Semiprime and Semiprimitive Rings as Subdirect Products

Q: How is (12.5) different from Birkhoff's theorem?

Birkhoff (12.3) applies to every nonzero ring but delivers subdirectly irreducible factors, which may be as complicated as R. (12.5) assumes a radical vanishes and delivers factors from a class with a real structure theory. The trade is generality for usable factors, and it is the trade that makes the reduction technique of (12.8) work.

Q: Can I always take finitely many factors?

No. Z is semiprimitive and every left primitive quotient is some F_p, so any decomposition into primitive factors needs infinitely many. Finiteness of the family is essentially the artinian case: a semiprimitive left artinian ring is semisimple and decomposes into finitely many simple artinian factors.

Executive Summary

Two radicals are defined as intersections: the lower nilradical ${Nil}_{*} R$ is the intersection of all prime ideals, and $rad R$ is the intersection of all left primitive ideals. Saying a radical vanishes is therefore the same as saying a family of ideals meets in zero — which is exactly the input a subdirect representation needs.

Lam's $(12.5)$ makes that observation into a two-part characterisation: semiprime = subdirect product of prime rings, semiprimitive = subdirect product of left primitive rings. The proofs are short. The value lies in what follows: any property inherited by subrings and quotients need only be checked on prime or primitive rings, where the structure theory is strong.

${Nil}_{*} R = 0$ Semiprime

$rad R = 0$ Semiprimitive

PrimeFactors in part (a)

Left primitiveFactors in part (b)

Overview

A subdirect representation of $R$ is the same thing as a family of ideals ${𝔄_{i}}$ with $⋂_{i} 𝔄_{i} = 0$ ; the factors are the quotients $R / 𝔄_{i}$ . To manufacture a useful representation one therefore wants a family of ideals that (i) meets in zero and (ii) has recognisable quotients. Prime ideals and left primitive ideals are the two standard supplies.

{Nil}_{*} R = ⋂_{𝔭 prime} 𝔭, rad R = ⋂_{𝔭 left primitive} 𝔭

(10.13), (11.5)

Both radicals are intersections of ideals with well-understood quotients. Vanishing of the radical is precisely the injectivity condition for the corresponding subdirect map.

Because every left primitive ideal is prime, and ${Nil}_{*} R \subseteq rad R$ , the semiprimitive statement is the sharper of the two: it produces better factors under a stronger hypothesis. Neither statement needs chain conditions, finiteness, or commutativity.

The parallel statement for reduced rings — with domains as factors — is genuinely different and needs a lemma about minimal primes; it is treated on Reduced Rings as Subdirect Products of Domains. The atoms of the general theory are on Subdirectly Irreducible Rings.

Learning Objectives

State both halves of $(12.5)$ with full hypotheses.
Prove each direction from the two radical identities.
Show that the minimal prime ideals already suffice for the semiprime half.
Explain why the semiprimitive half is insensitive to the left-right choice.
Decide, for a given ring, which of the four decompositions in this section applies.
Exhibit a ring admitting no subdirect decomposition into prime factors.

Definitions

Prime ring: $R \neq 0$ and for ideals $𝔄, 𝔅$ , $𝔄 𝔅 = 0 \Rightarrow 𝔄 = 0$ or $𝔅 = 0$ . Elementwise: $a R b = 0 \Rightarrow a = 0$ or $b = 0$ .
Semiprime ring: No nonzero nilpotent ideal; elementwise $a R a = 0 \Rightarrow a = 0$ ; equivalently ${Nil}_{*} R = 0$ .
Left primitive ring: $R$ has a faithful simple left module; equivalently $0$ is a left primitive ideal.
Semiprimitive ring: $rad R = 0$ , where $rad R$ is the Jacobson radical, also written $J (R)$ .
${Nil}_{*} R$: The lower nilradical (Baer radical): the smallest semiprime ideal, equal to the intersection of all prime ideals of $R$ .

Ideal means two-sided ideal. Rings are associative with identity and nonzero unless said otherwise.

Key Results

Theorem(12.5)Subdirect decomposition of semiprime and semiprimitive rings

Let $R$ be a nonzero ring.

$R$ is semiprime iff $R$ can be represented as a subdirect product of prime rings.
$R$ is semiprimitive iff $R$ can be represented as a subdirect product of left primitive rings.

