My solution of POW12 was accepted. [Remark of December 19, 2009: it is only stated that it was "completely or partially proved"]
(Tradução aqui)
“ Problem. Find, with proof, the maximum value of
where
and
is variable. In particular, your answer must be greater than or equal to all values obtained from other choices of
“
Here is the solution I submited.
Solution.
Let
,
,
and
. For a given
, with
, we know by the Lagrange multipliers method that
is a local extremum if for 


where
is the value of the multiplier
that is a solution of these
equations. Hence we have respectively

and

Solving this system of equations we find

and
,
the latter being a local extremum of
, for a fixed
. We transformed the initial maximizing problem in
continuous variables and the discrete variable
into the maximizing of
. Now we introduce the following function

The function
has a maximum at the same point
than the function

On the other hand
for
for
and
for
. Therefore

is a maximum of
. Since
, the maximum occurs at
. Thus, for
we have

Update (Dec., 19,2009): some errors (identified by Rod Carvalho here) corrected.