Projective Covers over Semiperfect Rings

Q: Why must the idempotents be local?

Locality of e is exactly what makes eR/eJ simple, so that the decomposition of M/MJ into simples can be matched summand by summand with projectives eR. A decomposition of 1 into non-local orthogonal idempotents gives projectives whose tops are not simple, and the matching fails. (23.6) says semiperfectness is precisely the availability of such a decomposition.

Q: Is the projective cover of M finitely generated when M is?

Yes, over a semiperfect ring. M/MJ is then a finitely generated semisimple module, hence a finite direct sum of simples, so the index set I is finite and P = _ ∈ I e_ R is finitely generated. This also shows the cover is finitely presented whenever R is right Noetherian.

Q: Where exactly does the proof break for infinitely generated modules over a semiperfect ring that is not perfect?

At the two upgrades. The lifted map only satisfies (P) + MJ = M and ⊆ PJ; both become the desired statements only if MJ and PJ are small in M and in P. Without finite generation, that smallness is equivalent to right T-nilpotence of J by (23.16), which is exactly what perfectness adds.

Q: How is (24.12) used to lift idempotents?

In the proof of the converse theorem (24.16): given covers of uR and of (1-u)R, their direct sum is a cover of R, and so is R → R. Uniqueness (24.10) produces an isomorphism P ⊕ Q ≅ R, and the corresponding decomposition 1 = e + f into orthogonal idempotents reduces to u and 1 - u modulo J.

Q: Can two non-isomorphic modules have isomorphic projective covers?

Yes. The cover depends only on M/MJ, so S_1 and e_1R over T_2(k) both have cover e_1R despite having different dimensions. The cover determines the top of M, not M itself; recovering M requires the kernel as well.

Executive Summary

Projective covers are scarce in general, and this page identifies the two classes of rings that supply them wholesale. If $R$ is semiperfect, every finitely generated right module has a projective cover; if $R$ is right perfect, every right module does, with no finiteness hypothesis at all.

The construction is a single lifting argument. Decompose $M / M J$ over the semisimple ring $R / J$ , cover each simple summand by its principal indecomposable $e_{α} R$ , lift the resulting map through $M ↠ M / M J$ by projectivity, and then verify surjectivity and smallness. Nakayama's Lemma does both verifications in the finitely generated case; right T-nilpotence does them in general.

$⨁_{α} e_{α} R$ Shape of every cover

f.g. $M$ Scope over a semiperfect ring

all $M$ Scope over a right perfect ring

(24.12)Lam's numbering

Overview

Write $J = rad R$ and $\bar{R} = R / J$ . Semiperfectness gives two things at once: $\bar{R}$ is semisimple, so every $\bar{R}$ -module splits into simples; and $1 = e_{1} + \dots + e_{n}$ for orthogonal local idempotents $e_{i}$ , by $(23.6)$ , so those simples are all of the form $\bar{e_{i}} \bar{R} ≅ e_{i} R / e_{i} J$ and each has an explicit projective cover $e_{i} R$ by $(24.11) (2)$ .

\bar{R} = \bar{e_{1}} \bar{R} \oplus \dots \oplus \bar{e_{n}} \bar{R}, \bar{e_{i}} \bar{R} ≅ e_{i} R / e_{i} J simple.

(24.12a)

A complete list of the simple right $R$ -modules, each with a known projective cover.

Given any $M$ , the quotient $M / M J$ is an $\bar{R}$ -module, hence a direct sum $⨁_{α \in I} \bar{e_{α}} \bar{R}$ with each $e_{α}$ drawn from ${e_{1}, \dots, e_{n}}$ . Setting $P = ⨁_{α \in I} e_{α} R$ produces a projective module mapping onto $M / M J$ , and projectivity lifts that map to $θ : P \to M$ .

The corollary $(24.14)$ then runs the argument backwards: a projective module is its own cover's target, so comparing with the constructed cover and invoking uniqueness classifies projectives as direct sums of principal indecomposables.

