Local Galois Representations

GaloisRepresentation(pi): RepLoc -> GalRep
Precision: RngIntElt                    Default: 12
WeilRepresentation(pi): RepLoc -> GalRep
Precision: RngIntElt                    Default: 12

Given a minimal representation \(\pi\) of \({\GL}_2({\Q}_p),\) this returns a representation of the Weil-Deligne group, as a local Galois representation, that agrees on inertia with the representation \(\rho_\pi\) associated to \(\pi\) under the local Langlands correspondence. (See Section The Local Langlands Correspondence and Chapter galrep.) Only for supercuspidal \(\pi\) are the Frobenius data of \(\rho_\pi\) recovered as well, as described below; for a principal series or Steinberg \(\pi\) the result can differ from \(\rho_\pi\) by an unramified twist of each constituent.

The parameter Precision is the \(p\)-adic precision of the field the representation is constructed over. It governs the supercuspidal case alone, where a request below the default is raised to it, or below \(40\) to \(40\) for an exceptional supercuspidal at \(p=2\), which has no admissible pair; a principal series is assembled from Artin representations and ignores it. A computation on the result may need more precision than constructing it does, and how much is not predictable from the conductor.

For supercuspidal \(\pi\) with admissible pair \((E,\chi)\) the Frobenius data of \(\rho_\pi\) are recovered as follows. When \(E/{\Q}_p\) is unramified the Loeffler-Weinstein rectifier is applied. When \(E/{\Q}_p\) is ramified the value of \(\chi\) at a uniformizer of \(E\) is obtained from the Atkin-Lehner eigenvalue together with the tame and wild \(\eps\)-factors, giving the full \(\rho_\pi\) when \(\pi\) arises from a newform of trivial character, \(p\) is odd, the conductor of \(\pi\) is \(p^3\) and those \(\eps\)-factors are implemented. In the remaining ramified cases an error is raised.

When \(E/{\Q}_p\) is ramified the representation agrees with the one this function returns for an elliptic curve only after a Tate twist. Pairing a weight \(2\) newform \(f\) with the elliptic curve \(C\) of the same \(L\)-series – level \(27\) with 27a1, the two newforms of level \(54\) with 54a1 and 54b1 – GaloisRepresentation(C,3) equals the Tate twist by \(-1\) of the representation built from LocalComponent(f,3) and not that representation itself. For these pairs the equality also fixes the action of Frobenius, so the two have the same EpsilonFactor and RootNumber. Other local components need not agree at any integer Tate twist: for those of 36a1 at \(3\) and 50a1 at \(5\), where \(E/{\Q}_p\) is unramified, no integer Tate twist of one has the EpsilonFactor of the other.

AdmissiblePair(pi): RepLoc -> RngPad, Map
Precision: RngIntElt                    Default: 12

Given an ordinary minimal supercuspidal representation \(\pi\) of \(\GL_2(\Q_p),\) this returns the associated admissible pair \((E,\chi).\) (See Section The Local Langlands Correspondence.) Two objects are returned: the ring of integers \(E\) of a quadratic field extension of \(\Q_p\), and a map \(\chi\) which is a character of the unit group of \(E.\) Note that \(E\) is the ring of integers and not the field itself; use FieldOfFractions(E) for the latter.

The parameter Precision is the \(p\)-adic precision of the field the pair is constructed over; a request below the default is raised to it. How much is needed depends on the input and is not predictable from its conductor, and a computation on the result may need more precision than constructing it does.