Is n or D affected by which minimal left ideal I choose?

No. All minimal left ideals of a simple artinian ring are isomorphic as left modules — there is only one simple left module up to isomorphism — so D = End(_RA) is determined up to ring isomorphism and n = _DA is determined outright. What is not canonical is the isomorphism R ≅ M_n(D) itself.

Simple Artinian Rings and Minimal One-Sided Ideals

Q: Does a simple ring always have a minimal left ideal?

No, and that is the point of (3.10). The Weyl algebra A_1(Q) is simple with zero socle: if it had a minimal left ideal it would be artinian, and the strictly descending chain A_1 x A_1x^2 … shows it is not. The same argument applies to _i M_2^i(D).

Q: Why does the theorem need simplicity rather than just primitivity?

A left primitive ring with nonzero socle need not be artinian — End(V_D) with _D V infinite is the standard example. What simplicity adds is that the socle, being a two-sided ideal, has nowhere to sit except 0 or R. Under primitivity the socle can be a proper nonzero ideal, and the correct statement is the density theorem instead.

Q: Does (3.11) really need no chain condition?

Correct. Rieffel's argument uses only that R is simple and A ≠ 0: simplicity kills the kernel and gives AR = R, which makes the image a left ideal of E containing 1_E. Applied to a non-artinian simple ring it still gives an isomorphism, but onto the endomorphism ring of an infinite-dimensional module over a ring D that need not be a division ring.

Q: Is a simple left noetherian ring necessarily right noetherian?

The methods of this section give no answer. (3.10) transfers the artinian condition across sides only because it produces the two-sided-symmetric normal form M_n(D), and there is no analogous normal form in the noetherian case. Simple noetherian rings have their own structure theory; Cozzens–Faith is the standard reference.

Q: What happens to (3.10) if R has no identity?

It breaks down. Without an identity, AR = R can fail, minimal left ideals may behave pathologically, and even R^2 = R is an extra assumption. The whole of this collection assumes rings with identity and unital modules.

Executive Summary

The Wedderburn–Artin Theorem describes a semisimple ring as a finite product $M_{n_{1}} (D_{1}) \times \dots \times M_{n_{r}} (D_{r})$ . When the ring is simple the product collapses to a single factor, and the interesting question becomes which hypothesis is really needed to reach that factor. Lam's answer $(3.10)$ is that for a simple ring the descending chain condition, semisimplicity, the existence of one single minimal left ideal, and the matrix form $M_{n} (D)$ are all the same condition.

The economical implication is $(3) \Rightarrow (2)$ : one minimal left ideal, in a simple ring, forces the whole ring to be a sum of simple left modules. Behind it is a construction that works in any ring — the sum $B_{𝔄}$ of all minimal left ideals isomorphic to a given one is always two-sided — and simplicity then leaves it no room to be anything but $R$ .

Rieffel's proof of $(3.11)$ goes further: for any simple ring and any nonzero left ideal $𝔄$ , the ring recovers itself as $End (𝔄_{D})$ with $D = End (_{R} 𝔄)$ . No chain condition is used. The chain condition enters only at the last step, to force $\dim_{D} 𝔄$ finite.

4Equivalent conditions in (3.10)

$M_{n} (D)$ Normal form

1Minimal left ideal needed

1965Rieffel's proof

Overview

Throughout, ring means associative ring with identity, modules are unital, and a simple ring is a nonzero ring whose only two-sided ideals are $0$ and $R$ . Simplicity alone is a weak hypothesis: the Weyl algebra $A_{1} (ℚ)$ , the skew Laurent ring $ℚ (t) [x, x^{- 1}; σ]$ for $σ$ of infinite order, and the quotient of $End (V_{D})$ by the finite-rank endomorphisms are all simple and none of them is artinian.

What $(3.10)$ isolates is exactly the extra input that turns a simple ring into a classified object. That input can be phrased as a chain condition, as a module-theoretic decomposition, or — most cheaply — as the assertion that the left regular module has at least one simple submodule.

R simple ⟹ [R left artinian \Leftrightarrow_{R} R semisimple \Leftrightarrow soc (_{R} R) \neq 0 \Leftrightarrow R ≅ M_{n} (D)]

(3.10)

The four conditions of Lam $(3.10)$ , with condition (3) rewritten in terms of the left socle.

A second, independent route to the same classification is Rieffel's. It proves the double centralizer property for every simple ring and every nonzero left ideal, and only then imposes the chain condition to bound a dimension. This is the version that generalises: the Jacobson Density Theorem and Morita theory are both descendants of it, as the pages on primitive rings make explicit.

Once $(3.10)$ is available, the phrase simple artinian ring becomes unambiguous — left artinian and right artinian coincide for simple rings — and by the results on matrix rings such a ring is also left and right noetherian.

