Convergence of formal logarithmic derivatives #
This file connects the formal logarithmic derivative of a complex power series to its analytic sum. If a power series has constant coefficient one and its analytic sum has no zero in a disk of convergence, then its formal logarithmic derivative converges throughout that disk.
The zero-free hypothesis is essential: the radius of the logarithmic derivative is limited by the
nearest zero of the original series, even when the original series converges farther.
Quantitatively, if the coefficients of f - 1 have absolute sum less than one on a circle, the
file gives an explicit bound for the absolute coefficient sum of f'/f there.
Main results #
PowerSeries.hasSum_coeff_logDeriv_mul_pow_of_zeroFree: evaluation of the formal logarithmic derivative throughout a zero-free convergence disk.PowerSeries.summable_norm_coeff_logDeriv_mul_pow_of_zeroFree: absolute convergence of the formal logarithmic derivative throughout a zero-free convergence disk.PowerSeries.tsum_norm_coeff_logDeriv_mul_pow_succ_le: an explicit bound for the absolute coefficient sum of the formal logarithmic derivative whenfis close to1.
The coefficients of a formal logarithmic derivative are the normalized iterated derivatives at zero of the analytic logarithmic derivative.
A formal logarithmic derivative sums to the analytic logarithmic derivative throughout a
zero-free disk. Let f be a complex power series with constant coefficient one. If its analytic
sum has no zero in a disk inside its disk of convergence, then the coefficient series of
f.logDeriv sums to the analytic logarithmic derivative at every point of the smaller disk.
A formal logarithmic derivative converges absolutely throughout a zero-free disk.
A formal logarithmic derivative converges throughout a zero-free disk.
Under the hypotheses of PowerSeries.tsum_norm_coeff_logDeriv_mul_pow_succ_le, the series
∑ m, |[Xᵐ] (f'/f)| r ^ (m + 1) converges.
A majorant for the coefficients of a formal logarithmic derivative. Write aₙ for the
coefficients of f, where a₀ = 1, and let r ≥ 0. If
T = ∑ n, n |aₙ| rⁿ converges and t = ∑_{n ≥ 1} |aₙ| rⁿ < 1, then
∑ m, |[Xᵐ] (f'/f)| r ^ (m + 1) ≤ T / (1 - t).
For the majorant F(X) = ∑ |aₙ| Xⁿ this reads ∑ m, |[Xᵐ] (X f'/f)| rᵐ ≤ r F'(r) / (2 - F(r)).
No zero-freeness hypothesis is needed beyond t < 1. For a family, the bound is uniform only
when t is bounded uniformly below 1 and the corresponding values of T are controlled.