Skip to content

Deformation Theory

Summary

Mazur's deformation theory asks: given a residual Galois representation \(\bar{\rho}: G_{\mathbb{Q}} \to \text{GL}_2(\mathbb{F}_p)\), which "lifts" to \(\text{GL}_2(A)\) for local rings \(A\) exist? The universal deformation ring \(R\) parametrises all admissible lifts, the Hecke algebra \(T\) the modular ones. Wiles' goal: \(R = T\).

Prerequisites


1. The Starting Point

What we have

From the results of Langlands–Tunnell (for \(p = 3\)) or via the 3-5 switch (for \(p = 5\)), we know: for a semistable elliptic curve \(E/\mathbb{Q}\), the residual representation

\[ \bar{\rho} = \bar{\rho}_{E,p}: G_{\mathbb{Q}} \to \text{GL}_2(\mathbb{F}_p) \]

is modular – it comes from a newform.

What we want to show

We must prove that the full \(p\)-adic representation

\[ \rho = \rho_{E,p}: G_{\mathbb{Q}} \to \text{GL}_2(\mathbb{Z}_p) \]

is also modular. The residual representation is the reduction modulo \(p\): \(\rho \bmod p = \bar{\rho}\).

The lifting question

The problem can be formulated as follows: among all representations \(\rho: G_{\mathbb{Q}} \to \text{GL}_2(\mathbb{Z}_p)\) that reduce modulo \(p\) to the given \(\bar{\rho}\), which are modular? Wiles' answer: all of them (under suitable conditions).

For this, one needs a systematic tool to describe the "space of all lifts" – and that is precisely what Mazur's deformation theory provides.


2. What Is a Deformation?

Local rings

A complete local Noetherian ring \(A\) with residue field \(\mathbb{F}_p\) is a ring of the form

\[ A = \varprojlim A/\mathfrak{m}^n, \]

where \(\mathfrak{m}\) is the maximal ideal and \(A/\mathfrak{m} \cong \mathbb{F}_p\). Examples:

  • \(A = \mathbb{F}_p\) (trivial lift – only the residual representation)
  • \(A = \mathbb{Z}_p\) (the \(p\)-adic integers)
  • \(A = \mathbb{Z}_p[[x_1, \ldots, x_n]]\) (formal power series rings)
  • \(A = \mathbb{Z}_p[x]/(x^2)\) (dual numbers – for infinitesimal deformations)

Lifts

A lift of \(\bar{\rho}\) to \(A\) is a continuous homomorphism

\[ \rho_A: G_{\mathbb{Q}} \to \text{GL}_2(A), \]

that reduces modulo \(\mathfrak{m}\) to the given representation \(\bar{\rho}\):

\[ \rho_A \pmod{\mathfrak{m}} = \bar{\rho}. \]

Deformations

Two lifts \(\rho_A\) and \(\rho_A'\) are called equivalent if they can be conjugated into each other by a matrix \(M \in \ker(\text{GL}_2(A) \to \text{GL}_2(\mathbb{F}_p))\):

\[ \rho_A' = M \rho_A M^{-1}, \qquad M \equiv I_2 \pmod{\mathfrak{m}}. \]

A deformation is an equivalence class of lifts. The passage from lifts to deformations eliminates the "inessential" degrees of freedom from the choice of basis.


3. The Universal Deformation Ring \(R\)

Mazur's representability theorem

The central result of Barry Mazur (1989) is:

Theorem (Mazur, 1989)

Let \(\bar{\rho}: G_{\mathbb{Q}} \to \text{GL}_2(\mathbb{F}_p)\) be a continuous, irreducible representation. Then there exists a universal deformation ring \(R\) (a complete local Noetherian ring with residue field \(\mathbb{F}_p\)) and a universal deformation $$ \rho^{\text{univ}}: G_{\mathbb{Q}} \to \text{GL}_2(R), $$ such that every deformation of \(\bar{\rho}\) to a ring \(A\) uniquely factors through a local ring homomorphism \(R \to A\).

In the language of category theory: the functor "deformations of \(\bar{\rho}\)" is representable, and \(R\) is the representing object.

What does this mean concretely?

The universal deformation ring \(R\) is the "largest possible" lift:

  • Every deformation of \(\bar{\rho}\) to \(\mathbb{Z}_p\) arises via a ring homomorphism \(R \to \mathbb{Z}_p\) (specialisation of the universal deformation).
  • The structure of \(R\) encodes all information about all possible lifts simultaneously.
  • \(R\) can be written as a \(\mathbb{Z}_p\)-algebra: \(R \cong \mathbb{Z}_p[[x_1, \ldots, x_r]] / (f_1, \ldots, f_s)\) for suitable \(r\) and relations \(f_i\).

Analogy

One can think of \(R\) as the coordinate ring of a moduli variety: points of \(\text{Spec}(R)\) (more precisely: \(\mathbb{Z}_p\)-valued points) correspond to deformations of \(\bar{\rho}\). The geometry of \(\text{Spec}(R)\) reflects the structure of the space of all deformations.


4. Deformation Conditions

Why conditions are necessary

The "bare" universal deformation ring \(R\) parametrises all deformations of \(\bar{\rho}\) – without any restriction. For Wiles' proof, this is too much: one needs deformations satisfying additional local conditions.

Local conditions at \(q \neq p\)

For every prime \(q \neq p\), one can require the deformation at \(q\) to have a particular form. The most important conditions:

  • Unramified: The inertia group \(I_q\) acts trivially. This is imposed at places of good reduction.
  • Steinberg: The representation at \(q\) has a special form corresponding to multiplicative reduction.
  • Minimal condition: The representation at \(q\) has the same type as \(\bar{\rho}\) – no additional ramification allowed.

