Can anybody explain this to me? (Analysis)

  • Thread starter Thread starter Artusartos
  • Start date Start date
  • Tags Tags
    Analysis Explain
Join the discussion
Ask a follow-up here, or get your own question answered by working scientists, mathematicians and engineers — people, not an autocomplete.
Real named experts · corrections over time · the nuance an AI answer skips
13 replies · 2K views
Artusartos
Messages
236
Reaction score
0
For x in [0, infinity), let [itex]f_n(x)= \frac{x}{n}[/itex]...

Determine whether [itex]f_n[/itex] converges uniformly to f (the limit, which is equal to 0) on [0,1].

Answer:

Let [itex]\epsilon > 0[/itex] be given. Let [itex]N= \frac{1}{\epsilon}[/itex]. Then for n>N, [itex]| f_n(x) - 0 | = | \frac{x}{n} | \leq \frac{1/ \epsilon}{n} = \frac{1}{n \epsilon} < \epsilon[/itex] as desired.

My questions:

1) Why is [itex]| \frac{x}{n} | \leq \frac{1/ \epsilon}{n}[/itex]? How do we know that x is less than 1 over epsilon?

2) Why is [itex]\frac{1}{n \epsilon} < \epsilon[/itex]?

3) Finally, how did they know that N was supposed to be 1/epsilon?

Thanks in advance
 
Physics news on Phys.org
This "proof" is false.This sequence does not converge uniformly in that interval,because the suprimum for x is not final so certainly does not converge to 0.
 
hedipaldi said:
This "proof" is false.This sequence does not converge uniformly in that interval,because the suprimum for x is not final so certainly does not converge to 0.


But this answer is from here (page 49, exercise 24.2 b)

http://www.scribd.com/doc/70434268/Ross-Solutions
 
hedipaldi said:
This "proof" is false.This sequence does not converge uniformly in that interval,because the suprimum for x is not final so certainly does not converge to 0.

I don't know what you mean by "not final". Since the sequence converges, it MUST converge uniformly on any closed and bounded (i.e. compact) interval.
 
Artusartos said:
But this answer is from here (page 49, exercise 24.2 b)

http://www.scribd.com/doc/70434268/Ross-Solutions

(I assume you mean page 47.) The proof for part (b) says "Claim: [itex]f_n \rightarrow f[/itex] uniformly on [0,1]." Part (c) shows by contradiction that it is not true on [itex][0,\infty)[/itex].

I am not sure which part you are trying to solve:

Artusartos said:
For x in [0, infinity), let [itex]f_n(x)= \frac{x}{n}[/itex]...

Determine whether [itex]f_n[/itex] converges uniformly to f (the limit, which is equal to 0) on [0,1].
 
jbunniii said:
(I assume you mean page 47.) The proof for part (b) says "Claim: [itex]f_n \rightarrow f[/itex] uniformly on [0,1]." Part (c) shows by contradiction that it is not true on [itex][0,\infty)[/itex].

I am not sure which part you are trying to solve:

I was actually talking about part (b)...about why it converges uniformly on [0,1].
 
I read through the proof at your link, and it's a mess. My comments below:

Artusartos said:
For x in [0, infinity), let [itex]f_n(x)= \frac{x}{n}[/itex]...

Determine whether [itex]f_n[/itex] converges uniformly to f (the limit, which is equal to 0) on [0,1].

Answer:

Let [itex]\epsilon > 0[/itex] be given. Let [itex]N= \frac{1}{\epsilon}[/itex]. Then for n>N, [itex]| f_n(x) - 0 | = | \frac{x}{n} | \leq \frac{1/ \epsilon}{n} = \frac{1}{n \epsilon} < \epsilon[/itex] as desired.

My questions:

1) Why is [itex]| \frac{x}{n} | \leq \frac{1/ \epsilon}{n}[/itex]? How do we know that x is less than 1 over epsilon?
In general, it's not true. All you know is that [itex]0 \leq x \leq 1[/itex]. If [itex]0 < \epsilon < 1[/itex] (assumption not stated in the proof), then [itex]x \leq 1 < 1/\epsilon[/itex].

