Mabuchi Energy and Kähler-Ricci Flow
K-energy as the variational principle on the celestial sphere, and its connection to spin memory via gradient flow
1. Kähler Geometry Setup
Let $(M^{2n}, \omega_0, J)$ be a compact Kähler manifold with Kähler form $\omega_0$and integrable complex structure $J$. In local holomorphic coordinates $(z^1, \ldots, z^n)$, the Kähler form is:
$$\omega_0 = \frac{\sqrt{-1}}{2}\, g_{i\bar{j}}\, dz^i \wedge d\bar{z}^j, \qquad g_{i\bar{j}} = \frac{\partial^2 K}{\partial z^i \,\partial \bar{z}^j}$$
where $K$ is the Kähler potential. The space of Kähler metrics in the cohomology class $[\omega_0]$ is parametrized by Kähler potentials:
$$\mathcal{H} = \bigl\{\varphi \in C^\infty(M) : \omega_\varphi = \omega_0 + \tfrac{\sqrt{-1}}{2}\,\partial\bar{\partial}\varphi > 0 \bigr\}$$
The space $\mathcal{H}$ is an infinite-dimensional Riemannian manifold with the Mabuchi metric:
$$\langle \psi_1, \psi_2 \rangle_\varphi = \int_M \psi_1\, \psi_2\, \frac{\omega_\varphi^n}{n!}$$
2. Kähler–Einstein Equation
A Kähler metric is Kähler–Einstein if $\mathrm{Ric}(\omega) = \lambda\, \omega$ for some constant $\lambda$. In terms of the Kähler potential, this is the complex Monge–Ampère equation:
$$\bigl(\omega_0 + \tfrac{\sqrt{-1}}{2}\,\partial\bar{\partial}\varphi\bigr)^n = e^{h_0 - \lambda\varphi}\, \omega_0^n$$
where $h_0$ is the Ricci potential defined by $\mathrm{Ric}(\omega_0) - \lambda\omega_0 = \frac{\sqrt{-1}}{2}\,\partial\bar{\partial} h_0$. This is a fully nonlinear elliptic PDE. For $\lambda \leq 0$, Aubin and Yau proved existence and uniqueness. For $\lambda > 0$ (Fano manifolds), existence requires vanishing of the Futaki invariant and K-stability.
3. The Mabuchi K-Energy
The Mabuchi K-energy is the unique functional on $\mathcal{H}$ whose critical points are Kähler–Einstein metrics. It is defined via its first variation along a path$\{\varphi_t\}_{t \in [0,1]}$ from $0$ to $\varphi$:
$$\mathcal{K}(\omega_\varphi) = -\int_0^1 \int_M \dot{\varphi}_t\, \bigl(R(\omega_t) - \bar{R}\bigr)\, \frac{\omega_t^n}{n!}\, dt$$
where $R(\omega_t)$ is the scalar curvature of $\omega_t$ and $\bar{R} = n\lambda$ is the average scalar curvature (a topological constant by Gauss–Bonnet). The key property is path independence:
Proof of path independence:
The variation $\delta \mathcal{K} / \delta \varphi = -(R - \bar{R})$ is an exact 1-form on $\mathcal{H}$. The integrability condition is:
$$\frac{\delta^2 \mathcal{K}}{\delta \varphi \,\delta \psi} = \int_M \psi\, \Delta_{\bar{\partial}} \bigl(\Delta_{\bar{\partial}} + \lambda\bigr)^{-1}\!\left(\frac{\delta R}{\delta \varphi}\right) \frac{\omega^n}{n!} = \frac{\delta^2 \mathcal{K}}{\delta \psi\, \delta \varphi}$$
which holds because the linearization of scalar curvature is self-adjoint with respect to the $L^2$ inner product.
4. Kähler-Ricci Flow
The Kähler–Ricci flow (KRF) preserves the Kähler condition and evolves$\omega$ toward the Kähler–Einstein metric:
$$\frac{\partial \omega}{\partial t} = -\mathrm{Ric}(\omega) + \lambda\, \omega$$
In terms of the Kähler potential, this reduces to a parabolic complex Monge–Ampère equation:
$$\frac{\partial \varphi}{\partial t} = \log \frac{(\omega_0 + \frac{\sqrt{-1}}{2}\,\partial\bar{\partial}\varphi)^n}{\omega_0^n} + \lambda\varphi + h_0$$
The Tian–Zhu monotonicity formula shows that $\mathcal{K}$ decreases along KRF:
$$\frac{d\mathcal{K}}{dt} = -\int_M \bigl|R_{i\bar{j}} - \lambda\, g_{i\bar{j}}\bigr|^2\, \frac{\omega^n}{n!} \;\leq\; 0$$
with equality if and only if $\omega$ is Kähler–Einstein. This is the exact analogue of Perelman's monotonicity $d\mathcal{F}/dt = 2\int |R_{ij} + \nabla_i\nabla_j f|^2\, e^{-f}\, d\mu \geq 0$for Ricci flow.
5. The Celestial Sphere as Kähler Manifold
The celestial sphere at null infinity is $S^2 \cong \mathbb{CP}^1$, the simplest compact Kähler manifold. In stereographic coordinates $z = e^{i\phi}\tan(\theta/2)$, the Fubini–Study metric is:
$$\omega_{\rm FS} = \frac{\sqrt{-1}}{2}\,\frac{dz \wedge d\bar{z}}{(1 + |z|^2)^2}, \qquad R(\omega_{\rm FS}) = 2 = \bar{R}$$
The round metric on $S^2$ is the unique Kähler–Einstein metric (up to isometry) with $\lambda = 1$. This is the Aubin–Yau theorem applied to $\mathbb{CP}^1$:
$$\mathcal{K}(\omega_{\rm FS}) = 0 \qquad \text{(absolute minimum)}$$
Any Kähler metric on $\mathbb{CP}^1$ has $\mathcal{K} \geq 0$, with equality only at the round metric. KRF on $S^2$ converges exponentially to $\omega_{\rm FS}$: this is the uniformization theorem realized as gradient flow.
