5

I searched around but was unable to find anything.

For the usual $2$-ellipse we have the parametrization $x(t) = a\cos(t)$ and $y(t) = b\sin(t)$ for $t\in [0,2\pi]$.

Is there anything similar for the more general $3$-ellipse or $4$-ellipse? If there general case is too difficult is there a parametrization for a constrained version, for example where some of the foci lie, say on the $x$-axis?

Edit: I mean in 2 dimensions, as here https://en.wikipedia.org/wiki/N-ellipse

Edit 2: Assume $u_1 = (-R,0)$ and $u_2 = (R,0)$ and $u_3 = (0, -H)$ are the foci of the $3$-ellipse, for $R,H > 0$, given some $d$ the $3$-ellipse is the set of points given by $$\left\{(x, y) \in \mathbf{R}^{2} : \sum_{i=1}^{3} \sqrt{\left(x-u_{i}\right)^{2}+\left(y-v_{i}\right)^{2}}=d\right\}$$

Is there a parametrization of this curve similar to the $2$-ellipse?

  • I had never heard of these multifocal ellipses, and find your question very interesting. I see that even a trifocal ellipse will be an algebraic curve of degree eight, which suggests to me that any parametrization will be extremely difficult to find. What if the foci are the vertices of an equilateral triangle? – Lubin Mar 15 '19 at 03:05
  • 1
    To avoid a confusion of the number of the foci with the number of dimensions, add a definition of n-ellipse and optionally, some examples (pictures) to the question. – g.kov Mar 15 '19 at 03:11
  • I will edit the question when I get back home, stuck with mobile for now – pureundergrad Mar 15 '19 at 10:45

1 Answers1

0

A Parametrization of the 3-Ellipse

Let $$x^2+y^2+z^2=\rho$$ and $$Q_k\equiv (u-u_k)^2+(v-v_k)^2$$ where $u,v$ are our plane variables; then we wish to parametrize $$\sqrt{Q_1}+\sqrt{Q_2}+\sqrt{Q_3}=\rho.$$ Let $$\sqrt{Q_1}=x^2\\ \sqrt{Q_2}=y^2,\\ \sqrt{Q_3}=z^2.$$

Then

$$u^2+v^2-2u_1u-2v_1v+u_1^2+v_1^2=x^4\\u^2+v^2-2u_2u-2v_2v+u_2^2+v_2^2=y^4\\u^2+v^2-2u_3u-2v_3v+u_3^2+v_3^2=z^4.$$

Subtract (2) from (1) and (3) from (2) and (1) from (3) to obtain

$$2(u_2-u_1)u+2(v_2-v_1)v+u_1^2+v_1^2-u_2^2-v_2^2=x^4-y^4,\\2(u_3-u_2)u+2(v_3-v_2)v+u_2^2+v_2^2-u_3^2-v_3^2=y^4-z^4,\\2(u_1-u_3)u+2(v_1-v_3)v+u_3^2+v_3^2-u_1^2-v_1^2=z^4-x^4.$$ From here we relabel $$U_1u+V_1v+W_1=x^4-y^4,\\U_2u+V_2v+W_2=y^4-z^4,\\U_3u+V_3v+W_3=z^4-x^4,$$ and we know that $$\begin{vmatrix}U_1&V_1&-W_1+x^4-y^4\\U_2&V_2&-W_2+y^4-z^4\\U_3&V_3&-W_3+z^4-x^4\end{vmatrix}=0$$ $$\begin{vmatrix}U_1&V_1&x^4-y^4\\U_2&V_2&y^4-z^4\\U_3&V_3&z^4-x^4\end{vmatrix}=\begin{vmatrix}U_1&V_1&W_1\\U_2&V_2&W_2\\U_3&V_3&W_3\end{vmatrix}$$ $$[U_2V_3]^-(x^4-y^4)-[U_1V_3]^-(y^4-z^4)+[U_1V_2]^{-}(z^4-x^4)=[U_1V_2W_3]^-$$ $$\frac{[U_2V_3]^--[U_1V_2]^{-}}{[U_1V_2W_3]^-}x^4-\frac{[U_1V_3]^-+[U_2V_3]^-}{[U_1V_2W_3]^-}y^4+\frac{[U_1V_2]^{-}+[U_1V_3]^-}{[U_1V_2W_3]^-}z^4=1.$$ Let's now parameterize $(x,y,z)$ using $$(x,y,z) = \sqrt{\rho}(\sin\theta\cos\phi,\sin\theta\sin\phi,\cos\theta)$$ and therefore our second equation restricts one parameter (we'll choose $\theta$):

