Let $F\subset K$ be finite field extension. It is clear $[K:F]$ is finite saying $K$ is a finite dimensional vector space. Consider the root of $x^2-a=0,a\in F$.
Here is a statement saying if $F,K$ finite fields, $[K:F]=2l,l\in N-\{0\}$, $x\not\in F$ and $x^2-a=0$ with $a\in F$, then $x\in K$. I cannot prove this directly using degree formula as this does not give me any quite useful information. I want to show $[K(x):K]=1$. So $[K(x):F]=[K(x):K][K:F]=[K(x):F(x)][F(x):F]\leq 4l$. So $[K(x):K]\leq 2$
However consider $Q\to Q[2^{\frac{1}{4}}]$ degree 4 extension by $x^4-2$ being irreducible through eisenstein criterion where $Q$ is rational number and the map is embedding $Q$ as the subfield of $Q[2^{\frac{1}{4}}]$. Say $x^2+1=0,1\in Q$ for sure. I do not have $i\in Q[2^{\frac{1}{4}}]$.
What is the reconcillation here? It seems finite field and infinite field behaves quite differently here.
The structure theorem of finite fields and finite extensions of finite fields is quite strong.
The first one is key here. Let's say $F = \operatorname{GF}(p^m)$.
If $p = 2$ then the map $x \mapsto x^2$ is an automorphism. Thus every element of $F$ has a square root.
If $p > 2$ then half the elements of $F^\times$ are squares.
(See The number of elements which are squares in a finite field. for details.)
So for $p = 2$ the equation $x^2 - a$ has all of its roots already in $F$. Next, suppose $p \ne 2$ and $a, b \in F^\times$ are two non-squares. Then
$$ F[x]/(x^2 - a) \text{ and } F[x]/(x^2 - b) $$
are fields of order $p^{2m}$ and hence by point (1.) above they are isomorphic.