Weak Doubling Property in Analysis

• Weak Doubling Property is a relaxation of the classical doubling condition, ensuring a reliable lower bound for measure growth.
• It is applied in metric geometry, elliptic homogenization, and boundary measure analysis to enable flexible control where strong doubling fails.
• The concept underpins techniques like quasisymmetric uniformization of spheres and refined bounds in periodic elliptic PDEs.

Searching arXiv for recent and foundational papers on weak doubling property and related notions. arxiv.search({"query":"all:\"weak doubling property\" OR all:\"weak metric doubling measure\" OR id:(Lohvansuu et al., 2017) OR id:(Rajala et al., 2020) OR id:(Kenig et al., 2021) OR id:(Korobenko et al., 2013) OR id:(Pathak, 17 Nov 2025)","max_results":10,"sort_by":"submittedDate","sort_order":"descending"}) Searching arXiv by paper IDs and topic keywords. {"query":"(Lohvansuu et al., 2017, Rajala et al., 2020, Kenig et al., 2021, Korobenko et al., 2013, Pathak, 17 Nov 2025) weak metric doubling weak doubling property","max_results":10,"sort_by":"relevance"} The weak doubling property refers, in the literature considered here, to several related but non-equivalent relaxations of standard doubling. In its classical measure-theoretic form, doubling means

$$\mu\!\bigl(B(y,2R)\bigr)\le C_D\,\mu\!\bigl(B(y,R)\bigr)$$

for all centers $y$ and radii $R>0$. By contrast, the weak variants that appear in geometric analysis and elliptic PDE are one-sided or partial analogues: in metric geometry they are formulated through lower bounds for chain-length functionals associated with a doubling measure; in periodic elliptic homogenization they take the form of scale-uniform $L^2$ doubling inequalities for weak solutions; and in elliptic measure they may require $\omega(2\Delta)\le C\,\omega(\Delta)$ only for most surface balls, or up to a set of small measure (Lohvansuu et al., 2017, Kenig et al., 2021, Pathak, 17 Nov 2025, Korobenko et al., 2013).

1. Standard doubling and weak variants

A basic reference point is the usual doubling condition for a measure $\mu$: $$\forall\,y\in X,\;\forall\,R>0:\qquad \mu\!\bigl(B(y,2R)\bigr)\;\le\;C_D\,\mu\!\bigl(B(y,R)\bigr). \tag{D}$$ Korobenko–Maldonado–Rios show that, in Euclidean, metric, Dirichlet-form, and subelliptic settings, a weak $(p\sigma,p)$-Sobolev inequality implies exactly this standard doubling condition, with $C_D$ depending only on $p,\sigma$, and the Sobolev constant $y$0 (Korobenko et al., 2013). Their account explicitly states that no genuinely weaker or “quasi-doubling” variants appear there.

In the metric-doubling framework of David–Semmes, a doubling Borel measure $y$1 is metric-doubling of dimension $y$2 if there exists another metric $y$3 on $y$4 and a constant $y$5 such that

$y$6

This is a two-sided deformation estimate: the measure deforms the original metric without too much shrinking or too much stretching (Lohvansuu et al., 2017).

The weak metric-doubling condition removes the a priori upper bound. Lohvansuu–Rajala–Rasimus define a weak metric-doubling measure of dimension $y$7 by requiring only the lower control

$y$8

where $y$9 is built from $R>0$0-chains and $R>0$1-weighted chain lengths. In this formulation, “weak” means that the lower bound survives in the $R>0$2 limit, but no upper bound is assumed in advance (Lohvansuu et al., 2017).

This suggests that, across the cited literature, “weak doubling” is best understood as a family of relaxations of doubling, rather than a single canonical definition.

2. Weak metric-doubling measures on metric spaces

Let $R>0$3 be a metric space, $R>0$4 a doubling Radon or Borel measure, and $R>0$5. For $R>0$6 and $R>0$7, set

$R>0$8

A finite sequence $R>0$9 is a $L^2$0-chain from $L^2$1 to $L^2$2 if

$L^2$3

Its $L^2$4-length is

$L^2$5

One then defines

$L^2$6

and

$L^2$7

A doubling measure $L^2$8 is a weak metric-doubling measure of dimension $L^2$9 if there exists $\omega(2\Delta)\le C\,\omega(\Delta)$0 such that

$\omega(2\Delta)\le C\,\omega(\Delta)$1

for all $\omega(2\Delta)\le C\,\omega(\Delta)$2 (Lohvansuu et al., 2017, Rajala et al., 2020).

The central distinction from strong metric doubling is that only the lower bound is postulated. To recover a genuine metric-doubling structure, one must prove a posteriori that $\omega(2\Delta)\le C\,\omega(\Delta)$3 also satisfies an upper control

$\omega(2\Delta)\le C\,\omega(\Delta)$4

and that $\omega(2\Delta)\le C\,\omega(\Delta)$5 is in fact a metric (Lohvansuu et al., 2017).

