Lifting Idempotents Modulo Nil Ideals

Q: Why the exponent 2m rather than m or m+1?

Because the cut has to leave every cross term divisible by a^mb^m. With total degree 2m and the split at k = m, a term from the left half has a-degree at least m+1 and a term from the right half has b-degree at least m, so their product clears the threshold. Smaller total degree leaves cross terms of insufficient degree in one of the two variables.

Q: Why is nil enough, when so many theorems require nilpotent?

The proof consumes nilpotency of a single element, namely a - a^2, and allows its exponent to be whatever that element needs. Statements requiring a uniform bound — such as the existence of a fixed N making an algorithm terminate in N steps — do need nilpotence.

Q: What exactly does (21.29)(2) buy that (21.28) does not?

The conclusion e ∈ aR in (21.28) is what confines the idempotent to a prescribed right ideal. Without it you could lift the idempotent generator of the image of A and land outside A, which would make the corollary vacuous.

Executive Summary

Lifting idempotents is not possible modulo an arbitrary ideal. It is always possible modulo a nil ideal, and the proof is constructive: expand $1 = (a + (1 - a))^{2 m}$ , cut the expansion in half, and the first half is the idempotent you want.

The construction gives more than existence. The lift lies inside $a R$ , it is a polynomial in $a$ with integer coefficients, and any two lifts of the same class are conjugate. Because every nil ideal sits inside $rad R$ , the whole machinery of the previous page — orthogonalisation, transport of primitivity, lifting countable families — becomes available at once.

$(21.28)$ The lifting theorem

$2 m$ Binomial exponent

$e \in a R$ Where the lift lives

Nil, not nilpotentHypothesis needed

Overview

Recall the obstruction. In $ℤ$ with $I = 6 ℤ$ , the class of $3$ is idempotent modulo $I$ and does not lift; here $I$ is not nil. Replace $ℤ$ by $ℤ / 12 ℤ$ and $I$ by $rad (ℤ / 12) = (6)$ , which satisfies $(6)^{2} = 0$ , and the same class $3$ now lifts — to $9$ .

The mechanism is simple to describe. If $\bar{a}$ is idempotent then $a (1 - a) = a - a^{2}$ lies in $I$ , hence is nilpotent, say $(a (1 - a))^{m} = 0$ . Since $a$ and $1 - a$ commute, expanding $(a + (1 - a))^{2 m} = 1$ splits the sum into terms heavy in $a$ and terms heavy in $1 - a$ , and any product of one from each group is annihilated by $(a (1 - a))^{m}$ .

1 = {\underset{⏟}{\sum_{k = 0}^{m - 1} (\binom{2 m}{k}) a^{2 m - k} (1 - a)^{k}}}_{e} + {\underset{⏟}{\sum_{k = m}^{2 m} (\binom{2 m}{k}) a^{2 m - k} (1 - a)^{k}}}_{f}

(21.28)

The two halves are orthogonal, so each is idempotent.

The payoff is structural. Left artinian rings, right artinian rings and semiprimary rings all have nilpotent — hence nil — radical, so every result on this page applies to them without further checking.

Learning Objectives

State $(21.28)$ with the hypothesis that $I$ is a nil ideal, and identify where nilness is used.
Verify that the two halves of the binomial expansion are orthogonal idempotents.
Show the lift lies in $a R$ and reduces to $\bar{a}$ modulo $I$ .
Prove $(21.29) (1)$ : semilocal plus nil radical plus no nontrivial idempotents implies local.
Prove $(21.29) (2)$ : a right ideal contains a nonzero idempotent exactly when it is not nil.
Show by example that the nilness hypothesis in $(21.29)$ cannot be dropped.

Definitions

Definition—Nil and nilpotent

An ideal $I \subseteq R$ is nil if every $x \in I$ satisfies $x^{n (x)} = 0$ for some integer $n (x) \geq 1$ depending on $x$ . It is nilpotent if $I^{n} = 0$ for a single $n$ , meaning every product of $n$ elements of $I$ vanishes. Nilpotent implies nil; the converse fails.

