---
title: 'Spacetime Brakke Flow: Geometric Measure View'
url: https://www.emergentmind.com/topics/spacetime-brakke-flow
type: topic
---

# Spacetime Brakke Flow: Geometric Measure View

Searching arXiv for the cited paper and closely related work on spacetime Brakke flow, parabolic rectifiability, and canonical space-time measures.
Spacetime Brakke flow is a space–time formulation of weak mean curvature flow in the varifold setting. In its most basic form, it is a Brakke flow tested against time-dependent weights \(\phi(x,t)\), so that the evolution inequality acquires the term \(\partial_t\phi\). In a stronger geometric-measure-theoretic form, the flow is encoded by a canonical measure on spacetime, either the “space-time track” \(d\|V_t\|\,dt\) or a Radon measure on \(J\times \mathbf G_k(U)\), whose pushforward to \(J\times U\) records the evolving mass distribution. In the multi-phase setting, this spacetime viewpoint can be coupled to a \(BV\) partition \(\{E_i(t)\}\), producing a canonical \(BV\)–Brakke flow in which the phase boundaries move with normal velocity \(h\cdot \nu_i\) and satisfy explicit transport identities [2109.14415] [2512.19227] [2606.22441].

## 1. Classical Brakke flow in spacetime form

For a \(k\)-dimensional Brakke flow \(\{V_t\}_{t\in[0,T)}\) in an open set \(U\subset \mathbf R^{n+1}\), Brakke’s inequality with a spacetime test function \(\phi\in C_c^1(U\times[0,T);\mathbf R^+)\) takes the form
\[
\|V_{t_2}\|(\phi(\cdot,t_2))-\|V_{t_1}\|(\phi(\cdot,t_1))
\le
\int_{t_1}^{t_2}\!\!\int_U
\Big(
-\phi |h(x,V_t)|^2+\nabla \phi\cdot h(x,V_t)+\partial_t\phi
\Big)\,d\|V_t\|(x)\,dt.
\]
If \(\Gamma(t)\) is a smooth mean curvature flow with \(v=h\) and \(V_t=\mathrm{var}(\Gamma(t),1)\), equality holds, recovering the classical transport identity [2109.14415].

The time-dependent-test-function formulation is not merely notational. It is one of the “usual” definitions of Brakke flow, and the pointwise, integrated, time-independent, and time-dependent formulations are equivalent under mild assumptions. A central point is the correction of Brakke’s \(\S 3.5\) argument for time-dependent test functions, which justifies the spacetime inequality and the passage between these formulations. In particular, the corrected pointwise spacetime inequality is
\[
V_t(\phi_t)\big|_{t=s}\le B(V_s,\phi_s,u_s)+V_s(\partial_t\phi|_{t=s}),
\]
and the integrated form with time-dependent \(\phi\) is equivalent to the classical differential inequality [1705.08789].

From the spacetime perspective, the canonical track of a classical Brakke flow is the measure
\[
d\mu=d\|V_t\|\,dt
\]
on \(\mathbf R^{n+1}\times \mathbf R^+\). Brakke’s inequality is then a spacetime distributional inequality: the term \(\partial_t\phi\) is the time derivative of a spacetime test function, and the support of \(\mu\) encodes the space-time track of the moving varifold. This viewpoint is central in Ilmanen-type clearing-out and monotonicity arguments, and it is the starting point for later space-time-Grassmann formulations [2109.14415].

## 2. Canonical spacetime measures and space-time-Grassmann formulations

A precise measure-theoretic realization of spacetime Brakke flow packages the evolving varifolds into a Radon measure on spacetime and the Grassmann bundle. For a \(k\)-dimensional Brakke flow over \(J\times U\), one defines
\[
V(\psi)=\int_J\Big(\int \psi(t,x,S)\,dV(t)_{(x,S)}\Big)\,dt,
\]
a Radon measure on \(J\times G_k(U)\), and its canonical spacetime weight
\[
\|V\|(f)=\int_J\Big(\int f(t,x)\,d\mu(t)_x\Big)\,dt,
\qquad d\|V\|(t,x)=d\|V(t)\|(x)\,dt.
\]
This gives a canonical space-time-Grassmann measure whose pushforward to \(J\times U\) is the space-time track [2606.22441] [2512.19227].

