-1

Please help to obtain.. in as elegant a form as possible.. the locus of point $P$ equations if its distances to two circles:

$$ (x-h)^2 + y^2 = a^2;\;(x+h)^2 + y^2 = b^2 ;\;$$

are in a constant ratio $e,$ or

$$\dfrac{ \sqrt{(x + h)^2 + y^2} - a} { \sqrt{(x - h)^2 + y^2} - b}=e.$$

I am looking at a generalization to conic sections with $ (0,\infty)$ domains for circle radii $(a,b)$.

When $ a=0,b=0 $ we have Apollonius Circles as loci. When $ a > 0,b >0 $ the fixed base focal points of Apollonius Circles are expanded to bigger circles and the circles themselves become distorted as shown below.

Required to obtain a parametrization of the Ovals or equations in terms of any known functions.

The set of new Ovals shown have been obtained by contour plots. Distances are positive when normal distance from point P is outside a circle ( and so $e$), and negative when inside.

EDIT1:

New Ovals const Dist ratio to two circles

Constants used for plot of new Ovals loci when $(2h < a+b)$ for intersecting circles:

$$\;h=1.8,\;a=3,\;b=2,\;e= (-2,-1,0,1,4).$$

In particular I wish to know if the case of equal distances $e=-1$ is an ellipse or if the circle centers are its foci.

Narasimham
  • 40,495
  • Does $e = \frac{\text{length from circle 1}}{\text{length from circle 2}} :$ or $: e = \frac{\text{length from circle 2}}{\text{length from circle 1}} :$ –  Jul 13 '20 at 20:47
  • The former. In new plot/ graph it is clearly seen for case $ e=4>1 $. There is a small error in the plot , shall correct it later. – Narasimham Jul 14 '20 at 15:06
  • Adding the parameterization requirement after @arthur has already put in so much effort on this problem (including addressing additional requests in comments) seems inconsiderate. It would be more appropriate to accept arthur's answer as-is, then ask about parameterizing the ovals in a separate question. – Blue Jul 14 '20 at 15:59
  • Indeed that would be proper after his several responses. As you suggest we go the whole hog in parameterizing etc.in a later question. Did you find it interesting? – Narasimham Jul 14 '20 at 20:52
  • Are you trying to work out a problem in electrostatics, by any chance? – rf1x Jul 17 '20 at 10:24
  • For equipotential lines between differently charged cylinders is one possibility. – Narasimham Jul 17 '20 at 10:46

1 Answers1

3

enter image description here

Let the center of $C1$ be $(x1,y1)$ and radius $r1$.

The center of $C2$ is $(x2,y2)$ and radius $r2$.

Z is the locus $(x,y)$.

The distance from $Z$ to the center of $C1$ is:

$$D1 = \sqrt{(x - x1)^2 + (y-y1)^2} \tag{1}$$

The closest point from $C1$ to $Z$ is extended from the radius.

The point on the perimeter is $P1$.

The distance from the perimeter of $C1$ to $Z$ is $L1$, from $P1$ to $Z$:

$$L1 = \left\lvert \sqrt{(x - x1)^2 + (y-y1)^2} - r1 \right\rvert \tag{2}$$

Similarly for $C2$:

$$L2 = \left\lvert \sqrt{(x - x2)^2 + (y-y2)^2} - r2 \right\rvert \tag{3}$$

$Z$ is at a distance ratio $e$ : $L1 = e L2$

$$\left\lvert \sqrt{(x - x1)^2 + (y-y1)^2} - r1 \right\rvert = e \left\lvert \sqrt{(x - x2)^2 + (y-y2)^2} - r2 \right\rvert \tag{4}$$

$(x1,y1) = (h,0)$ , $r1 = |a|$.

$(x2,y2) = (-h,0)$, $r2 = |b|$

The sign changes if $Z$ is inside the circle.


Let

$S1 = \sqrt{(x - x1)^2 + (y-y1)^2} \tag{5}$

$S2 = \sqrt{(x - x2)^2 + (y-y2)^2} \tag{6}$

For $Z$ outside both circles the absolute signs are both $+$.

