---
title: Bray's Conformal-Flow Method
url: https://www.emergentmind.com/topics/bray-s-conformal-flow-method
type: topic
---

# Bray's Conformal-Flow Method

Searching arXiv for the cited paper and closely related conformal-flow/Jang references.
Bray’s conformal-flow method is a geometric deformation scheme for proving Penrose-type lower bounds for the ADM mass in terms of horizon area. In the formulation reviewed and extended by Han and Khuri in "The conformal flow of metrics and the general Penrose inequality" [1409.0067], the method begins in the Riemannian, time-symmetric setting and is then adapted to general Einstein initial data by coupling it to a generalized Jang equation. The central idea is to deform a metric conformally so that the area of the outermost minimal surface remains fixed, the ADM mass is monotone nonincreasing, and the long-time limit approaches Schwarzschild. In the non-time-symmetric case, the same flow is used more subtly: it supplies the warping factor for the generalized Jang construction, converting the Penrose conjecture into a coupled PDE problem with “desirable properties” [1409.0067].

## 1. Penrose inequality setting and the Riemannian origin

The ambient problem is the Penrose inequality for an asymptotically flat initial data set $(M,g,k)$ for the Einstein equations, with energy density $\mu$ and momentum density $J$ satisfying the dominant energy condition $\mu \ge |J|$. For a surface $S \subset M$, the null expansions are
\[
\theta_\pm = H_S \pm \operatorname{Tr}_S k,
\]
where $H_S$ is the mean curvature with respect to the unit normal toward spatial infinity. Apparent horizons are characterized by $\theta_+=0$ or $\theta_-=0$ [1409.0067].

In the time-symmetric case $k=0$, the problem reduces to a purely Riemannian statement. Horizons are then minimal surfaces, the scalar curvature satisfies $R_g \ge 0$, and one can seek a deformation toward Schwarzschild that preserves the relevant boundary area while controlling mass. Han and Khuri emphasize that Bray’s conformal flow is especially suited to the multiple-horizon case, where Huisken–Ilmanen’s inverse mean curvature flow is not directly sufficient. In Bray’s method, the metric is deformed conformally in a way that keeps the area of the outermost minimal surface constant, makes the ADM mass monotone nonincreasing, and forces convergence to Schwarzschild asymptotically. This directly yields the Riemannian Penrose inequality.

The generalization studied in [1409.0067] drops the assumption $k=0$. That relaxation changes the geometric structure substantially: horizons are no longer minimal surfaces in $(M,g)$ but apparent horizons in $(M,g,k)$; the scalar curvature of the spatial metric alone need not be nonnegative; and the momentum density and extrinsic curvature enter essentially. The resulting argument is therefore conditional rather than complete: the Penrose conjecture without time symmetry is reduced to solving a coupled PDE system.

## 2. Definition of Bray’s conformal flow

The metric-side starting point is a 3-dimensional asymptotically flat Riemannian manifold $(M,g)$ with one end and with boundary an outerminimizing minimal surface with finitely many components. A point stressed in [1409.0067] is that the review of Bray’s flow is formulated without assuming $R_g \ge 0$, because in the later Jang setting one only obtains weak nonnegativity.

The evolving metric is
\[
g_t = u_t^4 g,
\]
where the conformal factor $u_t$ is generated by a velocity function $v_t$ through
\[
\partial_t u_t = v_t u_t, \qquad u_0=1,
\]
so that
\[
u_t(x)=\exp\left(\int_0^t v_s(x)\,ds\right).
\]
The function $v_t$ is defined on the exterior region $M_t$, namely the region outside the outermost minimal surface $\partial M_t$ in $(M,g_t)$, by
\[
\Delta_{g_t} v_t = 0 \quad \text{on } M_t,
\]
with boundary and asymptotic conditions
\[
v_t|_{\partial M_t}=0, \qquad v_t(x)\to -1 \quad \text{as } |x|\to\infty.
\]
This is the conformal flow of metrics in the formulation used by Han and Khuri [1409.0067].

