2

Let’s suppose I have two spheres $S_1$ and $S_2$. Let's suppose they have a non-trivial intersection, meaning its not a point. Then their intersection forms a circle.

Assume I have the centers of each sphere $P_1$ and $P_2$. Assume their radii are $r_1$ and $r_2$ respectively.

Consider this diagram:

enter image description here

This is a cross section view of the spheres along their axis of intersection. Importantly, I wish to find the radius. A similar question is asked here:

What is the easiest way to find the radius and center of the circle of intersection between two spheres?

The author suggests that we can get to the solution for the radius via $d_1^2 + d_2^2 = d^2$, but I don't understand how this is possible because, in this diagram, when you look at $d_{1}, d_{2}, d$ they are colinear, so I don't see how Pythagorean theorem would apply since that only applies to right triangles.

Can some explain why the formula is correct?

EDIT: it seems like this thread is relevant

https://mathworld.wolfram.com/Sphere-SphereIntersection.html

Let's assume momentarily that $P_1 = (0,0,0)$ and $P_2=(0,0,d)$.

Then we have the equations of the spheres given by

\begin{align*} x^2+y^2+z^2&=r_1^2\\ (x-d)^2+y^2+z^2&=r_2^2\\ \end{align*}

Combining and rearranging yields

$$x=\frac{d^2-r_2^2+r_1^2}{2d}$$

Then taking the first equation, we have

\begin{align*} x^2+y^2+z^2&=r_1^2\\ y^2+z^2&=r_1^2-x^2\\ y^2+z^2&= r_1^2-\left(\frac{d^2-r_2^2+r_1^2}{2d}\right)^2\\ y^2+z^2&= \frac{4d^2r_1^2-(d^2-r_2^2+r_1^2)^2}{4d^2}\\ \end{align*}

Calling $r^2\equiv \frac{4d^2r_1^2-(d^2-r_2^2+r_1^2)^2}{4d^2}$, you notice then, we can write this as

$$y^2+z^2 = r^2$$

implying the radius of this circle is

\begin{align*} r &= \sqrt{\frac{4d^2r_1^2-(d^2-r_2^2+r_1^2)^2}{4d^2}}\\ &=\frac{1}{2d} \sqrt{4d^2r_1^2-(d^2-r_2^2+r_1^2)^2} \end{align*}

User @peterwhy suggested the following test case. Let $r_1=r_2=d=1$ , then the radius of intersection common circle $r$ should be the height of an equilateral triangle of unit side: $r=\frac{\sqrt{3}}{2}$, which indeed this formula yields.

My question is, is this generalizable to any spheres with arbitrary centers so long as they intersect?

  • For more context, the linked answer says / claims $d_1^2+d_2^2=(r_1^2-r^2)+(r_2^2-r^2)=d^2$. – peterwhy Feb 20 '24 at 02:25
  • @peterwhy could you possibly explain why that's important to know? i'm just curious. is $(r^2_1-r^2) + (r_2^2-r^2)=d^2$ more useful / insightful? – Stan Shunpike Feb 20 '24 at 03:49
  • No, I was also confused after reading the answer, and I was just posting that line here for others. – peterwhy Feb 20 '24 at 04:15
  • @peterwhy I may have found a thread with a better answer. I updated my question with the relevant derivation. I'm still confused however if this is only applicable to the particular spheres described there or can be used for any two intersecting spheres – Stan Shunpike Feb 20 '24 at 04:36
  • @StanShunpike IMHO, the derivation is perfectly sound. Could you specify why you are confused about it? – Soham Saha Feb 20 '24 at 04:43
  • @SohamSaha Well, we assumed the centers of the two spheres initially so that there were cancellations. Does that lose generality for arbitrary centers of the spheres? – Stan Shunpike Feb 20 '24 at 04:51
  • @StanShunpike consider the 2 spheres with a distance d between their centers at the given locations. Do you think the radius of the interesting circle would change if you change the orientation of the 2 spheres? The centers don’t matter, as long as the distance between them is d. – Soham Saha Feb 20 '24 at 05:05
  • 1
    @SohamSaha Yup, perfect way to visualize it. Thank you so much! – Stan Shunpike Feb 20 '24 at 05:07
  • 1
    Forget all the 3D stuff and stick to the cross-section you’ve drawn — two intersecting circles. – Ted Shifrin Feb 20 '24 at 05:51

1 Answers1

0

HINT

Area of sectioned triangles shown is expressed in two ways:

The three sides of top half triangle are $ r_1,r_2, r$

$$ \frac 12 d~ r = (s(s-d)(s-r_1)(s-r_2))^{\frac12}$$

Narasimham
  • 40,495
  • I could not follow this hint, sorry! I didn't know what $s$ was. However, I did attempt to derive $r$ myself. I found a specific case and put it in my question. Do you know if the solution i found is generalizable? – Stan Shunpike Feb 20 '24 at 04:38
  • 1
    @StanShunpike, I think $s$ refers to semi-perimeter, and the RHS is an application of Heron’s formula. – Soham Saha Feb 20 '24 at 04:45
  • Yes, it was given as a hint ( standard notation) so you could calculate and compare with expansion of $ d=\sqrt{r_1^2-r^2}+ \sqrt{r_2^2-r^2}$ – Narasimham Feb 20 '24 at 05:20