Twisted Diophantine Approximation: Anisotropic Methods

• Twisted Diophantine approximation is a family of anisotropic methods that modifies classical number theory through fixed weights, directional constraints, and prescribed ambient data.
• It unifies various approaches—weighted exponents, inhomogeneous targets, and restricted denominators—to interpolate between simultaneous, dual, and multiplicative approximations.
• The theory highlights spectrum flexibility and the breakdown of a universal Jarník relation, leading to new insights in measure, dimension, and transference principles.

Searching arXiv for recent and foundational papers on twisted Diophantine approximation. Twisted Diophantine approximation is a family of non-isotropic Diophantine problems in which the approximation process is modified by fixed weights, fixed directions, fixed orbits, restricted denominator sets, or manifold constraints. In one major usage, it denotes weighted exponents in dimension $2$ interpolating between classical simultaneous/dual approximation and multiplicative approximation; in another, it denotes the inhomogeneous problem where a base point or matrix is fixed and the target varies; in more geometric variants, the admissible approximants are confined to cones, subspaces, subsequences, or nondegenerate manifolds. The subject is therefore unified less by a single definition than by a common structural feature: approximation is performed relative to a prescribed anisotropy or fixed ambient datum (Marnat, 2014, Harrap, 2010, Beresnevich et al., 6 Jul 2025).

1. Terminological scope and principal models

In current literature, the term “twisted Diophantine approximation” is used for several closely related paradigms. The weighted two-dimensional theory fixes $\boldsymbol\theta=(\theta_1,\theta_2)$ and distributes the height budget asymmetrically between the two coordinates through weights $i,j\ge 0$ with $i+j=1$. The twisted inhomogeneous theory fixes a vector $\alpha$ or a matrix $A$ and varies the target $\beta$, studying approximation by the orbit $\{q\alpha\}$ or $\{Aq\}$. Geometric versions impose projective or angular restrictions, replacing weighted boxes by cones around a subspace or direction. More recent work also treats restricted denominators, matrix-generated toral orbits, and manifold-constrained targets (Marnat, 2014, Süess, 2017, Champagne et al., 2022, Hauke et al., 2 Aug 2025, Chow et al., 18 Nov 2025).

Paradigm	Fixed datum	Typical object
Weighted exponents	$\boldsymbol\theta$, $\boldsymbol\theta=(\theta_1,\theta_2)$0	$\boldsymbol\theta=(\theta_1,\theta_2)$1, $\boldsymbol\theta=(\theta_1,\theta_2)$2
Twisted inhomogeneous	$\boldsymbol\theta=(\theta_1,\theta_2)$3 or $\boldsymbol\theta=(\theta_1,\theta_2)$4	$\boldsymbol\theta=(\theta_1,\theta_2)$5, $\boldsymbol\theta=(\theta_1,\theta_2)$6
Angular or cone-restricted	subspace $\boldsymbol\theta=(\theta_1,\theta_2)$7 or direction $\boldsymbol\theta=(\theta_1,\theta_2)$8	$\boldsymbol\theta=(\theta_1,\theta_2)$9, $i,j\ge 0$0 constraints
Restricted times or denominators	$i,j\ge 0$1 or $i,j\ge 0$2	$i,j\ge 0$3, $i,j\ge 0$4
Manifold-constrained targets	fixed $i,j\ge 0$5, $i,j\ge 0$6	$i,j\ge 0$7

A useful conceptual distinction runs between two broad branches. One branch studies exponents attached to a fixed $i,j\ge 0$8 and modified height conditions, as in the weighted interpolation between classical and multiplicative approximation. The other studies limsup sets of targets approximated by a fixed orbit, as in $i,j\ge 0$9, $i+j=1$0, or $i+j=1$1. The literature connects these branches through common tools—geometry of numbers, minimal points, ubiquity, mass transference, discrepancy, and shrinking-target methods—but their basic invariants are different.

2. Weighted exponents and the failure of a twisted Jarník relation

The foundational weighted two-dimensional theory is developed in “There is no analogue to Jarník’s relation for twisted Diophantine approximation” (Marnat, 2014). For $i+j=1$2 with $i+j=1$3 linearly independent over $i+j=1$4, and weights $i+j=1$5 with $i+j=1$6, the twisted exponents are defined by the systems

$i+j=1$7

and

$i+j=1$8

Their supremal exponents are denoted $i+j=1$9 and $\alpha$0. When $\alpha$1, they reduce to the classical exponents $\alpha$2.