$$\Delta_x(\sin\theta\cos\phi)^4+\Delta_y(\sin\theta\sin\phi)^4+\Delta_z(\cos\theta)^4=\frac{1}{\rho^2}$$

$$\left[\Delta_x(\cos\phi)^4+\Delta_y(\sin\phi)^4\right]\left(1-(\cos\theta)^2\right)^2+\Delta_z(\cos\theta)^4=\frac{1}{\rho^2}$$

and since the root $x$ of

$$a(1-x)^2+bx^2=c$$ is $$x_{\pm}=\frac{a\pm\sqrt{(a+b)c-ab}}{a+b}$$ we have $$\cos\theta=\sqrt{\frac{\rho(\cos\phi)^2(\Delta_x+\Delta_y(\tan\phi)^4)+\sqrt{(\Delta_x+\Delta_y(\tan\phi)^4+\Delta_z(\sec\phi)^4)-\rho^2(\Delta_x+\Delta_y(\tan\phi)^4)\Delta_z}}{\rho(\cos\phi)^2(\Delta_x+\Delta_y(\tan\phi)^4+\Delta_z(\sec\phi)^4)}}$$ for which $z=\sqrt{\rho}\cos\theta,$ and from that we can derive $\sin\theta=\sqrt{1-\cos^2\theta}.$ Thus our $u,v$ can be obtained from

$$U_1u+V_1v+W_1=x^4-y^4,\\U_2u+V_2v+W_2=y^4-z^4.$$

$$\begin{bmatrix}U_1&V_1\\U_2&V_2\end{bmatrix}\begin{bmatrix}u\\ v\end{bmatrix}=\begin{bmatrix}x^2-y^2-W_1\\ y^2-z^2-W_2\end{bmatrix}$$ or $$\begin{bmatrix}u\\ v\end{bmatrix}=\frac{1}{[U_1V_2]^-}\begin{bmatrix}V_2&-V_1\\-U_2&U_1\end{bmatrix}\begin{bmatrix}x^4-y^4-W_1\\ y^4-z^4-W_2\end{bmatrix}$$ and $(u,v)$ our parametrization is finally given by $$\boxed{\boxed{\left(\frac{V_2x^4-(V_1+V_2)y^4+V_1z^4+[V_1W_2]^-}{[U_1V_2]^-},\frac{-U_2x^4+(U_1+U_2)y^4-U_1z^4+[W_1U_2]^-}{[U_1V_2]^-}\right).}}$$



Trying to do it more directly in a previous attempt:

This is an attempt at the 3-ellipse:

Let's give it a try. Let $$Q_k=(x-x_k)^2+(y-y_k)^2$$

so that $$\sqrt{Q_1}+\sqrt{Q_2}+\sqrt{Q_3}=\rho.$$

Now let $$\sqrt{Q_3}=\rho \cos^2 \alpha\\ \sqrt{Q_1}+\sqrt{Q_2}=\rho\sin^2\alpha $$ so that

$$Q_3=\rho^2\cos^4\alpha \\ 4Q_1Q_2=\left(\rho^2\sin^4\alpha-Q_1-Q_2\right)^2.$$ For simplicity let $(x_3,y_3)=(0,0),$ and $(x_2,y_2)=(1,0)$ so $$x^2+y^2=\rho^2\cos^4\alpha\equiv g,\\ x^2=g-y^2,\\ \rho^2\sin^4\alpha=(\rho-\sqrt{g})^2\equiv G.$$ Then