Local conditions at \(p\)

At the prime \(p\) itself, there are particularly important conditions:

  • Flat: The representation comes from a flat group scheme over \(\mathbb{Z}_p\). This is the strongest condition and corresponds to good reduction.
  • Ordinary: The representation at \(p\) has an upper triangular form with unramified quotient.
  • Semistable: A generalisation that allows multiplicative reduction.

The restricted deformation ring \(R_{\mathcal{D}}\)

Combining a set \(\mathcal{D}\) of local conditions, one obtains a quotient of the universal deformation ring:

\[ R \twoheadrightarrow R_{\mathcal{D}}, \]

parametrising only those deformations that satisfy the conditions \(\mathcal{D}\). In what follows, we simply write \(R\) for the restricted ring \(R_{\mathcal{D}}\).


5. The Hecke Ring \(T\)

Modular deformations

Among all deformations of \(\bar{\rho}\), there is a special subset: the modular deformations – those coming from newforms.

For every newform \(f\) of weight 2 and level \(N\), there exists (by Eichler–Shimura) a Galois representation \(\rho_f: G_{\mathbb{Q}} \to \text{GL}_2(\mathcal{O}_f)\), where \(\mathcal{O}_f\) is the coefficient ring of \(f\). If \(\bar{\rho}_f \cong \bar{\rho}\), then \(\rho_f\) is a deformation of \(\bar{\rho}\).

The Hecke algebra

The Hecke algebra \(\mathbb{T}\) is generated by the Hecke operators \(T_q\) (for primes \(q \nmid N\)) and \(U_q\) (for \(q \mid N\)), acting on the space of cusp forms \(S_2(\Gamma_0(N))\).

The localised Hecke ring \(T\) is the quotient of \(\mathbb{T}\) parametrising the modular deformations of \(\bar{\rho}\):

\[ T = \mathbb{T}_{\mathfrak{m}}, \]

localised at the maximal ideal \(\mathfrak{m}\) determined by \(\bar{\rho}\) (concretely: \(T_q - \text{tr}(\bar{\rho}(\text{Frob}_q)) \in \mathfrak{m}\) for all \(q\)).

The modular deformation

The Hecke algebra \(T\) carries a universal modular deformation:

\[ \rho^{\text{mod}}: G_{\mathbb{Q}} \to \text{GL}_2(T), \]

capturing all modular deformations simultaneously.


6. The Natural Surjection \(R \twoheadrightarrow T\)

Why a surjection exists

Since every modular deformation is in particular a deformation, there exists (by the universal property of \(R\)) a natural ring homomorphism:

\[ \varphi: R \twoheadrightarrow T. \]

This homomorphism is surjective: the Hecke algebra \(T\) is generated by the traces \(\text{tr}(\rho^{\text{mod}}(\text{Frob}_q))\), and these are images of the corresponding traces of the universal deformation.

What the surjection means

\[ R \twoheadrightarrow T \]

means: \(T\) is a quotient of \(R\). Or geometrically: the "modular points" form a closed subset of the deformation space.

The decisive question is: is \(\varphi\) an isomorphism? That is: \(R = T\)?


7. Wiles' Goal: \(R = T\)

What \(R = T\) means

If \(R \cong T\) (as rings), then every admissible deformation of \(\bar{\rho}\) is automatically modular:

\[ R = T \implies \text{every deformation of } \bar{\rho} \text{ with the given conditions is modular.} \]

In particular: the representation \(\rho_{E,p}\) of the elliptic curve \(E\) is a deformation of \(\bar{\rho}\) with the right local conditions (because \(E\) is semistable). If \(R = T\), then \(\rho_{E,p}\) is modular – and hence \(E\) is modular.

The proof structure

\[ \boxed{\bar{\rho} \text{ modular}} \xrightarrow{R = T} \boxed{\rho_{E,p} \text{ modular}} \implies \boxed{E \text{ modular}} \implies \boxed{\text{FLT}} \]

Why \(R = T\) is hard

The surjection \(R \twoheadrightarrow T\) is "for free" – it follows from the universal property. The injectivity is the hard part: one must show that the kernel is trivial, i.e., that there are no non-modular deformations.

Wiles' great breakthrough was the development of a numerical criterion – a purely algebraic condition implying \(R = T\). This criterion and its proof are the subject of the next article.

Overview of the proof machinery

Object Description Parametrises
\(\bar{\rho}\) Residual representation Starting point
\(R\) Universal deformation ring All admissible deformations
\(T\) Hecke algebra Modular deformations
\(R \twoheadrightarrow T\) Natural surjection Modular ⊂ All
\(R = T\) Isomorphism All = Modular

Outlook

Deformation theory provides the conceptual framework for Wiles' proof. But the heart of the matter is the proof of \(R = T\) – a deep algebraic statement developed in the next article:

Article Topic
05 – R = T The numerical criterion, Selmer groups, and the proof
06 – The Taylor–Wiles Trick The patching argument that closed the gap

Sources

  • Andrew Wiles: Modular elliptic curves and Fermat's Last Theorem, Annals of Mathematics 141 (1995), §1.2–1.6
  • Barry Mazur: Deforming Galois representations, in: Galois Groups over \(\mathbb{Q}\), MSRI Publications 16 (1989) – The foundation of deformation theory
  • Nigel Boston: The Proof of Fermat's Last Theorem (2003), Chapter 11 – Deformation rings and Hecke algebras
  • Gebhard Böckle: Deformations of Galois representations, in: Clay Mathematics Proceedings 4 (2005) – Modern exposition