Let us consider (in spherical coordinates) the expression

Great arc distance between two points on a unit sphere

$$d({\bf v}_1,{\bf v}_2)=\cos^{-1}\left(\cos\theta_1\cos\theta_2+\sin\theta_1\sin\theta_2\cos\left(\varphi_1-\varphi_2\right)\right) \qquad (1)$$ that gives the geodesic distance between two points $${\bf v}_i=(\sin\theta_i\cos\varphi_i,\sin\theta_i\sin\varphi_i,\cos\theta_i)$$ on the unit sphere $S$.

A question:

Assume $\,{\bf v}_i\in C\,$ where $\,C\,$ is the circle $\, \left\{\theta=\pi/4 \right\}\cap S\,$. The pole (the origin of the spherical coordinate system) is still the center of the sphere of course.

Then $\,\theta_1=\theta_2=\frac{\pi}{4}\,$ and, by (1), we have $$d({\bf v}_1,{\bf v}_2)=\cos^{-1}\left(\frac{1}{2}+\frac{1}{2}\cos\left(\varphi_1-\varphi_2\right)\right)=\cos^{-1}\left(\frac{1}{2}+\frac{1}{2}\cos\left(\varphi_1-\varphi_2\right)\right)=\\ =\cos^{-1}\left(\cos^{2}\frac{\varphi_1-\varphi_2}{2}\right)$$.

On the other hand, the radius of $C$ is $\frac{1}{2}$. So the arc length between $\,{\bf v}_1,\,{\bf v}_2\,$ should also be $$\text{radius of}\;\;C\times |\varphi_1-\varphi_2|= \frac{|\varphi_1-\varphi_2|}{2}$$

My problem is $$\cos^{-1}\left(\cos^{2}\frac{\varphi_1-\varphi_2}{2}\right)\not\equiv \frac{|\varphi_1-\varphi_2|}{2}$$ Take for instance $\varphi_1=\pi/2$, $\varphi_2=0$. Then $\cos^{-1}\left(\cos^{2}\frac{\varphi_1-\varphi_2}{2}\right)=\frac{\pi}{3}$ while $\frac{\varphi_1-\varphi_2}{2}=\frac{\pi}{4}$.

Where am I mistaken ?


  • $\begingroup$ The geodesic between the points you are describing is not the circle $C$. Remember the geodesic is always a great circle (i.e. the intersection of a plane passing through the origin and the sphere). $\endgroup$ – Hamed Jun 25 '18 at 1:27

You've got two problems here.

  • Firstly, the radius of the circle $\theta = \pi/4$ is not $1/2$, it's $\cos{(\pi/4)} = 1/\sqrt{2}$.
  • Secondly, the circle $\theta=\pi/4$ is not a great circle, so it does not minimise the distance between the two points on the sphere. This is very clear in the case where $\phi_1-\phi_2=\pi$, because then the great circle passing through both points goes through the north pole of the sphere, and the distance along this is $\pi/2$, whereas the distance along $\theta=\pi/4$ is $\pi/\sqrt{2}$ which is rather larger. There's a comparison image of this sort of situation here, with example distances given. And here is a comparison of the great-circle distance (blue) with the "latitudinal" distance (orange) for the case you're considering:

enter image description here

Clearly for small distances the latitudinal distance is a reasonable approximation, but it gets worse and worse for large differences in angle.


This is not an answer. This is a comment in reply to the great answer of @Chappers above. I am using the answer environment because I do not know how insert a figure inside a comment.

@Chappers Thank you for the great answer.

It does not really matter, but the radius of the circle $C$ is $\frac{1}{2}$ I still believe. Consider the isosceles right triangle with the vertices at the origin $O$, the north pole $\,B (\theta=0, \varphi=0),\,$ and $\,A (\theta=\frac{\pi}{2}, \varphi=0).\,$ Let $A,B\in S$ so that $OA=OB=1$. Let the horizontal line segment $\overline{DC}$ represent a radius in $C$. Since $\theta=\angle COD=\frac{\pi}{4}$ then all the acute inner angles in the figure below are $\pi/4$. So $BC=OC=CD=\frac{1}{2}$.enter image description here


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.