The classical two-dimensional theory contains Jarník’s identity

$\alpha$3

The weighted theory shows that this phenomenon is exceptional. The paper constructs points

$\alpha$4

from $\alpha$5-sequences $\alpha$6, $\alpha$7 and proves explicit formulas

$\alpha$8

$\alpha$9

$A$0

under explicit inequalities on $A$1 and $A$2. The same paper proves that for fixed $A$3, fixed $A$4, and every

$A$5

there exist uncountably many $A$6 with

$A$7

Accordingly, $A$8 is not determined by $A$9, and no analogue of Jarník’s relation survives in the twisted weighted setting (Marnat, 2014).

This negative result is structurally important. It shows that the weighted interpolation between classical and multiplicative approximation does not preserve the rigid functional dependence of the classical two-dimensional spectrum. A common misconception is that a one-parameter deformation of the classical problem should inherit a one-equation transference law. The weighted formulas show the opposite: two free parameters $\beta$0 remain visible in the spectrum, and the uniform exponents can vary independently over nontrivial regions.

3. Twisted inhomogeneous approximation and weighted Kurzweil theory

A second major meaning of twisted Diophantine approximation fixes a base point $\beta$1 and varies the inhomogeneous target $\beta$2. In the weighted setting, for a weight vector $\beta$3 with $\beta$4, one studies

$\beta$5

This is the framework of weighted twisted inhomogeneous approximation developed by Harrap and extended in later work (Süess, 2017).

In dimension $\beta$6, “Twisted inhomogeneous Diophantine approximation and badly approximable sets” proves a weighted Kurzweil-type characterization. With

$\beta$7

the paper shows

$\beta$8

where $\beta$9 consists of irrational $\{q\alpha\}$0 such that the twisted limsup set $\{q\alpha\}$1 has full Lebesgue measure. It also proves that if $\{q\alpha\}$2, then the twisted badly approximable target set $\{q\alpha\}$3 has Hausdorff dimension $\{q\alpha\}$4 (Harrap, 2010).

The base-point restriction was subsequently removed. “Badly approximable points in twisted Diophantine approximation and Hausdorff dimension” proves that for every $\{q\alpha\}$5 and every weight vector $\{q\alpha\}$6 with $\{q\alpha\}$7,

$\{q\alpha\}$8

and in fact

$\{q\alpha\}$9

The construction uses weighted best approximations $\{Aq\}$0, the lacunarity estimate

$\{Aq\}$1

and a Cantor-set mass-distribution argument (Bengoechea et al., 2015).

The weighted Dirichlet-scale almost-everywhere theory was sharpened in “Simultaneous Diophantine approximation on affine subspaces and Dirichlet improvability”. If $\{Aq\}$2, where $\{Aq\}$3 denotes the weighted singular vectors, then for almost all $\{Aq\}$4,

$\{Aq\}$5

Equivalently, for every $\{Aq\}$6, the set $\{Aq\}$7, with $\{Aq\}$8, has full Lebesgue measure (Süess, 2017).

The matrix-orbit version appears in “Weighted twisted inhomogeneous Diophantine approximation”. For fixed $\{Aq\}$9, weights $\boldsymbol\theta$0, and coordinatewise approximation data $\boldsymbol\theta$1, the central set is

$\boldsymbol\theta$2

For $\boldsymbol\theta$3, the paper proves a zero-full law governed by

$\boldsymbol\theta$4

and obtains Hausdorff-dimension formulae in substantial weighted regimes via weighted ubiquity and weighted mass transference (Hussain et al., 2023).

4. Directional, angular, and cone-restricted twists

A different branch of the subject replaces coordinate weights by directional restrictions. “Diophantine approximation in angular domains” studies the classical linear-form problem

$\boldsymbol\theta$5

under the restriction that $\boldsymbol\theta$6 lie in an angular sector, notably the positive cone $\boldsymbol\theta$7. Schmidt had shown that the golden ratio

$\boldsymbol\theta$8

is admissible in the positive-quadrant problem. Roy proved that $\boldsymbol\theta$9 is in fact optimal: for any unbounded increasing $\boldsymbol\theta=(\theta_1,\theta_2)$00, there exist $\boldsymbol\theta=(\theta_1,\theta_2)$01 such that for all sufficiently large $\boldsymbol\theta=(\theta_1,\theta_2)$02,

$\boldsymbol\theta=(\theta_1,\theta_2)$03

while there are still infinitely many cone-constrained approximants of order $\boldsymbol\theta=(\theta_1,\theta_2)$04. In this sense, the directional restriction lowers the effective exponent from the classical value $\boldsymbol\theta=(\theta_1,\theta_2)$05 to $\boldsymbol\theta=(\theta_1,\theta_2)$06 (Roy, 2016).

