You could not validate step (1) by saying that any number squared is an even number (e.g., $3^2 = 9$, but 3 is an odd number). As for explicitly why your proof is wrong, the statement that $a^2 = 4c^2$ is a non sequitur; that is, it does not logically follow from what you start out assuming, namely that $a^2 = 2b$. If you really want a direct proof, you could do something as follows, although it does not fall out quite nicely.
Direct proof: Suppose $a^2$ is even. Then $a^2 = 2b$, where $b\in\mathbb{Z}$. Then $\frac{a^2}{2}=b\neq 0 \rightarrow \left[\frac{a^2}{2}\right] = [0]$, where $[0]$ denotes an equivalence class representative of the even integers. This is because there are two congruence classes mod 2, namely the even and odd integers (for which $[0]$ and $[1]$ are equivalence class representatives, respectively). I doubt this is the kind of answer you were looking for though.
Alternative direct proof: Suppose $a^2$ is even. Then let us represent $a^2$ as follows:
$$
a^2 = a\cdot a = a_1\cdot a_2,
$$
where we know $a_1 = a_2$. Well, when two integers $c$ and $d$ are multiplied together and yield an even integer, then $c$ and $d$ must both be even or one must be even and the other odd (i.e., they cannot both be odd; if you use this fact, then you may need to prove it in case you cannot take it for granted). Well, since $a_1 = a_2$ and $a_1\cdot a_2 = a^2$, we must necessarily have that $a$ is even (this due to the fact that a single number cannot be both even and odd; if $a$ is odd, then we are multiplying two odd numbers, yielding an odd number as a product--this means $a$ can only be even then). This is kind of an ugly proof, but it may be more along the lines of what you are looking for.
A proof by contraposition seems ideal here though in terms of ease and clarity.
Thus a = 2c and a is even
I can't quite see the problem here
– k m Jan 12 '15 at 03:47