Gromov–Hausdorff–Prokhorov Topology

• Gromov–Hausdorff–Prokhorov topology is a metric structure integrating both geometric and measure convergences for measured metric spaces.
• It employs embedding and correspondence techniques to rigorously analyze convergence in both compact and locally compact settings.
• The topology yields a Polish space framework, underpinning probabilistic limit theorems and scaling limits for random geometric models.

The Gromov–Hausdorff–Prokhorov (GHP) topology provides a foundational metric structure on classes of measured metric spaces, equipping the space of isometry classes of these objects with a robust notion of convergence that simultaneously captures both geometric and measure-theoretic aspects. It generalizes the classical Gromov–Hausdorff topology for metric spaces by integrating the Prokhorov metric for Borel measures, supporting intricate probabilistic limit theorems for random geometric objects, such as Lévy random trees and scaling limits of large combinatorial structures. Both the compact and noncompact (locally compact) settings are encompassed, yielding Polish spaces—complete, separable, and suitable for probability—crucial for stochastic process convergence and random metric measure analysis (Abraham et al., 2012).

1. Rooted Measured Metric Spaces and GHP-Isometry

A rooted measured metric space is formally denoted $X = (X, d, \rho, \mu)$, where $(X, d)$ is a complete separable metric space, $\rho \in X$ is a distinguished root, and $\mu$ is a Borel measure (finite or locally finite, according to context) (Abraham et al., 2012). Two such spaces $X$ and $X'$ are GHP-isometric if there exists an isometry $\varphi: X \to X'$ with $\varphi(\rho) = \rho'$ and $\varphi_* \mu = \mu'$. Spaces are identified up to GHP-isometry, ensuring the quotient space reflects genuine equivalence of structure and measure.

This rooted formalism is critical for localization in the non-compact context, enabling the definition of local convergence in terms of balls centered at the root and facilitating integration over radii in the general case.

2. Gromov–Hausdorff–Prokhorov Distance: Definition and Characterizations

Compact Case

For $K$, the set of GHP-isometry classes of compact rooted measured metric spaces with $(X, d)$0, the GHP distance is defined by

$(X, d)$1

where $(X, d)$2 ranges over all Polish metric spaces, and $(X, d)$3 and $(X, d)$4 are isometric embeddings (Abraham et al., 2012). Here, $(X, d)$5 and $(X, d)$6 are the Hausdorff and Prokhorov distances, respectively, defined for compact subsets and finite measures on $(X, d)$7.

An equivalent correspondence-based formula is

$(X, d)$8

where $(X, d)$9 is a correspondence (with surjective projections) measuring maximal distortion, and $\rho \in X$0 is a measure coupling with prescribed marginals; $\rho \in X$1 denotes the complement.

Non-Compact (Locally Compact) Case

For locally compact length spaces with locally finite measures, the distance is defined via localization: For $\rho \in X$2,

$\rho \in X$3

where $\rho \in X$4 denotes the restriction of $\rho \in X$5 to the closed ball of radius $\rho \in X$6 around the root, equipped with the restricted measure. The map $\rho \in X$7 is càdlàg (right-continuous with left limits), guaranteeing the integral metrizes the local GHP convergence (Abraham et al., 2012).

Key Properties

• $\rho \in X$8 is a metric satisfying symmetry, the triangle inequality, and positive definiteness.
• The topology induced by $\rho \in X$9 on isometry classes aligns with probabilistic convergence (weak, vague, Gromov–Hausdorff–vague as developed in (Athreya et al., 2014)).
• Equivalence between embedding- and correspondence-based characterizations is established via ambient construction techniques and gluing arguments.

3. Topological Properties: Compactness, Separability, Polishness

Compactness and Precompactness Criteria

• In the compact case, a set $\mu$0 is relatively compact iff
  1. $\mu$1
  2. For every $\mu$2, each $\mu$3 admits an $\mu$4-net of size at most $\mu$5 for some $\mu$6
  3. $\mu$7 (Abraham et al., 2012)

This combines Gromov's criterion for compactness with tightness of measures (Prokhorov’s theorem).

• In the non-compact case (e.g., boundedly-compact rooted spaces), precompactness is characterized analogously by uniform control over the metric entropy and masses of balls of all radii (Noda, 2024).

Separability and Completeness

Both the compact and non-compact spaces of GHP-isometry classes are separable, as countable dense sets are constructed by considering finite metric spaces with rational parameters and measures supported on finitely many points (Abraham et al., 2012, Noda, 2024, Khezeli, 2018). Completeness is established by diagonal arguments within balls of increasing radius and standard measure-theoretic techniques.

