# decimal representation of $2^m$ starts with a particular finite sequence of decimal digits

Given a finite sequence of decimal digits $a_1,a_2,...,a_n$ prove that there exists a natural number $m$ such that decimal representation of $2^m$ starts with that sequence of digits.

• .. with $a_1\ne 0$? – Hagen von Eitzen May 10 '14 at 13:36
• Do you know any relevant property of the rotations of the circle $S^1$? – Did May 10 '14 at 13:39
• @Did Thanks! Actually I prefer to solve this question using rotations of $S^1$, Could you give me a little more explanation ? Oh ! Do you mean rotations of the circle $[0,1]/0\~1$ by the map $R_{log2}$ which sends $x \in S^1$ to $x+log2 (mod 1)$ has dense orbits in the circle... – the8thone May 10 '14 at 14:33
• there is another answer here – user354674 Sep 8 '16 at 14:24

Hint: $\log 2$ is irrational. What can you say about $\{ n log 2 \}$, the fractional parts?

• the fractional parts of the sequence $\{m \log 2\}^{\infty}_{m=1}$ is $\{ \log 2,\log 4,\log 8,\log 1.6, \log 3.2 , \ldots \}$ which is a sequence of irrational numbers in $[0,1]$. If we know that this sequence is dense in $[0,1]$ we are done , but do we know that ?! – the8thone May 10 '14 at 14:26
• @the8thone Great. Your guess is true, now show it. Use the pigeonhole principle. – Calvin Lin May 10 '14 at 15:01
• I don't understand why are we done if the sequence $\{m \log 2\}$ is dense. Can you explain me, please? Thank you very much! – g.pomegranate Nov 19 '17 at 18:49
• @g.pomegranate More generally, for irrational $r$, $\{ m r \}$ is dense. For example, by the pigeonhole principle applied to the first 101 values (pigeons) and 100 intervals $[\frac{n}{100}, \frac{n+1}{100} ]$ (hole), we can find 2 values of $mr$ which lie in the same interval. Their positive difference is a value of the form $m'r$, and lies within $[0, 0.01]$. So, we have a value in every 0.01 interval. Extend to $100 \rightarrow \infty$. – Calvin Lin Nov 27 '17 at 16:09
• @CalvinLin Thank you very much! – g.pomegranate Nov 28 '17 at 18:44

It is not a proof, just a try to find by approximations the first solutions, without the guarantee that they exist.

I tried a small script to select candidates which were submited next to a large digital computer for example manually.

The principle is the decimal approximation. The algorithm keeps only the 11 digits of a number hoping that the truncature will not affect a lot the most left digits. To extend the search , it selects also a few approximative matches, always under the target, since the truncature makes the result smaller.

There is an example of implementation for the target 999 , easy to adapt to other ones:

function main()
{
var i, k = 68719476736,res="",res2="" ; /// begin with 2^36
for( i=37 ; i < 50000 ; i++ )
{
k = 2* ((""+k).substr(0,7)) ;

var kk = (""+k).substr(0,3) ;
if( kk=="999" ) // if( kk=="998" || kk=="999" )
{
res +=( k+" ("+i+"), " ) ;
res2 +=( "2^"+i+"," ) ;
}
}

return res+"\n"+res2+"\n" ;
}

// scratchpad formalism to output the result as the last seen variable
// type CTRL L to run
var tto , ttt = main() ;
tto = ttt ;


Truncated numbers selected with the power in parenthesis :

$99912223808 (2621), 99928497088 (4757), 99944773024 (6893),$ $99961051568 (9029), 99977332768 (11165), 99993616640 (13301),$ $99905846016 (15922),$ $99922118264 (18058), 99938393032 (20194), 99954670632 (22330),$ $99970950824 (24466), 99987233664 (26602),$ $99915739784 (31359), 99932013584 (33495), 99948290144 (35631),$ $99964569264 (37767), 99980851144 (39903),$ $99997135584 (42039), 99909361920 (44660), 99925634840 (46796),$ $99941910296 (48932); ...$

After external tests, $2^{2621}$, $2^{4757}$, $2^{6893}$ , $2^{9029}$ , $2^{11165}$ seem to work well with the target of the test $999$.

edit : keeping 11 significative digits and testing only the target, I get directly the same begining of the sequence solution.