---
title: Separated Doubling Property in Analysis
url: https://www.emergentmind.com/topics/separated-doubling-property
type: topic
---

# Separated Doubling Property in Analysis

Searching arXiv for the cited papers and related uses of “Separated Doubling Property.”
In current arXiv literature, the expression **Separated Doubling Property** does not denote a single universally standardized definition. Rather, it appears in several mathematically distinct settings in which a doubling condition is coupled to a separation requirement, a separated family of sets, or a weakened choice-dependent form of doubling. One explicit usage occurs for determinantal point processes associated with generalized Fock spaces, where an integral criterion characterizes when the random configuration is almost surely a separated sequence [2502.02237]. Related usages arise for measures that are doubling on several \(n\)-adic grids but fail to be globally doubling [2310.17615, 2009.03875], for weak variants of doubling that depend on a favorable representation of a ball [1902.02184], and for packing statements that bound the cardinality of well-separated sets in doubling metrics [1712.05007]. This suggests that the term functions as an umbrella for several separation-sensitive refinements of doubling rather than as a single fixed definition.

## 1. Foundational doubling notions

A metric space \((X,d)\) is **doubling** with constant \(D\) if every ball of finite radius can be covered by at most \(D\) balls of half the radius:
\[
B_r(x)=\{y\in X:d(x,y)\le r\}.
\]
A quasi-metric version replaces the triangle inequality by
\[
d(x,y)\le K\max\{d(x,z),d(z,y)\},
\]
whenever the right-hand side is defined and finite [1603.07521].

A Borel measure \(\mu\) on a metric space \((X,d)\) is **doubling** if there exists \(C>0\) such that for every \(x\in X\) and every \(r>0\),
\[
\mu(B(x,2r))\le C\,\mu(B(x,r)).
\]
In the measure-theoretic setting, the central issue is growth under radius doubling; in the metric setting, the central issue is covering by smaller balls [1902.02184].

Separation enters through additional structures. In the complex plane, a sequence \(\Lambda=(\lambda_k)_{k\ge 1}\subset\mathbb C\) is **separated** if
\[
\inf_{j\ne k}|\lambda_j-\lambda_k|>0.
\]
In metric geometry, separation often appears as an \(r\)-separation condition on finite sets, while in dyadic and \(n\)-adic analysis it appears through restrictions to families of intervals that are separated by the underlying grid [2502.02237].

## 2. Explicit use in generalized Fock-space determinantal processes

In [2502.02237], the term is used in a precise probabilistic-analytic setting. Let \(\phi:\mathbb C\to\mathbb R\) be subharmonic, let \(\nu=\Delta\phi\) be doubling in the sense that
\[
\nu(D(z,2r))\le C\,\nu(D(z,r)),\qquad z\in\mathbb C,\ r>0,
\]
and define the associated radius function \(\rho(z)>0\) by
\[
\nu(D(z,\rho(z)))=1.
\]
The paper imposes the natural assumption
\[
\lim_{|z|\to\infty}\rho(z)=+\infty.
\]

For the determinantal point process \(\Lambda_\phi\) associated to the generalized Fock space \(F_\phi\), the main characterization is:
\[
\mathbb P(\Lambda_\phi\text{ is separated})=
\begin{cases}
1,& \text{if }\int_{\mathbb C}\frac{dm(z)}{\rho(z)^2}<+\infty,\\[4pt]
0,& \text{if }\int_{\mathbb C}\frac{dm(z)}{\rho(z)^2}=+\infty.
\end{cases}
\]
In that paper, this integral criterion is termed the **Separated Doubling Property** [2502.02237].

The criterion is sharpened by comparison with a Poisson process having the same first intensity,
\[
d\sigma_\phi(z)=\frac{dm(z)}{\rho(z)^2}.
\]
For the canonical weight \(\phi_\alpha(z)=|z|^\alpha\), \(\alpha>0\), one has
\[
\rho(z)\asymp |z|^{1-\alpha/2},
\]
and the associated determinantal process \(\Lambda_\alpha\) is almost surely separated if and only if \(\alpha<4/3\), whereas the corresponding Poisson process \(\Lambda_\alpha^P\) is almost surely separated if and only if \(\alpha<1\) [2502.02237]. The paper interprets this gap as a manifestation of **intrinsic repulsion** in determinantal processes.