$rad R$: The Jacobson radical. Every nil one-sided ideal is contained in it, since $x$ nilpotent makes $1 - y x$ invertible by a finite geometric series.
$a R$: The principal right ideal generated by $a$ ; the theorem places the lifted idempotent inside it.
$ℤ [a]$: The commutative subring of $R$ generated by $a$ and $1$ ; all elements appearing in the construction live here.
Semiprimary: $rad R$ nilpotent and $R / rad R$ semisimple. Every one-sided artinian ring is semiprimary.

Nil ideals are automatically contained in the Jacobson radical, so every result of the page Lifting Idempotents Modulo an Ideal applies verbatim to the situation considered here.

Core Concepts

Everything happens in a commutative subring

Write $b = 1 - a$ . Then $a b = b a = a - a^{2}$ , and both $a$ and $b$ lie in the commutative subring $ℤ [a]$ of $R$ . The binomial theorem is therefore available in its usual form, and the noncommutativity of $R$ plays no role at all in the construction. This is why the theorem is so robust: it is really a statement about a single element.

{\bar{a}}^{2} = \bar{a}

a b = a - a^{2} \in I

(a b)^{m} = 0

split

(a + b)^{2 m}

e^{2} = e

Why cutting at $k = m$ works

In the term $(\binom{2 m}{k}) a^{2 m - k} b^{k}$ , the two exponents sum to $2 m$ . Terms with $k \leq m - 1$ carry $a$ -exponent at least $m + 1$ ; terms with $k \geq m$ carry $b$ -exponent at least $m$ . Multiply one of each and you obtain a monomial divisible by $a^{m} b^{m} = (a b)^{m} = 0$ . So the two halves annihilate each other, and since they sum to $1$ , each is idempotent.

Nil is the right hypothesis, not nilpotent

The proof needs a single element to be nilpotent, with an exponent allowed to depend on that element. That is exactly nilness. Requiring $I^{n} = 0$ would be a strictly stronger hypothesis and would exclude, for instance, the ideal of strictly upper triangular matrices of finite support in an infinite matrix ring, or the nil radical of a commutative ring with unbounded nilpotency indices such as $k [x_{1}, x_{2}, \dots] / (x_{1}, x_{2}^{2}, x_{3}^{3}, \dots)$ .

Key Results

Theorem(21.28)Lifting modulo a nil ideal

Let $R$ be a ring and $I \subseteq R$ a nil ideal, so that $I \subseteq rad R$ . Let $a \in R$ be such that its image $\bar{a} \in \bar{R} = R / I$ is idempotent. Then there exists an idempotent $e \in a R$ with $\bar{e} = \bar{a}$ . In particular, idempotents lift modulo any nil ideal.

Proof

Put $b = 1 - a$ , so that $a b = b a = a - a^{2}$ . Since ${\bar{a}}^{2} = \bar{a}$ , we have $a b \in I$ , hence $a b$ is nilpotent: choose $m \geq 1$ with $(a b)^{m} = 0$ . All elements below lie in the commutative subring $ℤ [a]$ , so the binomial theorem applies:

1 = (a + b)^{2 m} = \sum_{k = 0}^{2 m} (\binom{2 m}{k}) a^{2 m - k} b^{k} .

Split the sum at $k = m$ and set

e = \sum_{k = 0}^{m - 1} (\binom{2 m}{k}) a^{2 m - k} b^{k}, f = \sum_{k = m}^{2 m} (\binom{2 m}{k}) a^{2 m - k} b^{k},

so that $e + f = 1$ .

Orthogonality. A term of $e$ has $a$ -exponent $2 m - k \geq m + 1$ ; a term of $f$ has $b$ -exponent at least $m$ . Their product is an integer multiple of $a^{i} b^{j}$ with $i \geq m + 1$ and $j \geq m$ , hence divisible by $a^{m} b^{m} = (a b)^{m} = 0$ . Expanding, $e f = 0$ , and $f e = 0$ by commutativity.

