minimize $\displaystyle P(E)= \frac 1 2 \sum_{k=0}^{+\infty} P(E_k)$
with fixed $\lvert E_k \rvert$, $\displaystyle\sum_{k=1}^{+\infty} \lvert E_k\vert \lt +\infty$
Concentration compactness?
[Novaga - Paolini - Stepanov - Tortorelli
J. Geom. Anal. 2022]
$$\begin{aligned} \mathbf{E} &= (E_1,\dots,E_k,\dots), \qquad E_k\subset \mathbb{R}^d \\ \partial^* \mathbf E &= \bigcup_{k=1}^{+\infty} \bigcup_{j=0}^{k-1}\partial^* E_k \cap \partial^* E_j \\ \partial \mathbf E &= \left\{ x\mid \forall \rho\, \exists k\colon 0 \lt \lvert E_k\cap B_\rho(x)\rvert \lt \lvert B_\rho(x)\rvert \right\}\\ \end{aligned}$$
Lemma: $P(\mathbf E) = \mathcal{H}^{d-1}(\partial^* \mathbf E)$
Theorem. Given $a_k\gt 0$, $k=1,2,\dots$ such that $\sum \sqrt {a_k} \lt \infty$, there exists a minimal cluster $\mathbf E$ in $\mathbb R^2$ with $\lvert E_k\rvert = a_k$. Moreover $\mathcal H^1(\partial \mathbf E \setminus \partial^* \mathbf E) = 0$.
[Novaga - P. - Stepanov - Tortorelli
networks and heterogeneous media 2023]
proof.
$\mathbf E^n$ minimizing for $a_1,\dots,a_n$.
Uniform estimate on the diameter of the clusters.
$E_k^n \to E_k$, $\quad\mathbf E = (E_1,E_2\dots)$.
Blaschke: $\partial \mathbf E^n \to K$.
Goła̧b: $\mathcal H^1(K)\le \lim \mathcal H^1(\partial \mathbf E^n)$.
Finally
$
P(\mathbf E)
= \mathcal H^1(\partial^* \mathbf E)
\le \mathcal H^1(\partial \mathbf E)
\le \mathcal H^1(K)
\le \lim P(\mathbf E^n)
\le \lim P(\mathbf E)
$.
Remark. $\mathcal H^1(\partial \mathbf E \setminus \partial^* \mathbf E) = 0$ implies $P(\mathbf E) = \mathcal H^1(\partial \mathbf E)$ but does not imply $P(E_k) = \mathcal H^1(\partial E_k)$ for all $k$. Not even in a finite cluster:
[Ambrosio - Caselles - Masnou - Morel
JEMS 2001]
$\mathbf E = (E_1,\dots,E_N)$ is a partition of $\mathbb R^d$. Find $\mathbf E$ minimizing $P(\mathbf E, \Omega)$ on every bounded $\Omega$ with $\lvert E_k \rvert$ fixed, but possibly $\infty$.
[Alama - Bronsard - Vriend
to appear on: Transactions of the AMS ]
$\to$
$\to$
$\to$
[Foisy et al.
Pacific J. Math. 1993]
[Wichiramala
J. Reine Angew. Math 2004]
$\to$
$\to$
$\to$
[Hutchings - Morgan - Ritoré
Annals of Math. 2002]
[Reichardt
J. Geom. Anal. 2008]
[Milman - Neeman
to appear: acta mathematica]
$\mathbf E$ minimizes $P(\mathbf E, \Omega)$ with fixed $\lvert E_j\cap \Omega\rvert$ for all $j\in J \subset (1,2,\dots, N)$.
Theorem. $J$-isoperimetric partitions are closed with respect to $L^1_{\text{loc}}$ convergence.
Theorem. $J$-isoperimetric partitions are locally
isoperimetric if their boundary is flat at infinity
($d\le 7$)
[Novaga - P. - Tortorelli
to appear on: Transactions of the AMS]
[Bronsard - Novack
arXiv 2024]
Locally isoperimetric partitions allow the exchange of any
volume between infinite regions whereas
$J$-isoperimetric
partitions don't.
👉 Large exchange of volumes is required to pass from one notion to the other.
Show that in $\mathbb R^8$ there exists a locally isoperimetric partition with one finite region and two infinite regions which coincides with the Simons' cone at infinity.
$G$ discrete subgroup of $\mathbb R^2$ (lattice).
$\displaystyle \mathbf E = \bigcup_{g\in G}\{E_1+g,E_2+g\}
$ partition of $\mathbb R^d$.
minimize $P(E_1)+P(E_2)$
with $\lvert E_1 \rvert$ and $\lvert E_2\rvert$ fixed.
[Nobili - Novaga - P.
draft]
Minimizers exist and enjoy the same local regularity properties of isoperimetric clusters.
Regions might have many connected components with very small area. Regions can share edges with translated copies of themselves.
breakthrough:
paolini.github.io/2025-talk-Trento