---
title: Mass Transference Principle
url: https://www.emergentmind.com/topics/mass-transference-principle
type: topic
---

# Mass Transference Principle

Searching arXiv for recent and foundational papers on the Mass Transference Principle and its extensions.
The Mass Transference Principle (MTP) is a geometric measure-theoretic mechanism for converting a full measure statement for a limsup set into a Hausdorff measure or Hausdorff dimension statement for a related limsup set obtained by shrinking the underlying approximating sets. In its classical form, due to Beresnevich and Velani, it concerns sequences of balls in Euclidean space; subsequent work has extended the principle to general metric spaces, systems of linear forms, neighborhoods of sets satisfying a local scaling property, rectangles, arbitrary open subsets of balls, and dynamically defined shrinking targets [1704.06628], [2306.15535].

## 1. Classical formulation

For a sequence of sets \((A_i)\), the limsup set is
\[
\limsup_{i\to\infty} A_i
=
\bigcap_{N=1}^{\infty}\bigcup_{i\ge N} A_i,
\]
that is, the set of points belonging to infinitely many \(A_i\). In the original Euclidean theory, one starts with balls \(B_i=B(x_i,r_i)\subset \mathbb R^k\), a dimension function \(f\), and the transformed balls
\[
B_i^f = B\bigl(x_i,f(r_i)^{1/k}\bigr).
\]
The classical Beresnevich–Velani statement is that if, for every ball \(B\subset \mathbb R^k\),
\[
\mathcal H^k\!\left(B\cap \limsup B_i^f\right)=\mathcal H^k(B),
\]
then
\[
\mathcal H^f\!\left(B\cap \limsup B_i\right)=\mathcal H^f(B),
\]
provided \(r_i\to 0\) and \(r^{-k}f(r)\) is monotonic [1704.06628].

This formulation is powerful because \(\mathcal H^k\) is a constant multiple of Lebesgue measure on \(\mathbb R^k\). The principle therefore transfers a full Lebesgue measure statement for a limsup set of enlarged balls into a full Hausdorff \(f\)-measure statement for the limsup set of the original balls. In the frequently used power-law case, if
\[
\operatorname{diam}(E_i)=\bigl(\operatorname{diam}(B_i)\bigr)^a
\qquad (a>1),
\]
and
\[
\mathcal L(\limsup B_i)=1,
\]
then
\[
\dim_H \limsup E_i \ge \frac{d}{a},
\]
which is the form emphasized in the arbitrary-shape extension of Koivusalo and Rams [1812.08557].

The classical principle became central in metric Diophantine approximation because many exceptional sets arise naturally as limsup sets of balls. The surveys stress that it gives a systematic route from Lebesgue-measure zero-one laws, such as Khintchine-type statements, to Hausdorff-measure and Hausdorff-dimension results of Jarník type [1704.06628], [2306.15535].

## 2. Geometric content and scaling mechanisms

A major theme in later work is that the MTP is not tied to Euclidean balls alone. In a locally compact metric space \((X,d)\) carrying a reference Hausdorff \(g\)-measure comparable to \(g(r)\) on balls, the transformed ball is
\[
B^{f,g}=B\bigl(x,g^{-1}(f(r))\bigr),
\]
and full \(\mathcal H^g\)-measure for \(\limsup B_i^{f,g}\) implies full \(\mathcal H^f\)-measure for \(\limsup B_i\) under suitable assumptions [1803.02654]. This metric-space version already indicates that the relevant input is a scaling law for ambient measure rather than Euclidean symmetry.

Allen and Baker abstracted this through the local scaling property (LSP). For sets \(F_j\subset X\), the LSP with parameter \(0\le \kappa<1\) requires
\[
\mathcal H^g\bigl(B(x,r)\cap \Delta(F_j,\delta)\bigr)
\asymp
g(\delta)^{1-\kappa}g(r)^\kappa
\]
for \(x\in F_j\) and \(0<\delta<r\). This yields an MTP for limsup sets of shrinking neighborhoods \(\Delta(F_j,\Upsilon_j)\), covering points, affine planes, smooth compact manifolds, and self-similar sets satisfying the open set condition [1803.02654].