Learning Objectives

State the four conditions of $(3.10)$ with the simplicity hypothesis in place.
Prove that the isotypic sum $B_{𝔄}$ of a minimal left ideal is a two-sided ideal in any ring.
Deduce $(3) \Rightarrow (2)$ of $(3.10)$ and hence the left–right symmetry of the artinian condition for simple rings.
Reconstruct Rieffel's proof that $R ≅ End (𝔄_{D})$ for $𝔄$ any nonzero left ideal of a simple ring.
Explain where the descending chain condition is used to force $\dim_{D} 𝔄 < \infty$ .
Compute $𝔄$ , $D$ and $\dim_{D} 𝔄$ for $R = M_{2} (ℚ)$ and check the double centralizer property by hand.

Definitions

Definition—Simple ring

A ring $R$ is simple if $R \neq 0$ and the only two-sided ideals of $R$ are $0$ and $R$ . No finiteness or chain condition is part of the definition. Equivalently, $R a R = R$ for every $a \neq 0$ .

Definition—Minimal left ideal

A left ideal $𝔄 \subseteq R$ is minimal if $𝔄 \neq 0$ and the only left ideals contained in $𝔄$ are $0$ and $𝔄$ . This is precisely the statement that $𝔄$ is a simple submodule of $_{R} R$ .

$soc (_{R} R)$: The left socle: the sum of all minimal left ideals of $R$ , taken to be $0$ if there are none. It is a two-sided ideal of $R$ .
$B_{𝔄}$: For a minimal left ideal $𝔄$ , the sum of all minimal left ideals of $R$ that are isomorphic to $𝔄$ as left $R$ -modules — the isotypic component of $𝔄$ inside the socle.
$End (_{R} 𝔄)$: The ring of left $R$ -module endomorphisms of $𝔄$ , written on the right of its arguments, so that $𝔄$ becomes an $(R, D)$ -bimodule with $D = End (_{R} 𝔄)$ .
$End (𝔄_{D})$: The ring of right $D$ -module endomorphisms of $𝔄$ , written on the left. The natural map $R \to End (𝔄_{D})$ sends $r$ to left multiplication by $r$ .
Simple artinian ring: A simple ring satisfying the DCC on left ideals. By $(3.10)$ the same ring then satisfies the DCC on right ideals, so the adjective needs no side.

Endomorphisms of a left module are written on the right of the argument, and composed with the right-hand rule. This convention removes every opposite ring from the statements below; it is the convention used throughout this collection.

Core Concepts

Isotypic sums are two-sided

The whole of $(3) \Rightarrow (2)$ rests on one observation about right multiplication. If $𝔅$ is a left ideal and $r \in R$ , then the map $ρ_{r} : 𝔅 \to 𝔅 r$ , $b \mapsto b r$ , is a homomorphism of left $R$ -modules. So $𝔅 r$ is a homomorphic image of $𝔅$ and is itself a left ideal.

If $𝔅$ is minimal, its image under $ρ_{r}$ is either $0$ or a copy of $𝔅$ . Both possibilities land inside the isotypic sum $B_{𝔄}$ whenever $𝔅 ≅ 𝔄$ . That is the entire proof that $B_{𝔄}$ is closed under right multiplication.

Minimal left ideal

𝔄

Isotypic sum

B_{𝔄}

Two-sided ideal

B_{𝔄} = R

R

simple

_{R} R

semisimple

Socle reformulation

Condition (3) of $(3.10)$ says $soc (_{R} R) \neq 0$ . Since the socle is a two-sided ideal, in a simple ring it can only be $0$ or $R$ ; and $soc (_{R} R) = R$ is exactly the statement that $_{R} R$ is a sum of simple modules. So for simple rings, (2) and (3) are the two possible values of one invariant.

Two endomorphism rings, one bimodule

Fix a nonzero left ideal $𝔄$ of $R$ and put $D = End (_{R} 𝔄)$ , acting on the right. Then $𝔄$ is an $(R, D)$ -bimodule, and there are two rings acting on it: $R$ on the left, $D$ on the right. Each is contained in the centraliser of the other inside ${End}_{ℤ} (𝔄)$ .

Rieffel's theorem says that when $R$ is simple, the containment $R ↪ End (𝔄_{D})$ is an equality: $R$ is the full centraliser of $D$ . This is the algebraic ancestor of von Neumann's bicommutant theorem, and the reason the property is called the double centralizer property.

Where the chain condition finally enters

Nothing so far bounds $\dim_{D} 𝔄$ . If $𝔄$ is a minimal left ideal then $D$ is a division ring by Schur's Lemma and $𝔄$ is a genuine right $D$ -vector space, but that space could still be infinite-dimensional — and it is, for a simple non-artinian ring with nonzero socle, except that $(3.10)$ shows no such ring exists. The DCC is used exactly once, to rule out an infinite $D$ -independent sequence in $𝔄$ .

Key Results

Lemma(3.9)Isotypic components

Let $R$ be any ring and let $𝔄$ be a minimal left ideal of $R$ . Write $B_{𝔄}$ for the sum of all minimal left ideals of $R$ isomorphic to $𝔄$ as left $R$ -modules. Then:

$B_{𝔄}$ is a two-sided ideal of $R$ ;
if $𝔄$ and $𝔄'$ are minimal left ideals that are not isomorphic as left $R$ -modules, then $B_{𝔄} \cdot B_{𝔄'} = 0$ .

Proof

(1) $B_{𝔄}$ is a sum of left ideals, hence a left ideal. For the right-hand closure it suffices to show $𝔅 r \subseteq B_{𝔄}$ whenever $𝔅$ is a minimal left ideal with $𝔅 ≅ 𝔄$ and $r \in R$ . Right multiplication by $r$ is a map of left $R$ -modules $𝔅 \to 𝔅 r$ , and it is surjective, so $𝔅 r$ is a homomorphic image of the simple module $𝔅$ . Therefore $𝔅 r = 0$ or $𝔅 r ≅ 𝔅 ≅ 𝔄$ . In the second case $𝔅 r$ is a minimal left ideal isomorphic to $𝔄$ , so it lies in $B_{𝔄}$ by definition. Either way $𝔅 r \subseteq B_{𝔄}$ .

(2) It is enough to prove $𝔄 𝔄' = 0$ for the generating minimal left ideals. Suppose $𝔄 a' \neq 0$ for some $a' \in 𝔄'$ . Since $𝔄'$ is minimal and $R a'$ is a nonzero left ideal inside it, $R a' = 𝔄'$ , so $𝔄 a'$ is a nonzero left ideal contained in $𝔄'$ , forcing $𝔄 a' = 𝔄'$ . But right multiplication by $a'$ is a surjection of left $R$ -modules $𝔄 ↠ 𝔄'$ with $𝔄$ simple, hence an isomorphism $𝔄 ≅ 𝔄'$ — contradicting the hypothesis.

Theorem(3.10)Four conditions on a simple ring

Let $R$ be a simple ring. The following are equivalent:

$R$ is left artinian;
$R$ is left semisimple, that is $_{R} R$ is a direct sum of simple modules;
$R$ has a minimal left ideal;
$R ≅ M_{n} (D)$ for some $n \geq 1$ and some division ring $D$ .

When these hold, $n$ and $D$ are uniquely determined — $n$ as an integer and $D$ up to ring isomorphism.

Proof

**(2) $\Leftrightarrow$ (4).** The Wedderburn–Artin Theorem $(3.5)$ writes a left semisimple ring as $M_{n_{1}} (D_{1}) \times \dots \times M_{n_{r}} (D_{r})$ . A product with $r \geq 2$ has proper nonzero two-sided ideals (the factors), so simplicity forces $r = 1$ . Conversely $M_{n} (D)$ is left semisimple by $(3.3)$ .

**(2) $\Rightarrow$ (1).** A left semisimple ring is left artinian; this is $(2.6)$ .

**(1) $\Rightarrow$ (3).** The set of nonzero left ideals of $R$ is nonempty, since $R \neq 0$ contains $R$ itself. By the DCC it has a minimal element, which is a minimal left ideal.

**(3) $\Rightarrow$ (2).** Let $𝔄$ be a minimal left ideal and form $B_{𝔄}$ . By $(3.9) (1)$ this is a two-sided ideal, and it is nonzero because it contains $𝔄$ . Simplicity gives $B_{𝔄} = R$ . But $B_{𝔄}$ is by construction a sum of simple left submodules of $_{R} R$ , so $_{R} R$ is a sum of simple modules, hence semisimple.

Uniqueness of the pair $(n, D)$ is $(3.3)$ together with the Jordan–Hölder Theorem: $D ≅ End (_{R} V)$ for the unique simple left $R$ -module $V$ , and $n$ is the multiplicity of $V$ in $_{R} R$ .

Corollary—The artinian condition is side-neutral for simple rings

A simple ring is left artinian if and only if it is right artinian, and then it is also left and right noetherian. Indeed, condition (4) of $(3.10)$ is left–right symmetric: $M_{n} (D)$ has finite length as a module over itself on either side, since it is a $D$ -space of dimension $n^{2}$ and every one-sided ideal is a $D$ -subspace.

Theorem(3.11)Rieffel — double centralizer property

Let $R$ be a simple ring and let $𝔄$ be any nonzero left ideal of $R$ . Put $D = End (_{R} 𝔄)$ , viewed as a ring of right operators on $𝔄$ , so that $𝔄$ is a right $D$ -module. Then the natural map

f : R ⟶ End (𝔄_{D}), f (r) (a) = r a (a \in 𝔄),

is an isomorphism of rings. No chain condition is assumed on $R$ and $𝔄$ need not be minimal.

Proof

Write $E = End (𝔄_{D})$ , with elements written on the left. Each $f (r)$ really is right $D$ -linear, because the actions of $R$ and $D$ on $𝔄$ are on opposite sides, so $f$ is a well-defined ring homomorphism and $f (1) = {id}_{𝔄} = 1_{E}$ .

Injectivity. $\ker f$ is a two-sided ideal of $R$ , and $1 \notin \ker f$ . Simplicity gives $\ker f = 0$ .

