Limit Definition of Derivative as n Approaches Infinity

  • Thread starter Thread starter songoku
  • Start date Start date
  • Tags Tags
    Limit Proof
songoku
Messages
2,467
Reaction score
382
Homework Statement
Please see below
Relevant Equations
Limit
Sequence Theorem
Derivative from first principle
1664335070591.png


##f'(x_0)## is defined as:
$$f'(x_0)=\lim_{h \rightarrow 0} \frac{f(x_0+h)-f(x_0)}{h}$$
or
$$f'(x_0)=\lim_{x \rightarrow x_0} \frac{f(x)-f(x_0)}{x-x_0}$$

I can imagine that as ##n \rightarrow \infty## the value of ##f(b_n)## and ##f(a_n)## will approach ##f(x_0)## so the value of the limit will be like tangent to graph ##f(x)## at point ##x_0##

But I don't know how to do it mathematically. The definition I know for derivative is the limit approaches 0 while the question is n approaches infinity. How to relate the question to the definition?

Thanks
 
Physics news on Phys.org
How about
b_n-a_n:=h_n
h_n\rightarrow 0
 
  • Like
Likes songoku
anuttarasammyak said:
How about
b_n-a_n:=h_n
h_n\rightarrow 0
I think I can solve the question using this hint.

Thank you very much anuttarasammyak
 
songoku said:
Homework Statement:: Please see below
Relevant Equations:: Limit
Sequence Theorem
Derivative from first principle

View attachment 314745

##f'(x_0)## is defined as:
$$f'(x_0)=\lim_{h \rightarrow 0} \frac{f(x_0+h)-f(x_0)}{h}$$
or
$$f'(x_0)=\lim_{x \rightarrow x_0} \frac{f(x)-f(x_0)}{x-x_0}$$

I can imagine that as ##n \rightarrow \infty## the value of ##f(b_n)## and ##f(a_n)## will approach ##f(x_0)## so the value of the limit will be like tangent to graph ##f(x)## at point ##x_0##

But I don't know how to do it mathematically. The definition I know for derivative is the limit approaches 0 while the question is n approaches infinity. How to relate the question to the definition?

Thanks
Use ##\epsilon-\delta## to relate the two?
 
PeroK said:
Use ##\epsilon-\delta## to relate the two?
Sorry I don't understand the hint.

I followed hint in post#2 and this is what I did:
Let ##h_n=b_n -a_n## and as ##n \rightarrow \infty, h_n \rightarrow 0##
$$\lim_{n \rightarrow \infty} \frac{f(b_n)-f(a_n)}{b_n-a_n}$$
$$=\lim_{h_n \rightarrow 0} \frac{f(a_n+h_n)-f(a_n)}{h_n}$$

The expression is equal to ##f'(a_n)## and ##f'(a_n)## is equal to ##f'(x_0)##

By ##\epsilon-\delta##, do you mean using ##\epsilon-\delta## definition of limit?

Thanks
 
songoku said:
Sorry I don't understand the hint.

I followed hint in post#2 and this is what I did:
Let ##h_n=b_n -a_n## and as ##n \rightarrow \infty, h_n \rightarrow 0##
$$\lim_{n \rightarrow \infty} \frac{f(b_n)-f(a_n)}{b_n-a_n}$$
$$=\lim_{h_n \rightarrow 0} \frac{f(a_n+h_n)-f(a_n)}{h_n}$$

The expression is equal to ##f'(a_n)## and ##f'(a_n)## is equal to ##f'(x_0)##
That needs to be proved.
songoku said:
By ##\epsilon-\delta##, do you mean using ##\epsilon-\delta## definition of limit?

Thanks
Yes. What else could that mean?
 
PeroK said:
That needs to be proved.
You mean I have to prove ##f'(a_n)=f'(x_0)##?

PeroK said:
Yes. What else could that mean?
The definition is something that is not taught in class so maybe we are not expected to use that but I will try first and see what I can get from it

Thanks
 
songoku said:
You mean I have to prove ##f'(a_n)=f'(x_0)##?
That statement is clearly not true. You don't even know that for all ##n## ##f'(a_n)## exists.
songoku said:
The definition is something that is not taught in class so maybe we are not expected to use that but I will try first and see what I can get from it

Thanks
Then what definition of a limit are you using?
 
