Motivation
Restricting a representation to a subgroup is immediate: keep the same vector space and use fewer group elements. Moving in the other direction is harder. A subgroup representation gives no prescribed action for elements outside the subgroup, and an arbitrary extension on the same space may not exist. Induction enlarges the space in a controlled way, making one copy for each left coset and letting the original subgroup action resolve how those copies are joined.
The decisive property is a correspondence of maps. A map from the induced representation to a representation of the larger group is determined exactly by a subgroup-equivariant map from the original space. This is Frobenius reciprocity in its linear form. Over the complex numbers, character inner products convert that correspondence into a practical decomposition method.
Throughout, is a finite group, , and representations are finite-dimensional over a field . The construction and map correspondence work in every characteristic. Only the character-decomposition statements below impose and use the earlier complex orthogonality theorem. An averaging formula with division by is not asserted in modular characteristic.
Definitions
Definition 7.1
(Restriction and induction with the sides specified)
For a -module , its restriction is the same vector space with only the -action. For an -module , define
Here is a left -module and a right -module by multiplication, while is a left -module. Concretely, start with the ordinary vector-space tensor and quotient by the subspace generated by
The surviving symbols are still written . Linearity extends these balancing relations from group elements to all of .
The left -action is . It respects balancing because , and the group laws follow from multiplication in . Thus induction is a representation, not merely a vector space with a suggestive name. The sided definition matters because the group algebras need not be commutative; the previous chapter's commutative-base convention cannot simply be transferred without adjustment.
Theorems
Theorem 7.2
(Coset normal form and dimension)
Choose representatives for the left cosets . Every induced vector has a unique expression , and
If with , the action sends to .
Theorem 7.3
(Frobenius reciprocity as a correspondence of maps)
For an -module and a -module , there is a natural linear isomorphism
It sends to . The inverse sends an -map to the -map .
Theorem 7.4
(Induced characters and character reciprocity over the complex numbers)
For and a representation with character ,
With the class-function inner product linear in its first argument,
Thus the multiplicity of an irreducible in an induced representation can be calculated after restricting to the smaller group.
Proof sketch or proof idea
A direct-sum model verifies the normal form
Proof
(Coset normal form)
Every group element has a unique expression for . Define a linear map from to the induced module by . It is onto because balancing rewrites as .
For an inverse, send to the tuple with in slot and zeros elsewhere, extending linearly in the group-algebra argument. This respects the relation for any : the values of and are both in the same slot. It is also bilinear over , so it descends through the tensor quotient. The two maps are inverse on generators, proving uniqueness and the dimension formula. The action formula follows by rewriting in this normal form.
Changing representatives changes these coordinate descriptions, not the induced object defined by the quotient. In particular, a selected coset representative is not a group homomorphism from a quotient group, and need not be normal. The coset space serves as an index set regardless of whether it has group multiplication.
Concept lensAlgebraic
Induced vectors as equivariant functions on the group
Let be the space of maps with , where acts by . Send to the function whose value at is when and otherwise. This respects balancing and the -actions, and evaluation at the representatives recovers the coordinates . Both spaces have dimension , so : an induced vector is a function determined by its values on one representative per coset.
The reciprocity maps preserve exactly the required relations
Proof
(Frobenius reciprocity)
Let be a -map. For ,
so its proposed restriction is -linear. Conversely, if is -linear, the rule respects balancing:
It therefore defines a linear map on the quotient, and left multiplication shows it is -equivariant. Starting with and evaluating at recovers . Starting with and reconstructing gives , first for group elements and then by linearity for all . Thus the maps are inverse. Their linearity and naturality follow by composing and evaluating on these same generators.
Proof X-ray
Balancing is the exact extension condition
The proof has two directions, and each uses a different hypothesis. From to , the only inputs are the balancing relation and -equivariance of applied to ; together they give -linearity of . From to , the rule is first a map on , and the one thing to check is that it kills the generators of the quotient. That check, , is exactly -linearity of , so the hypothesis is used once and in full. Uniqueness comes from : the translates of span the induced module, so a -map is determined by its values there. No step divides by or , which is why the correspondence holds in every characteristic.
There is no averaging and no division by a group order in this proof. Its validity in modular characteristic does not imply complete reducibility there; it is a statement about Hom spaces, which continue to exist even when representations have nonsplit extensions.
Trace sees only fixed coset slots
Proof
(Induced characters and character reciprocity)
Use the coset direct-sum model. If moves slot to a different slot, that block contributes nothing to the matrix trace. The fixed slots are exactly those with , equivalently . On such a slot the action is the original operator on . Adding its trace gives the first character formula.
Replace a representative by . The resulting subgroup element is conjugated by , so its trace does not change. Every left coset has such representatives. Summing over all and dividing by therefore gives the second formula over . In particular, the induced character vanishes on a group element that is not conjugate to any element of . At the identity all slots contribute, recovering the dimension formula.
The complex character theorem gives
Take complex conjugates to obtain the stated orientation of the inner products. These particular values are real nonnegative integers, but the conjugation explains the convention without silently switching the linear argument. For irreducible , the left side in the stated orientation is its multiplicity, by orthogonality and complete reducibility over the complex numbers.
Comparing the maps does not make the operations inverse
There are canonical comparison maps , , and , . The first is injective: choose the identity as the representative of , so it is inclusion of that slot. The second is onto because maps to . These properties come directly from the construction.
