1

Given a unit vector $\hat{n}$ and angle $\theta_{nm}$, what is the general form for the unit vector $\hat{m}$, if $\theta_{nm}$ is the separation angle between $\hat{n}$ and $\hat{m}$?

Obviously they satisfy:

$$\hat{m}\cdot\hat{m}=1 \quad \hat{n}\cdot\hat{n}=1 \quad \hat{n}\cdot\hat{m}=\cos{\theta_{nm}}$$

This has puzzled me for a bit now. It's intuitive that the 'solution' lies on a cone enveloping $\hat{n}$, but it's unclear to me what its analytic expression is.

The case of $\hat{n}=\left(0, 0, 1\right)$ is trivial in $xyz$-coordinates,

$$ \hat{m} = \left( \cos\phi\sqrt{1-\cos\theta_{nm}},\,\, \sin\phi\sqrt{1-\cos\theta_{nm}},\,\, \cos\theta_{nm} \right) \quad $$

with $\phi$ being any angle. My initial approach was to transform $\hat{n}\rightarrow\left(0, 0, 1\right)$, and then transform $\hat{m}$ inversely, which however seems unnecessary (and I dislike rotation matrices). Given the independence of coordinate systems, there must be a general vector form, which could then be easily cast in Cartesian, spherical, or whatever systems. I am looking for this pure vector solution, not in a vector-component form.

I found a similar question whose solution is in a vector form. But I am unsure how to extend it to the problem at hand.

If you can show how it is applicable to a given $\left( n_x,\, n_y,\, n_z \right)$ or $\left( n_r,\, n_\theta,\, n_\phi \right)$ to get $\hat{m}$, that’d be fantastic! Thanks.

Comment: Here's how to visualize. With all unit vectors, we can think of them lying on a unit-sphere. Let $\hat{n}$ be defined by a point anywhere on the sphere. Then, $\hat{m}$ lies anywhere on the circle centered at $\hat{n}$ and with radius $\sin\theta$.

Quanto
  • 97,352
  • You need an additional vector $v$ orthogonal to $n$. Then $m=n\cos\theta+(v\cos\phi+n\times v\sin\phi)\sin\theta$ for any angle $\phi$. However there is no simple closed-form formula for $v$ in terms of $n$ thanks to the hairy ball theorem. –  Oct 28 '19 at 21:54
  • I'm not asking for a single vector, otherwise yes I'd need to define the vector of rotation. I'm asking for the solution of an arbitrary $\phi$. That is for example, if given $\hat{n}=\left< 0,, 0,, 1\right>$ then $\hat{m}=\left< \sin\phi\sin\theta,, -\cos\phi\sin\theta,, \cos\theta \right>$. This is of course the trivial case of $\hat{n}$ pointing straight up and would need to be generalized. $\phi$ is any angle and would not matter. – Captain Morgan Oct 28 '19 at 22:04
  • @Rahul Oops, I realize I more-or-less restated part of my question in response to you. Sorry. – Captain Morgan Oct 28 '19 at 22:06

1 Answers1

1

As visualized by the OP, the resulting vector $\vec{m}$ shall point to anywhere on a circle with the vector $\vec{n}$ as the center. In other words, the vector $\vec{m}$ is to be parametrized to trace out a circle around $\vec{n}$.

To construct $\vec{m}$ generally, we assume a parametrization vector $\vec{a}$ that is not parallel to $\vec{n}$, but otherwise arbitray. First, a vector that is perpendicular to $\vec{n}$ can be constructed as follows,

$$\vec{u}= \vec{a} - (\vec{a}\cdot\vec{n})\vec{n}$$

which satisfies $\vec{u}\cdot\vec{n}=0$. Next, the unit vector $\vec{m}$ that is at angle $\theta_{nm}$ with $\vec{n}$ can be constructed in the general form below,

$$\vec{m} = \cos \theta_{nm} \> \vec{n} + \sin\theta_{nm}\frac{\vec{u}}{|\vec{u}|}$$

It is straightforward to verify that $|\vec{m}|=1$ and $\vec{m}\cdot\vec{n}=\cos \theta_{nm}$, that is, $\vec{m}$ is a unit vector at the angle $\theta_{nm}$ with the vector $\vec{n}$.

By varying the parametrization vector $\vec{a}$, which is equivalent to, yet more general than, an arbitrary $\phi$, the resulting vector $\vec{m}$ rotates around the given vector $\vec{n}$. As pointed by the OP, the radius of the rotation is $\sin\theta_{nm}$, which can be obtained from the vector expression constructed above.

Quanto
  • 97,352
  • I'm guessing there's no way to do this without the generation of the dummy vector huh? It's basically what $\phi$ is so I guess it's not a huge deal. – Captain Morgan Oct 29 '19 at 20:58
  • @Captain Morgan - Yeah, since $m$ depicts a special curve, you need something to parametrize it. The most general form of parametrization is an arbitrary $a$ vector, for which the familiar $\phi$ is a special case. – Quanto Oct 29 '19 at 21:48