Learning Objectives

Assemble $P = ⨁_{α} e_{α} R$ from a decomposition of $M / M J$ and lift to $θ : P \to M$ .
Prove the two auxiliary facts $θ (P) + M J = M$ and $\ker θ \subseteq P J$ .
Complete the argument for finitely generated $M$ over a semiperfect ring using Nakayama's Lemma.
Complete the argument for arbitrary $M$ over a right perfect ring using right T-nilpotence.
State the local-ring form $(24.13)$ and identify the rank of the cover as $\dim_{R / J} M / M J$ .
Deduce $(24.14)$ and decompose the projectives of a concrete artinian ring.

Definitions

Semiperfect: $R / rad R$ is semisimple and idempotents lift modulo $rad R$ . Equivalently, by $(23.6)$ , $1$ is a finite sum of orthogonal local idempotents. The condition is left–right symmetric.
Right perfect: $R / rad R$ is semisimple and $rad R$ is right T-nilpotent. Strictly stronger than semiperfect, and not left–right symmetric.
Local idempotent $e$: $e R e$ is a local ring; equivalently $e R$ is indecomposable with $e R / e J$ simple.
$e_{α} R$: A principal indecomposable, or projective indecomposable, module. Over a semiperfect ring these are precisely the indecomposable projective modules with local endomorphism ring $End (e_{α} R) ≅ e_{α} R e_{α}$ .
$M / M J$: The top of $M$ ; a module over the semisimple ring $R / J$ , hence semisimple, and the object the cover is built from.

Modules are right $R$ -modules; $J = rad R$ throughout. Semiperfectness is symmetric, so the statement about finitely generated modules holds equally on the left; right perfectness is not, and the statement about arbitrary modules is genuinely one-sided.

Core Concepts

Why the top determines the cover

A projective cover $θ : P \to M$ satisfies $\ker θ \subseteq rad P = P J$ , so $θ$ induces an isomorphism $P / P J ≅ M / M J$ . The top of the cover is forced to be the top of $M$ — and over a semiperfect ring the top determines $P$ completely, because $P / P J$ is a direct sum of simples and each simple has a unique projective cover $e_{α} R$ .

P / P J ≅ M / M J ⟹ P ≅ ⨁_{α \in I} e_{α} R

(24.12b)

The multiset ${e_{α}}_{α \in I}$ is read off from the simple summands of $M / M J$ .

This is why the construction has no choices in it, and why $(24.10)$ can be invoked afterwards without any compatibility check.

Where the two hypotheses enter

The lifted map $θ$ automatically satisfies two weakened conclusions: $θ (P) + M J = M$ instead of $θ (P) = M$ , and $\ker θ \subseteq P J$ instead of $\ker θ \subseteq_{s} P$ . Upgrading each requires a smallness statement — of $M J$ in $M$ for the first, of $P J$ in $P$ for the second.

What each hypothesis buys
Needed	Finitely generated $M$ , semiperfect $R$	Arbitrary $M$ , right perfect $R$
$M J \subseteq_{s} M$ , giving $θ (P) = M$	Nakayama $(4.22)$ , since $M$ is finitely generated	$(23.16)$ , since $J$ is right T-nilpotent
$P J \subseteq_{s} P$ , giving $\ker θ \subseteq_{s} P$	Nakayama, since $I$ is finite so $P$ is finitely generated	$(23.16)$ again, for the arbitrary direct sum $P$
Index set $I$	Finite, because $M / M J$ is finitely generated semisimple	Arbitrary

Right T-nilpotence is precisely the hypothesis that makes $N J \subseteq_{s} N$ hold for every right module $N$ , which is what removes finiteness from both rows at once.

Key Results

Proposition(24.12)Existence of projective covers

Let $R$ be a semiperfect ring. Then every finitely generated right $R$ -module has a projective cover, and likewise every finitely generated left $R$ -module. If $R$ is right perfect, then every right $R$ -module — finitely generated or not — has a projective cover.

