2
$\begingroup$

In number theory, an odious number is a positive integer that has an odd number of $1$s in its binary expansion.

The first odious numbers are:

$1, 2, 4, 7, 8, 11, 13, 14, 16, 19, 21, 22, 25, 26, 28, 31, 32, 35, 37, 38...$

Let $a$ denote the increasing sequence of odious numbers where $a_0=1$, $a_1=2$, etc.

Whats the summatory function of the sequence $b$ where $b_i=a_i^k$?

Also: this is useful; but it didn't give out the actual function. Please help.

$\endgroup$
4
  • $\begingroup$ Perhaps one can use the fact that any permutation of the binary digits of an odious number gives rise to the binary expansion of another odious number, so group theoretic considerations may be useful. $\endgroup$ Commented Mar 27, 2021 at 21:07
  • $\begingroup$ Also, defining an "inner product" on odious numbers $n_{1}$ and $n_{2}$ as $\sum_{i}b_{1}(i)b_{2}(i)$ where $n_{j}=\sum_{i\geq 0}b_{j}(i)2^{i}$, one can easily figure out that the sum of an odd number of pairwise orthogonal odious numbers is an odious number. $\endgroup$ Commented Mar 27, 2021 at 21:15
  • $\begingroup$ I would hazard a guess that, for $n$ large, roughly half the positive integers up to $n$ are odious, and the sum of their $k$th powers would be roughly $(1/2)\sum_{r=1}^nr^k$. $\endgroup$ Commented Mar 28, 2021 at 5:52
  • $\begingroup$ The sum $a_n$ of the first $n$ odious numbers is tabulated at oeis.org/A173209 and the formula $a_n=n^2-(n/2)+O(1)$, attributed to Charles R Greathouse IV, is given without proof. $\endgroup$ Commented Mar 28, 2021 at 5:58

2 Answers 2

6
$\begingroup$

I think there is no simple formula here, although we can get some recurrence relations and related identities for generating functions as explained below.


Similarly to odious numbers, we have evil numbers that contain even number of 1s in their binary representation.

Let $S^o_k(n)$ and $S^e_k(n)$ denote the sum of $k$-th powers of all odious and evil numbers $\leq n$, respectively.

It can be easily seen that each odious number has one of the two forms: $2m$ where $m$ is an odious number, or $2m+1$ where $m$ is an evil number. It follows that for $n\geq 1$ $$S^o_k(n) = S^o_k(\lfloor\tfrac{n}2\rfloor)2^k + \sum_{i=0}^k \binom{k}{i} S^e_i(\lfloor\tfrac{n-1}2\rfloor) 2^i$$ and similarly $$S^e_k(n) = S^e_k(\lfloor\tfrac{n}2\rfloor)2^k + \sum_{i=0}^k \binom{k}{i} S^o_i(\lfloor\tfrac{n-1}2\rfloor) 2^i.$$ We also have $S^o_k(0)=S^e_k(0)=0$ for all $k\geq 1$, and $S^o_0(0)=0$ and $S^e_0(0)=1$.


In terms of generating functions $F^o(x,y):=\sum_{n,k\geq 0} S^o_k(n) x^n \frac{y^k}{k!}$ and $F^e(x,y):=\sum_{n,k\geq 0} S^e_k(n) x^n \frac{y^k}{k!}$, we have $$F^o(x,y) = (1+x)F^o(x^2,2y) + (x+x^2) F^e(x^2,2y) e^y,$$ $$F^e(x,y) = (1+x)F^e(x^2,2y) + (x+x^2) F^o(x^2,2y) e^y.$$ Defining $G(x,y):=F^e(x,y) - F^o(x,y)$, we have $$G(x,y) = (1+x)(1 - x e^y)G(x^2,2y).$$ Also, it is easy to get $$F^e(x,y) + F^o(x,y) = \frac{1}{(1-x)(1-xe^y)}.$$

Since $G(x,y) = \frac{1}{(1-x)(1-xe^y)} - 2F^o(x,y)$, we a standalong functional equation for $F^o$: $$F^o(x,y) = (1+x)(1 - x e^y)F^o(x^2,2y) + \frac{xe^y}{(1-x)(1-x^2e^{2y})}.$$

$\endgroup$
0
$\begingroup$

Too big to comment: If we take the sequence of Gray code numbers to be $\{ g(m) \}_{m=0}^{\infty}$ then the $m$-th Odious number $Q(m)$ ($Q(0)=1,Q(1)=2...$) is the $g(m)$-th number of in the sequence of $g(2k+1), k=0,1,2,...,$

Hence, $Q(m)=g(2g(m)+1)$.

A formula for $m$-th Gray number is $$g(m)=\sum_{k=0} \sigma_i(m)2^{k-1}$$ where $\sigma_i(m)=\lfloor{\frac{|r_i(m)|}{2^k}}\rfloor$.

And, $r_i(m)=m-(\lfloor{\frac{m}{2^k}}\rfloor -(-1)^{\lfloor{\frac{m}{2^k}}\rfloor})2^k$.

$\endgroup$
2
  • 1
    $\begingroup$ Perhaps you should avoid the notation $O(m)$ as it has a totally different meaning in number theory. $\endgroup$ Commented Mar 28, 2021 at 6:31
  • $\begingroup$ @Sylvain JULIEN ,Oh, yes, I am changing it. Thank you. $\endgroup$
    – Alapan Das
    Commented Mar 28, 2021 at 6:33

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .