2

I've recently been searching around the web for an inverse of the zeta function, $ \zeta^{-1}(s) $, somewhat unsuccessfully. I then came across $$ \zeta(s) = \frac{\eta(s)}{1-2^{1-s}}, $$ where $$\eta(s) = \sum^\infty_{k=1} \frac{-1^{k-1}}{k^s}$$ I then manipulated it like so:

$$\begin{align*} \zeta(s) &= \frac{\eta(s)}{1-2^{1-s}} \\[0.1in] \frac{1}{\zeta(s)} &= \frac{1-2^{1-s}}{\eta(s)} \\[0.1in] \frac{\eta(s)}{\zeta(s)} &= 1-2^{1-s} \\[0.1in] \frac{\eta(s)}{\zeta(s)} - 1 &= -2^{1-s} \\[0.1in] -\frac{\eta(s)}{\zeta(s)} - 1 &= 2^{1-s} \\[0.1in] \log_2\Bigl(-\frac{\eta(s)}{\zeta(s)} - 1\Bigr) &= 1-s \\[0.1in] -\log_2\Bigl(-\frac{\eta(s)}{\zeta(s)} - 1\Bigr) -1 &= s \end{align*}$$

So there you go? An inverse of the zeta function. However, when I input it into Wolfram Alpha, it doesn't work, as in:

$$ \zeta^{-1}(2) \approx -1.58496-4.53236i $$ $$ \zeta(-1.58496 + -4.53236i) = 0.282432... - 0.377775... i $$

Any help?

Zev Chonoles
  • 129,973
Olly Britton
  • 105
  • 5

1 Answers1

1

You made an arithmetic mistake in multiplying both sides by $-1$:

$$\frac{\eta(s)}{\zeta(s)} - 1 = -2^{1-s} \implies -\frac{\eta(s)}{\zeta(s)} + 1 = 2^{1-s} $$

(By the way, you haven't really produced an inverse of the Riemann zeta function, since your formula depends on the value of $s$ itself, in order to plug into $\eta$.)

Zev Chonoles
  • 129,973
  • Surely then I could rewrite it as – Olly Britton Feb 27 '17 at 18:31
  • $$ \eta(s) = (1-2^{1-s})\zeta(s) $$ – Olly Britton Feb 27 '17 at 18:31
  • 1
    @OllyBritton: That still depends on $s$. – Zev Chonoles Feb 27 '17 at 18:34
  • But surely to evaluate $\zeta^{-1}(s)$ all you need to do is plug it into the formula and you can calculate the values of $\zeta(s)$ yourself? – Olly Britton Feb 27 '17 at 18:57
  • @OllyBritton: Plug what into the formula? Remember, the whole point of the inverse of a function is that I can tell you a number, $x$, and you can tell me which number $s$ has the property that $\zeta(s)=x$. – Zev Chonoles Feb 27 '17 at 19:04
  • 1
    Note that using $\eta(s)=(1-2^{1-s})\zeta(s)$ also makes your formula a complete tautology, because the $\zeta(s)$ cancels and has no impact on anything: $$-\log_2\Bigl(-\frac{(1-2^{1-s})\zeta(s)}{\zeta(s)} + 1\Bigr) -1 = s$$ $$-\log_2\bigl(-(1-2^{1-s}) + 1\bigr) -1 = s$$ $$-\log_2(2^{1-s}) -1 = s$$ $$-(1-s)-1=s$$ $$s=s$$ Your formula takes in a value of $\zeta(s)$ and then never uses it for anything. – Zev Chonoles Feb 27 '17 at 19:07
  • 1
    Also, maybe the point I should have started with is that the Riemann zeta function is not injective, and therefore does not have an inverse everywhere. For example, there are many known values of $s$ for which $\zeta(s)=0$, so there is no way to define $\zeta^{-1}(0)$. – Zev Chonoles Feb 27 '17 at 19:09
  • @Zev Chonoles. We can find an inverse for the zeta function in the the interval $(1,\infty)$ right? – Star Alpha Aug 08 '21 at 07:27