In the codimension \(1\) formulation of Buet–Leonardi–Masnou–Sagueni, a spacetime Brakke flow is a finite Radon measure \(\lambda\) on \(\mathbf R^n\times G_{d,n}\times[0,1]\) with disintegration
\[
\lambda=\mu(t)\otimes \nu_{(x,t)}\otimes dt,
\]
where \(\mu(t)\) is the mass measure at time \(t\) and \(\nu_{(x,t)}\) is the tangent-plane distribution. The spacetime first variation is
\[
\delta\lambda(X)=\int \mathrm{div}_S X(y,t)\,d\lambda(y,S,t),
\]
and if \(\delta\lambda\) has no singular part with respect to \(\|\lambda\|\), then there exists a spacetime mean curvature \(h(\cdot,\cdot,\lambda)\) satisfying
\[
\int \mathrm{div}_S X\,d\lambda=-\int X\cdot h\,d(\mu(t)\otimes dt).
\]
The corresponding spacetime Brakke inequality is
\[
\mu(t_2)(\phi(\cdot,t_2))-\mu(t_1)(\phi(\cdot,t_1))
\le
-\int_{t_1}^{t_2}\!\int \phi\,|h|^2\,d\mu(t)\,dt
+\int_{t_1}^{t_2}\!\int S^\perp(\nabla\phi)\cdot h\,d\lambda
+\int_{t_1}^{t_2}\!\int \partial_t\phi\,d\mu(t)\,dt.
\]
Immediate consequences are mass decay and the \(L^2\)-bound \(\int_0^1\int |h|^2\,d\mu(t)\,dt\le \mu(0)(\mathbf R^n)\) [2509.06441].

This measure-theoretic packaging is not merely formal. It yields a new definition of Brakke flow as a spacetime measure satisfying a distributional Brakke inequality, and this definition is equivalent to the classical ones. Moreover, left and right time-slice representatives extracted from the spacetime measure are classical Brakke flows, and the mean curvature vector, density, and tangent map along the flow are measurable with respect to the spacetime weight measure [2512.19227]. A further consequence is that standard Ilmanen convergence of Brakke flows is equivalent to weak convergence of the corresponding space-time-Grassmann measures [2606.22441].

## 3. Multi-phase \(BV\)–Brakke flow and canonical selection

In the multi-phase setting of Stuvard–Tonegawa, one considers \(N\ge 2\) grains with an \(L^{n+1}\)-partition \(\{E_i(t)\}_{i=1}^N\) of \(\mathbf R^{n+1}\): pairwise disjoint open sets whose union fills space up to null sets, each \(E_i(t)\) having locally finite perimeter. Interfaces are the reduced-boundary intersections
\[
I_{i,j}(t):=\partial^*E_i(t)\cap \partial^*E_j(t),
\]
and the surface energy is the total variation or perimeter. The construction is isotropic and equal-tension; anisotropy is not addressed in that framework [2109.14415].

The main existence theorem states that if \(\Gamma_0\) is a closed countably \(n\)-rectifiable set with finite weighted area \(\int_{\Gamma_0}\Omega\,d\mathcal H^n<\infty\), where \(\Omega\in C^2(\mathbf R^{n+1})\), \(0<\Omega\le 1\), \(|\nabla\Omega|\le c_1\Omega\), and \(\|\nabla^2\Omega\|\le c_1\Omega\), and if \(E_{0,1},\dots,E_{0,N}\) form an \(L^{n+1}\)-partition with
\[
\Gamma_0=\mathbf R^{n+1}\setminus \bigcup_{i=1}^N E_{0,i},
\]
then there exist a Brakke flow \(\{V_t\}_{t\ge 0}\) and open sets \(\{E_i(t)\}_{t\ge 0}\) such that \(\|V_0\|=\mathcal H^n\llcorner \Gamma_0\), \(E_i(0)=E_{0,i}\), and with \(\Gamma(t):=\mathbf R^{n+1}\setminus \bigcup_i E_i(t)\),
\[
\mathcal H^{n-1+\delta}\big(\Gamma(t)\,\Delta\, \mathrm{spt}\|V_t\|\big)=0
\quad\text{for any }\delta>0\text{ and a.e. }t\ge 0,
\]
together with the distributional identity
\[
\frac{d}{dt}\int_{E_i(t)}\phi\,dx
=
\int_{\partial^*E_i(t)} \phi\, h\cdot \nu_i\,d\mathcal H^n
+\int_{E_i(t)}\partial_t\phi\,dx.
\]
If \(\{V_t\}\) is locally unit density, then locally
\[
\mathcal H^n\Big(\Gamma(t)\,\Delta\,\bigcup_{i=1}^N \partial^*E_i(t)\Big)=0,
\qquad
\|V_t\|=\mathcal H^n\llcorner \Big(\bigcup_i \partial^*E_i(t)\Big)
\]
for a.e. \(t\) [2109.14415].