2) Why is [itex]\frac{1}{n \epsilon} < \epsilon[/itex]?
This is certainly not true. Take [itex]\epsilon = 1/10[/itex], [itex]N = 10[/itex], and [itex]n = 11[/itex]. Then
[tex]\frac{1}{n \epsilon} = \frac{1}{11 /10} = \frac{10}{11}[/tex]
which is certainly not less than 1/10.

I suggest ignoring that "solution" altogether. The statement is true, but the proof is bogus.
 
jbunniii said:
I read through the proof at your link, and it's a mess. My comments below:


In general, it's not true. All you know is that [itex]0 \leq x \leq 1[/itex]. If [itex]0 < \epsilon < 1[/itex] (assumption not stated in the proof), then [itex]x \leq 1 < 1/\epsilon[/itex].


This is certainly not true. Take [itex]\epsilon = 1/10[/itex], [itex]N = 10[/itex], and [itex]n = 11[/itex]. Then
[tex]\frac{1}{n \epsilon} = \frac{1}{11 /10} = \frac{10}{11}[/tex]
which is certainly not less than 1/10.

I suggest ignoring that "solution" altogether. The statement is true, but the proof is bogus.


So can I prove it like this?

[itex]| f_n(x) - 0| = | \frac{x}{n} | \leq \frac{1}{n}[/itex]

So when is 1/n equal to epsilon? It is when n=1/epsilon...so n needs to be greater than 1/epsilon. Do you think my answer is correct?
 
Artusartos said:
So can I prove it like this?

[itex]| f_n(x) - 0| = | \frac{x}{n} | \leq \frac{1}{n}[/itex]

So when is 1/n equal to epsilon? It is when n=1/epsilon...so n needs to be greater than 1/epsilon. Do you think my answer is correct?

Yes, if [itex]n > 1/\epsilon[/itex], then [itex]1/n < \epsilon[/itex], which is what you need. The key is that this choice of [itex]n[/itex] works for any [itex]x \in [0,1][/itex].
 
jbunniii said:
Yes, if [itex]n > 1/\epsilon[/itex], then [itex]1/n < \epsilon[/itex], which is what you need. The key is that this choice of [itex]n[/itex] works for any [itex]x \in [0,1][/itex].

Thanks a lot. I also have another question, if you don't mind...

From the same link that I gave...for exercise 24.6 (b), we are asked if f_n converges uniformly on [0,1]. And the solution (in the link) says that [itex]| \frac{1-2xn}{n^2} | \leq \frac{1}{\sqrt{n}}[/itex]. I wasn't able to understand why that is true...
 
Artusartos said:
Thanks a lot. I also have another question, if you don't mind...

From the same link that I gave...for exercise 24.6 (b), we are asked if f_n converges uniformly on [0,1]. And the solution (in the link) says that [itex]| \frac{1-2xn}{n^2} | \leq \frac{1}{\sqrt{n}}[/itex]. I wasn't able to understand why that is true...
That's because it is false. Consider [itex]x = 1[/itex], [itex]n = 2[/itex]. Then
[tex]\left|\frac{1 - 2xn}{n^2}\right| = \left|\frac{1-4}{4}\right| = \frac{3}{4}[/tex]
which is not less than [itex]1/\sqrt{2} \approx 0.7071[/itex].

Whoever wrote the solution manual you are using is just plain wrong.

You can obtain a bound that will work as follows:
[tex]|1 - 2xn| \leq |1| + |2xn| = 1 + 2n|x| \leq 1 + 2n \leq n + 2n = 3n[/tex]
Try using this to solve the problem. You'll have to use a different N from what the solution chose, but it should work.
 
jbunniii said:
That's because it is false. Consider [itex]x = 1[/itex], [itex]n = 2[/itex]. Then
[tex]\left|\frac{1 - 2xn}{n^2}\right| = \left|\frac{1-4}{4}\right| = \frac{3}{4}[/tex]
which is not less than [itex]1/\sqrt{2} \approx 0.7071[/itex].

Whoever wrote the solution manual you are using is just plain wrong.

You can obtain a bound that will work as follows:
[tex]|1 - 2xn| \leq |1| + |2xn| = 1 + 2n|x| \leq 1 + 2n \leq n + 2n = 3n[/tex]
Try using this to solve the problem. You'll have to use a different N from what the solution chose, but it should work.



Thanks a lot :)