Let $d>1$, and let $\alpha_1,\dots,\alpha_d,\beta_1,\dots,\beta_d$ be positive real numbers. Is there an easy proof of the following claim:
If $\alpha_j-(\Pi_{i=1}^d \alpha_i)^{1/d}=\beta_j-(\Pi_{i=1}^d \beta_i)^{1/d} $ for every $1\le j\le d$ and not all of these differences are zero, then $\alpha_i=\beta_i$ for every $i$.
If we allow all of the differences to be zero, then uniqueness fails: set $\alpha_i=a,\beta_i=b$ for $a \neq b$.
I think I have a proof for this uniqueness, but it is very cumbersome. I am interested in seeing other approaches.