The Radical of a Left Artinian Ring Is Nilpotent

Q: Does (4.12) need artinian on both sides?

No. Left DCC alone gives that rad R is nilpotent, and nilpotency is a two-sided condition, so the radical is then simultaneously the largest nilpotent left ideal and the largest nilpotent right ideal. This matters because there are rings artinian on one side only.

Q: Why is the sum of all nilpotent ideals nilpotent here but not in general?

Because in a left artinian ring that sum is contained in rad R, which is itself nilpotent. Without a chain condition the sum can be an infinite ascending union of nilpotent ideals with no common bound on the exponent, and it is then nil at best.

Q: How large can the nilpotency index be relative to the dimension?

For a k-algebra of dimension n the index is at most n, and the bound is attained: k[x]/(x^n) has index exactly n. For T_n(k) the dimension is n(n+1)/2 while the index is only n, so the bound is far from tight in general.

Q: What replaces (4.12) for perfect rings?

Left T-nilpotency: for a left perfect ring, rad R need not be nilpotent, but every sequence a_1, a_2, … of radical elements has some product a_n … a_2 a_1 equal to zero. That condition is exactly what keeps Nakayama-style arguments alive without a uniform exponent.

Executive Summary

Jacobson's radical is defined for every ring, but only under a chain condition does it recover the classical object. Lam's $(4.12)$ is that recovery: if $R$ is left artinian, then $rad R$ is nilpotent, and it is the largest nilpotent left ideal as well as the largest nilpotent right ideal.

The immediate corollary $(4.13)$ is the first nil implies nilpotent theorem in the subject: in a left artinian ring every nil one-sided ideal is nilpotent, because it is trapped inside a nilpotent ideal. Everything downstream — Hopkins–Levitzki, Loewy series, block theory of finite-dimensional algebras — needs a nilpotent radical to filter with.

DCCThe only hypothesis, on left ideals

$J^{n} = 0$ Conclusion of (4.12)

BothLargest nilpotent left and right ideal

(4.13)Nil implies nilpotent

Overview

Write $J = rad R$ . The descending chain $J \supseteq J^{2} \supseteq J^{3} \supseteq \dots$ must stabilise under the DCC, so $J^{k} = J^{k + 1} = \dots$ for some $k$ . Everything hinges on showing that this stable value is zero, and the argument for that is a minimal-counterexample argument using the DCC a second time.

The result closes a historical gap. Wedderburn defined the radical of a finite-dimensional algebra as its largest nilpotent ideal; Jacobson defined a radical for every ring by quasi-regularity. $(4.12)$ says the two definitions agree exactly where Wedderburn's makes sense, so nothing was lost in the generalisation.

R left artinian ⟹ (rad R)^{n} = 0 for some n \geq 1,

(4.12)

and $rad R$ contains every nilpotent left ideal and every nilpotent right ideal of $R$ .

Learning Objectives

State $(4.12)$ with the hypothesis left artinian, not merely artinian.
Show that the powers of $J$ stabilise and prove that the stable value $I$ satisfies $I^{2} = I$ .
Run the minimal-counterexample argument that forces $I = 0$ .
Deduce $(4.13)$ : nil one-sided ideals of a left artinian ring are nilpotent.
Prove that a left artinian ring is semiprimary.
Compute the nilpotency index of $rad R$ for $T_{n} (k)$ , for $ℤ / 72 ℤ$ and for $𝔽_{p} C_{p}$ .

Definitions

Left artinian: Every descending chain of left ideals $𝔄_{1} \supseteq 𝔄_{2} \supseteq \dots$ stabilises; equivalently every nonempty family of left ideals has a minimal member.
Semiprimary: $rad R$ is nilpotent and $R / rad R$ is semisimple. Every left artinian ring is semiprimary; the converse fails.
$J^{n}$: The additive group generated by all products of $n$ elements of $J$ . Since $J$ is a two-sided ideal, each $J^{n}$ is a two-sided ideal.
Wedderburn radical: The largest nilpotent ideal, when it exists. It exists for left artinian rings and equals $rad R$ there; for general rings it need not exist at all.
Radical filtration: The chain $M \supseteq J M \supseteq J^{2} M \supseteq \dots$ for a module $M$ ; finite exactly when $J$ is nilpotent.

