3
$\begingroup$

Given a finite dimensional algebra $A$ with two indecomposable modules $M$ and $N$. Define $H(M,N)$ as the largest number of indecomposable summands of a module $X$ such that there exists a non-split short exact sequence $0 \rightarrow N \rightarrow X \rightarrow M \rightarrow 0$. Let $R_i$ be a basis of $Ext_A^1(M,N)$ such that $R_i$ is representated by short exact sequences with middle terms $X_i$. Let $|W|$ denote the number of indecomposable summands of a module $W$.

Question: Is it true that $H(M,N)$ is equal to the largest number of the $|X_i|$ independent of the choice of a basis?

I posted this here https://math.stackexchange.com/posts/3381428/edit with no answers.

$\endgroup$
1
$\begingroup$

I don't think so: If $A$ is of finite representation type, the space $\mathrm{Ext}^1(M, N)$ is algebraically stratified so that each stratum corresponds to an isomorphism class (after base change to algebraic closure) of $X$. Consequently, there is a dense stratum in $\mathrm{Ext}^1(M, N)$ which gives a maximal extension (and it tends to have a minimal number of indecomposables, morally).

For example, take the quiver with vertices $I = \{1, \ldots, 4\}$ and with adjacency matrix $$ Q = \begin{pmatrix}0 & 0 & 0 & 0 \\ 1 & 0 & 0 & 0 \\ 0 & 1 & 0 & 0 \\ 0 & 1 & 0 & 0\end{pmatrix} $$ In other words, it is an oriented Dynkin quiver of type $D_4$ where the vertex $1$ is a sink and $3, 4$ are sources. Let $M$ be the only indecomposable representation with dimension vector $(0, 1, 1, 1)$ and $N$ the indecomposable with dimension vector $(1, 1, 0, 0)$. For $i \in I$, let $P(i)$ denote the projective cover of the irreducible representation $L(i)$ corresponding to $i$.

Then $M$ has a projective resolution $0\to P(1)\oplus P(2)\to P(3)\oplus P(4)\to M\to 0$. Hence $\mathrm{Ext}^1(M, N)$ is of dimension $2$. We have the following two extensions of $M$ by $N$ who have $2$ indecomposables: $(1, 1, 1, 0)\oplus (0, 1, 0, 1)$ and $(1, 1, 0, 1)\oplus (0, 1, 1, 0)$. They represent two lines (minus the origin) in $\mathbf{C}^2$. However, there obviously exists an indecomposable extension of $M$ by $N$, which is the injective hull $I(1)$ of $L(1)$. In fact any element of the complement of $\mathbf{C}\times \{0\}\cup \{0\}\times\mathbf{C}$ is isomorphic to $I(1)$. Since this complement is Zariski-dense in $\mathbf{C}^2$, it follows that the largest number of indecompsables can be either $1$ or $2$, depending on the choice of basis.

I'm not expert of quiver representations, so please excuse me if the notations are terrible or if there is any error.

| cite | improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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