Reason for Gram's Law about Zeta Zeros

  • Level: Undergrad 
  • Thread starter Swamp Thing
  • Start date
  • Tags Tags
    Law Reason
4 replies · 2K views · last reply
I suppose you could phrase Gram's law as the statement that ##\Re(\zeta(\frac{1}{2}+it))## tends to be positive, or that ##Z(t)## is somehow positively correlated with ##\cos\theta(t)##. Although Gram's law only applies when ##\zeta(\frac{1}{2}+it)## is real, it seems to hold elsewhere on the critical line as well. The plot of ##\Re(\zeta(\frac{1}{2}+it))## is mostly positive, whereas the imaginary part appears to fluctuate rapidly:

Line graph of a rapidly oscillating function against t from 0 to about 70, with peaks reaching around 4 and dips below zero, with a shaded area near the start.
Line plot of a real-valued oscillating function against an unlabelled horizontal axis, with tick labels 20, 40, 60 and vertical values from -2 to 2; the curve oscillates around zero with growing amplitude and a few large peaks near 60.


In particular, ##\int_0^T\cos\theta(t)\cdot Z(t)\,dt## is positive and large, while ##\int_0^T\sin\theta(t)\cdot Z(t)\,dt## is relatively small and might be positive or negative.

Actually, apparently there is a formula ##\sum_{n\leq N} Z(g_{n-1})Z(g_n) \sim -2(\gamma + 1)N##, which I found a few pages into this paper. I plotted the autocorrelation function of ##Z(g_n)## (integrating from 0 to 100,000) using the approximate formula ##g_n=2\pi e^{1+W(\frac{8n+1}{8e})}##:

Normalized plot of ##f(\tau)=\int_0^{10^5} Z(g(t))Z(g(t+\tau))\,dt##:
Line graph of an oscillating function over x from -10 to 10, with y-axis ticks from -0.2 to 0.6; peaks near 0.6 at x about 0 and near 0.2 elsewhere, with dips down to about -0.24.

I don't really have a full understanding of what I'm doing, but this is all very intriguing to me.
 
  • Like
Likes   Reactions: Swamp Thing
Thank you.. I'm reading now on a mobile device but I'll probably get back later with a question or two.
 
suremarc said:
...
I plotted the autocorrelation function of ##Z(g_n)## (integrating from 0 to 100,000) using the approximate formula ##g_n=2\pi e^{1+W(\frac{8n+1}{8e})}##:
...

What is the "W" in this formula?
 
Swamp Thing said:
What is the "W" in this formula?
That's the Lambert W function. Nothing special, it's just there since ##\theta(t)\sim \frac{t}{2}\log\frac{t}{2\pi}## for ##t## large, and so its inverse function can be approximated using W.