Judy sent us the following
solution:
Suppose that we have a labelling with magic total
that looks like this:
Then we can replace each entry
by
and get another labelling (because
if
is one of 1, 2, ..., 9, then
will also be). Also, this
label will have magic total
, because instead of adding up
, for
example, we'll add up
, and all of the sums will be the
same. So if we have a labelling with magic total
, then we certainly have
one with magic total
.
Using the solution to the Magic W problem, we know that there is 1 magic
labelling for
, and there are 5 for
. So (using what we worked out
above) there's also one for
and 5 for
. (There can't be any more
for
, for example, because any labelling of
is also one of
.)
When do magic labellings exist? Well, again using the ideas from the solution
to the Magic W problem, we must have
. But
, so
, so
. Also,
, so
so
. So we only need to check whether there are any magic labellings
for
. Suppose that there is one. Then we have
, and also
, so
. But that's not allowed, so there are no
labellings of
.
To summarise, there are magic labellings for
, 14, 16 and 17 (and no
others), and there are
magic labellings.