Suppose for each $s$ in an open interval, $P_s(x)=\sum_{k=0}^\infty a_k(s) x^k$ is a power series with radius of convergence greater than R, where each $a_k(s)$ is differentiable. My question is: Is the radius of convergence of $\sum\limits_{k=0}^\infty a_k'(s) x^k$ at least R?


The radius of convergence can be $0$ for some $s$. Let $h \colon \mathbb{R}\to\mathbb{R}$ a continuous (or smooth, if you want your coefficients smooth) function with support in $[-1,1]$, $h(0) = 1$, $h \geqslant 0$ and $\int_{-1}^1 h(t)\,dt = 1$. Let

$$a_k(s) = k!\int_0^s h((k!)^2\cdot t)\,dt.$$

Then you have $\lvert a_k(s)\rvert \leqslant \frac{1}{k!}$, so the radius of convergence is $\infty$ for all $s$, but $a_k'(0) = k!$, so

$$\sum_{k=0}^\infty a_k'(0)x^k$$

has radius of convergence $0$.

  • $\begingroup$ Is it possible if a_k's are analytic? $\endgroup$ – TCL May 9 '14 at 17:33
  • $\begingroup$ You can take $a_k(s) = (k!s) e^{-(k!s)^2}$, then the $a_k(s)$ are uniformly bounded, so the radius of convergence is at least $1$ for all $s$, but $a_k'(0)= k!$. If $f(s,x) = \sum a_k(s)\cdot x^k$ is analytic in both variables jointly, then it cannot happen. $\endgroup$ – Daniel Fischer May 9 '14 at 17:43
  • $\begingroup$ Perhaps your last comment here answer my other question? since the coefficents A_i there are derivatives of rational functions in s. If so, can you answer the question there?:math.stackexchange.com/questions/787379/… $\endgroup$ – TCL May 9 '14 at 21:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.