Weighted Mass Transference Principle

• The weighted Mass Transference Principle is an extension of the classical MTP that replaces isotropic balls with coordinatewise scaled rectangles governed by a weight vector.
• It employs anisotropic geometry and local scaling properties to transfer full-measure or ubiquity statements into precise Hausdorff measure and dimension estimates.
• Its applications span weighted Diophantine approximation, multiplicative approximation, and analysis on fractals and manifolds.

The weighted Mass Transference Principle is a family of anisotropic extensions of the Beresnevich–Velani Mass Transference Principle in which isotropic balls are replaced by coordinatewise scaled rectangles, or more generally by products of neighbourhoods whose radii are governed by a weight vector. In this setting, “weighted” does not mean weighting the measure; it means anisotropic scaling in different directions, equivalently non-uniform Diophantine exponents. The central theme is unchanged from the classical theory: a full-measure or ubiquity statement for a limsup set built from comparatively large sets is transferred to Hausdorff measure or Hausdorff dimension information for a limsup set built from smaller sets, but now in a geometry adapted to weights (Allen et al., 2023).

1. Classical antecedent and the transference paradigm

The point of departure is the original Mass Transference Principle in Euclidean space. If $\{B_i=B(x_i,r_i)\}$ is a sequence of balls in $\mathbb R^k$ with $r_i\to 0$, and $f$ is a dimension function, the associated shrunk ball is

$B^f:=B\bigl(x,f(r)^{1/k}\bigr).$

When $f(r)=r^s$, this becomes $B^s(x,r)=B(x,r^{s/k})$. The Beresnevich–Velani principle states that if, for every ball $B\subset\mathbb R^k$,

$$\mathcal H^k\Bigl(B\cap \limsup_{i\to\infty} B_i^f\Bigr)=\mathcal H^k(B),$$

then, again for every ball $B\subset\mathbb R^k$,

$\mathbb R^k$0

Informally, full Lebesgue measure for a limsup set of shrunk balls is converted into full Hausdorff $\mathbb R^k$1-measure for the limsup set of the original balls (Allen et al., 2023).

This mechanism is especially powerful in metric Diophantine approximation. A Khintchine-type theorem often provides a zero-full law for Lebesgue measure, whereas a Jarník-type theorem seeks Hausdorff measure or dimension. The classical MTP bridges these regimes without redoing the covering argument at Hausdorff scale. The same philosophy persists in weighted variants, but the geometry is no longer isotropic.

The classical statement also has a metric-space form. In a locally compact metric space $\mathbb R^k$2 equipped with a doubling reference gauge $\mathbb R^k$3 and Ahlfors-type control

$\mathbb R^k$4

balls are modified by

$\mathbb R^k$5

and the conclusion becomes full $\mathbb R^k$6-measure for the original limsup set whenever the limsup set of the modified balls has full $\mathbb R^k$7-measure (Allen et al., 2023).

2. Meaning of “weighted” and the anisotropic geometry

In the weighted theory, the basic Diophantine inequalities take the form

$\mathbb R^k$8

with a weight vector $\mathbb R^k$9. Geometrically, these inequalities define coverings by rectangles whose side lengths scale like different powers of a common parameter. This is the decisive shift from isotropic ball geometry to anisotropic geometry. The survey literature emphasizes that, in this context, weighted MTP means anisotropic scaling defined by weights in different directions, not weighting the ambient measure (Allen et al., 2023).

For $r_i\to 0$0 and $r_i\to 0$1, the basic weighted rectangle is

$r_i\to 0$2

A limsup set generated by such rectangles records infinitely many weighted approximation events. The weighted MTP asks how a full-measure statement for an associated isotropic or weakly anisotropic limsup set implies Hausdorff information for the weighted limsup set (Allen et al., 2023).