Left artinian is not symmetric: there are right artinian rings that are not left artinian, so both conclusions of (4.12) are statements about a ring assumed artinian on the left only.

Core Concepts

Two uses of the chain condition

The DCC is invoked twice and for different purposes. First on the chain of powers $J \supseteq J^{2} \supseteq \dots$ , to produce an idempotent ideal $I$ with $I^{2} = I$ . Second on the family of left ideals $𝔄$ with $I 𝔄 \neq 0$ , to produce a minimal such $𝔄$ . Neither use can be replaced by the ACC.

StabiliseDCC gives

k

with

I : = J^{k} = J^{k + 1} = \dots

, so

I^{2} = I

Suppose $I e q 0$ Then the family

{𝔄 : I 𝔄 \neq 0}

is nonempty, since

I \cdot I = I \neq 0

MinimiseDCC gives a minimal member

𝔄_{0}

; pick

a \in 𝔄_{0}

with

I a \neq 0

Collapse

I (I a) = I^{2} a = I a \neq 0

and

I a \subseteq 𝔄_{0}

force

I a = 𝔄_{0} ∋ a

, so

a = y a

with

y \in I

Contradict

(1 - y) a = 0

with

1 - y

a unit gives

a = 0

, contradicting

I a \neq 0

What nilpotency buys

A nilpotent radical turns $R$ into a finite tower $R \supseteq J \supseteq J^{2} \supseteq \dots \supseteq J^{n} = 0$ whose successive quotients are modules over the semisimple ring $R / J$ . Every structural theorem for artinian rings is proved by inducting up this tower: Hopkins–Levitzki, the existence of composition series, idempotent lifting, and the block decomposition all follow this pattern.

Key Results

Theorem(4.12)The radical of a left artinian ring

Let $R$ be a left artinian ring. Then $rad R$ is nilpotent. Moreover $rad R$ is the largest nilpotent left ideal of $R$ and also the largest nilpotent right ideal of $R$ .

Proof

Write $J = rad R$ . Applying the DCC to $J \supseteq J^{2} \supseteq J^{3} \supseteq \dots$ gives an integer $k$ with $I : = J^{k} = J^{k + 1} = \dots$ ; in particular $I^{2} = J^{2 k} = J^{k} = I$ .

Suppose $I \neq 0$ . Then $I \cdot I = I \neq 0$ , so the family $ℱ = {𝔄 : 𝔄 a left ideal of R, I 𝔄 \neq 0}$ is nonempty. By the DCC choose $𝔄_{0}$ minimal in $ℱ$ , and fix $a \in 𝔄_{0}$ with $I a \neq 0$ .

Now $I a$ is a left ideal contained in $𝔄_{0}$ , and $I (I a) = I^{2} a = I a \neq 0$ , so $I a \in ℱ$ . Minimality forces $I a = 𝔄_{0}$ . Since $a \in 𝔄_{0} = I a$ , there is $y \in I \subseteq rad R$ with $a = y a$ , that is $(1 - y) a = 0$ . But $1 - y$ is a unit by the maximality property of the radical, so $a = 0$ — contradicting $I a \neq 0$ .

Hence $I = J^{k} = 0$ and $J$ is nilpotent. It is therefore a nilpotent left ideal, and every nilpotent left ideal — indeed every nil one-sided ideal — is contained in $rad R$ by $(4.11)$ . So $J$ is the largest nilpotent left ideal, and by the same containment applied on the other side, the largest nilpotent right ideal.

Corollary(4.13)Nil implies nilpotent

In a left artinian ring $R$ , every nil left ideal and every nil right ideal is nilpotent.

Proof

Let $𝔄$ be a nil one-sided ideal. By $(4.11)$ , $𝔄 \subseteq rad R$ , and by $(4.12)$ there is $n$ with $(rad R)^{n} = 0$ . Then $𝔄^{n} \subseteq (rad R)^{n} = 0$ .

Corollary—Left artinian rings are semiprimary

If $R$ is left artinian then $rad R$ is nilpotent and $R / rad R$ is semisimple.

Proof

Nilpotency is $(4.12)$ . For the quotient: $R / rad R$ is left artinian, being a quotient of a left artinian ring, and it is semiprimitive because $rad (R / rad R) = 0$ . A left artinian semiprimitive ring is semisimple, which is the equivalence $(4.14)$ proved on The Hopkins–Levitzki Theorem.

Proposition—A dimension bound on the index

Let $A$ be a finite-dimensional algebra over a field $k$ with $\dim_{k} A = n \geq 1$ and $J = rad A$ . Then $J^{n} = 0$ ; the nilpotency index of $J$ is at most $\dim_{k} A$ .

Proof

If $J^{i} = J^{i + 1}$ for some $i \geq 1$ , then $J \cdot J^{i} = J^{i}$ with $J^{i}$ finitely generated as a module, so Nakayama's Lemma gives $J^{i} = 0$ . Hence the chain $J \supseteq J^{2} \supseteq \dots$ is strictly decreasing until it hits $0$ , and each step drops the $k$ -dimension by at least one. Since $1 \notin J$ we have $\dim_{k} J \leq n - 1$ , so at most $n - 1$ strict drops are available and $J^{n} = 0$ .

Remark—The converse fails in both directions worth noting

A ring with nilpotent radical need not be left artinian: $ℤ$ has zero radical, which is nilpotent, and is not artinian. Even semiprimary is not enough — there exist semiprimary rings that are neither left nor right artinian, and neither left nor right noetherian; Lam constructs one in Exercise 20.5.

Proof Techniques and Method

How the proof works and what to reuse elsewhere.

Move 1

Stabilise, then kill the idempotent ideal

Under a DCC, any descending chain of ideals stabilises at an idempotent ideal $I = I^{2}$ . Showing $I = 0$ is the recurring task; inside the radical it is done by the unit trick. This is the template for T-nilpotency arguments too.

Move 2

Minimal counterexample on the annihilating family

Rather than argue about $I$ directly, minimise over the left ideals it fails to annihilate. Minimality then upgrades a containment $I a \subseteq 𝔄_{0}$ to an equality, and equality is what produces the equation $a = y a$ .

Move 3

Nakayama in place of the DCC

For finite-dimensional algebras the same conclusion follows from Nakayama plus a dimension count, giving the explicit bound $J^{\dim A} = 0$ that the DCC argument does not supply.

The two arguments answer different questions. The DCC proof shows nilpotency with no bound; the Nakayama proof gives a bound but needs finite dimension. In practice the second is what a computation uses.

Worked Example

Triangular matrices: index exactly n

Let $k$ be a division ring and $R = T_{n} (k)$ . Then $R$ is left artinian, being finite-dimensional over $k$ on the left, and $rad R = J$ is the ideal of strictly upper triangular matrices.

J^{m} = {(a_{i j}) \in T_{n} (k) : a_{i j} = 0 whenever j - i < m}, J^{n} = 0, J^{n - 1} = k e_{1 n} \neq 0 .

(E.1)

Each multiplication by $J$ pushes the surviving band one step further from the diagonal.

So the nilpotency index is exactly $n$ , while $\dim_{k} R = n (n + 1) / 2$ . The dimension bound of the previous section is satisfied with plenty of room, which is typical: the index is usually far below the dimension.

A finite ring

Take $R = ℤ / 72 ℤ$ , which is finite and hence artinian. Its maximal ideals are $(2)$ and $(3)$ , so $rad R = (2) \cap (3) = (6)$ . Powers: $6^{2} = 36 \neq 0$ in $ℤ / 72 ℤ$ , and $6^{3} = 216 = 3 \cdot 72 = 0$ . The index is exactly $3$ , and $R / rad R ≅ ℤ / 6 ℤ ≅ 𝔽_{2} \times 𝔽_{3}$ is semisimple, as $(4.12)$ and the semiprimary corollary require.

A modular group algebra

Let $p$ be prime, $G = C_{p}$ the cyclic group of order $p$ , and $k = 𝔽_{p}$ . Writing $G = ⟨ g ⟩$ and $u = g - 1$ ,

k G ≅ k [t] / (t^{p} - 1) = k [u] / (u^{p}), u = t - 1,

(E.2)

using $t^{p} - 1 = (t - 1)^{p}$ in characteristic $p$ .