Idempotency. $e = e \cdot 1 = e (e + f) = e^{2} + e f = e^{2}$ .

Location. Every term of $e$ has $a$ -exponent at least $m + 1 \geq 1$ , so $e \in a R$ (indeed $e = a \cdot g (a)$ for a polynomial $g$ with integer coefficients).

Reduction. Modulo $I$ we have $\bar{a} \bar{b} = \bar{a} (1 - \bar{a}) = \bar{a} - {\bar{a}}^{2} = 0$ , so every term of $e$ with $k \geq 1$ vanishes in $\bar{R}$ , since it contains both a positive power of $a$ and a positive power of $b$ . Hence $\bar{e} = {\bar{a}}^{2 m} = \bar{a}$ , the last step because $\bar{a}$ is idempotent. $■$

Remark—What the construction adds

Three features are not automatic from abstract liftability. The lift $e$ lies in $a R$ , which is what makes $(21.29) (2)$ work. It is a polynomial in $a$ , so it commutes with every element that commutes with $a$ — the key point when lifting central idempotents. And $f = 1 - e$ lies in $(1 - a) R$ by the same argument applied to $b$ .

Corollary—Uniqueness up to conjugacy

Let $I$ be a nil ideal of $R$ and let $e, f$ be idempotents of $R$ with $e - f \in I$ . Then $f = u^{- 1} e u$ where $u = e f + (1 - e) (1 - f) \in 1 + I$ . If in addition $e$ and $f$ commute, then $e = f$ .

Proof

The conjugacy statement holds for any ideal inside $rad R$ and is proved on the page Lifting Idempotents Modulo an Ideal. For the commuting case, expand using $e f = f e$ , $e^{2} = e$ , $f^{2} = f$ :

(e - f) (1 - e - f) = e - e^{2} - e f - f + f e + f^{2} = 0 .

Now $(1 - 2 e)^{2} = 1 - 4 e + 4 e^{2} = 1$ , so $1 - 2 e$ is a unit equal to its own inverse, and

1 - e - f = (1 - 2 e) + (e - f) = (1 - 2 e) [1 + (1 - 2 e) (e - f)] .

The bracket lies in $1 + I \subseteq 1 + rad R \subseteq U (R)$ , so $1 - e - f$ is a unit. Cancelling it in the displayed identity gives $e - f = 0$ . $■$

Corollary(21.29)Semilocal rings with nil radical

Let $R$ be a semilocal ring such that $I = rad R$ is nil. Then:

(1) if $R \neq (0)$ and $R$ has no idempotents other than $0$ and $1$ , then $R$ is a local ring;
(2) a right ideal $𝔄 \subseteq R$ contains a nonzero idempotent if and only if $𝔄$ is not nil.

The hypotheses hold, in particular, for every semiprimary ring and every left or right artinian ring.

Proof

(1) Suppose $\bar{R} = R / I$ had an idempotent other than $\bar{0}$ and $\bar{1}$ . By $(21.28)$ it lifts to an idempotent $e \in R$ , and $e \notin {0, 1}$ because its image is neither $\bar{0}$ nor $\bar{1}$ — contradiction. So $\bar{R}$ has only the trivial idempotents. But $\bar{R}$ is semisimple and nonzero, so by Wedderburn–Artin it is a finite product of matrix rings over division rings; a product of two nonzero rings has nontrivial central idempotents, and $M_{n} (D)$ with $n \geq 2$ has the nontrivial idempotent $E_{11}$ . Hence $\bar{R} ≅ D$ for a division ring $D$ , which says exactly that $R$ is local.

(2) If $𝔄$ contains a nonzero idempotent $e$ , then $e^{n} = e \neq 0$ for all $n$ , so $e$ is not nilpotent and $𝔄$ is not nil.

