• Welcome to the new Internet Infidels Discussion Board, formerly Talk Freethought.

Infinite Sum

SLD

Contributor
Joined
Feb 25, 2001
Messages
5,138
Location
Birmingham, Alabama
Basic Beliefs
Freethinker
Take the sum of x from 1 to infinity of 1/x^n, can anyone prove that it always converges for all n > 1? It does not for n=1. Obviously it converges for n=2 and greater. By experiment it also converges for n=1.01. But will it converge for some n infinitesimally close to 1? And a proof of such.

666116A3-AD76-42EB-B5BE-38386375C4BD.jpeg
 
Take the sum of x from 1 to infinity of 1/x^n, can anyone prove that it always converges for all n > 1? It does not for n=1. Obviously it converges for n=2 and greater. By experiment it also converges for n=1.01. But will it converge for some n infinitesimally close to 1? And a proof of such.

View attachment 40116
Eventually when adding smaller and smaller numbers which all operations of this equation must, the numbers you add become too small to throw it off of convergence.

Any shrinkage in the process will cause later iterations to lose their significance over time, and when that significance is lost, convergence begins to happen.

It happens because for any constant nonzero amount of "loss of significance over time", eventually significance will move towards zero, so at "infinity", significance of additional iterations is going to be infinitely small.

The only way to prevent the loss of significance of interactions would be to shift the exponent over time.

The system of how much significance is lost, as N increases, asymptotically approaches 0.

It might be better to look at your equation from the standpoint of 1/(n^(x+1)), and map the asymptotes of the graph of this system of infinite sums.
 
Take the sum of x from 1 to infinity of 1/x^n, can anyone prove that it always converges for all n > 1? It does not for n=1. Obviously it converges for n=2 and greater. By experiment it also converges for n=1.01. But will it converge for some n infinitesimally close to 1? And a proof of such.

View attachment 40116
I'd have to look up how to test a series for convergence. Can't do it off the top of my head.

But limit as n > 1 from +inf (x^n) = x.

So in the limit it is sum 1/x 1 to inf. As x gets large the terms asymptotically go to zero so I'd think it converge but maybe not.

You can look up on the net how to test a series for convergence.
 
Proof of convergence for

\( \displaystyle{ \zeta(s) = \sum_{n=1}^\infty \frac{1}{n^s} } \)

for Re(s) > 1:

I will use that definition and not analytical continuations of that definition for Re(s) <= 1.

First off, ζ(s) > 0 for all s, supplying a lower bound. So we must look for an upper bound. Since this series diverges for Re(s) <= 0, we look at Re(s) > 0. Finding a lower bound for each n thus gives an upper bound for the series. We can also find a lower bound by finding an upper bound for each n, something that will improve our convergence estimate.

Lower bound for each n: (1) >= 1, (2, 3) >= 2, (4, 5, 6, 7) >= 4, ... giving 2k of n = 2k

Upper bound for each n: (1) <= 1, (2) <= 2, (3, 4) <= 4, (5, 6, 7, 8) <= 8, ... giving 1, then 2k of each n = 2k+1

Thus,

\( \displaystyle{ 1 + \sum_{k=0}^\infty \frac{2^k}{2^{s(k+1)}} < \zeta(s) < \sum_{k=0}^\infty \frac{2^k}{2^{sk}} } \)

Both bounds are geometric series:

\( \displaystyle{ 1 + \sum_{k=0}^\infty \frac{2^k}{2^{s(k+1)}} < \zeta(s) < \sum_{k=0}^\infty \frac{2^k}{2^{sk}} } \)

reducing to

\( \displaystyle{ 1 + 2^{-s} \sum_{k=0}^\infty 2^{-(s-1)k} < \zeta(s) < \sum_{k=0}^\infty 2^{-(s-1)k}} \)

This means that the series form of ζ(s) converges for Re(s) > 1 and diverges for Re(s) <= 1.

There's a lot more at  Riemann zeta function
 
  • Like
Reactions: SLD
Ipetrich,

Are you saying 1/(x^1) does not converge?
 
Ipetrich,

Are you saying 1/(x^1) does not converge?
I'm guessing that you mean

\( \displaystyle{ \zeta(1) = \sum_{n=1}^\infty \frac{1}{n} } \)

That series does not converge. In fact, partial sums have a limit:

\( \displaystyle{ \sum_{n=1}^\N \frac{1}{n} = \log N + \gamma + O(1/N) } \)

where γ = 0.577... is the Euler-Mascheroni constant.
 
