Simplifying $\arcsin\frac{\sin a}{\sin c}+\arccos\frac{\tan a}{\tan c}-\frac\pi2$, the area of a spherical right triangle w/hypotenuse $c$ and leg $a$

81 Views Asked by At

Given an unit sphere and on this sphere a right angled triangle and given the length of the hypotenuse as $c$ and a leg as $a$.

The area $E$ can calculated via the Napier rules as: (using radians as angle unit)

$$ E= \arcsin \left(\frac{\sin a}{\sin c}\right) + \arccos \left(\frac{\tan a}{\tan c}\right) - \frac{\pi}{2} $$

I cannot believe that this formula cannot be simplified, but I don't know how.

Can somebody show me (with steps) how to simplify this formula?

Following the suggestions below: Using $p= sin a/ sin c $ and $ q = tan a / tan c $ the formula gets to

$$ E = \arcsin p + \arccos q - \frac{\pi}{2} $$

To be continued ....

1

There are 1 best solutions below

2
On

"Raw" areas and lengths tend to be ugly. The trick to an elegant relation is usually to wrap the values in appropriate trig functions.

For instance, here you could write $$\cos E = \frac{\cos^2a + \cos c}{\cos a (1 + \cos c)}$$ This isn't super-terrible; at least everything's in terms of cosines.

Nevertheless, my experience has taught me that non-Euclidean trig seems to prefer half-angle arguments. (In fact, it happens so often that I've taken to using $x_2$ as a compact way to denote $x/2$. I'll use that notation here.) Consider ... $$\begin{align} \tan^2E_2 = \frac{1-\cos E}{1+\cos E} &= \frac{(1-\cos a)(\cos a-\cos c)}{(1+\cos a)(\cos a+\cos c)} \\[6pt] &= \frac{2\sin^2a_2 \cdot 2 \sin(c_2-a_2) \sin(c_2+a_2)}{2\cos^2a_2\cdot 2 \cos(c_2-a_2) \cos(c_2+a_2)} \\[6pt] &= \tan^2a_2 \tan(c_2-a_2) \tan(c_2+a_2) \tag{$\star$} \end{align}$$ This form is rather nice; everything's in terms of (half-angle) tangents! (And, if you like, you can re-solve explicitly for $E$.) In particular, it allows for convenient "infinitesimal sanity-checking" (not to be confused with "infinitesimal-sanity checking", a service I wish my bank would offer).

For vanishingly-small $x$, we have $\tan x \approx x$. So, for an infinitesimal spherical right triangle, relation $(\star)$ indicates $$\left(\frac12E\right)^2 \approx \left(\frac12 a\right)^2 \cdot \frac12 (c-a) \cdot \frac12(c+a) \quad\to\quad E \approx \frac12 a \sqrt{c^2-a^2}$$ which is, of course, the formula for the area of a Euclidean right triangle with legs $a$ and $\sqrt{c^2-a^2}$.


Note. Expanding-upon my comment about getting to my first equation ... Take the cosines of both sides of OP's equation. Defining $p := \sin a/\sin c$, $q := \tan a/\tan c$ for convenience, we have ... $$\begin{align} \cos E &= \cos\left(\arcsin p + \arccos q - \pi/2\right) \\ &=\sin(\arcsin p+\arccos q) \\ &=\sin(\arcsin p)\cos(\arccos q) + \cos(\arcsin p) \sin(\arcsin q) \\ &= p q + \sqrt{ (1-p^2)(1-q^2) } \end{align}$$ Now, observe that $$\begin{align} 1 - p^2 &= \frac{\sin^2c-\sin^2a}{\sin^2c} = \frac{(1-\cos^2c)-(1-\cos^2a)}{\sin^2c} = \frac{\cos^2a-\cos^2c}{\sin^2c} \\[8pt] 1-q^2 &= \frac{\tan^2c-\tan^2a}{\tan^2c} = \frac{(\sec^2c-1)-(\sec^2a-1)}{\tan^2c} = \frac{\sec^2c-\sec^2a}{\tan^2c} \\[4pt] &=\frac{\cos^2a-\cos^2c}{\tan^2c\cos^2c\cos^2a}=\frac{\cos^2a-\cos^2c}{\sin^2c\cos^2a} \end{align}$$ Therefore, we can write $$\begin{align} \cos E &= \frac{\sin a}{\sin c}\cdot\frac{\tan a}{\tan c} + \frac{\cos^2a-\cos^2c}{\cos a\sin^2 c} = \frac{\sin^2 a \cos c+\cos^2a-\cos^2c}{\cos a\sin^2 c} \\[4pt] &= \frac{(1-\cos^2 a) \cos c+\cos^2a-\cos^2c}{\cos a\sin^2 c} = \frac{(1-\cos^2 a) \cos c+\cos^2a-\cos^2c}{\cos a\sin^2 c} \\[4pt] &= \frac{(1 + \cos c) (\cos^2 a + \cos c)}{\cos a (1-\cos^2c)} \\[4pt] &= \frac{\cos^2 a+\cos c}{\cos a(1+\cos c)} \end{align}$$

As I mentioned in my comment: It's a bit of a slog. ... Moreover, it's not entirely obvious at first what trig manipulations should be done. Again, my experience has guided me here: in particular, I've found that a bias for cosines tends to serve me pretty well in simplifications like this, so that gave me a direction to start.

Plus, experience has instilled in me a faith that there must be an elegant relation behind all of this mess ... which is why I wasn't satisfied with the rather clunky cosine formula.

The messy denominator suggested to me that I might want to divide by, say, $\sin E$, whose denominator would match and therefore cancel; but the overall result isn't any better. Recalling that non-Euclidean trigs seem to prefer half-angle arguments, I calculated $1+\cos E$ and $1-\cos E$ (most of the way to computing $\cos^2E_2$ and $\sin^2E_2$), which happened to have nicely-factored numerators (and still-messy-but-cancelable denominators), leading me to the initial $\tan^2 E_2$ formula.

The steps from there to the final form were perhaps a bit less mysterious, and the elegance of that final form told me that that was the proper place to stop.