Paper XIIReading edition

Compatible systems and companion completion

Reading tools
Reading libraryFind statementsInspect Lean
Switch work

Saved reading

Read side by side

Start reading

Resources & source

Original TeX with line links · Download TeX

The wording and mathematics follow the retained manuscript. This reading format does not add a proof review or extend Lean coverage.

Skip contents and start reading

10 Compatible systems and companion completion

Pointwise automorphy at one coefficient place and existence of companions at other coefficient places are different statements. They meet only after a common algebraic Frobenius spine has been fixed.

Definition 10.1: Algebraic Frobenius spine

An algebraic Frobenius spine consists of a number field \(L\), a finite set of places \(S\) of \(F\), and polynomials

\[ P_v(X)\in L[X] \]

for every finite place \(v\notin S\), together with any additional local labels required by the declared compatibility problem.

Definition 10.2: Finite companion forest

Fix a coefficient place \(\lambda\) of \(L\). A finite companion forest for the spine is an inverse system of finite sets

\[ X_{\lambda,1} \longleftarrow X_{\lambda,2} \longleftarrow X_{\lambda,3} \longleftarrow\cdots \]

whose points are admissible realizations modulo \(\lambda^m\), with edges given by reduction. A complete realization is admissible for this forest when all its finite reductions belong to the displayed levels; the empty level obstruction below concerns this declared admissibility problem.

Theorem 10.3: Companion branch criterion

Suppose every level \(X_{\lambda,m}\) is finite and nonempty, and every point reduces along the displayed forest to the fixed residual root. Then

\[ \boxed{ \varprojlim_mX_{\lambda,m} \neq\varnothing. } \]

Proof of Theorem 10.3

The disjoint union of the levels is an infinite rooted tree. It is locally finite because every level is finite. It contains a vertex at every depth. König's infinity lemma gives an infinite branch [10]. That branch is a point of the inverse limit.

Corollary 10.4: Empty finite companion obstruction

If \(X_{\lambda,m}=\varnothing\) for some \(m\), then no complete \(\lambda\) adic realization admissible for this forest exists.

Proof of Corollary 10.4

A complete realization would reduce to a point at every finite level.

Finite forests alone give compatible classes. To obtain actual continuous representations, we use the following explicit coefficient and admissibility data. They also delimit the meaning of completion in the next theorem.

Fix a topological group \(\Gamma\), a finite matrix index set \(D\), and a set \(\Lambda\) of coefficient indices. For each \(\lambda\), let \({\mathcal O}_\lambda\) be a commutative ring, complete and separated for a finitely generated ideal \(I_\lambda\), with finite residue ring \({\mathcal O}_\lambda/I_\lambda\). Give \({\mathcal O}_\lambda\) its ideal adic topology and every quotient its discrete topology. Let

\[ i_\lambda:{\mathcal O}_\lambda\longrightarrow K_\lambda, \qquad e_\lambda:L\longrightarrow K_\lambda \]

be supplied ring homomorphisms, where \(L\) and \(K_\lambda\) are fields, \(K_\lambda\) has a declared topology, and \(i_\lambda\) is continuous.

For every positive integer \(n\), let \(X_{\lambda,n}\) be a finite nonempty set of conjugacy classes of actual continuous representations

\[ \Gamma\longrightarrow{\operatorname{GL}}_D({\mathcal O}_\lambda/I_\lambda^n). \]

Conjugacy means change of basis by an invertible matrix over that quotient. Assume reduction preserves these sets and that \(X_{\lambda,1}\) consists of the declared residual class. Each admissible class must have a continuous representative. The auxiliary level zero is the unique class over the trivial quotient and imposes no extra occupancy hypothesis.

Choose elements \(\gamma_v\in\Gamma\) on a sample set \(\Omega\), with declared good subsets \(\Omega_\lambda\subseteq\Omega\). In the arithmetic application these are the chosen Frobenius elements at the permitted places. For each \(v\in\Omega_\lambda\), supply an integral polynomial \(P_{\lambda,v}\in{\mathcal O}_\lambda[X]\) such that

\[ i_\lambda(P_{\lambda,v})=e_\lambda(P_v)\quad\text{in }K_\lambda[X]. \]

Here and below a ring homomorphism applied to a polynomial acts on its coefficients. Assume every representative of every class in \(X_{\lambda,n}\) has characteristic polynomial \(P_{\lambda,v}\bmod I_\lambda^n\) at \(\gamma_v\). Also supply a subgroup \(H_\lambda\leq\Gamma\) on which all these finite representations are trivial. This is an inertia condition when \(H_\lambda\) is the specified inertia subgroup.

Finally, put

\[ B_\lambda={\mathcal O}_\lambda[x_{g,a,b}:g\in\Gamma,\ a,b\in D]. \]

For each declared local condition index \(u\), supply an ideal \(J_{\lambda,u}\subseteq B_\lambda\). Evaluation at a representation sends \(x_{g,a,b}\) to its \((a,b)\) matrix entry at \(g\). Assume that evaluation at every admissible finite representative kills \(J_{\lambda,u}\), with polynomial coefficients reduced modulo \(I_\lambda^n\). For this class based theorem, the equations must hold in every admissible frame. A condition imposed only in one chosen frame needs a separate compatible frame argument. No arbitrary arithmetic condition is identified with such an ideal without a separate argument.