Key identity. For $r \in 𝔄$ and $h \in E$ we claim $h \circ f (r) = f (h (r))$ . Fix $a \in 𝔄$ . Right multiplication $ρ_{a} : x \mapsto x a$ maps $𝔄$ into $𝔄$ (because $𝔄$ is a left ideal and $a \in 𝔄$ ) and is left $R$ -linear, so $ρ_{a} \in D$ . Since $h$ is $D$ -linear, $h (r a) = h (ρ_{a} (r)) = ρ_{a} (h (r)) = h (r) a$ . Therefore

(h \circ f (r)) (a) = h (r a) = h (r) a = f (h (r)) (a),

which is the claim. In particular $E \cdot f (𝔄) \subseteq f (𝔄)$ .

Surjectivity. $𝔄 R$ is a two-sided ideal of $R$ and is nonzero because $𝔄 = 𝔄 \cdot 1 \subseteq 𝔄 R$ ; simplicity gives $𝔄 R = R$ . Applying $f$ , $f (R) = f (𝔄) f (R)$ . Hence

E \cdot f (R) = E \cdot f (𝔄) f (R) \subseteq f (𝔄) f (R) = f (R),

so $f (R)$ is a left ideal of $E$ . It contains $1_{E} = f (1)$ , so $f (R) = E$ .

Corollary(3.13)Structure of a simple left artinian ring

Let $R$ be a simple left artinian ring. Then $R ≅ M_{n} (D)$ for some integer $n \geq 1$ and some division ring $D$ , and both $n$ and $D$ are uniquely determined.

Proof

By the DCC choose a minimal left ideal $𝔄$ . Schur's Lemma $(3.6)$ makes $D : = End (_{R} 𝔄)$ a division ring, so $𝔄$ is a right $D$ -vector space, and $(3.11)$ gives $R ≅ E : = End (𝔄_{D})$ .

Suppose $\dim_{D} 𝔄$ were infinite and pick a $D$ -independent sequence $a_{1}, a_{2}, \dots$ in $𝔄$ . The sets

I_{m} = {g \in E : g (a_{1}) = \dots = g (a_{m}) = 0} (m = 1, 2, \dots)

are left ideals of $E$ , since $g (a_{i}) = 0$ implies $(h g) (a_{i}) = 0$ . Extending ${a_{1}, \dots, a_{m + 1}}$ to a $D$ -basis produces $g \in E$ vanishing on $a_{1}, \dots, a_{m}$ with $g (a_{m + 1}) \neq 0$ , so $I_{1} ⊋ I_{2} ⊋ \dots$ is strictly descending — contradicting the DCC in $R ≅ E$ .

Hence $n : = \dim_{D} 𝔄$ is finite and $E ≅ M_{n} (D)$ by the matrix representation of $D$ -linear maps. Uniqueness is $(3.3)$ .

R ≅ End (𝔄_{D}) ≅ M_{n} (D), D = End (_{R} 𝔄), n = \dim_{D} 𝔄

(3.13)

Every ingredient is read off from a single minimal left ideal $𝔄$ ; different choices of $𝔄$ give isomorphic $D$ and the same $n$ .

Proof Techniques and Method

How these proofs work, and which moves transfer to other arguments.

Four moves carry the section. Each reappears in the density theorem and in Morita theory.

Move 1

Right multiplication is a left-module map

For a left ideal $𝔅$ and $r \in R$ , the map $b \mapsto b r$ is left $R$ -linear. Simplicity of $𝔅$ then forces $𝔅 r$ to be $0$ or a copy of $𝔅$ . This single fact makes $B_{𝔄}$ and the socle two-sided.

Move 2

Build a two-sided ideal, then invoke simplicity

Simplicity is only usable against two-sided objects. The recurring strategy is to manufacture a two-sided ideal — $B_{𝔄}$ , $\ker f$ , $𝔄 R$ — and let simplicity collapse it to $0$ or $R$ .

Move 3

A left ideal of $E$ containing $1_{E}$ is $E$

Surjectivity arguments for maps into an endomorphism ring almost always end this way: show the image absorbs left multiplication, then observe it contains the identity.

Move 4

Turn independence into a strict chain

An infinite independent sequence $a_{1}, a_{2}, \dots$ yields the strictly descending annihilator chain $I_{m} = {g : g (a_{i}) = 0, i \leq m}$ . This converts a dimension statement into a chain condition and is the only place the DCC is used.

Move 4 is worth isolating because it is reversible in practice: whenever a structure theorem needs finite dimension, look for the annihilator chain that the chain condition is supposed to break.

Worked Example

Verifying $(3.11)$ inside $M_{2} (ℚ)$

Take $R = M_{2} (ℚ)$ and let $𝔄$ be the set of matrices whose second column is zero:

𝔄 = {(\begin{matrix} a & 0 \\ b & 0 \end{matrix}) : a, b \in ℚ} .

