Motivation
A decomposition of a representation into irreducibles describes one action. A decomposition of its algebra into matrix blocks describes all its simple actions at once. For example, the three complex irreducibles of , of dimensions one, one and two, suggest an algebra of the form . The matrix block records the two-dimensional action and its multiplicity in the regular representation; it is not a product of four unrelated scalar actions.
The proof proceeds through a change of viewpoint. Decompose the regular module into simple pieces, compute its endomorphisms using Schur's lemma, and then recover the original ring from those endomorphisms. The order of multiplication matters: endomorphisms of a left regular module correspond to right multiplication, so an opposite ring appears. Tracking it makes the theorem valid over general division rings, where a scalar-field shortcut would be wrong.
Rings and modules are unital, with . Modules in this unit are left modules. The Maschke and Schur results and complex character orthogonality are proved prior background. No classification of real division algebras is assumed.
Definitions
Definition 13.1.1
(Simple modules and a semisimple regular module)
A nonzero module is simple if its only submodules are zero and itself. A ring is semisimple here if its regular left module is a direct sum of simple modules. This module-theoretic definition is the hypothesis of the structural theorem below, not a claim that every ring is of this form.
A division ring has a multiplicative inverse for every nonzero element, without requiring commutative multiplication. The opposite ring has the same additive group as , with multiplication . The matrix ring uses the usual ordered row-by-column products.
A finite direct sum of simple modules has useful closure properties. They are proved here, so that they are not an additional assumption.
Lemma 13.1.2
(Submodules of a finite sum of simple modules)
Let be a finite direct sum of simple modules. Every submodule has a complement, and both and are again finite direct sums of simple modules.
Proof
(Induction on the number of summands)
Induct on the number of summands; with no summands and the claim is trivial. Write , with simple. For , the submodule of is either or . If , then ; by induction has a complement in , and is then a complement of in . If , projection onto is injective on and identifies with a submodule of . By induction , and is the graph of a homomorphism . Its complement is . In both cases induction also identifies and its complement as finite sums of simple modules. The quotient is isomorphic to the complement.
Theorems
Theorem 13.1.3
(Artin–Wedderburn, with the handedness visible)
A semisimple ring is isomorphic to a finite product
where each is a division ring and . Conversely, every such finite product is semisimple. If are the distinct simple summand types of the regular module and their multiplicities, one may take . Every simple left -module is isomorphic to one of the .
Theorem 13.1.4
(Finite-dimensional algebras over an algebraically closed field)
If is a finite-dimensional semisimple algebra over an algebraically closed field , with simple modules of dimensions , then
Consequently a finite group over has this group-algebra decomposition whenever does not divide its order. The assertion about dimensions uses algebraic closure; it cannot be transferred unchanged to the real field.
Concept lensStructural
Counting blocks through the center
The center of is the scalar matrices, so over an algebraically closed the center of is , one dimension per block. For a group algebra the center has a second basis: commutes with every exactly when , so the conjugacy-class sums form a basis. When is algebraically closed and does not divide , the number of simple modules therefore equals the number of conjugacy classes; for both are three.
Proof sketch or proof idea
Why only finitely many simple pieces occur
Proof
(Finitely many summands and their endomorphism blocks)
Write as simple left ideals. The element one has only finitely many nonzero coordinates, lying in a finite subsum . Since is a left submodule, every lies in . Thus . Group its finitely many summands by isomorphism type:
Schur's lemma says is a division ring and for distinct types: a nonzero map between simple modules is both injective and surjective. A homomorphism on the displayed sum is a block matrix of component homomorphisms. The off-type blocks vanish, and composition within one type is matrix multiplication in . Therefore
Recovering the ring, including the opposite
Proof
(Recovering the ring, the converse and the list of simple modules)
Every left-module endomorphism satisfies , so it is right multiplication by . Moreover . Hence . Taking opposite rings in the last display, and using the transpose identification
gives the first structural assertion. To check the transpose rule, reverse the product first and then transpose: its entry is , exactly the entry obtained by multiplying the transposed matrices with scalar multiplication reversed. No commutativity of has been inserted.
