Frobenius reciprocity

Intuition [Beginner]

Frobenius reciprocity is one of the most influential identities in mathematics: it states that the operations of induction (lifting a subgroup-representation to a $G$-representation) and restriction (forgetting a $G$-representation down to an $H$-representation) are adjoint. Specifically, $G$-equivariant maps from an induced representation match $H$-equivariant maps to a restricted representation:

$${\mathrm{Hom}}_G({\mathrm{Ind}}_H^{G}W,V)\cong {\mathrm{Hom}}_H(W,{\mathrm{Res}}_H^{G}V).$$

The intuition: induction and restriction are paired, like adjoints in linear algebra. To compute "how many copies of an irreducible $V$ of $G$ appear in the induced ${\mathrm{Ind}}_H^{G}W$", you can equivalently compute "how many copies of $W$ appear inside the restriction of $V$ to $H$." This duality reduces a calculation on the larger group $G$ to a (typically simpler) calculation on the subgroup $H$.

Frobenius proved this for finite groups in 1898, giving in concrete form the inner-product identity $⟨{\chi }_{\mathrm{Ind}W},{\chi }_V{⟩}_G=⟨{\chi }_W,{\chi }_{\mathrm{Res}V}{⟩}_H$. He could not have known that he was discovering an instance of a vastly more general phenomenon: adjoint functors in category theory, formalised by Daniel Kan in 1958. Frobenius reciprocity is the prototype of all categorical adjunctions in mathematics, predating category theory by 60 years.

Visual [Beginner]

The Frobenius reciprocity diagram: a bijection between two Hom spaces, one over $G$ on the induced side, one over $H$ on the restricted side.

Worked example [Beginner]

Take $G=S_3$ and $H=\mathbb{Z}/3\mathbb{Z}=\{e,(123),(132)\}$. We saw in unit 07.01.07 that the induced representation ${\mathrm{Ind}}_H^{S_3}{\chi }_{\omega }$ from the 1-dimensional character ${\chi }_{\omega }$ on $H$ has character $(2,0,-1)$ on the conjugacy classes $\{e\},\{$trans$\},\{$3-cycle$\}$ of $S_3$.

To compute how many copies of the standard 2-dim representation $V_{\mathrm{std}}$ of $S_3$ appear in ${\mathrm{Ind}}_H^{S_3}{\chi }_{\omega }$, Frobenius reciprocity says: instead, compute how many copies of ${\chi }_{\omega }$ appear in ${\mathrm{Res}}_H^{S_3}{V}_{\mathrm{std}}$.

The restriction of $V_{\mathrm{std}}$ to $H$: looking at the character of $V_{\mathrm{std}}$ on $H$, it takes values ${\chi }_{V_{\mathrm{std}}}(e)=2$, ${\chi }_{V_{\mathrm{std}}}((123))=-1$, ${\chi }_{V_{\mathrm{std}}}((132))=-1$. The 1-dim characters of $H=\mathbb{Z}/3\mathbb{Z}$ are ${\chi }_0=(1,1,1)$, ${\chi }_{\omega }=(1,\omega ,{\omega }^2)$, ${\chi }_{{\omega }^2}=(1,{\omega }^2,\omega )$.

Inner product $⟨{\chi }_{V_{\mathrm{std}}}{|}_H,{\chi }_{\omega }{⟩}_H=\frac{1}{3}(2\cdot 1+(-1)\overline{\omega }+(-1)\overline{{\omega }^2})=\frac{1}{3}(2-{\omega }^2-\omega )=\frac{1}{3}(2-(-1))=1$.

What this tells us: ${\chi }_{\omega }$ appears once in ${\mathrm{Res}}_H{V}_{\mathrm{std}}$. By Frobenius reciprocity, $V_{\mathrm{std}}$ appears once in ${\mathrm{Ind}}_H^{S_3}{\chi }_{\omega }$. Computing both sides: ${\mathrm{Ind}}_H^{S_3}{\chi }_{\omega }$ has dimension 2 = $\dim V_{\mathrm{std}}$, so the induction is exactly $V_{\mathrm{std}}$, with the multiplicity 1 confirmed.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Let $G$ be a group, $H\le G$ a subgroup, $W$ a representation of $H$, and $V$ a representation of $G$, all over a field $k$. The restriction ${\mathrm{Res}}_H^{G}V$ is $V$ with the $G$-action restricted to $H$; the induction ${\mathrm{Ind}}_H^{G}W$ is the $G$-representation $kG{\otimes }_{kH}W$ (see unit 07.01.07).