Proof

Put $J = rad R$ and $\bar{R} = R / J$ . By $(23.6)$ write $1 = e_{1} + \dots + e_{n}$ with the $e_{i}$ orthogonal local idempotents; then $\bar{R} = ⨁_{i} \bar{e_{i}} \bar{R}$ and every simple right $\bar{R}$ -module is isomorphic to some $\bar{e_{i}} \bar{R}$ . Note $e_{i} R \cap J = e_{i} J$ , so $\bar{e_{i}} \bar{R} ≅ e_{i} R / e_{i} J$ .

Let $M$ be a right $R$ -module. Since $\bar{R}$ is semisimple, the $\bar{R}$ -module $M / M J$ decomposes as $⨁_{α \in I} \bar{e_{α}} \bar{R}$ for some index set $I$ , each $e_{α}$ being one of $e_{1}, \dots, e_{n}$ . Set

P = ⨁_{α \in I} e_{α} R, P / P J = ⨁_{α \in I} e_{α} R / e_{α} J ≅ M / M J,

(P.1)

$P$ is projective, being a direct sum of direct summands of $R_{R}$ .

Composing $P \to P / P J$ with this isomorphism gives a map $P \to M / M J$ ; since $P$ is projective and $M \to M / M J$ is onto, it lifts to $θ : P \to M$ with $π θ$ equal to that map, where $π : M \to M / M J$ . Two consequences are immediate:

(A) $θ (P) + M J = M$ , because $π θ$ is onto;
(B) $\ker θ \subseteq \ker (π θ) = P J$ .

**Case 1: $M$ finitely generated, $R$ semiperfect.** Then $M / M J$ is a finitely generated semisimple $\bar{R}$ -module, so $I$ may be taken finite and $P$ is finitely generated. By Nakayama's Lemma $(4.22)$ , $M J \subseteq_{s} M$ , so (A) forces $θ (P) = M$ . Again by Nakayama, $P J \subseteq_{s} P$ , so (B) together with $(24.2) (3)$ gives $\ker θ \subseteq_{s} P$ . Hence $θ$ is a projective cover.

**Case 2: $M$ arbitrary, $R$ right perfect.** Now $J$ is right T-nilpotent, so by $(24.2) (2)$ we have $N J \subseteq_{s} N$ for every right $R$ -module $N$ . Applying this to $N = M$ turns (A) into $θ (P) = M$ , and to $N = P$ turns (B) into $\ker θ \subseteq_{s} P$ . Again $θ$ is a projective cover.

The left-module statement in the semiperfect case follows because semiperfectness is left–right symmetric, so the same argument applies to $R^{op}$ .

Remark(24.13)The local case

Suppose $R$ is a local ring, so $\bar{R} = R / J$ is a division ring and $n = 1$ with $e_{1} = 1$ . For $M$ finitely generated, choose $m_{1}, \dots, m_{k} \in M$ whose images form an $\bar{R}$ -basis of the vector space $M / M J$ . Then the map $⨁_{i = 1}^{k} R \to M$ sending the $i$ th standard basis vector to $m_{i}$ is onto with small kernel, hence is a projective cover. The cover is therefore free, of rank $k = \dim_{\bar{R}} M / M J$ .

Corollary(24.14)Structure of projective modules

If $R$ is semiperfect, every finitely generated projective right $R$ -module $P$ is isomorphic to a finite direct sum $⨁_{α} e_{α} R$ of principal indecomposables.
If $R$ is right perfect, every projective right $R$ -module $P$ is isomorphic to a direct sum $⨁_{α \in I} e_{α} R$ , with $I$ of arbitrary cardinality.

Proof

In case (1), $P$ is finitely generated projective, so $P \to P / P J$ is a projective cover by $(24.11) (1)$ . In case (2), $J$ is right T-nilpotent, so $P \to P / P J$ is a projective cover for arbitrary projective $P$ , again by $(24.11) (1)$ .

