Normalized binomial mid-coefficients and power means (Q2581374)

For a non-negative integer \(n\) let \[ \mu_n=2^{-n}{2n\choose n} \] be the normalized binomial mid-coefficient. In [\textit{T. Bang, B. Fuglede}, J. Number Theory 35, 345--349 (1990; Zbl 0702.11017)] it was shown that the the only solutions of the diophantine equation \(\mu_p\mu_q=\mu_r\mu_s\) in non-negative integers \(p,q,r,s\) are the trivial ones \((p,q)=(r,s)\) or \((s,r)\). For a real number \(t\) and positive real numbers \(x_1,\ldots,x_n\) let \[ M_t(x_1,\ldots,x_n)=\biggl({{1}\over {n}}\sum_{j=1}^n x_j\biggr)^{1/t} \] if \(t\neq 0\) and \[ M_0(x_1,\ldots,x_n)=\biggl(\prod_{j=1}^n x_j\biggr)^{1/n} \] be the unweighted power mean of order \(t\) of the numbers \(x_1,\ldots,x_n\). In this paper, the authors show that the only solutions of the diophantine equation \[ M_k(\mu_p,\mu_q)=M_\ell(\mu_r,\mu_s) \] in integers \(k\) and \(\ell\) non-zero and \(p,q,r,s\) non-negative are the trivial ones \(k=\ell\) and \((p,q)=(r,s)\) or \((s,r)\) or \(k\neq \ell\) and \(p=q=r=s\). The proof is elementary and is based on a careful analysis of the \(2\)-adic valuation of the involved quantities. They apply the same technique to study a variant of the above diophantine equation for triples of the \(\mu_n\)'s. In this case, the result is that if \(k\neq 0,~-1\), \(0\leq a_1\leq a_2\leq a_3\) and \(0\leq b_1\leq b_2\leq b_3\) and furthermore \(a_1\leq b_1\), then \[ M_k(\mu_{a_1},\mu_{a_2},\mu_{a_3})=M_k(\mu_{b_1},\mu_{b_2},\mu_{b_3}) \] holds only in trivial cases except when \(k=1\), \((a_1,a_2,a_3)=(1,3,3)\) and \((b_1,b_2,b_3)=(2,2,2)\). The cases \(k=0,~-1\) are left as open problems. The authors also show that the equation \[ \mu_p^k+\mu_q^k=\mu_r^k \] has only the trivial solutions \((k,p,q,r)=(1,1,1,0)\) and \((-1,0,0,1)\) and ask for all the solutions of the arithmetic-geometric mean equation \[ {{1}\over {n}}(\mu_{a_1}+\ldots+\mu_{a_n})=(\mu_{b_1}\ldots\mu_{b_n})^{1/n}. \]