Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $f:\mathbb{R}\rightarrow\mathbb{R}$ be an everywhere twice continuously differentiable function, and suppose $f(1)=f'(1)=f(-1)=f'(-1)=0$. What is the maximum possible value for $f(0)$ given $|f''(x)|\leq 1$ for all $x\in\mathbb{R}$?

Thanks for the help!

share|cite|improve this question
Hint: use limited development up to second order in points $x=-1$ and $x=1$. – TZakrevskiy Feb 14 '14 at 14:45
Could you elaborate on what "limited development" means? I am not familiar with that term. – user88849 Feb 14 '14 at 14:50
$f(x+y) = f(x)+f'(x)y+f''(\xi)y^2/2$ where $\xi\in (\min(x,x+y),\max(x,x+y))$. – TZakrevskiy Feb 14 '14 at 14:55
I can see how that immediately gives $f(0)\leq\frac{1}{2}$, so if I can realize that with an example, the proof would be complete. I do not see a way to lower this to anything like 1/4, but then the trouble is that I cannot find such a function. Mostly I've tried playing around with quartic polynomials and the like, but obviously these would have to be modified away from $[-1,1]$ to control the second derivative on all of $\mathbb R$. – user88849 Feb 14 '14 at 15:07
up vote 4 down vote accepted

By continuity, $f$ assumes its maximum on $[-1,1]$ at some point $a$. Wlog. $a\ge 0$. By the condition, $f(1-h)\le \frac12h^2$. From $f'(a)=0$, also by the condition $f(a+h)\ge f(a)-\frac12h^2$. This implies $$ f(a)-\frac{(1-a)^2}8\le f\left(a+\frac{1-a}2\right)= f\left(1-\frac{1-a}2\right)\le \frac{(1-a)^2}8$$ hence $$f(0)\le f(a)\le \frac{(1-a)^2}{4}\le \frac14.$$ This is sharp as we see from $$ f(x)=\begin{cases}\frac12(1+x)^2&\text{if }x\le-\frac12\\ \frac14-\frac12x^2&\text{if }-\frac12\le x\le \frac12\\ \frac12(1-x)^2&\text{if }x\ge \frac12\end{cases}$$ which is twice dfferentiable except at $\pm\frac12$, but can be approximated arbitrarily well with a smoother function:

For $0<\epsilon<\frac12$ let $$ g(x)=\begin{cases}1&\text{if }x<-\frac12-\epsilon\text{ or } x>\frac12+\epsilon\\ \frac1\epsilon(x+\frac12)&\text{if }-\frac12-\epsilon\le x\le-\frac12+\epsilon\\ -1&\text{if }-\frac12+\epsilon<x<\frac12+\epsilon\\ \frac1\epsilon(\frac12-x)&\text{if }+\frac12-\epsilon\le x\le+\frac12+\epsilon,\end{cases}$$ then $G(x)=\int_{-1}^x g(t)\,\mathrm dt$ and $F(x)=\int_{-1}^x G(t)\,\mathrm dt$. Then $F(x)$ fulfills the conditions and almost coincides with $f(x)$. In fact $|G(x)-f'(x)|\le \int_{-\epsilon}^\epsilon \frac t\epsilon\,\mathrm dt =\epsilon$ and therefore $|F(0)-f(0)|\le \epsilon$ (because $F(-1)=f(-1)=0$).

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.