$$4((x-1)^2+y^2)((x-x_1)^2+(y-y_1)^2)=\left(\rho^2\sin^4\alpha-Q_1-Q_2\right)^2$$

$$4(g-2x+1)(g-2xx_1+x^2_1-2yy_1+y^2_1)=\left(G-(g-2x+1)-(g-2xx_1+x^2_1-2yy_1+y^2_1)\right)^2;$$ let $x_1^2+y_1^2=Z$ and collecting like terms in $x,y$ we obtain

$$16x_1x^2+16y_1xy-8((1+g)x_1+Z+g)x-8(1+g)y_1y+4(Z+g)(g+1)=\left(G-2g-Z-1+2(1+x_1)x+2y_1y\right)^2$$ which we can relabel as $$ax^2+bxy+cx+dy+e=(u+vx+wy)^2\\ ag+e+dy-ay^2+(c+by)\sqrt{g-y^2}=\left(u+v\sqrt{g-y^2}+wy\right)^2$$ or $$ag+e+dy-ay^2+(c+by)\sqrt{g-y^2}=u^2+v^2g+2uwy+(w^2-v)y^2+2(uv+vwy)\sqrt{g-y^2}$$

$$0=u^2+(v^2-a)g-e+(2uw-d)y+(w^2-v+a)y^2+2(uv-c+(vw-b)y)\sqrt{g-y^2}$$ which can be rephrased as

$$0=A+By+Cy^2+2(D+Ey)\sqrt{g-y^2}$$ $$(A+By+Cy^2)^2-4(D+Ey)^2(g-y^2)=0$$ $$A^2+B^2y^2+C^2y^4+2y(AB+CAy+BCy^2)-4(D+Ey)^2(g-y^2)=0$$ $$A^2+2ABy+(B^2+CA)y^2+2BCy^3+C^2y^4-4(D+Ey)^2(g-y^2)=0$$ $$A^2+2ABy+(B^2+CA)y^2+2BCy^3+C^2y^4+(-4gD^2-8gDEy+4(D^2-gE^2)y^2+8DEy^3+4E^2y^4)=0$$ $$A^2-4gD^2+2(AB-4gDE)y+(B^2+CA+4(D^2-gE^2))y^2\\+2(BC+4DE)y^3+(C^2+4E^2)y^4=0$$ which can again be rephrased as

$$P+Qy+Ry^2+Sy^3+Ty^4=0$$

I'm probably not going to go into detail trying to find out if, for example, $Q,S=0$, or if $T=0$, or if under some transformation I can make enough terms vanish: for now, it's sufficient know that the problem of parametrizing this octic (deg 8) curve needs only the solution to a quartic, and not a septic.


$$\begin{matrix}\text{Coefficient}&\text{Definition}\\ \\ P&A^2-4gD^2\\ Q&2(AB-4gDE)\\ R&B^2+CA+4(D^2-gE^2)\\ S&2(BC+4DE)\\ T&C^2+4E^2\\ \\ A&u^2+(v^2-a)g-e\\ B&2uw-d\\ C&w^2-v+a\\ D&uv-c\\ E&vw-b \\ \\ a&16x_1 \\ b&16y_1\\ c&-8(Z+g+(1+g)x_1)\\ d&-8(1+g)y_1\\ e&4(Z+g)(g+1)\\ \\ u&G-2g-Z-1\\v&2(1+x_1)\\ w&2y_1\\ \\ Z&x_1^2+y_1^2\\G&(\rho-\sqrt{g})^2\\g& \rho^2\cos^4\alpha\end{matrix}$$


If I would do it over again, I'd wager the substitution $X=x^2+y^2,Y=x^2-y^2.$