_{ij}$ is the Ricci curvature tensor. Think of it as a heat equation for geometry — it smooths out irregularities in curvature over time.
Fixed Points: Einstein Metrics
A metric is a fixed point of the Ricci flow (up to rescaling) when:
$$R_{ij} = \lambda g_{ij}$$
for some constant $\lambda \in \mathbb{R}$. This is the Einstein condition. Such a metric is called an Einstein metric, and the constant $\lambda$ is called the Einstein constant.
Taking the trace:
$$R = \lambda n$$
where $R$ is the scalar curvature and $n = \dim(M)$.
The three cases are:
- $\lambda > 0$: positive Einstein metric (e.g., round sphere $S^n$)
- $\lambda = 0$: Ricci-flat metric (e.g., flat torus $T^n$, Calabi-Yau)
- $\lambda < 0$: negative Einstein metric (e.g., hyperbolic space $\mathbb{H}^n$)
Concrete Example: The Round 2-Sphere $S^2$
The simplest and most instructive example is $S^2$ with the round metric in spherical coordinates $(\theta, \phi)$:
$$g = R^2 \begin{pmatrix} 1 & 0 \ 0 \sin^2\theta \end{pmatrix}$$
The Ricci tensor of the round sphere of radius $R$ is:
$$R_{ij} = \frac{1}{R^2} g_{ij}$$
This is exactly the Einstein condition with $\lambda = \frac{1}{R^2} > 0$.
Under the Ricci flow:
$$\frac{\partial g}{\partial t} = -2R_{ij} = -\frac{2}{R^2} g_{ij}$$
The sphere shrinks homothetically — it stays round but its radius changes. Setting $g(t) = r(t)^2 g_0$ where $g_0$ is the unit sphere metric:
$$2r \dot{r} g_0 = -\frac{2}{r^2} r^2 g_0 = -2 g_0$$
$$\Rightarrow r\dot{r} = -1 \Rightarrow r(t) = \sqrt{R^2 - 2t}$$
The sphere collapses to a point at $t^* = R^2/2$.
Python Implementation
We simulate and visualize the Ricci flow on $S^2$ numerically. We track the metric coefficient evolution, Ricci curvature, the Einstein condition residual, and the geometry in 3D.
1 | import numpy as np |
Code Walkthrough
Section 1 — Analytical Ricci Flow on $S^2$
The round sphere is an exact Einstein metric. We use the closed-form solution:
$$r(t) = \sqrt{R_0^2 - 2t}, \quad t \in \left[0,, \frac{R_0^2}{2}\right)$$
From this we derive:
| Quantity | Formula |
|---|---|
| Gaussian curvature | $\kappa = 1/r^2$ |
| Scalar curvature | $R = 2/r^2$ |
| Einstein constant | $\lambda = 1/r^2$ |
| Area | $A = 4\pi r^2$ |
We also introduce a sinusoidal perturbation $r_\epsilon = r(t)(1 + \epsilon \sin(\cdots))$ to simulate a near-Einstein metric and measure how far it deviates from the Einstein condition. The Einstein residual:
$$\text{Res} = \frac{|\lambda_\epsilon - \lambda|}{\lambda}$$
should be small for near-Einstein metrics and goes to zero as the metric relaxes.
Section 2 — Numerical Ricci Flow on Rotationally Symmetric Ellipsoids
For a rotationally symmetric metric on $S^2$ parametrized by equatorial radius $a$ and polar radius $c$, the Ricci flow reduces to the ODE system:
$$\frac{da}{dt} = -\frac{1}{c^2}, \qquad \frac{dc}{dt} = -\frac{1}{a^2}$$
We integrate this with a simple explicit Euler method over 8000 steps (sufficient for stability on this timescale). We track three initial conditions:
- Prolate ($a = 1, c = 3$): elongated along the $z$-axis
- Round ($a = c = 1$): already at the fixed point
- Oblate ($a = 3, c = 1$): flattened at the poles
The key observable is the isotropy ratio $a(t)/c(t) \to 1$, which signals convergence to the Einstein metric.
Section 3 — 3D Ellipsoid Visualization
Three snapshots of the prolate ellipsoid at $t = 0$, $t = 0.4,t_\text{end}$, and $t = 0.85,t_\text{end}$ are rendered as 3D surfaces, color-mapped with plasma. You can watch the shape “round out” — the Ricci flow literally sculpts the manifold toward its optimal geometry.
Graph-by-Graph Analysis
Panels 1–3 (top row): Analytical $S^2$ under Ricci flow
- $r(t)$ decreases monotonically from $R_0 = 2$ to $0$ at $t^* = 2$.
- $R(t) = 2/r^2$ blows up as $t \to t^*$ — this is the Type I singularity studied by Hamilton and Perelman.
- The area $A(t) = 4\pi r^2$ contracts linearly in $r^2$, reaching zero at collapse.
Panels 4–5 (middle-left): Einstein constant and residual
- $\lambda(t) = 1/r^2$ grows monotonically, confirming the metric remains Einstein (just rescaled) throughout the flow.
- The residual plot shows the perturbed metric deviating from the Einstein condition — oscillating at early times, then growing as the sphere shrinks. A true Einstein metric would give a flat zero line.
Panels 6–7 (middle-center/right): Ellipsoid axis evolution
- All three initial configurations converge: $a(t)$ and $c(t)$ approach each other before collapse.
- The isotropy ratio $a/c \to 1$ is the smoking gun — the Ricci flow is driving the geometry toward the Einstein fixed point.
Panel 8: Phase portrait
- The diagonal $a = c$ is the Einstein locus (fixed points up to rescaling).
- All trajectories flow toward this diagonal and then collapse along it — precisely the behavior predicted by Hamilton’s theorem.
Panel 9: Scalar curvature
- $\bar{R}(t) = 1/c^2 + 1/a^2$ increases and converges for all initial data, consistent with the curvature blow-up at the Einstein fixed point (which is also a finite-time singularity for the normalized flow).
Panels 10–12 (bottom row): 3D geometry
- The prolate ellipsoid ($a=1, c=3$) visibly rounds out across the three snapshots.
- By $t = 0.85,t_\text{end}$, $a \approx c$: the shape is nearly a round sphere — the Einstein metric.
Figure saved.
Summary
| Concept | Mathematical Statement |
|---|---|
| Ricci flow | $\partial_t g_{ij} = -2R_{ij}$ |
| Einstein condition (fixed point) | $R_{ij} = \lambda g_{ij}$ |
| Round sphere (positive Einstein) | $\lambda = 1/r^2 > 0$ |
| Collapse time | $t^* = r_0^2/2$ |
| Convergence criterion | $a(t)/c(t) \to 1$ |
The Ricci flow is a gradient flow for the normalized Einstein–Hilbert functional, and its fixed points are exactly the Einstein metrics. On $S^2$, the unique (up to scaling) Einstein metric is the round sphere — and the Ricci flow finds it, even from a lumpy or anisotropic start. This is the geometric heart of Hamilton’s theorem on surfaces, and a precursor to Perelman’s proof of the Geometrization Conjecture.