$a+b+c = 4$. What is the maximum value of $ab+ac+bc$?

Could this be solved by a simple application of Jensen's inequality? If so, I am unsure what to choose for $f(x)$. If $ab+ac+bc$ is treated as a function of $a$ there seems no easy way to express $bc$ in terms of $a$.

EDIT: The context of the question is maximising the surface area of a rectangular prism. Also I might have misinterpreted the question, because it says "the sum of the length of the edges (side lengths are a,b,c) is 4", and gives the options $\frac1{3}, \frac{2}{3}, 1, \frac{4}{3}$.

Otherwise, how would this be done?

  • $\begingroup$ How about mathworld.wolfram.com/LagrangeMultiplier.html $\endgroup$ – lab bhattacharjee Sep 30 '14 at 3:06
  • 1
    $\begingroup$ So you have $4(a+b+c)=4$ and you want to maximize $2(ab+bc+ca)$. That should give you $\frac23$ in the end. $\endgroup$ – Macavity Sep 30 '14 at 4:04
  • $\begingroup$ James as Macavity pointed out, the actual expression is $4(a+b+c)=4$. Draw a diagram, you will see that there are 4 sides length a, b and c each. $\endgroup$ – Sherlock Holmes Sep 30 '14 at 6:54

@James, first we square both sides:

$$(a+b+c)^2=1 \implies 1=a^2+b^2+c^2+2(ab+bc+ac)\tag{1}$$ $$\because a^2+b^2+c^2 \geq ab +ac + bc\tag{2}$$

For a,b,c>0

$$ \\ 1 \geq 3(ab+bc+ac) \tag{3}$$ $$ \frac{1}{3} \geq ab +bc+ca\tag{4} \\ \square$$

$(2)$ can be easily proven by considering: $(a-b)^2 \geq 0 \\ (b-c)^2 \geq 0 \\ (a-c)^2 \geq 0 $, and adding the 3 inequalities and rearranging.

$(3)$ follows by using $(2)$ to replace $a^2+b^2+c^2$ in $(1)$ with the smaller $ab+ac+bc$.

Surface area is $2(ab+bc+ca)$, so maximum surface area is simply $\frac{2}{3}$

  • $\begingroup$ How do you get $a^2+b^2+c^2\ge ab+ac+bc$? $\endgroup$ – robjohn Oct 2 '14 at 20:05
  • $\begingroup$ @robjohn I have edited it in. $\endgroup$ – Sherlock Holmes Oct 3 '14 at 11:35
  • $\begingroup$ I hope you don't mind. I added some to your answer to make it easier to follow. Furthermore, there is no real reason that $a,b,c\gt0$. $\endgroup$ – robjohn Oct 3 '14 at 14:38
  • $\begingroup$ @robjohn "The context of the question is maximising the surface area of a rectangular prism", so shouldn't $a,b,c>0$, or, at the very least, $a,b,c \geq 0$? $\endgroup$ – Sherlock Holmes Oct 4 '14 at 1:49
  • $\begingroup$ The question asked does not mention surface areas, only the motivation for the question asked. Another interesting bound is $$-\frac12(a^2+b^2+c^2)\le ab+bc+ca\le a^2+b^2+c^2$$ $\endgroup$ – robjohn Oct 4 '14 at 3:41

Jensen's Inequality gives $$ \left[\frac13(a+b+c)\right]^2\le\frac13(a^2+b^2+c^2) $$ and we know that $$ \begin{align} ab+bc+ca &=\frac12\left[(a+b+c)^2-(a^2+b^2+c^2)\right]\\ &\le\frac12\left[(a+b+c)^2-\frac13(a+b+c)^2\right]\\ &=\frac13(a+b+c)^2 \end{align} $$ and equality can be achieved when $a=b=c$. Therefore, if $a+b+c=4$, the maximum of $ab+bc+ca$ is $\frac{16}{3}$ which can be achieved if $a=b=c=\frac43$.


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.