So $k G$ is local with $rad (k G) = (u)$ , nilpotent of index exactly $p$ , and $k G / rad (k G) ≅ k$ . This is Maschke's theorem failing as loudly as possible: the group algebra is as far from semisimple as a $p$ -dimensional algebra can be, and its Loewy length equals its dimension.

Process and Workflow

You need to know whether $rad R$ is nilpotent.

R is left artinianYes, by

(4.12)

, and it is the largest nilpotent one-sided ideal on either side.

R is left perfectOnly left T-nilpotent in general, which is strictly weaker than nilpotent; see T-Nilpotency.

R is semilocalNo conclusion.

R / rad R

is semisimple but the radical can be as badly behaved as in

k [[x]]

R is a finite-dimensional algebraYes, with the explicit bound

J^{\dim_{k} R} = 0

from Nakayama.

Compute or guess $J$ Use the trace form in characteristic zero, or read off a nilpotent ideal from a triangular or graded presentation.

Confirm nilpotencyForm

J^{2}, J^{4}, \dots

by repeated squaring until the product vanishes or the dimension stops dropping.

Check the quotientVerify that

R / J

is semisimple; combined with

J

nilpotent this certifies

J = rad R

FilterUse

R \supseteq J \supseteq \dots \supseteq J^{n} = 0

to reduce module questions to questions over the semisimple ring

R / J

Comparison and Classification

How the radical behaves across chain conditions
Hypothesis on $R$	$rad R$ is…	Witness or reference
left artinian	nilpotent	$(4.12)$ ; $T_{n} (k)$ has index $n$
semiprimary	nilpotent by definition	Lam Exercise 20.5 gives a non-artinian example
left perfect	left T-nilpotent, not always nilpotent	see T-Nilpotency
semilocal	unrestricted	$k [[x]]$ is local with non-nil radical
left noetherian	unrestricted	$ℤ_{(p)}$ is noetherian local with non-nil radical
algebraic algebra over $k$	the largest nil ideal, not always nilpotent	$k [x_{1}, x_{2}, \dots] / (x_{i}^{2})$
arbitrary	only a quasi-regular ideal	$(4.1)$ and $(4.5)$

Standard rings against the properties in play
	Left artinian	$rad R$ nilpotent	$rad R$ nil	Semiprimary
$T_{n} (k)$	yes	yes	yes	yes
$ℤ / 72 ℤ$	yes	yes	yes	yes
$𝔽_{p} C_{p}$	yes	yes	yes	yes
$M_{n} (D)$	yes	yes	yes	yes
$ℤ$	no	yes	yes	no
$k [[x]]$	no	no	no	no
$ℤ_{(p)}$	no	no	no	no
$k [x_{1}, x_{2}, \dots] / (x_{i}^{2})$	no	no	yes	no

Standard rings against the properties in play

Relationship Map

Left artinianSemiprimaryLeft perfectSemilocal

Each implication is strict. Semiprimary but not artinian is Lam's Exercise 20.5; left perfect but not semiprimary happens as soon as the radical is T-nilpotent without being nilpotent; semilocal but not perfect is $ℤ_{(p)}$ .

Semilocal

R / rad R

semisimple

Left perfect…and

rad R

left T-nilpotent

Semiprimary…and

rad R

nilpotent

Left artinian…and DCC on left ideals — this page

Semisimple…and

rad R = 0

Reading inwards, each band adds one finiteness condition on the radical. $(4.12)$ is the statement that the innermost-but-one band contains the left artinian rings, which is what licenses every filtration argument in artinian ring theory.

Applications and Industry Use

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

Modular representation theory

Loewy series of group algebras

For $k G$ with $char k = p$ dividing $| G |$ the radical is nilpotent and its powers give the Loewy filtration of every module. The nilpotency index of $rad (k G)$ is a genuine invariant of the group and the characteristic.

Coding theory

Finite chain rings

A finite chain ring is artinian and local, so its radical $(π)$ is nilpotent; codes over $ℤ / p^{n}$ and over Galois rings are graded by the finitely many powers of $(π)$ , and decoding algorithms lift solutions layer by layer.