On the other hand the proof of $(24.12)$ constructs a projective cover $θ : ⨁_{α} e_{α} R \to P / P J$ of the same module. By the uniqueness statement $(24.10)$ , the two covers are isomorphic over $P / P J$ ; in particular $P ≅ ⨁_{α} e_{α} R$ .

Remark—Uniqueness of the decomposition

Each $e_{α} R$ has endomorphism ring ${End}_{R} (e_{α} R) ≅ e_{α} R e_{α}$ , which is local because $e_{α}$ is a local idempotent. By the Krull–Schmidt–Azumaya theorem, a direct sum decomposition into modules with local endomorphism rings is unique up to isomorphism and permutation of the summands. So the multiplicities of $e_{1} R, \dots, e_{n} R$ in $P$ are invariants of $P$ .

Proof Techniques and Method

How these proofs work, and which move to reuse.

Move 1

Solve the problem modulo the radical

Over $R / J$ everything is semisimple, so decomposition is free of charge. Build the answer there, where there is no obstruction, and only then lift.

Move 2

Lift by projectivity, correct by Nakayama

Projectivity produces a map whose reduction is right; Nakayama upgrades *right modulo $M J$ * to right on the nose. The pattern is the module-theoretic analogue of Hensel lifting.

Move 3

Compare two covers of the same module

Uniqueness $(24.10)$ turns any second construction into an isomorphism. $(24.14)$ is nothing but this move applied to $P$ and to the constructed cover of $P / P J$ .

Move 3 is used again in the converse direction in $(24.16)$ : there, comparing the cover $R \to \bar{R}$ with a cover assembled from covers of $\bar{u} \bar{R}$ and $\bar{v} \bar{R}$ produces an isomorphism $P \oplus Q ≅ R$ , and the induced decomposition of $1$ lifts the idempotent $\bar{u}$ .

Worked Example

Upper triangular $2 \times 2$ matrices

Let $k$ be a field and $R = T_{2} (k) = {(\begin{matrix} a & b \\ 0 & c \end{matrix}) : a, b, c \in k}$ . This is finite-dimensional, hence artinian, hence both semiperfect and right (and left) perfect. Take the matrix units $e_{1} = E_{11}$ and $e_{2} = E_{22}$ : they are orthogonal idempotents with $e_{1} + e_{2} = 1$ , and each $e_{i} R e_{i} ≅ k$ is local.

e_{1} R = (\begin{matrix} k & k \\ 0 & 0 \end{matrix}), e_{2} R = (\begin{matrix} 0 & 0 \\ 0 & k \end{matrix}), J = rad R = (\begin{matrix} 0 & k \\ 0 & 0 \end{matrix}) .

(E.1)

Dimensions over $k$ : $\dim e_{1} R = 2$ , $\dim e_{2} R = 1$ , $\dim J = 1$ .

Computing the radicals of the blocks: $e_{1} J = J$ and $e_{2} J = 0$ . Hence the two simple right modules are $S_{1} = e_{1} R / e_{1} J$ and $S_{2} = e_{2} R / e_{2} J = e_{2} R$ , both one-dimensional, and their projective covers are $e_{1} R$ and $e_{2} R$ respectively.

Projective covers over $T_{2} (k)$
Module $M$	$M / M J$	Cover $P$	$\ker θ$	$\dim_{k}$
$S_{1}$	$S_{1}$	$e_{1} R$	$e_{1} J ≅ S_{2}$	$2 \to 1$
$S_{2}$	$S_{2}$	$e_{2} R$	$0$	$1 \to 1$
$e_{1} R$	$S_{1}$	$e_{1} R$	$0$	$2 \to 2$
$R_{R}$	$S_{1} \oplus S_{2}$	$e_{1} R \oplus e_{2} R = R$	$0$	$3 \to 3$
$S_{1} \oplus S_{1}$	$S_{1} \oplus S_{1}$	$e_{1} R \oplus e_{1} R$	$e_{1} J \oplus e_{1} J$	$4 \to 2$

The first row is the interesting one. $e_{1} R$ is uniserial of length $2$ : its unique proper submodule is $e_{1} J$ , on which the ring acts through $c$ , so $e_{1} J ≅ S_{2}$ . Thus $e_{1} R$ has top $S_{1}$ and socle $S_{2}$ , and the cover $e_{1} R \to S_{1}$ has small kernel $S_{2}$ — small because $e_{1} J = (e_{1} R) J$ and $e_{1} R$ is finitely generated.