For arbitrary shapes inside balls, Koivusalo and Rams replaced radius rescaling by a generalized singular value function
\[
\varphi^s(E)=\sup_{\mu}\inf_{x\in E}\inf_{r>0}\frac{r^s}{\mu(B_r(x))},
\]
defined for Borel sets \(E\subset \mathbb R^d\). A crucial structural fact is
\[
\varphi^s(E)\le \mathcal H_\infty^s(E)\le 6^s\varphi^s(E),
\]
so \(\varphi^s\) is quantitatively equivalent to Hausdorff content [1812.08557]. In this form, the MTP becomes a statement about shape-sensitive \(s\)-dimensional content rather than about explicit powers of radii.

Persson introduced a different but related size condition based on the \(t\)-dimensional Riesz energy
\[
I_t(U)=\iint_{U\times U}|x-y|^{-t}\,dx\,dy.
\]
For open sets \(U_j\subset B(x_j,r_j)\), the controlling quantity is
\[
\frac{I_t(U_j)\,\lambda(B(x_j,r_j))}{\lambda(U_j)^2},
\]
and the resulting lower bound is expressed through the supremum of those \(t\) for which this ratio remains uniformly bounded [1911.10399]. These formulations collectively show that the transfer mechanism is governed by a geometric size functional adapted to the target geometry: radius powers for balls, singular values or content for anisotropic sets, and energy for general open subsets.

## 3. Major extensions beyond the original ball setting

The modern theory comprises several distinct extensions.

| Extension | Representative result | Main conclusion |
|---|---|---|
| Balls to neighborhoods of planes | Allen–Beresnevich [1703.10015] | Full \(\mathcal H^f\)-measure for \(\Lambda(\Upsilon)\) |
| Neighborhoods of LSP sets | Allen–Baker [1803.02654] | Full \(\mathcal H^f\)-measure in locally compact metric spaces |
| Balls to arbitrary open sets | Koivusalo–Rams [1812.08557] | \(\dim_H \limsup E_i\ge s\), \(\dim_P \limsup E_i=d\) |
| Balls to arbitrary open sets with large intersection | Persson [1911.10399] | \(\limsup U_j\in \mathscr G^s(\mathbb T^d)\) |
| Rectangles to rectangles | Wang–Wu [1909.00924] | Full Hausdorff \(s\)-measure or dimension lower bounds |
| Unbounded rectangle exponents | Wang–Wu extension [2410.18578] | Dimension formulas with some exponents \(+\infty\) |
| Hausdorff-content unification | He [2402.00513] | Full Hausdorff measure and large intersection from content bounds |
| Dynamical balls to open sets | He [2508.03359] | \(\mathscr G^s\)-membership for dynamical shrinking targets |

For systems of linear forms, the resonant sets are \(l\)-dimensional planes \(R_j\subset \mathbb R^k\), and one studies
\[
\Lambda(\Upsilon)=\{x:x\in \Delta(R_j,\Upsilon_j)\ \text{for infinitely many }j\}.
\]
Allen and Beresnevich proved that if
\[
\mathcal H^k\bigl(B\cap \Lambda(g(\Upsilon)^{1/m})\bigr)=\mathcal H^k(B)
\]
for all balls \(B\) in a fixed ball \(\Omega\), then
\[
\mathcal H^f\bigl(B\cap \Lambda(\Upsilon)\bigr)=\mathcal H^f(B),
\]
with \(g(r)=r^{-l}f(r)\) and \(k=m+l\) [1703.10015]. This resolves the systems-of-linear-forms version in full generality.

For arbitrary open subsets \(E_i\subset B_i\subset [0,1]^d\), Koivusalo and Rams proved that if
\[
\varphi^s(E_i)\ge \mathcal L(B_i)
\]
for all \(i\), and \(\mathcal L(\limsup B_i)=1\), then
\[
\dim_H \limsup E_i \ge s
\qquad\text{and}\qquad
\dim_P \limsup E_i = d.
\]
They also obtained the equivalent Hausdorff-content formulation
\[
\mathcal H_\infty^s(E_i)\ge \mathcal L(B_i)
\quad\Longrightarrow\quad
\dim_H \limsup E_i\ge s
\]
[1812.08557].

Persson’s extension is similar in geometry but stronger in conclusion: if
\[
s = \sup\left\{ t>0: \sup_j \frac{I_t(U_j)\lambda(B(x_j,r_j))}{\lambda(U_j)^2} <\infty \right\},
\]
and \(\lambda(\mathbb T^d\setminus \limsup B(x_j,r_j))=0\), then
\[
\dim_H \limsup U_j\ge s
\qquad\text{and}\qquad
\limsup U_j\in \mathscr G^s(\mathbb T^d),
\]
where \(\mathscr G^s\) denotes Falconer’s class of sets with large intersections [1911.10399].