For the converse, the left regular module of is the direct sum of its column ideals . Each column ideal is simple. Indeed, for any nonzero column choose with . Given any column , the matrix whose th column has entries and whose other entries vanish sends to . Thus any submodule containing contains every column. The central coordinate idempotents split a finite product into these matrix factors, proving semisimplicity.
If is any simple module, choose . The map , , is onto. At least one simple regular summand has nonzero image, hence maps isomorphically onto . This proves completeness of the list of types. It also shows why a new simple module cannot be hiding outside the regular representation. Every finitely generated module is a quotient of some and is consequently semisimple by the finite-sum complement argument above.
Algebraic closure removes the division rings
Proof
(Division rings collapse over an algebraically closed field)
When is finite-dimensional over , every regular summand and its endomorphism division ring are finite-dimensional over . For , the linear map given by left multiplication on has an eigenvalue . If were nonzero, it would be invertible in , making that linear map minus times the identity invertible, contradicting its eigenvector. Thus . Every element of is scalar, so .
Evaluation at one gives as -vector spaces. Computing the same Hom space from the regular decomposition gives dimension , because only the copies of contribute, each by a one-dimensional endomorphism space. Hence . The ring decomposition and dimension identity follow. Maschke's theorem makes semisimple when the characteristic does not divide , giving the group-algebra consequence.
Proof X-ray
Why the regular module determines multiplication
The argument has one structural input and one key move. The input is semisimplicity of : it lets Schur's lemma, which holds over any ring, cut into blocks with no off-type terms. The key move is evaluation at one: an endomorphism is determined by , so the endomorphism ring is itself acting by right multiplication. Because , this identification reverses products, and the opposite ring records the switch; the transpose rule then moves the opposite onto the division rings without assuming commutativity. Finite dimension and algebraic closure enter only in the last step, through an eigenvalue of left multiplication on : they make , and then turns the multiplicity into the dimension . Over that last step fails, as the cyclic example below shows, while the block structure survives with a division ring .
Character formulas locate the complex blocks
The block decomposition over can be located explicitly by characters.
Proposition 13.1.5
(Central idempotents from characters)
Let be a finite group with distinct complex irreducible characters of degrees , and set
Then acts as the identity on the th simple type and as zero on the others. The are the primitive central idempotents of : , for , and .
Proof
(Traces and character orthogonality)
Characters are constant on conjugacy classes, so is central. On an irreducible it is scalar, say , by Schur. Its trace is
by character orthogonality. Thus it acts as identity on and zero on every other simple type. The regular representation is a sum of these types and is faithful, so , for , and . Under the matrix decomposition, is the identity of its single block. It is a primitive central idempotent: a central idempotent in a matrix ring over is scalar and hence zero or identity. These identities provide exact projectors onto all copies of an irreducible type in any representation.
Worked examples
Worked example 13.1.6
(The complex algebra of )
The trivial, sign and standard characters have dimensions . Thus and its regular module is . For the standard character, values on identity, transpositions and three-cycles are . Its block projector is
The other projectors are and . Their sum is one and their block dimensions are one, one and four, totaling six. A two-dimensional simple appears twice in the regular module, while it gives one matrix block; these are different counts.
Worked example 13.1.7
(The real algebra of a cyclic group of order three)
Write . Then . The factors and are coprime because
CRT therefore gives , identifying the second quotient by sending to a nonreal cube root of one. The two simple real modules have dimensions one and two, each with regular multiplicity one. The latter has endomorphism division ring , of real dimension two. The complex sum-of-squared-dimensions formula would give five rather than three if incorrectly applied over . The actual division-ring blocks have real dimensions one and two, so their sum is correct.
Worked example 13.1.8
(What fails in characteristic dividing the group order)
Over , the cyclic group algebra is
The class is nonzero and nilpotent. A commutative semisimple ring from the structural theorem must be a product of fields: a matrix block of size at least two is noncommutative, and a size-one division block must be commutative. A product of fields has no nonzero nilpotent elements. Thus this group algebra is not semisimple. Commutativity is essential to that last obstruction; matrix algebras themselves contain nonzero nilpotent matrices.