Two basic conformal formulas control the evolution. First, the scalar curvature transforms by
\[
R_{g_t} = u_t^{-4} R_g.
\]
Equivalently, the quantity $u_t^5R_{g_t}$ evolves linearly because $v_t$ is $g_t$-harmonic. Second, for a conformal change $\tilde g = w^4 g$, mean curvature transforms by
\[
H_{\tilde g}=w^{-2}\left(H_g + 4w^{-1}\nu(w)\right).
\]
These formulas are not merely formal identities: they are the analytic core of the later doubling argument and the mass monotonicity mechanism.

A crucial preserved quantity is the area of the outermost minimal surface. Because $v_t=0$ on $\partial M_t$ and $\partial M_t$ is minimal, the area function $A(t)$ remains constant in time. This fixed-area property is one of the reasons conformal flow is adapted to Penrose-type inequalities.

## 3. Doubling argument, mass monotonicity, and the Schwarzschild limit

A core technical step is to pass to a doubled manifold across the outermost minimal surface. Writing
\[
\bar M_t = M_t^- \cup M_t^+,
\]
one defines on the two sides
\[
g_t^\pm = (w_t^\pm)^4 g_t, \qquad w_t^\pm = \frac{1\pm v_t}{2}.
\]
The doubled metric is $\bar g_t = g_t^- \cup g_t^+$. Since $v_t=0$ on the boundary, the two sides glue smoothly across $\partial M_t$. The mean curvature transformation gives
\[
H_t^- = - H_t^+,
\]
and this cancellation is central to the doubling construction [1409.0067].

Han and Khuri review two complementary mass formulas. One is spinorial: harmonic spinors on the doubled manifold, together with the Lichnerowicz formula, yield an integral representation for the doubled ADM energy in terms of gradient terms and scalar-curvature terms. The other is non-spinorial: one solves
\[
\Delta_{g_t} z_t - \frac18 R_{g_t} z_t =0 \quad \text{on } M_t,
\]
with $z_t \to 1$ at infinity and scalar-flat conformal metric $z_t^4 g_t$. After integration by parts this gives a lower bound for the doubled energy by a nonnegative bulk integral. In both versions, if $R_g \ge 0$, or more generally if the relevant scalar-curvature contribution is nonnegative, then the doubled energy is nonnegative.

Under harmonic flatness at infinity, the asymptotic expansion of the flow yields a differential relation for the physical mass $m_t$,
\[
\dot m_t = -2\tilde m_t,
\]
where $\tilde g_t=(1-v_t)^4g_t$ and $\tilde m_t$ is its ADM energy. Integrating gives the refined identity
\[
E_{\mathrm{ADM}} - E_{\mathrm{ADM}}(\infty) = 2\int_0^\infty \tilde E_{\mathrm{ADM}}(t)\,dt.
\]
Since the flow converges, after rescaling, to the exterior region of a constant-time Schwarzschild slice, and since the horizon area is constant, one obtains
\[
E_{\mathrm{ADM}} - \sqrt{\frac{|\partial M|}{16\pi}} = 2\int_0^\infty \tilde E_{\mathrm{ADM}}(t)\,dt.
\]
Han and Khuri emphasize that this is stronger than merely the inequality, because it identifies the mass defect as an integral of a nonnegative quantity.

The long-time Schwarzschild limit explains the geometric logic of the method. The flow is arranged so that the initial data are deformed toward the explicit equality model while preserving the horizon area. Once positivity of the instantaneous energy along the flow is secured, the Penrose inequality follows by comparing the initial mass to the Schwarzschild end state.

