Is there any kind of known pattern to $\sqrt 2$ in base 2? Is there any classification categories for decimal digits of numbers that for example would put $\sqrt 2, \sqrt 3 \cdots \sqrt n$ into separate category than $\sqrt[3] 2, \sqrt[3] 3 \cdots \sqrt[3] n$

Or given a decimal expansion with arbitrary precision, or definition of digit in the nth place (not necessarily decimal expansion) to get a probability of which class of numbers the number is likely to be? e.g. transcedental or algebraic of degree k?

  • $\begingroup$ If there was one on base $2$, then there would be one on base $10$ as well, making $\sqrt2$ rational. $\endgroup$ Nov 9, 2015 at 9:47
  • 5
    $\begingroup$ @barakmanos : The question did not ask for a "repeating pattern", e.g. liouville's number has a simple pattern and it is transcendental. $\endgroup$
    – jimjim
    Nov 9, 2015 at 9:50
  • 2
    $\begingroup$ I believe it remains unknown whether $\sqrt {2} $ is "normal" in any base i.e. for each natural number $n$ having uniform distribution of all $n$-tuples of digits in that base. $\endgroup$
    – Simon
    Nov 9, 2015 at 10:00
  • 1
    $\begingroup$ The wikipedia page is pretty interesting, and points out that it is unknown if $\sqrt 2$ is normal. In fact, we have never proven that a number is normal, except for numbers which were constructed to be normal. en.wikipedia.org/wiki/Normal_number $\endgroup$
    – pancini
    Nov 9, 2015 at 10:14
  • 1
    $\begingroup$ See this :community.wolfram.com/web/community/groups/-/m/t/1063480 $\endgroup$ Apr 17, 2017 at 8:03

2 Answers 2


One expression for the square root of 2 is given by $\sqrt{2} = \sum_{k=0}^{\infty} (2k+1) \binom{2k}{k} 2^{-3k -1} $ which is rather useful for computing the digits exactly. It also means that any pattern in the binary digits of $\sqrt{2}$ should be somewhat related to the the digits of the central binomial coefficients. There's probably a combinatorial interpretation of this.

Incidentally, $(2k+1) \binom{2k}{k}$ is asymptotic to $\frac{2^{2k + 1}}{\sqrt{\pi}}$ so the number of terms affecting the kth digit is linear in k. Also, the greatest power of 2 dividing the central binomial coefficient is known as Gould's sequence, i,e, $\binom{2k}{k} = 2^{2^{h(k)}}$ where $h(k)$ is the hamming weight of k.


there is a method to calculate decimal places of root2, probably that could help find a pattern with it. Although this is not a very obvious one, it probably could (very loosely) help to find a pattern. The method works a bit like dividing, so you have to calculate the next decimal place, then the next. It does not work on approximations by the way.


On this link you will see a few methods, but scroll down past the cyan shortcut box and you will see the one shown by Frank La Fontaine.

I'm sorry that I can't show an actual pattern, but hopefully this helps.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.