Does I ⊆ rad R imply that R is I-adically complete?

No, and the failure is common. Z_(p) has pZ_(p) = radZ_(p) but its p-adic completion is the strictly larger ring Z_p. Completeness is a genuine extra hypothesis, and it is the one that produces lifts.

Idempotents in I-Adically Complete Rings

Q: Is the lifted idempotent unique?

No. If e is a lift then so is any conjugate of e by a unit congruent to 1 modulo I, and there are generally many. What is canonical is the isomorphism class of the summand: by (21.21), for I ⊆ rad R two idempotents of R generate isomorphic right ideals exactly when their images do.

Q: Why only countable orthogonal families?

The proof of (21.25) is a recursion: having lifted e_1, …, e_n, one adjusts the next lift to be orthogonal to their sum. The adjustment uses conjugation by a unit built from the previous idempotents, and there is no transfinite version at this level of generality, so the statement is made for countable families.

Executive Summary

Idempotents of a quotient $R / I$ record decompositions of $R / I$ ; lifting one to $R$ recovers a decomposition of $R$ itself. This is impossible in general — $3 \in ℤ / 6 ℤ$ is idempotent and $ℤ$ has no idempotent but $0$ and $1$ — so one looks for hypotheses on $I$ that force liftability. Nilpotence of $I$ is one. Completeness of $R$ in the $I$ -adic topology is the other, and it is strictly more general.

The proof is successive approximation. Lifting across $I^{n} / I^{n + 1}$ is easy because that ideal has square zero; doing it for every $n$ produces a coherent sequence, and completeness converts the sequence into an actual element. Completeness also forces $I \subseteq rad R$ , which is what makes the lift essentially unique and lets primitivity be tested downstairs.

$I \subseteq rad R$ Automatic consequence

$ℵ_{0}$ Orthogonal families liftable

$ℤ_{p}$ Motivating example

(21.31)The lifting theorem

Overview

Let $I$ be an ideal of a ring $R$ . The powers of $I$ give an inverse system of quotient rings, each mapping onto the previous one:

R / I

R / I^{2}

R / I^{3}

\dots

\hat{R} = {\hat{R}}_{I} = \lim_{⟵} R / I^{n}, ι : R ⟶ \hat{R} .

(21.30a)

The $I$ -adic completion of $R$ , with its canonical map. $R$ is called $I$ -adically complete when $ι$ is an isomorphism.

Think of $I^{n}$ for large $n$ as the small elements. Then $ι$ being injective says that the only element smaller than everything is $0$ , and $ι$ being surjective says that every Cauchy sequence already has a limit inside $R$ . Completeness is exactly these two statements together.

The archetypes are $ℤ_{p}$ complete for $I = p ℤ_{p}$ , and $k [[x]]$ complete for $I = (x)$ . Noncommutative examples are typically manufactured from commutative ones: if $k$ is a complete commutative ring and $R$ is a $k$ -algebra finitely generated as a $k$ -module, then $R$ is complete for $I R$ — that construction is treated on the Idempotents in Complete Algebras page.

Learning Objectives

State the inverse system defining ${\hat{R}}_{I}$ and the two conditions equivalent to completeness.
Prove $(21.30)$ : nilpotent $\Rightarrow$ complete $\Rightarrow$ $I \subseteq rad R$ , and show both implications are strict.
Prove $(21.31)$ : idempotents of $R / I$ lift to $R$ whenever $R$ is $I$ -adically complete.
Use the square-zero lifting formula $e = 3 a^{2} - 2 a^{3}$ at each stage of the induction.
Deduce $(21.32)$ : primitivity is detected in $R / I$ , and countable orthogonal families lift.
Recognise when a ring is not complete even though $I \subseteq rad R$ .

Definitions

Definition(21.30a)

I

-adic completion and completeness

For an ideal $I \subseteq R$ , the ** $I$ -adic completion** is $\hat{R} = \lim_{⟵} R / I^{n}$ , the inverse limit taken along the natural surjections $R / I^{n + 1} ↠ R / I^{n}$ . Concretely its elements are sequences $(a_{1}, a_{2}, \dots)$ with $a_{n} \in R / I^{n}$ and $a_{n + 1} \mapsto a_{n}$ . The ring $R$ is ** $I$ -adically complete** if the canonical map $ι : R \to \hat{R}$ is an isomorphism, which amounts to:

(Hausdorff / injectivity) $⋂_{n \geq 1} I^{n} = 0$ ;
(Convergence / surjectivity) for every sequence $a_{1}, a_{2}, \dots \in R$ with $a_{n + 1} \equiv a_{n} (\mod I^{n})$ for all $n$ , there exists $a \in R$ with $a \equiv a_{n} (\mod I^{n})$ for all $n$ .

Definition—Lifting idempotents

Given an ideal $I \subseteq R$ , an idempotent $x \in R / I$ **can be lifted to $R$ ** if there is an idempotent $e \in R$ with $e + I = x$ . We say *idempotents of $R / I$ can be lifted to $R$ * when this holds for every idempotent of $R / I$ .

$I$ -adic topology: The topology on $(R, +)$ with ${I^{n} : n \geq 0}$ a fundamental system of neighbourhoods of $0$ ; it is Hausdorff exactly when condition (1) holds.
Formal limit: When $R$ is not complete, a Cauchy sequence still determines an element of $\hat{R}$ ; one writes $a = \lim_{n} a_{n}$ in $\hat{R}$ .
Nil ideal: An ideal every element of which is nilpotent. Weaker than nilpotent, and also sufficient for lifting — see the companion result (21.28).
$U (R)$: The unit group of $R$ . The link between completeness and the radical runs through geometric series landing in $U (R)$ .

All rings have an identity. Powers $I^{n}$ are ideal powers, not the set of $n$ -th powers of elements.

Core Concepts

Completeness as a construction principle

Condition (2) is a licence to construct. If a desired element $a$ can be pinned down modulo $I$ , then modulo $I^{2}$ , and so on compatibly, completeness produces $a$ . Condition (1) says the result is unique. Every use of completeness in ring theory is an instance of this: solve the problem in the artinian-like quotients $R / I^{n}$ , where obstructions vanish because $I / I^{n}$ is nilpotent, then assemble.

Why square-zero ideals are the base case

Suppose $N \subseteq R$ is an ideal with $N^{2} = 0$ and $a \in R$ satisfies $a^{2} - a \in N$ . Put $n = a^{2} - a$ , so $n^{2} = 0$ and $n$ commutes with $a$ . Then

e = a - (2 a - 1) (a^{2} - a) = 3 a^{2} - 2 a^{3}

(C.1)

A universal polynomial lift across a square-zero ideal — no choices, no hypotheses beyond $n^{2} = 0$ .

satisfies $e \equiv a (\mod N)$ and $e^{2} = e$ . Indeed $e^{2} - e = (a^{2} - a) - (4 a^{2} - 4 a + 1) n = n - (4 n + 1) n = - 4 n^{2} = 0$ . This one formula, applied to the square-zero ideals $I^{n} / I^{n + 1} \subseteq R / I^{n + 1}$ , drives the whole induction.

Completeness sits between nilpotence and the radical

The chain below is $(21.30)$ . It positions completeness precisely: strong enough to lift idempotents, weak enough to cover $ℤ_{p}$ and $k [[x]]$ , and it always places $I$ inside the radical, so the lifting theory of $§ 21$ applies verbatim.

I

nilpotent

R

I

-adically complete

I \subseteq rad R

Key Results

Proposition(21.30)Nilpotent, complete, radical

Let $I$ be an ideal of a ring $R$ . If $I$ is nilpotent then $R$ is $I$ -adically complete. If $R$ is $I$ -adically complete then $I \subseteq rad R$ .