Conversely suppose $𝔄$ is not nil. If $𝔄 \subseteq I$ then $𝔄$ would be nil, so the image $\bar{𝔄}$ is a nonzero right ideal of the semisimple ring $\bar{R}$ . Every right ideal of a semisimple ring is a direct summand, hence generated by an idempotent, so $\bar{𝔄} = \bar{a} \bar{R}$ for some $a \in 𝔄$ with $\bar{a}$ a nonzero idempotent. By $(21.28)$ there is an idempotent $e \in a R \subseteq 𝔄$ with $\bar{e} = \bar{a} \neq \bar{0}$ ; in particular $e \neq 0$ . $■$

Counterexample—Nilness cannot be dropped from (21.29)

Let $p \neq q$ be primes and $R = ℤ_{(p)} \cap ℤ_{(q)}$ , the subring of $ℚ$ of fractions with denominator coprime to $p q$ . Then $R$ is semilocal with $rad R = p q R$ and $R / rad R ≅ 𝔽_{p} \times 𝔽_{q}$ . Being a domain, $R$ has no idempotents besides $0$ and $1$ , yet $R$ is not local: it has two maximal ideals. So $(21.29) (1)$ fails.

The same ring defeats $(21.29) (2)$ : the right ideal $p R$ is not nil, because a domain has no nonzero nilpotents, and it contains no nonzero idempotent. The missing hypothesis is precisely that $rad R = p q R$ is not nil.

Proof Techniques and Method

How these proofs work, and which move to reuse.

The construction is a template, not a one-off.

Move 1

Restrict to $ℤ [a]$

A condition on a single element can be analysed inside the commutative subring it generates. Noncommutativity of the ambient ring becomes irrelevant, and classical identities become available.

Move 2

Split a partition of unity

Write $1$ as a sum whose terms fall into two groups that annihilate each other. Each group is then automatically idempotent, and the groups are orthogonal complements.

Move 3

Overshoot the degree

Choose the exponent $2 m$ so that every cross term is forced past the vanishing threshold. Doubling is the cheapest choice that works uniformly.

Move 2 is the reusable idea. It reappears in the proof that a projection can be split off a direct sum, and in the construction of orthogonal idempotents from a partition of a semisimple quotient. Whenever you can write $1 = e + f$ with $e f = 0$ , you have a decomposition of the ring, free of charge.

Worked Example

Lifting $3$ from $ℤ / 6$ to $ℤ / 12$

Take $R = ℤ / 12 ℤ$ . Its radical is $I = (6) = {0, 6}$ , and $6^{2} = 36 \equiv 0$ , so $I^{2} = 0$ and $I$ is nil. The quotient is $R / I ≅ ℤ / 6 ℤ$ , in which $\bar{3}$ is idempotent because $9 \equiv 3 (\mod 6)$ .

Apply the construction with $a = 3$ . Then $b = 1 - a = - 2 \equiv 10$ , and

a b = 3 \cdot 10 = 30 \equiv 6 (\mod 12), (a b)^{2} \equiv 36 \equiv 0,

(E.1)

so we may take $m = 2$ and expand to the power $2 m = 4$ . The lift is the sum of the terms with $k \leq m - 1 = 1$ :

e = (\binom{4}{0}) a^{4} + (\binom{4}{1}) a^{3} b = 3^{4} + 4 \cdot 3^{3} \cdot 10 = 81 + 1080 .

(E.2)

Reduce modulo $12$ : $81 = 72 + 9 \equiv 9$ and $1080 = 90 \cdot 12 \equiv 0$ , so $e = 9$ . Check the three conclusions of the theorem: $9^{2} = 81 \equiv 9$ , so $e$ is idempotent; $9 \equiv 3 (\mod 6)$ , so $e$ lifts $\bar{a}$ ; and $9 = 3 \cdot 3 \in 3 R$ , so $e \in a R$ as promised.