“Diophantine approximation with constraints” studies a higher-dimensional angular version. If $\boldsymbol\theta=(\theta_1,\theta_2)$07 is a subspace of dimension $\boldsymbol\theta=(\theta_1,\theta_2)$08, the problem is to approximate a linear form $\boldsymbol\theta=(\theta_1,\theta_2)$09 using integer vectors $\boldsymbol\theta=(\theta_1,\theta_2)$10 satisfying

$\boldsymbol\theta=(\theta_1,\theta_2)$11

The optimal constrained exponent is

$\boldsymbol\theta=(\theta_1,\theta_2)$12

and the paper proves that for every $\boldsymbol\theta=(\theta_1,\theta_2)$13 with $\boldsymbol\theta=(\theta_1,\theta_2)$14-linearly independent coordinates, and every $\boldsymbol\theta=(\theta_1,\theta_2)$15, there exists $\boldsymbol\theta=(\theta_1,\theta_2)$16 with

$\boldsymbol\theta=(\theta_1,\theta_2)$17

while conversely $\boldsymbol\theta=(\theta_1,\theta_2)$18 is sharp. A striking feature is that $\boldsymbol\theta=(\theta_1,\theta_2)$19 depends only on $\boldsymbol\theta=(\theta_1,\theta_2)$20, not on the ambient dimension $\boldsymbol\theta=(\theta_1,\theta_2)$21 (Champagne et al., 2022).

These geometric variants are not weighted in the $\boldsymbol\theta=(\theta_1,\theta_2)$22-box sense, but they are twisted in a precise projective sense: admissible approximants are restricted to a cone around a prescribed direction set. This suggests a broader taxonomy in which “twist” may be implemented either by anisotropic scaling or by anisotropic directionality.

5. Restricted denominators, arithmetic subsequences, and toral dynamics

Twisted approximation with prescribed time sets or denominator sets has become a distinct subfield. “Twisted approximation with restricted denominators” fixes an increasing integer sequence $\boldsymbol\theta=(\theta_1,\theta_2)$23, a real $\boldsymbol\theta=(\theta_1,\theta_2)$24, and studies

$\boldsymbol\theta=(\theta_1,\theta_2)$25

For Lebesgue-almost every $\boldsymbol\theta=(\theta_1,\theta_2)$26, the paper proves the divergence criterion

$\boldsymbol\theta=(\theta_1,\theta_2)$27

with no growth condition on $\boldsymbol\theta=(\theta_1,\theta_2)$28 and no monotonicity assumption on $\boldsymbol\theta=(\theta_1,\theta_2)$29. It also proves a Jarník-style Hausdorff-measure law: for a dimension function $\boldsymbol\theta=(\theta_1,\theta_2)$30 with $\boldsymbol\theta=(\theta_1,\theta_2)$31 monotonic,

$\boldsymbol\theta=(\theta_1,\theta_2)$32

for almost every $\boldsymbol\theta=(\theta_1,\theta_2)$33, and deduces

$\boldsymbol\theta=(\theta_1,\theta_2)$34

for $\boldsymbol\theta=(\theta_1,\theta_2)$35 (Hauke et al., 2 Aug 2025).

A toral-dynamical matrix analogue is developed in “Twisted Diophantine approximation for matrix transformations of tori”. For a sequence of integral matrices $\boldsymbol\theta=(\theta_1,\theta_2)$36 and side-length functions $\boldsymbol\theta=(\theta_1,\theta_2)$37, the target set is

$\boldsymbol\theta=(\theta_1,\theta_2)$38

where $\boldsymbol\theta=(\theta_1,\theta_2)$39 is the axis-aligned box with side lengths $\boldsymbol\theta=(\theta_1,\theta_2)$40. If $\boldsymbol\theta=(\theta_1,\theta_2)$41 for $\boldsymbol\theta=(\theta_1,\theta_2)$42, then for Lebesgue-almost every $\boldsymbol\theta=(\theta_1,\theta_2)$43,

$\boldsymbol\theta=(\theta_1,\theta_2)$44

The same paper proves a Jarník-type Hausdorff measure theorem with critical quantity

$\boldsymbol\theta=(\theta_1,\theta_2)$45

and, for $\boldsymbol\theta=(\theta_1,\theta_2)$46, derives the exact dimension formula

$\boldsymbol\theta=(\theta_1,\theta_2)$47

for almost every $\boldsymbol\theta=(\theta_1,\theta_2)$48 (Chow et al., 18 Nov 2025).

