Intereting Posts

Countability of boundary points
Supremum of an union of bounded sets
Does $\mathbb{R}^\mathbb{R}$ have a basis?
Backwards epsilon
Show that an infinite number of triangles can be inscribed in either of the parabolas $y^2=4ax$ and $x^2=4by$ whose sides touch the other parabola.
Sheafification: Show that $\tilde{\mathscr{F}_x}=\mathscr{F}_x$.
Direct proof of empty set being subset of every set
If $n$ is a positive integer greater than 1 such that $3n+1$ is perfect square, then show that $n+1$ is the sum of three perfect squares.
Notation problem in integration: write $dx$ or ${\mathrm{d}}x$?
What is the difference between matrix theory and linear algebra?
Are there statements that are undecidable but not provably undecidable
Is there a simple, constructive, 1-1 mapping between the reals and the irrationals?
Matrix reciprocal positive to prove $λ_{max}=n$
A tiling puzzle/question
Prove that $X'$ is a Banach space

I have very little understanding on how complex functions work but was wondering if someone could show what the summation of the zeta function simplifies to when $s$ is the first non-trivial zero of the Riemann zeta function, $\zeta(s)$.

In other words, show how one gets 0 when plugging in the first non-trivial zero of the zeta function into the zeta function.

- Proof of an inequality about $\frac{1}{z} + \sum_{n=1}^{\infty}\frac{2z}{z^2 - n^2}$
- How to choose a proper contour for a contour integral?
- Let $A_1,A_2,..,A_n$ be the vertices of n sides of a regular polygon such that $1/A_1.1/A_2=1/A_1.1/A_3+1/A_1.1/A_4$ then value of $n$ must be?
- How to prove $\prod _{a=0}^{9}\prod_{b=0}^{100}\prod_{c=0}^{100}(w^a+z^b+z^c)\equiv?\pmod {101}$
- Solve $z^6=(z-1)^6$.
- Complex power of a complex number

- Why this proof $0=1$ is wrong?(breakfast joke)
- Deriving an expression for $\cos^4 x + \sin^4 x$
- Closed form for $ S(m) = \sum_{n=1}^\infty \frac{2^n \cdot n^m}{\binom{2n}n} $ for integer $m$?
- Analytic Continuation of Zeta Function using Bernoulli Numbers
- Are these zeros equal to the imaginary parts of the Riemann zeta zeros?
- Find $\sin\frac{\pi}{3}+\frac{1}{2}\sin\frac{2\pi}{3}+\frac{1}{3}\sin\frac{3\pi}{3}+\cdots$
- Residues and poles proof
- Complex number trigonometry problem
- Hard Definite integral involving the Zeta function
- Am I wrong in thinking that $e^{i \pi} = -1$ is hardly remarkable?

The series $\zeta(s)=\sum_{n=1}^{\infty}n^{-s}$ is only valid for $Re (s)>1$, so it is not possible to use this series at $s=\rho$, where $\rho=\frac{1}{2}+it_0$ is a non-trivial zero. We need to use an analytic continuation, see What is the analytic continuation of the Riemann Zeta Function. One way, for $Re(s)>0$, except for $s=1$, is

$$

\zeta(s)=\frac{1}{1-2^{1-s}}\sum_{n=1}^{\infty}\frac{(-1)^{n-1}}{n^s},

$$

see Zeta function zeros and analytic continuation.

*[update: see updated image/consideration at the end]*

The convergence of the above summation, using the alternating zeta (or “Dirichlet’s eta”), based on the terms of the series can be much improved using Euler-summation to some order. For my own exercise with this I’ve made an excel-file to experiment with it and I show here a couple of pictures which illustrate the acceleration-power of Euler-summation, when even its order can be optimally adapted to fractional or complex orders.

With Euler-summation/-acceleration in principle one calculates the finitely truncated Dirichlet-series with some weighting coefficients $e_n(o,t)$ which are determined by the order of the Euler-summation. This can approximate the final result much better than the “unaccelerated” series with the same number of terms.