Polish Space Structure

As the spaces are both complete and separable under $\mu$8 and its local variants, the full space of GHP-isometry classes is Polish, making it suitable for probabilistic limit theorems and stochastic processes (Abraham et al., 2012, Athreya et al., 2014, Khezeli, 2018, Khezeli, 2019, Noda, 2024).

4. Dimensional and Manifold Structure

The topological dimension of Gromov–Hausdorff and Gromov–Prokhorov spaces has been precisely established. For spaces of isometry classes of finite metric spaces with $\mu$9 points, $X$0, and for metric measure spaces with $X$1 points, $X$2 (Nakajima et al., 17 Feb 2025). Spaces with unrestricted cardinality are strongly infinite-dimensional, as the Hilbert cube embeds in them topologically. Compact classes are strongly countable-dimensional, being countable unions of finite-dimensional closed subspaces.

Additionally, strongly non-embeddability results are obtained: finite-dimensional subspaces cannot be embedded in lower-dimensional metric spaces, and the full spaces contain every compact metrizable space.

5. Applications in Probability, Random Geometry, and Scaling Limits

The GHP topology is the canonical ambient for convergence of random geometric objects with measure, covering discrete-to-continuum limits, random planar maps, and various random trees (notably Lévy, Brownian, and $X$3-stable continuum trees) (Abraham et al., 2012, He et al., 2017, Addario-Berry et al., 2015). Coding functions mapping excursions to real trees are continuous for the GHP topology, supporting the analysis of measured continuum random trees as limits of discrete structures.

Applications encompass:

• Scaling limits of random quadrangulations and their cores: joint convergence in the GHP topology to the Brownian map (Addario-Berry et al., 2015).
• Joint convergence of random trees and their associated cut-trees to CRT and its fragmentation in GHP sense (He et al., 2017).
• Markov process invariance principles on tree-like metric measure spaces, exploiting GHP-vague and Gromov-vague topologies (Athreya et al., 2014).

The lower mass-bound property is pivotal for equivalence between sampling-based and support-based (Gromov–vague vs. Gromov–Hausdorff–vague) convergence (Athreya et al., 2014).

6. Generalizations and Functorial Extensions

Recent frameworks formalize decorated Gromov–Hausdorff–type metrics, providing a functorial approach that extends the GHP topology to spaces equipped with additional random objects—e.g., stochastic processes, marked points, or field laws—by viewing these as functors from the base space to suitable metric spaces (Noda, 2024, Khezeli, 2018). For boundedly-compact rooted metric spaces,

$X$4

where $X$5 is a suitable functor (Borel measures, paths, etc.).

This allows metrization of spaces such as random measured metric spaces carrying processes indexed by the space, with a unified Polish topology supporting classical and process-level tightness arguments.

7. Proof Architecture and Key Lemmas

Key technical results include:

• The equivalence of embedding-based and correspondence/coupling formulations (Abraham et al., 2012), exploiting ambient metric space constructions and lifting arguments.
• Triangle inequality by concatenation of embeddings/correspondences and controlled gluing.
• Positive definiteness: from vanishing GHP distance constructing root- and measure-preserving isometries.
• Completeness and separability via diagonalization on finite-radius balls and tightness of measures.
• Strassen-type theorems: coupling-based characterizations for measures and spaces with approximate equalities (Khezeli, 2019).

These foundations enable probabilistic analysis, invariance principles, and stochastic process convergence in a broad class of random geometric models.

---

References:

• "A note on Gromov–Hausdorff–Prokhorov distance between (locally) compact measure spaces" (Abraham et al., 2012)
• "Topological dimension of the Gromov-Hausdorff and Gromov-Prokhorov spaces" (Nakajima et al., 17 Feb 2025)
• "The gap between Gromov-vague and Gromov-Hausdorff-vague topology" (Athreya et al., 2014)
• "Joint convergence of random quadrangulations and their cores" (Addario-Berry et al., 2015)
• "Metrization of the Gromov-Hausdorff (-Prokhorov) Topology for Boundedly-Compact Metric Spaces" (Khezeli, 2019)
• "A Unified Framework for Generalizing the Gromov-Hausdorff Metric" (Khezeli, 2018)
• "Metrization of Gromov-Hausdorff-type topologies on boundedly-compact metric spaces" (Noda, 2024)
• "Gromov-Hausdorff-Prokhorov convergence of vertex cut-trees of n-leaf Galton-Watson trees" (He et al., 2017)