Rajala–Rasimus use the same construction in the finitely connected planar setting. In their notation, for $\omega(2\Delta)\le C\,\omega(\Delta)$6 one writes $\omega(2\Delta)\le C\,\omega(\Delta)$7, and a $\omega(2\Delta)\le C\,\omega(\Delta)$8-WMDM is defined by

$\omega(2\Delta)\le C\,\omega(\Delta)$9

They also prove a reverse local bound: for each $\mu$0 there is $\mu$1 such that

$\mu$2

where

$\mu$3

This local upper estimate is obtained by an annulus-separation argument using the one-sided estimate and the doubling hypothesis (Rajala et al., 2020).

3. Quasispheres, circle domains, and the geometric role of weak doubling

For metric two-spheres, Lohvansuu–Rajala–Rasimus prove a sharp quasisphere criterion. A metric two-sphere $\mu$4 is a quasisphere if and only if it is linearly locally connected and carries a weak metric-doubling measure of dimension $\mu$5 (Lohvansuu et al., 2017). Here linear local connectivity means that there is $\mu$6 such that, for every $\mu$7 and $\mu$8, any two points in $\mu$9 can be joined by a continuum in $$\forall\,y\in X,\;\forall\,R>0:\qquad \mu\!\bigl(B(y,2R)\bigr)\;\le\;C_D\,\mu\!\bigl(B(y,R)\bigr). \tag{D}$$0, and any two points in $$\forall\,y\in X,\;\forall\,R>0:\qquad \mu\!\bigl(B(y,2R)\bigr)\;\le\;C_D\,\mu\!\bigl(B(y,R)\bigr). \tag{D}$$1 can be joined by a continuum in $$\forall\,y\in X,\;\forall\,R>0:\qquad \mu\!\bigl(B(y,2R)\bigr)\;\le\;C_D\,\mu\!\bigl(B(y,R)\bigr). \tag{D}$$2.

The proof of the nontrivial implication starts from

$$\forall\,y\in X,\;\forall\,R>0:\qquad \mu\!\bigl(B(y,2R)\bigr)\;\le\;C_D\,\mu\!\bigl(B(y,R)\bigr). \tag{D}$$3

which automatically satisfies the lower bound

$$\forall\,y\in X,\;\forall\,R>0:\qquad \mu\!\bigl(B(y,2R)\bigr)\;\le\;C_D\,\mu\!\bigl(B(y,R)\bigr). \tag{D}$$4

Proposition 3.1 then shows that $$\forall\,y\in X,\;\forall\,R>0:\qquad \mu\!\bigl(B(y,2R)\bigr)\;\le\;C_D\,\mu\!\bigl(B(y,R)\bigr). \tag{D}$$5 is actually a metric and that there exists $$\forall\,y\in X,\;\forall\,R>0:\qquad \mu\!\bigl(B(y,2R)\bigr)\;\le\;C_D\,\mu\!\bigl(B(y,R)\bigr). \tag{D}$$6 such that

$$\forall\,y\in X,\;\forall\,R>0:\qquad \mu\!\bigl(B(y,2R)\bigr)\;\le\;C_D\,\mu\!\bigl(B(y,R)\bigr). \tag{D}$$7

Lemma 3.2 and Proposition 3.1 imply that the identity map

$$\forall\,y\in X,\;\forall\,R>0:\qquad \mu\!\bigl(B(y,2R)\bigr)\;\le\;C_D\,\mu\!\bigl(B(y,R)\bigr). \tag{D}$$8

is quasisymmetric and that $$\forall\,y\in X,\;\forall\,R>0:\qquad \mu\!\bigl(B(y,2R)\bigr)\;\le\;C_D\,\mu\!\bigl(B(y,R)\bigr). \tag{D}$$9 is Ahlfors $(p\sigma,p)$0-regular. Bonk–Kleiner’s theorem then yields a quasisymmetric parametrization of $(p\sigma,p)$1 by the round sphere, and composition gives the desired parametrization of $(p\sigma,p)$2 (Lohvansuu et al., 2017).

The key technical step is the missing upper bound. According to the proof outline, it is established by constructing short $(p\sigma,p)$3-chains in annuli and patching them across a cover of $(p\sigma,p)$4. The ingredients singled out in the paper are Lemma 4.3 on short chains in annuli, Proposition 5.2 on patching continua, and Lemma 5.3 giving the upper bound for $(p\sigma,p)$5 (Lohvansuu et al., 2017).

In the finitely connected planar setting, Rajala–Rasimus generalize the uniformization of Ahlfors $(p\sigma,p)$6-regular spaces. If $(p\sigma,p)$7 is homeomorphic to a finitely connected planar domain, linearly locally connected with constant $(p\sigma,p)$8, carries a $(p\sigma,p)$9-WMDM $C_D$0 with doubling constant $C_D$1, and $C_D$2 is compact with $C_D$3 boundary components separated by ratio $C_D$4, then there is an $C_D$5-quasisymmetric homeomorphism

$C_D$6

onto a circle domain, where $C_D$7 depends only on $C_D$8, $C_D$9, $p,\sigma$0, $p,\sigma$1, and $p,\sigma$2 (Rajala et al., 2020).

