Presumably you already have a notation to indicate the $j$-the element of _any_ vector $\mathbf{v}$, and you should be consistent with this.
Also, for this particular algorithm (or for iterative algorithms in general), you probably have a way to indicate the value of variable $x$ during the $k$-th iteration. You ought to be consistent with this, too.
My personal choices would be:
(1) $\mathbf{v}[j]$ denotes the $j$-the element of the vector $\mathbf{v}$.
(2) $x_k$ denotes the value of the variable $x$ during the $k$-th iteration.
So, putting these two together, $\mathbf{a}_k[j]$ or $\mathbf{a}[j]_k$ is the value of $\mathbf{a}[j]$ during the $k$-th iteration. I prefer the former.
There are many other choices, of course; it's the consistency that's important, in my view.