Is anything known about the following question? Given a positive integer $p$ and a real number $0<\alpha<1$, what partition $\lambda$ whose parts sum to $\alpha p^2$ (asymptotically) and whose diagram fits in a $p\times p$ square maximizes the number of semistandard Young tableaux of shape $\lambda$ and largest part at most $p$? In particular, is there a limiting curve like there is for the partition $\lambda$ of $n$ maximizing the number of standard Young tableaux of shape $\lambda$? By the Frame-Robinson-Thrall hook length formula, this latter question is related to the variational problem $$ \min\int_A \log(f(x)+f^{-1}(y)-x-y)dx\,dy, $$ over all nonincreasing nonnegative $f$ on $(0,\infty)$ satisfying $\int_0^\infty f(x)dx =1$, and where $A$ is the region between the $x$-axis and the graph of $f(x)$. See for instance http://stat.wharton.upenn.edu/~shepp/publications/55.pdf. For the first question, I believe that the corresponding variational problem is by the hook-content formula $$ \max\int_A (\log(1+x-y)-\log(f(x)+f^{-1}(y)-x-y))dx\,dy, $$ where $f\colon (0,1)\to(0,1)$, $f$ is nonincreasing, and $\int_0^1 f(x)dx = \alpha$. A number of similar questions can be formulated.
$\begingroup$
$\endgroup$
2
asked Mar 3, 2015 at 21:12
-
$\begingroup$ Interesting, this would also, in the limit, maximize the volume of the corresponding Gelfand-Tsetlin polytope, with boundary values given by the restriction on the partitions... $\endgroup$– Per AlexanderssonCommented Mar 3, 2015 at 21:37
-
$\begingroup$ I am almost sure that the limit curve exists, may be calculated less or more explicitly, and the maximal diagram is also near this curve. $\endgroup$– Fedor PetrovCommented Mar 3, 2015 at 23:33
Add a comment
|