Three essays on Machin's type formulas

Basic Information

• Paper ID: 2302.00154
• Title: Three essays on Machin's type formulas
• Authors: Armengol Gasull (Universitat Autònoma de Barcelona), Florian Luca (Wits University & UNAM), Juan L. Varona (Universidad de La Rioja)
• Classification: math.NT (Number Theory)
• Journal: Indagationes Mathematicae (to appear)
• Paper Link: https://arxiv.org/abs/2302.00154

Abstract

This paper investigates three problems related to Machin-type formulas. First, it provides a complete classification of all two-term Machin formulas where arctangent functions are evaluated at 2-integers (values of the form $b/2^a$), which are computationally useful since multiplication or division by powers of 2 are very fast operations on most computers. Second, it presents a method for finding infinitely many formulas with $N$ terms, which is particularly useful for $N=2$, recovering most known formulas, providing new ones, and enabling simple proof of the existence of two-term Machin formulas with arbitrarily small Lehmer measure. Finally, it corrects an oversight in previous results and provides all two-term Machin-type formulas involving arctangents of powers of the golden ratio.

Research Background and Motivation

Historical Background

Machin formulas originate from the identity discovered by John Machin in 1706: $4 \arctan\frac{1}{5} - \arctan\frac{1}{239} = \frac{\pi}{4}$

Combined with the arctangent series expansion discovered by Gregory in 1671: $\arctan x = \sum_{m=0}^{\infty} \frac{(-1)^m}{2m+1} x^{2m+1}, |x| < 1$

Machin used this formula to compute 100 decimal places of π.

Research Motivation

1. Computational Efficiency: Finding more efficient Machin-type formulas for computing π, particularly those involving powers of 2, since computers perform such operations extremely efficiently
2. Theoretical Completeness: Systematically classifying and constructing various Machin-type formulas
3. Lehmer Measure Minimization: Finding formulas with arbitrarily small Lehmer measure to improve computational efficiency
4. Correcting Historical Gaps: Perfecting the classification of Machin formulas involving the golden ratio

Limitations of Existing Methods

• Limited number of known Machin-type formulas
• Lack of systematic construction methods
• Insufficient research on special forms (e.g., formulas involving powers of 2)
• Previous research on the golden ratio contains gaps

Core Contributions

1. Complete classification of two-term Machin formulas involving 2-integers: Provides 10 isolated solutions and 2 parametric families
2. Proposes the "Machin formula machine" method: Systematically generates Machin-type formulas with $N$ terms
3. Proves the existence of two-term Machin formulas with arbitrarily small Lehmer measure: Provides constructive proof
4. Corrects and completes the classification of Machin formulas involving powers of the golden ratio: Discovers 16 essentially different identities

Detailed Methods

First Problem: 2-Integer Machin Formulas

Task Definition

Find all solutions of the form: $x_1 \arctan(z_1) + x_2 \arctan(z_2) = \frac{\pi}{4}$ where $x_1, x_2 \in \mathbb{Q}$, $z_k \in (0,1)$ and $z_k = 2^{a_k}/b_k$ or $b_k/2^{a_k}$.

Core Technique

Using complex number representation, the problem is transformed to: $(1 + iz_1)^{12u_1}(1 + iz_2)^{12|u_2|} = (1 - iz_1)^{12u_1}(1 - iz_2)^{12|u_2|}$

By analyzing factorization in Gaussian integers, combined with known Diophantine equation solutions:

• $x^2 + 1 = 2y^n$
• $x^2 + 2^a = y^n$

Second Problem: Machin Formula Machine

Core Idea

Define rational functions $R_j(n,x)$ ($j = 0,1,2,3$) such that: $\frac{d}{dx}\arctan(R_j(n,x)) = \frac{n}{1+x^2}$

Function Definition

Based on De Moivre's formula, define: $R_j(n,x) = \tan(n\arctan(x) + j\pi/4)$

Specific expressions:

• $R_0(n,x) = \frac{\text{numer}_n(x)}{\text{denom}_n(x)}$
• $R_1(n,x) = \frac{\text{denom}_n(x) + \text{numer}_n(x)}{\text{denom}_n(x) - \text{numer}_n(x)}$
• $R_2(n,x) = -\frac{1}{R_0(n,x)}$
• $R_3(n,x) = -\frac{1}{R_1(n,x)}$

Main Theorem

For integers $r_k$ satisfying $\sum_{k=1}^N r_k = 0$: $\sum_{k=1}^N \frac{r_k}{n_k} \arctan(R_{j_k}(n_k, x)) = \frac{r}{s}\pi$

Third Problem: Golden Ratio Formulas

Problem Description

Classify all formulas of the form: $x_1 \arctan(\phi^{a_1}) + x_2 \arctan(\phi^{a_2}) = \frac{\pi}{4}$ where $\phi = \frac{1+\sqrt{5}}{2}$ is the golden ratio.