In the general rectangles-to-rectangles framework, the ambient space is a product

$r_i\to 0$3

where each $r_i\to 0$4 is a bounded locally compact metric probability space with $r_i\to 0$5 $r_i\to 0$6-Ahlfors–David regular: $r_i\to 0$7 Resonant sets $r_i\to 0$8 are thickened coordinatewise by a common scale $r_i\to 0$9, giving weighted product neighbourhoods

$f$0

A second vector $f$1 then defines the shrunk rectangles with exponents $f$2. The weighted transference question becomes: what Hausdorff measure or dimension can be guaranteed for the limsup set of the $f$3-rectangles, assuming full measure or ubiquity for the limsup set of the base $f$4-rectangles (Allen et al., 2023)?

A further structural ingredient is the local scaling property. In the survey formulation, each coordinate family satisfies an LSP with exponent $f$5, so that neighbourhoods of resonant sets have the prescribed two-parameter scaling in thickness and ambient radius. This is the main mechanism by which codimension enters the weighted theory (Allen et al., 2023).

3. The first weighted principle: from balls to rectangles

The first weighted MTP identified in the recent literature is the Wang–Wu–Xu passage from balls to rectangles. Starting from a sequence of points $f$6 and radii $f$7, one defines

$f$8

and, for a weight vector $f$9 with $B^f:=B\bigl(x,f(r)^{1/k}\bigr).$0,

$B^f:=B\bigl(x,f(r)^{1/k}\bigr).$1

The hypothesis is a full Lebesgue measure statement for the isotropic limsup set: $B^f:=B\bigl(x,f(r)^{1/k}\bigr).$2 The conclusion is a lower bound for $B^f:=B\bigl(x,f(r)^{1/k}\bigr).$3 and, under the additional condition $B^f:=B\bigl(x,f(r)^{1/k}\bigr).$4, an infinite Hausdorff measure statement at the corresponding critical exponent. In the survey, both conclusions are expressed by explicit minima over the ordered weights $B^f:=B\bigl(x,f(r)^{1/k}\bigr).$5, so the anisotropic exponents directly determine the Hausdorff scale (Allen et al., 2023).

This result is the prototype of a weighted MTP because its input is isotropic and its output is anisotropic. Full Lebesgue measure of a limsup set of balls is transferred to Hausdorff dimension and Hausdorff measure information for a limsup set of weighted rectangles. The construction is tailored to weighted simultaneous approximation, where the rectangles represent coordinatewise inequalities with distinct exponents.

The same development was already highlighted in the earlier survey “The Mass Transference Principle: Ten Years On”, where balls-to-rectangles results were singled out as one of the main deterministic extensions beyond the original balls-to-balls theory (Allen et al., 2017).

4. The general weighted principle: from rectangles to rectangles

The decisive generalization is the Wang–Wu mass transference principle from rectangles to rectangles. In its direct form, the framework consists of product spaces, Ahlfors regular factor measures, resonant sets with a local $B^f:=B\bigl(x,f(r)^{1/k}\bigr).$6-scaling property, and a rectangular ubiquity condition. For each coordinate $B^f:=B\bigl(x,f(r)^{1/k}\bigr).$7, the $B^f:=B\bigl(x,f(r)^{1/k}\bigr).$8-scaling property requires

$B^f:=B\bigl(x,f(r)^{1/k}\bigr).$9

This gives precise control of the measure of neighbourhoods of $f(r)=r^s$0 inside ambient balls (Wang et al., 2019).

Local ubiquity for rectangles is the condition that there exists $f(r)=r^s$1 such that for every ball $f(r)=r^s$2,

$f(r)=r^s$3

Uniform local ubiquity strengthens this to eventual lower bounds for all large $f(r)=r^s$4. The shrunk limsup set is

$f(r)=r^s$5

with $f(r)=r^s$6 and $f(r)=r^s$7 (Wang et al., 2019).

The critical Hausdorff exponent $f(r)=r^s$8 is explicit. It is obtained by minimizing over the finite set

$f(r)=r^s$9

after partitioning the coordinates according to whether the active scale lies above $B^s(x,r)=B(x,r^{s/k})$0, between $B^s(x,r)=B(x,r^{s/k})$1 and $B^s(x,r)=B(x,r^{s/k})$2, or below both. The resulting expression depends on the ordered anisotropic exponents, the factor dimensions $B^s(x,r)=B(x,r^{s/k})$3, and the scaling parameter $B^s(x,r)=B(x,r^{s/k})$4. Under local ubiquity, Wang–Wu prove

$B^s(x,r)=B(x,r^{s/k})$5

and, at the critical value, full Hausdorff measure on every ball: $B^s(x,r)=B(x,r^{s/k})$6 (Wang et al., 2019).

