Page 1 of 1

Problema 2 ONM 2008

Posted: Wed Apr 30, 2008 1:10 pm
by Cezar Lupu
Se considera functia \( f:[0,1]\to\mathbb{R} \), derivabila, cu derivata continua pe \( [0,1] \). Sa se arate ca daca \( f(1/2)=0 \), atunci

\( \int_0^1(f^{\prime}(x))^{2}dx\geq 12\left(\int_0^1f(x)dx\right)^{2} \).

Cezar Lupu, Tudorel Lupu

Posted: Thu May 01, 2008 1:38 pm
by o.m.
Use Fubini theorem
\( \int_{0}^{1}f(x)dx=\int_{0}^{1}(\int_{1/2}^{x}f^{\prime}(t)dt)dx=\int_{0}^{1}k(t)f^{\prime}(t)dt \)

where \( k(t)=-t \) on \( [0;1/2] \) and \( k(t)=1-t \) on \( [1/2;1] \).

With Cauchy-Schwarz
\( (\int_{[0,1]}f)^2 \leq (\int_{[0;1]}k^2)(\int_{[0;1]}f^{\prime2}) \)

and
\( \int_{[0;1]}k^2=\int_{[0;1/2]}t^2dt+\int_{[1/2;1]}(1-t)^2dt=1/24+1/24=1/12 \)

Posted: Thu May 01, 2008 8:23 pm
by aleph
Your inequality with "24" is valid only for small values of 24 :D
More seriously, the second "=" is wrong.
The question if 12 is the best constant remains.

Posted: Fri May 02, 2008 12:27 pm
by Cezar Lupu
aleph wrote:Your inequality with "24" is valid only for small values of 24 :D
More seriously, the second "=" is wrong.
The question if 12 is the best constant remains.
Yes, the best constant is 12. The problem is not that easy as it seems at first glance. :)
My solution (which is elementary) is based on Cauchy-Schwarz inequality in the following way:

\( \int_0^{1/2}(f^{\prime}(x))^{2}dx\cdot\int_0^{1/2}x^{2}dx\geq\left(\int_0^{1/2}xf^{\prime}(x)\right)^{2} \)
and
\( \int_{1/2}^{1}(f^{\prime}(x))^{2}dx\cdot\int_{1/2}^{1}(1-x)^{2}dx\geq\left(\int_{1/2}^{1}(1-x)f^{\prime}(x)\right)^{2} \).

By adding these two inequalities and using the elementary inequality \( \alpha^2+\beta^2\geq\frac{(\alpha+\beta)^{2}}{2} \) and integration by parts, we get the desired result. \( \qed \)