Counting number of ways to sum integers and half integers to a specific integer modulo N

291 Views Asked by At

I came across this while working on a Physics problem.

I want to count the number of ways I can get a nonnegative integer $k \in$ {$0,1, \cdots, N-1 $} where $$k \equiv \sum_{i=1}^{N} x_{i} n_{i}\mod N$$ where $x_{i}$ $\in$ {0,1} and $n_{i} = i-1$ when the number of non-zero $x_{i}$'s are odd and $n_{i} = \frac{2i-1}{2}$ when the number of non-zero $x_{i}$'s are even.

I would like to find a function (or an approximate form) for $A(k,N)$ (number of such ways to get $k$ for fixed $N$) as some sort of a "nice" expression in terms of $k$ and $N$, in the sense I would like to be able to compare $A(k,N)$ for different values of $k$ as opposed to just developing an strategy to compute it. Is this doable? I'd greatly appreciate any suggestions or insights on how to think about this.

1

There are 1 best solutions below

3
On BEST ANSWER

It's natural to work with the case split.

The sums $\sum_{i=1}^{N} x_{i} (i - 1)$ with an odd number of $x_i$s are the sums $0x_1 + \sum_{i=2}^{N} x_{i} (i - 1)$: since we can adjust $x_1$ to fix the parity, they're just the sums of all the subsets of $\{1, 2, \ldots, N-1\}$, which distribute almost uniformly over the $N$ values modulo $N$.[1] So an approximate form for the contribution of this case would be $\frac{2^N}{N}$.

The sums $\sum_{i=1}^{N} x_{i} (i - \frac12)$ with an even number of $x_i$s can be rewritten (as Michael Burr raised in comments) as $\frac12 \sum_{i=1}^{N} x_{i} (2i - 1)$. If $N$ is odd then the odd numbers $1$ to $2N-1$ alias modulo $N$ to the numbers $0$ to $N-1$ and the same observations apply as in the previous case. (Note that since $2k \equiv k \bmod N \iff k \equiv 0 \bmod N$, in general the bias of the sum of odd and even contributions will tend to cancel out). But if $N$ is even then the odd numbers $1$ to $2N-1$ alias modulo $N$ to the odd numbers $1$ to $N-1$ each with multiplicity $2$. The ($2^{N-1}$) sums of even numbers of them can only be even, but again distribute almost uniformly over the $\frac{N}2$ possible values, so here the contribution is approximately $\frac{2^N}N$.

In short: to first order, we find that $A(k, N) \approx \frac{2^{N+1}}{N}$ except for odd $k$ and even $N$, when it's half that.


[1] For more precision, see section 5 of When are subset sums equidistributed modulo m?, Wagon and Wilf, Electronic Journal of Combinatorics 1.1 (1994).