Theorem (Frobenius reciprocity, Frobenius 1898). For any $H$-representation $W$ and $G$-representation $V$, there is a natural isomorphism of $k$-vector spaces

$${\mathrm{Hom}}_G({\mathrm{Ind}}_H^{G}W,V)\cong {\mathrm{Hom}}_H(W,{\mathrm{Res}}_H^{G}V).$$

This is the assertion that the induction functor ${\mathrm{Ind}}_H^G:\mathrm{Rep}(H)\to \mathrm{Rep}(G)$ is left adjoint to the restriction functor ${\mathrm{Res}}_H^G:\mathrm{Rep}(G)\to \mathrm{Rep}(H)$.

Explicit form of the isomorphism. A $G$-equivariant map $\Phi :{\mathrm{Ind}}_H^{G}W\to V$ is determined by its restriction to the "base copy" $1\otimes W\subset kG{\otimes }_{kH}W$, which gives a $k$-linear map $\phi :W\to V$. The $G$-equivariance of $\Phi$ together with the $H$-equivariance of the embedding $1\otimes \cdot$ forces $\phi$ to be $H$-equivariant: $\phi ({\rho }_W(h)w)={\rho }_V(h)\phi (w)$ for $h\in H$. Conversely, any $H$-equivariant $\phi :W\to {\mathrm{Res}}_H^{G}V$ extends uniquely to a $G$-equivariant $\Phi :{\mathrm{Ind}}_H^{G}W\to V$ via $\Phi (g\otimes w):={\rho }_V(g)\phi (w)$.

Theorem (Character form, Frobenius 1898). For finite $G$ and $H\le G$, irreducible characters $\psi$ of $H$ and $\chi$ of $G$,

$$⟨{\chi }_{{\mathrm{Ind}}_H^G\psi },\chi {⟩}_G=⟨\psi ,\chi {|}_H{⟩}_H.$$

Equivalently, the multiplicity of an irreducible $V$ of $G$ in ${\mathrm{Ind}}_H^{G}W$ equals the multiplicity of $W$ in ${\mathrm{Res}}_H^{G}V$ ^{[Serre §7.2]}.

Right adjoint. The restriction functor ${\mathrm{Res}}_H^G$ also has a right adjoint, the coinduced representation

$${\mathrm{Coind}}_H^{G}W:={\mathrm{Hom}}_{kH}(kG,W),$$

with $G$-action by right translation. For finite $G$, ${\mathrm{Ind}}_H^{G}W\cong {\mathrm{Coind}}_H^{G}W$ (the two adjoints agree); this isomorphism is the self-duality of finite-group induction. For infinite groups (or in the homological algebra of $kG$-modules), the two diverge in general.

Key theorem with proof [Intermediate+]

Theorem (Frobenius reciprocity). Let $G$ be a group, $H\le G$, $W$ an $H$-representation and $V$ a $G$-representation, all over a field $k$. Then there is a natural $k$-linear isomorphism

$${\mathrm{Hom}}_G({\mathrm{Ind}}_H^{G}W,V)\stackrel{\sim }{\to }{\mathrm{Hom}}_H(W,{\mathrm{Res}}_H^{G}V).$$

Proof. Define maps in both directions and check they are mutually inverse.

Forward map $\Phi :{\mathrm{Hom}}_G({\mathrm{Ind}}_H^{G}W,V)\to {\mathrm{Hom}}_H(W,V{|}_H)$. For $\Psi \in {\mathrm{Hom}}_G({\mathrm{Ind}}_H^{G}W,V)$, define $\Phi (\Psi ):W\to V$ by $\Phi (\Psi )(w):=\Psi (1\otimes w)$. We check $H$-equivariance: for $h\in H$,

$$\Phi (\Psi )({\rho }_W(h)w)=\Psi (1\otimes {\rho }_W(h)w)=\Psi (h\otimes w)={\rho }_V(h)\Psi (1\otimes w)={\rho }_V(h)\Phi (\Psi )(w),$$

using $1\otimes {\rho }_W(h)w=h\otimes w$ in ${\mathrm{Ind}}_H^{G}W=kG{\otimes }_{kH}W$ (since the tensor is over $kH$) and the $G$-equivariance of $\Psi$.

