[NT] more zeta stuff for the fans
#1
On mathstack the following conjecture about a dirichlet series was written.

https://math.stackexchange.com/questions...dot3-s-5-c

Since there are " zeta fans " here ...


Regards

tommy1729
#2
(02/15/2023, 12:32 AM)tommy1729 Wrote: On mathstack the following conjecture about a dirichlet series was written.

https://math.stackexchange.com/questions...dot3-s-5-c

Since there are " zeta fans " here ...


Regards

tommy1729

I don't think this is extendable beyond \(\Re(s) > 2\)--and if it were it would be very non-obvious. To ask for zeroes before analytically continuing is a recipe for disaster. The sum DOES NOT converge for \(1 < \Re(s) < 2\)--so to say there are zeroes here; when the sum diverges; without an analytic continuation is... nonsense. Shy

The standard way to analytically continue is useless here too. If I write:

\[
P(x) = \sum_{n\le x} p_n\\
\]

Then the tight bounds are:

\[
P(x) = \frac{x^2}{2} \left(\log(x) + \log \log(x) - 3/2 + o(1)\right)\\
\]

So from here we have:

\[
\zeta_P(s) = -s\int_1^\infty P(x)x^{-s-1}\,ds\\
\]

This integral converges for \(\Re(s) > 2\); but once you account for \(P(x)\)'s asymptotics--there's no obvious way to extend it. The error term \(o(1)\) decays too slowly-- about \(1/log^2 x\) to meaningfully allow for a simple continuation. Unless mick has a magical reflection formula up his sleeve, this is pretty pointless...
#3
It can still make sense if you consider the zero's for every truncated sum.

So zero's outside analytic continuation can make sense.

And a summability method for that integral might help.

I remember as a kid , I somewhat rediscovered the Riemann Hypothesis.

I was looking for zero's of the zeta function , being aware it had a pole.

At that time I did not know complex analysis or even got officially introduced to complex numbers.

I also missed the whole number theory thing.

So i just wanted to find zero's of a function that looked nice, without knowing analytic continuation , calculus , formal complex numbers or number theory.

And I did.

With the truncated sum for the zeta and some more tricks I rediscovered that the zero's were on the critical line.

The intuition I had was that every function apart from of the form exp(..) and gamma(..) had a zero.
This probably followed from rediscovering of the complex numbers.

Therefore I said all those functions had zero's.

I considered a function were the zero was hard to find and the function easy to define.
so ended with zeta.
probably inspired by exp because they did not have a 0.

So I tested my ideas and found critical zero's with 123lotus or whatever that predeccessor of excel was called.

I did not tell anyone because I thought I had something new and special lol.


regards

tommy1729
#4
(02/16/2023, 01:00 PM)tommy1729 Wrote: It can still make sense if you consider the zero's for every truncated sum.

So zero's outside analytic continuation can make sense.

And a summability method for that integral might help.

I remember as a kid , I somewhat rediscovered the Riemann Hypothesis.

I was looking for zero's of the zeta function , being aware it had a pole.

At that time I did not know complex analysis or even got officially introduced to complex numbers.

I also missed the whole number theory thing.

So i just wanted to find zero's of a function that looked nice, without knowing analytic continuation , calculus , formal complex numbers or number theory.

And I did.

With the truncated sum for the zeta and some more tricks I rediscovered that the zero's were on the critical line.

The intuition I had was that every function apart from of the form exp(..) and gamma(..) had a zero.
This probably followed from rediscovering of the complex numbers.

Therefore I said all those functions had zero's.

I considered a function were the zero was hard to find and the function easy to define.
so ended with zeta.
probably inspired by exp because they did not have a 0.

So I tested my ideas and found critical zero's with 123lotus or whatever that predeccessor of excel was called.

I did not tell anyone because I thought I had something new and special lol.


regards

tommy1729

Okay, Tommy

That's all well and good. But, sorry. That doesn't explain the situation at hand, not at all.

Guessing:

\[
\sum_{n=1}^M p(n)n^{-s} = A_M(s)\\
\]

And saying zeroes of \(A_M(s)\) are within \(1 \le \Re(s) \le 2\) has no statement on the zeroes of \(\lim_{M\to\infty} A_M(s) = A(s)\). Especially when your formula of \(A(s)\) only exists for \(\Re(s) > 2\). The zeroes as you observe \(M\to \infty\)--will obviously be in \(1 \le \Re(s) \le 2\); because \(A(s)\) is non-zero for \(\Re(s) > 2\)....

I'm sorry, Tommy; this doesn't make sense. None of it does.
#5
(02/16/2023, 05:07 AM)JmsNxn Wrote:
(02/15/2023, 12:32 AM)tommy1729 Wrote: On mathstack the following conjecture about a dirichlet series was written.

https://math.stackexchange.com/questions...dot3-s-5-c

Since there are " zeta fans " here ...


Regards

tommy1729

I don't think this is extendable beyond \(\Re(s) > 2\)--and if it were it would be very non-obvious. To ask for zeroes before analytically continuing is a recipe for disaster. The sum DOES NOT converge for \(1 < \Re(s) < 2\)--so to say there are zeroes here; when the sum diverges; without an analytic continuation is... nonsense. Shy

The standard way to analytically continue is useless here too. If I write:

\[
P(x) = \sum_{n\le x} p_n\\
\]

Then the tight bounds are:

\[
P(x) = \frac{x^2}{2} \left(\log(x) + \log \log(x) - 3/2 + o(1)\right)\\
\]

So from here we have:

\[
\zeta_P(s) = -s\int_1^\infty P(x)x^{-s-1}\,ds\\
\]

This integral converges for \(\Re(s) > 2\); but once you account for \(P(x)\)'s asymptotics--there's no obvious way to extend it. The error term \(o(1)\) decays too slowly-- about \(1/log^2 x\) to meaningfully allow for a simple continuation. Unless mick has a magical reflection formula up his sleeve, this is pretty pointless...

But that integral transform is designed not to have analytic continuation.

If P(x) = x^3 this would also not converge , yet its related zeta function does as it can be expressed by riemann zeta functions.



In fact if we write p_n as a taylor series 

p(n) = a_0 + a_1 n + a_2 n^2 + ...

Then we get expressions like

A(s) = zeta(s) + a_1 zeta(s-1) + a_2 zeta(s-2) + a_3 zeta(s-3) + ...

which might have analytic continuation then, since zeta does.


Notice this is not like an euler product or the prime zeta function.



And even if we have an natural boundary , that does not explain why the zero's approach the boundaries.


regards

tommy1729
#6
(02/20/2023, 12:38 AM)tommy1729 Wrote:
(02/16/2023, 05:07 AM)JmsNxn Wrote:
(02/15/2023, 12:32 AM)tommy1729 Wrote: On mathstack the following conjecture about a dirichlet series was written.

https://math.stackexchange.com/questions...dot3-s-5-c

Since there are " zeta fans " here ...


Regards

tommy1729

I don't think this is extendable beyond \(\Re(s) > 2\)--and if it were it would be very non-obvious. To ask for zeroes before analytically continuing is a recipe for disaster. The sum DOES NOT converge for \(1 < \Re(s) < 2\)--so to say there are zeroes here; when the sum diverges; without an analytic continuation is... nonsense. Shy

The standard way to analytically continue is useless here too. If I write:

\[
P(x) = \sum_{n\le x} p_n\\
\]

Then the tight bounds are:

\[
P(x) = \frac{x^2}{2} \left(\log(x) + \log \log(x) - 3/2 + o(1)\right)\\
\]

So from here we have:

\[
\zeta_P(s) = -s\int_1^\infty P(x)x^{-s-1}\,ds\\
\]

This integral converges for \(\Re(s) > 2\); but once you account for \(P(x)\)'s asymptotics--there's no obvious way to extend it. The error term \(o(1)\) decays too slowly-- about \(1/log^2 x\) to meaningfully allow for a simple continuation. Unless mick has a magical reflection formula up his sleeve, this is pretty pointless...

But that integral transform is designed not to have analytic continuation.

