Suppose that $f(x)$ is differentiable on $[0,1]$ and $f(0) = f(1) = 0$. It is also known that $|f''(x)| \le A$ for every $x \in (0,1)$. Prove that $|f'(x)| \le A/2$ for every $x \in [0,1]$.

I'll explain what I did so far. First using Rolle's theorem, there is some point $c \in [0,1]$ so $f'(c) = 0$.

EDIT: My first preliminary solution was wrong so I tried something else. EDIT2: Another revision :\

I define a Taylor series of a second order around the point $1$: $$ f(x) = f(1) + f'(1)(x-1) + \frac12 f''(d_1)(x-1)^2 $$ $$ f(0) = f(1) + f'(1)(-1) + \frac12 f''(d_1)(-1)^2 $$ $$ |f'(1)| = \frac12 |f''(d_1)| <= \frac12 A $$

Now I develop a Taylor series of a first order for $f'(x)$ around $1$: $$ f'(x) = f'(1) + f''(d_2)(x-1) $$ $$ |f'(x)| = |f'(1)| + x*|f''(d_2)|-|f''(d_2)| \leq \frac{A}{2} + A - A = \frac{A}{2} $$

It looks correct to me, what do you guys think?

Note: I cannot use integrals, because we have not covered them yet.

  • 1
    $\begingroup$ $c$ does not need to be $1/2$. $\endgroup$ – Andrés E. Caicedo Jan 9 '14 at 18:38
  • $\begingroup$ Yeah I thought about that, but I tried drawing a sample function graph a number of times, and it seemed like it had to be going through the middle if its uniformly continuous. Perhaps I misunderstood the concept of uniform continuity? Does it even play a role here? $\endgroup$ – Xsy Jan 9 '14 at 22:18
  • $\begingroup$ I've edited the question. My first solution was wrong so I tried something else, would love to know if its correct. $\endgroup$ – Xsy Jan 12 '14 at 0:52
  • $\begingroup$ I think it's alright. Don't forget to mention about $d$ that it is between $c$ and $x$. $\endgroup$ – GinKin Jan 12 '14 at 17:54
  • 1
    $\begingroup$ The new edit isn't right either. Since $x<1$, the right expression is $$ |f'(x)|=|f'(1)|+|f''(d_2)|(1-x)\le\frac A2+A(1-x), $$ and this is not enough to conclude what you want. $\endgroup$ – Andrés E. Caicedo Jan 12 '14 at 18:15

I've already give an answer without Integral to this question in another post. I'll duplicate the answer here. If it's against policy please tell me what I should do.

I manage an answer using both time a taylor expansion center at a point $x \in \left [0, 1 \right ]$. I get $$ f(h) = f(x)+f'(x)(h-x) + \frac{1}{2} f''(\xi(h))(h-x)^2 $$

now for h = 0 and h = 1, I get

$$0=f(x)-xf'(x)+\frac{1}{2}x^2f''(\xi(0))$$ and $$0=f(x)+f'(x)-xf'(x)+\frac{1}{2}(1-x)^2f''(\xi(1)).$$ Subtracting one to the other to get $f(x)$ out and a little manipulation yield: $$|f'(x)| = \frac{1}{2}|x^2f''(\xi(0))-(1-x)^2f''(\xi(1))|\leq\frac{1}{2}(|x^2f''(\xi(0))|+|(1-x)^2f''(\xi(1))|)\leq \frac{A}{2}(|x^2|+|1-x|^2) $$

since $x \in \left [0, 1 \right ], (|x^2|+|1-x|^2) \leq1 $ and we get the result.


If $f''$ exists on $[0,1]$ and $|f''(x)| \le A$, then $f'$ is Lipschitz continuous, which implies that $f'$ is absolutely continuous. So $f'(a)-f'(b)=\int_{a}^{b}f''(t)\,dt$ for any $0 \le a, b \le 1$. And, of course, the same is true of $f$ because $f$ is continuously differentiable.

Because $f(0)=0$, one has the following for $0 \le x \le 1$: $$ f(x) = \int_{0}^{x}f'(t)\,dt = tf'(t)|_{0}^{x}-\int_{0}^{x}tf''(t)\,dt=xf'(x)-\int_{0}^{x}tf''(t)\,dt. $$ Similarly, because $f(1)=0$, one has the following for $0 \le x \le 1$: $$ f(x)=\int_{1}^{x}f'(t)\,dt = (t-1)f'(t)|_{1}^{x}-\int_{1}^{x}(t-1)f''(t)\,dt=(x-1)f'(x)-\int_{1}^{x}(t-1)f''(t)\,dt. $$ Subtracting the second equation from the first gives $$ 0=f'(x)-\int_{0}^{x}tf''(t)\,dt+\int_{1}^{x}(t-1)f''(t)\,dt. $$ Therefore, $$ |f'(x)| \le A\left(\int_{0}^{x}t\,dt + \int_{x}^{1}(1-t)\,dt\right)=\frac{A}{2}(x^{2}+(1-x)^{2}). $$ The maximum of the expression on the right occurs at $x=0$ and $x=1$, with a value of $A/2$. Therefore, $|f'(x)| \le A/2$.

  • 2
    $\begingroup$ Thank's for the solution but we haven't covered integrals yet, so I can't really use/understand your solution at the moment. $\endgroup$ – Xsy Jan 10 '14 at 11:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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