Symbolic computation

Structure of finite-dimensional algebras

Wedderburn decomposition routines rely on the radical being nilpotent so that idempotents can be lifted from $A / rad A$ through the finite tower of powers.

Numerical and control settings

Nilpotent parts of finite-dimensional operator algebras

The algebra generated by a single nilpotent operator is artinian with nilpotent radical, and its Loewy structure is exactly the Jordan block structure — the linear-algebra shadow of $(4.12)$ .

Computational Notes

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

For a finite-dimensional algebra $A$ over a field, once $rad A$ is known as a subspace, its nilpotency index is found by computing $J, J^{2}, J^{3}, \dots$ as subspaces until the zero space appears — at most $\dim_{k} A$ products, each a spanning-set multiplication followed by a rank computation.
Repeated squaring finds a vanishing power in about $\log_{2} \dim_{k} A$ products, but returns only a power of two above the true index; a final linear search recovers the exact index.
The Loewy layers $J^{i} M / J^{i + 1} M$ of a module are the standard output of a decomposition routine; their number is the Loewy length, bounded by the nilpotency index of the radical.
GAP's RadicalOfAlgebra and Magma's JacobsonRadical return the radical of a finite-dimensional algebra; nilpotency is then a certificate one can verify cheaply, which is why implementations use the verification route rather than trusting the general algorithm blindly.

Failure Modes and Common Mistakes

Do not expect the nilpotency index to be computable from $\dim_{k} A$ alone; the dimension only bounds it, and the bound is rarely attained.
Do not use $(4.13)$ outside a chain condition. The general statement nil implies nilpotent is false and its failures are the content of the Koethe circle of problems.
Do not conclude artinian from semiprimary; the implication runs one way only.

Quick Reference

HypothesisDCC on left ideals

Conclusion

rad R

nilpotent

Maximalitylargest nilpotent left ideal and largest nilpotent right ideal

(4.13)nil one-sided ideals are nilpotent

Semiprimaryleft artinian

\Rightarrow

radical nilpotent and

R / rad R

semisimple

Index bound

J^{\dim_{k} A} = 0

for a finite-dimensional algebra

Standard indices

T_{n} (k)

n

;

ℤ / p^{n} ℤ

n

;

𝔽_{p} C_{p}

p

Failure without DCC

rad k [[x]] = (x)

, not nil

Worked radicals and their indices
$R$	$rad R$	Index	$R / rad R$
$T_{n} (k)$ , $k$ a division ring	strictly upper triangular	$n$	$k \times \dots \times k$
$ℤ / 72 ℤ$	$(6)$	$3$	$𝔽_{2} \times 𝔽_{3}$
$ℤ / p^{n} ℤ$	$(p)$	$n$	$𝔽_{p}$
$𝔽_{p} C_{p}$	$(g - 1)$	$p$	$𝔽_{p}$
$M_{n} (D)$	$0$	$1$	$M_{n} (D)$
$k [x] / (x^{m})$	$(x)$	$m$	$k$

Frequently Asked Questions

Does

(4.12)

need artinian on both sides?

No. Left DCC alone gives that $rad R$ is nilpotent, and nilpotency is a two-sided condition, so the radical is then simultaneously the largest nilpotent left ideal and the largest nilpotent right ideal. This matters because there are rings artinian on one side only.

Why is the sum of all nilpotent ideals nilpotent here but not in general?

Because in a left artinian ring that sum is contained in $rad R$ , which is itself nilpotent. Without a chain condition the sum can be an infinite ascending union of nilpotent ideals with no common bound on the exponent, and it is then nil at best.

How large can the nilpotency index be relative to the dimension?

For a $k$ -algebra of dimension $n$ the index is at most $n$ , and the bound is attained: $k [x] / (x^{n})$ has index exactly $n$ . For $T_{n} (k)$ the dimension is $n (n + 1) / 2$ while the index is only $n$ , so the bound is far from tight in general.

Is the converse of

(4.13)

interesting?