If P(x) = x^3 this would also not converge , yet its related zeta function does as it can be expressed by riemann zeta functions.



In fact if we write p_n as a taylor series 

p(n) = a_0 + a_1 n + a_2 n^2 + ...

Then we get expressions like

A(s) = zeta(s) + a_1 zeta(s-1) + a_2 zeta(s-2) + a_3 zeta(s-3) + ...

which might have analytic continuation then, since zeta does.


Notice this is not like an euler product or the prime zeta function.



And even if we have an natural boundary , that does not explain why the zero's approach the boundaries.


regards

tommy1729

I mean, I kind of see where your coming from.

But if \(|A_M(s) - A(s)| < \epsilon/2\) and \(|A(s)| > \epsilon\)--then:

\[
\begin{align}
|A(s)| &\le |A_M(s)| + |A_M(s) - A(s)|\\
|A_M|(s) &\ge \epsilon/2\\
\end{align}
\]

Where \(\epsilon\) only depends on \(M\) and \(\Re(s)>2\).



This is exactly what they teach you NOT TO DO! to solve the Riemann Hypothesis. If I take the sum:

\[
\zeta^M(s) = \sum_{n=1}^M n^{-s}\\
\]

The zeroes of this function coallesce near \(\Re(s) = 1\); because the sum goes to infinity and back near here. But it has no marker on the actual zeroes of \(\zeta(s)\). Because \(\zeta\) NEEDS TO BE analytically continued first... Without an analytic continuation; we can't speak on the zeroes. The partial sums tell us nothing about the zeroes; because the partial sums converge to infinity as \(M\to\infty\) in the critical strip--there is no domain of normality.
#7
(02/21/2023, 06:23 AM)JmsNxn Wrote:
(02/20/2023, 12:38 AM)tommy1729 Wrote:
(02/16/2023, 05:07 AM)JmsNxn Wrote:
(02/15/2023, 12:32 AM)tommy1729 Wrote: On mathstack the following conjecture about a dirichlet series was written.

https://math.stackexchange.com/questions...dot3-s-5-c

Since there are " zeta fans " here ...


Regards

tommy1729

I don't think this is extendable beyond \(\Re(s) > 2\)--and if it were it would be very non-obvious. To ask for zeroes before analytically continuing is a recipe for disaster. The sum DOES NOT converge for \(1 < \Re(s) < 2\)--so to say there are zeroes here; when the sum diverges; without an analytic continuation is... nonsense. Shy

The standard way to analytically continue is useless here too. If I write:

\[
P(x) = \sum_{n\le x} p_n\\
\]

Then the tight bounds are:

\[
P(x) = \frac{x^2}{2} \left(\log(x) + \log \log(x) - 3/2 + o(1)\right)\\
\]

So from here we have:

\[
\zeta_P(s) = -s\int_1^\infty P(x)x^{-s-1}\,ds\\
\]

This integral converges for \(\Re(s) > 2\); but once you account for \(P(x)\)'s asymptotics--there's no obvious way to extend it. The error term \(o(1)\) decays too slowly-- about \(1/log^2 x\) to meaningfully allow for a simple continuation. Unless mick has a magical reflection formula up his sleeve, this is pretty pointless...

But that integral transform is designed not to have analytic continuation.

If P(x) = x^3 this would also not converge , yet its related zeta function does as it can be expressed by riemann zeta functions.



In fact if we write p_n as a taylor series 

p(n) = a_0 + a_1 n + a_2 n^2 + ...

Then we get expressions like

A(s) = zeta(s) + a_1 zeta(s-1) + a_2 zeta(s-2) + a_3 zeta(s-3) + ...

which might have analytic continuation then, since zeta does.


Notice this is not like an euler product or the prime zeta function.



And even if we have an natural boundary , that does not explain why the zero's approach the boundaries.


regards

tommy1729

I mean, I kind of see where your coming from.

But if \(|A_M(s) - A(s)| < \epsilon/2\) and \(|A(s)| > \epsilon\)--then:

\[
\begin{align}
|A(s)| &\le |A_M(s)| + |A_M(s) - A(s)|\\
|A_M|(s) &\ge \epsilon/2\\
\end{align}
\]

Where \(\epsilon\) only depends on \(M\) and \(\Re(s)>2\).