A local example: $R = ℤ / p^{3} ℤ$

Here $R$ is local with $J = (p)$ and $\bar{R} = 𝔽_{p}$ . By $(24.13)$ , the projective cover of a finitely generated module $M$ is free of rank $\dim_{𝔽_{p}} M / p M$ . For $M = ℤ / p^{2} ℤ \oplus ℤ / p ℤ$ we get $M / p M ≅ 𝔽_{p}^{2}$ , so the cover is $R^{2} \to M$ , of order $p^{6}$ mapping onto a module of order $p^{3}$ , with kernel of order $p^{3}$ contained in $R^{2} J = (p R)^{2}$ — small, as required.

Process and Workflow

Reduce modulo the radicalForm

M / M J

and regard it as a module over the semisimple ring

R / J

Decompose into simplesWrite

M / M J ≅ ⨁_{α \in I} \bar{e_{α}} \bar{R}

; the multiset of labels is the only data the cover depends on.

Assemble the projectiveSet

P = ⨁_{α \in I} e_{α} R

. This is projective and has the correct top by construction.

Lift the mapUse projectivity of

P

to lift

P \to M / M J

along

M ↠ M / M J

, obtaining

θ : P \to M

Upgrade with NakayamaConclude

θ (P) = M

from

θ (P) + M J = M

, and

\ker θ \subseteq_{s} P

from

\ker θ \subseteq P J

, using finite generation or right T-nilpotence.

Which version of the theorem applies?

$R$ semiperfect, $M$ finitely generatedUse

(24.12)

Case 1. The index set

I

is finite and both smallness statements come from Nakayama's Lemma.

$R$ semiperfect, $M$ not finitely generatedThe construction still produces

θ

with (A) and (B), but neither can be upgraded: nothing forces

M J

P J

to be small. A cover may genuinely fail to exist.

$R$ right perfect, any $M$ Use

(24.12)

Case 2. Right T-nilpotence supplies smallness uniformly, and

(24.18)

shows this property characterises right perfect rings.

$R$ localUse

(24.13)

: the cover is free of rank

\dim_{R / J} M / M J

, and no idempotent bookkeeping is needed.

Comparison and Classification

Supply of projective covers by class of ring
	Semisimple	Local	Semiperfect	Right perfect	Right artinian	General
Covers for finitely generated $M$	yes	yes	yes	yes	yes	no
Covers for all $M$	yes	partial	partial	yes	yes	no
$rad R$ right T-nilpotent	yes	partial	partial	yes	yes	no
Finitely generated projectives are $⨁ e_{α} R$	yes	yes	yes	yes	yes	no
All projectives are $⨁ e_{α} R$	yes	partial	partial	yes	yes	no

Supply of projective covers by class of ring

The part entries in the local and semiperfect columns record that these classes contain both perfect and non-perfect rings: $ℤ_{(p)}$ is local, hence semiperfect, but $rad ℤ_{(p)} = p ℤ_{(p)}$ is not T-nilpotent, since no product $p \cdot p \dots p$ vanishes.

The three statements compared
Result	Hypothesis on $R$	Hypothesis on $M$ or $P$	Conclusion
(24.12) Case 1	semiperfect	$M$ finitely generated	$M$ has a projective cover
(24.12) Case 2	right perfect	none	$M$ has a projective cover
(24.13)	local	$M$ finitely generated	cover is free of rank $\dim_{R / J} M / M J$
(24.14)(1)	semiperfect	$P$ finitely generated projective	$P ≅ ⨁_{finite} e_{α} R$
(24.14)(2)	right perfect	$P$ projective	$P ≅ ⨁_{α \in I} e_{α} R$