Non-nil right ideals contain idempotents

Stay in $R = ℤ / 12$ , which is artinian and hence semilocal with nil radical, so $(21.29) (2)$ applies. Consider two ideals:

Testing $(21.29) (2)$ in $ℤ / 12$
Ideal	Nil?	Contains a nonzero idempotent?
$(6) = {0, 6}$	Yes: $6^{2} = 0$	No — consistent with the corollary
$(4) = {0, 4, 8}$	No: $4^{2} = 4$	Yes: $4$ itself
$(2) = {0, 2, 4, 6, 8, 10}$	No: $2^{2} = 4$ , $2^{3} = 8$ , $2^{4} = 4, \dots$	Yes: $4 \in (2)$
$(3) = {0, 3, 6, 9}$	No: $3^{2} = 9$ , $9^{2} = 9$	Yes: $9 \in (3)$

The ideal $(2)$ illustrates the point cleanly: no power of $2$ vanishes modulo $12$ , and the corollary predicts an idempotent, which the construction locates at $4$ .

A local ring by $(21.29) (1)$

Let $R = ℤ / p^{n} ℤ$ for a prime $p$ . It is artinian, hence semilocal with nilpotent radical $(p)$ . Its only idempotents are $0$ and $1$ , since $x (x - 1) \equiv 0 (\mod p^{n})$ with $x$ and $x - 1$ coprime forces $p^{n} ∣ x$ or $p^{n} ∣ x - 1$ . Corollary $(21.29) (1)$ therefore certifies that $ℤ / p^{n}$ is local — as it is, with maximal ideal $(p)$ and residue field $𝔽_{p}$ .

Process and Workflow

Confirm the hypothesisCheck that

I

is nil. For a finite-dimensional algebra this is automatic once

I \subseteq rad A

Pick a representativeChoose any

a \in R

with

\bar{a}

the target idempotent. No care is needed at this stage; different choices give conjugate lifts.

Find the nilpotency indexCompute the least

m

with

(a - a^{2})^{m} = 0

Expand and cutForm

\sum_{k < m} (\binom{2 m}{k}) a^{2 m - k} (1 - a)^{k}

. This is the lift.

VerifyCheck

e^{2} = e

e - a \in I

, and

e \in a R

. All three are guaranteed but cheap to confirm.

Orthogonalise if neededWhen lifting several idempotents at once, apply the orthogonalisation lemma from Lifting Idempotents Modulo an Ideal after each lift.

Your right ideal $𝔄$ in a semilocal ring with nil radical — does it contain a nonzero idempotent?

Some element of $𝔄$ is not nilpotentYes. Reduce modulo the radical, take the idempotent generator of the image, and lift it into

𝔄

via

(21.28)

Every element of $𝔄$ is nilpotentNo. A nonzero idempotent is never nilpotent, so

𝔄

can contain none.

You cannot tellTest whether

𝔄 \subseteq rad R

. If yes,

𝔄

is nil; if no, its image in the semisimple quotient is a nonzero summand.

Comparison and Classification

Sufficient conditions for lifting, compared
	Constructive?	Lift lands in $a R$	Needs $R$ semilocal	Typical setting
$I$ nilpotent	yes	yes	no	Finite-dimensional algebras
$I$ nil	yes	yes	no	Semiprimary and artinian rings
$R$ is $I$ -adically complete	yes	no	no	Complete local rings, power series
$R$ semiperfect, $I = rad R$	no	no	yes	Definition, not a criterion

Sufficient conditions for lifting, compared

Nil versus nilpotent
Property	Nil ideal	Nilpotent ideal
Definition	Every element is nilpotent, exponent may vary	$I^{n} = 0$ for one fixed $n$
Contained in $rad R$	Yes	Yes
Idempotents lift	Yes, by (21.28)	Yes, as a special case
Bounded index	Not required	Required
Automatic for artinian $R$	Yes	Yes, by Hopkins–Levitzki
Closed under sums of ideals	Yes	Not in general for infinite sums

For a left artinian ring the two notions coincide on the radical, which is why the distinction is invisible in the finite-dimensional theory and becomes real only for infinitely generated examples.

Relationship Map

The hypotheses of this page sit inside the following hierarchy of classes of rings.