A key feature is the explicit volume-change identity. For any bounded open \(U\) and \(0\le t_1<t_2<\infty\),
\[
\mathcal L^{n+1}(U\cap E_i(t_2))-\mathcal L^{n+1}(U\cap E_i(t_1))
=
\int_{t_1}^{t_2}\!\!\int_{U\cap \partial^*E_i(t)} h\cdot \nu_i\,d\mathcal H^n\,dt.
\]
The identity remains valid even if \(\partial^*E_i(t)\) carries portions where \(\|V_t\|\) has higher multiplicity or if \(\partial E_i(t)\setminus \partial^*E_i(t)\) contains interior boundary. In \(BV\) spacetime form, with \(\chi_i=\chi_{E_i(t)}\),
\[
\left.\int \chi_i\,\phi\,dx\right|_{t=t_1}^{t_2}
=
\int_{t_1}^{t_2}\!\!\int \chi_i\,\partial_t\phi\,dx\,dt
-
\int_{t_1}^{t_2}\!\!\int \phi\, h\cdot d\nabla\chi_{E_i(t)}\,dt,
\]
and
\[
\nabla' \chi_i(x,t)=\big(-\nu_i(x,t),\, h(x,V_t)\cdot \nu_i(x,t)\big)\,
d\mathcal H^n\llcorner \partial^*E_i(t)\,dt.
\]
This upgrades Brakke’s inequality to a transport equality for the phases [2109.14415].

The same theorem provides global energy estimates. In the weighted case,
\[
\|V_t\|(\Omega)\le \|V_0\|(\Omega)\,\exp(c_1^2 t/2),
\qquad
\int_0^t\int |h|^2\,\Omega\,d\|V_s\|\,ds<\infty,
\]
and if \(\mathcal H^n(\Gamma_0)<\infty\), then
\[
\|V_t\|(\mathbf R^{n+1})+\int_0^t\!\!\int |h|^2\,d\|V_s\|\,ds
\le \mathcal H^n(\Gamma_0).
\]
This suggests a canonical selection principle: Brakke’s inequality alone allows non-unique behaviors such as mass drop, whereas the coupled \(BV\) partition and the explicit transport identity constrain the motion by tying it to \(h\cdot \nu_i\) [2109.14415].

## 4. Geometry of the spacetime track: tangents, rectifiability, and densities

Recent work sharpens the geometric structure of the canonical spacetime measure. For a \(k\)-dimensional Brakke flow, the support of the canonical space-time measure \(\|V\|\) is a vertical parabolic \((k+2)\)-rectifiable set. The scaling is exactly the parabolic one: a \(k\)-dimensional spatial measure scales like \(\lambda^k\), while \(dt\) scales like \(\lambda^2\), so \(d\|V(t)\|\,dt\) scales like \(\lambda^{k+2}\) [2606.22441].

