Automorphic representation theory reorganises highly symmetric functions into irreducible representation spaces. Its local pieces carry Hecke and harmonic-analysis data; its global objects produce L-functions. The Langlands program asks how this analytic representation world corresponds to arithmetic symmetry.
The subject is large enough that no single article can replace a graduate course. The purpose here is architectural and computational: show why adeles are useful, how local components arise, what “unramified” and “spherical” mean, how Hecke eigenvalues become Satake parameters, how Euler products are assembled, and where Galois representations enter the picture.
This guide continues directly from Hecke Algebras, Double Cosets and Modular Forms and Galois Representations. The first supplies arithmetic eigenvalues; the second supplies Frobenius matrices. Automorphic representation theory builds the analytic side on which those local data can be compared.
Automorphic forms · Adeles · Local components · Satake parameters · L-functions · Langlands architecture · Practice
Begin with harmonic analysis on a quotient
Classical Fourier analysis studies functions on a group such as the circle and decomposes them into irreducible frequency modes. Automorphic harmonic analysis studies functions on quotients built from reductive groups and arithmetic subgroups.
The basic representation-theoretic question is familiar: which irreducible representations occur when a group acts on a function space? What changes is the group, the quotient and the measure-theoretic setting.
The regular representation may now be infinite-dimensional. Discrete and continuous spectral pieces can coexist. Cusp forms isolate an important discrete part. Hecke operators commute with the group action in suitable settings and supply additional arithmetic labels.
From classical modular forms to automorphic forms
A classical modular form f(z) is a holomorphic function on the upper half-plane satisfying a transformation law under a congruence subgroup of SL₂(Z), together with growth conditions at the cusps. This is already an automorphy condition.
The automorphic perspective asks us to stop treating the complex variable z as the entire object. Instead, encode the symmetry using functions on a group, typically GL₂ or a related reductive group, modulo rational points and compact subgroups.
This reformulation has a major advantage: the same object can be analysed place by place. The real or complex place produces the archimedean representation. Every prime p produces a p-adic local representation. The global automorphic representation packages all of these local behaviours together.
Adeles put every completion into one object
For Q, the adele ring A is a restricted product of R and all p-adic fields Qp. An adele is a tuple
(x∞,x₂,x₃,x₅,…)
with x∞∈R and xp∈Qp, subject to the restriction that xp lies in Zp for all but finitely many p.
The restriction is essential. Without it, multiplication, topology and Haar measure would not assemble in the intended arithmetic way. The adele ring contains Q diagonally: a rational number is viewed simultaneously inside every completion.
For an algebraic group G over Q, the adelic group G(A) combines the real and p-adic groups G(R),G(Q₂),G(Q₃),… in one global object.
global arithmetic object → all local completions at once → one adelic representation.
Why Q-points are quotiented out
The diagonal subgroup G(Q) records the global rational symmetry shared by all places. Automorphic forms are functions on, or related to, the quotient G(Q)\G(A), often with central-character, smoothness, finiteness and growth conditions depending on the precise theory.
The quotient means two adelic points differing by one rational group element represent the same global arithmetic configuration. Local components remain visible, but global rational redundancy has been removed.
Do not treat G(Q)\G(A) as an ordinary finite quotient. It is a topological and measure-theoretic space. Harmonic analysis here requires Haar measure, unitary representations and functional analysis in addition to algebra.
Automorphic representations arise from invariant subspaces
G(A) acts by right translation on suitable functions on G(Q)\G(A). An automorphic representation is, broadly, an irreducible representation occurring in an appropriate automorphic function space.
One should not identify one automorphic form with one automorphic representation. A single form can generate a representation under translations. Conversely, a representation may contain many vectors corresponding to different normalisations, levels or local test choices.
This is the same distinction found earlier between a vector and the invariant subspace it generates.
Local components form a restricted tensor product
An irreducible automorphic representation π of G(A) is typically described as a restricted tensor product
π≈⊗′vπv,
where v runs over all places of the number field. For Q these are ∞ and the prime numbers. The factor πv is a representation of G(Qv).
The prime on the tensor product records a restriction: at almost every finite place, πp is unramified and contains a distinguished vector fixed by a maximal compact subgroup such as GL₂(Zp) in the GL₂ setting.
This is the automorphic analogue of the fact that arithmetic objects have good reduction or unramified Galois behaviour at all but finitely many primes.
