2
$\begingroup$

I've got a problem in which I'm trying to find the area of an ellipse which is given by the intersection of an elliptic cylinder with a plane. Nothing here is parallel to the coordinate axes, which is making it kind of annoying to work with.

The plane is given by the equation $x+ay+a^2z=0$, and the cylinder is given by $(x-a^2z)^2-a(x-a^2z)(y-az)+a^2(y-az)^2=L^2$.

I can think of some complicated ways to do it with integrals, and I'm wondering if there's something simpler that I'm missing. If I could transform into some coordinates in the plane I'm trying to work with, and I knew that the coordinate transformation would preserve areas, then I would be good, because I know how to get the area of an ellipse in the form $Ax^2+Bxy+Cy^2=1$. I'm really not sure how to do this transformation, though, and whether this is even the best way to proceed.

Maybe I should be using Langrange multipliers with two constraints to just obtain the lengths of the semi-major and semi-minor axes? That sounds like a pain, but doable. At least we're centered at the origin.

Thanks in advance for any assistance.


Edit: If I take $u=(x-a^2z)$ and $v=(y-az)$, then my cylinder becomes $u^2-auv+a^2v^2=L^2$, which is pretty great, but then I've basically cut the cylinder with a plane parallel to $z=0$. In that case, my given plane becomes $u+av+3a^2z=0$, and I'm not sure how that's helpful.

$\endgroup$
  • $\begingroup$ This looks suspiciously like you are already very close to your required form, where you get $$\frac 1{L^2}u^2-\frac a{L^2}uv+\frac {a^2}{L^2}v^2=1$$ Perhaps this $u,v$ setup can take over in the planar equation as well? $\endgroup$ – abiessu Feb 9 '15 at 21:52
  • $\begingroup$ By graphic view ( not established ) the plane is cut normally by a right circular cylinder for special case a =1. In this case only direction cosines of cylinder ( equally inclined to x,y,z axes) axis need be considered. $\endgroup$ – Narasimham Feb 10 '15 at 10:25
  • $\begingroup$ Is the axis of the cylinder $x=ay=a^2z$, and is this axis perpendicular to the cutting plane? $\endgroup$ – David K Feb 11 '15 at 15:56
  • $\begingroup$ That is the axis of the cylinder, but it is not perpendicular to the cutting plane. A line normal to the plane would be given by $a^2x=ay=z$ instead. $\endgroup$ – G Tony Jacobs Feb 15 '15 at 1:08
4
$\begingroup$

Expanding the equation of the cylinder, we get $$ x^2-axy+a^2y^2-a^2xz-a^3yz+a^4z^2 = L^2 $$ Writing this as a quadratic form: $$ \begin{bmatrix}x\\y\\z\end{bmatrix}^T \begin{bmatrix}1 & -\frac{a}{2} & -\frac{a^2}{2} \\ -\frac{a}{2} & a^2 & -\frac{a^3}{2} \\ -\frac{a^2}{2} & -\frac{a^3}{2} & a^4 \end{bmatrix} \begin{bmatrix}x\\y\\z\end{bmatrix} = L^2 $$ Diagonalizing this matrix, we find that the eigenvalue $0$ is associated with the eigenvector $\vec{v} = (a^2,a,1)^T$, giving the axis of the cylinder. The other two eigenvalues $\lambda_1$ and $\lambda_2$ are related to the semi-major axes of the ellipse by $s_{1,2} = L/\sqrt{ \lambda_{1,2}}$. The other eigenvalues are: $$ \lambda_{1,2} = \frac{1}{2}\left(1+a^2+a^4\pm\sqrt{(a^2-1)^2(a^4+a^2+1)} \right)$$

The area of the cross section of this cylinder is $$A_0=\pi s_1s_2 = \frac{\pi L^2}{\sqrt{\lambda_1 \lambda_2}} = \frac{2\pi L^2}{a\sqrt{3(1+a^2+a^4)}}$$

