Multiplicative Diophantine Approximation

• Multiplicative Diophantine approximation is the study of approximating points by considering the product of coordinate errors rather than individual discrepancies.
• It establishes precise measure laws using limsup sets, where convergence or divergence of tailored series determines full or null Lebesgue measures and Hausdorff dimensions.
• The theory integrates mass transference principles and Fourier analytic methods to reveal non-Salem properties and a unique product formula for dimensional analysis.

Multiplicative Diophantine approximation concerns the study of how sets of points in Euclidean or toric spaces can be approximated by rational points—with a focus not on coordinatewise (“additive”) errors, but on products of approximation errors across several coordinates or linear forms. The core objects of interest are “limsup sets” of points for which some product of Diophantine errors falls below a given function infinitely often. Recent work has produced sharp metric, dimensional, and analytic criteria to classify the size and properties of such sets, with notable distinctions from additive and simultaneous approximation, and significant connections to mass transference, Fourier dimension, and the Salem property.

1. Definitions and Formulation

The general multiplicative limsup set is defined as follows (He, 13 Apr 2025):

Let $n, m \ge 1$ and an approximating function $\psi:\mathbb{Z}^n \to (0, 1/2]$ (often, $\psi$ is assumed to depend only on $|q| = \max_i |q_i|$ and to be non-increasing in $|q|$). Write the distance to the nearest integer as $\|x\|$, and for $x_j \in \mathbb{R}^n$, $\|q x_j\| := \|q_1 x_{j,1} + \dots + q_n x_{j,n}\|$. Then,

$$M_X(n, m; \psi) := \Big\{x = (x_1, ..., x_m) \in [0,1]^{nm} : \prod_{j=1}^m \|q x_j\| < \psi(q) \text{ for infinitely many } q \in \mathbb{Z}^n \setminus \{0\}\Big\}$$

Equivalently, $$M_X(n, m; \psi) = \limsup_{|q| \to \infty} M(q, \psi(q))$$ where $$M(q, \psi(q)) := \{x : \prod_{j=1}^m \|q x_j\| < \psi(q)\}$$.

This structure generalizes the classical multiplicative approximation set in $\mathbb{R}^m$, i.e., $$\{x \in [0,1]^m : \prod_j \|q x_j\| < \psi(q) \text{ i.o.}\}$$, and includes higher-dimensional and weighted settings.

2. Metric Laws: Zero–One Laws for Lebesgue Measure

A central question is: for which $\psi$ does $M_X(n, m; \psi)$ have full Lebesgue measure, and when is it null?

The main result, fully generalizing and extending previous work, is [(He, 13 Apr 2025), Thm. 1.7]:

• For any $n, m \ge 1$ and $\psi$ non-increasing,

$\lambda^{nm}\big(M_X(n, m; \psi)\big) = \begin{cases} 0, & \sum_{q \neq 0} \psi(q)\left(\log_2 \psi(q)^{-1}\right)^{m-1} < \infty, \[1.5ex] 1, & \sum_{q \neq 0} \psi(q)\left(\log_2 \psi(q)^{-1}\right)^{m-1} = \infty. \end{cases}$

Thus, the Lebesgue measure is determined by a series whose terms involve the $m-1$st power of the binary logarithm of the error threshold.

A distinguished feature is the absence of further “extra” log factors beyond $\left(\log_2 \psi(q)^{-1}\right)^{m-1}$ for $m > 2$, resolving a question of Hussain and Simmons.

3. Hausdorff Measure and Dimension Criteria

The fine-scale size of $M_X(n, m; \psi)$ is captured by its Hausdorff $f$-measure for a dimension function $f$ [(He, 13 Apr 2025), Thm. 1.8, 1.9]. Two regimes appear:

• Non-critical dimensions (for $f(r) \asymp r^s$, $(n-1)m < s \le nm-1$):

$\mathcal{H}^f\big(M_X(n, m; \psi)\big) = \begin{cases} 0, & \sum_{q \ne 0} f\left(\frac{\psi(q)}{|q|}\right)\left(\frac{\psi(q)}{|q|}\right)^{1-nm} < \infty, \[1.5ex] \mathcal{H}^f([0,1]^{nm}), & \sum_{q \ne 0} \left(\frac{\psi(q)}{|q|}\right)^{1-nm} = \infty. \end{cases}$

• Critical log-dimension (for $f(r)=r^{nm}\log(1/r)$):

$\mathcal{H}^f\big(M_X(n, m; \psi)\big) = \begin{cases} 0, & \sum_{q \ne 0} \psi(q)|q|^{-n} \left(\log_2 \psi(q)^{-1}\right)^{m-1} < \infty, \[1.5ex] \mathcal{H}^f([0,1]^{nm}), & \sum_{q \ne 0} \psi(q)|q|^{-n} \left(\log_2 \psi(q)^{-1}\right)^{m-1} = \infty. \end{cases}$

These correspond to a sharp “zero–full” ($0$ or maximal) phenomenon in Hausdorff measure, and the divergence/convergence point precisely tracks the natural $L^1$-sums as in classical metrical Diophantine theory, except for the expected extra log factors—ultimately reflecting the hyperbolic structure of the multiplicative setup.

