2

I have noticed that in many functions with one or more variables we can get the maxima or minima if we assume all the variables to be equal. Are there certain properties a function must satisfy for this to be true, or is this completely random?

Take a look at these two functions.

  1. Assume ABC is an acute angled triangle and $\displaystyle p=\frac{\sqrt3+\sin A+\sin B+\sin C}{2\sin A\sin B\sin C}$ .If we assume $A=B=C=60^\circ$ we can get the minimum value of $p$.

  2. Similarly $y=\sum_{cyc (a,b,c)}{}\frac{a}{b^2+1} \geq \frac{3}{2}$ where $a+b+c=3$ .If we assume $a=b=c$, we can get the minimum value of $y$

There are countless other examples like this, is there anything going on here?

Oshawott
  • 4,028
Tatai
  • 755
  • 3
    Symmetry of the expression could explain this fact in many cases but it is not a general fact for all functions of course. – user Aug 31 '21 at 12:35
  • Related: https://math.stackexchange.com/q/2315419/42969, https://math.stackexchange.com/q/911276/42969, https://math.stackexchange.com/q/448274/42969 – Martin R Aug 31 '21 at 12:56

1 Answers1

1

I think it's completely random in the general.

If we want to use AM-GM, we know that the equality occurs for equality case of any variables.

For example, for non-negatives $a$ and $b$ we have $\frac{a+b}{2}\geq\sqrt{ab}$,

which says that a minimal value of $\frac{a+b}{2}-\sqrt{ab}$ is equal to $0$

and the maximal value of $\sqrt{ab}-\frac{a+b}{2}$ is equal to $0$ and this maximum occurs for $a=b$.

Now, let we need to find a maximum of $$(a^2-ab+b^2)(a^2-ac+c^2)(b^2-bc+c^2)$$ for non-negatives $a$, $b$ and $c$ such that $a+b+c=3$.

For $a=b=c=1$ we get a value $3$. But it's not a maximal value.

The maximal value it's $12$ and occurs for $(a,b,c)=(2,1,0)$.

  • In the second question, how did you get this intuition (that the maximal value will occur at (2,1,0). Was it just good luck, or did you follow some specific process – Tatai Sep 01 '21 at 09:33
  • @Sunaina The variables are non-negative. For $c=0$ we have $a+b=3$ and we got one variable. From here we can get a number $12$ and we can think about nice solution. It does not always exist. But for this inequality there is a very nice solution. – Michael Rozenberg Sep 01 '21 at 09:56