Without losing generality we assume that \(ab\geq 0 \). From the 1st condition we have \(a+b=-c\) replace this into the 2nd condition
\[a^2+b^2+(a+b)^2=1\]
By C-S \(a^2+b^2\geq\frac{(a+b)^2}{2}\), so \(1\geq\frac{3}{2}(a+b)^2 \Rightarrow c^2\leq\frac{2}{3}\). Now we rewrite the 2nd condition
\[2(a^2+b^2)+2ab=1\]
We have this inequality for all reals \(a^2+b^2\geq 2ab\), therefore \(6ab\leq 1\Rightarrow a^2b^2\leq\frac{1}{36}\)
\[\therefore a^2b^2c^2\leq\frac{1}{54}\]
The equality holds when \((a,b,c)=\bigg(\frac{1}{\sqrt{6}},\frac{1}{\sqrt{6}},\frac{-2}{\sqrt{6}}\bigg);\bigg(\frac{-1}{\sqrt{6}},\frac{-1}{\sqrt{6}},\frac{2}{\sqrt{6}}\bigg)\)

Now, as \[0=(a+b+c)^2=a^2+b^2+c^2+2(ab+bc+ca)\\\implies 0=1+2(ab+bc+ca)\\\implies ab+bc+ca=-\frac{1}{2}\]

Therefore, \(a, b, c\) are the (real) roots of the following function:

\[f(x)=x^3-\frac{1}{2}x-\alpha\]

For the existence of three real roots, we must have: \[f(x_1)f(x_2) \leq 0\] where \(x_1\) and \(x_2\) are the points of extrema of the function \(f(x)\).

Note that \(f'(x_1)=f'(x_2)=0\).

Since \(f'(x)\) is a quadratic polynamial, it can be solved.

Solving, plugging it into the above inequality, we get: \[|\alpha| \leq \frac{1}{3\sqrt6} \\\implies \alpha^2 \leq \frac{1}{54} \]

With equality holding if and only if two of the real numbers \(a, b, c\) are equal.

Right, actually, if the three are non-negative, then the three are equal to 0. But this can be proved by cases, for example, WLOG let's assume \(a\) and \(b\) are positive and \(c\) is negative, then \(-c\) is positive, and so on.

I got a question... if you post a problem as a note, why to give yourself the solution?

## Comments

Sort by:

TopNewestWithout losing generality we assume that \(ab\geq 0 \). From the 1st condition we have \(a+b=-c\) replace this into the 2nd condition \[a^2+b^2+(a+b)^2=1\] By C-S \(a^2+b^2\geq\frac{(a+b)^2}{2}\), so \(1\geq\frac{3}{2}(a+b)^2 \Rightarrow c^2\leq\frac{2}{3}\). Now we rewrite the 2nd condition \[2(a^2+b^2)+2ab=1\] We have this inequality for all reals \(a^2+b^2\geq 2ab\), therefore \(6ab\leq 1\Rightarrow a^2b^2\leq\frac{1}{36}\) \[\therefore a^2b^2c^2\leq\frac{1}{54}\] The equality holds when \((a,b,c)=\bigg(\frac{1}{\sqrt{6}},\frac{1}{\sqrt{6}},\frac{-2}{\sqrt{6}}\bigg);\bigg(\frac{-1}{\sqrt{6}},\frac{-1}{\sqrt{6}},\frac{2}{\sqrt{6}}\bigg)\)

Log in to reply

Let \(\alpha=abc\).

Now, as \[0=(a+b+c)^2=a^2+b^2+c^2+2(ab+bc+ca)\\\implies 0=1+2(ab+bc+ca)\\\implies ab+bc+ca=-\frac{1}{2}\]

Therefore, \(a, b, c\) are the (real) roots of the following function:

\[f(x)=x^3-\frac{1}{2}x-\alpha\]

For the existence of three real roots, we must have: \[f(x_1)f(x_2) \leq 0\] where \(x_1\) and \(x_2\) are the points of extrema of the function \(f(x)\).

Note that \(f'(x_1)=f'(x_2)=0\).

Since \(f'(x)\) is a quadratic polynamial, it can be solved.

Solving, plugging it into the above inequality, we get: \[|\alpha| \leq \frac{1}{3\sqrt6} \\\implies \alpha^2 \leq \frac{1}{54} \]

With equality holding if and only if two of the real numbers \(a, b, c\) are equal.

Log in to reply

Note: It's slightly more straightforward to just take the cubic discriminant condition for 3 real roots.

Log in to reply

The problem was I didn't remeber the cubic discriminant formula then. So, I used calculus in the solution.

Log in to reply

Comment deleted Jun 07, 2016

Log in to reply

Btw, equality holds only if two of the three numbers \(a, b, c\) are equal.

Log in to reply

You can't use am-gm inequality as \(a, b\) and \(c\) aren't non negative (if they are, then the inequality to be proved holds trivially).

Since \(a, b, c\) are real, \(a^2b^2c^2 \geq 0\).

Log in to reply

Right, actually, if the three are non-negative, then the three are equal to 0. But this can be proved by cases, for example, WLOG let's assume \(a\) and \(b\) are positive and \(c\) is negative, then \(-c\) is positive, and so on.

I got a question... if you post a problem as a note, why to give yourself the solution?

Log in to reply

Log in to reply

ya you are right

Log in to reply

Log in to reply

ok @Deeparaj Bhat

Log in to reply