# Limit on infinity

1. Apr 5, 2013

### mathsciguy

I have a question about limits at infinity, particularly, about a limit I have seen in the context of infinite series convergence.

Let's say we have an infinite series where the the sequence of partial sums is given by {S(n)} and also, it is convergent and the sum is equal to S. Then we know that the limit of S(n) as n approaches infinity is S, but from this, can we also say the the limit of S(n+1) is equal to S?

Well, based from the textbook I was reading, it is. I actually have an intuitive idea of why, but I'd rather see something more of a 'formal' explanation, probably something which has epsilons. I tried verifying with an epsilon-N proof but I'm not that confident.

2. Apr 5, 2013

### Staff: Mentor

Yes. Both limits result in S.
It sounds like you've made a good start. People who are new to this try to do things with a $\delta\ \epsilon$ argument. If you can show that for a given $\epsilon$ a number N can be found so that if n ≥ N, then |Sn - S| < $\epsilon$, you're set. Keep in mind that Sn+1 is the next number in the sequence. As long as the index is past the magic number N, every term in the sequence satisfies the inequality above.

3. Apr 5, 2013

### mathsciguy

Hm, well I could think that obviously if n≥N then this implies n+1≥N (if N>0 and n is restricted to positive integers). Does that mean I can say: if for any ε>0, and N>0 such that |S(n)-S|<ε whenever n≥N is true by hypothesis, then |S(n+1)-S|<ε whenever n+1≥N is also true?

The catch is I've used both N for S(n) and S(n+1).

Last edited: Apr 5, 2013
4. Apr 5, 2013

### Staff: Mentor

It's simpler than that: If n ≥ N, then for sure n+1 ≥ N. So Sn is within $\epsilon$ of S, Sn+1 is within $\epsilon$ of S, Sn+2 is within $\epsilon$ of S, ...

5. Apr 5, 2013

### mathsciguy

I think that's same things as what I've said. Haha, sometimes I have a hard time wrapping my head around stuff like this. Thanks anyway.

Last edited: Apr 5, 2013
6. Apr 5, 2013

### Bacle2

Maybe this summarizes the ideas here:
You want to know :

Limn→oo[S(n+1)-S(n)]

But, since the series converges, the sequence of partial sums is Cauchy, so that
the difference in the parentheses goes to zero, i.e., the limit is 0. If you show quickly that both limits exist (more precisely, you can show that if S(n) exists as n→∞ , then S(n+1) also exists). Then you can say:

Limn→oo[S(n+1)-S(n)]=0 . Since both the limits exist, the limit distributes and

Limn→∞S(n)=Limn→∞S(n+1)

Last edited: Apr 5, 2013
7. Apr 21, 2013

### mathsciguy

What if the limit of S(n) as n approaches infinity is known to be S, I've read that the function S(n-1) as n approaches infinity is also S. How do we verify this? My idea is that, after finding an N such that: |S(n)-S|<ε whenever n>N; and we adjust ε so that we find a 1<N.

So that: |S(n-1)-S|<ε' whenever (n-1)>(N-1).

8. Apr 21, 2013

### Bacle2

Basically, |S_n- S|<e/2 (S_n converges) , and |S_(n-1)-S_n|<e/2 ( convergence implies

Cauchy) and then the triangle ineq.

kicks-in :

You can use that every convergent sequence is Cauchy, which means that for every e>0 there

is a positive N with |S_n-S_m| <e for all n,m>N . Choose, then, some N > n for any e>0

so that |S_n-S_(n-1)| <e . Then |S_(n-1) -S |=

|S_(n-1) -S_(n)+S_(n)-S | ≤ .....(triangle ineq.)

Last edited: Apr 21, 2013