On the local constancy of certain mod p Galois representations
Loading...
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
IISER Mohali
Abstract
In this thesis, we study local constancy in the weight space of the mod p reduction of
certain irreducible 2-dimensional crystalline representations V k,a p of Gal Q̄ p /Q p
using the mod p local Langlands correspondence for GL 2 (Q p ). Indeed all irreducible
2-dimensional crystalline representations of Gal Q̄ p /Q p are a twist by a crystalline
character of some V k,a p . Latter is uniquely determined by the Hodge-Tate weights
(0, k − 1) and the trace of the Frobenius a p . Let V k,a p denote the mod p reduction
of a Gal Q̄ p /Q p -stable lattice up to semisimplification. For a fixed a p , we study
local constancy of V k,a p in the weight space. In order to prove local constancy, we
show that V k ′ ,a p is constant for all k ′ ∈ k + p t (p − 1)Z >0 (t sufficiently large) by
explicitly computing V k ′ ,a p . Latter is obtained using the mod p local Langlands
correspondence for GL 2 (Q p ). This gives local constancy in the punctured disk
{k ′ | k ′ ∈ k + p t (p − 1)Z >0 } around k. Next, we determine the structure of V k,a p by
applying a result of Berger-Li-Zhu together with Berger’s local constancy theorem
to establish constancy in the whole disk. This gives a lower bound on the radius of
local constancy solely in terms of a p .
Let k = b + c(p − 1) + 2 where 2 ≤ b ≤ p and 0 ≤ c ≤ p − 2. Broadly speaking,
our results fall into two regimes (c, p − 1) and (1, c + 1) of the slope ν(a p ). In
the former situation, we establish local constancy under certain constraints on b, c
and ν(a p ). We also in this case improve on the lower bound for k in Berger’s
theorem guaranteeing local constancy. When ν(a p ) < c + 1, we establish local
constancy (in the irreducible cases) for non-integral slope under the condition that
b ≥ c + ⌊ν(a p )⌋ − 1 holds. Lastly, assuming a certain technical conjecture (for
which we provide good evidence), we establish local constancy for certain subsets of
b ≤ c + ⌊ν(a p )⌋ − 2 and such that the slope is non-integral, lying in (1, c − 1).