Abstract:We construct a local deformation problem for residual Galois representations $\bar{\rho}$ valued in an arbitrary reductive group $\hat{G}$ which we use to develop a variant of the Taylor-Wiles method. Our generalization allows Taylor-Wiles places for which the image of Frobenius is semisimple, a weakening of the regular semisimple constraint imposed previously in the literature. We introduce the notion of $\hat{G}$-adequate subgroup, our corresponding 'big image' condition. When $\hat{G}$ is a simply connected simple group of type $\mathrm{C}$ or of exceptional type and $\hat{G} \to \mathrm{GL}_n$ is a faithful irreducible representation of minimal dimension, we show that a subgroup is $\hat{G}$-adequate if it is $\mathrm{GL}_n$-irreducible and the residue characteristic is sufficiently large.
We apply our ideas to the case $\hat{G} = \mathrm{GSp}_4$ and prove a modularity lifting theorem for abelian surfaces over a totally real field $F$ which holds under weaker hypotheses than in the work of Boxer-Calegari-Gee-Pilloni. We deduce some modularity results for elliptic curves over quadratic extensions of $F$.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to generalize the Taylor - Wiles method to arbitrary reductive groups in order to prove the modularity theorem of abelian surfaces over totally real fields. Specifically, the author introduced a new construction method of local deformation problems, which allows the lifting of Galois representations of specific types, and developed a variant of the Taylor - Wiles method on this basis. This generalization enables the author to prove the modularity lifting theorem under weaker assumptions.
### Main contributions
1. **Generalization of the Taylor - Wiles method**: The author generalizes the Taylor - Wiles method to an arbitrary reductive group \(\hat{G}\), allowing the image of the Frobenius element to be a semisimple element, rather than the regular semisimple element required in previous literature.
2. **Introduction of the concept of \(\hat{G}\)-ample subgroups**: The author defines \(\hat{G}\)-ample subgroups, which is a weakened version of the "big image" condition. For some simple linear algebraic groups, the author proves that absolute irreducibility implies \(\hat{G}\)-ampleness.
3. **Application to the modularity of abelian surfaces and elliptic curves**: The author applies these theories to the modularity problems of abelian surfaces and elliptic curves, and in particular, proves the modularity lifting theorem over totally real fields, relaxing the previous assumptions.
### Specific problem descriptions
- **Local deformation problems**: The author considers the lifting problem of the Galois representation \(\bar{\rho}\) in the reductive group \(\hat{G}\) and introduces local deformation problems, allowing ramification at some auxiliary prime points.
- **Semisimple Frobenius images**: Different from previous literature, the author allows the image of the Frobenius element to be a semisimple element, rather than a regular semisimple element. This allows more Galois representations to be considered.
- **\(\hat{G}\)-ample subgroups**: The author defines \(\hat{G}\)-ample subgroups, which is a weakened version of the "big image" condition, ensuring the existence of a set of Taylor - Wiles prime points that satisfy certain Galois cohomology conditions.
### Application examples
- **Modularity of abelian surfaces**: The author proves that over a totally real field \(F\), the modularity theorem of abelian surfaces can hold under weaker assumptions.
- **Modularity of elliptic curves**: By applying the results of abelian surfaces to elliptic curves, the author obtains the modularity lifting theorem of elliptic curves over the quadratic extension \(K/F\).
### Summary
The main objective of this paper is to provide a new framework for studying the modularity problems of Galois representations by generalizing the Taylor - Wiles method and to prove the modularity lifting theorem under more general conditions. This work not only extends the application range of existing theories but also provides new tools and methods for future research.