6. KRF Convergence and Uniformization
On $\mathbb{CP}^1$, the Kähler–Ricci flow with $\lambda = 1$ converges exponentially to the Fubini–Study metric. The convergence rate is controlled by the first positive eigenvalue of $\Delta_{\bar{\partial}}$:
$$\|\omega(t) - \omega_{\rm FS}\|_{C^k} \leq C_k\, e^{-2t}, \qquad \lambda_1(\Delta_{\bar{\partial}}) = 2 \;\text{ on } \mathbb{CP}^1$$
This is the Kähler analogue of Hamilton's convergence theorem for Ricci flow on $S^2$. The exponential rate $e^{-2t}$ is sharp: the eigenvalue $\lambda_1 = 2$ corresponds to the$\ell = 1$ spherical harmonics (the conformal Killing vectors of $S^2$).
For a general initial Kähler metric $\omega_0$ on $\mathbb{CP}^1$, the potential satisfies:
$$\frac{\partial \varphi}{\partial t} = \log\frac{(1 + \partial_z\partial_{\bar{z}}\varphi)}{(1 + |z|^2)^{-2}} + \varphi + h_0$$
The maximum principle applied to $R - \bar{R}$ shows that $|R - 2| \leq C e^{-2t}$, confirming that scalar curvature uniformizes. This is the two-dimensional uniformization theorem proved via parabolic methods rather than elliptic (Poincaré–Koebe).
$$\mathcal{K}(\omega(t)) = \mathcal{K}(\omega_0)\, e^{-2t} + \mathcal{O}(e^{-4t}) \quad \text{as } t \to \infty$$
7. Futaki Invariant and Obstructions
The Futaki invariant is the obstruction to existence of Kähler–Einstein metrics on Fano manifolds. For a holomorphic vector field $X$ on $(M, [\omega])$, it is defined as:
$$\mathrm{Fut}(X) = \int_M X(h_\omega)\, \frac{\omega^n}{n!}$$
where $h_\omega$ is the Ricci potential. The Futaki invariant is independent of the choice of representative $\omega \in [\omega_0]$ and vanishes if a KE metric exists. On $\mathbb{CP}^1$, the only holomorphic vector fields are the $\mathfrak{sl}(2, \mathbb{C})$ generators, and:
$$\mathrm{Fut}(X) = 0 \quad \text{for all } X \in \mathfrak{sl}(2, \mathbb{C})$$
confirming the existence of the round metric as the unique KE metric. The Futaki invariant is the first variation of $\mathcal{K}$ at a critical point along the holomorphic vector field direction:
$$\mathrm{Fut}(X) = -\frac{d}{dt}\bigg|_{t=0} \mathcal{K}\bigl(\sigma_t^*\omega\bigr), \qquad \sigma_t = \exp(tX)$$
In the gravitational context, a nonzero Futaki invariant on the deformed celestial sphere would signal a topological obstruction to the return of the round metric after a GW burst — this cannot happen for $\mathbb{CP}^1$, but is relevant for compactifications with more complex asymptotic geometry.
8. Spin Memory as K-Energy Displacement
A gravitational wave burst passing through $\mathscr{I}^+$ deforms the conformal metric on the celestial sphere away from the round metric. The spin memory $\Delta\Psi$ is the net angular displacement after the burst. In Kähler language:
$$\Delta\Psi = \mathcal{K}(\omega_{\rm after}) - \mathcal{K}(\omega_{\rm before})$$
Before the burst, the celestial metric is round ($\mathcal{K} = 0$). During the burst, the news tensor $N_{AB} \neq 0$ drives the metric away from round, increasing $\mathcal{K}$. After the burst, KRF drives $\mathcal{K}$ back toward zero, but the residual displacement encodes the spin memory:
$$\Delta\Psi = \int_{u_1}^{u_2} \frac{d\mathcal{K}}{du}\, du = -\int_{u_1}^{u_2} \int_{S^2} \bigl|R_{z\bar{z}} - g_{z\bar{z}}\bigr|^2\, d^2z\, du$$
The KRF on $S^2$ then uniformizes the deformation, restoring the round metric on a timescale set by the first nonzero eigenvalue of $\Delta_{\bar{\partial}}$ on $\mathbb{CP}^1$.
9. Three Lagrangians, One Structure
The Perelman $\mathcal{F}$-functional, the Bondi–Sachs action $S_{\rm BS}$, and the Mabuchi K-energy $\mathcal{K}$ share a common gradient flow structure:
| Functional | Domain | Gradient Flow | Critical Points |
|---|---|---|---|
| $\mathcal{F}(g, f)$ | Spatial slice $\Sigma$ | Ricci flow | Ricci solitons |
| $S_{\rm BS}$ | $\mathscr{I}^+$ | Bondi mass loss | Zero news $N_{AB} = 0$ |
| $\mathcal{K}(\omega)$ | $S^2 \cong \mathbb{CP}^1$ | Kähler–Ricci flow | Round metric |
$$\frac{d\mathcal{F}}{dt} \geq 0, \qquad \frac{dm_B}{du} \leq 0, \qquad \frac{d\mathcal{K}}{dt} \leq 0$$
All three are monotone along their respective flows, with the same underlying mechanism: the gradient of a convex functional on an infinite-dimensional space of metrics.