Proof

First implication. Say $I^{m} = 0$ . Then $I^{n} = 0$ for $n \geq m$ , so $⋂_{n} I^{n} = 0$ and the inverse system is eventually the constant system $R \leftarrow R \leftarrow \dots$ . A Cauchy sequence $(a_{n})$ satisfies $a_{n + 1} \equiv a_{n} (\mod I^{n}) = (\mod 0)$ for $n \geq m$ , hence is eventually constant, and its eventual value is the required limit.

Second implication. Let $b \in I$ ; we show $1 - b \in U (R)$ . Put $s_{n} = 1 + b + \dots + b^{n - 1}$ . Then $s_{n + 1} - s_{n} = b^{n} \in I^{n}$ , so $(s_{n})$ is Cauchy and completeness supplies $u \in R$ with $u \equiv s_{n} (\mod I^{n})$ for every $n$ . Now

(1 - b) u - 1 = (1 - b) (u - s_{n}) + [(1 - b) s_{n} - 1] = (1 - b) (u - s_{n}) - b^{n} \in I^{n}

(21.30b)

for every $n$ , so $(1 - b) u - 1 \in ⋂_{n} I^{n} = 0$ and $(1 - b) u = 1$ . The same computation on the other side gives $u (1 - b) = 1$ , so $1 - b \in U (R)$ . Since $I$ is an ideal, $x b \in I$ for all $x \in R$ , so $1 - x b \in U (R)$ for all $x$ ; by the unit characterisation of the Jacobson radical, $b \in rad R$ .

Remark—Both implications are strict

$k [[x]]$ is $(x)$ -adically complete but $(x)$ is not nilpotent, indeed not nil. And $ℤ_{(p)}$ has $p ℤ_{(p)} = rad ℤ_{(p)}$ yet is not $p$ -adically complete: its completion is $ℤ_{p}$ , strictly larger. So neither arrow reverses.

Theorem(21.31)Lifting across a complete ideal

Let $I$ be an ideal of a ring $R$ such that $R$ is $I$ -adically complete. Then every idempotent of $R / I$ can be lifted to an idempotent of $R$ .

Proof

Let $a_{1} \in R / I$ be idempotent. Regard $R / I$ as $(R / I^{2}) / (I / I^{2})$ . The ideal $I / I^{2} \subseteq R / I^{2}$ has square zero, so by the formula $(C . 1)$ there is an idempotent $a_{2} \in R / I^{2}$ mapping to $a_{1}$ .

Inductively, having found an idempotent $a_{n} \in R / I^{n}$ lifting $a_{n - 1}$ , apply the same step to the square-zero ideal $I^{n} / I^{n + 1} \subseteq R / I^{n + 1}$ to obtain an idempotent $a_{n + 1} \in R / I^{n + 1}$ lifting $a_{n}$ . The resulting sequence is by construction compatible, so it defines an element $a = (a_{1}, a_{2}, \dots) \in \lim_{⟵} R / I^{n} = \hat{R}$ .

Since each $a_{n}$ is idempotent, $a^{2} = (a_{1}^{2}, a_{2}^{2}, \dots) = (a_{1}, a_{2}, \dots) = a$ . Completeness identifies $\hat{R}$ with $R$ , so $a$ is an idempotent of $R$ , and its image in $R / I$ is $a_{1}$ .

Proposition(21.22)Primitivity descends and ascends

Let $e \in R$ be an idempotent and $I \subseteq rad R$ an ideal. If $\bar{e}$ is primitive in $R / I$ then $e$ is primitive in $R$ . The converse holds provided idempotents of $R / I$ can be lifted to $R$ .

Proposition(21.25)Countable orthogonal families lift

Let $I \subseteq rad R$ be an ideal such that idempotents of $R / I$ can be lifted to $R$ . Then for any countable (possibly finite) family ${x_{1}, x_{2}, \dots}$ of pairwise orthogonal idempotents of $R / I$ there is a family ${e_{1}, e_{2}, \dots}$ of pairwise orthogonal idempotents of $R$ with ${\bar{e}}_{i} = x_{i}$ for every $i$ .