In both cases the factors may be taken to be quotients of $R$ : the prime rings $R / 𝔭$ with $𝔭$ prime, respectively the left primitive rings $R / 𝔭$ with $𝔭$ left primitive.

Proof

**( $\Rightarrow$ ), part (a).** Let ${𝔭_{i}}$ be the family of all prime ideals of $R$ . Since $R \neq 0$ it has a maximal ideal, which is prime, so the family is nonempty. Semiprimeness says ${Nil}_{*} R = 0$ , and ${Nil}_{*} R = ⋂_{i} 𝔭_{i}$ , so the quotient maps $R ↠ R / 𝔭_{i}$ have kernels intersecting in zero. That is exactly a subdirect representation $R ↪ \prod_{i} R / 𝔭_{i}$ , and each $R / 𝔭_{i}$ is prime by definition of a prime ideal.

**( $\Rightarrow$ ), part (b).** Identical with ${𝔭_{i}}$ the family of left primitive ideals and $rad R = ⋂_{i} 𝔭_{i}$ . Each $R / 𝔭_{i}$ is left primitive because $𝔭_{i}$ is the annihilator of a simple left $R$ -module, which becomes a faithful simple $R / 𝔭_{i}$ -module.

**( $\Leftarrow$ ), both parts.** Let $ε : R ↪ \prod_{i} R_{i}$ be a subdirect representation and set $𝔄_{i} = \ker (R \to R_{i})$ , so that $R / 𝔄_{i} ≅ R_{i}$ and $⋂_{i} 𝔄_{i} = 0$ . If every $R_{i}$ is prime then every $𝔄_{i}$ is a prime ideal, so ${Nil}_{*} R \subseteq ⋂_{i} 𝔄_{i} = 0$ and $R$ is semiprime. If every $R_{i}$ is left primitive then every $𝔄_{i}$ is a left primitive ideal, so $rad R \subseteq ⋂_{i} 𝔄_{i} = 0$ and $R$ is semiprimitive.

Corollary(12.5)′Minimal primes suffice

Let $R$ be a nonzero semiprime ring and let ${𝔭_{i}}$ be the set of minimal prime ideals of $R$ . Then $⋂_{i} 𝔭_{i} = 0$ and $R ↪ \prod_{i} R / 𝔭_{i}$ is a subdirect representation by prime rings.

Proof

Every prime ideal contains a minimal prime: order the primes inside a given prime $𝔭$ by reverse inclusion and apply Zorn's Lemma, the point being that the intersection of a descending chain of prime ideals is again prime. Hence the intersection of the minimal primes equals the intersection of all primes, which is ${Nil}_{*} R = 0$ .

Corollary(12.5)″Embedding consequences

A nonzero semiprime ring embeds in a direct product of prime rings. A nonzero semiprimitive ring embeds in a direct product of left primitive rings, and hence — applying the Jacobson Density Theorem to each factor — in a direct product of rings of linear transformations, each dense in $End (M_{i})$ for a right vector space $M_{i}$ over a division ring $D_{i}$ .

RemarkLeft, right, and the asymmetry that is not there

Left primitivity is genuinely one-sided: there exist left primitive rings that are not right primitive. Yet $(12.5)$ (b) is side-neutral, because the radical is: $rad R$ is simultaneously the intersection of the left primitive ideals and of the right primitive ideals. So a semiprimitive ring is a subdirect product of left primitive rings and, by a different family of ideals, also of right primitive rings.

Proof Techniques and Method

How this proof works, and where the effort really lies.

The proof of $(12.5)$ occupies a few lines because all the work was done earlier, in establishing the two radical identities. Isolating the pattern makes it reusable.

Pick a class of quotientsChoose the kind of factor you want: prime, primitive, domain, simple.

Name the matching idealsPrime ideals, left primitive ideals, completely prime ideals. The class of factors determines the class of ideals.

Identify the intersectionThat intersection is a radical:

{Nil}_{*} R

rad R

, or

{Nil}_{*} R

again in the reduced case.

Assume the radical vanishesThe hypothesis of the theorem is exactly the statement that the intersection is zero.

Read off the embeddingZero intersection of kernels equals injectivity; surjectivity of coordinate maps is free for quotient maps.