Backward map ${\Phi }^{\prime }:{\mathrm{Hom}}_H(W,V{|}_H)\to {\mathrm{Hom}}_G({\mathrm{Ind}}_H^{G}W,V)$. For $\phi \in {\mathrm{Hom}}_H(W,V{|}_H)$, define ${\Phi }^{\prime }(\phi ):kG{\otimes }_{kH}W\to V$ by ${\Phi }^{\prime }(\phi )(g\otimes w):={\rho }_V(g)\phi (w)$, extended $k$-linearly.

This is well-defined on the tensor product over $kH$: for $h\in H$,

$${\Phi }^{\prime }(\phi )((gh)\otimes w)={\rho }_V(gh)\phi (w)={\rho }_V(g){\rho }_V(h)\phi (w)={\rho }_V(g)\phi ({\rho }_W(h)w)={\Phi }^{\prime }(\phi )(g\otimes {\rho }_W(h)w),$$

using $H$-equivariance of $\phi$. So ${\Phi }^{\prime }$ respects the equivalence $gh\otimes w=g\otimes hw$ in the tensor product.

$G$-equivariance of ${\Phi }^{\prime }(\phi )$: for $g^{\prime }\in G$,

$${\Phi }^{\prime }(\phi )(g^{\prime }\cdot (g\otimes w))={\Phi }^{\prime }(\phi )((g^{\prime }g)\otimes w)={\rho }_V(g^{\prime }g)\phi (w)={\rho }_V(g^{\prime }){\rho }_V(g)\phi (w)={\rho }_V(g^{\prime }){\Phi }^{\prime }(\phi )(g\otimes w).$$

Inverse property. Compute $\Phi \circ {\Phi }^{\prime }$ and ${\Phi }^{\prime }\circ \Phi$:

$\Phi ({\Phi }^{\prime }(\phi ))(w)={\Phi }^{\prime }(\phi )(1\otimes w)={\rho }_V(1)\phi (w)=\phi (w)$, so $\Phi \circ {\Phi }^{\prime }=\mathrm{id}$.

${\Phi }^{\prime }(\Phi (\Psi ))(g\otimes w)={\rho }_V(g)\Phi (\Psi )(w)={\rho }_V(g)\Psi (1\otimes w)=\Psi (g\otimes w)$ (by $G$-equivariance of $\Psi$), so ${\Phi }^{\prime }\circ \Phi =\mathrm{id}$.

Both maps are $k$-linear by construction, and naturality (functoriality in $W$ and $V$) follows from the explicit formulas. $\square$

Corollary (Character form, finite groups). For finite $G$ and complex representations,

$$⟨{\chi }_{{\mathrm{Ind}}_H^{G}W},{\chi }_V{⟩}_G=\dim {\mathrm{Hom}}_G({\mathrm{Ind}}_H^{G}W,V)=\dim {\mathrm{Hom}}_H(W,V{|}_H)=⟨{\chi }_W,{\chi }_V{|}_H{⟩}_H.$$

This is the version Frobenius proved in 1898; the categorical-adjunction formulation became standard only after Eilenberg-Mac Lane (1945) and Kan (1958).

Bridge. The construction here builds toward later units of the strand, where the same pattern is taken up at higher structure. The defining pattern appears again in those units in a sharpened form, where the local data is glued or quotiented. Putting these together, the foundational insight is that the data of this unit gives the structural signature that the rest of the strand reads off.

Exercises [Intermediate+]

Lean formalization [Intermediate+]

lean_status: partial — Mathlib has the basic adjunction structure for induction-restriction in Mathlib.RepresentationTheory.Induced, mediated through tensor-Hom adjunctions for modules over the group algebra.

[object Promise]

Advanced results [Master]

Frobenius reciprocity for compact Lie groups. For a compact Lie group $G$ with closed subgroup $H$, the same adjunction holds:

$${\mathrm{Hom}}_G({\mathrm{Ind}}_H^{G}W,V)\cong {\mathrm{Hom}}_H(W,{\mathrm{Res}}_H^{G}V),$$

where ${\mathrm{Ind}}_H^{G}W$ is constructed as smooth sections of the homogeneous vector bundle $G{\times }_{H}W$ over $G/H$. The multiplicity of an irreducible $V$ of $G$ in ${\mathrm{Ind}}_H^{G}W$ equals the multiplicity of $W$ in $V{|}_H$ — the branching multiplicity.