Spherical vectors identify the unramified places
Let G(Qp) have a chosen maximal compact subgroup Kp. A local representation πp is spherical, in the standard unramified setting, when it has a nonzero Kp-fixed vector.
The spherical Hecke algebra consists of compactly supported Kp-bi-invariant functions under convolution. It acts on the Kp-fixed vectors. For an irreducible unramified representation, this fixed-vector space is typically one-dimensional in the standard reductive setting.
Therefore every spherical Hecke operator acts on the distinguished line by a scalar. The local representation has again been compressed to a Hecke eigenvalue system.
Satake parameters are local spectral coordinates
For an unramified representation of GLn(Qp), the spherical Hecke eigenvalues can be encoded by an unordered n-tuple of nonzero complex parameters αp,1,…,αp,n, up to the relevant normalisation.
These are Satake parameters. They play a role analogous to eigenvalues of a Frobenius matrix. For GL₂ one often writes αp,βp.
For a classical normalised Hecke eigenform, the local quadratic polynomial can be written
1−apT+ε(p)pk−1T²=(1−αpT)(1−βpT).
Thus αp+βp=ap and αpβp=ε(p)pk−1 in the classical normalisation. Representation theorists often rescale parameters to unitary normalisation. Again, the normalisation must travel with the formula.
Worked local example from Ramanujan Delta
For Δ of weight 12 and level 1, a₂=τ(2)=−24. The local polynomial at p=2 is
1+24T+2¹¹T²=1+24T+2048T².
The two Satake roots satisfy α₂+β₂=−24 and α₂β₂=2048 in this arithmetic normalisation.
The previous Hecke article used the same polynomial to produce prime-power coefficient recurrences. The automorphic viewpoint interprets the two roots as local representation parameters. The Galois viewpoint interprets the corresponding polynomial as a Frobenius characteristic polynomial after choosing an ℓ-adic realisation.
The same local polynomial can support several languages
- Hecke language: ap is the Tp eigenvalue.
- Automorphic language: αp,βp encode the unramified local representation.
- Galois language: ap and ε(p)pk−1 are Frobenius trace and determinant.
- L-function language: the reciprocal polynomial is the local Euler factor.
The languages are not automatically equivalent in arbitrary settings. Theorems and conjectures are required to establish correspondences. But when a correspondence is known, matching local polynomials is one of the strongest visible signs that the two constructions are describing the same arithmetic structure.
Automorphic L-functions are built locally
For an unramified local GLn representation with Satake parameters αp,1,…,αp,n, the standard local factor has the form
Lp(s,π)=∏i=1n(1−αp,ip−s)−1,
subject to the normalisation being used. At ramified places, local factors require a more careful definition.
The global L-function is assembled from local factors. Analytic continuation and functional equations are central questions. Langlands emphasized that standard automorphic L-functions form a broad class with structured analytic behaviour, while functoriality predicts relationships among many further L-functions. His writings listed below give the primary conceptual source for this viewpoint. [1–3]
GL1 recovers classical abelian harmonic analysis
For G=GL₁, the group is multiplicative. Automorphic representations are one-dimensional characters of the idele class group under suitable conditions. These are Hecke characters.
This case includes Dirichlet characters after translating between classical and adelic language. The associated L-functions generalise the Riemann zeta function and Dirichlet L-functions.
GL1 is an important check because local-global representation language reduces to familiar scalar characters. The non-abelian difficulty begins when local components have dimension greater than one in the dual-group parameter space and the group action no longer commutes.
GL2 contains classical modular forms
A normalised cuspidal Hecke eigenform can be translated into an adelic cusp form generating an automorphic representation π of GL₂(A). Its archimedean component records weight information. Its finite local components record level and Hecke data.
At almost every prime p, the local component is unramified and is determined by the Hecke/Satake parameters. At primes dividing the level, the local component is ramified and contains more delicate conductor information.
Thus the classical q-expansion is not discarded. It becomes one coordinate system on a global representation carrying much richer local structure.
Cuspidal representations isolate the discrete arithmetic spectrum
Automorphic forms include Eisenstein series and other non-cuspidal contributions. Cuspidality imposes vanishing of certain integrals along unipotent radicals and removes continuous-spectrum contributions generated by parabolic induction in the simplest conceptual picture.
