The Lerch Phi Function and a possible Mathematica Bug [Solved]

284 Views Asked by At

The Lerch Phi Function is defined as

$$\Phi(-s, \alpha, \nu) = \sum_{k = 0}^{+\infty} \frac{(-s)^k}{(k+\nu)^{\alpha}}$$

Now, in my special case I have $\alpha = 1$, hence it does simply reduce to

$$\Phi(-s, 1, \nu) = \sum_{k = 0}^{+\infty} \frac{(-s)^k}{(k + \nu)}$$

Clearly if I put $\nu = 0$ the series shall be infinite.

But when I compute it with Mathematica (Wolfram Serious Mathematica software, not Online Alpha) it says that

$$\Phi(-s, 1, \nu) = -\ln(1+s)$$

How is this possible?

EDIT

There are also other problems related to that. For example setting $\nu = -1$ and Mathematica says

$$\Phi(-s, 1, -1) = 1+ s\ln(1+s)$$

Which is actually not true, since by its definition (both it starts from $k = 0$ or $k = 1$ there is a point in which the series is infinite).

1

There are 1 best solutions below

3
On BEST ANSWER

According to the reference, we have

$$\Phi(-s, \alpha, \nu) = \sum_{k = 0}^{+\infty} \frac{(-s)^k}{(k+\nu)^{\alpha}}$$

for $\Re(\nu)>0$. For $\Re(\nu)\le0$, we have

$$\Phi(-s, \alpha, \nu) = \sum_{k = 0}^{+\infty} \frac{(-s)^k}{[(k+\nu)^2]^{\alpha/2}}$$

where any term $k+\nu=0$ is eliminated.

References:


http://reference.wolfram.com/language/ref/LerchPhi.html