Parabolic blow-up is defined by
\[
V^{t_0,x_0;r}(\alpha)=
r^{-(k+2)}\!\int
\alpha\big(r^{-2}(t-t_0),r^{-1}(x-x_0),S\big)\,dV(t,x,S),
\]
with weight
\[
\|V^{t_0,x_0;r}\|
=
r^{-(k+2)}(T_{(t_0,x_0),r})_\#\|V\|,
\qquad
T_{(t_0,x_0),r}(s,y)=\big(r^{-2}(s-t_0),r^{-1}(y-x_0)\big).
\]
At \(\|V\|\)-almost every \((t,x)\), there exists a unique static planar tangent flow,
\[
\Big\{
\Theta^k(\|\langle V,t\rangle\|,x)\cdot
\mathcal H^k\llcorner \mathrm{Tan}^k(\langle V,t\rangle,x)
:\ s\in \mathbf R
\Big\},
\]
and the standard densities agree:
\[
\Theta(V,t,x)
=
\mathbf\Theta^k(\|\langle V,t\rangle\|,x)
=
(2\boldsymbol{\alpha}(k))^{-1}\Theta^{k+2}_\rho(\|V\|,t,x)
=
\boldsymbol{\beta}(k)^{-1}\Theta^{k+2}_d(\|V\|,t,x).
\]
Equivalently, \(\|V\|\) is represented by parabolic Hausdorff measure with density weight [2606.22441].

The parabolic-rectifiability result complements older spacetime track arguments. In the multi-phase setting, Ilmanen’s clearing-out lemma yields
\[
\mathcal H^{n-1+\delta}\big((\mathrm{spt}\,\mu)_t\setminus \mathrm{spt}\|V_t\|\big)=0
\quad\text{a.e. }t,
\]
and \(\Gamma(t)\) is \(\mathcal H^n\)-equivalent to \(\mathrm{spt}\|V_t\|\). At \(\mu\)-almost every \((x,t)\), the spacetime track has a distinguished tangent direction,
\[
\begin{pmatrix}
h(x,V_t)\\
1
\end{pmatrix}
\in T_{(x,t)}\mu,
\]
and on the spacetime reduced boundary of phase \(i\),
\[
T_{(x,t)}\mu
=
\big(T_x(\partial^*E_i(t))\times\{0\}\big)
\oplus
\mathrm{span}
\begin{pmatrix}
h(x,V_t)\\
1
\end{pmatrix}.
\]
These formulas express the geometric fact that the spacetime tangent contains both the spatial tangent plane and the velocity direction [2109.14415].

A plausible implication is that the spacetime track is not only an auxiliary bookkeeping device but a genuine rectifiable object with tangent geometry, density theory, and convergence notions of its own. That interpretation is explicit in the space-time-Grassmann approach, where convergence of Brakke flows is equivalent to weak convergence of the associated spacetime measures [2606.22441] [2512.19227].

## 5. Regularity, avoidance, and agreement with smooth flow

In codimension \(1\), spacetime Brakke flows satisfy an avoidance principle analogous to the smooth one. If \(V_0\in V_{n-1}(\mathbf R^n)\), \((\mu(t))_{t\in[0,T]}\) is the mass measure of a spacetime Brakke flow starting from \(V_0\), and \((M_t)_{t\in[0,T]}\) is the smooth mean curvature flow of a compact \(C^2\) hypersurface \(M\), then
\[
\mathrm{spt}\,\mu(0)\cap \mathrm{spt}\,M_0=\emptyset
\quad\Longrightarrow\quad
\mathrm{spt}\,\mu(t)\cap \mathrm{spt}\,M_t=\emptyset
\quad\text{for all }t\in[0,T].
\]
If the initial hypersurface is a smooth closed \(C^2\) manifold and the classical mean curvature flow exists smoothly on \([0,s]\), then provided the spacetime flow is nontrivial on \([0,s]\),
\[
\mathrm{spt}\,\mu(t)=\mathrm{spt}\,M_t
\qquad\text{for }t\in[0,s].
\]
The argument uses a spacetime test function supported in a tubular neighborhood of the smooth flow and the signed-distance identity \(r(\partial_t r-\Delta r)\ge 0\) [2509.06441].

Local regularity results also have a genuinely spacetime formulation. For general \(k\)-dimensional Brakke flows in \(\mathbf R^n\), if the flow is locally close to a \(k\)-plane in the sense of measure, then it is locally represented as a smooth graph over that plane with estimates on all derivatives up to the end-time. Moreover, if the Gaussian density at \((x_0,t_0)\) lies in \([1,1+\varepsilon_{\mathrm{White}})\), then in a neighborhood of \((x_0,t_0)\) the flow is a smooth mean curvature flow and extends smoothly up to time \(t_0\). This extends White’s local regularity theorem to general Brakke flows, including forced ones [2212.07727].