Theorem 10.5: Conditional simultaneous companion completion

Under the coefficient and finite admissibility hypotheses above, one can choose continuous representations

\[ \rho_\lambda:\Gamma\longrightarrow{\operatorname{GL}}_D({\mathcal O}_\lambda) \qquad(\lambda\in\Lambda) \]

whose reduction class belongs to \(X_{\lambda,n}\) for every positive \(n\). They retain the declared residual class. After applying \(i_\lambda\), they are continuous, trivial on \(H_\lambda\), and satisfy

\[ \det\!\left(X- i_\lambda(\rho_\lambda(\gamma_v))\right) =e_\lambda(P_v),\qquad v\in\Omega_\lambda. \]

Their integral evaluation points factor through every actual quotient \(B_\lambda/J_{\lambda,u}\).

Proof of Theorem 10.5

The finitely generated ideal and finite residue ring give finite coefficient quotients at every precision. An individual conjugacy class has finitely many frames because the group of invertible basis matrices over each quotient is finite. Thus the finitely many admissible classes give a finite nonempty set of actual continuous representations at each positive level. Continuity is invariant under conjugation. Reduction gives total maps between these sets. Compactness, as in Theorem 10.3, selects strictly compatible representatives. This step neither assumes transition surjectivity nor lifts a prescribed conjugating matrix.

Completeness and separatedness identify \({\operatorname{GL}}_D({\mathcal O}_\lambda)\) with the compatible inverse limit of the quotient matrix groups. Lift entries and inverse entries to construct this identification. The two reduction and reconstruction identities hold entrywise. Finite quotient continuity implies continuity for the ideal adic topology, including the matrix inverse, so the reconstructed representation is continuous.

Each coefficient of its characteristic polynomial agrees with the prescribed integral coefficient modulo every positive power of \(I_\lambda\). Separatedness gives equality of the full polynomials. The same argument applied to the matrix entries of \(\rho_\lambda(g)-1\) gives triviality on \(H_\lambda\). Continuous coefficient change and the matching equations for \(P_{\lambda,v}\) give the asserted common spine over the local fields.

For \(p\in J_{\lambda,u}\), evaluation commutes with reduction. Its value is zero modulo every positive power of \(I_\lambda\), hence is zero integrally by separatedness. Therefore the evaluation homomorphism kills the whole ideal and factors uniquely through \(B_\lambda/J_{\lambda,u}\). Choose a realization for each \(\lambda\).

The quotient conclusion is a proved effectivity statement for the declared polynomial relations. For local conditions presented by continuous maps from represented local rings, Corollary 3.2 also accounts for the closed ideal generated by their defining ideals. Neither statement constructs a deformation ring for a condition that has not been shown representable. Hodge conditions, Weil–Deligne conditions, integral lifts of arbitrary Frobenius polynomials, and the finite occupancy hypotheses still require their own arithmetic inputs. The theorem retains a residual conjugacy class, not a prescribed residual matrix frame, and asserts no uniqueness up to integral conjugacy.

Standard fact 10.6: Strict compatibility and one member determination

A strictly compatible semisimple system is determined at every coefficient place by its common unramified characteristic polynomials. If an automorphic representation gives a compatible system with the same polynomials, then Chebotarev density and Brauer Nesbitt identify the two systems member by member after semisimplification [23], [3].

Theorem 10.7: One member propagation in a fixed system

Let

\[ \{\rho_\lambda\}_{\lambda\in\Lambda} \]

be a semisimple strictly compatible system with Frobenius spine \((P_v)\). Suppose that for one coefficient place \(\lambda_0\), the representation \(\rho_{\lambda_0}\) is identified with the Galois representation attached to an automorphic representation \(\pi\). Assume that the compatible system attached to \(\pi\) has the same polynomials \(P_v(X)\). Then for every \(\lambda\in\Lambda\),

\[ \boxed{ \rho_\lambda \cong \rho_{\pi,\lambda}^{\mathrm{ss}}. } \]

Proof of Theorem 10.7

For every \(\lambda\), the two semisimple representations have the same unramified characteristic polynomials at all places outside one finite set. Apply Standard fact 9.2.

Remark 10.8

The theorem propagates automorphy inside a compatible system that already exists. It does not construct a missing coefficient place member and does not prove universal companion existence.

Proposition 10.9: Finite Frobenius separator

Let \(t_1,\ldots,t_r\) be linearly independent functions on a set \(\Omega\) with values in a field \(E'\). Then there are points \(\omega_1,\ldots,\omega_r\in\Omega\) such that the matrix

\[ \left(t_i(\omega_j)\right)_{j,i} \]

is invertible. In particular, when the trace functions of finitely many compatible constituents are linearly independent on the unramified Frobenius set, finitely many Frobenius traces recover their multiplicities in any known semisimple carrier.

Proof of Proposition 10.9

Let

\[ W = \operatorname{span} \left\{ (t_1(\omega),\ldots,t_r(\omega)):\omega\in\Omega \right\} \subseteq E'^r. \]

If \(W\neq E'^r\), a nonzero linear functional on \(E'^r\) would vanish on every evaluation vector, giving a nontrivial linear relation among the functions. Thus \(W=E'^r\). Choose \(r\) evaluation vectors forming a basis.

Under the coefficient and effectivity hypotheses of cref thm:simultaneous companion completion , local companion branches realize the Frobenius spine fixed in advance.
Under the coefficient and effectivity hypotheses of Theorem 10.5, local companion branches realize the Frobenius spine fixed in advance.

View full-size figure (new tab)

Source and conversion record

Source SHA-256: f5137c0fffed58acc1590165c77a75848e4d97671dca9caafb7c147e6aa1112a

Unsupported prose commands stop the build. The mathematical expression sequence is checked against the source, with manuscript macros expanded for display.

Conversion records

Notation in Paper XII

Look up a symbol without leaving this chapter. Select text before opening help to search for it.

Search this work’s notation.

    Symbols can have different meanings in different passages. Read each result in its own context.

    Open the full notation index