Abstract:We prove the existence of a projective good moduli space of principal $\mathcal{G}$-bundles under nonconnected reductive group schemes $\mathcal{G}$ over a smooth projective curve $C$. We also prove that the moduli stack of $\mathcal{G}$-bundles decomposes into finitely many substacks $\text{Bun}_{\mathcal{P}}$ each admitting a torsor $\text{Bun}_{\mathcal{G}_\mathcal{P}}\to \text{Bun}_{\mathcal{P}}$ under a finite group, for some connected reductive group schemes $\mathcal{G}_\mathcal{P}$ over $C$. We use this for the second purpose of the article: for any constant connected reductive group $G$, the subset of essentially finite $G$-bundles in the moduli of degree 0 semistable $G$-bundles over $C$ is not dense, unless $G$ is a torus or the genus of $C$ is smaller than 2. We do this by giving an upper bound on the dimension of the closure of the subset of essentially finite $G$-bundles.
What problem does this paper attempt to address?
This paper attempts to solve two main problems:
1. **Existence of the moduli space of principal bundles under the action of non - connected groups**:
- The paper proves that on a smooth projective curve \(C\), for a non - connected reductive group scheme \(G\), there exists a well - projected moduli space which contains all \(G\)-principal bundles.
- Specifically, the paper proves that the moduli stack \(\text{Bun}_G\) of \(G\)-principal bundles can be decomposed into finitely many sub - stacks \(\text{Bun}_P\), and each \(\text{Bun}_P\) is a torsor under a finite group \(\text{Bun}_{G_P}\to\text{Bun}_P\), where \(G_P\) is a connected reductive group scheme on \(C\).
- This result further proves that in the case of characteristic zero, the moduli stack \(\text{Bun}^{\text{ss}}_G\) of semi - stable \(G\)-principal bundles has a well - projected moduli space.
2. **Non - density of essentially finite \(G\)-principal bundles**:
- The paper solves the density problem of essentially finite \(G\)-principal bundles in the moduli space under the connected reductive group \(G\).
- Specifically, the paper proves that when the characteristic is zero and the genus \(g\) of \(C\) is \(g\geq2\), for any connected reductive group \(G\), the set \(M^{\text{ef}}_G\) of essentially finite \(G\)-principal bundles is not dense in the moduli space \(M^{\text{ss},0}_G\) of semi - stable \(G\)-principal bundles, unless \(G\) is a torus or the genus of \(C\) is less than 2.
- To prove this, the paper gives an upper bound for the dimension of the closure of essentially finite \(G\)-principal bundles.
### Summary of main results
1. **Existence of the moduli space**:
- **Theorem 1.1**: Suppose \(\text{char}(k) = 0\), and \(G\) is a (not necessarily connected) reductive group scheme on \(C\). Then the moduli stack \(\text{Bun}^{\text{ss}}_G\) of semi - stable \(G\)-principal bundles has a well - projected moduli space.
- **Theorem 1.2**: Suppose \(G\) is a (not necessarily connected) reductive group on \(C\). If \(\text{char}(k)>0\), assume that \(G / G^{\circ}\) is finite, and \(\text{char}(k)\) does not divide \(|G / G^{\circ}|\). Then there exists a finite decomposition \(\text{Bun}_G\cong\coprod_P\text{Bun}_P\), where each \(\text{Bun}_P\) is a torsor under a finite group \(\text{Bun}_{G_P}\to\text{Bun}_P\), and \(G_P\) is a connected reductive group scheme on \(C\).
2. **Non - density of essentially finite principal bundles**:
- **Theorem 1.5**: Let \(\text{char}(k) = 0\), let \(g\) denote the genus of \(C\), and assume \(g\geq2\). For any connected reductive group \(G\), we have
\[
\dim M^{\text{ef}}_G\leq g\cdot\text{rk}(G).
\]
In particular, if \(G\) is not a torus, then \(M^{\text{ef}}_G\) is not