(E.1)

$𝔄$ is a left ideal, and it is minimal: if $0 \neq A \in 𝔄$ has first column $(a, b)^{T} \neq 0$ , then left multiplying by suitable matrices sends $(a, b)^{T}$ to any prescribed column vector, so $R A = 𝔄$ . As a left $R$ -module, $𝔄 ≅ ℚ^{2}$ , the column space.

The division ring $D$

An endomorphism of $_{R} 𝔄$ commutes with left multiplication. Right multiplication by $d E_{11} = (\begin{matrix} d & 0 \\ 0 & 0 \end{matrix})$ sends $(\begin{matrix} a & 0 \\ b & 0 \end{matrix})$ to $(\begin{matrix} a d & 0 \\ b d & 0 \end{matrix})$ and is left $R$ -linear. Conversely every element of $End (_{R} 𝔄)$ arises this way, so

D = End (_{R} 𝔄) ≅ ℚ, \dim_{D} 𝔄 = 2 .

(E.2)

Therefore $End (𝔄_{D}) ≅ End (ℚ_{ℚ}^{2}) ≅ M_{2} (ℚ) = R$ , exactly as $(3.11)$ predicts, and $(3.13)$ returns $n = 2$ , $D = ℚ$ .

A simple ring with no minimal left ideal

Fix a division ring $D$ and set $R_{i} = M_{2^{i}} (D)$ , embedding $R_{i} ↪ R_{i + 1}$ by $M \mapsto (\begin{matrix} M & 0 \\ 0 & M \end{matrix})$ . All the embeddings preserve the identity, so the union $R = ⋃_{i \geq 0} R_{i}$ is a ring.

$R$ is simple: a nonzero ideal $I$ meets some $R_{i}$ in a nonzero ideal of $R_{i}$ , which is all of $R_{i}$ by $(3.1)$ , so $1 \in I$ . But $R$ is not left artinian. Let $e_{i} \in R_{i}$ be the matrix unit $E_{11}$ of size $2^{i}$ , viewed in $R$ . Then $e_{i + 1} = e_{i + 1} e_{i}$ , so

R e_{0} ⊋ R e_{1} ⊋ R e_{2} ⊋ \dots

(E.3)

Strictness follows by comparing the $(2^{i} + 1, 2^{i} + 1)$ entry, which is $1$ for $e_{i}$ and $0$ for every element of $R e_{i + 1}$ .

By $(3.10)$ , this ring therefore has no minimal left ideal at all: its left socle is zero. That is a genuinely non-obvious conclusion drawn from the theorem rather than from a direct search.

Comparison and Classification

Simple rings, with and without the chain condition
Ring	Simple?	Minimal left ideal?	Left artinian?	Left noetherian?
$M_{n} (D)$ , $D$ a division ring	yes	yes	yes	yes
$A_{1} (k_{0})$ , $k_{0}$ a field of characteristic $0$	yes	no	no	yes
$k [x, x^{- 1}; σ]$ , $k$ a field, $σ$ of infinite order	yes	no	no	yes
$⋃_{i} M_{2^{i}} (D)$	yes	no	no	no — it is von Neumann regular but not semisimple
$End (V_{D}) / (finite rank)$ , $\dim_{D} V = ℵ_{0}$	yes	no	no	no
$End (V_{D})$ , $\dim_{D} V = ℵ_{0}$	no — finite-rank ideal	yes	no	no
$ℤ$	no	no	no	yes

Which hypotheses (3.10) actually needs
	Simple	Has minimal left ideal	Semisimple	Isomorphic to some $M_{n} (D)$
$M_{3} (ℍ)$	yes	yes	yes	yes
$M_{2} (ℚ) \times ℚ$	no	yes	yes	no
$A_{1} (ℚ)$	yes	no	no	no
$End (V_{D})$ , $V$ infinite-dimensional	no	yes	no	no
Upper triangular $T_{2} (k)$	no	yes	no	no

Which hypotheses (3.10) actually needs

The third and fourth rows are the two ways $(3.10)$ can be misremembered. Dropping simplicity leaves a ring with minimal left ideals that is far from semisimple; keeping simplicity but dropping the socle hypothesis leaves a ring with no classification at all.

Relationship Map

For a simple ring the four conditions form a cycle rather than a hierarchy — which is the content of the theorem.