Branching laws. For semisimple Lie groups $G\supset H$, the decomposition of $V{|}_H$ into irreducible $H$-representations is encoded in branching laws. For ${\mathrm{GL}}_n\supset {\mathrm{GL}}_{n-1}$ (the "Gelfand-Tsetlin" case), the branching is multiplicity-free: each irreducible of $H$ appears at most once in $V{|}_H$, indexed by Gelfand-Tsetlin patterns. For ${\mathrm{SO}}_n\supset {\mathrm{SO}}_{n-1}$, similar branching laws hold. These are central to Gelfand-Tsetlin bases and the geometry of flag manifolds.

Iwahori-Matsumoto isomorphism. For a $p$-adic reductive group $G$ with Iwahori subgroup $I$, the Hecke algebra $H(G,I)$ is isomorphic to a finite Hecke algebra (Iwahori-Matsumoto 1965). Frobenius reciprocity relates representations of $G$ with Iwahori-fixed vectors to representations of this finite Hecke algebra, providing a bridge between $p$-adic representation theory and finite Hecke combinatorics.

Bernstein-Zelevinsky parabolic induction. For $G={\mathrm{GL}}_n(F)$ with parabolic subgroup $P=MN$ (Levi-unipotent factorisation), the induced representation ${\mathrm{Ind}}_P^G\sigma$ from a representation $\sigma$ of the Levi $M$ (extended to $P$ as the identity on $N$) gives the parabolically induced representation. By Frobenius reciprocity, ${\mathrm{Hom}}_G({\mathrm{Ind}}_P^G\sigma ,\pi )\cong {\mathrm{Hom}}_M(\sigma ,{\mathrm{Jac}}_P\pi )$ where ${\mathrm{Jac}}_P$ is the Jacquet functor (the $N$-coinvariants). The Bernstein-Zelevinsky classification (1976–77) of irreducible admissible representations of ${\mathrm{GL}}_n(F)$ proceeds by analysing parabolic induction and Jacquet functors via this adjunction.

Geometric induction. In geometric representation theory (Beilinson-Bernstein localisation, perverse sheaves on flag varieties), the adjunction $\mathrm{Ind}⊣\mathrm{Res}$ is replaced by a six-functor formalism: pullback and pushforward along maps of flag varieties induce convolution operations on perverse sheaves, with adjunctions implementing Frobenius reciprocity in the geometric setting. Kazhdan-Lusztig polynomials and the Kazhdan-Lusztig conjecture (proved by Beilinson-Bernstein and Brylinski-Kashiwara, 1981) sit at the apex of this geometric Frobenius reciprocity.

Synthesis. This construction generalises the pattern fixed in 07.01.07 (induced representation), with the symmetric data replaced by its skew or twisted analogue. Read in the opposite direction, the construction is dual to the metric story: complements and orthogonality are taken with respect to the bilinear datum of this unit, not a metric, and the resulting category of subobjects is the one the rest of the strand classifies. The central insight is that this datum identifies algebra with geometry: functions become vector fields, subspaces become quotients, and invariants become cohomology classes — and that identification is the engine driving every theorem downstream.

Full proof set [Master]

The categorical proof of Frobenius reciprocity is given in the key-theorem section. The character form follows by taking dimensions. The unit and counit of the adjunction are constructed in Exercise 2; the triangle identities are routine verification. The projection formula (Exercise 5) and its applications to Mackey theory occupy chapters of Curtis-Reiner Vol. I and Serre Part III. Lie-theoretic Frobenius reciprocity for compact groups uses the Peter-Weyl theorem (unit 07.07.02) and is in Knapp Ch. VIII; the parabolic version for $p$-adic groups is in Bushnell-Henniart and the Bernstein notes. The geometric version (Beilinson-Bernstein localisation) is beyond the scope of this unit; standard references are Hotta-Takeuchi-Tanisaki D-Modules, Perverse Sheaves, and Representation Theory and the surveys of Soergel.

Connections [Master]

• Induced representation 07.01.07 — Frobenius reciprocity is the structural result organising the relationship between induction and restriction.

• Group representation 07.01.01 — restriction ${\mathrm{Res}}_H^G$ is the canonical functor $\mathrm{Rep}(G)\to \mathrm{Rep}(H)$, and Frobenius reciprocity identifies its left adjoint.

• Character of a representation 07.01.03 — the character form $⟨{\chi }_{\mathrm{Ind}},{\chi }_V{⟩}_G=⟨{\chi }_W,{\chi }_V{|}_H{⟩}_H$ reduces multiplicity computation to subgroup arithmetic.

• Schur's lemma 07.01.02 — Schur's lemma gives the relationship between $\dim \mathrm{Hom}$ and multiplicity.