  • Like
Likes songoku
anuttarasammyak said:
How about
b_n-a_n:=h_n
h_n\rightarrow 0
I'm not convinced this helps. We only know the derivative exists at ##x_0##, which is a key point.
 
  • #10
@songoku I think you need to expand the expression and pull out a constant term of ##f'(x_0)## and show that what's left converges to zero. There's quite a lot of algebra.
 
  • Like
Likes songoku
  • #11
If f is differentiable at x_0 then for any h, <br /> f(x_0 + h) = f(x_0) + hf&#039;(x_0) + E(h) where \lim_{h \to 0} \frac{E(h)}{h} = 0. If you set h_n = a_n - x_0 and k_n = b_n - x_0 then the algebra is easier, although I did have to resort to proof by cases depending on whether |h_n/k_n| remains bounded as n \to \infty.
 
Last edited:
  • Like
Likes songoku
  • #12
PeroK said:
I'm not convinced this helps. We only know the derivative exists at x0, which is a key point.
Let us see it.
\lim_{n \rightarrow \infty} \frac{f(a_n+h_n)-f(a_n)}{h_n}
If we make it
=\lim_{n\rightarrow \infty} \frac{f(x_0+h_n)-f(x_0)}{h_n}
It works. Can we make it so ?
 
  • #13
anuttarasammyak said:
Let us see it.
\lim_{n \rightarrow \infty} \frac{f(a_n+h_n)-f(a_n)}{h_n}
If we make it
=\lim_{n\rightarrow \infty} \frac{f(x_0+h_n)-f(x_0)}{h_n}
It works. Can we make it so ?
Okay, but you still have to justify that step.
 
  • #14
\lim_{n\rightarrow \infty}\frac{f(b_n)-f(a_n)}{b_n-a_n}=\lim_{n\rightarrow \infty}\frac{[f(\beta_n+x_0)-f(x_0)]+[f(x_0)-f(x_0-\alpha_n)]}{\beta_n+\alpha_n}
where
\beta_n=b_n-x_0
\alpha_n=x_0-a_n, decomposing ##h_n## before. We know
\lim_{n\rightarrow \infty}\frac{f(\beta_n+x_0)-f(x_0)}{\beta_n}=\lim_{n\rightarrow \infty}\frac{f(x_0)-f(x_0-\alpha_n)}{\alpha_n}=f&#039;(x_0)
 
Last edited:
  • #15
anuttarasammyak said:
\lim_{n\rightarrow \infty}\frac{f(b_n)-f(a_n)}{b_n-a_n}=\lim_{n\rightarrow \infty}\frac{[f(\beta_n+x_0)-f(x_0)]+[f(x_0)-f(x_0-\alpha_n)]}{\beta_n+\alpha_n}
where
\beta_n=b_n-x_0
\alpha_n=x_0-a_n, decomposing ##h_n## before. We know
\lim_{n\rightarrow \infty}\frac{f(\beta_n+x_0)-f(x_0)}{\beta_n}=\lim_{n\rightarrow \infty}\frac{f(x_0)-f(x_0-\alpha_n)}{\alpha_n}=f&#039;(x_0)
Two things. You have a problem with the denominator and this isn't your homework.
 
  • #16
To give a big hint. The first step here is to rewrite the expression as:
$$\frac{f(b_n) - f(a_n)}{b_n - a_n} = \frac{f(b_n) - f(x_0)}{b_n - a_n} + \frac{f(x_0) - f(a_n)}{b_n - a_n}$$$$ = \frac{f(b_n) - f(x_0)}{b_n - x_0}\bigg [\frac{b_n - x_0}{b_n - a_n} \bigg ] + \frac{f(x_0) - f(a_n)}{x_0 - a_n} \bigg [\frac{x_0 - a_n}{b_n - a_n} \bigg ]$$And then to introduce the quantity ##f'(x_0)## ...
 