Cuspidal automorphic representations therefore form an especially important class of irreducible constituents of the discrete automorphic spectrum.
The phrase “automorphic representation” is broader than “cuspidal automorphic representation.” Keeping the adjective matters, particularly when discussing L-functions and correspondence statements.
The trace formula is a non-abelian spectral accounting system
The Selberg and Arthur trace formulas compare two ways of taking a trace of suitable operators on automorphic function spaces.
- The spectral side organises irreducible representations and their multiplicities.
- The geometric side organises conjugacy classes, orbital integrals and related geometric data.
This is a vast generalisation of the representation-theoretic habit seen earlier: one operator can be understood either from its action on spectral components or from its kernel and symmetry classes.
The full Arthur trace formula is advanced and technically demanding. The structural point here is that representation theory creates a language in which global harmonic analysis can be compared across groups.
The Langlands program is a correspondence architecture, not one theorem
The phrase “Langlands program” covers a network of conjectures, constructions and theorems linking automorphic representations, Galois representations or related arithmetic parameters, L-functions and harmonic analysis on reductive groups.
There is no universal theorem saying that every automorphic representation of every group is already known to correspond to a Galois representation. Results depend on the group, field, regularity, algebraicity, local conditions and many other hypotheses.
The architecture is nevertheless clear enough to state:
arithmetic parameters ↔ automorphic representations, with matching local data and L-functions.
Langlands’s original problems and later expositions emphasise functoriality and L-functions as central organising ideas. [1–3]
Local Langlands first: classify local representations by local parameters
At a local field F such as Qp, local Langlands seeks a correspondence between irreducible admissible representations of a reductive group and suitable parameters built from the local Galois/Weil side into the Langlands dual group.
For GLn over p-adic fields, the local Langlands correspondence is a theorem. Its formulation includes precise compatibility with L-functions, epsilon factors and other local invariants.
For more general groups the parameterisation naturally involves packets rather than a simple one-representation-to-one-parameter rule. This is one reason the dual group and L-group are necessary.
The Langlands dual group reverses the root data
A reductive group G has a dual group Ĝ whose root datum is obtained by exchanging roots with coroots. The dual group is not usually the same concrete group acting on the original automorphic forms.
For GLn, the dual group is again GLn(C), making the notation deceptively simple. For other classical groups, the dual can change type.
Local Satake parameters naturally live as semisimple conjugacy classes in this dual group. This explains why Hecke eigenvalues can be treated as coordinates on a dual-group conjugacy class rather than merely a list of unrelated scalars.
Functoriality moves representations between groups
Suppose there is an appropriate homomorphism between L-groups. Langlands functoriality predicts a transfer of automorphic representations from one group to another compatible with the map on local parameters.
At an unramified place, the prediction is especially concrete: apply the dual-group map to the Satake conjugacy class and read off the target local parameters.
Examples and special cases include base change, automorphic induction, symmetric-power lifts and endoscopic transfers, but the general functoriality principle remains far broader than the currently proved collection of cases.
A tiny symmetric-square parameter calculation
Suppose an unramified GL₂ local parameter has eigenvalues α and β. Apply the symmetric-square representation of GL₂(C). On Sym²(C²), the induced eigenvalues are
α², αβ, β².
The corresponding degree-3 local factor is therefore formally
[(1−α²T)(1−αβT)(1−β²T)]−1.
This calculation illustrates functoriality at the parameter level without pretending to prove global automorphy of every symmetric-power lift. The local linear-algebra transformation is elementary; the global automorphic theorem is much deeper.
Matching L-functions is a powerful checksum
If two constructions are predicted to correspond, compare their unramified local characteristic polynomials first. A mismatch at one good prime disproves a claimed exact correspondence in that normalisation.
Agreement at finitely many primes is encouraging but not conclusive. Global equality may require strong multiplicity-one theorems, density arguments, analytic continuation or arithmetic uniqueness results.
This hierarchy prevents an important reasoning error: computational evidence and structural theorem are different levels of support.
Strong multiplicity one shows how local data can control a global GL_n representation
For cuspidal automorphic representations of GLn, strong multiplicity-one theorems say, in suitable formulations, that equality of local components outside a sufficiently small exceptional set can determine the global representation.
This is the automorphic analogue of reconstructing a global object from dense local information. It explains why Hecke eigenvalues at almost all primes carry extraordinary classification power.
