2

I stumbled across this problem which goes as follows:

given |z-3| + |z| + |z+3| = 12 ,we need to find maximum and minimum value of [|z|] where [a] is greatest integer less than or equal to a.

my approach:

well for expressions like |z-b| + |z-c| = k it is easy to observe geometrically the locus as ellipse,hyperbola,pair of rays,line segment etc. and then |z| is distance of z from origin so optimizing is easy. however here I don't know the geometric interpretation of |z-3| + |z| + |z+3| = 12 so the only thing I am left with is put z = x+ iy but it still requires atleast 3 rounds of squaring and I could not optimize it that way.

hence I am stuck.Is there any geometrical essence to this expression? what is the best way to optimize this? kindly help me out.

1 Answers1

3

To get the answer without proof, this is kind of like an ellipse (in fact, it's a 3-ellipse) with an axis on the $x$-axis. So it's intuitively obvious we'll have extremes when $x=0$ or $y=0$. Below is a proof of the answer.

Particular points

Note that if $x=0$ and $y>0$ we have $12=|z-3|+|z|+|z+3|=|y|+2\sqrt{9+y^2}=y+2\sqrt{9+y^2}$. Thus, $(12-y)^2=4(9+y^2)$ so $144-24y+y^2=36+4y^2$, so that $y^2+8y-36=0$. Combined with $y>0$ we have $y=2(\sqrt{13}-2)<2(\sqrt{16}-2)=4$ so that $|z|<4$ at this point.

And if $y=0$ and $x>0$ we have $12=|x-3|+x+x+3$. If there is a solution with $x\ge3$, then $12=3x$ and we must have $x=4$, which works. So we have $|z|=4$ at this point.

Upper bound and max

Note that the triangle inequality tells us that $12=|z-3|+|z|+|z+3|\ge|z|+|(z-3)+(z+3)|=|z|+|2z|=3|z|$, so that $|z|\le 4$ no matter what. Therefore, $z=4$ attains the maximum $[|z|]=4$.

Lower bound and min

I suspect there is a more elegant way to show the minimum, but I couldn't find one in a reasonable amount of time.

We will show that $12=|z-3|+|z|+|z+3|$ does not intersect the disk $|z|\le3$, by finding where $|z-3|+|z|+|z+3|$ is maximized on the disk and that the maximum value is less than $12$.

Note that the The $y$-partial of $|z-3|+|z|+|z+3|$ is $\dfrac{y}{|z-3|}+\dfrac{y}{|z|}+\dfrac{y}{|z+3|}$ $=y*\dfrac{|z+3||z|+|z-3||z+3|+|z-3||z|}{|z-3||z||z+3|}$. So a potential critical point where this is $0$ or undefined has to have $y=0$ or $z$ at $0$ or $\pm3$ (all of which have $y=0$ anyway), since the three numerator terms are all nonnegative and can't be $0$ simultaneously.

Since we optimize by looking at critical points or the boundary, $|z-3|+|z|+|z+3|$ is maximized over the disk when $y=0$ or when $|z|=3$, In the former case, $|z-3|+|z|+|z+3|=|x|+3-x+x+3=|x|+6\le|z|+6\le3+6=9$. In the latter case, we can use $(x,y)=3(\cos t,\sin t)$ to reduce to a single variable optimization or Lagrange Multipliers. Either way, we find that we need to look at $(\pm3,0)$ and $(0,\pm3)$. We already looked at the points on the $x$-axis, and at $(0,\pm3)$ we have $|z-3|+|z|+|z+3|=2*3\sqrt{2}+3<6*\frac32+3=12$.

Therefore, $|z|\le3$ is impossible, so the lowest possible value of $[|z|]$ is $3$, occurring when $|z|<4$. We found a point with $|z|<4$ already, so the minimum value is indeed $3$.

Mark S.
  • 25,893
  • +1, very nice! for the lower bound, would it be simpler to show non-intersection with the $|z|=3$ circle, instead of the disk $|z| \le 3$? assuming they intersect reduces to $|z+3| + |z-3| = 9$ which is an ellipse and a quick contradiction. admittedly, this argument assumes the original curve is a closed loop... – antkam May 03 '20 at 22:21
  • @antkam I like the idea. Do you have a quick proof that the original curve is connected (e.g. it's not two or three closed loops)? And do you have a way to get that "quick contradiction" from the ellipse that is faster than the $|z|=3$ calculus work I skipped over in my answer? – Mark S. May 03 '20 at 23:58
  • @antkam I see, the algebra to show $|z|=3$ and $|z+3|+|z-3|=9$ has no real solutions is a hair easier. – Mark S. May 04 '20 at 00:09
  • 1
    sorry i dont have a quick way to show the original 3-ellipse is a single closed loop. assuming it is, and further assuming it intersects the circle, then the intersections must be on the 2-ellipse, but this 2-ellipse's minor axis length $> 6$, so it cant intersect the circle (i think that's a well known fact about 2-ellipses, right?) $\implies$ a contradiction is reached. – antkam May 04 '20 at 04:29