Corollary(21.32)The complete case

Let $I$ be an ideal of $R$ with $R$ $I$ -adically complete. Then an idempotent $e \in R$ is primitive in $R$ if and only if $\bar{e}$ is primitive in $R / I$ , and every countable set of pairwise orthogonal idempotents of $R / I$ lifts to a set of pairwise orthogonal idempotents of $R$ .

Proof. $(21.30)$ gives $I \subseteq rad R$ and $(21.31)$ gives liftability; now apply $(21.22)$ and $(21.25)$ .

Remark(21.21)Uniqueness of the lift

For an ideal $I \subseteq rad R$ and idempotents $e, f \in R$ , one has $e R ≅ f R$ as right $R$ -modules if and only if $\bar{e} \bar{R} ≅ \bar{f} \bar{R}$ over $R / I$ . So a lift is never literally unique — conjugating by a unit in $1 + I$ produces another — but it is unique up to isomorphism of the resulting summand, which is all a decomposition theory needs.

Proof Techniques and Method

How these proofs work, and which move to reuse.

Move 1

Reduce to square zero

Never lift across $I$ in one go. Filter by $I \supseteq I^{2} \supseteq \dots$ so that each step crosses an ideal with square zero, where an explicit polynomial formula does the work.

Move 2

Assemble by completeness

A compatible sequence of solutions in the $R / I^{n}$ is an element of $\hat{R}$ . Any polynomial identity satisfied at every level is satisfied by the limit, because the inverse limit is computed componentwise.

Move 3

Geometric series for units

To show $1 - b$ is invertible for $b$ small, exhibit the partial sums $1 + b + \dots + b^{n - 1}$ as a Cauchy sequence and let completeness supply the inverse. This is how completeness implies $I \subseteq rad R$ .

Move 1 is the same idea as Hensel's lemma; Move 2 is the same idea as Newton's method converging in a complete metric space. Nothing in either argument uses commutativity, which is exactly why the technique survives into noncommutative ring theory.

Worked Example

Splitting a quadratic extension of $ℤ_{7}$

Let $R = ℤ_{7} [x] / (x^{2} - 2)$ and $I = 7 R$ , so $R$ is $I$ -adically complete because $ℤ_{7}$ is $7$ -adically complete and $R$ is free of rank $2$ over it. Modulo $7$ we have $x^{2} - 2 = (x - 3) (x + 3)$ in $𝔽_{7} [x]$ , since $3^{2} = 9 \equiv 2$ . Hence

R / 7 R ≅ 𝔽_{7} [x] / (x - 3) \times 𝔽_{7} [x] / (x + 3) ≅ 𝔽_{7} \times 𝔽_{7},

(E.1)

and the idempotent cutting out the first factor is $a = 6 (x + 3) = 6 x + 4$ in $R / 7 R$ : at $x = 3$ it takes the value $6 \cdot 6 = 36 \equiv 1$ , at $x = - 3$ the value $0$ . Check directly: $a^{2} = 36 (x^{2} + 6 x + 9) = 36 (6 x + 11) \equiv 6 x + 4 (\mod 7)$ , using $x^{2} = 2$ .

Step one: lift modulo $49$

Take the naive lift $a = 6 x + 4 \in R$ and compute in $ℤ [x] / (x^{2} - 2)$ : $a^{2} = 48 x + 88$ and $a^{3} = 720 x + 928$ . The obstruction is $a^{2} - a = 42 x + 84 = 7 (6 x + 12)$ , which indeed lies in $7 R$ and squares into $49 R$ . Applying $(C . 1)$ ,

e_{2} = 3 a^{2} - 2 a^{3} = 3 (48 x + 88) - 2 (720 x + 928) = - 1296 x - 1592 \equiv 27 x + 25 (\mod 49) .

(E.2)

Verify: $(27 x + 25)^{2} = 729 x^{2} + 1350 x + 625 = 1350 x + 2083$ , and $1350 = 27 \cdot 49 + 27$ , $2083 = 42 \cdot 49 + 25$ . So $(27 x + 25)^{2} \equiv 27 x + 25 (\mod 49)$ , and reducing mod $7$ returns $6 x + 4$ . Both checks pass.

Step two: the limit

Iterating produces $e_{n} \in R / 7^{n} R$ for every $n$ , and completeness assembles them into a genuine idempotent $e \in R$ . One can name it: Hensel's lemma gives $α \in ℤ_{7}$ with $α^{2} = 2$ and $α \equiv 3 (\mod 7)$ — the next approximation is $α \equiv 10 (\mod 49)$ , since $10^{2} = 100 = 2 + 2 \cdot 49$ — and then

e = \frac{x + α}{2 α}, R ≅ e R \times (1 - e) R ≅ ℤ_{7} \times ℤ_{7} .

(E.3)

The idempotent evaluates to $1$ at $x = α$ and to $0$ at $x = - α$ .

Contrast $R' = ℤ_{(7)} [x] / (x^{2} - 2)$ with the same ideal. Here $\bar{e}$ still exists in $R' / 7 R'$ , and $7 R' \subseteq rad R'$ , but $R'$ is not $7$ -adically complete — and in fact $R'$ is a domain, so it has no nontrivial idempotent and the lift genuinely fails. Completeness, not the radical condition, is what does the work.

Process and Workflow

Check the Hausdorff conditionVerify

⋂_{n} I^{n} = 0

. Without it, distinct elements of

R

become indistinguishable in

\hat{R}

and no uniqueness statement holds.

Check convergenceVerify that every compatible sequence has a limit in

R

. For

ℤ_{p}

and

k [[x]]

this is the definition; for a module-finite algebra over a complete base it follows from Krull's intersection theorem and Nakayama.

Lift stage by stageStart from the idempotent in

R / I

and apply

e = 3 a^{2} - 2 a^{3}

across each

I^{n} / I^{n + 1}

Assemble and read offThe compatible sequence is an idempotent of

R

. Then

(21.32)

transfers primitivity, and

(21.21)

pins the summand

e R

down up to isomorphism.

Comparison and Classification

Hypotheses on $I$ that guarantee lifting
Hypothesis on $I$	Lifting holds?	$I \subseteq rad R$ ?	Typical example
$I$ nilpotent	yes, (21.28)	yes	strictly upper triangular matrices
$I$ nil	yes, (21.28)	yes	nil radical of an algebraic algebra
$R$ $I$ -adically complete	yes, (21.31)	yes, (21.30)	$p ℤ_{p} \subseteq ℤ_{p}$
$I \subseteq rad R$ only	not in general	yes	$7 ℤ_{(7)} [x] / (x^{2} - 2)$
$I$ arbitrary	no	no	$6 ℤ \subseteq ℤ$

The two sufficient conditions — nil and complete — are genuinely independent. The ideal $(x) \subseteq k [[x]]$ is complete but not nil; a nil ideal of infinite nilpotency index in a non-complete ring is nil but gives no completeness. A ring that is semiperfect is by definition semilocal with idempotents lifting modulo the radical, so both conditions are ways of certifying semiperfectness when $R / rad R$ is semisimple.

Relationship Map

Ideals $I$ of $R$ no lifting in general

$I \subseteq rad R$ primitivity descends; lifts, when they exist, are unique up to isomorphism

(21.21)

$R$ is $I$ -adically completeidempotents lift

(21.31)

; countable orthogonal families lift

(21.32)

$I$ nilpotentthe inverse system stabilises; every Cauchy sequence is eventually constant

Reading outwards: the smaller the ideal in this hierarchy, the more decomposition data of $R$ is already visible in $R / I$ . At the innermost level $R$ and $R / I$ have literally the same idempotent theory up to conjugacy.

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

Blocks over a complete DVR

Brauer theory is set up over a complete discrete valuation ring $𝒪$ with residue field of characteristic $p$ . Completeness of $𝒪$ is exactly what lets the block idempotents of $k G$ be lifted to $𝒪 G$ , so that blocks in characteristic $p$ and in characteristic $0$ can be compared.

Number theory

Hensel's lemma

Factoring a polynomial over $ℤ_{p}$ from a factorisation over $𝔽_{p}$ is the commutative shadow of $(21.31)$ : coprime factorisations correspond to idempotents of the quotient algebra, and lifting the factorisation is lifting the idempotent.

Symbolic computation

$p$ -adic and Hensel lifting in CAS

Multivariate factorisation, linear solving over $ℚ$ , and Gröbner basis reconstruction all work modulo a prime and lift $p$ -adically. The lifting step is the polynomial recursion of this page, usually run with quadratic rather than linear convergence.

Deformation theory

Rigidity of idempotents

That idempotents lift across nilpotent ideals says decompositions do not deform: a first-order deformation of an algebra carries its decomposition along. This formal-smoothness statement is used when arguing that a family of algebras has locally constant block structure.

The honest summary: this is infrastructure for working with a hard object by working with its easy reductions. Almost every $p$ -adic or formal method in algebra rests on the ability to lift solutions of $x^{2} = x$ and of Hensel-type equations.

Computational Notes

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

The lift $e = 3 a^{2} - 2 a^{3}$ is linearly convergent: it gains one power of $I$ per iteration, so reaching precision $I^{n}$ costs $n$ steps.
Iterating the same formula gives quadratic convergence in practice: if $a^{2} - a \in I^{m}$ then $3 a^{2} - 2 a^{3}$ satisfies $e^{2} - e \in I^{2 m}$ , so precision doubles per step and only $O (\log n)$ steps are needed. This is the Newton iteration for $x^{2} - x$ .
Each step costs a constant number of multiplications in $R / I^{n}$ ; for a $k$ -algebra of dimension $d$ over $ℤ / p^{n}$ that is $O (d^{3} n^{2})$ bit operations with schoolbook arithmetic, less with fast multiplication.
Systems that work $p$ -adically — Magma's pAdicRing, Sage's Zp, GAP's PadicNumbers — carry an explicit precision parameter; every idempotent computed is only correct to that precision, and no exact test $e^{2} = e$ is available.
Completeness is not a decidable property of a presentation. In practice one certifies it structurally: a module-finite algebra over a complete noetherian commutative base is complete.

Failure Modes and Common Mistakes

Do not confuse $I^{n}$ (the $n$ -th power of the ideal) with the set of $n$ -th powers of elements of $I$ ; for noncommutative $I$ the former is generated by products $b_{1} \dots b_{n}$ .
Do not expect $(21.25)$ to cover uncountable orthogonal families — the recursion in its proof is genuinely countable.
Do not assume the lifted idempotent lies in the same one-sided ideal as your original element; the guarantee $e \in a R$ comes from the nil-ideal theorem $(21.28)$ , not from $(21.31)$ .
Do not forget that condition (1) of completeness can fail silently: $⋂_{n} I^{n} \neq 0$ makes $ι$ non-injective, and then $R$ is not complete even if every Cauchy sequence converges.

Historical Notes and Lessons Learned

1897–1908Hensel's $p$ -adic numbersHensel introduces $ℤ_{p}$ and the lifting lemma for coprime factorisations, the commutative prototype of every argument on this page.
1930sKrull's completionsKrull develops $I$ -adic topologies and the intersection theorem for noetherian rings, providing the Hausdorff condition that makes limits unique.
1950sSemiperfect ringsIdempotent lifting modulo the radical is isolated as the defining property, alongside semilocality, of a semiperfect ring; complete rings become the standard supply of examples beyond the artinian case.
1960sIntegral representation theoryWork of Swan, Curtis and Reiner puts complete local rings at the base of the theory, precisely so that Krull–Schmidt and block decompositions behave.

