On the distribution of the sum of digits of sums \(a+b\) (Q2422102)

For \(N>2\) let \(\mathcal{A}, \mathcal{B}\) be subsets of \(\{1,2, \ldots,N\}\), and denote by \(s(n)\) the binary sum of digits of the integer \(n\). The authors investigate the distribution of \(s(a+b)\) for \((a,b)\in \mathcal{A}\times \mathcal{B}\). They prove that, as \(N\rightarrow \infty\), \begin{align*} \frac{1}{|\mathcal{A}| |\mathcal{B}|} &\left |\{(a,b)\in \mathcal{A}\times \mathcal{B},\;\; s(a+b)<z\}\right|\\ &=\Phi(y_{2N})+O\left(\frac{N}{\sqrt{|\mathcal{A}| |\mathcal{B}|}} \frac{(\log \log N) (\log \log \log N)^{1/2}}{(\log N)^{1/4}}\right), \end{align*} uniformly in \(z\) with \(0<z<\log(2N)/\log 2\), where \(\Phi(u)=(2\pi)^{-1/2}\int_{-\infty}^u e^{-t^2/2} dt\), and \(y_{2N}=y_{2N}(z)\) is defined by \[z=\frac{\log 2N}{\log 4}+y_{2N} \cdot \frac{1}{2}\left(\frac{\log 2N}{\log 2}\right)^{1/2}.\] As the authors note, this result implies that, as long as \[\sqrt{|\mathcal{A}| |\mathcal{B}|}\cdot \frac{(\log N)^{1/4}}{N \log \log N} \rightarrow \infty,\quad N\rightarrow \infty,\] the distribution of \(s(a+b)\) for \((a,b) \in \mathcal{A}\times \mathcal{B}\) resembles the distribution of \(s(n)\) for \(n\leq 2N\).