• Regular representation 07.01.05 — the regular-representation decomposition is a direct application of Frobenius reciprocity to ${\mathrm{Ind}}_{\{e\}}^G\mathbb{C}$.

• Tensor product of representations 07.01.06 — the projection formula combines induction and tensor product.

• Cartan-Weyl classification 07.04.01 — branching laws for compact semisimple Lie groups are governed by Frobenius reciprocity.

• Highest weight representation 07.03.01 — parabolic induction (Lie-group analogue) is a primary source of $G$-representations from Levi subgroup data.

Historical & philosophical context [Master]

Ferdinand Frobenius proved the reciprocity that bears his name in his 1898 paper Über Relationen zwischen den Charakteren einer Gruppe und denen ihrer Untergruppen (Sitzungsberichte der königlich preussischen Akademie der Wissenschaften zu Berlin), the same paper that introduced the induced representation construction. Frobenius stated reciprocity in its character form: for finite $G$ and $H\le G$, irreducible characters $\psi$ of $H$ and $\chi$ of $G$,

$$⟨{\chi }_{\mathrm{Ind}\psi },\chi {⟩}_G=⟨\psi ,\chi {|}_H{⟩}_H.$$

This was a striking duality: the multiplicity of $\chi$ inside the induced character of $\psi$ equals the multiplicity of $\psi$ inside the restricted character of $\chi$. Frobenius derived it by direct manipulation of the induced-character formula, without any modern category-theoretic vocabulary.

The structural significance of Frobenius's discovery only became clear in the mid-20th century. Daniel Kan's 1958 paper Adjoint functors (Trans. AMS) introduced the general categorical notion, abstracting the pattern shared by tensor-Hom, free-forgetful, sheafification-forgetful, induction-restriction, and dozens of other classical constructions. Saunders Mac Lane's textbook Categories for the Working Mathematician (1971) consolidated the categorical perspective and identified Frobenius reciprocity as a paradigmatic example. The recognition that Frobenius's 1898 character identity was a 60-year-old special case of a then-emerging general theory of adjunctions retroactively elevated reciprocity to one of the most important early instances of category-theoretic thinking in mathematics.

The 20th-century extension to Lie groups began with Hermann Weyl (1925–26) for compact Lie groups, where the unitarian trick reduced the analysis to a Peter-Weyl-style decomposition of $L^2(G)$. Harish-Chandra's mid-century classification of admissible representations of real reductive Lie groups (1950s–1970s) and Bernstein-Zelevinsky's 1976–77 papers on ${\mathrm{GL}}_n({\mathbb{Q}}_p)$ both centrally relied on parabolic induction and Frobenius reciprocity in the form ${\mathrm{Hom}}_G({\mathrm{Ind}}_P\sigma ,\pi )\cong {\mathrm{Hom}}_M(\sigma ,{\mathrm{Jac}}_P\pi )$, with the Jacquet functor as the right-adjoint companion to parabolic induction. The geometric Langlands programme, developed since the 1980s by Drinfeld, Beilinson-Drinfeld, Frenkel-Gaitsgory and others, lifts Frobenius reciprocity to the level of categories of perverse sheaves on flag varieties — the Eisenstein and constant-term functors satisfy a six-functor reciprocity that is a far-reaching geometric analogue of Frobenius's 1898 identity.

Bibliography [Master]

• Frobenius, Über Relationen zwischen den Charakteren einer Gruppe und denen ihrer Untergruppen, Sitzungsberichte (1898) — the original.
• Kan, Adjoint functors, Trans. AMS 87 (1958) — the categorical formulation.
• Mac Lane, Categories for the Working Mathematician (1971) — adjunctions as a unifying concept.
• Serre, Linear Representations of Finite Groups — §7.2.
• Fulton & Harris, Representation Theory: A First Course — §3.3.
• Knapp, Lie Groups Beyond an Introduction — Ch. VIII, Frobenius reciprocity for Lie groups.
• Bernstein & Zelevinsky, Induced representations of reductive p-adic groups, I, Ann. Sci. ENS 10 (1977) — parabolic induction and Jacquet adjunction.
• Bushnell & Henniart, The Local Langlands Conjecture for $\mathrm{GL}(2)$ — modern $p$-adic application.
• Hotta, Takeuchi, Tanisaki, D-Modules, Perverse Sheaves, and Representation Theory — geometric Frobenius reciprocity.
• Curtis, Pioneers of Representation Theory (1999) — historical account.