Counterexample mode
Nilpotent elements do not by themselves disprove semisimplicity
The algebra is semisimple by its column decomposition, yet and . The useful general obstruction is a nonzero nilpotent two-sided ideal, not an individual nilpotent element. An exercise below proves that such an ideal cannot occur in a finite product of matrix division rings. Keeping the word “ideal” repairs a frequently overextended field-product intuition.
Common mistakes
Common mistake
(Copies of a simple module are not ring factors)
A regular decomposition counts copies of simple modules; a ring product counts distinct simple types. They are linked by matrix sizes, not by assigning one scalar factor per vector-space dimension. Schur's lemma gives a division ring in general, and only the stated finite-dimensional algebraically closed argument reduces it to the field.
Common mistake
(The opposite ring cannot be dropped)
Right multiplication reverses composition: , not . Deleting the opposite ring is unjustified for noncommutative coefficients; an exercise below shows the reversal already in .
Common mistake
(A complex character table does not repair modular characteristic)
Maschke's theorem has a characteristic hypothesis. Its failure is not repaired by quoting a complex character table in modular characteristic, where the averaging denominators may not exist; the algebra above is not semisimple at all.
Summary
A semisimple regular module has finitely many simple summands. Schur's lemma turns its endomorphisms into matrix blocks, and right multiplication recovers the original ring with the required opposite. Algebraic closure identifies the division rings with the scalar field and ties block sizes to irreducible dimensions. Complex characters then construct the central block projectors explicitly.
Exercises
Exercise 13.1.9
(In , check the multiplication reversal for right-multiplication maps using and . Why does not give a homomorphism from with its original multiplication?)
Foundation
Solution(3)
- 1
Compose two right multiplications
, whereas .
- 2
Evaluate at
Key step
Here and , so .
- 3
Identify the correct source ring
The identity says that is a homomorphism from , precisely because that ring reverses the product.
Exercise 13.1.10
(Prove that a finite-dimensional commutative semisimple algebra over is . Deduce the block structure of .)
Core
Solution(3)
- 1
No matrix block of size at least two
A block with contains and , and .
Because: The algebra is commutative.
- 2
Every division block is
Key step
Each remaining factor is a division ring finite-dimensional over , hence equal to . Thus .
Because: Finite dimension and algebraic closure collapse the division rings.
- 3
Apply this to
, and has the four distinct roots . Polynomial CRT gives four scalar blocks, ; equivalently, the four one-dimensional characters exhaust the dimension four.
Exercise 13.1.11
(For , identify the simple left modules, their regular multiplicities and the primitive central idempotents. Describe the two column ideals in its matrix factor.)
Core
Solution(4)
- 1
List the simple modules
The simple types are , on which only the first factor acts, and , with the standard matrix action and zero first-factor action.
Because: Every simple module occurs in the regular module.
- 2
Count regular multiplicities
, and is the sum of its two columns, each a copy of . The multiplicities are one and two.
- 3
Find the central idempotents
Key step
A central idempotent of the product is a pair of central idempotents of the factors, each or the identity. The primitive ones are and .
- 4
Describe the column ideals
consists of matrices with only the first column possibly nonzero, and of those with only the second. They are distinct isomorphic simple left ideals, not two-sided block ideals.
Exercise 13.1.12
(Prove that has no proper nonzero two-sided ideals for any division ring . Deduce that a finite product of these rings has no nonzero nilpotent two-sided ideal.)
SynthesisProof
Solution(4)
- 1
Isolate one entry
Let a two-sided ideal contain a matrix with . Then lies in .
- 2
Produce every matrix
Key step
Multiplying on the left by the diagonal matrix with in position produces . Thus every matrix unit lies in , and left multiples by scalar diagonal matrices and sums give every matrix: .
- 3
Split ideals of a product
For a two-sided ideal of a finite product, , and each is an ideal of the th factor, so it is zero or the whole factor.
Because: The central coordinate idempotents lie in the ring.
- 4
Rule out nilpotence
∎A nonzero ideal therefore contains the identity of some block. Since for every , no power of the ideal is zero.