It appears that the real part of Riemann zeta is related to the imaginary part by this formula:

$$\Re\left(\zeta \left(\frac{1}{2}+i t\right)\right)=\frac{\Im\left(\zeta \left(\frac{1}{2}+i t\right)\right) (\cos (2 \vartheta (t))+1)}{-\sin (2 \vartheta (t))}$$

and vice versa:

$$\Im\left(\zeta \left(\frac{1}{2}+i t\right)\right)=\frac{\Re\left(\zeta \left(\frac{1}{2}+i t\right)\right) (-\sin (2 \vartheta (t)))}{\cos (2 \vartheta (t))+1}$$

where $\vartheta (t)$ is the Riemann Siegel theta function.

How does one prove this formula above?

I guess a starting point should be these formulas for the real and imaginary parts of the Riemann zeta function:

$$\Re\left(\zeta \left(a+ib\right)\right)=\frac{1}{2} \left(\zeta \left(a+ib\right)+\zeta \left(a-ib\right)\right)$$

$$\Im\left(\zeta \left(a+ib\right)\right)=\frac{1}{2 i}\left(\zeta \left(a+ib\right)-\zeta \left(a-ib\right)\right)$$

What I would do is to look at the series expansion of:

$$-\frac{\sin (2 \vartheta (t))}{\cos (2 \vartheta (t))+1}$$

Normal[Series[(-Sin[RiemannSiegelTheta[t]*2])/(1 +
Cos[RiemannSiegelTheta[t]*2]), {t, 0, 1}]]
N[%, 20]

This gives: $$\frac{1}{2} t \left(\log (\pi )-\psi ^{(0)}\left(\frac{1}{4}\right)\right)$$ and numerically: $$2.6860917096128327911 t$$ where $\psi ^{(0)}\left(\frac{1}{4}\right)$ is the Polygamma function.

Then I would compare it to the series expansion of: $$\frac{\zeta \left(\frac{1}{2}+i t\right)-\zeta \left(\frac{1}{2}-i t\right)}{\frac{1}{2} (2 i) \left(\zeta \left(\frac{1}{2}+i t\right)+\zeta \left(\frac{1}{2}-i t\right)\right)}$$

which Mathematica:

Normal[Series[((Zeta[1/2 + I*t] - Zeta[1/2 - I*t])/(2*
       I))/((Zeta[1/2 + I*t] + Zeta[1/2 - I*t])/2), {t, 0, 1}]]
N[%, 20]

tells me is: $$\frac{t \zeta '\left(\frac{1}{2}\right)}{\zeta \left(\frac{1}{2}\right)}$$ and numerically: $$2.6860917096128327911 t$$

I then asked Mathematica to compare the two numerically equivalent expressions symbolically with the True Query command TrueQ:

TrueQ[1/2 t (Log[\[Pi]] - PolyGamma[0, 1/4]) == (
  t Derivative[1][Zeta][1/2])/Zeta[1/2]]

But the computer says no, Mathematica returns False although numerically the values so far are identical.

So is the real question to show that: $$\frac{1}{2} t \left(\log (\pi )-\psi ^{(0)}\left(\frac{1}{4}\right)\right)=\frac{t \zeta '\left(\frac{1}{2}\right)}{\zeta \left(\frac{1}{2}\right)}$$ ?

I know that: $$\log (\pi )=\frac{\zeta '(0)}{\zeta (0)}-\log (2)$$ but that won't take me far will it?


I misread the Mathematica manual at first and missed that this is an equality:

$$\frac{1}{2} t \left(\log (\pi )-\psi ^{(0)}\left(\frac{1}{4}\right)\right)=\frac{1}{2} t \left(\log (\pi )-\frac{\Gamma '\left(\frac{1}{4}\right)}{\Gamma \left(\frac{1}{4}\right)}\right)$$

Therefore the task is to show that: $$\frac{1}{2} t \left(\log (\pi )-\frac{\Gamma '\left(\frac{1}{4}\right)}{\Gamma \left(\frac{1}{4}\right)}\right)=\frac{t \zeta '\left(\frac{1}{2}\right)}{\zeta \left(\frac{1}{2}\right)}$$

TrueQ[1/2 t (Log[\[Pi]] - Gamma'[1/4]/Gamma[1/4]) == (
  t Derivative[1][Zeta][1/2])/Zeta[1/2]]

returns False.

N[1/2 t (Log[\[Pi]] - Gamma'[1/4]/Gamma[1/4]) == (
  t Derivative[1][Zeta][1/2])/Zeta[1/2], 20]

(2.6860917096128327911 t == 2.6860917096128327911 t)


Will you look at the definition of $Z(t)$ and $\vartheta(t)$ ?

$$\text{For } t \in \mathbb{R}, \ Z(t) \in \mathbb{R}, \qquad \zeta(\tfrac{1}2+it) = Z(t) e^{-i \vartheta(t)} $$ thus $$\Re(\zeta(\tfrac{1}2+it)) = Z(t) \cos(\vartheta(t)), \qquad Z(t) = \frac{\Re(\zeta(\frac{1}2+it))}{\cos(\vartheta(t))}$$ $$\Im(\zeta(\tfrac{1}2+it)) = - Z(t) \sin(\vartheta(t)) = -\frac{\Re(\zeta(\frac{1}2+it))}{\cos(\vartheta(t))}\sin(\vartheta(t)) =\frac{ -\Re(\zeta(\frac{1}2+it))\sin(2\vartheta(t))}{1+\cos(2\vartheta(t))}$$ where I used $$\frac{\sin(x)}{\cos(x)} =\frac{\cos(x)\sin(x)}{\cos(x)^2}=\frac{\sin(2x)}{1+\cos(2x)}$$


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.