The Stern Sequence and Moments of Minkowski's Question Mark Function

Roland Bacher

DOI: https://doi.org/10.48550/arXiv.1703.07268

2017-03-21

Abstract:1 : We use properties of the Stern Sequence for numerical computations of moments $\int^1_0 t^n d?(t)$ associated to Minkowski's Question Mark function.

Number Theory

What problem does this paper attempt to address?

The core problem that this paper attempts to solve is to calculate the moments related to Minkowski's Question Mark Function through the properties of Stern sequences. Specifically, the paper focuses on the integral \(\int_0^1 t^n d?(t)\), where \(d?(t)\) is the probability measure related to Minkowski's Question Mark Function.

### Problem Background

Minkowski's Question Mark Function \(?(x)\) and its inverse function Conway's box function \(\Box(x)\) are related to continued fraction expansions, transcendental properties, and the probability distribution of rational numbers in the Calkin - Wilf tree. Denjoy proved that \(?(x)\) is monotonically continuous and singular (on the set where it is almost everywhere differentiable but the derivative is zero). Alkauskas studied the moment sequence \(m_n=\int_0^1 x^n d?(s)\) of the probability density \(d?\) and obtained some properties of these moments.

### Research Methods

The author Roland Bacher uses the properties of Stern sequences to study these moments. The Stern sequence is defined as follows:
\[
s(0) = 0, \quad s(1) = 1, \quad s(2n) = s(n), \quad s(2n + 1) = s(n)+s(n + 1), \quad n\geq1.
\]
By relating the moment \(m_n\) to the Stern sequence, the author provides some new proofs and discovers some new properties.

### Main Contributions

1. **Connecting Stern Sequences and Conway's Box Function**: The author shows how to use Stern sequences to define Conway's box function \(\Box(x)\).
2. **Linear Relationships of Moments**: By considering Riemann sums, the author obtains a set of linear relationships, which are different from Alkauskas' previous work and can be more easily subjected to spectral analysis.
3. **Asymptotic Formulas of Moments**: The author derives asymptotic formulas for the moment \(m_n\), which are more complex than Alkauskas' formulas but are more accurate experimentally.
4. **Moment Values at Negative Integer Moments**: The author also discusses the moment values at negative integer moments and gives a sequence identity.
5. **Numerical Computation**: The author proposes an efficient numerical algorithm based on the above theoretical results for calculating the moment \(m_n\).

### Conclusion

By introducing Stern sequences, the author not only provides new insights into the moments of Minkowski's Question Mark Function but also improves the existing asymptotic formulas and numerical computation methods. This provides new tools and perspectives for in - depth understanding of Minkowski's Question Mark Function and its related properties.

### Related Formulas

- Definition of Minkowski's Question Mark Function:
\[
?(x)=- 2\sum_{k = 1}^{\infty}\frac{(-1)^k}{2^{a_1+\cdots+a_k}}
\]
where \(x = [0; a_1, a_2, a_3,\ldots]\) is the continued fraction expansion of \(x\).

- Definition of Moments:
\[
m_z=\int_0^1\Box(t)^z dt=\int_0^1 x^z d?(x)
\]

- Asymptotic Formula:
\[
m_n\sim\lambda\sum_{h = 2}^{\infty}\frac{1}{2^h}\left(1-\frac{1}{h}\right)^n
\]
where \(\lambda=\sum_{n = 0}^{\infty}\frac{(\log 2)^n}{n!}m_n\).

These formulas and conclusions together constitute the main contributions of this paper.