0
$\begingroup$

Given the expression defined for $x,y \in {[}0,2\pi{)} $ $$\begin{align} E(x,y)&=\cos(2x)+\cos(2y)+\cos(2x+2y) \\ &+4\cos{x}+4\cos y+4\cos(x+y)\\ &-2\cos(x-y)-2\cos(2x+y)-2\cos(2y+x) \end{align}$$ Wolfram-Alpha tells me that the minimum of this multivariable expression is $-\frac{27}{2}$, for $x=y=\frac{2\pi}{3}$.

How can this be proved (easily)?

I've tried using multivariable calculus but the calculations are really "messy". Using formulae that transform sums of trigonometric functions to products did not help either, in my case.

$\endgroup$

1 Answer 1

Reset to default
2
$\begingroup$

It may not be very simple but i'll try.

First observe that it's a continuous function and it should have minimum and maximum on a closed interval. Next, observe that the function is symmetrical relative to $x$ and $y$ so if there is a minimum inside of the interval it will be achieved at some point $(a,a)$ because partial derivatives will be symmetrical as well. So we just need to find minimum of $$f(x)=E(x,x)=\cos 4x-4\cos 3x+6\cos 2x+8\cos x-2$$ Using multiple angle formulas, we can write this function as a polynomial of $\cos x$: $$f(x)=8\cos^4x-16\cos^3x+4\cos^2x+20\cos x-7$$ Let's find derivative $f'(x)=-\sin x \cdot (32\cos^3x-48\cos^2x+8\cos x+20)$ or $f'(x)=-4\sin x \cdot(2\cos x+1)(4\cos^2x-8\cos x+5)$.

We have two critical points to check: $x=0$ and $\cos x=-0.5$ ((plus the boundaries).

$f(0)=f(2\pi)=9$, when $x_m=\frac{2\pi}{3}$ we have our minimum $f(x_m)=-\frac{27}{2}$

$\endgroup$
2
  • $\begingroup$ Thank you very much for this, however I don't completely understand the reasoning behind the fact that $x=y$, can't the partial derivative be $0$ (condition for maximum or minimum) in a point where $x \ne y$, for example $(x,y)$ - and then the partial derivative will be $0$ for the point at $(y,x)$ too. Could you help me clear that out? $\endgroup$
    – Parallelism Alert
    Jul 17, 2019 at 10:06
  • $\begingroup$ @ParallelismAlert: This is a very good and a valid point and there is no easy answer. If you want to learn more, look up Purkiss Principle and these references: math.stackexchange.com/questions/681153/… mathoverflow.net/questions/58721/… In this particular case the function is convex, that's the key point here. $\endgroup$
    – Vasili
    Jul 17, 2019 at 12:07

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.