The zeta function, as it's called, can be extended beyond Re(s) > 1 by analytic continuation.

It can be expressed as an integral:

\( \displaystyle{ \zeta(s) = \frac{1}{\Gamma(s)} \int_0^\infty \frac{x^{s-1}}{e^x - 1} dx } \)

One can extend convergence to Re(s) > 0 with some rearrangement:

\( \displaystyle{ \zeta(s) = \frac{1}{1-2^{1-s}} \sum_{n=1}^\infty \frac{(-1)^{n+1}}{n^s} = \frac{1}{1-2^{1-s}} \frac{1}{\Gamma(s)} \int_0^\infty \frac{x^{s-1}}{e^x + 1} dx } \)

where Γ(x) is a generalization of the factorial function: Γ(x) = (x-1)! for positive integer x.

For s near 1,

\( \displaystyle{ \zeta(s) = \frac{1}{s-1} + \gamma + \dots } \)
 
One can define convergence of a sequence without reference to the limit that a sequence converges to. A convergent sequence defined in this way is a "Cauchy sequence", named after early 19th cy. French mathematician Augustin-Louis Cauchy.

First, when a series has a limit. Consider a sequence a(n) for n = 1, 2, 3, ... with limit a(lim). Then for every e > 0, there is some N such that |a(n) - a(lim)| < e for all n > N.

Now the Cauchy definition. For every e > 0, there is some N such that |a(n) - a(m)| < e for all n, m > N.

The nice thing about the Cauchy definition is that one defined convergence without reference to the sequence limit, because the limit may not be a member of some set that the sequence members are defined in.

Consider these sequences of positive numbers: 1, 1/2, 1/3, 1/4, 1/5, ... and 1, 1/2, 1/4, 1/8, 1/16, ...

They both have limit 0, and 0 is not a positive number.

Real numbers can be defined in terms of rational numbers as the limits of all Cauchy sequences of rational numbers. Real numbers are also the limits of all Cauchy sequences of real numbers, making them Cauchy-closed.
 
I guess I did not fully comprehend the meaning of convergence and divergence.

I took convergence to mean asymptotically approaching a value as the series goes to infinity. I may be confusing limits with convergence and divergence.

From this link 1/x flattens out, but the area goes to infinity, diverges.


Learn something new every day....
That’s right. If you integrate 1/x from 1 to infinity You get infinity. If some one said for you to paint the wall of 1/x from 1 to infinity, you might say it can’t be done. But you’d be wrong. What is the volume of the bounded trumpet formed by rotating 1/x around the x axis? The answer is pi! Try it. It’s a simple integration. So all you have to do is create a circular barrier around the function and fill it with pi units of paint and voila, you’ve painted an infinite area wall!
 
I guess I did not fully comprehend the meaning of convergence and divergence.

I took convergence to mean asymptotically approaching a value as the series goes to infinity. I may be confusing limits with convergence and divergence.

From this link 1/x flattens out, but the area goes to infinity, diverges.


Learn something new every day....
That’s right. If you integrate 1/x from 1 to infinity You get infinity. If some one said for you to paint the wall of 1/x from 1 to infinity, you might say it can’t be done. But you’d be wrong. What is the volume of the bounded trumpet formed by rotating 1/x around the x axis? The answer is pi! Try it. It’s a simple integration. So all you have to do is create a circular barrier around the function and fill it with pi units of paint and voila,

I think I get it now:D

Sonids le you knew the answer before you asked the question.
 
I guess I did not fully comprehend the meaning of convergence and divergence.

I took convergence to mean asymptotically approaching a value as the series goes to infinity. I may be confusing limits with convergence and divergence.

From this link 1/x flattens out, but the area goes to infinity, diverges.


Learn something new every day....
That’s right. If you integrate 1/x from 1 to infinity You get infinity. If some one said for you to paint the wall of 1/x from 1 to infinity, you might say it can’t be done. But you’d be wrong. What is the volume of the bounded trumpet formed by rotating 1/x around the x axis? The answer is pi! Try it. It’s a simple integration. So all you have to do is create a circular barrier around the function and fill it with pi units of paint and voila,

I think I get it now:D

Sonids le you knew the answer before you asked the question.
Well it seemed intuitively obvious, but I didn’t know what the proof was. Just experimentally it seems to approach infinity as n goes to 1. I knew 1 didn’t converge. The series gets infinitesimally closer to 1 the higher the number. But all even n take the from of pi^n/(natural number).

