# A time parameterization of geodesics on the sphere

So I know that all the geodesics on the sphere ($\mathbb{S}^2\subseteq\mathbb{R}^3$) lie on great circles. However, I have been having a bit of trouble coming up with a time parameterization of these great circle arcs. Specifically, if I have 2 points $(\theta_1,\phi_1)$ and $(\theta_2,\phi_2)$ that lie on $\mathbb{S}^2$ what is the function $\gamma:[0,1]\to\mathbb{S}^2$ such that $\gamma([0,1])$ is the geodesic that connects these two points?

This doesn't seem like it should be terribly difficult; however, I've been getting stuck. The reason that I care about this parameterization is that I am trying to get some visualization working in Mathematica.

• Maybe use spherical coordinates? – IAmNoOne Aug 6 '16 at 0:32
• What do you mean? If I just take $\gamma(t)=(1-t)(\theta_1,\phi_1)+t(\theta_2,\phi_2)$ I don't get a curve that lies on the great circle. – yousuf soliman Aug 6 '16 at 0:40
• Well you are parametrizing a straight line. I think if you just fix one of the coordinates, you would get your great circle. – IAmNoOne Aug 6 '16 at 2:03
• The coordinate function $\mathbf{x} = (\cos\theta \sin \phi, \sin\theta \sin \phi, \cos\phi )$ describes 1 patch of $S^2$. So if I fix $\theta$, I get one great circle was my idea. But I guess that isn't what you really want since you want to map from $[0,1]$. So actually we mayhave to do this in Euclidean coordinates. The idea should be the same. – IAmNoOne Aug 6 '16 at 2:13
• @Nameless that makes a lot of sense and gives me the intuition on how to come up with what John Hughes answered below. – yousuf soliman Aug 6 '16 at 3:42

Use $$\mathbf{x} = (\cos\theta \sin \phi, \sin\theta \sin \phi, \cos\phi )$$

to get vector forms $v_1$, $v_2$ for your two points.

Let $$w = v_2 - (v_2 \cdot v_1) v_1 \\ u = \frac{1}{\|w\|} w$$

(This is basically just Gram-Schmidt on the basis $\{v_1, v_2 \}$.)

Now let $$\alpha(t) = \cos(t) v_1 + \sin(t) u$$ As $t$ goes from $0$ to $2\pi$, you'll traverse the great circle containing $v_1$ and $v_2$, starting from $v_1$, passing through $v_2$ before you get to angle $\pi$, and continuing on back to $v_1$.

If you want to stop at $v_2$, just let $t$ run from $0$ to $c$, where $$c = \cos^{-1} (v_2 \cdot v_1).$$

• Why did you need to normalize $w$? It doesn't seem that you use $u$ anywhere else. Should I use $u$ instead of $w$ in the definition of $\alpha$? – yousuf soliman Aug 6 '16 at 3:47
• Sorry 'bout that. Yes, use $u$ instead of $w$. I'll edit. Normalizing $w$ is to put it back on the unit sphere, rather than being a much shorter vector. Notice that the whole thing fails if $v_2 = \pm v_1$, but in the case where $v_2 = v_1$, the constant path works, and when $v_2 = -v_1$, there's no single geodesic between them, so there's no easy way to pick just one. (Example: from north pole to south pole, any line of longitude will suffice as a great-circle arc.) – John Hughes Aug 6 '16 at 11:15