Calculate sum $$ \sum_{k=2}^{2^{2^n}} \frac{1}{2^{\lfloor \log_2k \rfloor} \cdot 4^{\lfloor \log_2(\log_2k )\rfloor}} $$
I hope to solve this in use of Iverson notation:
my try
$$ \sum_{k=2}^{2^{2^n}} \frac{1}{2^{\lfloor \log_2k \rfloor} \cdot 4^{\lfloor \log_2(\log_2k )\rfloor}} = \sum_{k,l,m}2^{-l}4^{-m} [2^l \le k < 2^{l+1}][2^{2^m} \le k < 2^{2^m+1}] $$
and now: $$ [2^l \le k < 2^{l+1}][2^{2^m} \le k < 2^{2^m+1}] \neq 0 $$ if and only if $$2^l \le k < 2^{l+1} \wedge 2^{2^m} \le k < 2^{2^m+1} $$
I can assume that $l$ is const (we know value of $l$) and treat $m$ as variable depence from $l$. Ok so:
$$2^l \le 2^{2^m} \wedge 2^{2^m+1} \le 2^{l+1} $$
but it gives me that $l=2^m$
I think that it is not true (but also I don't see mistake). Even if it is true, how can be it finished?
Here is an answer following rather closely OP's approach.
Comment:
In (1) we use Iverson brackets to get rid of the floor function.
In (2) we use an equivalent representation of the floor function.
In (3) we see the intervals inside the Iverson brackets suggest a partioning of the interval $\left[2,2^{2^n}\right]$. We do a first partitioning with respect to $m$ as union of right half-open intervals and add the value for $k=2^{2^n}$. This way we see $m$ takes precisely one value, namely $m=p$.
In (4) we continue similarly as we did in (3). This way we see $l$ takes precisely one value, namely $l=q$.
In (5) we use the finite geometric series formula.