Do we distinguish cases for n?

  • Context: MHB 
  • Thread starter Thread starter evinda
  • Start date Start date
Click For Summary

Discussion Overview

The discussion revolves around the bounds of the set defined by the sum of a geometric series with a variable parameter $\alpha$, specifically focusing on the cases when $-1 < \alpha < 0$ and $0 < \alpha < 1$. Participants explore the conditions under which the least upper bound can be determined and whether different cases for $n$ affect the bounds.

Discussion Character

  • Exploratory, Technical explanation, Debate/contested, Mathematical reasoning

Main Points Raised

  • One participant presents a formula for the sum of a geometric series and seeks to establish bounds for the set when $-1 < \alpha < 0$.
  • Another participant suggests using specific series as lower and upper bounds but questions the generality of these bounds.
  • A participant argues that $1/(1 - \alpha)$ cannot be the least upper bound for $-1 < \alpha < 0$ and proposes a different approach using $\beta = -\alpha$.
  • It is noted that the upper bounds differ based on whether $n$ is odd or even, leading to a discussion on the implications of this distinction.
  • Participants engage in clarifying the conditions under which certain inequalities hold, particularly for small values of $n$.
  • One participant attempts to prove that a specific value is the least upper bound by assuming the existence of a smaller upper bound and deriving contradictions.
  • There is a discussion about the nature of upper bounds in relation to the set and the conditions that define them.
  • Questions arise regarding the correctness of the lower and upper bounds established for the case when $0 < \alpha < 1$.

Areas of Agreement / Disagreement

Participants express differing views on the bounds of the set, particularly concerning the cases for $n$ and the implications of $\alpha$ being negative. There is no consensus on the least upper bound when $-1 < \alpha < 0$, and the discussion remains unresolved regarding the lower bounds for this case.

Contextual Notes

Limitations include the dependence on the specific values of $n$ and the assumptions made about $\alpha$. The discussion highlights the complexity of establishing bounds based on the parity of $n$ and the behavior of the series as $\alpha$ varies.

evinda
Gold Member
MHB
Messages
3,741
Reaction score
0
Hello! (Wave)

Let $\alpha \neq 1$. I have shown that

$$1+\alpha+ \dots+ \alpha^k+\dots+ \alpha^n=\frac{1-\alpha^{n+1}}{1-\alpha}.$$

Now I want to show that, when $0<|\alpha|<1$, then the set $\{1+ \alpha+ \dots+ \alpha^k+\dots+ \alpha^n \mid n \in \mathbb{N} \}$ is bounded. And I want to find its least upper bound.

I have shown that when $0< \alpha<1$, then the lower bound of the set is $0$ and the least upper bound is $\frac{1}{1-\alpha}$.

So it suffices to find the upper and lower bound of the set when $-1< \alpha<0$.

In this case, we cannot find a general inequality for $\alpha^{n+1}$. Do we distinguish cases for $n$ ? (Thinking)
 
Physics news on Phys.org
evinda said:
In this case, we cannot find a general inequality for $\alpha^{n+1}$.

Hey evinda! (Smile)

Can't we use $-1-|\alpha|-|\alpha|^2-...$ and $1+|\alpha|+|\alpha|^2+...$ as lower and upper bounds? (Wondering)
 
I like Serena said:
Hey evinda! (Smile)

Can't we use $-1-|\alpha|-|\alpha|^2-...$ and $1+|\alpha|+|\alpha|^2+...$ as lower and upper bounds? (Wondering)

Ah I see... But can we also find a least upper bound for the general case? (Thinking)
 
Note $1/(1 - \alpha)$ cannot be the least upper bound when $-1 < \alpha < 0$. For if $\alpha = -1/2$, $1/(1 - \alpha) = 2/3$, but $1 + \alpha = 3/2 > 2/3$.

To handle the negative case, let $\beta = -\alpha$ so that $0 < \beta < 1$. Then

$$1 + \alpha + \cdots + \alpha^n = \begin{cases}\dfrac{1 - \beta^{n+1}}{1+\beta}&\text{if $n$ is odd}\\\dfrac{1 + \beta^{n+1}}{1 + \beta}&\text{if $n$ is even}\\\end{cases}$$

Since $0 < \beta < 1$, in the both cases above, the sum on the left-hand side is bounded above by $\sigma := (1 + \beta^3)/(1 + \beta) = (1 -\alpha^3)/(1 - \alpha)$. Thus, $\sigma$ is an upper bound for your set, call it $S$. Since $1 + \alpha + \alpha^2 = (1 - \alpha^3)/(1 - \alpha) = \sigma$, then $\sigma$ is in fact the least upper bound of $S$.
 
Euge said:
Note $1/(1 - \alpha)$ cannot be the least upper bound when $-1 < \alpha < 0$. For if $\alpha = -1/2$, $1/(1 - \alpha) = 2/3$, but $1 + \alpha = 3/2 > 2/3$.

To handle the negative case, let $\beta = -\alpha$ so that $0 < \beta < 1$. Then

$$1 + \alpha + \cdots + \alpha^n = \begin{cases}\dfrac{1 - \beta^{n+1}}{1+\beta}&\text{if $n$ is odd}\\\dfrac{1 + \beta^{n+1}}{1 + \beta}&\text{if $n$ is even}\\\end{cases}$$

Since $0 < \beta < 1$, in the both cases above, the sum on the left-hand side is bounded above by $\sigma := (1 + \beta^3)/(1 + \beta) = (1 -\alpha^3)/(1 - \alpha)$. Thus, $\sigma$ is an upper bound for your set, call it $S$. Since $1 + \alpha + \alpha^2 = (1 - \alpha^3)/(1 - \alpha) = \sigma$, then $\sigma$ is in fact the least upper bound of $S$.

