Message Boards Message Boards

0
|
2934 Views
|
6 Replies
|
3 Total Likes
View groups...
Share
Share this post:

Wolfram Alpha Pro takes too long to solve a sum problem

Posted 1 year ago
6 Replies
Posted 1 year ago

Many thanks!

POSTED BY: Chris Bianchi

Well, this is not MMA-related - and really trivial: The sums are of the form:

$\sum_{k=0}a_{k+2}-\sum_{k=0}a_{k+1}$

which gives

$\sum_{k=2}a_{k}-\sum_{k=1}a_{k} = \sum_{k=2}a_{k}-\left(a_1+\sum_{k=2}a_{k}\right) = -a_1$

In this case we have

$a_k=\frac{\ln k}{2k} \quad\Longrightarrow\quad a_1=0$

EDIT: OK, after some thinking I guess in a strong mathematical sense my approach is not really correct, because effectively I am doing a term wise cancellation - but in infinite sums the commutative law does not hold ...

POSTED BY: Henrik Schachner

In fact, the two terms you input do not converge. So, if you enter the expression in that form, the program will meet two infinities and can't add them up.

In addition, if a limit $ \underset{n \rightarrow \infty}{\lim} a_n - b_n $ exists (converges), it doesn't mean series $\underset{n\rightarrow \infty}{\lim}a_n$ and $\underset{n\rightarrow \infty}{\lim}b_n$ both exist (converge).

The two $\infty$ in you expression may be different precisely. The expression you input is equal to $ \underset{n\rightarrow \infty}{\lim}\sum_{k=0}^{n} \frac{ln(k+2)}{2(k+2)} - \underset{m\rightarrow \infty}{\lim}\sum_{k=0}^{m}\frac{ln(k+1)}{2(k+1)} $

where two series both don't converge actually. So the sum of them don't converge.

But if you sum each term together like the reply before, you will get a series that converge and is easy to calculate the result.
$ \underset{n\rightarrow \infty}{\lim}(\sum_{k=0}^{n} \frac{ln(k+2)}{2(k+2)} - \sum_{k=0}^{n}\frac{ln(k+1)}{2(k+1)}) = \underset{n\rightarrow \infty}{\lim}\sum_{k=0}^{n}( \frac{ln(k+2)}{2(k+2)} - \frac{ln(k+1)}{2(k+1)}) $
Here the two $\infty$ are the same and we can add each term of two series up. The series above converge and is equal to 0.

Addition: the two infinite series are different actually. The first one doesn't converge but the second one converges to a finite value.

POSTED BY: Bowen Ping
Posted 1 year ago

Many thanks!

POSTED BY: Chris Bianchi
Posted 1 year ago

If you think very carefully then can you justify rewriting your problem as

Sum(Ln(k+2)/(2(k+2))-Ln(k+1)/(2(k+1)),{k,0,n})

Writing your problem in that form means WA will give you an answer in a faction of a second.

Then think very very carefully about what happens with that result as n becomes very very large.

Depending on how much background you have, you might be able to use WA to determine what that result will be as n goes to infinity.

POSTED BY: Bill Nelson
Posted 1 year ago

Many thanks Bill!

POSTED BY: Chris Bianchi
Reply to this discussion
Community posts can be styled and formatted using the Markdown syntax.
Reply Preview
Attachments
Remove
or Discard

Group Abstract Group Abstract