7
$\begingroup$

Consider an entire function $f:\mathbb C \to \mathbb C$ that is real on the real line and even.

This function has a Taylor series of the form $$f(z) = \sum_{i=0}^{\infty} a_i z^{2i} \text{ with } a_i \in \mathbb R.$$

I have an estimate of the form

$$\vert e^{z^2 \mu} f(z) \vert \le e^{\vert z \vert^4 \nu}$$ for all $z \in \mathbb C$ and $\mu, \nu >0.$

I could now bound the Taylor coefficients of this power series using Cauchy estimates, see for instance here, by using that

$$\vert f(z) \vert \le e^{\vert z \vert^4\nu + \vert z \vert^2 \mu}.$$

However, I feel that I am wasting something here, as the initial estimate already implies that on the real line $f$ grows at most like $e^{z^4\nu}$ when multiplied by a Gaussian.

So the bound I derived for my Cauchy estimate is very conservative on the real line.

Of course I cannot immediately improve upon this estimate since when $z$ is purely imaginary, that bound may be optimal and Cauchy estimates are supposed to hold on a circle in the complex domain. However, since the function is real on the real line, the growth should really be determined by my initial estimate somehow.

Therefore my question is: Can I do anything better here to bound the Taylor coefficients than applying Cauchy estimates to the estimate
$$\vert f(z) \vert \le e^{\vert z \vert^4\nu + \vert z \vert^2 \mu}.$$

$\endgroup$

1 Answer 1

8
$\begingroup$

Let $z=re^{i\theta}$. Then your original inequality implies $$|f(re^{i\theta})|\leq \exp(\nu r^4-\mu r^2\cos2\theta).$$ Then instead of Cauchy estimate you can use the exact formula for the coefficient: $$|a_{2n}|=\left|\frac{1}{2\pi i}\int_{|z|=r}\frac{f(z)}{z^{2n+1}}dz\right|\leq\frac{1}{2\pi} \int_0^{2\pi}|f(re^{i\theta})|r^{-2n}d\theta$$ $$\leq r^{-2n}\exp(\nu r^4)\frac{1}{2\pi}\int_0^{2\pi}e^{-\mu r^2\cos2\theta}d\theta,$$ and now take the minimum, with respect to $r$. (The last integral can be expressed in terms of the modified Bessel function $I_0$, if needed.)

However, this can give you an advantage only for small $n$, since for large $n$, the term with $r^2$ in the first estimate is negligible.

$\endgroup$
0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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