Left artinianMinimal left ideal exists

B_{𝔄} = R

_{R} R

semisimple

R ≅ M_{n} (D)

Left artinian

Semiprime ringsno nonzero nilpotent ideals

Prime rings

𝔄 𝔅 = 0 \Rightarrow 𝔄 = 0

𝔅 = 0

Simple ringsonly ideals are

0

and

R

; always left and right primitive

Simple noetherian ringsWeyl algebras, skew Laurent rings over fields

Simple artinian rings

M_{n} (D)

; equivalently simple with nonzero socle

Simple ring $R$ — always true
- consequences with no extra hypothesis
  - $rad R = 0$ , since the radical is a proper two-sided ideal
  - $R$ is prime, hence semiprime
  - $R$ is left and right primitive
  - $Z (R)$ is a field
  - $R ≅ End (𝔄_{D})$ for every nonzero left ideal $𝔄$
- consequences requiring $soc (_{R} R) \neq 0$
  - $R ≅ M_{n} (D)$ with $n$ , $D$ unique
  - $R$ is left and right artinian and noetherian
  - $R$ has exactly one simple left module up to isomorphism
  - every left $R$ -module is projective and injective
- things that can still fail
  - $R$ need not be noetherian on either side
  - $R$ need not be finite-dimensional over $Z (R)$
  - $R$ need not have any minimal one-sided ideals

Applications and Industry Use

Applications here means where this structure is used — inside mathematics and in the engineering and computing disciplines that consume it.

Division algebras

Brauer groups and space-time codes

Classifying simple artinian algebras finite-dimensional over a field $k$ reduces, via $M_{n} (D)$ , to classifying the division algebras $D$ . Cyclic division algebras built this way supply the fully diverse codebooks used in MIMO space-time block coding.

Representation theory

Block and isotypic decomposition

The $B_{𝔄}$ construction is the isotypic decomposition of a group algebra. For $k G$ with $char k ∤ | G |$ it produces the Wedderburn components, one per irreducible representation.

Operator algebras

Ancestor of the bicommutant theorem

$(3.11)$ says a simple ring is its own double centraliser on any nonzero left ideal. Von Neumann's bicommutant theorem is the topological analogue, and type I factors are the operator-algebra counterpart of $M_{n} (D)$ .

Symbolic computation

Wedderburn decomposition in CAS

GAP's wedderga package returns the simple components of a semisimple group algebra; Magma and Sage expose the analogous decomposition for finite-dimensional algebras. Each component is exactly an $M_{n} (D)$ produced by $(3.13)$ .

Coding theory

Codes as one-sided ideals

A cyclic code over a finite field is a left ideal in a semisimple group algebra; the $M_{n} (D)$ decomposition converts code design into a choice of components, which is how minimum-distance bounds are computed.

Category theory

Morita equivalence

$R$ and $M_{n} (R)$ have equivalent module categories. $(3.11)$ is a hands-on instance of the general Morita correspondence and motivates it historically.

The honest summary: this theorem is infrastructure inside algebra. Its downstream uses are almost all mediated by a classification step — you decompose an algebra into matrix rings first, and the engineering payload is computed inside the components.

Design Considerations

Design considerations here means the choices made when modelling a problem with these algebraic structures.

Which left ideal to centralise. $(3.11)$ holds for every nonzero left ideal, but the answer is only $M_{n} (D)$ with $D$ a division ring when $𝔄$ is minimal. Choosing a larger $𝔄$ gives a correct but less useful $D$ .
Side conventions. Writing endomorphisms of left modules on the right removes opposite rings from every statement. If you write them on the left instead, $(3.13)$ becomes $R ≅ M_{n} (D^{op})$ and half the literature's discrepancies are explained by this choice alone.
Simple versus semisimple as the working hypothesis. Model with semisimple when you want a finite product and per-component analysis; model with simple when you want a single division ring and are willing to prove the chain condition separately.
Do not assume finite dimension over the centre. A simple artinian ring is $M_{n} (D)$ , but $D$ may be infinite-dimensional over $Z (D)$ . If your argument needs finite dimension, that is an extra hypothesis and belongs in the statement.

Standards and Notation

Standards here covers notation, symbol and markup standards, and reference implementations, rather than material or design codes.

Matrix ring

M_{n} (D)

in this collection;

{Mat}_{n} (D)

and

D^{n \times n}

elsewhere

Socle

soc (_{R} R)

for the left socle; some authors write

Soc (R)

without a side

Endomorphisms

End (_{R} M)

written on the right;

{End}_{R} (M)

elsewhere, usually on the left

Simple ringNonzero, only trivial two-sided ideals; DCC is not implied

MarkupPresentation MathML per ISO/IEC 40314; symbol conventions per ISO 80000-2

GAPWedderburnDecomposition in the wedderga package; RadicalOfAlgebra

Magma / SageWedderburnDecomposition, A.wedderburn_decomposition()

Computational Notes

Computational notes cover algorithms, cost and library behaviour rather than manufacturing process.

For a finite-dimensional algebra $A$ over a field $k$ given by structure constants, with $\dim_{k} A = m$ :

Deciding simplicity is easy once the radical is known: $A$ is simple iff $rad A = 0$ and the Wedderburn decomposition has one factor. Both steps are polynomial time in $m$ over a finite field.
Producing an explicit isomorphism $A ≅ M_{n} (D)$ needs a rank-one idempotent, equivalently a minimal left ideal. Over a finite field this is polynomial time (the Meataxe and Rónyai's algorithms); over $ℚ$ it is not known to be easy — Rónyai showed that constructing an explicit isomorphism $A ≅ M_{n} (ℚ)$ is at least as hard as factoring integers.
The invariant $n$ and the isomorphism type of $D$ can sometimes be read off without constructing the isomorphism, using the reduced characteristic polynomial or local invariants at each place.
For infinite-dimensional presentations there is no algorithm at all: the word problem for finitely presented associative algebras is undecidable, so even testing whether a given element is zero can fail to terminate, and no general simplicity test exists.