Technical Method

By analyzing norms and prime factorization of elements in $K = \mathbb{Q}(i,\sqrt{5})$, utilizing the essential prime divisor theorem for Fibonacci and Lucas numbers.

Experimental Results

Main Results

Theorem 1: 2-Integer Machin Formulas

Discovers 10 isolated solutions and 2 parametric families:

Examples of isolated solutions:

• $(-1, \frac{1}{239}, 4, \frac{1}{5})$ (classical Machin formula)
• $(1, \frac{1}{41}, 2, \frac{2}{5})$

Parametric families:

• $(1, \frac{1}{2^{a^2}+1}, 1, \frac{2^{a^2}}{2^{a^2}+1})$
• $(1, \frac{1}{2^{a^2}-1}, 1, \frac{2^{a^2}-1}{2^{a^2}})$

Theorem 3: Small Lehmer Measure Formulas

Proves that for any $\varepsilon > 0$, there exists a Machin formula: $n \arctan\frac{1}{b_1} - \arctan\frac{a_2}{b_2} = \frac{\pi}{4}$ with Lehmer measure less than $\varepsilon$.

The constructive proof is based on convergents of the continued fraction of π, with specific results shown in Tables 1 and 2.

Specific Numerical Examples

Table 1 presents formulas based on convergents of the continued fraction of π, with Lehmer measure decreasing from 0.901429 to 0.172125.

Table 2 presents formulas based on $2^m$, achieving minimum Lehmer measure of 0.199544.

Golden Ratio Formulas

Discovers 16 essentially different identities, including:

• $\frac{\pi}{4} = \frac{1}{3}\arctan(\phi^3) + \frac{1}{3}\arctan(\phi)$
• $\frac{\pi}{4} = -\frac{1}{2}\arctan(\phi^5) + \frac{3}{2}\arctan(\phi)$

Related Work

Historical Development

• Störmer (1895): Proved that formulas of the form $x_1\arctan(1/m_1) + x_2\arctan(1/m_2) = \pi/4$ number only 4
• Lehmer (1938): Introduced Lehmer measure as a metric for computational efficiency
• Modern Development: Chudnovsky algorithm and other faster π computation methods

Related Research

The paper's "Machin formula machine" extends results from 4, which used recursive relations of Rédei polynomials to obtain special cases of the $R_3(n,x)$ function.

Conclusions and Discussion

Main Conclusions

1. Completely resolves the classification problem of two-term Machin formulas involving 2-integers
2. Provides a systematic method for constructing Machin formulas
3. Proves that formulas with arbitrarily small Lehmer measure can be constructed
4. Corrects and completes the classification of formulas related to the golden ratio

Limitations

1. For cases with $N > 2$, the method may not cover all known formulas
2. Although small Lehmer measure formulas can be constructed, the numerators and denominators have many digits, requiring trade-offs in practical computational efficiency
3. Modern π computation has more efficient algorithms (e.g., Chudnovsky algorithm), limiting the practical value of Machin-type formulas

Future Directions

1. Extension to Machin formulas with more terms
2. Investigation of other special forms of arctangent parameters
3. Exploration of connections with elliptic functions and modular functions

In-Depth Evaluation

Strengths

1. Theoretical Completeness: Systematically resolves several important classification problems
2. Methodological Innovation: The proposed "Machin formula machine" is general and practical
3. Constructive Proofs: Not only proves existence but also provides concrete construction methods
4. Historical Correction: Carefully addresses and corrects gaps in previous research

Technical Highlights

1. Clever Complexification: Transforms real number problems into algebraic problems in the complex domain
2. Application of Diophantine Equations: Effectively utilizes known solutions to Diophantine equations
3. Application of Continued Fraction Theory: Constructs small Lehmer measure formulas using properties of the continued fraction of π

Shortcomings

1. Limited Practical Value: These formulas show no obvious advantages in modern high-precision π computation
2. Complexity Issues: Constructed small Lehmer measure formulas involve integers with many digits
3. Generalization Limitations: Method applicability to more general cases is restricted

Impact

1. Theoretical Contribution: Provides a systematic framework for Machin-type formula theory
2. Computational Mathematics: Offers new tools for research on special function identities
3. Number Theory Applications: Demonstrates the power of algebraic number theory methods in concrete problems

References

The paper cites 24 important references, spanning from classical works by Machin (1706) and Gregory (1671) to modern related research, particularly foundational work by Störmer, Lehmer, and others in this field.