The plane cuts the cylinder at an angle, and the area of the slanted cut is $A = A_0/\cos \theta$ where $\theta$ is the angle between the plane's normal axis and the cylinder axis. The plane's normal vector is $\vec{n} = (1,a,a^2)^T$. Using a dot product, $$\cos\theta = \frac{\vec{n}\cdot\vec{v}}{\left\lVert \vec{n} \right\rVert \left\lVert \vec{v} \right\rVert} = \frac{3a^2}{1+a^2+a^4}$$ So then, $$ A = \frac{2\pi L^2}{a\sqrt{3(1+a^2+a^4)}} \frac{1+a^2+a^4}{3a^2} = \frac{2\pi L^2}{3a^3\sqrt{3}} \sqrt{1+a^2+a^4}$$


Ignore this stuff: As OP pointed out, projecting out $x$ doesn't make any sense.

Substituting away $x$ using the plane equation $x=-ay-a^2 z$ into the cylinder equation gives $$ a^2(y+2az)^2+a^2(y+2az)(y-az)+a^2(y-az)^2=L^2 $$ $$ (y+2az)^2+(y+2az)(y-az)+(y-az)^2=\left(\frac{L}{a}\right)^2 $$ Let $u=y+2az$ and $v=y-az$, so $$ u^2+uv+v^2 = \left(\frac{L}{a}\right)^2 $$ The Jacobian of this transformation is $-3a$. The area of the ellipse in $u,v$ coordinates can be found using the second formula here. To obtain the area in the original $x,y,z$ coordinates, multiply by the Jacobian of the transformation.

$\endgroup$
  • 1
    $\begingroup$ If you substitute away the $x$, then you get an equation that relates the $y$- and $z$-coordinates of each point on the ellipse, i.e., an equation of its projection onto the $yz$-plane. Once I have an equation in two variables, I'm well able to calculate the area. However, that's not the ellipse I'm looking for, or am I missing something? $\endgroup$ – G Tony Jacobs Feb 10 '15 at 21:33
  • $\begingroup$ @GTonyJacobs: Indeed, you are correct. I have edited my answer with a full solution. Let me know if you want clarification on any steps. $\endgroup$ – Victor Liu Feb 10 '15 at 23:31
  • $\begingroup$ This is excellent; thank you. The only thing that's not 100% clear is that I don't know where the equation $s_{1,2}=L/\sqrt{\lambda_{1,2}}$ comes from. This must be a general fact about ellipses and quadratic forms? $\endgroup$ – G Tony Jacobs Feb 15 '15 at 2:08
  • 1
    $\begingroup$ @GTonyJacobs: Indeed. The diagonal form of the quadratic form $(x,y)^T diag(a,b) (x,y) = L^2$ leads to $ax^2 + by^2 = L^2$, which you compare to an ellipse with semimajor axes $x^2/s_1^2 + y^2/s_2^2 = L^2$. $\endgroup$ – Victor Liu Feb 15 '15 at 17:52
0
$\begingroup$

I think this is less difficult than it appears. If you take any figure with area $S$ in a plane and projected it orthogonally onto another plane, the area of the image of your figure will be $S \cos \theta$, where $\theta$ is the angle between the planes. (Equivalently, $\theta$ is the angle your direction of projection makes with a vector normal to the plane you projected from.)

More generally, if you project the figure obliquely onto another plane, then the image has area $S \cos\theta\sec\phi$ where $\phi$ is the angle between the direction of projection and a vector normal to the plane you projected to.

Now consider the intersection of the cylinder with the $x,y$-plane. This is an ellipse whose formula you can derive simply by setting $z = 0$ in the formula of the cylinder. You know how to find the area $S$ of that ellipse. A projection parallel to the axis of the cylinder takes that ellipse onto your desired ellipse. I believe that projection is an orthogonal projection onto the plane $x+ay+a^2z=0$, in which case all you need to do is find the angle $\theta$ between the axis of the cylinder and the $z$ axis, and then the area of the ellipse you want is $S \cos\theta$.

If I'm wrong about the axis of the cylinder being orthogonal to the plane $x+ay+a^2z=0$, then you also have to find the angle $\phi$ between the axis and that plane, and then the answer is $S \cos\theta\sec\phi$.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.