This is exactly what they teach you NOT TO DO! to solve the Riemann Hypothesis. If I take the sum:

\[
\zeta^M(s) = \sum_{n=1}^M n^{-s}\\
\]

The zeroes of this function coallesce near \(\Re(s) = 1\); because the sum goes to infinity and back near here. But it has no marker on the actual zeroes of \(\zeta(s)\). Because \(\zeta\) NEEDS TO BE analytically continued first... Without an analytic continuation; we can't speak on the zeroes. The partial sums tell us nothing about the zeroes; because the partial sums converge to infinity as \(M\to\infty\) in the critical strip--there is no domain of normality.

Yes yes very good point.

And it might destroy that way of thinking for the riemann zeta.

But in general just because the partial sums converge to infinity locally does not mean we cannot detect zero's beyond it.
We may get some fake zero's and we might get only approximations of where the actual zero's are.

But with some luck it might still work.

I must say that the method i * actually * used was a bit more complex than the story.
Which probably explains why it worked.

Numerically it is very unstable because the strip converges to infinity almost everywhere , but maybe not at all the zero's.

So if we find a zero is it surrounded by large numbers going to infinity.
Which means derivatives going to infinity as well, resembling discontinu , chaos and divergeance in root finding algorithms etc.

and phenomena like double limits and numerical errors and such.

I remember that my algorithm gave 1/2 + 14.135 i as a zero but an overflow error at 0.49999999999 + 14.135 i.

It was crazy.

I remember the algorithm did not give zero's beyond a natural boundary.

I tested it on functions where the natural boundary was the unit circle like f(x)  = x^2 + x^2^2 + x^2^3 + x^2^4 + ...

Maybe I should look into those old methods again.

I have to give it more thought that is for sure, I guess we agree on that lol.
Im very busy at the moment though.

regards

tommy1729
#8
oh yeah i remember , i had some kind of error bound.

Guess that was equivalent to analytic continuation or summability methods in a hidden way then.

regards

tommy1729
#9
I did some thinking and calculations , and yeah it seems it is better to give up the original idea.

Some coincidences lead to wrong conclusions it seems.

But I think I can modify the idea to be more meaningful and get a new conjecture.

This might take a while though.


regards

tommy1729
#10
(02/21/2023, 11:52 PM)tommy1729 Wrote: I did some thinking and calculations , and yeah it seems it is better to give up the original idea.

Some coincidences lead to wrong conclusions it seems.

But I think I can modify the idea to be more meaningful and get a new conjecture.

This might take a while though.


regards

tommy1729

Ok at first it seems if we add a factor (-1)^n 

f(x) = 1 - 2/2^s + 3/3^s - 5/4^s + ...

, then we get no zeros with Re(s) > 5/2.

Not sure how (un)remarkable that is yet.


regards

tommy1729


Possibly Related Threads…
Thread Author Replies Views Last Post
  [special] binary partition zeta function tommy1729 1 653 02/27/2023, 01:23 PM
Last Post: tommy1729
  [NT] Caleb stuff , mick's MSE and tommy's diary functions tommy1729 0 460 02/26/2023, 08:37 PM
Last Post: tommy1729
  toy zeta function tommy1729 0 526 01/20/2023, 11:02 PM
Last Post: tommy1729
  Half-iterates and periodic stuff , my mod method [2019] tommy1729 0 3,193 09/09/2019, 10:55 PM
Last Post: tommy1729
  Some slog stuff tommy1729 15 31,789 05/14/2015, 09:25 PM
Last Post: tommy1729
  Zeta iterations Balarka Sen 11 26,398 02/26/2013, 09:49 AM
Last Post: Gottfried
  zeta and sinh tommy1729 0 4,462 05/30/2011, 12:07 PM
Last Post: tommy1729
  1960s stuff on extending tetration to real height? mike3 7 19,776 03/25/2010, 08:58 PM
Last Post: mike3



Users browsing this thread: 1 Guest(s)