There is also a full-measure formulation, closer in spirit to the original Beresnevich–Velani theorem. If

$B^s(x,r)=B(x,r^{s/k})$7

then the same lower bound

$B^s(x,r)=B(x,r^{s/k})$8

follows for the limsup set built from the shrunk rectangles (Wang et al., 2019). This is the canonical rectangles-to-rectangles weighted MTP.

The survey of recent extensions presents the same theorem in a slightly different notation, with a local scaling exponent $B^s(x,r)=B(x,r^{s/k})$9, partitions $B\subset\mathbb R^k$0, and an explicit minimum over $B\subset\mathbb R^k$1. In that presentation, the crucial fact is the structural dependence of the Hausdorff exponent on $B\subset\mathbb R^k$2, $B\subset\mathbb R^k$3, $B\subset\mathbb R^k$4, and $B\subset\mathbb R^k$5, rather than on a single isotropic radius exponent (Allen et al., 2023).

5. Diophantine applications and weighted dimension formulae

The natural home of the weighted MTP is weighted Diophantine approximation. In simultaneous approximation, one studies

$B\subset\mathbb R^k$6

which is a limsup set of rectangles with side lengths $B\subset\mathbb R^k$7. In the power-law regime $B\subset\mathbb R^k$8, the Wang–Wu machinery yields exact Hausdorff dimension formulae. The paper states, for ordered exponents $B\subset\mathbb R^k$9 with $$\mathcal H^k\Bigl(B\cap \limsup_{i\to\infty} B_i^f\Bigr)=\mathcal H^k(B),$$0,

$$\mathcal H^k\Bigl(B\cap \limsup_{i\to\infty} B_i^f\Bigr)=\mathcal H^k(B),$$1

for an appropriate $$\mathcal H^k\Bigl(B\cap \limsup_{i\to\infty} B_i^f\Bigr)=\mathcal H^k(B),$$2 determined by the exponent configuration (Wang et al., 2019).

The same rectangular formalism applies to systems of linear forms. For

$$\mathcal H^k\Bigl(B\cap \limsup_{i\to\infty} B_i^f\Bigr)=\mathcal H^k(B),$$3

the paper gives the exact dimension

$$\mathcal H^k\Bigl(B\cap \limsup_{i\to\infty} B_i^f\Bigr)=\mathcal H^k(B),$$4

Here the weights are the exponents $$\mathcal H^k\Bigl(B\cap \limsup_{i\to\infty} B_i^f\Bigr)=\mathcal H^k(B),$$5 attached to the different rows of the linear-form system (Wang et al., 2019).

A further application is multiplicative approximation. Sets defined by

$$\mathcal H^k\Bigl(B\cap \limsup_{i\to\infty} B_i^f\Bigr)=\mathcal H^k(B),$$6

decompose into unions of anisotropic approximation sets with coordinate exponents summing to the multiplicative parameter. The survey stresses that the dimensional theory for limsup sets generated by rectangles underpins the dimensional theory in multiplicative Diophantine approximation, where the usually used methods or even their generalizations fail to work (Wang et al., 2019).

The weighted framework also extends beyond Euclidean Lebesgue space. Wang–Wu apply it to shrinking target problems on products of Cantor sets and to multiplicative approximation on Cantor sets, obtaining exact dimension formulae in terms of the associated Ahlfors exponents and anisotropic shrinking rates (Wang et al., 2019). The survey notes more broadly that weighted MTPs have already been applied to weighted simultaneous approximation on manifolds, in $$\mathcal H^k\Bigl(B\cap \limsup_{i\to\infty} B_i^f\Bigr)=\mathcal H^k(B),$$7-adic settings, and in dynamical Diophantine approximation (Allen et al., 2023).

