2

Given a bundle of k linear independent vectors, each vector coordinates is some permutation of $\{1, \ldots, n\}$. Prove that linear span of these vectors contains not more than $k!$ vectors whose coordinates are permutation of $\{1,\ldots,n\}$.

My observations

  • call k linear independent vectors $\sigma_1, \ldots, \sigma_k$. Then if $\sigma=a_1\cdot\sigma_1+\cdots+a_k\cdot\sigma_k$(*), then $a_1+\cdots+a_k=1$ (write $(*)$ as linear system, sum all n rows and divide for $(1+\cdots+n)$
  • $k!$ is achievable. Let take $\sigma_1=\operatorname{id}$, $\sigma_i$ is $\operatorname{id}$ with cyclically shifted first $k$ coordinates ($k=3$: $(1,2,3,4,5) \rightarrow (3,1,2,4,5) \rightarrow (2,3,1,4,5)$ they are independent (easy to calculate determinant)). Let's take $k!$ permutations of $\{1,\ldots,k\}$ and just add $\{k+1,\ldots,n\}$ at the end. It is obvious that rank of this bundle is less than or equal $k$. (But we need it is at least $k$.)

I don't need full solution, but I really ask for hint. Problem is beautiful but i'm at a dead end. The only one idea is induction. But it just don't work, because different bundles of k linear independent vectors have so different spans.

joriki
  • 238,052
veirab
  • 41
  • Hint: First induct on $k\in \mathbb N$ ($k=1$ trivially holds). Then induct on $n$ w/ Base Case $n=k$ the bound is trivial [however the fact that there must exist of $k$ linearly independent vectors from permuting your ordered set is not obvious and the proofs I have in mind use representation theory]. Inductive Case: $n \gt k$ first prove the nice case where last coordinate is identically $=n$, then do the more general case. Note: the use of 'bundle of vectors' is unfortunate as a vector bundle is a defined term in topology but evidently what you mean is 'collection of vectors' here. – user8675309 Feb 11 '23 at 18:53
  • Could you please provide me with a link containing a proof of this fact, using representation theory. Perhaps you also know what I could read about representation theory so that I can understand the proof? – veirab Apr 10 '23 at 07:55
  • This 2 months old so some of the details I had in mind have faded. I suggest you go through the Representation Theory Chapter in Artin's Algebra. The proof of the fact comes from solving its exercises. In the 1st edition: ex 9.6.9; in the 2nd edition ex 10.6.5. For a character theory solution to said exercises: https://math.stackexchange.com/questions/4240957/fixed-points-of-permutation-groups/ however the 2nd edition suggests skipping rep/character theory arguments and using the solution of ex 4.M.1 instead [which does not require rep theory though I have forgotten its direct solution]. – user8675309 Apr 10 '23 at 16:56

0 Answers0