In the anisotropic setting, Wang and Wu established a rectangles-to-rectangles principle in product spaces with Ahlfors regular measures and a rectangle ubiquity property, yielding full Hausdorff measure at a critical exponent \(s(\mathbf t)\) and, under a full-measure hypothesis, dimension lower bounds of the same form [1909.00924]. Later work incorporated the unbounded setup in which some effective exponents are \(+\infty\), giving, for example,
\[
\dim_{\rm H} S(\tau)=\min\left\{1,\frac{3}{1+\tau}\right\}
\]
for
\[
S(\tau)=\left\{(x_1,x_2)\in \mathbb R^2:\|q x_1\|<q^{-\tau},\ \|q x_2\|<e^{-q}\ \text{for infinitely many }q\in\mathbb N\right\}
\]
[2410.18578].

## 4. Proof architectures

The original MTP is classically proved by covering arguments and the mass distribution principle. Later extensions preserve this strategy but adapt the geometry.

For arbitrary open sets, Koivusalo and Rams do not directly adapt the balls-to-balls argument. They first use a selection lemma derived from the full-measure hypothesis \(\mathcal L(\limsup B_i)=1\), then replace each \(E_i\) by a structured support set \(\widetilde E_i\subset E_i\) carrying an absolutely continuous probability measure of bounded density, and finally build a Cantor-type subset
\[
F=\bigcap_{j=0}^\infty F_j \subset \limsup E_i
\]
together with a measure \(\mu\) whose local dimensions can be controlled in two scale regimes. The conclusion \(\dim_H F\ge s\) follows from the mass distribution principle [1812.08557].

Persson’s proof proceeds differently. It uses Vitali’s covering lemma to extract disjoint subfamilies of balls and Riesz energy estimates to build measures \(\mu_n\) supported on unions of the \(U_j\). A criterion for Falconer classes states that if \(\mu_n\) satisfy local density comparability on balls and have uniformly bounded \(s\)-energy, then
\[
\limsup E_n\in \mathscr G^s(\mathbb T^d).
\]
This converts bounded energy directly into large intersection [1911.10399].

Allen and Baker use the LSP to run a Cantor construction inside neighborhoods of the \(F_j\), with a \(K_{G,B}\)-lemma supplying abundant good pieces at each stage and a mass distribution argument producing the final \(\mathcal H^f\)-measure statement [1803.02654].

A different proof architecture appears in He’s unified Hausdorff-content approach. The key input is a Frostman-type measure built from weighted Hausdorff content, together with a new class \(\mathsf g^f(X)\) of \(G_\delta\)-sets satisfying
\[
\mathcal H_\infty^f(A\cap B)\ge c\,\mathcal H_\infty^f(B)
\]
for every ball \(B\). This class is closed under countable intersections, and every \(A\in \mathsf g^f(X)\) satisfies
\[
\mathcal H^f(A)=\mathcal H^f(X).
\]
The resulting theorem transfers Hausdorff-content lower bounds on open sets \(E_n\subset B_n\) directly to \(\limsup E_n\in \mathsf g^f(X)\), thereby unifying MTP and large intersection property in one framework [2402.00513].

For dynamical Diophantine approximation, He’s later work introduces quasi-self-conformal measures and proves a dimensional MTP from balls to open sets: if
\[
\mathcal H_\infty^s(E_n)\gg r_n^{\overline{\dim}_H\nu}
\]
for open sets \(E_n\subset B(x_n,r_n)\), then
\[
\limsup E_n\in \mathscr G^s(X),
\]
with applications to \(\beta\)-transformations and the Gauss map [2508.03359].

## 5. Applications

A standard application of the classical principle is the derivation of Jarník-type Hausdorff measure statements from Khintchine-type Lebesgue measure theorems; the surveys explicitly describe the MTP as the mechanism by which Khintchine’s theorem implies Jarník’s theorem and Dirichlet’s theorem implies the Jarník–Besicovitch theorem [1704.06628].