## 4. Generalized Jang deformation for non-time-symmetric data

For general initial data $(M,g,k)$, the main obstacle is that the scalar curvature of $g$ need not be nonnegative. The method in [1409.0067] addresses this by passing to a generalized Jang graph
\[
\Sigma = \{t=f(x)\}\subset M\times \mathbb R
\]
inside the warped product
\[
(M\times \mathbb R,\ g+\phi^2dt^2),
\]
where the warping factor $\phi \ge 0$ is to be chosen. The induced metric on the graph is
\[
\bar g = g + \phi^2 df^2.
\]
The graph function satisfies the generalized Jang equation
\[
H_\Sigma - \operatorname{Tr}_\Sigma K =0.
\]

The essential scalar-curvature identity for the Jang metric is
\[
\bar R = 16\pi(\mu - J(w)) + |h-K|_\Sigma|^2 + 2|q|^2 - 2\phi^{-1}\operatorname{div}_{\bar g}(\phi q).
\]
Because $\mu \ge |J|$, the first three terms are nonnegative except for the divergence term. Han and Khuri therefore describe $\bar R$ as “weakly nonnegative.” This formula is the bridge between the Einstein constraint quantities and a Riemannian positivity argument.

The paper also reviews the boundary and asymptotic behavior relevant for blow-up solutions near apparent horizons. Near $\partial M$, assumptions of the form
\[
|\theta_\pm(S_\tau)| \le c\,\tau^l
\]
and
\[
\phi(x)=\tau^b\widetilde\phi(x), \qquad \widetilde\phi>0
\]
are used, while at spatial infinity
\[
\phi(x)=1+\frac{C}{|x|}+O(|x|^{-2}).
\]
Solutions then satisfy
\[
\nabla^m f(x)=O(|x|^{-1/2-m}), \qquad m=0,1,2,
\]
which ensures equality of the ADM energies of $(M,g,k)$ and the Jang surface $(\Sigma,\bar g)$.

For the Penrose problem, Han and Khuri do not work on all of $M$ but on the region $M_0$ outside the outermost minimal area enclosure of $\partial M$. The boundary condition becomes
\[
H_{\partial \Sigma_0}=0,
\]
with a conjectured Neumann-type condition on $\partial M_0$ designed to enforce the correct horizon behavior. This setup ensures that $\partial\Sigma_0$ is minimal, so Bray’s conformal flow can be run on the Jang surface.

## 5. Coupling the conformal flow to the Jang surface

The central innovation of [1409.0067] is to perform Bray’s conformal flow on the Jang surface $(\Sigma_0,\bar g)$ and to use the flow itself to define the warping factor $\phi$. Starting from a generalized Jang solution over $M_0$ with $H_{\partial\Sigma_0}=0$, one runs
\[
\bar g_t = u_t^4 \bar g
\]
with the same flow equations and boundary conditions as in the Riemannian case. The scalar curvature transforms as
\[
R_{\bar g_t}=u_t^{-4}\bar R.
\]
Applying the conformal-flow mass identity then gives
\[
E_{\mathrm{ADM}} - \sqrt{\frac{|\partial \Sigma_0|}{16\pi}} \ge 2\int_0^\infty \tilde E_{\mathrm{ADM}}(t)\,dt.
\]

The remaining obstruction is still the divergence term in $\bar R$. The new idea is to choose $\phi$ from the time-integrated weights appearing in the conformal-flow mass identity. With the scalar auxiliary formulation, Han and Khuri set
\[
\phi = 2\int_0^\infty \chi_t\left[(w_t^-z_t^-)^2 + (w_t^+z_t^+)^2\right]u_t^2\,dt,
\]
and a corresponding spinorial formula is also given. Here $\chi_t$ is the characteristic function of the exterior flow region $\Sigma_t$. The paper records the key properties
\[
\phi(x)>0 \ \text{away from } \partial\Sigma_0, \qquad \phi(x)\to 1 \ \text{as } |x|\to\infty, \qquad \phi|_{\partial\Sigma_0}=0.
\]

These properties are decisive. The positivity of $\phi$ in the interior means that the generalized Jang equation becomes nondegenerate away from the boundary. The vanishing of $\phi$ on $\partial\Sigma_0$ is exactly what is needed to annihilate the boundary term after integrating the divergence term by parts. The limit $\phi \to 1$ at infinity matches asymptotically flat warped-product geometry and preserves ADM energy.