The methodological lesson: completeness is not a finiteness condition, and that is its value. It buys the same lifting statements as nilpotence while allowing rings with elements of infinite order in the filtration, so the theory extends from artinian algebras to $p$ -adic orders without change of proof.

Quick Reference

Completion

\hat{R} = \lim_{⟵} R / I^{n}

Complete

ι : R \to \hat{R}

an isomorphism

Condition 1

⋂_{n \geq 1} I^{n} = 0

Condition 2Cauchy sequences have limits in

R

Consequence

I \subseteq rad R

, so

1 + I \subseteq U (R)

Square-zero lift

e = 3 a^{2} - 2 a^{3}

Main theoremComplete

\Rightarrow

idempotents of

R / I

lift

BonusPrimitivity is detected in

R / I

; countable orthogonal families lift

Statements and their hypotheses
Statement	Hypotheses	Reference
$I$ nilpotent $\Rightarrow$ $R$ complete	$I$ an ideal	(21.30)
$R$ complete $\Rightarrow$ $I \subseteq rad R$	$I$ an ideal	(21.30)
Idempotents of $R / I$ lift	$R$ $I$ -adically complete	(21.31)
$e$ primitive $\Leftrightarrow$ $\bar{e}$ primitive	$R$ $I$ -adically complete	(21.32)
Countable orthogonal families lift	$I \subseteq rad R$ , idempotents lift	(21.25)
$e R ≅ f R \Leftrightarrow \bar{e} \bar{R} ≅ \bar{f} \bar{R}$	$I \subseteq rad R$	(21.21)

Frequently Asked Questions

Why does the induction go through

I^{n} / I^{n + 1}

instead of lifting across

I

directly?

Because there is no formula that lifts an idempotent across an arbitrary ideal — $3 \in ℤ / 6$ is the standard obstruction. Across a square-zero ideal there is one, namely $e = 3 a^{2} - 2 a^{3}$ . Filtering $I$ by its powers turns an impossible single step into a sequence of trivial ones, and completeness is precisely the hypothesis that lets the sequence be assembled.

Is the lifted idempotent unique?

No. If $e$ is a lift then so is any conjugate of $e$ by a unit congruent to $1$ modulo $I$ , and there are generally many. What is canonical is the isomorphism class of the summand: by $(21.21)$ , for $I \subseteq rad R$ two idempotents of $R$ generate isomorphic right ideals exactly when their images do.

Does

I \subseteq rad R

imply that

R

I

-adically complete?

No, and the failure is common. $ℤ_{(p)}$ has $p ℤ_{(p)} = rad ℤ_{(p)}$ but its $p$ -adic completion is the strictly larger ring $ℤ_{p}$ . Completeness is a genuine extra hypothesis, and it is the one that produces lifts.

How does this relate to the nil-ideal lifting theorem?

They are independent sufficient conditions with the same conclusion. $(21.28)$ says idempotents lift across a nil ideal, and it gives the extra information $e \in a R$ ; $(21.31)$ says they lift across a complete ideal, and gives no such containment. A nilpotent ideal satisfies both hypotheses; $(x) \subseteq k [[x]]$ satisfies only the second.

Where do noncommutative complete rings come from?

Almost always by base change: take a complete commutative noetherian ring $k$ with ideal $I$ , and a $k$ -algebra $R$ that is finitely generated as a $k$ -module. Then $R$ is $I R$ -adically complete. Group rings $𝒪 G$ over a complete discrete valuation ring and orders in semisimple $ℚ_{p}$ -algebras are the standard instances.

Why only countable orthogonal families?

The proof of $(21.25)$ is a recursion: having lifted $e_{1}, \dots, e_{n}$ , one adjusts the next lift to be orthogonal to their sum. The adjustment uses conjugation by a unit built from the previous idempotents, and there is no transfinite version at this level of generality, so the statement is made for countable families.