$S1 - r1 = e(S2 - r2) \tag{7}$

$S1-e S2 = r1 - e r2 \tag{8}$

$(S1-e S2)^2 = (r1 - er2)^2 \tag{9}$

$S1^2 + e^2 S2^2 -2 e S1 S2 = (r1 - e r2)^2 \tag{10}$

$S1^2 + e^2 S2^2 - (r1 - e r2)^2 = 2 e S1 S2 \tag{11}$

$(S1^2 + e^2 S2^2 - (r1 - e r2)^2)^2 = 4 e^2 S1^2 S2^2 \tag{12}$

$(S1^2 + e^2 S2^2 - (r1 - e r2)^2)^2 - 4 e^2 S1^2 S2^2 = 0 \tag{13}$

Maxima code:

S1 : sqrt((x-x1)^2 + (y - y1)^2);
S2 : sqrt((x-x2)^2 + (y - y2)^2);
R : r1 - e*r2;
E1 : (S1^2 + e^2*S2^2 - R^2)^2 - 4*e^2*S1^2*S2^2;
E2 : expand(E1);
E3 : facsum(E2,y,y^2,y^3,y^4,x,x^2,x^3,x^4);
E4 : subst(b,r2,subst(a,r1,subst(0,y2,subst(-h,x2,subst(0,y1,subst(h,x1,E2))))));
E5 : facsum(E4,y,y^2,y^3,y^4,x,x^2,x^3,x^4);
tex(E5);

Finally with all the substitutions:

$$\left(e-1\right)^2\,\left(e+1\right)^2\,y^4+2\,\left(e-1\right)^2\, \left(e+1\right)^2\,x^2\,y^2+4\,\left(e-1\right)\,\left(e+1\right)\, \left(e^2+1\right)\,h\,x\,y^2+2\,\left(e^4\,h^2-2\,e^2\,h^2+h^2-b^2 \,e^4+2\,a\,b\,e^3-b^2\,e^2-a^2\,e^2+2\,a\,b\,e-a^2\right)\,y^2+ \left(e-1\right)^2\,\left(e+1\right)^2\,x^4+4\,\left(e-1\right)\, \left(e+1\right)\,\left(e^2+1\right)\,h\,x^3+2\,\left(3\,e^4\,h^2+2 \,e^2\,h^2+3\,h^2-b^2\,e^4+2\,a\,b\,e^3-b^2\,e^2-a^2\,e^2+2\,a\,b\,e -a^2\right)\,x^2+4\,\left(e-1\right)\,\left(e+1\right)\,h\,\left(e^2 \,h^2+h^2-b^2\,e^2+2\,a\,b\,e-a^2\right)\,x+\left(e\,h-h-b\,e+a \right)\,\left(e\,h-h+b\,e-a\right)\,\left(e\,h+h-b\,e+a\right)\, \left(e\,h+h+b\,e-a\right) = 0$$


Maxima can solve quartics:

Yall : solve(E5,y);
Y1 : part(Yall,1)^2;
Y2 : part(Yall,2)^2;
Y3 : part(Yall,3)^2;
Y4 : part(Yall,4)^2;

tex(Y1)$ tex(Y2)$ tex(Y3)$ tex(Y4)$


$$y^2={{2\,b\,e^2\,\sqrt{-4\,e^2\,h\,x+4\,h\,x+b^2\,e^2-2\,a\,b\,e+a^ 2}-2\,a\,e\,\sqrt{-4\,e^2\,h\,x+4\,h\,x+b^2\,e^2-2\,a\,b\,e+a^2}-e^4 \,x^2+2\,e^2\,x^2-x^2-2\,e^4\,h\,x+2\,h\,x-e^4\,h^2+2\,e^2\,h^2-h^2+ b^2\,e^4-2\,a\,b\,e^3+b^2\,e^2+a^2\,e^2-2\,a\,b\,e+a^2}\over{\left(e ^2-1\right)^2}}$$