Semilocal

R / rad R

semisimple; idempotents may fail to lift

Semilocal with nil radical

(21.29)

applies: local criterion and the nil/idempotent dichotomy

Semiprimaryradical nilpotent, quotient semisimple

One-sided artinianradical nilpotent by Hopkins–Levitzki

Finite-dimensional algebrathe working case in computation

I

nilpotent

I

nil

I \subseteq rad R

The first arrow reverses only under extra finiteness; the second never reverses, as the semilocal domain $ℤ_{(p)} \cap ℤ_{(q)}$ shows. Lifting is guaranteed at the first two stages and not at the third.

Applications and Industry Use

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

Structure theory

Artinian and semiprimary rings

Since a one-sided artinian ring has nilpotent radical, idempotents always lift there. This is the reason every left artinian ring is semiperfect and admits a decomposition of $1$ into orthogonal primitive idempotents.

Computer algebra

Splitting an algebra

Wedderburn decomposition algorithms compute the radical, split the semisimple quotient into matrix blocks, and lift the block idempotents back through the nilpotent radical using exactly this construction or its Newton-style variant.

Modular representation theory

Block idempotents

Central idempotents of $k G$ in characteristic $p$ are lifted to a complete discrete valuation ring; the polynomial form of the lift is what keeps centrality intact along the way.

Coding theory

Idempotent generators of cyclic codes

A cyclic code of length $n$ over $𝔽_{q}$ is generated by an idempotent when $\gcd (n, q) = 1$ . In the repeated-root case the group algebra has nilpotent radical, and idempotent generators for the semisimple part are lifted by the same mechanism.

The honest description is that this theorem is a workhorse. It is the step that lets an entire subject compute in a semisimple quotient and then return, and it is invoked far more often than it is stated.

Computational Notes

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

Let $A$ be a finite-dimensional algebra over a field $k$ with $\dim_{k} A = n$ , and let $I \subseteq rad A$ with $I^{N} = 0$ .

The binomial construction needs $m \leq N$ and therefore up to $2 N$ multiplications in $A$ , each costing $O (n^{3})$ scalar operations with the naive algorithm on structure constants.
The Newton-style iteration $a \mapsto 3 a^{2} - 2 a^{3}$ reaches an exact idempotent in $lceil \log_{2} N rceil$ steps, with three multiplications per step. This is the version implemented in practice.
Both routines are numerically exact over a field; there is no stability question, and both work over $ℤ / p^{k}$ and over complete local rings truncated to finite precision.
Computing $rad A$ dominates the total cost. Over characteristic zero the trace-form method is $O (n^{3})$ ; in characteristic $p$ the Friedl–Rónyai style algorithms are more involved.
GAP, Magma and Sage all provide primitive idempotent computation for finite-dimensional algebras, and the lifting stage is invisible to the user.

Failure Modes and Common Mistakes

Do not confuse a nil ideal with a nilpotent one when the ring is not artinian; only the weaker hypothesis is needed here, but only the stronger one gives a uniform exponent for algorithms.
Independent lifts of orthogonal idempotents are not orthogonal. The construction says nothing about compatibility between two separate applications.
$(21.29) (2)$ is about right ideals; the left-handed statement is the mirror image, proved the same way, but the two are separate statements about separate objects.
In $(21.28)$ the conclusion $e \in a R$ is genuinely used later. Weakening the theorem to bare existence loses the ability to place the idempotent inside a prescribed right ideal.

Best Practices

State which hypothesis you are using: nil, nilpotent, or merely inside the radical. The three give different theorems and the middle one is what $(21.28)$ needs.
When you need the lift inside a prescribed right ideal, quote $(21.28)$ in the form that places $e$ in $a R$ rather than a bare existence statement.
If centrality matters, use the fact that the lift is a polynomial in $a$ ; that is the cheapest way to preserve commutation with a prescribed set.
When lifting several idempotents, lift first and orthogonalise second; trying to arrange orthogonality during the lift complicates the bookkeeping without benefit.
For a concrete finite ring, cross-check a computed lift against a brute-force enumeration of idempotents. The arithmetic in the binomial sum is easy to get wrong by one term.

