I am looking for a way to easily discover how to go from a nested root to a sum of roots. For example,
$$\sqrt{10-2\sqrt{21}}=\sqrt{3}-\sqrt{7}$$
I know that if I set $\alpha=\sqrt{10-2\sqrt{21}}$, square both sides, I get
$$\alpha^2=10-2\sqrt{21}$$
Now I recognize that we have a situation where $10=3+7$ and $21=7\cdot 3$, so I can immediately see that we have
$$\alpha^2=10-2\sqrt{21}=3-2\sqrt{21}+7=\sqrt{3}^2-2\sqrt{3}\sqrt{7}+\sqrt{7}^2=(\sqrt{3}-\sqrt{7})^2$$
My question is, it this the only way to approach this problem? This approach mirrors basic algebra 1 methods of factoring quadratics, but I was curious to know if thre are other techniques that can be used to quickly deduce that a nested radical can be simplified to the sum of two radicals. Mathematically, suppose $a,b,c,m,n,r,s\in\mathbb{N}$. Is there a way to quickly determine $m,n,r,s$ in the equation
$$\sqrt{a\pm b\sqrt{c}}=m\sqrt{r}\pm n\sqrt{s}$$
Suppose $a+b\sqrt{c}=(m\sqrt{r}+n\sqrt{s})^2$, where $a,b,c,m,n,r,s\in\mathbb{Z}$ and $r,s,c>0$. Then you have the following: $$a=m^2r+n^2s\\ b=2mn\\ c=rs\\ a+b\sqrt{c}>0$$
This shows that $b$ must be even, and the gcd's of $m,n$ and $r,s$ each divide $a$. I see no reason why this set of diophantine equations has a particularly clean solution. As an example of a case where you cannot get the desired form we have $\sqrt{64+35\sqrt{3}}$.