The converse — nilpotent implies nil — is trivially true in any ring. What is not available in general is $(4.13)$ itself; it is the first of a family of nil implies nilpotent theorems that all require a finiteness hypothesis, such as Levitzki's theorem for nil one-sided ideals of left noetherian rings.

What replaces

(4.12)

for perfect rings?

Left T-nilpotency: for a left perfect ring, $rad R$ need not be nilpotent, but every sequence $a_{1}, a_{2}, \dots$ of radical elements has some product $a_{n} \dots a_{2} a_{1}$ equal to zero. That condition is exactly what keeps Nakayama-style arguments alive without a uniform exponent.

Does the theorem say anything about modules?

Indirectly, and decisively. A nilpotent radical gives every module a finite radical filtration $M \supseteq J M \supseteq \dots \supseteq J^{n} M = 0$ with semisimple factors over $R / rad R$ ; the Hopkins–Levitzki theorem is precisely the exploitation of that filtration.

References

T. Y. Lam, A First Course in Noncommutative Rings, Graduate Texts in Mathematics 131, Springer-Verlag, 1991, §4, (4.12)–(4.13) (pp. 56–58), with Exercise 20.5.
N. Jacobson, Structure of Rings, American Mathematical Society Colloquium Publications 37, revised edition, 1964, Chapter I.
F. W. Anderson and K. R. Fuller, Rings and Categories of Modules, 2nd edition, Graduate Texts in Mathematics 13, Springer-Verlag, 1992, §15.
C. W. Curtis and I. Reiner, Representation Theory of Finite Groups and Associative Algebras, Wiley-Interscience, 1962, Chapter V.
L. H. Rowen, Ring Theory, Volume I, Academic Press, 1988, Chapter 2.

AI Suggested Questions

Give a complete proof that a left artinian ring is left noetherian, using $(4.12)$ as the input.
Compute the nilpotency index of $rad (k G)$ for $G$ an elementary abelian $p$ -group of rank $r$ over $𝔽_{p}$ .
Construct a semiprimary ring that is neither left nor right artinian.
How does the nilpotency index of the radical relate to the Loewy length of the regular module?
What is the analogue of $(4.12)$ for rings satisfying the DCC on principal left ideals?
Show that the radical of a finite ring is nilpotent and bound the index in terms of the cardinality.

Radical of Artinian Rings

Executive Summary

Overview

Learning Objectives

Definitions

Core Concepts

Two uses of the chain condition

What nilpotency buys

Key Results

Proof Techniques and Method

Stabilise, then kill the idempotent ideal

Minimal counterexample on the annihilating family

Nakayama in place of the DCC

Worked Example

Triangular matrices: index exactly n

A finite ring

A modular group algebra

Process and Workflow

Comparison and Classification

Relationship Map

Applications and Industry Use

Loewy series of group algebras

Finite chain rings

Structure of finite-dimensional algebras

Nilpotent parts of finite-dimensional operator algebras

Computational Notes

Failure Modes and Common Mistakes

Quick Reference

Frequently Asked Questions

References

AI Suggested Questions

Continue learning

01Executive Summary

02Overview

03Learning Objectives

04Definitions

05Core Concepts

Two uses of the chain condition

What nilpotency buys

06Key Results

07Proof Techniques and Method

Stabilise, then kill the idempotent ideal

Minimal counterexample on the annihilating family

Nakayama in place of the DCC

08Worked Example

Triangular matrices: index exactly n

A finite ring

A modular group algebra

09Process and Workflow

10Comparison and Classification

11Relationship Map

12Applications and Industry Use

Loewy series of group algebras

Finite chain rings

Structure of finite-dimensional algebras

Nilpotent parts of finite-dimensional operator algebras

13Computational Notes

14Failure Modes and Common Mistakes

15Quick Reference

16Frequently Asked Questions

17Related KEVOS Topics

18References

19AI Suggested Questions

Continue learning

Executive Summary

Overview

Learning Objectives

Definitions

Core Concepts

Key Results

Proof Techniques and Method

Worked Example

Process and Workflow

Comparison and Classification

Relationship Map

Applications and Industry Use

Computational Notes

Failure Modes and Common Mistakes

Quick Reference

Frequently Asked Questions

Related KEVOS Topics

References

AI Suggested Questions