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. For $k G$ with $char k = p$ dividing $| G |$ , the algebra is artinian hence semiperfect; the principal indecomposables $e_{α} k G$ are the projective covers of the simple modules, and the Cartan matrix records their composition factors. Brauer characters and decomposition matrices are computed from exactly this decomposition.
Quiver algebras and computer algebra. Software for finite-dimensional algebras (GAP's QPA, Magma, Sage) represents a basic algebra by its projective indecomposables; $(24.14)$ is the theorem guaranteeing that a finitely generated projective is determined by a vector of multiplicities, which is how such modules are stored.
Minimal resolutions and Betti numbers. Iterating $(24.12)$ gives a minimal projective resolution, whose ranks are invariants; this is the standard route to $Ext$ groups over group algebras and over local rings in commutative algebra.
Lifting and deformation problems. The pattern solve modulo the radical, then lift is the algebraic core of Hensel-style arguments used in computational number theory and in the lifting of idempotents that produces block decompositions.

Honest summary: this material is internal to algebra, but it is the engine behind the data structures and invariants that representation-theoretic software actually manipulates.

Failure Modes and Common Mistakes

Do not conflate indecomposable projective with simple: $e_{α} R$ is usually not simple, and its top $e_{α} R / e_{α} J$ is what is simple.
Do not forget to check that the idempotents used are local; a decomposition of $1$ into orthogonal idempotents that are not local does not give the simple modules, and $(23.6)$ is where locality comes from.
Do not read $(24.14)$ as a Krull–Schmidt theorem on its own: uniqueness of the multiplicities needs the local endomorphism rings and the Krull–Schmidt–Azumaya theorem.
Do not assume the decomposition of $M / M J$ is unique as a set of submodules; only the multiset of isomorphism types of simple summands is well defined, and that is all the construction uses.

Quick Reference

Semiperfect

R / J

semisimple and idempotents lift;

1 = \sum e_{i}

e_{i}

local

Right perfect

R / J

semisimple and

J

right T-nilpotent

Cover recipe

M / M J ≅ ⨁_{α} \bar{e_{α}} \bar{R} \Rightarrow P = ⨁_{α} e_{α} R

Two facts about $θ$

θ (P) + M J = M

and

\ker θ \subseteq P J

Upgrade toolNakayama (f.g. case) or right T-nilpotence (general case)

Local ringCover is free of rank

\dim_{R / J} M / M J

Projectives, semiperfectFinitely generated

P ≅

finite

⨁ e_{α} R

Projectives, right perfectEvery

P ≅ ⨁_{α \in I} e_{α} R

UniquenessMultiplicities are invariants (Krull–Schmidt–Azumaya)

Checklist for building a cover
Step	Object	Justification
1	$M / M J$ over $R / J$	$R / J$ semisimple
2	$⨁_{α} \bar{e_{α}} \bar{R}$	classification of simples via $(23.6)$
3	$P = ⨁_{α} e_{α} R$	each $e_{α} R$ projective
4	$θ : P \to M$	projectivity of $P$
5	$θ$ onto	$M J \subseteq_{s} M$
6	$\ker θ \subseteq_{s} P$	$P J \subseteq_{s} P$ and $(24.2) (3)$

Frequently Asked Questions

Why must the idempotents be local?

Locality of $e$ is exactly what makes $e R / e J$ simple, so that the decomposition of $M / M J$ into simples can be matched summand by summand with projectives $e R$ . A decomposition of $1$ into non-local orthogonal idempotents gives projectives whose tops are not simple, and the matching fails. $(23.6)$ says semiperfectness is precisely the availability of such a decomposition.

Is the projective cover of

M

finitely generated when

M

is?

Yes, over a semiperfect ring. $M / M J$ is then a finitely generated semisimple module, hence a finite direct sum of simples, so the index set $I$ is finite and $P = ⨁_{α \in I} e_{α} R$ is finitely generated. This also shows the cover is finitely presented whenever $R$ is right Noetherian.

