9

Does the following series have a closed-form expression:

$$\sum_{k=0}^{\infty} \frac{z^k}{(k!)^2}$$

I know that it must converge because:

$$\sum_{k=0}^{\infty} \frac{z^k}{k!} = e^z$$

and the $(k!)^2$ denominator obviously increases more quickly than the $k!$ denominator.

This problem came up in computing the probability of a draw in a football match with each team's goal scoring modeled as a Poisson process.

Thanks,

John

JohnT
  • 103

2 Answers2

16

Denote

$$f(z)=\sum_{n=0}^\infty\frac{z^n}{(n!)^2}$$

By term-wise differentiation, we find that

$$f'(z)+zf''(z)=f(z)$$

A rather simple differential equation with the general solution

$$f(z)=c_1I_0(2\sqrt z)+c_2K_0(2\sqrt z)$$

where $I_n$ is a modified Bessel function of the first kind, and $K_n$ is a modified Bessel function of the second kind. By using

$$f(0)=f'(0)=1$$

we find that

$$f(z)=I_0(2\sqrt z)$$

  • Your answer is accurate, however: Is there any proof that this particular Bessel function of the first kind has no closed form in terms of the elementary functions?(finite combination of polynomials, trigonometrics, exponentials and logarithms) or does it has some? – Dr Potato Jun 08 '21 at 18:22
  • None that I know of. If such a closed form existed and was known however, it would certainly appear in e.g. WolframAlpha or Wikipedia. – Simply Beautiful Art Jun 09 '21 at 00:01
1

Another approach: Consider this sequence as the direct product of the reciprocal of the factorial function with itself.

Given that the exponential function is the generating function of the reciprocal of the factorial $$e^z=\sum_{n\in\mathbb{N}} \frac{z^n}{n!},$$ take $$f(z)=\sum_{n\in\mathbb{N}}\frac{z^n}{(n!)^2}$$ as the generating function of the termwise product of the former sequence with itself and use the Hadamard product of two generating functions formula: $$f(z)= e^z\cdot e^z = \sum_{n\in\mathbb{N}} \frac{1}{n!}\frac{1}{n!}z^n = \frac{1}{2\pi i}\oint_{x=0} e^{x\sqrt{z}}e^{\frac{\sqrt{z}}{x}} \frac{dx}{x},$$ so $$\boxed{\sum_{n\in\mathbb{N}}\frac{z^n}{(n!)^2} = \frac{1}{2\pi i}\oint_{x=0} e^{\sqrt{z}(x+\frac{1}{x})} \frac{dx}{x}}$$ which indeed matches with the $0$-th Modified Bessel Function of the First Kind asserted in the previous answer.

Now the task is to find a closed form of this function or to prove that this modified Bessel function of the first kind has no closed expression in terms of the standard elementary functions.

Dr Potato
  • 772