It does not say that the first few coefficients always determine a form. The theorem’s hypotheses and exceptional-set conditions are part of the conclusion.
A local-to-global verification workflow
- Specify the global group G and number field.
- Identify the automorphic space: cuspidal, discrete, Eisenstein or another setting.
- Record the archimedean component and every ramified finite place.
- At an unramified prime, compute Hecke eigenvalues and Satake parameters in one fixed normalisation.
- Build the local Euler factor from those parameters.
- If comparing to a Galois representation, match trace, determinant and characteristic polynomial before making a global claim.
- Separate a conjectural functorial transfer from a proved case.
- Use global theorems only with their stated hypotheses.
Common mistakes
- Equating a form with a representation: a form is usually a vector generating or lying inside a representation.
- Ignoring ramified primes: spherical Satake parameters describe unramified places.
- Mixing arithmetic and unitary normalisations: local parameters can be rescaled by powers of p.
- Calling the Langlands program fully proved: it contains many theorems and many open conjectures.
- Assuming every group is self-dual: the Langlands dual group is defined from dual root data.
- Inferring a global theorem from one local calculation: parameter arithmetic is not the same as proving automorphy.
Practice with worked answers
1. Places of Q
Which places occur in the adelic description of Q? Answer: the real place ∞ and one p-adic place for every prime p.
2. Restricted product
Why is the finite adelic product restricted? Answer: almost all p-components must lie in the standard compact subrings/subgroups, allowing a workable topology, measure and tensor-product structure.
3. Spherical vector
What does a Kp-fixed vector satisfy? Answer: πp(k)v=v for every k∈Kp.
4. GL2 local polynomial
If α+β=7 and αβ=10, write the local polynomial. Answer: 1−7T+10T²=(1−αT)(1−βT).
5. Standard local factor
For Satake parameters 2 and 3, what is the formal GL2 local factor? Answer: [(1−2p−s)(1−3p−s)]−1, assuming these numbers are already in the chosen normalisation.
6. Symmetric square
If α=2 and β=5, list the symmetric-square local parameters. Answer: 4,10,25.
7. Hecke bridge
What automorphic quantity equals ap for a classical normalised unramified eigenform? Answer: the appropriate normalised Tp Hecke eigenvalue, equivalently the sum of the two local parameters in the classical GL2 normalisation.
8. Galois bridge
Under a known modular-form/Galois correspondence, what does ap become on the arithmetic side? Answer: the Frobenius trace at an unramified prime.
9. Dual-group caution
Does a Satake parameter generally live as a conjugacy class in the original real or p-adic group? Answer: the standard formulation places it in the complex Langlands dual group, after choosing the relevant normalisation.
10. Proof versus prediction
You transform a local GL2 parameter by Sym² and obtain degree-3 parameters. Have you proved a global symmetric-square automorphic lift? Answer: no. You have computed the locally predicted target parameters; a global theorem requires substantially more.
The representation-theoretic lesson
Automorphic representation theory is a large-scale version of the same pattern seen throughout this BTT series. Choose the correct carrier space. Decompose by symmetry. Replace difficult operators by scalar or matrix parameters on irreducible components. Preserve local information. Reassemble globally.
The Langlands program adds an extraordinary further claim: different mathematical worlds may be organised by compatible representation parameters and L-functions.
global quotient → automorphic spectrum → local representations → Satake parameters → L-functions → arithmetic correspondence.
Mathematical references
[1] Robert P. Langlands, L-Functions and Automorphic Representations, for the motivic/automorphic L-function perspective. [2] Robert P. Langlands, Problems in the Theory of Automorphic Forms, for the early functoriality architecture. [3] Institute for Advanced Study, Automorphic Forms: Concepts, Techniques, Applications and Influence, including Langlands’s introductory remarks and lectures on L-functions, spectral theory and functoriality. The local parameter and Delta examples above are worked from the explicit Hecke polynomial already developed in this BTT series.
Continue Representation Mathematics — Batch 04
Build the arithmetic side in Galois Representations. Recover the eigenvalue algebra in Hecke Algebras and Modular Forms. Then ask a different foundational question—whether the symmetry group itself can be reconstructed from its tensor category of representations—in Tannakian Categories and Fibre Functors. Return to the BTT Mathematics Learning Hub.