Canonical: https://knowlpedia.com/langlands/arthur-multiplicity-formula/

# Arthur multiplicity formula

The component-group character formula selecting representations and multiplicities in a global Arthur packet.

Let \(\psi\) be a discrete global [Arthur
parameter](https://knowlpedia.com/langlands/arthur-parameter/), let \(\mathcal S_\psi\) be its finite component group, and let
\(\epsilon_\psi:\mathcal S_\psi\to\{\pm1\}\) be Arthur's canonical global
character. For a member

\[
\pi=\bigotimes_v'\pi_v
\]

of the global [\(A\)-packet](https://knowlpedia.com/langlands/a-packet/), the local packet pairings
give a global character

\[
\langle s,\pi\rangle
=
\prod_v \langle s_v,\pi_v\rangle .
\]

The **Arthur multiplicity formula** selects the representations for which
this character matches \(\epsilon_\psi\). In a standard multiplicity-free
form,

\[
m_\psi(\pi)
=
\frac{1}{|\mathcal S_\psi|}
\sum_{s\in\mathcal S_\psi}
\epsilon_\psi(s)\,
\langle s,\pi\rangle .
\]

By [character orthogonality](https://knowlpedia.com/algebra-representation-theory/character-orthogonality/) this is \(1\) when the two characters agree and
\(0\) otherwise.

## Meaning

Local packets supply many possible
[restricted
tensor products](https://knowlpedia.com/langlands/restricted-tensor-product-automorphic-representation/). The formula
is a global reciprocity constraint: local component-group labels must
multiply to the distinguished global sign. It therefore explains why a
global packet is not simply the [Cartesian product](https://knowlpedia.com/shared-foundations/cartesian-product/) of its local packets.

## Source of the sign

The character \(\epsilon_\psi\) is built from
[global root numbers](https://knowlpedia.com/langlands/root-number/) of pairs of the
[cuspidal
general-linear-group constituents](https://knowlpedia.com/langlands/cuspidal-automorphic-representation/) of \(\psi\). It is not an
arbitrary choice and can force a representation out of the discrete
spectrum even when all of its local components lie in the prescribed
packets.

## Scope and variants

The displayed formula is the clean form for the
[quasi-split](https://knowlpedia.com/algebraic-geometry-foundations/quasi-split-reductive-group/)
classical
settings of Arthur's classification. [Central quotients](https://knowlpedia.com/lie-groups/central-quotient-of-a-lie-group/), even orthogonal
outer automorphisms, non-quasi-split
[inner forms](https://knowlpedia.com/langlands/rigid-inner-twist/), and parameters with
multiplicity require refinements of the packet and coefficient conventions.

The formula is a theorem for the symplectic and
[orthogonal groups](https://knowlpedia.com/lie-groups/orthogonal-group/) treated by Arthur and in
corresponding established extensions, such as quasi-split unitary groups. It
remains part of the conjectural general theory for
arbitrary [reductive groups](https://knowlpedia.com/algebraic-geometry-foundations/reductive-algebraic-group/).

## References

1. James Arthur, *The Endoscopic Classification of Representations:
   Orthogonal and Symplectic Groups*, Theorem 1.5.2 and Chapter 8, AMS,
   2013. [AMS](https://bookstore.ams.org/COLL/61).
2. Chung Pang Mok, “Endoscopic classification of representations of
   quasi-split unitary groups,” *Memoirs of the AMS* 235 (2015).
   [AMS](https://bookstore.ams.org/memo-235-1108).