They need not be isomorphisms. When and the space is nonzero, induction multiplies its dimension by an index larger than one, whereas restriction leaves the dimension unchanged. The correspondence of Hom spaces is an adjunction, meaning precisely that maps of one displayed type correspond naturally to maps of the other. It is not an assertion that inducing and restricting undo each other.
Worked examples
Worked example 7.5
(Inducing a trivial character gives a coset permutation action)
If has trivial -action, balancing makes the induced space the vector space with basis , and permutes that basis. Its character over counts fixed left cosets. For and , the cosets can be identified with three points: is the stabilizer of point three, and is well defined and bijective. The induced character on the identity, transpositions and three-cycles is . Using the earlier characters and , it is .
Worked example 7.6
(A nontrivial subgroup character changes the decomposition)
Keep but let its nontrivial character send to minus one. At the identity the induced trace is three. A transposition fixes one coset, and on that slot acts by the nontrivial element of , contributing minus one. A three-cycle fixes no coset. Thus .
The standard representation restricts to on , since its transposition has eigenvalues one and minus one. Frobenius reciprocity therefore gives multiplicity one for the standard constituent in either induced representation. The trivial and sign constituents are distinguished by their restrictions, not by the common induced dimension.
Worked example 7.7
(Induction on a cyclic group can split over the complex numbers)
Let have order four and . Give the action . With representatives , the induced basis is , , and , . Thus the matrix of is
Its eigenvalues are , so it splits into the two one-dimensional characters and . Reciprocity gives the same answer: among the four characters , exactly those with odd restrict to the chosen nontrivial -character. The induced character on is . Irreducibility of a subgroup representation does not imply irreducibility after induction.
Counterexample mode
A correspondence of maps is not an inverse construction
Restrict the one-dimensional trivial -module to a proper subgroup and induce it back. The result has dimension , and is the coset permutation representation rather than the original line. The repaired assertion is Frobenius's correspondence of the specified Hom spaces, with its canonical comparison maps, not an isomorphism between every original and twice-transformed representation.
Working with a character calculation
Before summing the induced-character formula, identify which cosets can be fixed. A vanishing result outside conjugates of can save most of the calculation. At a fixed coset, compute the actual subgroup element and use the subgroup character; counting fixed cosets alone works only for induction of the trivial character. Check the identity value against the dimension formula.
For a decomposition, restriction may be shorter than direct induction. List the known irreducible characters of , restrict them to , and pair with the original -character. Each resulting multiplicity must be a nonnegative integer, and the weighted sum of constituent dimensions must equal . These are consequences of the theorem as well as useful arithmetic checks; a near-integer numerical result is not itself an exact multiplicity proof.
Common mistakes
Common mistake
(Mixing left and right coset conventions)
Use left cosets to decompose the right -module into the summands . Mixing a right-coset convention with the displayed formula changes where subgroup elements act. The tensor relation is exactly what transfers a right subgroup factor to the left action on .
Common mistake
(Counting fixed cosets for a nontrivial character)
For a nontrivial -character, the number of fixed cosets is not the induced character value. In the second worked example a transposition fixes one coset, but the value there is , not . Each fixed coset contributes ; only for the trivial character is every such contribution equal to one.
Induction enlarges dimension by the index; restriction preserves the underlying dimension and may split an irreducible.
Common mistake
(Importing complex formulas into modular characteristic)
The Hom adjunction works over arbitrary fields, but the complex inner-product multiplicity calculation uses complete reducibility and the complex character theorem. A modular trace computation is not certified by inserting division by a group order that vanishes in the field.
Summary
Coset normal form constructs induction with explicit subgroup actions on each slot. Balancing proves the natural Hom correspondence, and fixed-slot traces yield the induced-character formula. Over the complex numbers, reciprocity turns decomposition into restriction calculations, while the examples separate the adjunction from any claim that induction and restriction are inverse or preserve irreducibility.
Exercises
Quick check
(For the trivial subgroup , prove by explicit inverse maps, over any field.)
Evaluate a -map at the identity.
Solution · Answer
Send a -map to . In the other direction, send to the linear map . It is -equivariant, the composites are identity on and on each group-basis element, and no characteristic restriction was used. The Hom-space dimension is consequently .
Quick check
(If , prove induction in stages: .)
Multiply the two group-algebra factors together.
Solution · Answer
Send to . Moving an factor between and leaves the product unchanged, and moving a factor from to is exactly the target balancing relation. Thus the map is well defined and -linear. Its inverse sends to ; the balancing relation holds by moving first through the outer balance and then through the inner balance. One composite plainly fixes ; the other rewrites by outer balancing. Generators span, so these maps are inverse.
Quick check
(In the cyclic order-four example, induce the trivial character of instead. Give the matrix of , its four character values, and its decomposition.)
Recompute with the trivial action of .
Solution · Answer
The relations now give , , so the matrix is . Its eigenvalues are one and minus one; the character is . These are precisely the even-index characters whose restriction to is trivial.
Quick check
(Let and let a nontrivial complex character of send to , where and . Compute its induced character and decide irreducibility.)
Conjugating by a transposition gives .
Solution · Answer
The identity value is two. A transposition cannot be conjugate into , so its value is zero. At a three-cycle the two cosets contribute and , whose sum is minus one. Thus the character is , the standard irreducible character of . Its squared norm using class sizes is , confirming irreducibility. The conjugate nontrivial -character induces the same representation, so induction need not distinguish inequivalent subgroup representations.