Where $a$ and $b$ are any given real number. I have tried solving it using partial derivative. $$ s=a^2+b^2+(1-a-b)^2$$ $$\frac{\partial s}{\partial a}=2a-2(1-a-b) \tag{1}$$ $$\frac{\partial s}{\partial b}=2b-2(1-a-b) \tag{2}$$ for maxima both (1) and (2) are 0..from here we get two equations from where we get values of $a$ and $b$ $$2a-2(1-a-b)=0 \tag{3}$$ $$2b-2(1-a-b)=0 \tag{4}$$ putting the values of $a$ and $b$ we find from (3) and (4) in the function the maximum value of the function should be $\frac {1}{3}$.which in this case is not
-
The mean of the square is greater or equal to the square of the mean. – Bill Kleinhans Jul 05 '13 at 19:55
6 Answers
Hint: Apply Cauchy-Schwarz inequality to the vectors $$ \vec{u}=(1,1,1)\qquad\text{and}\qquad\vec{v}=(a,b,1-a-b). $$
Hint: Look at your equations $(3)$ and $(4)$. Can you deduce that $a=b$ at a critical point? Can you then solve for $a$ and $b$?
- 133,153
-
Elementary methods surely work, too. C-S is just damn useful with problems of this type, so I recommend that you learn it (not implying that C-S wouldn't be counted as elementary as well). – Jyrki Lahtonen Jul 05 '13 at 07:11
-
firstly, the maximum value of the function in that case is 1/2 so even if it shows the function is greater than 1/3 it does not show the function is equal to 1/3. Secondly, each time i have used ds/da=0..i am not sure if its maxima or minima ..so is the method in that case mathematically correct? – Mustahid Jul 05 '13 at 07:37
-
@Mustahid: Are you sure? I got $a=b=1/3$ from your equations $(3)$ and $(4)$. Plugging them into the formula gives value $3/9=1/3$. Of course, you still need to prove that it is a global minimum. – Jyrki Lahtonen Jul 05 '13 at 07:40
Using Titu Andreescu's Lemma, we have
$\dfrac{a^2}{1}+\dfrac{b^2}{1}+\dfrac{(1-a-b)^2}{1} \ge \dfrac{(a+b+1-a-b)^2}{3}=\dfrac{1}{3}$
- 7,881
Taking it from where you left it:
$$a=1-a-b\implies a=\frac{1-b}2\\ b=1-a-b\implies b=\frac{1-a}2$$
So substituting the first eq. into the second one and using the symmetry of both:
$$b=\frac{1-\frac{1-b}2}2=\frac{1+b}4\implies b=\frac13\;\;\text{, and thus also}\;\;a=\frac13$$
Continuing with the partial derivatives
$$\frac{\partial^2s}{\partial a^2}=4=\frac{\partial^2s}{\partial b^2}\;\;,\;\;\frac{\partial^2s}{\partial a\partial b}=\frac{\partial^2s}{\partial b\partial a}=2$$
$$\text{Thus, at}\;\;\left(\frac13\,,\,\frac13\right)\;,\;\;\text{the Hessian is positive definite and we get a minimum there.}$$
From here, we get that
$$\forall\,a,b\in\Bbb R\;,\;\;a^2+b^2+(1-a-b)^2\ge\frac29+\left(1-\frac23\right)^2=\frac29+\frac19=\frac13\;\ldots$$
- 211,718
- 17
- 136
- 287
-
Thanks to both for helping me.. :) I have studied Cauchy Schwarz inequality right now.and it has made the math really easy..i do not know how to input mathematical notation..could you please tell me how?? – Mustahid Jul 05 '13 at 08:45
-
There was a section called FAQ until some very few weeks ago where you could read about how to use LaTeX to write mathematics properly in this site. Weirdly enough, that section disappeared and I've no idea where that tutorial was put to rest...Perhaps you should ask a question in the META setcion... – DonAntonio Jul 05 '13 at 08:51
One more approach. We start with a bit simpler inequality:
$$a^2+(1-a)^2 > \frac{1}{2},$$
which can be proven without words by the following picture:
$\hspace{100pt}$
(if you insist on explanation: you want to make dark gray as small as possible, but clearly as long as $a > b$ then the red part is bigger than green).
Now, if there are three squares (as in the question), and some two are of different sizes, then you can make them of equal size without changing the third square, but at the same time decreasing the dark gray area. This process converges to three squares of the same size and gives an intuition why the result holds ;-)
$\hspace{100pt}$
I know it's not a proper proof, but I hope it helps ;-)
- 37,381
LHS$ =a^2+b^2+1-2(a+b)+(a+b)^2 \ge \dfrac{(a+b)^2}{2}+1-2(a+b)+(a+b)^2=\dfrac{3}{2}(a+b)^2-2(a+b)+1=\left(\sqrt{\dfrac{3}{2}}(a+b)-\sqrt{\dfrac{2}{3}}\right)^2+\dfrac{1}{3} \ge \dfrac{1}{3} $
first "=" is $a=b$,2nd "=" is $a+b=\dfrac{2}{3} \to a=b=\dfrac{1}{3}$ to get $\dfrac{1}{3}$
- 7,581
You want to estimate $R=a^2+b^2+c^2$, where $a+b+c=1$. Note that $R$ is the squared length of the radius vector to $(a,b,c)$. The equation $a+b+c=1$ gives you some plane, and the shortest radius vector to the plane is the perpendicular on it. It's easy to calculate that its length(height of a pyramid) is equal to $\frac{1}{ \sqrt{3}}$, so $R\ge\frac 1 3$.
Anyway, I like the way with Cauchy-Schwarz inequality more - it is the most obvious and easy way.
- 1,500