6. Comparison with the unweighted theory and later generalizations

Relative to the unweighted MTP, the weighted theory replaces isotropic balls by axis-parallel rectangles or product neighbourhoods, so rotational symmetry is lost and singular value–type behavior becomes relevant. The hypotheses are correspondingly stronger. In addition to radii tending to zero, one typically requires product structure of the ambient space, Ahlfors regularity of the factor measures, a local scaling property for the resonant sets, and either a ubiquity condition or a full-measure limsup statement. The dimension bound is no longer a function of a single shrink exponent; it is governed by minima over ordered weights and shrink parameters, reflecting the fact that different coordinates can dominate at different scales (Allen et al., 2023).

The proof architecture nonetheless remains recognizably Beresnevich–Velani. At a high level, the weighted proofs use carefully selected disjoint basic blocks at each scale, a Cantor-type construction, measure estimates on rectangular coverings, and a mass distribution principle. What changes are the geometric lemmas: one needs covering and packing results for anisotropic rectangles, precise scaling estimates for neighbourhoods of resonant sets, and slicing or Fubini-type arguments to recombine coordinatewise information (Allen et al., 2023).

Beyond rectangles, several general-shape and large-intersection extensions place weighted MTPs inside a broader landscape. Allen and Baker proved a general MTP for limsup sets defined via neighbourhoods of sets satisfying a local scaling property in locally compact metric spaces; the modified radii

$$\mathcal H^k\Bigl(B\cap \limsup_{i\to\infty} B_i^f\Bigr)=\mathcal H^k(B),$$8

already encode a nontrivial weighted interaction between the ambient gauge $$\mathcal H^k\Bigl(B\cap \limsup_{i\to\infty} B_i^f\Bigr)=\mathcal H^k(B),$$9, the target gauge $B\subset\mathbb R^k$0, and the geometric parameter $B\subset\mathbb R^k$1 (Allen et al., 2018). Koivusalo and Rams replaced the small balls in the classical theorem by arbitrary open sets $B\subset\mathbb R^k$2, controlled by a generalized singular value function $B\subset\mathbb R^k$3, and thereby obtained a balls-to-arbitrary-shapes form of the MTP that is naturally compatible with anisotropic geometry (Koivusalo et al., 2018).

Persson developed an energy-based MTP for arbitrary open sets $B\subset\mathbb R^k$4 on the torus, with the critical exponent

$B\subset\mathbb R^k$5

and proved not only $B\subset\mathbb R^k$6 for $B\subset\mathbb R^k$7, but also

$B\subset\mathbb R^k$8

placing weighted anisotropic limsup sets inside Falconer’s large-intersection classes. In corollaries on ellipsoids and rectangles, the relevant control is expressed through singular value function estimates, which makes the connection to weighted MTPs explicit (Persson, 2019).

He’s content-based unification goes further by treating balls-to-open-sets, Allen–Baker’s local-scaling MTP, Wang–Wu’s rectangle-to-rectangle theory, and dynamical variants within a single Hausdorff-content framework. In that setting, if open sets $B\subset\mathbb R^k$9 satisfy a uniform content lower bound

$\mathbb R^k$00

and the limsup of the $\mathbb R^k$01 has full $\mathbb R^k$02-measure, then $\mathbb R^k$03 belongs to $\mathbb R^k$04, hence has both full Hausdorff $\mathbb R^k$05-measure and large intersection property. This yields a simpler proof scheme that avoids explicit Cantor-set constructions and also covers rectangle-based weighted settings (He, 2024).

Current open directions are described in the survey literature with some precision: more general shapes beyond axis-parallel rectangles, weighted variants on manifolds and fractals, anisotropic scaling combined with inhomogeneous measures, and extensions under weaker regularity assumptions than Ahlfors regularity and the LSP. A further active line concerns inhomogeneous and dynamical weighted approximation, where many applications remain at the Lebesgue-measure level and a full Hausdorff-scale weighted transference theory is still being developed (Allen et al., 2023).