In higher-dimensional Diophantine approximation, the systems-of-linear-forms MTP turns Lebesgue-measure Khintchine–Groshev theorems into Hausdorff-measure statements in homogeneous, inhomogeneous, and constrained settings. Allen and Beresnevich derive Hausdorff-measure counterparts of Khintchine–Groshev type theorems with primitivity constraints, as well as general inhomogeneous transference results expressed through transformed approximation functions \(\Theta\) [1703.10015].

Persson’s large-intersection theorem has explicit arithmetic consequences. For weighted simultaneous approximation,
\[
W(\tau_1,\ldots,\tau_d)
=
\left\{ (x_1,\ldots,x_d)\in \mathbb T^d :
\max\{ q^{\tau_i}\|q x_i\| \}<1 \text{ for infinitely many } q\in\mathbb N \right\},
\]
the paper obtains
\[
W(\tau)\in \mathscr G^{D(\tau)}(\mathbb T^d),
\]
and therefore for any at most countable family \(T\subset [1,\infty)^d\),
\[
\dim_H \bigcap_{\tau\in T} W(\tau) = \inf_{\tau\in T} D(\tau)
\]
[1911.10399].

The rectangles-to-rectangles theory was developed partly because limsup sets generated by rectangles arise naturally from Minkowski’s theorem, weighted simultaneous approximation, linear forms, shrinking targets on products of Cantor sets, and multiplicative Diophantine approximation. Wang and Wu emphasize that the dimensional theory for rectangle-generated limsup sets underpins multiplicative approximation, where covering-and-slicing methods can fail to identify the exact answer [1909.00924].

In dynamics, He’s balls-to-open-sets theorem yields large intersection results for shrinking target sets associated with the \(\beta\)-transformation and the Gauss map. For
\[
W(T_\beta,f,\{h_n\})=\{x\in[0,1): |T_\beta^n x-h_n(x)|<e^{-S_n f(x)} \text{ for infinitely many }n\},
\]
the limsup set belongs to \(\mathscr G^s([0,1])\), where \(s\) is the unique solution of
\[
P(-s(f+\log\beta),T_\beta)=0.
\]
An analogous statement holds for the Gauss map, and the same framework also gives large intersection results for sets defined by large products of consecutive partial quotients [2508.03359].

## 6. Limitations, misconceptions, and current directions

A basic misconception is that positive measure of the ambient limsup set should suffice for transference. Koivusalo and Rams provide a sharp counterexample: there exist intervals \(B_w\subset[0,1]\) with
\[
\mathcal L(\limsup B_w)>0,
\]
but for every \(a>1\), if \(E_w\) is the concentric interval with \(|E_w|=|B_w|^a\), then
\[
\limsup E_w=\varnothing.
\]
Thus the full-measure hypothesis
\[
\mathcal L(\limsup B_i)=1
\]
is essential in their theorem and cannot simply be weakened to positivity [1812.08557].

Another recurring restriction is openness. The arbitrary-shape theorem of Koivusalo and Rams is stated for open sets \(E_i\subset B_i\), and this is used to build absolutely continuous measures supported on finite unions of disjoint cubes contained in \(E_i\). The paper explicitly identifies the weakening of openness to general Borel or analytic sets as a natural open direction [1812.08557].

There is also an important distinction between dimension and measure conclusions. The original Beresnevich–Velani principle is a Hausdorff-measure theorem, but some later arbitrary-shape extensions are primarily dimensional. Koivusalo and Rams prove Hausdorff dimension and packing dimension statements, not a full Hausdorff-measure analogue for general shapes [1812.08557]. This suggests that general content-sensitive formulations remain more tractable at the level of dimension than at the level of critical gauge measure.

A further conceptual shift is the strengthening from dimension lower bounds to large intersection conclusions. Membership in Falconer’s class \(\mathscr G^s\), or in He’s class \(\mathsf g^f(X)\), is strictly stronger than the assertion \(\dim_H A\ge s\) because these classes are stable under countable intersections [1911.10399], [2402.00513]. Much recent work therefore treats the MTP not only as a tool for computing dimension, but also as a route to robust intersection properties of limsup sets.

Taken together, these developments suggest a broad re-interpretation already explicit in the arbitrary-shape literature: the MTP is not merely a theorem about shrinking radii, but a theory of transference for geometric size quantities—Hausdorff content, singular value functionals, Riesz energies, and local scaling laws—across limsup constructions [1812.08557].

Source: https://www.emergentmind.com/topics/mass-transference-principle