$$y^2={{2\,b\,e^2\,\sqrt{-4\,e^2\,h\,x+4\,h\,x+b^2\,e^2-2\,a\,b\,e+a^ 2}-2\,a\,e\,\sqrt{-4\,e^2\,h\,x+4\,h\,x+b^2\,e^2-2\,a\,b\,e+a^2}-e^4 \,x^2+2\,e^2\,x^2-x^2-2\,e^4\,h\,x+2\,h\,x-e^4\,h^2+2\,e^2\,h^2-h^2+ b^2\,e^4-2\,a\,b\,e^3+b^2\,e^2+a^2\,e^2-2\,a\,b\,e+a^2}\over{\left(e ^2-1\right)^2}}$$


$$y^2={{-2\,b\,e^2\,\sqrt{-4\,e^2\,h\,x+4\,h\,x+b^2\,e^2-2\,a\,b\,e+a ^2}+2\,a\,e\,\sqrt{-4\,e^2\,h\,x+4\,h\,x+b^2\,e^2-2\,a\,b\,e+a^2}-e^ 4\,x^2+2\,e^2\,x^2-x^2-2\,e^4\,h\,x+2\,h\,x-e^4\,h^2+2\,e^2\,h^2-h^2 +b^2\,e^4-2\,a\,b\,e^3+b^2\,e^2+a^2\,e^2-2\,a\,b\,e+a^2}\over{\left( e^2-1\right)^2}}$$


$$y^2={{-2\,b\,e^2\,\sqrt{-4\,e^2\,h\,x+4\,h\,x+b^2\,e^2-2\,a\,b\,e+a ^2}+2\,a\,e\,\sqrt{-4\,e^2\,h\,x+4\,h\,x+b^2\,e^2-2\,a\,b\,e+a^2}-e^ 4\,x^2+2\,e^2\,x^2-x^2-2\,e^4\,h\,x+2\,h\,x-e^4\,h^2+2\,e^2\,h^2-h^2 +b^2\,e^4-2\,a\,b\,e^3+b^2\,e^2+a^2\,e^2-2\,a\,b\,e+a^2}\over{\left( e^2-1\right)^2}}$$


There is an $x$ under the square root?.

Does this mean the general case is not a conic section?


For $a= 0$, $b = 0$

Maxima code:

E6: subst(0,b,subst(0,a,E5));
tex(E6);

$$\left(e-1\right)^2\,\left(e+1\right)^2\,y^4+2\,\left(e-1\right)^2\, \left(e+1\right)^2\,x^2\,y^2+4\,\left(e-1\right)\,\left(e+1\right)\, \left(e^2+1\right)\,h\,x\,y^2+2\,\left(e^4\,h^2-2\,e^2\,h^2+h^2 \right)\,y^2+\left(e-1\right)^2\,\left(e+1\right)^2\,x^4+4\,\left(e- 1\right)\,\left(e+1\right)\,\left(e^2+1\right)\,h\,x^3+2\,\left(3\,e ^4\,h^2+2\,e^2\,h^2+3\,h^2\right)\,x^2+4\,\left(e-1\right)\,\left(e+ 1\right)\,h\,\left(e^2\,h^2+h^2\right)\,x+\left(e\,h-h\right)^2\, \left(e\,h+h\right)^2 = 0$$


A circle is expected for the previous equation:

Try to fit the form : ${a}^2((y -y_0)^2 + (x-x_0)^2 - r^2)^2 = 0$ to it:

Expand this expression and equate the coefficients of each $x^ny^m$:

Maxima code (all variables are solved by calculation):

X1 : a^2*( (y - y0)^2 + (x - x0)^2 - r^2)^2;
X2 : expand(X1);
X3 : facsum(X2,y,y^2,y^3,y^4,x,x^2,x^3,x^4);

X4 : X3 - E6; X5 : facsum(X4,y,y^2,y^3,y^4,x,x^2,x^3,x^4);

X6 : subst(0,y0,X5);

Yx0 : solve(part(X6,3),x0);