Arithmetic subsequences produce further twisted phenomena. “The Prime times of twisted Diophantine approximation” studies

$\boldsymbol\theta=(\theta_1,\theta_2)$49

for multiplicatively structured sets $\boldsymbol\theta=(\theta_1,\theta_2)$50. Under sieve-type hypotheses on $\boldsymbol\theta=(\theta_1,\theta_2)$51, the paper proves

$\boldsymbol\theta=(\theta_1,\theta_2)$52

with applications to primes, sums of two squares, Löschian integers, and square-free numbers. In particular, for badly approximable $\boldsymbol\theta=(\theta_1,\theta_2)$53, the Kurzweil-type zero-one law holds along the primes, and the square-free case yields a characterization of $\boldsymbol\theta=(\theta_1,\theta_2)$54 via restricted denominators (Hauke, 26 Mar 2026).

6. Manifold constraints, conceptual formalisms, and methodological structure

The manifold-constrained theory treats twisted approximation with $\boldsymbol\theta=(\theta_1,\theta_2)$55 restricted to a nondegenerate analytic manifold $\boldsymbol\theta=(\theta_1,\theta_2)$56. For fixed $\boldsymbol\theta=(\theta_1,\theta_2)$57, “Twisted Diophantine approximation on manifolds” studies

$\boldsymbol\theta=(\theta_1,\theta_2)$58

If

$\boldsymbol\theta=(\theta_1,\theta_2)$59

then for any doubling $\boldsymbol\theta=(\theta_1,\theta_2)$60 with

$\boldsymbol\theta=(\theta_1,\theta_2)$61

the set $\boldsymbol\theta=(\theta_1,\theta_2)$62 has zero measure on any nondegenerate analytic manifold. Under a divergence assumption and a strong non-approximability hypothesis expressed through the functions

$\boldsymbol\theta=(\theta_1,\theta_2)$63

the same paper proves full measure on nondegenerate analytic manifolds for Hardy $\boldsymbol\theta=(\theta_1,\theta_2)$64-functions $\boldsymbol\theta=(\theta_1,\theta_2)$65. It also shows that the set $\boldsymbol\theta=(\theta_1,\theta_2)$66 of badly $\boldsymbol\theta=(\theta_1,\theta_2)$67-approximable targets has zero measure on manifolds when $\boldsymbol\theta=(\theta_1,\theta_2)$68 is nonsingular, full measure when $\boldsymbol\theta=(\theta_1,\theta_2)$69 is very singular, and is absolute winning when $\boldsymbol\theta=(\theta_1,\theta_2)$70 is badly approximable (Beresnevich et al., 6 Jul 2025).

The subject is methodologically diverse. The weighted-exponent theory of (Marnat, 2014) relies on explicit lacunary constructions and minimal points in the sense of Davenport–Schmidt and Jarník. The full-dimension theorem for $\boldsymbol\theta=(\theta_1,\theta_2)$71 uses weighted best approximations, lacunarity, a Cantor construction, and the mass distribution principle (Bengoechea et al., 2015). Angularly constrained approximation uses parametric geometry of numbers and an angularly constrained realization theorem for rigid $\boldsymbol\theta=(\theta_1,\theta_2)$72-systems (Champagne et al., 2022). Restricted-denominator and toral-dynamical results use pairwise independence, second Borel–Cantelli, Fubini, discrepancy, and the Mass Transference Principle (Hauke et al., 2 Aug 2025, Chow et al., 18 Nov 2025). Weighted matrix-orbit results are formulated through weighted ubiquity and weighted mass transference (Hussain et al., 2023).

Two broader frameworks illuminate the subject’s conceptual range. “Diophantine Approximation Groups, Kronecker Foliations and Independence” associates to a real matrix $\boldsymbol\theta=(\theta_1,\theta_2)$73 the groups

$\boldsymbol\theta=(\theta_1,\theta_2)$74

and shows that planarity of the associated Kronecker foliations characterizes $\boldsymbol\theta=(\theta_1,\theta_2)$75-linear independence of columns, while density and minimality encode row independence (Gendron, 2012). “Local positivity and effective Diophantine approximation” is not formulated as a twisted-exponent paper, but it develops a weighted index

$\boldsymbol\theta=(\theta_1,\theta_2)$76

and a local-positivity method on blow-ups of $\boldsymbol\theta=(\theta_1,\theta_2)$77 for effective constrained simultaneous approximation, showing how geometric positivity can control anisotropic approximation conditions (Nickel, 2020).

Taken together, these developments show that twisted Diophantine approximation is not a single theorem or a single spectrum problem. It is a collection of anisotropic approximation theories whose behavior depends on the nature of the twist: weights, cones, fixed orbits, arithmetic subsequences, matrix actions, or manifold geometry. The classical two-dimensional identity

$\boldsymbol\theta=(\theta_1,\theta_2)$78

survives only in the balanced case; outside that case, the subject is governed by spectrum flexibility, directional rigidity, metric zero-full laws, and geometric transference rather than by a universal formula (Marnat, 2014).