Call the number of used terms *t* then we get the approximation

$$ \eta(z)=\lim_{t\to \infty} \sum_{n=1}^t e_n(o,t) {(-1)^{n-1}\over n^z} \tag 1$$ where the $e_n(o,t)$ are the coefficients by the Euler-summation-procedure of order $o$ for number-of-terms $t$. Then even at the first nontrivial root $z=\rho_0$ we need few terms of the partial sums to “see/extrapolate the result” .

For the trajectories of the partial sums $s_n$ we simply compute the above (1) with the same order $o$ to increasing number $t$ of terms :

$$ s_t=\sum_{n=1}^t e_n(o,t) {(-1)^{n-1}\over n^z} \tag 2$$

and the trajectory is then given by the sequence $s_1,s_2,s_3,…,s_{48}$

Here is the picture of the trajectory of the partial sums in the complex plane *without* any Euler-summation (or $o=0$) first. The trajectory begins at $s_1=1+0i$ which is the partial sum using only the first term, then proceeds along the blue line from point to point to arrive eventually at $s_\infty=0+0i$ *(update: upps, I see I’ve the argument in the pictures the letter $s$ which should be $z$ for this discussion here because $s$ denotes here the partial sums, sorry)*

We see the spiralling around the expected final value of $0+0i$ and we need hundreds of terms to be near to only three or four decimal digits.

Euler-summation with the weighting coefficients can improve this image dramatically.

Here is the same computation, but with a small Euler-order $o=0.1$ and number of terms $t=48$:

Obviously the convergence to the final value appears much faster and even smoother.

A higher order of Eulersummation ($o=0.5$) improves this further:

This gives even the idea, that the spiralling may be defeated!

Here is the detail around the origin, where I rescaled the absolute distance to the origin logarithmically so we see, that the spiralling even seem to stop after only *32* or so terms of the partial sums (but this might be a numerical artifact here).

The standard Eulersummation uses order $o=1$ and this gives this impression:

which does not give much improvement (if at all). In the logarithmic rescale around the origin we have no more artifact but can still observe that the arclengthes per step seem to decrease.

Finally, it seems that there is an “optimal” Eulerorder, but which is even complex. I’ve the impression, that $o=0.5 – 0.3i$ is in some sense optimal: the final steps of the *48*-term-trajectory seem to approach the final value in a nearly linear curve:

and the detail around the origin (here just zoomed in) looks very promising:

Similarly this can be used to work with the summation for the nontrivial roots with larger imaginary value – and as well for arguments $\eta(z)$ with smaller (and even negative) real part.

Here is an overlay of the numerical more precise computation focusing the angular and the absolute deviation from the origin in rectancular, but logarithmically scaled terms, for the $o=1/2$ and $o=1/2 – 0.3 î$ orders. We see, that both orders approximate similarly in terms of absolute deviation (radii decrease very similarly) but with the $o=0.5$-order the sequence of difference of angles seems to decrease nicely:

- If every real-valued continuous function is bounded on $X$ (metric space), then $X$ is compact.
- On $\int_0^1\arctan\,_6F_5\left(\frac17,\frac27,\frac37,\frac47,\frac57,\frac67;\,\frac26,\frac36,\frac46,\frac56,\frac76;\frac{n}{6^6}\,x\right)\,dx$
- Revisted$_2$: Are doubling and squaring well-defined on $\mathbb{R}/\mathbb{Z}$?
- Linear function between finite-dimensional vectorspaces is continuous
- The number of ones in a binary representation of an integer
- How to prove Bonferroni inequalities?
- Convergence of $\int_0^\infty \sin(t)/t^\gamma \mathrm{d}t$
- Description of $\mathrm{Ext}^1(R/I,R/J)$
- Determining the rank of a matrix based on its minors
- Show that a matrix $A$ is singular if and only if $0$ is an eigenvalue.
- Multivariable calculus – Implicit function theorem
- roots of a polynomial inside a circle
- $f, g$ entire functions with $f^2 + g^2 \equiv 1 \implies \exists h $ entire with $f(z) = \cos(h(z))$ and $g(z) = \sin(h(z))$
- Formulas for the (top) coefficients of the characteristic polynomial of a matrix
- Geometric proof for inequality