X7 : facsum(subst(part(Yx0,1,2),x0,X6),y,y^2,y^3,y^4,x,x^2,x^3,x^4);

Yr2 : solve(part(X7,1,6),[r^2]);

X8 : subst(part(Yr2,1,2),r^2,X7); X9 : facsum(expand(X8),y,y^2,y^3,y^4,x,x^2,x^3,x^4);

Ya2 : solve(expand(part(X9,1,1)/a^2/y^4),[a^2]);

X10 : subst(part(Ya2,1,2),a^2,expand(X9));

X11 : facsum(expand(X10),y,y^2,y^3,y^4,x,x^2,x^3,x^4);

Yx0 : ratsimp(subst(part(Ya2,1,2),a^2,Yx0));

Yr2 : ratsimp(subst(part(Ya2,1,2)^2,a^4,Yr2));

EQN : y^2 + (x - part(Yx0,1,2))^2 = part(Yr2,1,2);

The substitutions were:

$$y0 = 0 \tag{14}$$

$${\it x_0}=-{{\left(e^2+1\right)\,h}\over{e^2-1}} \tag{15}$$

$$r^2={{4\,e^2\,h^2}\over{e^4-2\,e^2+1}} \tag{16}$$

$$a^2=e^4-2\,e^2+1 \tag{17}$$

The result equation is:

$$y^2+\left(x+{{\left(e^2+1\right)\,h}\over{e^2-1}}\right)^2={{4\,e^2 \,h^2}\over{e^4-2\,e^2+1}} \tag{18}$$

Centers $(h,0)$ and $(0,0)$ and inverting $\displaystyle e \rightarrow \frac1{e} \:$ produced the standard form:

$$\boxed{ y^2+\left(x+{{he^2}\over{1-e^2}}\right)^2={{e^2\,h^2}\over{(1 - e^2)^2}}} \tag{19}$$

Maxima centers $(0,0)$ $(h,0)$ , $eL1 = L2$

S1 : sqrt((x-x1)^2 + (y - y1)^2);
S2 : sqrt((x-x2)^2 + (y - y2)^2);
R : e*r1 - r2;
E1 : (e^2*S1^2 + S2^2 - R^2)^2 - 4*e^2*S1^2*S2^2;
E2 : expand(E1);
E3 : facsum(E2,y,y^2,y^3,y^4,x,x^2,x^3,x^4);
E4 : subst(b,r2,subst(a,r1,subst(0,y2,subst(0,x2,subst(0,y1,subst(h,x1,E2))))));
E5 : facsum(E4,y,y^2,y^3,y^4,x,x^2,x^3,x^4);
tex(E5);

E6: subst(0,b,subst(0,a,E5)); tex(E6);

X1 : a^2*( (y - y0)^2 + (x - x0)^2 - r^2)^2; X2 : expand(X1); X3 : facsum(X2,y,y^2,y^3,y^4,x,x^2,x^3,x^4);

X4 : X3 - E6; X5 : facsum(X4,y,y^2,y^3,y^4,x,x^2,x^3,x^4);

X6 : subst(0,y0,X5);

Yx0 : solve(part(X6,3),x0);

X7 : facsum(subst(part(Yx0,1,2),x0,X6),y,y^2,y^3,y^4,x,x^2,x^3,x^4);

Yr2 : solve(part(X7,1,6),[r^2]);

X8 : subst(part(Yr2,1,2),r^2,X7); X9 : facsum(expand(X8),y,y^2,y^3,y^4,x,x^2,x^3,x^4);

Ya2 : solve(expand(part(X9,1,1)/a^2/y^4),[a^2]);

X10 : subst(part(Ya2,1,2),a^2,expand(X9));

X11 : facsum(expand(X10),y,y^2,y^3,y^4,x,x^2,x^3,x^4);

Yx0 : ratsimp(subst(part(Ya2,1,2),a^2,Yx0));

Yr2 : ratsimp(subst(part(Ya2,1,2)^2,a^4,Yr2));