The same three-step template produces the reduced case with completely prime ideals, and it is what fails for arbitrary rings: there is no radical whose vanishing characterises being a subdirect product of simple rings, because simplicity is not detected by an intersection of ideals of a fixed type.

Worked Example

The integers: both halves at once

$ℤ$ is a domain, hence prime, hence semiprime; and $rad ℤ = ⋂_{p} (p) = 0$ , so it is semiprimitive. The two decompositions are quite different.

Prime factors. The minimal prime of $ℤ$ is $(0)$ , so $(12.5)'$ gives the one-factor representation $ℤ ≅ ℤ$ — trivial, as it must be for a prime ring.
Primitive factors. A commutative ring is left primitive iff it is a field, so the left primitive ideals of $ℤ$ are exactly the $(p)$ , and the representation is $ℤ ↪ \prod_{p} 𝔽_{p}$ , which is genuinely nontrivial.

A group ring: $ℤ C_{2}$

Let $G = C_{2} = ⟨ g ⟩$ and $R = ℤ G$ . The two maps $g \mapsto 1$ and $g \mapsto - 1$ give surjections $π_{\pm} : R \to ℤ$ with kernels $(g - 1)$ and $(g + 1)$ , both nonzero. If $a + b g$ dies under both then $a + b = 0$ and $a - b = 0$ , so $2 a = 0$ and $a = b = 0$ .

ℤ C_{2} ↪ ℤ \times ℤ, a + b g \mapsto (a + b, a - b)

(E.1)

A subdirect representation by two prime rings; the image is ${(u, v) : u \equiv v (\mod 2)}$ , of index $2$ in the product.

So $ℤ C_{2}$ is semiprime, with $(g - 1)$ and $(g + 1)$ as its two minimal primes. It is also semiprimitive: composing $π_{\pm}$ with $ℤ ↠ 𝔽_{p}$ gives maximal — hence left primitive — ideals, and their intersection is $\ker π_{+} \cap \ker π_{-} = 0$ because $⋂_{p} p ℤ = 0$ .

A ring with no prime decomposition

Let $k$ be a field and $R = T_{2} (k)$ the upper triangular $2 \times 2$ matrices. The strictly upper triangular matrices form an ideal $𝔑$ with $𝔑^{2} = 0$ and $𝔑 \neq 0$ , so $R$ is not semiprime.

𝔑 = (\begin{matrix} 0 & k \\ 0 & 0 \end{matrix}), 𝔑^{2} = 0, {Nil}_{*} R = rad R = 𝔑 \neq 0 .

(E.2)

Directly: any homomorphism from $R$ onto a prime ring must kill $𝔑$ , since the image of a nilpotent ideal is a nilpotent ideal and a prime ring has none that is nonzero. So every prime ideal of $R$ contains $𝔑$ , no family of them can meet in zero, and $(12.5)$ correctly refuses to apply. The two prime ideals here are $(\begin{matrix} 0 & k \\ 0 & k \end{matrix})$ and $(\begin{matrix} k & k \\ 0 & 0 \end{matrix})$ , meeting in $𝔑$ , and $R / rad R ≅ k \times k$ .

Process and Workflow

Which subdirect decomposition should I use for a given ring $R \neq 0$ ?

$rad R = 0$ Use

(12.5)

(b): factors are left primitive rings, and the Density Theorem then describes each one as a dense ring of linear transformations over a division ring. This is the strongest routinely available decomposition.

$R$ reducedUse

(12.7)

: factors are domains, obtained from the minimal primes, which are completely prime by

(12.6)

. Better factors than prime rings, at the cost of a stronger hypothesis.

${Nil}_{*} R = 0$ onlyUse

(12.5)

(a): factors are prime rings; take the minimal primes to keep the family small.

None of theseFall back on Birkhoff

(12.3)

: subdirectly irreducible factors, always available, but the factors may be no simpler than

R

. Alternatively pass to

R / rad R

and lift afterwards.

In the third and fourth branches it is often worth computing $rad R$ first: if $rad R \neq 0$ one works with $R / rad R$ , decomposes that, and then asks separately whether the conclusion lifts. That two-stage strategy is the content of the reduction chart $(12.8)$ used on The Jacobson and Herstein Commutativity Theorems.