4. Mass Transference Principle: Balls-to-Rectangles

A key methodological advance is the systematic use of the “balls-to-rectangles” mass transference principle to transfer full-measure statements from Lebesgue measure and balls to more structured sets and Hausdorff measures (He, 13 Apr 2025).

The approach is as follows:

• Given full Lebesgue measure in the limsup of rectangles $M(q, \psi(q))$, one can cover these rectangles with balls of comparable small radius $\asymp \psi(q)/|q|$.
• Inside each such ball, one can find a rectangle with lower-bounded Hausdorff content.
• The mass transference principle then lifts the full-measure result for balls to a corresponding Hausdorff $f$-measure result for the limsup set of rectangles.

This device unifies and extends the dimension theory of multiplicative sets beyond what was possible with direct Borel–Cantelli/probabilistic methods, and is essential for capturing the critical log scaling consistently.

5. Fourier Dimension and Non-Salem Phenomenon

The analytic complexity of multiplicative limsup sets is measured by their Fourier dimension. For the class $M_X(n, m; \psi)$ (assuming univariable and not necessarily monotonic $\psi$), a sharp formula is available [(He, 13 Apr 2025), Thm. 1.15]:

• Let

$$T(\psi) := \inf\left\{ t \ge 0 : \sum_{q \ne 0} \left(\psi(q)^{1/m} |q|^{-1}\right)^t < \infty \right\}.$$

Then,

$\dim_F(M_X(n, m; \psi)) = 2 T(\psi).$

Except in the trivial $n=m=1$ case (the classical one-dimensional setting), this Fourier dimension strictly falls below the Hausdorff dimension, so these sets are typically non-Salem. This breaks from the additive theory, where Salem sets are generic in many limsup constructions (e.g., Kaufman's result for additive approximation), and is a distinctive rigidity of the multiplicative regime (Tan et al., 2024).

A further structural result is the product formula for Fourier dimension [(He, 13 Apr 2025), Thm. 1.13]: For Borel sets $E \subset \mathbb{R}^{d-k}$, $F \subset \mathbb{R}^k$ with zero Lebesgue measure for $E \times F$, $\dim_F(E \times F) = \min\{ \dim_F E, \dim_F F \}$. This result both explains the scarcity of Salem sets in multiplicative approximation and unifies diverse special cases that are approached via “product set” decompositions.

6. Resolution of Long-Standing Problems and Further Implications

Several open questions have now been conclusively answered:

• Hussain–Simmons asked whether further log factors are genuinely required in the measure dichotomy for $m > 2$. The answer is negative; only $(\log_2 \psi(q)^{-1})^{m-1}$ is needed (He, 13 Apr 2025).
• The full dimension, measure, and Fourier analysis of multiplicative limsup sets—including in weighted and inhomogeneous variants—now fit naturally into a balls-to-rectangles transference framework, which also aligns these results with recent perspectives in general weighted approximation theory (He, 13 Apr 2025).
• The strict non-Salem property in higher dimensions, and the minimal product formula for Fourier dimension, provide a clear obstruction to additive-multiplicative transfer of many analytic methods, and suggest further lines of inquiry into exceptional sets and intersections.

7. Context and Outlook

Multiplicative Diophantine approximation thus distinguishes itself as a setting where geometry, harmonic analysis, and dynamical systems intersect. The metric and dimensional theory for limsup sets is now mature for full-measure and Hausdorff-dimension problems—both in the classical and p-adic settings (Datta et al., 2019). Open areas remain for inhomogeneous and fibred analogs of Duffin–Schaeffer-type results (Yu, 2020), the distribution of multiplicative errors in higher codimension (and on submanifolds or fractals (Chow et al., 2024, Chow et al., 2024)), and the analytic structure in function fields and non-Archimedean geometries.

Comprehensive coverage of the measure, Hausdorff measure, and Fourier dimensions of multiplicative Diophantine sets—together with the powerful balls-to-rectangles mass transference mechanism—now enables precise exploration of exceptional sets, fine-scale distribution of rational approximants, and the search for Salem sets within restricted frameworks, marking both a culmination of decades of research and the basis for new structural and analytic questions (He, 13 Apr 2025, Tan et al., 2024, Frühwirth et al., 2024).