Where exactly does the proof break for infinitely generated modules over a semiperfect ring that is not perfect?

At the two upgrades. The lifted map only satisfies $θ (P) + M J = M$ and $\ker θ \subseteq P J$ ; both become the desired statements only if $M J$ and $P J$ are small in $M$ and in $P$ . Without finite generation, that smallness is equivalent to right T-nilpotence of $J$ by $(23.16)$ , which is exactly what perfectness adds.

Does

(24.14)

imply that projective modules over a semiperfect ring are free?

No — only over a local ring, where there is a single principal indecomposable $R$ itself. Over $T_{2} (k)$ the module $e_{1} R$ is projective and indecomposable of dimension $2$ , while free modules have dimension a multiple of $3$ ; so $e_{1} R$ is projective and not free.

How is

(24.12)

used to lift idempotents?

In the proof of the converse theorem $(24.16)$ : given covers of $\bar{u} \bar{R}$ and of $(1 - \bar{u}) \bar{R}$ , their direct sum is a cover of $\bar{R}$ , and so is $R \to \bar{R}$ . Uniqueness $(24.10)$ produces an isomorphism $P \oplus Q ≅ R$ , and the corresponding decomposition $1 = e + f$ into orthogonal idempotents reduces to $\bar{u}$ and $1 - \bar{u}$ modulo $J$ .

Can two non-isomorphic modules have isomorphic projective covers?

Yes. The cover depends only on $M / M J$ , so $S_{1}$ and $e_{1} R$ over $T_{2} (k)$ both have cover $e_{1} R$ despite having different dimensions. The cover determines the top of $M$ , not $M$ itself; recovering $M$ requires the kernel as well.

References

T. Y. Lam, A First Course in Noncommutative Rings, Graduate Texts in Mathematics 131, Springer-Verlag, 1991, §24, (24.12)–(24.14) (pp. 363–365); see also §23, (23.6) and (23.16).
H. Bass, “Finitistic dimension and a homological generalization of semi-primary rings”, Transactions of the American Mathematical Society 95 (1960), 466–488.
F. W. Anderson and K. R. Fuller, Rings and Categories of Modules, 2nd edition, Graduate Texts in Mathematics 13, Springer-Verlag, 1992, §27 (semiperfect and perfect rings).
C. W. Curtis and I. Reiner, Methods of Representation Theory, Volume I, Wiley-Interscience, 1981, §6 (principal indecomposable modules).
G. Azumaya, “Corrections and supplementaries to my paper concerning Krull–Remak–Schmidt's theorem”, Nagoya Mathematical Journal 1 (1950), 117–124.

AI Suggested Questions

Give an explicit example of a right perfect ring that is not left perfect and exhibit a left module with no projective cover.
Compute the projective indecomposables and the Cartan matrix of the group algebra $𝔽_{2} S_{3}$ .
Prove that ${End}_{R} (e R) ≅ e R e$ and use it to justify the Krull–Schmidt–Azumaya step in $(24.14)$ .
Show that a projective module over a local ring is free, and compare the argument with $(24.13)$ .
What is the projective cover of the trivial module over a group algebra in characteristic $p$ , and how does its structure reflect the Sylow subgroup?
Describe an algorithm that, given structure constants for a finite-dimensional algebra, returns the projective cover of a specified module.
For which semiperfect rings that are not right perfect does some infinitely generated module still fail to have a projective cover?

01Executive Summary

02Overview

03Learning Objectives

04Definitions

05Core Concepts

Why the top determines the cover

Where the two hypotheses enter

06Key Results

07Proof Techniques and Method

Solve the problem modulo the radical

Lift by projectivity, correct by Nakayama

Compare two covers of the same module

08Worked Example

Upper triangular 2×2 matrices

A local example: R=ℤ/p3ℤ

09Process and Workflow

10Comparison and Classification

11Applications and Industry Use

12Failure Modes and Common Mistakes

13Quick Reference

14Frequently Asked Questions

15Related KEVOS Topics

16References

17AI Suggested Questions

Continue learning