Comparison and Classification

The four subdirect decompositions of §12
Hypothesis on $R$	Family of ideals	Intersection equals	Factors	Reference
none ( $R \neq 0$ )	$𝔪_{a}$ maximal with $a \notin 𝔪_{a}$	$0$ by construction	subdirectly irreducible rings	(12.3)
semiprime	prime ideals (or just the minimal ones)	${Nil}_{*} R$	prime rings	(12.5)(a)
semiprimitive	left primitive ideals	$rad R$	left primitive rings	(12.5)(b)
reduced	minimal prime ideals	${Nil}_{*} R$	domains	(12.7)

Where familiar rings sit
	Semiprime	Semiprimitive	Prime	Left primitive
$ℤ$	yes	yes	yes	no
$ℤ / 6$	yes	yes	no	no
$ℤ / 8$	no	no	no	no
$k [[x]]$	yes	no	yes	no
$T_{2} (k)$	no	no	no	no
$M_{2} (ℚ)$	yes	yes	yes	yes
Weyl algebra $A_{1} (ℂ)$	yes	yes	yes	yes

Where familiar rings sit

Each column is a property of the ring itself, not of a factor. Reading down the first two columns tells you which half of (12.5) is available.

Relationship Map

The classes involved are strictly nested, and each containment corresponds to trading a hypothesis for better factors.

All nonzero ringsBirkhoff: subdirect product of subdirectly irreducible rings

Semiprime

{Nil}_{*} R = 0

; subdirect product of prime rings

Semiprimitive

rad R = 0

; subdirect product of left primitive rings

Von Neumann regularevery

a

satisfies

a = a x a

; always semiprimitive

Semisimple

rad R = 0

and left artinian; a finite direct product of matrix rings over division rings

**Reduced $\Rightarrow$ semiprime**, strictly: $M_{2} (k)$ is prime but has nonzero nilpotents.
**Semiprimitive $\Rightarrow$ semiprime**, strictly: $k [[x]]$ is a domain with nonzero radical.
**Prime $\Rightarrow$ semiprime** and **left primitive $\Rightarrow$ prime**, so the factors produced by part (b) are automatically legitimate factors for part (a).
**Semisimple $\Rightarrow$ semiprimitive** with only finitely many factors, all of them simple artinian — the finite, fully solved case of $(12.5)$ (b).

Applications and Industry Use

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

Commutativity theorems

Reduction to primitive rings

The Jacobson–Herstein theorem is proved by verifying its hypothesis on division rings, extending to left primitive rings by density, and then to all semiprimitive rings by $(12.5)$ (b). Part (b) is the step that makes the reduction legitimate.

Polynomial identities

PI theory

Kaplansky's theorem on primitive PI-rings becomes a theorem about semiprimitive PI-rings via the same decomposition, because polynomial identities are inherited by quotients and by subrings of products.

Commutative algebra

Minimal primes and irreducible components

For a reduced commutative noetherian ring the minimal primes are finite in number and the subdirect embedding $R ↪ \prod_{i} R / 𝔭_{i}$ is the algebraic form of decomposing a variety into irreducible components.

Symbolic computation

Radical and decomposition pipelines

Computer algebra systems compute the nilradical, split a reduced ring along its minimal primes, and work component by component. GAP, Singular, Macaulay2 and Sage all follow this pattern for commutative input.

The honest summary: $(12.5)$ is a licence, not a construction. It tells you that arguing on prime or primitive factors loses nothing, which is what allows the strong structure theory of §10 and §11 to be applied to rings that satisfy no structural hypothesis beyond a vanishing radical.

Design Considerations

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

Which family of ideals? All primes give an enormous index set; the minimal primes give the smallest family that still works. For commutative noetherian rings the minimal set is finite, which is usually decisive.
Which radical do you actually need to vanish? If the property you are transporting is inherited by subrings and quotients but fails for nilpotents, semiprime is enough. If your argument needs simple modules — density, for instance — you must have $rad R = 0$ .
Quotient first or decompose first? If $rad R \neq 0$ , decompose $R / rad R$ and treat lifting as a separate problem. The lifting step is where results genuinely fail, as $(12.11)$ shows.
Which side? Part (b) can be run with left or with right primitive ideals; choose whichever side matches the modules you intend to use. The resulting factor rings are generally different.
Do not over-decompose. A prime ring is already a one-factor subdirect product of prime rings. Applying $(12.5)$ to it produces nothing, and the redundant factors can obscure the argument.