How did you find these upper bounds?
I found that the upper bound is $\frac{1}{1+\beta}$ when $n$ is odd, and $\frac{2}{1+\beta}$ when $n$ is even. (Thinking)

I found the upper bounds as follows.We have that $\beta>0$. So $1-\beta^{n+1} <1 \Rightarrow \frac{1-\beta^{n+1}}{1+\beta}<\frac{1}{1+\beta}$.

We also have that $\beta<1 \Rightarrow \beta^{n+1}<1$. So $\frac{1+\beta^{n+1}}{1+\beta}<\frac{2}{1+\beta}$.
 
First, let me make a correction above: In my example where $\alpha = -1/2$, I meant to say $1/(1 - \alpha) = 2/3$, but $1 + \alpha + \alpha^2 = 1 - 1/2 + 1/4 = 3/4 > 2/3$.

Now since $0 < \beta < 1$, then when $n$ is even, $\beta^{n+1} \le \beta^{2 + 1} = \beta^3$, so then $(1 + \beta^{n+1})/(1 + \beta) \le (1 + \beta^3)/(1 + \beta)$ for even $n$. Since $\beta > 0$, then for all $n$, $(1 - \beta^{n+1})/(1 + \beta) < (1 + \beta^3)/(1 + \beta)$. This is how I obtain $\sigma := (1 + \beta^3)/(1 + \beta) = (1 - \alpha^3)/(1 - \alpha)$ as an upper bound for $S$. Since $\sigma \in S$, then $\sigma$ is the least upper bound for $S$.

Note that your bounds are not the upper bounds, as there are infinitely many upper bounds in the even and odd cases (in fact, there are infinitely many upper bounds for any nonempty bounded subset of the reals).
 
Euge said:
Now since $0 < \beta < 1$, then when $n$ is even, $\beta^{n+1} \le \beta^{2 + 1} = \beta^3$, so then $(1 + \beta^{n+1})/(1 + \beta) \le (1 + \beta^3)/(1 + \beta)$ for even $n$.

This holds for $n \geq 3$. It does not hold for $n=1$, right?
 
evinda said:
This holds for $n \geq 3$. It does not hold for $n=1$, right?

n=1 isn't even is it?
And doesn't it hold for n=2? (Wondering)
 
I like Serena said:
n=1 isn't even is it?
And doesn't it hold for n=2? (Wondering)

Oh yes, you are right.

To show that $\frac{1+\beta^3}{1+\beta}$ is indeed the least upper bound, I supposed that there is an $\epsilon>0$ such that $\frac{1+\beta^3}{1+\beta}-\epsilon$ is a smaller upper bound.

When $n$ is even, then we will have that $\frac{1+\beta^{n+1}}{1+\beta}<\frac{1+\beta^3}{1+\beta}-\epsilon, \forall n \in \mathbb{N}$.

Then we get that $n<\frac{\ln{(\beta^3-\epsilon (1+\beta))}}{\ln{(\beta)}}-1$, which is a contradiction.

When $n$ is odd, from $\frac{1-\beta^{n+1}}{1+\beta}< \frac{1+\beta^3}{1+\beta}$ we get that $\beta^{n+1}>-\beta^3+\epsilon(1+\beta)$.

Is $-\beta^3+\epsilon(1+\beta)$ positive? Or how do we find a restriction for $n$ in order to get a contradiction?
 
  • #10
I actually gave a proof that $\sigma$ is the least upper bound in my first post, adding more details in my second post.
 
  • #11
Euge said:
Since $\sigma \in S$, then $\sigma$ is the least upper bound for $S$.

How do we know that there is no other element in $S$ that is a lower upper bound of the set? (Thinking)
 
  • #12
evinda said:
How do we know that there is no other element in $S$ that is a lower upper bound of the set? (Thinking)

If you have a nonempty bounded set $X$ with an upper bound $s$ such that $s\in X$, then $s$ is the least upper bound of $X$. For since $s\in X$, $s\le \sup X$. On the other hand, since $s$ is an upper bound for $X$, $s \ge \sup X$. This shows $\sup X = s$.
 
  • #13
Euge said:
If you have a nonempty bounded set $X$ with an upper bound $s$ such that $s\in X$, then $s$ is the least upper bound of $X$. For since $s\in X$, $s\le \sup X$. On the other hand, since $s$ is an upper bound for $X$, $s \ge \sup X$. This shows $\sup X = s$.

Ah I see... And how can we find the lower bound of the set when $-1<\alpha<0$ ?Also.. are my lower and upper bound of the set correct for the case when $0<\alpha<1$ ?
 

Similar threads

  • · Replies 11 ·
Replies
11
Views
2K
  • · Replies 8 ·
Replies
8
Views
3K
  • · Replies 7 ·
Replies
7
Views
2K
  • · Replies 1 ·
Replies
1
Views
2K
  • · Replies 2 ·
Replies
2
Views
1K
  • · Replies 29 ·
Replies
29
Views
3K
  • · Replies 4 ·
Replies
4
Views
1K
  • · Replies 3 ·
Replies
3
Views
5K
  • · Replies 1 ·
Replies
1
Views
3K
Replies
4
Views
2K