The Bombieri--van der Poorten Formula for Partial Quotients of Higher Degree Algebraic Irrationals

1. Problem Statement

The paper addresses the problem of determining the partial quotients ($b_{n+1}$) of the regular continued fraction expansion for algebraic irrationals of the form $\alpha = \sqrt[m]{k}$, where $m \geq 3$.

While the Bombieri–van der Poorten formula provides a relationship between the partial quotients and the algebraic distance $d_n = |p_n^m - k q_n^m|$, the explicit analytical behavior of the remainder term ($R_n$) for degrees $m \geq 3$ has not been fully characterized. Specifically, the paper seeks to:

• Derive a closed-form expression for the error term $R_n$ in the decomposition $\theta_n = H_n + R_n$, where $\theta_n$ is the complete quotient and $H_n$ is a leading algebraic approximation.
• Determine the conditions under which $|R_n| < 1$, which is the critical threshold required to uniquely determine $b_{n+1}$ via the floor function $\lfloor H_n + R_n \rfloor$.
• Investigate whether the stability observed in lower-degree cases holds for higher degrees ($m > 3$).

2. Methodology

The author employs a rigorous analytical approach combining continued fraction theory with algebraic number theory.

• Derivation of the Remainder:
  Starting from the identity for the difference of $m$-th powers, the author expresses the complete quotient $\theta_n$ in terms of the convergent $x_n = p_n/q_n$ and the algebraic distance $d_n$. By linearizing the sum of powers (approximating $x_n \approx \alpha$), the author defines a leading term $H_n = \frac{m p_n^{m-1}}{d_n q_n}$.
  The remainder $R_n$ is derived explicitly as:
  $$R_n = \frac{q_n^{m-2}}{d_n} \left( \sum_{j=0}^{m-1} x_n^j \alpha^{m-1-j} - m x_n^{m-1} \right) - \frac{q_{n-1}}{q_n}$$
  This formulation isolates the algebraic correction term (arising from the non-linearity of the power sum) and the standard continued fraction term ($q_{n-1}/q_n$).

• Case Analysis (Cubic vs. Higher Degree):

  • Cubic Case ($m=3$): The author performs an exact algebraic factorization of the correction term for $m=3$. By analyzing the sign and magnitude of this term relative to $q_n$, a global stability proof is constructed.
  • Higher Degree ($m > 3$): The author utilizes the Mean Value Theorem to bound the algebraic error term, showing it decays as $O(q_n^{-2})$. However, the paper argues that for higher degrees, the linear approximation $H_n$ may deviate significantly from the actual sum of powers before the $q_n^{-2}$ decay dominates, potentially violating the $|R_n| < 1$ condition for small $n$.

• Counterexample Construction:
  To test the limits of the theory, the author constructs a specific counterexample using a 10th-degree irrational ($\alpha = \sqrt[10]{50}$) to demonstrate a failure of the stability bound.

3. Key Contributions and Results

A. Closed-Form Error Representation

The paper provides a precise analytical formula for the remainder $R_n$ for any degree $m \geq 3$. This moves beyond asymptotic estimates to an exact representation involving the convergent $x_n$, the root $\alpha$, and the algebraic distance $d_n$.

B. Global Stability for Cubic Irrationals ($m=3$)

Theorem 2.1 is the paper's central result. It proves that for $\alpha = \sqrt[3]{k}$, the remainder term satisfies $|R_n| < 1$ for all convergents where $q_n \geq 2$.

• Implication: For cubic irrationals, the next partial quotient is uniquely determined by the leading term $H_n$ within a window of length two.
• Explicit Bounds: If $x_n > \alpha$, then $H_n - 2 < b_{n+1} \leq H_n$.
• Floor Formula: The paper establishes that $b_{n+1} = \lfloor H_n - \frac{q_{n-1}}{q_n} \rfloor + \varepsilon_n$, where $\varepsilon_n \in \{0, 1\}$, and the ambiguity vanishes if the fractional part of the argument is non-zero.

C. Generalization to Higher Degrees ($m > 3$)

Theorem 3.1 extends the framework to arbitrary degrees $m$. It establishes that $b_{n+1}$ is constrained by $H_n$ provided $q_n$ is sufficiently large (specifically $q_n^2 \geq C(m, \alpha)$).

• The paper notes that while the asymptotic decay $O(q_n^{-2})$ holds, the constant factor depends on $m$ and $\alpha$.
• For small denominators in higher degrees, the linear approximation may fail to bound the remainder within $(-1, 1)$.

D. Counterexample for Higher Degrees

Section 5 presents a critical counterexample for $m=10$ ($\alpha = \sqrt[10]{50}$).

• For the convergent $p_1/q_1 = 3/2$ ($q_1=2$), the calculated remainder is $R_1 \approx -1.27$.
• Since $|R_1| > 1$, the heuristic $b_2 \approx \lfloor H_1 \rfloor$ fails (predicting 12 instead of the correct 11).
• This confirms that the "global stability" (valid for all $q_n \geq 2$) is a unique structural property of cubic irrationals and does not automatically extend to higher degrees.

4. Significance and Impact

• Algorithmic Efficiency: The results eliminate the need for exhaustive numerical verification of partial quotients for cubic irrationals. Researchers can now rely on the Bombieri–van der Poorten formula with a guaranteed error bound for all $q_n \geq 2$.
• Refinement of Diophantine Approximation: The paper clarifies the relationship between the algebraic distance $d_n$ and the continued fraction partial quotients, providing a tighter "Diophantine window" for high-degree algebraic numbers.
• Theoretical Boundaries: By identifying the specific failure mode for $m > 3$ at small denominators, the paper sets clear boundaries for the applicability of effective constants in Roth's Theorem and related algorithmic problems.
• Foundation for Future Work: The rigorous derivation of the error term provides a robust framework for investigating effective versions of Roth's Theorem and the algorithmic determination of continued fractions for high-degree algebraic numbers, as referenced in the conclusion.

In summary, Müller's paper rigorously validates the Bombieri–van der Poorten framework for cubic irrationals, proving strict stability, while simultaneously delineating the precise conditions under which this stability breaks down for higher-degree algebraic numbers.