What are we asked to find? Do we need to find x itself?
We could just cube the expression for ax and then work out a−x and so on. But that seems quite painful. I wonder whether there is a simpler way?
Solution
The question asks about ax, not about x itself. So if we write u=ax, the problem will look a little simpler. As a>1 and x>0, it follows that u>1. Then the question becomes:
Given that 2y=u+1/u, where u>1, prove that u=y+y2−1. If we multiply both sides of the equation by u, we get 2yu=u2+1, which rearranges to u2−2yu+1=0. Then using the quadratic formula we get u=22y±(−2y)2−4=y±y2−1.
There are two roots to the equation 2y=u+1/u, and they will be reciprocals of each other, as replacing u by 1/u gives exactly the same equation. So one of the roots will be greater than 1, the other will be less than 1. Since we require u>1, we need to take the positive sign.
So ax=u=y+y2−1.
Another way of arguing which sign to use goes like this.
Since u>1, 1/u<1<u, so 2y=u+1/u<u+u=2u, and thus y<u or u>y. So we have to use the positive sign.
Continuing to write u=ax, we are told that 2z=u3+u31. If we cube 2y=u+u1, we obtain (2y)3=(u+u1)3=u3+3u+u3+u31=2z+3(2y), so 8y3=2z+6y. Dividing by 2 and rearranging gives 4y3−3y=z.
This is related to the hyperbolic functions. If a=e, then y=coshx and z=cosh3x, while in general, since ax=exlna, we have y=cosh(xlna) and z=cosh(3xlna). The second result then shows that cosh3x=4cosh3x−3coshx, which is the same as the corresponding formula for (ordinary) cosines.