Failure Modes and Common Mistakes

Do not conclude $D ≅ D'$ from $M_{n} (R) ≅ M_{m} (S)$ for general rings $R$ , $S$ — the uniqueness in $(3.13)$ uses that $D$ and $D'$ are division rings.
Do not assume that a simple left noetherian ring is right noetherian. $(3.10)$ gives side-symmetry for the artinian condition only, and the argument passes through the matrix form, which the noetherian case does not supply.
Do not identify $End (_{R} 𝔄)$ with a subring of $R$ . It acts on the right and is generally not realisable inside $R$ ; conflating the two produces spurious opposite rings.
Do not expect the isomorphism $R ≅ M_{n} (D)$ to be canonical. It depends on a choice of minimal left ideal and of $D$ -basis; only $n$ and the isomorphism class of $D$ are invariants.

Best Practices

State which of the four conditions you are assuming; proofs that slide between them without naming the hypothesis are hard to audit.
When you need $M_{n} (D)$ , verify the DCC on one side explicitly — for a finite-dimensional algebra it is free, otherwise it is the real content.
Prefer the socle formulation $soc (_{R} R) \neq 0$ when arguing with other structure theorems; it composes better with the primitive-ring results than 'has a minimal left ideal'.
Record the side convention for endomorphism rings once, at the top of the argument, and never change it.

Historical Notes and Lessons Learned

1893MolienT. Molien classifies finite-dimensional semisimple algebras over the complex numbers, obtaining products of full matrix algebras — the earliest form of the result.
1907WedderburnJ. H. M. Wedderburn proves the structure theorem for finite-dimensional algebras over an arbitrary field, with the radical defined as the largest nilpotent ideal.
1927ArtinE. Artin replaces finite dimensionality by the descending chain condition on left ideals, creating the class of left artinian rings and the modern statement.
1945JacobsonThe density theorem generalises the argument to left primitive rings; a simple artinian ring becomes the special case in which the socle is the whole ring.
1958–65Morita and RieffelMorita's equivalence theory subsumes the matrix classification categorically, and Rieffel gives the short double-centralizer proof reproduced in (3.11), valid for every simple ring.

The methodological lesson repeats the one from radical theory: the classical statement was tied to a finiteness assumption, and each generalisation came from replacing the assumption by a property of the module category. Rieffel's proof is the extreme point — it uses no finiteness whatever, and the chain condition reappears only as a dimension bound at the very end.

Quick Reference

Hypothesis

R

simple:

R \neq 0

, only ideals

0

and

R

(3.10)(1)Left artinian

(3.10)(2)

_{R} R

semisimple

(3.10)(3)Some minimal left ideal exists, i.e.

soc (_{R} R) \neq 0

(3.10)(4)

R ≅ M_{n} (D)

D

a division ring

(3.11)

R ≅ End (𝔄_{D})

for every left ideal

𝔄 \neq 0

Recipe for $D$

D = End (_{R} 𝔄)

𝔄

minimal

Recipe for $n$

n = \dim_{D} 𝔄

Side symmetryLeft artinian and right artinian agree for simple rings

What each hypothesis buys
Assume	Conclude	Reference
$R$ simple	$R$ prime, primitive, $rad R = 0$ , $Z (R)$ a field	§3, Exercise 3.4
$R$ simple, $𝔄 \neq 0$ a left ideal	$R ≅ End (𝔄_{D})$	(3.11)
$R$ simple with a minimal left ideal	$_{R} R$ semisimple, $R$ artinian	(3.10)
$R$ simple left artinian	$R ≅ M_{n} (D)$ with $n$ , $D$ unique	(3.13)
$R$ semisimple	$R ≅ \prod_{i} M_{n_{i}} (D_{i})$ , $r$ factors unique	(3.5)
$R$ any ring, $𝔄$ minimal left ideal	$B_{𝔄}$ is a two-sided ideal	(3.9)

Frequently Asked Questions

Does a simple ring always have a minimal left ideal?

No, and that is the point of $(3.10)$ . The Weyl algebra $A_{1} (ℚ)$ is simple with zero socle: if it had a minimal left ideal it would be artinian, and the strictly descending chain $A_{1} x ⊋ A_{1} x^{2} ⊋ \dots$ shows it is not. The same argument applies to $⋃_{i} M_{2^{i}} (D)$ .

Why does the theorem need simplicity rather than just primitivity?

A left primitive ring with nonzero socle need not be artinian — $End (V_{D})$ with $\dim_{D} V$ infinite is the standard example. What simplicity adds is that the socle, being a two-sided ideal, has nowhere to sit except $0$ or $R$ . Under primitivity the socle can be a proper nonzero ideal, and the correct statement is the density theorem instead.

n

D

affected by which minimal left ideal I choose?

