Abstract
In Diophantine approximation, the two-dimensional case of simultaneous rational approximation remains a significant open problem. This paper identifies specific pairs of irrational numbers that are exceptionally difficult to approximate, approaching the conjectured theoretical limit of $2/7$. By performing high-precision computations on the continued fraction of $\theta = 2\cos(2\pi/7)$, the author discovers pairs with an approximation constant $c^* \approx 0.2857082$, the largest yet recorded.
Introduction
The study of how closely irrational numbers can be approximated by rationals is a cornerstone of number theory. While the one-dimensional case is well-understood through continued fractions—where the golden ratio is known to be the "most irrational"—the two-dimensional case involving simultaneous approximation with a common denominator is far more complex. We seek to find the constant $c_2$ that represents the supremum of the lower bounds of approximation quality for all pairs. Mathematical folklore suggests $c_2 = 2/7$, but explicit examples reaching this bound are notoriously difficult to find.
Methodology
The search focuses on the cubic number field $\mathbb{Q}(\theta)$, where $\theta = 2\cos(2\pi/7)$. This field is chosen for its minimal positive discriminant (49), making it a prime candidate for finding "worst-case" numbers. The author employs a construction based on Cusick’s theorem, which relates the approximation constant of an integral basis $\{1, \alpha, \beta\}$ to specific patterns in the continued fraction expansion of $\theta$.
To find these patterns, over 2 million partial quotients of the continued fraction of $\theta$ were computed using high-precision arithmetic. The resulting candidates were verified using the Furtwängler algorithm to ensure their status as "best approximations" and to calculate their approximation constants accurately.
The Long Initial Transient
A critical discovery in this work is the phenomenon of the "long initial transient." Many pairs that are theoretically badly approximable do not exhibit this property for small denominators $q$. Instead, they appear easily approximable until very large denominators are reached.
As shown in Figure 1, the approximation quality can fluctuate wildly before settling into the characteristic "bad" behavior. For the record-breaking Pair (G), the integers defining the pair are of the order $2^{3,000,000}$, requiring denominators of staggering size before the approximation constant stabilizes near the $2/7$ limit.
Results and Conclusion
The investigation yielded seven specific pairs (labeled A through G) with exceptionally large constants. The most significant, Pair (G), achieved $c^* \approx 0.2857082$. This value is remarkably close to $2/7 \approx 0.2857143$, providing strong numerical evidence for the $2/7$ conjecture. These results demonstrate that while badly approximable pairs are rare and computationally difficult to isolate, Cusick's theoretical framework provides a robust path toward identifying the limits of Diophantine approximation in higher dimensions.
Summary updated