EQN : y^2 + (x - part(Yx0,1,2))^2 = part(Yr2,1,2);

tex(%);

$$y^2+\left(x-{{e^2\,h}\over{e^2-1}}\right)^2={{e^2\,h^2}\over{e^4-2 \,e^2+1}} \tag{20}$$

  • Thanks, is it still possible to arrive at a conic-like recognizable form after the radicals are removed by squaring? – Narasimham Jul 13 '20 at 12:07
  • I can add that it will be messy. –  Jul 13 '20 at 12:09
  • Thanks. Looking for a form comparable to the canonical. – Narasimham Jul 13 '20 at 12:15
  • If the solution is in terms of $x^4$ and $y^4$ I'm not sure it can be reduced. –  Jul 13 '20 at 12:16
  • One can make some minor refinements, but I don't see a particularly "elegant" way to write the equation. (The lack of horizontal symmetry precludes anything truly "comparable" to a conic.) ... Just to mention: Your equation $(8)$ can be written $$-s_1+es_2+k=0$$ A straightforward way to eliminate the square roots is to complete what I call the "Heronic product", which in general looks like $$(a+b+c)(-a+b+c)(a-b+c)(a+b-c)$$ and expands to $$-a^4-b^4-c^4+2a^2b^2+2b^2c^2+2c^2a^2$$ As all elements are raised to even powers, any square-root elements yield regular polynomial expressions. – Blue Jul 13 '20 at 13:44
  • @Blue - thanks, I was not aware of that technique for removing square roots. –  Jul 13 '20 at 13:54
  • @arthur: As a "technique", I don't think it's all that well-known. I see the Heronic product a lot in my research, so I've come to appreciate how it eliminates square roots, and thus I like to pass this knowledge along. :) This answer of mine notes the meaningfulness of the Heronic product in deriving the standard eqn for the ellipse/hyperbola. (BTW: "Heronic" recalls Heron's formula for the area of a triangle: $A^2=s(s-a)(s-b)(s-c)$ where $s:=(a+b+c)/2$ is the semi-perimeter. Without $s$, this is $16A^2=(a+b+c)(-a+b+c)(a-b+c)(a+b-c)$.) – Blue Jul 13 '20 at 14:08
  • @arthur:Finally do you get the special case of the Apollonius circle ? – Narasimham Jul 13 '20 at 14:13
  • @Narasimham : No. I can't see it. I had a quick look at wikipedia but I'm not familiar with Apollonius circle. –  Jul 13 '20 at 14:21
  • Fixed ratio of distances to two fixed points is an Apollonius circle. If one circle is replaced by a straight line and the other circle shrinks to a point ( its center ) then the locus is a conic. – Narasimham Jul 13 '20 at 14:30
  • @Narasimham : thats interesting. Do you think the last equation can be reduced/factored to a circle? –  Jul 13 '20 at 14:32
  • @arthur: I suppose it can be simplified. Do the coefficients of $ (x^4,y^4 )$ vanish on simplification? – Narasimham Jul 13 '20 at 14:37
  • @Narasimham : I think the form could be ${\alpha}^2((y -y_0)^2 + (x-x_0)^2 - r^2)^2 = 0$ –  Jul 13 '20 at 14:45
  • The circle centers are $(h,0)$ and $(-h,0)$, the equations might match the standard form if the centers were $(h,0)$ and $(0,0)$. –  Jul 13 '20 at 18:30
  • Centers $(h,0)$ and $(0,0)$ produced: $$y^2+\left(x+{{h}\over{e^2-1}}\right)^2={{e^2,h^2}\over{e^4-2,e^2+ 1}}$$ –  Jul 13 '20 at 20:04
  • @arthur:Thanks for your careful replies. Seeing that it is not a conic curve in general, we should take the algebraic curve's equation/ parametrization further on in another linked question that I would put in shortly. – Narasimham Jul 14 '20 at 21:07
  • @Narasimham : thanks. I'm using maxima more often and letting it do the error prone algebra. –  Jul 14 '20 at 21:11