![](/files/happy5.png)
A nowhere-zero point in a linear mapping
![$ {\mathbb F} $](/files/tex/0996310fb2d1a49c39f098ca07ee5323bf728b79.png)
![$ A $](/files/tex/7a8d9782350e8eb5a84c149576d83160492cbdd3.png)
![$ n \times n $](/files/tex/fd981d449b91b1f4889d87406e6aa7d8acfb5d68.png)
![$ {\mathbb F} $](/files/tex/0996310fb2d1a49c39f098ca07ee5323bf728b79.png)
![$ x,y \in {\mathbb F}^n $](/files/tex/817584f802a24b69508ae913bf47b34d6ff78ca8.png)
![$ Ax=y $](/files/tex/6fc49872cdb030ca09b427d20b9fd3b1d3873641.png)
The motivation for this problem comes from the study of nowhere-zero flows on graphs. If is the directed incidence matrix of a graph
, then a nowhere-zero
-flow on
is precisely a vector
so that
has all entries nonzero, and
. The above conjecture is similar, but is for general (invertible) matrices. Alon and Tarsi have resolved this conjecture for all fields not of prime order using their polynomial technique.
Definition: Say that a matrix
is
-choosable if for all
with
and for all
with
, there exists a vector
and a vector
so that
. Note that every matrix is
-choosable, but that an
matrix is
-choosable if and only if it is invertible.
Alon and Tarsi actually prove a stronger property than Jaeger conjectured for fields not of prime order. They prove that if has characteristic
, then every invertible matrix over
is
-choosable. This result has been extended by DeVos [D] who showed that every such matrix is
-choosable. Yang Yu [Y] has verified that the conjecture holds for
matrices with entries in
when
.
Jaeger's conjecture is true in a very strong sense for fields of characteristic 2. DeVos [D] proved that every invertible matrix over such a field is -choosable for every
. The following conjecture asserts that invertible matrices over fields of prime order have choosability properties nearly as strong.
![$ {\mathbb Z}_p $](/files/tex/e8c94ceb5a9d688bff114c12f7fe9fe47ef955fc.png)
![$ {\mathbb Z}_p $](/files/tex/e8c94ceb5a9d688bff114c12f7fe9fe47ef955fc.png)
![$ p $](/files/tex/928cd9d544fdea62f88a627aaee28c416c4366c0.png)
![$ (k+2,p-k) $](/files/tex/24a35960088c1598742842b2d45173281b1a652f.png)
![$ k $](/files/tex/c450c3185f7285cfa0b88d3a903c54f7df601201.png)
This is essentially the strongest choosability conjecture one might hope to be true over fields of prime order. I (M. DeVos) don't have any experimental evidence for this at all, so it could be false already for some small examples. However, I suspect that if The permanent conjecture is true, that this conjecture should also be true. In any case, I (M. DeVos) am offering a bottle of wine for this conjecture.
Bibliography
[A] N. Alon, Combinatorial Nullstellensatz, Combinatorics Probability and Computing 8 (1999) no. 1-2, 7-29. MathSciNet
[AT] N. Alon, M. Tarsi, A Nowhere-Zero Point in Linear Mappings, Combinatorica 9 (1989), 393-395. MathSciNet
[BBLS] R. Baker, J. Bonin, F. Lazebnik, and E. Shustin, On the number of nowhere-zero points in linear mappings, Combinatorica 14 (2) (1994), 149-157. MathSciNet
[D] M. DeVos, Matrix Choosability, J. Combinatorial Theory, Ser. A 90 (2000), 197-209. MathSciNet
[Y] Y. Yu, The Permanent Rank of a Matrix, J. Combinatorial Theory Ser. A 85 (1999), 237-242. MathSciNet
* indicates original appearance(s) of problem.