No. All minimal left ideals of a simple artinian ring are isomorphic as left modules — there is only one simple left module up to isomorphism — so $D = End (_{R} 𝔄)$ is determined up to ring isomorphism and $n = \dim_{D} 𝔄$ is determined outright. What is not canonical is the isomorphism $R ≅ M_{n} (D)$ itself.

Does

(3.11)

really need no chain condition?

Correct. Rieffel's argument uses only that $R$ is simple and $𝔄 \neq 0$ : simplicity kills the kernel and gives $𝔄 R = R$ , which makes the image a left ideal of $E$ containing $1_{E}$ . Applied to a non-artinian simple ring it still gives an isomorphism, but onto the endomorphism ring of an infinite-dimensional module over a ring $D$ that need not be a division ring.

Is a simple left noetherian ring necessarily right noetherian?

The methods of this section give no answer. $(3.10)$ transfers the artinian condition across sides only because it produces the two-sided-symmetric normal form $M_{n} (D)$ , and there is no analogous normal form in the noetherian case. Simple noetherian rings have their own structure theory; Cozzens–Faith is the standard reference.

What happens to

(3.10)

R

has no identity?

It breaks down. Without an identity, $𝔄 R = R$ can fail, minimal left ideals may behave pathologically, and even $R^{2} = R$ is an extra assumption. The whole of this collection assumes rings with identity and unital modules.

References

T. Y. Lam, A First Course in Noncommutative Rings, Graduate Texts in Mathematics 131, Springer-Verlag, 1991, §3, results (3.9)–(3.13) (pp. 37–41).
M. A. Rieffel, “A general Wedderburn theorem”, Proceedings of the National Academy of Sciences USA 54 (1965), 1513.
N. Jacobson, Structure of Rings, American Mathematical Society Colloquium Publications 37, revised edition, 1964, Chapter IV.
E. Artin, “Zur Theorie der hyperkomplexen Zahlen”, Abhandlungen aus dem Mathematischen Seminar der Universität Hamburg 5 (1927), 251–260.
F. W. Anderson and K. R. Fuller, Rings and Categories of Modules, 2nd edition, Graduate Texts in Mathematics 13, Springer-Verlag, 1992, §13 and §22.
J. H. Cozzens and C. Faith, Simple Noetherian Rings, Cambridge Tracts in Mathematics 69, Cambridge University Press, 1975.

AI Suggested Questions

Work through the proof that $soc (_{R} R)$ is a two-sided ideal in an arbitrary ring.
Give a simple ring whose socle is zero but which is left and right noetherian, and prove both claims.
How does Rieffel's proof of $(3.11)$ specialise to give the Jacobson Density Theorem for left primitive rings?
Construct two non-isomorphic division rings $D$ , $D'$ with $M_{2} (D) ≅ M_{3} (D')$ , or prove no such pair exists.
Explain how the isotypic decomposition $R = B_{1} \oplus \dots \oplus B_{r}$ recovers the block decomposition of a group algebra.
What is the socle of $End (V_{D})$ for $\dim_{D} V$ infinite, and why is the quotient by it simple?
Compare $(3.11)$ with the Morita theorem: which hypotheses does each need, and what does Morita give that Rieffel does not?

01Executive Summary

02Overview

03Learning Objectives

04Definitions

05Core Concepts

Isotypic sums are two-sided

Socle reformulation

Two endomorphism rings, one bimodule

Where the chain condition finally enters

06Key Results

07Proof Techniques and Method

Right multiplication is a left-module map

Build a two-sided ideal, then invoke simplicity

A left ideal of E containing 1E is E

Turn independence into a strict chain

08Worked Example

Verifying (3.11) inside M2(ℚ)

The division ring D

A simple ring with no minimal left ideal

09Comparison and Classification

10Relationship Map

11Applications and Industry Use

Brauer groups and space-time codes

Block and isotypic decomposition

Ancestor of the bicommutant theorem

Wedderburn decomposition in CAS

Codes as one-sided ideals

Morita equivalence

12Design Considerations

13Standards and Notation

14Computational Notes

15Failure Modes and Common Mistakes

16Best Practices

17Historical Notes and Lessons Learned

18Quick Reference

19Frequently Asked Questions

20Related KEVOS Topics

21References

22AI Suggested Questions

Continue learning

Executive Summary

Overview

Learning Objectives

Definitions

Core Concepts

Key Results

Proof Techniques and Method

A left ideal of $E$ containing $1_{E}$ is $E$

Worked Example

Verifying $(3.11)$ inside $M_{2} (ℚ)$

The division ring $D$

Comparison and Classification

Relationship Map

Applications and Industry Use

Design Considerations

Standards and Notation

Computational Notes

Failure Modes and Common Mistakes

Best Practices

Historical Notes and Lessons Learned

Quick Reference

Frequently Asked Questions

Related KEVOS Topics

References

AI Suggested Questions