2
$\begingroup$

The circles are all four combinations of $(x\pm60)^2+(y\pm25)^2=5^2$ (see pic at end).

The ellipse I've got is one I found via trial and error but there must be an analytical way to solve this, right?

I looked at Find the area of largest rectangle that can be inscribed in an ellipse - and the $a=x\sqrt 2$ and $b=y\sqrt 2$ relationships seem important but I don't understand how they were derived (It seems as though many steps were omitted because it felt disjointed) nor how they could be applied to my situation.

Ellipse and Four Circles Pic

$\endgroup$
3
  • 2
    $\begingroup$ Why the uncommented downvotes? It is a perfectly reasonable question. $\endgroup$
    – almagest
    Commented May 22, 2016 at 10:52
  • 1
    $\begingroup$ @almagest: And, even more worryingly, the $4$ uncommented close votes. $\endgroup$
    – joriki
    Commented May 23, 2016 at 21:43
  • $\begingroup$ This is a complex problem and needs the solution of a quartic... at least that's what I found. $\endgroup$
    – N74
    Commented May 24, 2016 at 12:11

2 Answers 2

1
$\begingroup$

The first conclusion you have to make is that the problem is symmetric on both the $x$ and the $y$ axis, so the equation of the ellipse we want to find is of the form $$ {x^2 \over a^2}+{y^2 \over b^2}=1.$$

To find a solution to this problem we can, then, force the ellipse to be tangent to one of the circles, and the ellipse will be tangent to all the circles.

Let's focus on the circle in the first quadrant. $x_0=60$ and $y_0=25$ are the coordinates of its center and the radius is $r=5$. We can see that, if the ellipse touches the circle at $x=x_0+r$ the conic becomes a plane strip $-x_0-r \le x \le x_o+r$. In the same way, if the ellipse touches the circle at $y=y_0+r$ the conic becomes the plane strip $-y_0-r \le y \le y_0+r$.

So, let's take a point on the circle: $$x_c=x_0+r \cos \phi, y_c=y_0+r \sin \phi$$ $\phi$ has to be limited in the range $0 < \phi < {\pi \over 2}$, as we have seen that the ellipse degenerates if we pick a point on the extremes.

To force the ellipse to be tangent to the circle we will put: $$ {x_c^2 \over a^2}+{y_c^2 \over b^2}=1$$ (as the point on the circle has to be also on the ellipse).

The line tangent to the ellipse in $(x_c, y_c)$ has the equation $${x \cdot x_c \over a^2}+{y \cdot y_c \over b^2}=1 \Leftrightarrow y={b^2 \over y_c}-x{x_c \over y_c}{b^2 \over a^2}$$

The line tangent to the circle in $(x_c, y_c)$ is in the form $$y=k-{1 \over \tan \phi}x$$ so, equating the angular coefficients, we have: $${1 \over \tan \phi}={x_c \over y_c}{b^2 \over a^2}.$$

This, together with $$ {x_c^2 \over a^2}+{y_c^2 \over b^2}=1$$ gives us $$a^2=x_c^2+x_c y_c \tan\phi$$ and $$b^2=y_c^2+{x_c y_c \over \tan\phi}.$$

So we can draw an ellipse tangent to the four circles for each value of $\phi$

We have to minimize the area: that is $A=\pi a b$. To make things easier we will minimize $$f(\phi)=a^2b^2$$ we can do so as the areas are always positive, so squaring will not add minima.

Substituting everything into the equation for $f$ and differentiating in the $\phi$ variable, Wolfram Alfa gives us the nice equation: $$f'(\phi)={1 \over 64 } \csc^2 \left( {\pi \over 4}- {\phi \over 2} \right)\csc^2 \left( {\pi \over 2}+ {\phi \over 2} \right)\csc^2 \left( {\phi \over 2} \right)\sec^2 \left( {\phi \over 2} \right)\left( r+x_0 \cos \phi + y_0 \sin \phi \right)\left( -3rx_0 \sin \phi - rx_0 \sin 3 \phi -3ry_0 \cos \phi -ry_0 \cos 3 \phi -4x_0y_0 \right)\left( r \cos 2 \phi + x_0 \cos \phi - y_0 \sin \phi \right)$$

The only term that changes sign in this equation in our range of interest is the last one, so the ellipse with the minimum area (we know it is a minima as the area becomes plus infinite at the extremes of the range) has $\phi$ that is solution to $$r \cos 2 \phi + x_0 \cos \phi - y_0 \sin \phi = 0.$$ This leads to the solution of a fourth order equation, so I will not give you its closed-form solution.

The approximate value, for your numbers, is $\phi = 0.56369$ giving an area $A=13612.7$.

EDIT. It seems Jens is true. I trusted too much in Wolfram Alfa and did not double check my values.

Anyway the equation to solve was the right one and its solution gives the correct value of $\phi = 1.127$ thus implying $A=11525.3$.

Graph of the derivative and the area

In the graph a plot of the derivative (only the sign changing part) and of the area (scaled by 100) is shown. Just to show that the minima and the zeros are aligned.

$\endgroup$
6
  • $\begingroup$ Why does the range stop at ${\pi \over 4}$? Wouldn't $y=y_0+r$ occur at ${\pi \over 2}$? $\endgroup$
    – Jens
    Commented May 24, 2016 at 15:20
  • $\begingroup$ @Jens You're right, a cut and paste error. I'll correct it. $\endgroup$
    – N74
    Commented May 24, 2016 at 16:39
  • $\begingroup$ Does your assumption regarding which term changes sign in the Wolfram Alfa equation still hold true? $\endgroup$
    – Jens
    Commented May 24, 2016 at 17:57
  • $\begingroup$ @Jens Yes: they were stated assuming both $\sin$ and $\cos$ positive. $\endgroup$
    – N74
    Commented May 24, 2016 at 17:59
  • $\begingroup$ I'm convinced. Very well done! $\endgroup$
    – Jens
    Commented May 24, 2016 at 18:56
0
$\begingroup$

This is not an answer but an illustration of the correct ellipse based on N74's answer. It seems the OP's ellipse was in fact right on target. Below is the ellipse I get when the double of the angle given by N74 is used. The area $A = 11525.27$. Circle and ellipse2

$\endgroup$
2
  • $\begingroup$ I was wrong also about the value of the solution of the quartic equation... I changed my answer and credited you for your help. $\endgroup$
    – N74
    Commented May 25, 2016 at 8:32
  • $\begingroup$ @N74 Thanks. I just had to implement your equations (in Excel) so I could play around with them. One last quibble: the area you quoted in your edit is missing a decimal point :-) $\endgroup$
    – Jens
    Commented May 26, 2016 at 14:55

You must log in to answer this question.

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