Computational Notes

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

For a commutative ring finitely presented over a field, the minimal primes are computable by primary decomposition over Gröbner bases; the cost is doubly exponential in the number of variables in the worst case, and the decomposition is the expensive step, not the subdirect embedding.
For a finite-dimensional algebra over a field, $rad A$ is computable in polynomial time — by the trace form in characteristic $0$ , by the Friedl–Rónyai method in characteristic $p$ — so semiprimitivity is decidable and the primitive factors are the Wedderburn components of $A / rad A$ .
For finitely presented noncommutative rings there is no algorithm: the word problem is undecidable, so neither ${Nil}_{*} R$ nor $rad R$ can be computed in general, and $(12.5)$ gives no effective procedure.
The index set in $(12.5)$ is typically infinite even for small rings — $ℤ$ needs one factor per prime — so the theorem is a structural statement rather than a data structure.

Failure Modes and Common Mistakes

Do not assume the decomposition is unique. Different families of prime ideals meeting in zero give different, equally valid representations.
Do not expect the factors to inherit finiteness. A finitely generated semiprimitive ring can have primitive quotients that are infinite-dimensional over their centres.
Do not use the word semisimple for semiprimitive without saying so; the older literature conflates them, and $(12.5)$ is false with the artinian meaning.
Do not forget the nonzero hypothesis: the zero ring is vacuously radical-free but has no prime ideals at all.

Best Practices

State which radical you are assuming to vanish before invoking $(12.5)$ ; the two halves have different strengths and different factors.
Prefer minimal primes to all primes — same theorem, far smaller index set, and in the commutative noetherian case a finite one.
When transporting a property to the factors, check explicitly that it is inherited by quotients and by subrings of products; that is precisely what the argument consumes.
Record whether your conclusion is about $R$ or about $R / rad R$ . Lifting across the radical is a separate theorem, and sometimes a false one.

Quick Reference

Part (a)semiprime

iff

subdirect product of prime rings

Part (b)semiprimitive

iff

subdirect product of left primitive rings

Engine (a)

{Nil}_{*} R = ⋂ {𝔭 : 𝔭 prime}

Engine (b)

rad R = ⋂ {𝔭 : 𝔭 left primitive}

Smaller familyminimal primes already intersect to

{Nil}_{*} R

Nestingsemiprimitive

\Rightarrow

semiprime, strictly

Sides(b) holds with right primitive ideals too

Obstructiona nonzero nilpotent ideal blocks all prime factorisations

Checklist before applying (12.5)
Question	If yes	If no
Is $R \neq 0$ ?	proceed	the theorem does not apply
Is there a nonzero nilpotent ideal?	no prime decomposition exists	$R$ is semiprime; use part (a)
Is $rad R = 0$ ?	use part (b) and then density	decompose $R / rad R$ and lift separately
Is $R$ reduced?	upgrade to domains via $(12.7)$	prime factors are the best available

Frequently Asked Questions

Why is the theorem stated with left primitive rings rather than primitive rings?

Because primitivity is genuinely one-sided — there are left primitive rings that are not right primitive — so the word primitive alone is ambiguous. The statement is nonetheless side-neutral in effect, since $rad R$ is the intersection of the left primitive ideals and also of the right primitive ideals. Running the proof on the other side gives a decomposition by right primitive rings.

How is

(12.5)

different from Birkhoff's theorem?

Birkhoff $(12.3)$ applies to every nonzero ring but delivers subdirectly irreducible factors, which may be as complicated as $R$ . $(12.5)$ assumes a radical vanishes and delivers factors from a class with a real structure theory. The trade is generality for usable factors, and it is the trade that makes the reduction technique of $(12.8)$ work.

Can I always take finitely many factors?

No. $ℤ$ is semiprimitive and every left primitive quotient is some $𝔽_{p}$ , so any decomposition into primitive factors needs infinitely many. Finiteness of the family is essentially the artinian case: a semiprimitive left artinian ring is semisimple and decomposes into finitely many simple artinian factors.

Does the theorem say anything about how big the image is inside the product?