Substituting the Jang scalar-curvature formula into the time-integrated conformal-flow identity and using Fubini, the divergence term is converted into a boundary term
\[
2\int_{\partial\Sigma_0}\phi\, q(N)\,d\sigma_{\bar g},
\]
which vanishes because $\phi|_{\partial\Sigma_0}=0$. One is left with a bulk integral weighted by
\[
16\pi(\mu-J(w))+|h-K|_\Sigma|^2+2|q|^2,
\]
hence nonnegative under the dominant energy condition. The resulting inequality is
\[
E_{\mathrm{ADM}} - \sqrt{\frac{|\partial\Sigma_0|}{16\pi}} \ge \text{nonnegative bulk terms},
\]
and since $|\partial\Sigma_0| \ge A_{\min}$, the Penrose inequality follows provided the coupled PDE system can be solved.

This synthesis is the distinctive feature of the method. The conformal flow is no longer only a mass-monotonicity device; it also manufactures the correct warping factor for the generalized Jang equation.

## 6. Conditional theorem, equality case, and significance

The principal theorem of [1409.0067] is explicitly conditional: the Penrose conjecture, together with its rigidity statement, is reduced to solving a naturally coupled PDE system involving the generalized Jang equation and the conformal flow of metrics. The unknowns are the Jang graph function $f$, the warping factor $\phi$, the flow variables $u_t$ and $v_t$ for each $t\ge 0$, and optionally the auxiliary spinorial or scalar variables $\psi_t^\pm$ or $z_t^\pm$. The generalized Jang equation is quasilinear elliptic when $\phi>0$, and the conformal-flow variables are defined by harmonic or elliptic problems at each time.

The paper is equally clear about what is not proved. It does not establish a full existence theory for the coupled system; it does not give a complete regularity theory for all steps needed in the coupled setting; and it does not claim that long-time existence and convergence of the conformal flow in the generalized Jang environment are already resolved. A common misconception is therefore avoided explicitly: the work is not a proof of the full non-time-symmetric Penrose inequality, but a reduction of that conjecture to a PDE problem with favorable structure.

The equality case is analyzed through the nonnegative bulk identity. If equality holds, then all nonnegative terms vanish. The paper concludes that both spinors are parallel, hence $|\psi_t^\pm|=1$, and therefore
\[
\mu=|J|=0,\qquad h=K|_\Sigma,\qquad q=0,
\]
so $\bar R=0$. The doubled conformal-flow manifolds are flat, and $(\Sigma_0,\bar g)$ must be isometric to the exterior $t=0$ Schwarzschild slice
\[
SC^4=\left(\mathbb R\times(\mathbb R^3\setminus B_{m_0/2}(0)),\ -\phi_{sc}^2dt^2+g_{sc}\right),
\]
where
\[
\phi_{sc}(x)=\frac{1-\frac{m_0}{2r}}{1+\frac{m_0}{2r}}, \qquad
g_{sc}=\left(1+\frac{m_0}{2r}\right)^4\delta.
\]
In fact, the explicit conformal flow in Schwarzschild yields
\[
\phi(x)=\phi_{sc},
\]
showing that the coupled construction reproduces the Schwarzschild lapse exactly.

Within the Bray–Khuri program, the significance of the method lies in the replacement of earlier inverse-mean-curvature-flow-based choices of $\phi$ by a conformal-flow-defined warping factor. Han and Khuri note that the IMCF-based choice could vanish in the interior when the flow jumps, causing severe degeneracy in the generalized Jang equation. By contrast, the conformal-flow choice gives an explicit $\phi$ that is strictly positive in the interior, vanishes exactly where needed on the boundary, tends to $1$ at infinity, and reproduces Schwarzschild exactly. This suggests that Bray’s conformal flow is not merely an alternative to inverse mean curvature flow in the Riemannian case, but a structurally better coupling mechanism for the generalized Jang equation in the full Penrose problem.

Source: https://www.emergentmind.com/topics/bray-s-conformal-flow-method