Near triple junctions, spacetime regularity becomes more delicate. For planar network flows weakly close in a spacetime region to a static multiplicity-\(1\) triple junction \(J\), Tonegawa–Wickramasekera proved that in a smaller region the flow is classical: three curves come smoothly together at a single point at \(120^\circ\), remain smoothly close to \(J\), and move smoothly. Combined with White’s stratification theorem, this yields a closed singular set \(\Sigma\) of parabolic Hausdorff dimension at most \(1\), such that outside \(\Sigma\) the network flow is classical [1504.01212].

A higher-dimensional analogue is now available. For \(k\)-dimensional possibly forced Brakke flows near a static multiplicity-one triple junction cone \({\bf C}\), an \(\varepsilon\)-regularity theorem gives a \(C^{1,\alpha}\) spine \(\xi\) and three \(C^{1,\alpha}\) sheets \(f_i\) meeting along that spine, provided the flow lies in a small parabolic \(L^2\)-neighborhood of \({\bf C}\) and satisfies a structural slice assumption. The assumption is automatic for two classes singled out in the literature: codimension-one multi-phase \(BV\)-Brakke flows and mod \(3\) current flows arising from Ilmanen’s elliptic regularization [2510.02969]. In dimension \(n=1\), this is consistent with the canonical multi-phase scheme, where for a.e. \(t>0\) the support is locally a finite union of \(W^{2,2}\) curves meeting at angles \(0^\circ,60^\circ,120^\circ\) for \(N\ge 3\) and only \(0^\circ\) for \(N=2\) [2109.14415].

## 6. Constructions, approximation schemes, and variants

Several constructions realize spacetime Brakke flow as a limit of approximations. In the canonical multi-phase theory, the flow is built by a time-discrete algorithm with two steps in each epoch: a Lipschitz regularization \(f_1\in \mathcal E^{vc}(\mathcal E,j)\), which strictly decreases weighted perimeter while controlling volume change, and a mean-curvature step
\[
f_2(x)=x+\Delta t\, h_\varepsilon(x,\partial \mathcal E^*),
\]
where \(h_\varepsilon\) is defined by convolution with a localized heat kernel \(\Phi_\varepsilon\). Compactness of varifolds and \(BV\) partitions, together with uniform energy bounds and Ilmanen’s monotonicity and clearing-out arguments, produces a Brakke flow with associated \(BV\) partition [2109.14415].

A more general construction begins from an arbitrary compactly supported varifold of finite mass, even a point cloud varifold. For fixed \(\varepsilon>0\), one defines an approximate mean curvature \(h_\varepsilon(\cdot,V)\) by convolution and then a time-discrete flow by pushforwards
\[
V_{\varepsilon,T}(t_i)=(f_i)_\# V_{\varepsilon,T}(t_{i-1}),
\qquad
f_i=\mathrm{id}+\Delta t_i\, h_\varepsilon(\cdot,V_{\varepsilon,T}(t_{i-1})).
\]
As the time step tends to \(0\), this yields a unique time-continuous \(\varepsilon\)-approximate flow \(V_\varepsilon(t)\), and for time-dependent test functions \(\phi\) it satisfies the exact equality
\[
\|V_\varepsilon(b)\|(\phi(\cdot,b))-\|V_\varepsilon(a)\|(\phi(\cdot,a))
=
\int_a^b \delta\!\big(V_\varepsilon(t),\phi(\cdot,t)\big)\big(h_\varepsilon(\cdot,V_\varepsilon(t))\big)\,dt
+\int_a^b \|V_\varepsilon(t)\|(\partial_t\phi(\cdot,t))\,dt.
\]
Coupling with \(dt\) gives a spacetime measure \(\lambda_\varepsilon=dt\otimes V_\varepsilon(t)\); along \(\varepsilon_j\downarrow 0\), a subsequence converges to a spacetime Radon measure with \(L^2\)-bounded spacetime mean curvature, and under rectifiability of the time slices the limit is a spacetime Brakke flow [2510.00746].