18D842B4-F6D1-4781-B542-9C683943A3F5.jpeg
0FE486B2-598D-47D7-ABC3-F2D6A3207C12.jpeg
5052C71A-99C1-4606-8443-BAD68219F824.jpeg

But odd numbers don’t show such a relationship.
 
When I was confronted by a work related problem It was more expedient to just simulate it.
 
That’s right. If you integrate 1/x from 1 to infinity You get infinity. If some one said for you to paint the wall of 1/x from 1 to infinity, you might say it can’t be done. But you’d be wrong. What is the volume of the bounded trumpet formed by rotating 1/x around the x axis? The answer is pi! Try it. It’s a simple integration. So all you have to do is create a circular barrier around the function and fill it with pi units of paint and voila, you’ve painted an infinite area wall!
Mathematically yes, but the engineering side in me is screaming "Gabriel's horn becomes narrower than a paint molecule's maximum dimension so you cannot fill it with paint." :p
 
That’s right. If you integrate 1/x from 1 to infinity You get infinity. If some one said for you to paint the wall of 1/x from 1 to infinity, you might say it can’t be done. But you’d be wrong. What is the volume of the bounded trumpet formed by rotating 1/x around the x axis? The answer is pi! Try it. It’s a simple integration. So all you have to do is create a circular barrier around the function and fill it with pi units of paint and voila, you’ve painted an infinite area wall!
Mathematically yes, but the engineering side in me is screaming "Gabriel's horn becomes narrower than a paint molecule's maximum dimension so you cannot fill it with paint." :p
Maybe that's why it rings so loudly, on account of the molecule being compressed infinitely.

Gabriel as infinite blowhard?
 
That’s right. If you integrate 1/x from 1 to infinity You get infinity. If some one said for you to paint the wall of 1/x from 1 to infinity, you might say it can’t be done. But you’d be wrong. What is the volume of the bounded trumpet formed by rotating 1/x around the x axis? The answer is pi! Try it. It’s a simple integration. So all you have to do is create a circular barrier around the function and fill it with pi units of paint and voila, you’ve painted an infinite area wall!
Mathematically yes, but the engineering side in me is screaming "Gabriel's horn becomes narrower than a paint molecule's maximum dimension so you cannot fill it with paint." :p
F$&#$king engineers. You have ruined mathematics! :mad:
 
SLD posted on values for even powers. In general,

\( \displaystyle{ \zeta(2n) = \frac{ (-1)^{n+1} B_{2n} (2\pi)^{2n} }{ 2 (2n)! } } \)

where B(2n) is a "Bernoulli number", with this generating function:

\( \displaystyle{ \frac{t}{e^t - 1} = \sum_{n=0}^\infty \frac{B_n t^n}{n!} } \)

For all nonnegative integers n, B(n) is a rational number. For all odd n > 1, B(n) = 0. There are two conventions for the value of B(1): -1/2 and +1/2, and I've used the first one.

From the generating function, one can derive a sum formula,

\( \displaystyle{ \sum_{k=0}^n \binom{n+1}{k} B_k = \delta_{n,0} } \)

giving recurrence

\( \displaystyle{ B_n = \delta_{n,0} - \sum_{k=0}^{n-1} \binom{n}{k} \frac{B_k}{n-k+1} } \)

Bernoulli numbers are all rational numbers, and the zeta function for an even positive arg is (some rational number) * (pi)^(arg)
 
For n > 1, the denominators of B(2n) all divide 2*(2n+1)!! where (2n+1)!! = 1*3*5*...*(2n+1).

Relatives of the  Riemann zeta function
\( \displaystyle{ \zeta(s) = \sum_{n=1}^\infty \frac{1}{n^s} } \)

 Dirichlet eta function
\( \displaystyle{ \eta(s) = \sum_{n=1}^\infty \frac{(-1)^{n+1}}{n^s} = (1 - 2^{1-s}) \zeta(s) } \)

Dirichlet lambda function -  Lambda function
\( \displaystyle{ \lambda(s) = \sum_{n=1}^\infty \frac{1}{(2n+1)^s} = (1 - 2^{-s}) \zeta(s) } \)

 Dirichlet beta function
\( \displaystyle{ \beta(s) = \sum_{n=1}^\infty \frac{(-1)^{n+1}}{(2n+1)^s} } \)

I first learned about these functions long ago, from Abramowitz and Stegun: Handbook of Mathematical Functions (scanned print version)

[Edited as per request]
 
Last edited by a moderator:
Necessity is the mother of invention.

Mathematics came about as an aid to engineering.

Probability theory arose as an aid to gain an advantage at gambling.
 
Back
Top Bottom