This question elaborates on a previous one of mine.

Let $F$ be a free group of finite rank $r>1$.

**Question.** Is it possible to choose some basis for $F$ which refines the lower central series and in which the expression for every $g\in F$ is finite?

*Explanation.*
The Hall basis of commutators $C=\{c_1,c_2,\ldots\}$ for $F$ (as defined in the
collecting process of M. Hall) can be used to express every element $g\in F$ in the form
$g=c_1^{n_1}c_2^{n_2}\ldots c_{t}^{n_t}g_k$ with
$g_k\in \gamma_k$, the $k$-th term of the lower central series for $F$.
This expression (for some $g\in F$) may grow infinitely as $k\to\infty$.
Yet, sometimes it is possible to change the basis $C$ to make this expression finite (for a fixed $g\in F$).

*Example.* If $F=\langle a,b \rangle$ and $g=(ab)^3$ then

$g=a^3b^3[b,a]^3g_3$.

We may choose a new basis $C_1=\{a,b,[b,a],g_3,\ldots\}$ in which the expression for $g$ becomes finite. (Note that $g_3$ may be included in a basis, because its expression in $C$ is

$g_3=[b,a,b]^5[b,a,a]g_4$

with coprime exponents of the terms of weight $3$.)
So, the question is whether this sort of "basis correction" can be made for all $g\in F$ *uniformly*.

*Remark.* By a *basis* I mean a set $c_1,c_2,...$ of elements of $F$ whose images (in the appropriate terms) form a basis of the free abelian group

$ \gamma_1/\gamma_2\oplus\gamma_2/\gamma_3\oplus\ldots$

This should explain the phrase "refines the lower central series".