References

T. Y. Lam, A First Course in Noncommutative Rings, Graduate Texts in Mathematics 131, Springer-Verlag, 1991, §21, results (21.21), (21.22), (21.25), (21.30)–(21.32).
F. W. Anderson and K. R. Fuller, Rings and Categories of Modules, 2nd edition, Graduate Texts in Mathematics 13, Springer-Verlag, 1992, §27.
C. W. Curtis and I. Reiner, Methods of Representation Theory, Volume I, Wiley-Interscience, 1981, §6 and §30.
N. Jacobson, Basic Algebra II, W. H. Freeman, 1980, Chapter 9 (completions and Hensel's lemma).
H. Matsumura, Commutative Ring Theory, Cambridge University Press, 1986, Chapter 8 (I-adic completion, Krull's intersection theorem).

AI Suggested Questions

Give a full proof that $e = 3 a^{2} - 2 a^{3}$ is idempotent whenever $(a^{2} - a)^{2} = 0$ , and find the analogous formula for lifting across an ideal with $N^{3} = 0$ .
Exhibit a non-noetherian commutative ring whose $I$ -adic completion is not $I$ -adically complete.
How does the Newton iteration for idempotents achieve quadratic convergence, and what is the exact precision gain per step?
Compare the lifting theorems for nil ideals and for complete ideals: is there a common generalisation?
Describe how block idempotents of $k G$ are lifted to $𝒪 G$ over a complete discrete valuation ring, and what this buys in Brauer theory.
For which ideals $I$ of a noncommutative noetherian ring does the Artin–Rees lemma hold, and what does it give for completions?
Show that a countable orthogonal family of nonzero idempotents in $R$ produces an infinite direct sum of nonzero right ideals, and connect this to Dedekind-finiteness.

Idempotents in Complete Rings

Executive Summary

Overview

Learning Objectives

Definitions

Core Concepts

Completeness as a construction principle

Why square-zero ideals are the base case

Completeness sits between nilpotence and the radical

Key Results

Proof Techniques and Method

Reduce to square zero

Assemble by completeness

Geometric series for units

Worked Example

Splitting a quadratic extension of $ℤ_{7}$

Step one: lift modulo $49$

Step two: the limit

Process and Workflow

Comparison and Classification

Relationship Map

Applications and Industry Use

Blocks over a complete DVR

Hensel's lemma

$p$ -adic and Hensel lifting in CAS

Rigidity of idempotents

Computational Notes

Failure Modes and Common Mistakes

Historical Notes and Lessons Learned

Quick Reference

Frequently Asked Questions

References

AI Suggested Questions

Continue learning

01Executive Summary

02Overview

03Learning Objectives

04Definitions

05Core Concepts

Completeness as a construction principle

Why square-zero ideals are the base case

Completeness sits between nilpotence and the radical

06Key Results

07Proof Techniques and Method

Reduce to square zero

Assemble by completeness

Geometric series for units

08Worked Example

Splitting a quadratic extension of ℤ7

Step one: lift modulo 49

Step two: the limit

09Process and Workflow

10Comparison and Classification

11Relationship Map

12Applications and Industry Use

Blocks over a complete DVR

Hensel's lemma

p-adic and Hensel lifting in CAS

Rigidity of idempotents

13Computational Notes

14Failure Modes and Common Mistakes

15Historical Notes and Lessons Learned

16Quick Reference

17Frequently Asked Questions

18Related KEVOS Topics

19References

20AI Suggested Questions

Continue learning

Executive Summary

Overview

Learning Objectives

Definitions

Core Concepts

Key Results

Proof Techniques and Method

Worked Example

Splitting a quadratic extension of $ℤ_{7}$

Step one: lift modulo $49$

Process and Workflow

Comparison and Classification

Relationship Map

Applications and Industry Use

$p$ -adic and Hensel lifting in CAS

Computational Notes

Failure Modes and Common Mistakes

Historical Notes and Lessons Learned

Quick Reference

Frequently Asked Questions

Related KEVOS Topics

References

AI Suggested Questions