Bounding the L-Infty Norm of a Diffble Fn

  • Context: Graduate 
  • Thread starter Thread starter laonious
  • Start date Start date
  • Tags Tags
    Norm
Click For Summary

Discussion Overview

The discussion revolves around proving an inequality involving the L-Infinity norm of a twice continuously differentiable function on the interval [-1, 1]. Participants explore the implications of the inequality and the use of Taylor expansion in the proof process.

Discussion Character

  • Technical explanation
  • Mathematical reasoning
  • Debate/contested

Main Points Raised

  • One participant presents the inequality |f'(0)|^2 ≤ 4 ||f||_∞ (||f||_∞ + ||f''||_∞) and expresses difficulty in proving it, suggesting Taylor expansion as a potential approach.
  • Another participant suggests assuming WLOG that f(0), f'(0) ≥ 0 and ||f||_∞ = 1 to simplify the proof, leading to a reformulation of the inequality.
  • A further elaboration indicates that for x in (0, 1), f(x) can be bounded using the Taylor expansion, leading to a derived expression for f'(0) in terms of f(0) and ||f''||_∞.
  • Concerns are raised about the validity of a specific step in the proof, questioning whether the expression (1 - f(0) + ||f''||_∞/2)^2 ≤ 4 holds true, especially in light of functions with large second derivatives.
  • Another participant reminds others of the relationship x_0^2 = 2(1 - f(0))/||f''||_∞, which is relevant to the proof process.

Areas of Agreement / Disagreement

Participants do not reach a consensus on the validity of certain steps in the proof, particularly regarding the bounding of the expression involving f(0) and ||f''||_∞. Multiple competing views remain on the approach to proving the inequality.

Contextual Notes

The discussion highlights the assumptions made about the function f and its derivatives, as well as the implications of these assumptions on the proof. There are unresolved concerns regarding the bounding of specific expressions and the conditions under which they hold.

laonious
Messages
9
Reaction score
0
Hello,

I would appreciate any assistance with the following question: Suppose f \in C^2[-1,1] is twice continuously differentiable. Prove that

|f'(0)|^2 \leq 4 ||f||_\infty (||f||_\infty + ||f''||_\infty), where ||f||_infty is the standard sup norm. At first I thought Taylor expansion seemed promising, but I am pretty stymied.

Thank you!
 
Physics news on Phys.org
Fix your latex expression.
 
laonious said:
Hello,

I would appreciate any assistance with the following question: Suppose [tex]f \in C^2[-1,1][/tex] is twice continuously differentiable. Prove that

[tex]|f'(0)|^2 \leq 4 ||f||_\infty (||f||_\infty + ||f''||_\infty),[/tex] where [tex]||f||_{\infty}[/tex] is the standard sup norm. At first I thought Taylor expansion seemed promising, but I am pretty stymied.

Thank you!

Sorry, here's a fixed-up rewrite.
 
First, as the inequality is invariant under the transformations [itex]f \mapsto cf[/itex], [itex]x \mapsto -x[/itex], we may assume WLOG that [itex]f(0), \ f'(0) \ge 0[/itex], and [itex]||f||_\infty = 1[/itex]. Thus, we must prove
[tex]|f'(0)|^2 \le 4(1 + ||f''||_\infty)[/tex] subject to [itex]||f||_\infty = 1[/itex]. Now, for all [itex]x \in (0, 1)[/itex] we have
[tex]1 \ge f(x) \ge f(0) + f'(0)x - \frac{x^2}{2} ||f''||_\infty,[/tex] so [tex]f'(0) \le \frac{1 - f(0)}{x} + \frac{x}{2} ||f''||_\infty.[/tex] The RHS is minimized for [itex]x^2 = x^2_0 = \frac{2(1 - f(0))}{||f''||_\infty}[/itex] and decreasing for [itex]0 < x < x_0[/itex]. If [itex]x_0 \le 1[/itex] then [tex]|f'(0)|^2 \le \frac{1}{x_0^2} \left( 1 - f(0) + \frac{x^2_0}{2} ||f''||_\infty \right)^2 = 2(1 - f(0))||f''||_\infty \le 4||f''||_\infty[/tex].
If [itex]x_0 > 1[/itex] then [tex]|f'(0)|^2 \le \left(1 - f(0) + \frac{||f''||_\infty}{2} \right)^2 \le 4[/tex].

In either case [itex]|f'(0)|^2 \le 4 + 4||f''||_\infty[/itex].
 
namphcar,
Thanks for the great response, it was quite helpful. Would you mind being a little more explicit in your last step?

It's not clear to me that
[tex]\left(1-f(0)+\frac{||f''||_{\infty}}{2}\right)^2\leq 4,[/tex]
since a function of the form [itex]e^{kx}[/itex] would have an arbitrarily large second derivative.
Thanks!
 
Remember that [itex]x_0^2 = \frac{2(1 - f(0))}{||f''||_\infty}[/itex].
 

Similar threads

  • · Replies 3 ·
Replies
3
Views
4K
  • · Replies 3 ·
Replies
3
Views
4K
Replies
2
Views
2K
  • · Replies 10 ·
Replies
10
Views
4K
  • · Replies 0 ·
Replies
0
Views
4K
  • · Replies 1 ·
Replies
1
Views
3K
  • · Replies 9 ·
Replies
9
Views
8K
  • · Replies 2 ·
Replies
2
Views
2K
  • · Replies 1 ·
Replies
1
Views
2K
  • · Replies 5 ·
Replies
5
Views
2K