Abstract:Many upper bounds for the moduli of polynomial roots have been proposed but reportedly assessed on selected examples or restricted classes only. Regarding quality measured in terms of worst-case relative overestimation of the maximum root-modulus we establish a simple, nearly optimal result.
What problem does this paper attempt to address?
### The problems the paper attempts to solve
This paper aims to solve the problem of estimating the upper bound of the modulus of polynomial roots, especially the upper bound of the largest root modulus. Specifically, the author hopes to find a simple and nearly optimal result for evaluating the maximum modulus of polynomial roots.
#### Main problem description:
1. **Limitations of existing methods**:
- There have been many proposals regarding the upper bound of the modulus of polynomial roots, but these methods are usually evaluated only on specific examples or restricted classes of polynomials.
- The common method of evaluating these upper bounds is by choosing specific polynomials or batch - testing polynomial sets with a given coefficient distribution, which lacks universal applicability.
2. **Quality measurement criteria**:
- In 1970, van der Sluis proposed a method for measuring the quality of the upper bound of the root modulus, which avoids the selection of examples or coefficient distributions.
- This measurement criterion is based on the worst - case relative overestimation (\( \frac{B(p)}{\mu(p)} \)), where \( B(p) \) is the given upper bound of the root modulus and \( \mu(p) \) is the maximum root modulus of the polynomial.
3. **Theoretically optimal result**:
- According to van der Sluis' result, the Cauchy upper bound (\( \rho(p) \)) has the smallest worst - case relative overestimation among all absolute root - modulus upper bounds that use only the absolute values of the coefficients.
- For any non - trivial polynomial \( p(z) \), its worst - case relative overestimation is \( \frac{1}{\sqrt[n]{2}-1}\approx1.442n \).
4. **Improvement goals**:
- The author hopes to obtain an upper bound closer to the theoretically optimal result by introducing new methods, especially methods based on parameterized Cassini ovals.
- The goal is to make the worst - case relative overestimation of the new upper bound as close as possible to \( 1.442n \) and not exceed 2% of the theoretical lower limit.
#### Method overview:
- **Introducing Cassini ovals**: By considering the Frobenius companion matrix of the polynomial and using similarity transformation and row - sum estimation, the author derives a new inclusion - circle method.
- **Defining a new upper bound \( \Gamma(p) \)**: Based on the above method, a new upper bound of the root modulus \( \Gamma(p) \) is defined, and it is proved that its worst - case relative overestimation is very close to the theoretically optimal value.
#### Conclusion:
Through this method, the author has successfully found a polynomial root - modulus upper bound \( \Gamma(p) \) close to the theoretical optimum. Its worst - case relative overestimation does not exceed \( 1.4655n \). For high - degree polynomials (such as \( n\geq85 \)), the gap between this result and the theoretical lower limit does not exceed 2%.
In summary, the main contribution of this paper is to provide a new, nearly optimal method for estimating the upper bound of the modulus of polynomial roots, which solves the deficiencies of existing methods in universality and accuracy.