## 3. Adic and grid-restricted forms of separated doubling

A distinct strand of the literature concerns measures that satisfy doubling only along selected grids. For \(n\ge 2\), the \(n\)-adic system on \(\mathbb R\) consists of intervals
\[
I=\left[\frac{k-1}{n^m},\,\frac{k}{n^m}\right),\qquad m,k\in\mathbb Z.
\]
A measure \(\mu\) is **\(n\)-adic doubling** if there exists \(C>0\) such that for all sibling \(n\)-adic intervals \(J_1,J_2\),
\[
C^{-1}\le \frac{\mu(J_1)}{\mu(J_2)}\le C.
\]
Every global doubling measure is \(n\)-adic doubling for each \(n\), but the reverse is not true [2310.17615].

The main theorem of [2310.17615] states that for any finite set \(\{n_1,\dots,n_k\}\) of integers \(\ge 2\), there exists a measure that is \(n_i\)-adic doubling for each \(i\), but not globally doubling. An earlier result proves the analogous statement for any finite collection of primes: there exists an infinite family of measures on \(\mathbb R\) that are \(p_i\)-adic doubling for each \(i\), but not globally doubling [2009.03875]. In both papers, the phenomenon is presented as a failure of local-to-global reconstruction.

The constructions are recursive and highly nonuniform. In [2009.03875], weights \(a<1<b\) are assigned to children with
\[
(q-1)a+b=q,
\]
and after \(\alpha\) recursive steps one obtains
\[
\mu(H^{(\alpha)})=\frac{a^\alpha|I|}{q^\alpha},\qquad
\mu(G^{(\alpha)})=\frac{b^\alpha|I|}{q^\alpha},
\]
so that
\[
\frac{\mu(H^{(\alpha)})}{\mu(G^{(\alpha)})}=\left(\frac{a}{b}\right)^\alpha\to 0.
\]
Thus adjacent intervals of equal length can have unbounded measure ratios, even though all selected adic sibling ratios remain uniformly bounded [2009.03875].

This usage motivates a natural interpretation of **separated doubling** as doubling verified on several separated or grid-constrained families of intervals, but not globally. The same mechanism yields applications to reverse Hölder classes, Muckenhoupt \(A_p\) weights, Hardy spaces, BMO, and VMO: finite intersections of \(n\)-adic versions do not recover the full global classes [2310.17615].

## 4. Weak doubling and choice-dependent representations of balls

A third usage emerges from the non-uniqueness of centers and radii in arbitrary metric spaces. In [1902.02184], the paper does **not** introduce a single explicit “Separated Doubling Property” by name, but it studies variants and weakenings of doubling in precisely this direction. The key point is that a ball may admit several different “names,” meaning different choices of center and radius that generate the same set.

Standard doubling requires the inequality
\[
\mu(B(x,2r))\le C\,\mu(B(x,r))
\]
for **all** representations \(B(x,r)\) of a ball. By contrast, **weak doubling** or **weak \(t\)-bling** requires that for every ball \(B\), there exists some representation \(B=B(x,r)\) such that
\[
\mu(B(x,tr))\le C\,\mu(B(x,r)).
\]
When \(t=2\), this is weak doubling; when \(t=3\), weak tripling [1902.02184].

The significance of this weakening is that it exploits non-uniqueness. In ultrametric spaces, every point of a ball can serve as its center, so favorable names can exist even when uniform control over all names fails. The paper states that weak doubling is strictly weaker than standard doubling, and that none of the Besicovitch covering properties is equivalent to weak or standard doubling in arbitrary metric spaces [1902.02184].

The analytic consequences are selective. If \((X,d,\mu)\) is weakly \((1+\sqrt2)\)-bling, then the uncentered Hardy-Littlewood maximal operator is of weak type \((1,1)\) with constant at most \(C^2\). For the centered maximal operator, weak tripling is required, and weak \(t\)-bling with \(t<2\) fails to guarantee weak type \((1,1)\) [1902.02184]. In this sense, the separated or weak form captures genuine geometric information, but less than full doubling.