Nothing at all, and that is the main limitation. The image can be vanishingly small — $ℤ$ is countable inside a product of continuum cardinality. Only coordinatewise surjectivity is guaranteed, which is why properties are transported to the factors rather than recovered from them without further argument.

Why do nilpotent ideals block a decomposition into prime factors?

A prime ring has no nonzero nilpotent ideal, so any surjection onto a prime ring kills every nilpotent ideal of the source. Hence every prime ideal contains ${Nil}_{*} R$ , and no family of prime ideals can meet in zero unless ${Nil}_{*} R = 0$ . The failure is structural, not a defect of the chosen family.

Is the analogous statement true with simple rings as factors?

No. Being a subdirect product of simple rings is a much stronger and less natural condition, and it is not characterised by the vanishing of any of the usual radicals. The maximal ideals of $R$ intersect in the Brown–McCoy radical, so vanishing of that radical characterises subdirect products of simple rings — but that radical is not one of the two used in $(12.5)$ , and it is much less well behaved.

References

T. Y. Lam, A First Course in Noncommutative Rings, Graduate Texts in Mathematics 131, Springer-Verlag, 1991 — §12, theorem (12.5), pp. 206–207; §10 (definition 10.13) for the lower nilradical and §11 (corollary 11.5) for primitive ideals.
N. Jacobson, Structure of Rings, American Mathematical Society Colloquium Publications 37, revised edition 1964 — the radical as an intersection of primitive ideals.
I. N. Herstein, Noncommutative Rings, Carus Mathematical Monographs 15, Mathematical Association of America, 1968 — semiprime rings and reduction arguments.
L. H. Rowen, Ring Theory, Volume I, Academic Press, 1988 — prime and semiprime radicals, minimal primes.
N. Divinsky, Rings and Radicals, University of Toronto Press, 1965 — comparison of the classical radicals and their subdirect interpretations.

AI Suggested Questions

Write out the subdirect decomposition of $ℤ / n$ into primitive factors for a general $n$ .
Show directly that the Brown–McCoy radical characterises subdirect products of simple rings.
Give an example of a semiprime ring with infinitely many minimal primes and describe the resulting embedding.
How does $(12.5)$ combine with the Jacobson Density Theorem to describe an arbitrary semiprimitive ring?
Does a semiprime ring with finitely many minimal primes decompose as a direct product? Under what extra hypotheses?
Compare the size of the Birkhoff family of factors with the family of minimal primes for a commutative noetherian ring.
What is the analogue of $(12.5)$ for rings without identity, where maximal ideals may not exist?

Subdirect Decomposition Theorems

Executive Summary

Overview

Learning Objectives

Definitions

Key Results

Proof Techniques and Method

Worked Example

The integers: both halves at once

A group ring: $ℤ C_{2}$

A ring with no prime decomposition

Process and Workflow

Comparison and Classification

Relationship Map

Applications and Industry Use

Reduction to primitive rings

PI theory

Minimal primes and irreducible components

Radical and decomposition pipelines

Design Considerations

Computational Notes

Failure Modes and Common Mistakes

Best Practices

Quick Reference

Frequently Asked Questions

References

AI Suggested Questions

Continue learning

01Executive Summary

02Overview

03Learning Objectives

04Definitions

05Key Results

06Proof Techniques and Method

07Worked Example

The integers: both halves at once

A group ring: ℤC2

A ring with no prime decomposition

08Process and Workflow

09Comparison and Classification

10Relationship Map

11Applications and Industry Use

Reduction to primitive rings

PI theory

Minimal primes and irreducible components

Radical and decomposition pipelines

12Design Considerations

13Computational Notes

14Failure Modes and Common Mistakes

15Best Practices

16Quick Reference

17Frequently Asked Questions

18Related KEVOS Topics

19References

20AI Suggested Questions

Continue learning

Executive Summary

Overview

Learning Objectives

Definitions

Key Results

Proof Techniques and Method

Worked Example

A group ring: $ℤ C_{2}$

Process and Workflow

Comparison and Classification

Relationship Map

Applications and Industry Use

Design Considerations

Computational Notes

Failure Modes and Common Mistakes

Best Practices

Quick Reference

Frequently Asked Questions

Related KEVOS Topics

References

AI Suggested Questions