Their Section 3 also contains growth estimates for the deformed metric $p,\sigma$3: $p,\sigma$4 and, provided $p,\sigma$5,

$p,\sigma$6

From these inequalities they deduce that $p,\sigma$7 is comparable to the two-dimensional Hausdorff measure $p,\sigma$8 of $p,\sigma$9, so $y$00 becomes Ahlfors $y$01-regular (Rajala et al., 2020).

4. Dimension sharpness, examples, and failure of strong doubling

The geometric theory is dimension-sensitive. Lohvansuu–Rajala–Rasimus record that if $y$02, then no linearly locally connected two-sphere can carry a weak metric-doubling measure of that smaller dimension. If $y$03, there are linearly locally connected spheres with weak metric-doubling measures of dimension $y$04 that admit no quasisymmetric parametrization; Rickman rugs on $y$05 are given as an example (Lohvansuu et al., 2017).

The same source emphasizes that weak and strong metric doubling are genuinely different. In fractal constructions such as Rickman rugs and certain warped products, the lower bound

$y$06

holds by construction, but the corresponding upper bound fails unless an additional regularity hypothesis is imposed (Lohvansuu et al., 2017).

Rajala–Rasimus likewise point to examples showing that the weak condition is strictly broader than the classical two-sided theory. They refer to earlier work for constructions of fractal surfaces whose natural Hausdorff measure fails strong metric-doubling but still satisfies the one-sided condition

$y$07

They also note that in the classical Ahlfors-regular setting the surface measure is not only WMDM but in fact two-sided metric doubling, so the weak theory properly generalizes earlier quasisymmetric uniformization results (Rajala et al., 2020).

A common misconception is that the adjective “weak” merely reflects a technical reformulation of standard doubling. In these examples it does not: the one-sided condition can persist in settings where strong metric-doubling fails.

5. Weak doubling in periodic elliptic homogenization

In periodic elliptic homogenization, the weak doubling property has a different meaning. Consider

$y$08

where $y$09 is $y$10-periodic, symmetric, uniformly elliptic, and Lipschitz. If $y$11 solves $y$12 in $y$13, the doubling index of a ball $y$14 is

$y$15

The paper defines a doubling inequality, or weak doubling property, by the estimate

$y$16

where $y$17 is independent of $y$18 and $y$19 is an a priori bound on the global doubling index (Kenig et al., 2021).

Under the normalized doubling condition

$y$20

for some fixed $y$21, the main theorem gives explicit all-scale bounds. For $y$22, for any $y$23 there exists $y$24 such that for every $y$25,

$y$26

For $y$27, there is $y$28 such that

$y$29

Thus the constant is $y$30-independent and grows as a double-exponential in $y$31 for $y$32, or sub-exponential in $y$33 for $y$34 (Kenig et al., 2021).

The proof splits into three regimes. At small scale, $y$35, one approximates $y$36 by its homogenized harmonic limit and obtains

$y$37

At intermediate scale, $y$38, one uses the Armstrong–Kuusi–Smart three-ball inequality with exponential tail and iteration to reach

$y$39

At large scale, $y$40, one rescales to a problem with Lipschitz constant $y$41, applies Almgren-type frequency monotonicity, and concludes

$y$42

The stated ingredients are convergence rates in homogenization, a three-ball inequality from large-scale analyticity, and a refined Almgren frequency argument (Kenig et al., 2021).

6. Elliptic measure, Carleson drifts, and the boundary of the concept

For elliptic measure, the cited 2025 work studies

$y$43

on a $y$44-sided chord-arc domain $y$45, $y$46, with Ahlfors–David regular boundary, corkscrew condition, Harnack chain condition, bounded uniformly elliptic principal part, and drift satisfying both $y$47 and the Carleson-measure smallness condition

$y$48

If $y$49, then for every surface ball $y$50 and every $y$51,

$y$52

with $y$53 depending only on the structural parameters. Equivalently,

$y$54

and doubling follows by replacing $y$55 (Pathak, 17 Nov 2025).

The argument proceeds through three stages: a local Hardy inequality and Carleson-box Caccioppoli estimate; boundary Hölder regularity and a Bourgain estimate; and two-sided pointwise Green-function bounds. The paper states that one also obtains a weak-doubling property, which only demands $y$56 for most surface balls, or up to a set of small measure, under a slightly looser Carleson condition on $y$57 (Pathak, 17 Nov 2025).

Placed beside the Sobolev-to-doubling theorem of Korobenko–Maldonado–Rios, this yields a useful distinction. In the Sobolev paper, “weak” modifies the Sobolev hypothesis, while the conclusion is exact standard doubling: $y$58 for all balls (Korobenko et al., 2013). In the geometric and elliptic-measure papers, by contrast, “weak doubling” refers to a genuinely relaxed doubling notion. A plausible implication is that the term acquires its precise meaning only after the ambient structure—metric chains, solution norms, or boundary measure—has been fixed.