  • Like
Likes songoku and FactChecker
  • #17
PeroK said:
Two things. You have a problem with the denominator and this isn't your homework.
Though this is not my homework let me investigate the problem you find. With ## \lim_{n\rightarrow \infty} ##, may we equate the fractions as
\frac{f(\beta_n+x_0)-f(x_0)}{\beta_n}=\frac{f(x_0)-f(x_0-\alpha_n)}{\alpha_n}=\frac{[f(\beta_n+x_0)-f(x_0)]+[f(x_0)-f(x_0-\alpha_n)]}{\beta_n+\alpha_n}?[EDIT] For any n, may we say
\min \{\frac{f(\beta_n+x_0)-f(x_0)}{\beta_n},\frac{f(x_0)-f(x_0-\alpha_n)}{\alpha_n}\}
\leq\frac{[f(\beta_n+x_0)-f(x_0)]+[f(x_0)-f(x_0-\alpha_n)]}{\beta_n+\alpha_n}\leq
\max \{\frac{f(\beta_n+x_0)-f(x_0)}{\beta_n},\frac{f(x_0)-f(x_0-\alpha_n)}{\alpha_n}\}?
 
Last edited:
  • #18
PeroK said:
That statement is clearly not true. You don't even know that for all ##n## ##f'(a_n)## exists.
Ah yes, I see.
PeroK said:
Then what definition of a limit are you using?
I am not sure. The teacher said: "If we want to prove limit as x approaches 5 of (x + 1) is 6, then we need to use precise definition of limit but I won't cover that"

So maybe I am not using any definition of limit right now, I am not really sure.

PeroK said:
To give a big hint. The first step here is to rewrite the expression as:
$$\frac{f(b_n) - f(a_n)}{b_n - a_n} = \frac{f(b_n) - f(x_0)}{b_n - a_n} + \frac{f(x_0) - f(a_n)}{b_n - a_n}$$$$ = \frac{f(b_n) - f(x_0)}{b_n - x_0}\bigg [\frac{b_n - x_0}{b_n - a_n} \bigg ] + \frac{f(x_0) - f(a_n)}{x_0 - a_n} \bigg [\frac{x_0 - a_n}{b_n - a_n} \bigg ]$$And then to introduce the quantity ##f'(x_0)## ...
I understand this hint
 
  • #19
PeroK said:
To give a big hint. The first step here is to rewrite the expression as:
$$\frac{f(b_n) - f(a_n)}{b_n - a_n} = \frac{f(b_n) - f(x_0)}{b_n - a_n} + \frac{f(x_0) - f(a_n)}{b_n - a_n}$$$$ = \frac{f(b_n) - f(x_0)}{b_n - x_0}\bigg [\frac{b_n - x_0}{b_n - a_n} \bigg ] + \frac{f(x_0) - f(a_n)}{x_0 - a_n} \bigg [\frac{x_0 - a_n}{b_n - a_n} \bigg ]$$And then to introduce the quantity ##f'(x_0)## ...
As in my post #2 you leave ##b_n-a_n## in the formula. Now I think it was not a good strategy because of independent convergence of ##a_n \rightarrow x_0 ## and ##b_n \rightarrow x_0 ##.
The first term of RHS in the limit is
f&#039;(x)\lim_{n \rightarrow \infty}(1+\frac{x_0-a_n}{b_n-x_0})^{-1}
The second term of RHS in the limit is
f&#039;(x)\lim_{n \rightarrow \infty}(1+\frac{b_n-x_0}{x_0-a_n})^{-1}=f&#039;(x)(1-\lim_{n \rightarrow \infty}(1+\frac{x_0-a_n}{b_n-x_0})^{-1})
They cancel to bring f'(x) but I wonder according to independent convergence of ##a_n \rightarrow x_0 ## and ##b_n \rightarrow x_0 ##, though we know that the value for any n is between 0 and 1 anyway, I am not sure there exist the limit value for infinite n. Does the cancellation work even when there exist the no limit value ? I prefer a discussion without comparing the convergences of ##a_n## and ##b_n## .

[EDIT] Now I understand even if there exists no limit, as it is bounded between 0 and 1, we can get the result. Thanks for teachings guys.
 
Last edited:
  • #20
anuttarasammyak said:
The first term of RHS in the limit is
f&#039;(x)\lim_{n \rightarrow \infty}(1+\frac{x_0-a_n}{b_n-x_0})^{-1}
The first term of RHS in the limit is
f&#039;(x)\lim_{n \rightarrow \infty}(1+\frac{b_n-x_0}{x_0-a_n})^{-1}
That was not my idea.
 
  • Like
Likes anuttarasammyak
  • #21
Thank you very much anuttarasammyak, PeroK, pasmith
 
Back
Top