7
$\begingroup$

Let $ABCD$ be a regular tetrahedron with center $O.$ Consider two points $M,N,$ such that $\overrightarrow{NO}=-3\overrightarrow{MO}.$ Prove or disprove that $$NA+NB+NC+ND\geq MA+MB+MC+MD$$

I tried to use CS in the Euclidean space $E_3$, but it does not help, because the minoration is too wide.
Note: I also posted this on the Mathematics Stack Exchange, but not much progress has been made on this question. This is why I thought that posting here too would be all right (this problem is open in the sense that its proposer doesn't have a proof, so I guess it is fit the for this forum).
EDIT: The bounty expired, so this may be reopened.

$\endgroup$
10
  • $\begingroup$ what are these $NA$, $NB$, etc.? Distances? $\endgroup$ Commented Dec 30, 2019 at 16:12
  • $\begingroup$ @DimaPasechnik Yes, they are distances. $\endgroup$
    – JoMath
    Commented Dec 30, 2019 at 18:44
  • $\begingroup$ I think it’s inappropriate to post this here while there’s a bounty on the question at the other site, with upvotes and comments and an answer. $\endgroup$
    – user44143
    Commented Dec 30, 2019 at 19:45
  • $\begingroup$ @MattF. I pondered over this aspect before posting, as I know that crossposting is not encouraged, but I reached the conclusion that it would be all right since it has been there for a while and little progress has been made(the only answer there was for the 2D case and I encouraged that user to post it since I had got no response to the problem and I thought that maybe it would somehow prove to be useful for the 3D case in which I am interested). I hoped that by posting here I would draw more attention to the problem. I am sorry if I were wrong, but I find this problem really interesting. $\endgroup$
    – JoMath
    Commented Dec 30, 2019 at 20:06
  • 2
    $\begingroup$ This may be reopened after the bounty expires. $\endgroup$ Commented Dec 30, 2019 at 22:49

1 Answer 1

8
$\begingroup$

Following suggestions on Stack Exchange, we use the homothety with respect to $O$ and factor $-3$. If $X$ is a point, then $X'$ denotes the point for which $\overrightarrow{XO}= -3 \overrightarrow{X'O}$. So $M=N'$ and $A'$ is the midpoint of the face opposite to $A$. One has $XY=3X'Y'$ and the desired inequality becomes $3MA'+3MB'+3MC'+3MD'\geq MA+MB+MC+MD$.

It thus suffices to prove inequalities of the type $MB'+MC'+MD'\geq MA$. Thus put $f(X)=XB'+XC'+XD'-XA$. We have to show that $f(M)\geq0$.

Let us fix our tetrahedron as follows.

$A=\left(0,0,\sqrt{\frac{2}{3}}-\frac{1}{2 \sqrt{6}}\right)$,

$B= \left(-\frac{1}{2 \sqrt{3}},-\frac{1}{2},-\frac{1}{2 \sqrt{6}}\right)$,

$C=\left(-\frac{1}{2 \sqrt{3}},\frac{1}{2},-\frac{1}{2 \sqrt{6}}\right)$,

$D=\left(-\frac{1}{2 \sqrt{3}},\frac{1}{2},-\frac{1}{2 \sqrt{6}}\right)$.

Put $f_1(X)=XB'+XC'$, $f_2(X)=XD'-XA$, so that $f(X)=f_1(X)+f_2(X)$.

Notice that $f(O)=0$ and notice that $f(M)$ is positive if $M$ is far away from $O$.

Now take $M$ so that $f(M)$ is an absolute minimum. Note that $f$ is positive at $A$, $B'$, $C'$, $D'$, so that $M$ is none of those points. We have $f_2(M)<0$ and the gradients of $f_1$, $f_2$ cancel each other at $M$. That means that the level sets through $M$ of $f_1$ and $f_2$ touch at $M$. (There is no point where both gradients vanish.)

The level set through $M$ of $f_1$ is an ellipsoid, with $f_1$ smaller inside, and the level set through $M$ of $f_2$ is the lower sheet of a hyperboloid of two sheets, with $f_2$ more negative inside the sheet. As each level set is the curved boundary of a convex region, there is no other point where the two level sets touch.

The level sets are symmetric with respect to the plane $ \left\{X\mid XB'=XC' \right\}$, so $M$ must lie on that plane. We have shown that $MB'=MC'$. Similarly $MC'=MD'$ and $M$ must be of the form $(0,0,x)$, so that $f(M)=\frac{\sqrt{72 x^2-4 \sqrt{6} x+3}-\sqrt{8 x^2-4 \sqrt{6} x+3}}{2 \sqrt{2}}\geq0$.

$\endgroup$
4
  • $\begingroup$ Why is $f_2(M)<0$ in your proof ? I see that $f_2$ can take negative values, but couldn't it be the case that the minimum of $f$ is reached at some point $M$ with $f_1(M)$ very small and $f_2(M)$ non-negative ? $\endgroup$
    – M. Dus
    Commented Jan 8, 2020 at 10:38
  • $\begingroup$ @M.Dus. As it is a sum of two lengths, $f_1$ is positive. $\endgroup$ Commented Jan 8, 2020 at 12:44
  • $\begingroup$ Of course, but what I had not seen is that $f(O)=0$ so that you know the minimum is non-positive. Thank you anyway $\endgroup$
    – M. Dus
    Commented Jan 8, 2020 at 14:24
  • $\begingroup$ I have clarified the text. $\endgroup$ Commented Jan 9, 2020 at 8:32

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .