MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I would be grateful for any info concerning the following question.

Given an n-tuple z=(z_1,...,z_n) of nonzero complex numbers, let X(z) denote the GL(n,C)-conjugacy class of the diagonal matrix diag(z_1,...,z_n).

Let Y(z) be an open subset of X(z) formed by the matrices g in X(z) such that each of the principal minors d_1(g), d_2(g),...,d_n(g), of g, is nonzero (by definition, d_k(g) := determinant of the upper-left k by k block of the matrix g).

I am interested in the following

Problem. Compute the first Betti number, i.e. the rank of the first homology group with rational coefficients, of the manifold Y(z), for a sufficiently general n-tuple z=(z_1,...,z_n).

share|cite|improve this question

Below now is a proof that for sufficiently generic $z$, $b_1(Y(z))=n-1$.

First we show $b_1(Y(z))\geq n-1$.

Let $d=(d_1,d_2..,d_{n-1}): Y(z)\to({\mathbb C}^*)^{n-1}$. One can explicitly lift the loops in $H_1(({\mathbb C}^*)^{n-1})$ as follows. Fix $i$ between $1$ and $n-1$. Let $e(t)=exp(2\pi \sqrt{-1} t)$ and $$B(t):=\left(\begin{array}{cc} e(t) & 1\\\ e(t)(z_i+z_{i+1}-e(t))-z_iz_{i+1} & z_i+z_{i+1}-e(t) \end{array}\right).$$ Then the characteristic polynomial of $B(t)$ is $(\lambda-z_i)(\lambda-z_{i+1})$. Form an $n\times n$ matrix $A(t)$ as having $z_1,\dots,z_{i-1}$ on the main diagonal, then $B(t)$, then $z_{i+2},\dots,z_{n}$ and zero everywhere else. This matrix has the property that $A(t)\in Y(z)$ (as we assumed generic z) and $d_j(A(t))=z_1\cdot\dots\cdot z_j$ unless $j=i$ when $d_i(A(t))=e(t)z_1\cdot\dots\cdot z_{i-1}$. Thus $A(t)$ is a loop in $Y(z)$ covering the $i$th generator of $H_1(({\mathbb C}^*)^{n-1})$.

Thus we find that $d_*:H_1(Y(z))\to H_1(({\mathbb C}^*)^{n-1})$ is surjective so $b_1(Y(z))\geq n-1$.

To argue that $b_1\leq n-1$ we first show that the divisors $D_i:=d_i^{-1}(0)$ are irreducible on $X(z)$ for sufficiently generic $z$. We note that $D_i$ is irreducible in $M_{nxn}$ which can be easily seen over a finite field by counting points on $D_i$. This implies that for a sufficiently generic $z$ the divisor $D_i$ is irreducible on $X(z)$. When $z_i\neq z_j$ for $i\neq j$ we see that $X(z)$ fibers over the full flag variety with contractible fibers, so it is simply connected. From the homology Gysin sequence we get that removing an irreducible divisor from a smooth variety can increase $b_1$ by at most one. By induction we can conclude that $b_1(Y(z))\leq n-1$.

As an example take $n=2$. Here we are removing a copy of ${\mathbb C}^*$ from $X(z)$ an affine line bundle over ${\mathbb P}^1$. This will have Betti numbers $b_1=1$ and $b_2=2$ with weights $q,q,q^2$ respectively, to give Serre-polynomial $q^2+1$. We will have the map $d=d_1:Y(z) \to {\mathbb C}^{*}$. The fibers of this map will generically be a ${\mathbb C}^{*}$ and will have two singular fibers, which will give the non-trivial fundamental group. The point is that one can lift the loop generating $\pi_1({\mathbb C}^*)$ as generically $d_1$ is a fibration with connected fibers.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.