Let $\mathbb{S}^n$ be the round sphere, $p_0$ be the north-pole and $T$ be the equator whose center is $p_0$. Given a point $p$ on the sphere, how do we calculate the (spherical, intrinsic) distance of $p$ to $T$? It certainly has to do with the inner product $\langle p, p_0 \rangle$, but I am unsure about the explicit formula. If $p$ is next to $\pm p_0$, $d(p, T)$ is next to $\pi/2$, and if $p$ is next to $T$, obviously $d(p, T)$ is next to $0$. And it is easy to see that $d(p, T)$ is even with respect to $\langle p, p_0 \rangle$. I suspect $d(p, T) = |\arcsin \langle p, p_0 \rangle|$. Is this correct?


Let $p$ and $q$ be two points on the surface of the sphere, of radius $R$, subtending a central angle of $\theta$ (along any great circle through $p$ and $q$). Then $$ \frac{\langle p,q \rangle}{||p|| \, ||q||} = \frac{\langle p,q \rangle}{R^2} = \cos \theta \text{.} $$

Any great circle through $p_0$ meets $T$ perpendicularly. So the angle $\theta$ satisfying $$ \frac{\langle p_0,p \rangle}{R^2} = \cos \theta $$ is the angle between $p_0$ and $p$ along a great circle perpendicular to $T$. If $0 \leq \theta\leq \pi/2$, the (smallest) angle between $p$ and $T$ is $\varphi = \frac{\pi}{2} - \theta$. If $\pi/2 \leq \theta \leq \pi$, then $\varphi = \theta - \pi/2$. (It can help to draw a picture here, where in one case $p$ is in the same hemisphere bounded by $T$ as is $p_0$ and in the other case, $p$ is in the opposite hemisphere.) Combining all this, $$ \varphi = \left| \frac{\pi}{2} - \theta \right| = \left| \frac{\pi}{2} - \cos^{-1}\left( \frac{\langle p_0,p \rangle}{R^2} \right) \right| \text{.} $$

Finally, the distance along the great circle through $p$ perpendicular to $T$ is $\varphi R$, so $$ d(p,T) = R \left| \frac{\pi}{2} - \cos^{-1}\left( \frac{\langle p_0,p \rangle}{R^2} \right) \right| \text{.} $$

If this is the unit sphere, $R = 1$ and this simplifies a bit.

Suppose we now want to solve that for $\langle p_0,p \rangle$. \begin{align*} \frac{d(p,T)}{R} &= \left| \frac{\pi}{2} - \cos^{-1}\left( \frac{\langle p_0,p \rangle}{R^2} \right) \right| \\ &\in \pm \left( \frac{\pi}{2} - \cos^{-1}\left( \frac{\langle p_0,p \rangle}{R^2} \right) \right) \text{.} \end{align*} Then \begin{align*} \cos\left( \frac{d(p,T)}{R} \right) &= \cos \pm \left( \frac{\pi}{2} - \cos^{-1}\left( \frac{\langle p_0,p \rangle}{R^2} \right) \right) \\ &= \cos \left( \frac{\pi}{2} - \cos^{-1}\left( \frac{\langle p_0,p \rangle}{R^2} \right) \right) && \text{cosine is even} \\ &= \cos \frac{\pi}{2} \cos \cos^{-1}\left( \frac{\langle p_0,p \rangle}{R^2} \right) + \sin \frac{\pi}{2} \sin \cos^{-1}\left( \frac{\langle p_0,p \rangle}{R^2} \right) \\ &= \sin \cos^{-1}\left( \frac{\langle p_0,p \rangle}{R^2} \right) \text{.} \end{align*} The range of arccosine is $[0, \pi]$, and sine is nonnegative for all those angles, so \begin{align*} \cos\left( \frac{d(p,T)}{R} \right) &= \frac{+\sqrt{R^4 - \langle p_0,p \rangle^2}}{R^2} \\ &= \sqrt{1 - \left( \frac{\langle p_0,p \rangle}{R^2} \right)^2} \text{,} \end{align*} so $$ \left( \frac{\langle p_0,p \rangle}{R^2} \right)^2 = 1 - \cos^2 \left( \frac{d(p,T)}{R} \right) $$ and then \begin{align*} \langle p_0,p \rangle &= \pm R^2 \sqrt{1 - \cos^2 \left( \frac{d(p,T)}{R} \right)} \\ &= \pm R^2 \sqrt{\sin^2 \left( \frac{d(p,T)}{R} \right)} \\ &= \pm R^2 \sin \left( \frac{d(p,T)}{R} \right) \text{.} \end{align*} This is the best we can do, because $d(p,T)$ doesn't contain information about which hemisphere $p$ is in. (This missing information was first indicated by the "$\pm$".) We choose the positive root if $p$ is in the same hemisphere as $p_0$ and the negative root if $p$ is in the other hemisphere (and the square root has the value zero if $p$ is on $T$, so either choice of sign is fine).

  • $\begingroup$ Can we solve for $\langle p,p_0 \rangle$? $\endgroup$ – Eduardo Longa Dec 1 '17 at 14:38
  • 1
    $\begingroup$ @EduardoLonga : THat would be a new question. However, I'll edit in a second answer. $\endgroup$ – Eric Towers Dec 1 '17 at 14:51
  • $\begingroup$ we can further simplify the last square root to get a sine, right? $\endgroup$ – Eduardo Longa Dec 1 '17 at 20:12
  • 1
    $\begingroup$ @EduardoLonga : Yes we can and I have (now, together with a dropped square of $R$ near the end). The result shouldn't be so surprising; The angle between $p$ and $p_0$ is complementary to the angle between $p$ and $T$, so while $\langle p, p_0 \rangle$ is proportional to $\cos \theta$, it is also proportional to the complementary cosine (that is, sine) of the complementary angle. $\endgroup$ – Eric Towers Dec 1 '17 at 21:37

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.