Allen–Cahn approximations provide another route. On Riemannian manifolds, the energy density measures of the parabolic Allen–Cahn equation converge to rectifiable Radon measures that evolve by Brakke flow in the sense of mean curvature. The spacetime track is the measure \(p=\mu_t\,dt\), and a local almost monotonicity formula, together with clearing-out and vanishing discrepancy, is used to obtain density bounds and the Brakke inequality in the limit [1308.0570]. This route is particularly relevant to elliptic-regularization and diffuse-interface constructions.

There are also constrained variants. For volume-preserving mean curvature flow, a Brakke-type inequality is obtained in which the velocity law is
\[
\vec v=\vec h-\lambda(t)\,\vec \nu
\]
in the smooth case, and in the weak \(BV\)-varifold formulation becomes
\[
\int_0^T\!\int_\Omega \vec v\cdot \phi\,d\mu_t\,dt
=
\int_0^T\!\int_\Omega
\Big(
\vec h-\lambda(t)\,\frac{d\|\nabla\psi(\cdot,t)\|}{d\mu_t}\,\vec \nu
\Big)\cdot \phi\,d\mu_t\,dt.
\]
The resulting flow fits the spacetime Brakke framework with time-dependent test functions, but the \(\lambda^2\)-term is replaced by a local scale-dependent error controlled through the phase-field approximation [2505.23222].

These constructions support different geometric regimes. The canonical \(BV\)–Brakke flow is tailored to multi-phase partitions and grain growth; the approximate-mean-curvature scheme applies to very general varifolds and discrete data; Allen–Cahn connects spacetime Brakke flow to phase-field limits; and the volume-preserving variant shows that the spacetime inequality can be adapted to nonlocal constraints [2109.14415] [2510.00746] [1308.0570] [2505.23222].

## 7. Uniqueness, non-uniqueness, and current scope

Non-uniqueness is intrinsic to Brakke flow: mass drop and sudden vanishing are compatible with Brakke’s inequality alone. The multi-phase canonical construction removes part of this redundancy by coupling the varifold flow to \(BV\) grains whose boundaries move with normal velocity \(h\cdot \nu_i\). In dimension \(n+1=2\), if a strong network flow exists initially as in Fischer–Laux–Simon–et al., then the canonical \(BV\)–Brakke flow agrees with it until the first topology change, resolving non-uniqueness in that regime [2109.14415] [2003.05478].

At the support level, codimension \(1\) spacetime Brakke flow satisfies strong comparison with smooth mean curvature flow through the avoidance principle, but uniqueness of the varifold-valued flow is not asserted, and limit spacetime measures may depend on subsequences in approximation procedures [2509.06441] [2510.00746]. More generally, the space-time-Grassmann theory identifies a canonical equivalence class of classical Brakke flows rather than a single representative: left and right time-slice representatives encode jump discontinuities, and any time-slice choice lying between them yields the same spacetime measure [2512.19227].

Several limitations remain explicit. In the canonical multi-phase construction, anisotropic surface tensions are not addressed, general uniqueness beyond the network case remains open, and fine regularity at higher-codimension junctions is unresolved [2109.14415]. In avoidance theory, codimension \(1\) is essential; higher-codimension avoidance can fail [2509.06441]. In the approximate-mean-curvature construction, uniqueness in the \(\varepsilon\to 0\) limit is not guaranteed, stationary varifolds remain stationary at fixed \(\varepsilon\), and identification as a spacetime Brakke flow requires rectifiability of the limit slices [2510.00746].

At the same time, the spacetime viewpoint has clarified the generic geometry of the flow. The support of the canonical spacetime measure is parabolic \((k+2)\)-rectifiable, tangent flows are uniquely static and planar at \(\|V\|\)-almost every point, and Gaussian, spatial, and parabolic densities agree there [2606.22441]. This suggests that the most robust current interpretation of spacetime Brakke flow is not simply “Brakke flow with a \(\partial_t\phi\) term,” but a parabolic geometric-measure object whose time slices, spacetime support, tangent structure, and transport identities can all be studied within a single framework.

Source: https://www.emergentmind.com/topics/spacetime-brakke-flow