Squaring each side,
x^2\sec^2 y - 2x\sec y \sin y + \sin^2 y = \sin^2 y + \sin^2 \alpha
sec y sin y = sin y * 1/cos y = tan y, so
x^2\sec^2 y - 2x\tan y = \sin^2 \alpha
Then, there's a trig identity 1 + tan^2 x = sec^2 x, so
x^2(1 + \tan^2 y) - 2x\tan y - \sin^2 \alpha = 0
Distributing and rearranging,
x^2\tan^2 y - 2x\tan y + x^2 - \sin^2 \alpha = 0
You might recognize this as a quadratic equation, not in x, but in tan y. We can use the quadratic formula:
\tan y = \frac{2x \pm \sqrt{4x^2 - 4(x^2)(x^2 - \sin^2 \alpha)}}{2(x^2)}
\tan y = \frac{2x \pm \sqrt{4x^2 - 4x^4 + 4x^2\sin^2 \alpha}}{2x^2}
Factoring out an 4x^2 from the square root,
\tan y = \frac{2x \pm \sqrt{x^2(1 - x^2 + \sin^2 \alpha}}{2x^2}
Then, \sqrt{a^2b} = a\sqrt{b}, so
\tan y = \frac{2x \pm 2x\sqrt{1+\sin^2 \alpha - x^2}}{2x^2}
Simplifying this,
\tan y = \frac{1 \pm \sqrt{1+\sin^2 \alpha - x^2}}{x}
Then, (tan y = tan(y + pi) = tan(y + 2pi) = ...),
y = \arctan\left(\frac{1 \pm \sqrt{1+\sin^2 \alpha - x^2}}{x} \right) + k\pi
where arctan(x) is the principal value of the inverse-tangent.
And we can finally see what the domain will be.
Domain of arctan(x): all reals
Domain of inside parts:
- The value under the square root has to be non-negative:
1+\sin^2 \alpha - x^2 \geq 0
x^2 \leq 1+\sin^2 \alpha
For these kinds of inequalities (x^2 <= constant), the interval is always [-sqrt(constant), sqrt(constant)], so
-\sqrt{1 + \sin^2 \alpha} \leq x \leq \sqrt{1+\sin^2 \alpha}
- The denominator cannot be zero.
x \neq 0
So the domain of y is simply where both of these is true. But we need to figure out what sin^2(α) is. The bounds are completely irrelevant, since sin^2(x) has range [0, 1]. So the values under the square roots is 1 + [0, 1] = [1, 2]. The smallest negative value in the range is when sin^2(α) = 1, and -sqrt(1+1) = -sqrt(2). The largest positive value in the range is also when sin^2(α) = 1 and sqrt(1+1) = sqrt(2). But we can't include zero. So the range is [-\sqrt{2}, 0) \cup (0, \sqrt{2}].
I've looked at WolframAlpha on this problem, but the graph does go through zero (which is easy to see by setting f(x) = 0. cos x is zero at pi/2 + 2pi*k, so it goes through zero infinitely many times). The actual range is [-sqrt(2), sqrt(2)] with zero, but I don't know why there is an x in the denominator leaving it out of the domain.