Zoltán M. Balogh,Gergely Kiss,Tamás Titkos,Dániel Virosztek
Abstract:We study $p$-Wasserstein spaces over the branching spaces $\mathbb{R}^2$ and $[0,1]^2$ equipped with the maximum norm metric. We show that these spaces are isometrically rigid for all $p\geq1,$ meaning that all isometries of these spaces are induced by isometries of the underlying space via the push-forward operation. This is in contrast to the case of the Euclidean metric since with that distance the $2$-Wasserstein space over $\mathbb{R}^2$ is not rigid. Also, we highlight that $1$-Wasserstein space is not rigid over the interval $[0,1]$, while according to our result, its two-dimensional analog with the more complicated geodesic structure is rigid. To the best of our knowledge, our results are the first ones concerning the structure of isometries of Wasserstein spaces over branching spaces. Accordingly, our proof techniques substantially deviate from those used for non-branching underlying spaces.
What problem does this paper attempt to address?
The main problem that this paper attempts to solve is to study whether the \(p\)-Wasserstein spaces on the branching spaces \(\mathbb{R}^2\) and \([0,1]^2\) under the maximum - norm metric have isometric rigidity. Specifically, the authors explore whether all isometries in these spaces can be induced by isometries of the underlying spaces. This is in contrast to the situation under the Euclidean metric, where the 2 - Wasserstein space on \(\mathbb{R}^2\) does not possess this rigidity.
### Summary of the main problems:
- **Research objects**: \(p\)-Wasserstein spaces, especially the branching spaces \(\mathbb{R}^2\) and \([0,1]^2\) under the maximum - norm metric.
- **Core problem**: Whether these Wasserstein spaces are isometrically rigid, that is, whether all isometries are induced by isometries of the underlying spaces.
- **Comparative study**: Comparison with Wasserstein spaces under the Euclidean metric, the latter does not possess rigidity in some cases.
### Specific problem description:
The key problems mentioned in the paper are:
1. **Isometric rigidity**: For \(p\geq1\), whether \(W_p(\mathbb{R}^2, d_m)\) and \(W_p([0,1]^2, d_m)\) are isometrically rigid.
2. **Diagonal rigidity**: The concept of diagonal rigidity is introduced, that is, whether all isometries keep the measures with diagonal support unchanged.
3. **The influence of different \(p\) values**: Distinguish between the cases of \(p = 1\) and \(p>1\) because they are different in nature.
### Main conclusions of the paper:
- For \(p\geq1\), \(W_p(\mathbb{R}^2, d_m)\) and \(W_p([0,1]^2, d_m)\) are isometrically rigid.
- It is proved that diagonal rigidity can imply overall rigidity.
- Specific proof methods for diagonal rigidity are proposed, especially for the different cases of \(p = 1\) and \(p>1\).
### Related formulas:
- Maximum - norm metric:
\[
d_m((x_1, x_2), (y_1, y_2))=\max\{|x_1 - y_1|, |x_2 - y_2|\}
\]
- \(p\)-Wasserstein distance:
\[
d_{W_p}(\mu, \nu)=\left(\inf_{\pi\in C(\mu, \nu)}\int_{X\times X}d_m(x, y)^p\,d\pi(x, y)\right)^{\frac{1}{p}}
\]
Through these studies, the authors provide a new perspective for understanding the structure of Wasserstein spaces on branching spaces and show the differences from non - branching spaces.