## 5. Separation, packing, and combinatorial characterizations of doubling

In geometric applications, the phrase “separated doubling property” often refers less to a formal definition than to the consequences of doubling for separated configurations. In [1712.05007], the term is not given as a separate definition; its role is identified with the **packing property** of doubling metrics. If a metric space has doubling dimension \(d\), then every ball of radius \(r\) can be covered by \(2^d\) balls of radius at most \(r/2\), and any \(r\)-separated subset \(X\) contained in a ball of radius \(R\) satisfies
\[
|X|\le \left(\frac{4R}{r}\right)^d.
\]
This bound is the core tool in the proof that greedy spanners are optimal in doubling metrics [1712.05007].

The same paper uses the packing estimate to bound the degree of a cluster in a cluster graph. When neighboring cluster centers are at distance \(>\epsilon\ell\) but lie within a ball of radius \(3\ell\), the number of neighbors is at most
\[
\left(\frac{12}{\epsilon}\right)^d.
\]
Thus separation plus doubling yields polynomial control in the scale ratio, with exponent governed by doubling dimension [1712.05007].

A related combinatorial characterization appears in [1211.0216]. For a finite subset \(C\subset X\), let
\[
\rho_w=\min\{d(w,v):v\in C\setminus\{w\}\}.
\]
A point \(w\in C\) is \((\delta,s)\)-supported if
\[
\inf_{p\in X}\bigl|(B(w,\rho_w/\delta)\setminus B(p,\delta\rho_w))\cap C\bigr|\ge s.
\]
The main theorem states that a metric space carries a Benjamini-Schramm lemma if and only if it is doubling [1211.0216]. Non-doubling spaces admit arbitrarily large finite sets that are well separated yet contained in a bounded region, so many points become highly supported. This gives a finitary characterization of doubling through separation patterns in finite subsets.

## 6. Invariance, robustness, and related geometric properties

The qualitative robustness of doubling under geometric transformations is established in [1603.07521]. If \((X,d)\) is doubling and \(f:(X,d)\to (Y,d')\) is a quasi-Möbius homeomorphism, then \((Y,d')\) is also doubling. The proof uses a factorization theorem: any quasi-Möbius homeomorphism can be written as
\[
f=f_2^{-1}\circ f'\circ f_1,
\]
where \(f'\) is quasi-symmetric and \(f_1,f_2\) are either metric inversions or the identity. Since doubling is already known to be invariant under quasi-symmetric maps, the essential technical step is invariance under metric inversion [1603.07521].

For a point \(p\in X\), metric inversion is defined by
\[
d_p(x,y)=\inf\left\{\sum_{i=1}^k
\frac{d(x_{i-1},x_i)}{d(p,x_{i-1})\,d(p,x_i)}\right\}.
\]
The paper proves that if \((X,d)\) is doubling with constant \(D\), then \((X,d_p)\) is also doubling, with a controlled new doubling constant. An appendix extends the argument from metric spaces to quasi-metric spaces [1603.07521].

The same paper shows that **uniform disconnectedness** is invariant under quasi-Möbius maps, while **Assouad dimension** is not preserved, even though finite Assouad dimension is equivalent to doubling [1603.07521]. This suggests that qualitative covering and separation properties are more stable under quasi-Möbius equivalence than quantitative dimension data. The paper also gives an application: complete, doubling, uniformly perfect, and uniformly disconnected spaces are quasi-Möbius equivalent to a symbolic Cantor set [1603.07521].

## 7. Conceptual synthesis

Across these literatures, the recurring theme is the tension between **local doubling control** and **global geometric separation**. In generalized Fock-space determinantal processes, separated doubling is an exact integral criterion for almost sure separation of random point sets [2502.02237]. In adic analysis, it marks the fact that a measure can be doubling on several prescribed grids while failing to be globally doubling [2310.17615]. In general metric spaces, weak doubling isolates the dependence of doubling estimates on the choice of center and radius, exploiting the non-uniqueness of ball representations [1902.02184]. In doubling metrics, separation manifests through packing bounds and through the scarcity of highly supported points [1712.05007, 1211.0216].

No single formal definition subsumes all of these uses. What is common is the replacement of unrestricted doubling by a version sensitive to separation: separation of points, separation of scales, separation of interval systems, or separation of admissible representations. In that sense, the expression **Separated Doubling Property** identifies a class of phenomena rather than a unique axiom, and its precise meaning must be read from the surrounding framework.

Source: https://www.emergentmind.com/topics/separated-doubling-property