Quick Reference

Theorem

I

nil,

\bar{a}

idempotent

\Rightarrow

there is an idempotent

e \in a R

with

\bar{e} = \bar{a}

The lift

e = \sum_{k < m} (\binom{2 m}{k}) a^{2 m - k} (1 - a)^{k}

where

(a - a^{2})^{m} = 0

UniquenessAny two lifts are conjugate by a unit of

1 + I

; commuting lifts are equal.

Corollary (1)Semilocal, radical nil, no nontrivial idempotents, nonzero

\Rightarrow

local.

Corollary (2)A right ideal contains a nonzero idempotent iff it is not nil.

ScopeApplies to every semiprimary ring and every one-sided artinian ring.

Checklist before applying the results
To conclude	You must know	Reference
An idempotent lifts	$I$ is nil	(21.28)
The lift lies in $a R$	Same hypothesis; it is part of the statement	(21.28)
$R$ is local	Semilocal, radical nil, only trivial idempotents, $R \neq 0$	(21.29)(1)
$𝔄$ has a nonzero idempotent	Semilocal, radical nil, $𝔄$ not nil	(21.29)(2)
Countable families lift orthogonally	Radical nil, plus the orthogonalisation lemma	(21.25) with (21.28)

Frequently Asked Questions

Why the exponent

2 m

rather than

m

m + 1

Because the cut has to leave every cross term divisible by $a^{m} b^{m}$ . With total degree $2 m$ and the split at $k = m$ , a term from the left half has $a$ -degree at least $m + 1$ and a term from the right half has $b$ -degree at least $m$ , so their product clears the threshold. Smaller total degree leaves cross terms of insufficient degree in one of the two variables.

Does the theorem need

R

to be commutative anywhere?

No. All computations take place in the commutative subring generated by $a$ and $1$ , so the binomial theorem applies without any hypothesis on $R$ . That is the whole trick: a statement about one element is a statement about a commutative ring.

Is the lift unique?

Not as an element, but almost. Any two idempotents congruent modulo an ideal inside the radical are conjugate by a unit lying in $1 + I$ , and if they happen to commute they are equal. So the lift is unique up to inner automorphism, and canonical in the commutative case.

Why is nil enough, when so many theorems require nilpotent?

The proof consumes nilpotency of a single element, namely $a - a^{2}$ , and allows its exponent to be whatever that element needs. Statements requiring a uniform bound — such as the existence of a fixed $N$ making an algorithm terminate in $\log N$ steps — do need nilpotence.

What exactly does

(21.29) (2)

buy that

(21.28)

does not?

The conclusion $e \in a R$ in $(21.28)$ is what confines the idempotent to a prescribed right ideal. Without it you could lift the idempotent generator of the image of $𝔄$ and land outside $𝔄$ , which would make the corollary vacuous.

Does

(21.29)

have a left-handed version?

Yes, and it is proved identically: a left ideal of a semilocal ring with nil radical contains a nonzero idempotent if and only if it is not nil. Part (1) is already side-neutral, because being local, being semilocal and the radical being nil are all left-right symmetric conditions.

How does this relate to the completeness criterion?

They are the two independent sufficient conditions in the theory. Nilness makes the correction series terminate; completeness makes it converge. Neither implies the other: a complete discrete valuation ring has non-nil radical, and a nilpotent ideal gives completeness trivially but is far more special.

References

T. Y. Lam, A First Course in Noncommutative Rings, Graduate Texts in Mathematics 131, Springer-Verlag, 1991, §21, (21.28)–(21.29) (pp. 329–330); nil ideals and the radical in §4, (4.11).
N. Jacobson, Structure of Rings, American Mathematical Society Colloquium Publications 37, revised edition, 1964, Chapter I.
I. N. Herstein, Noncommutative Rings, Carus Mathematical Monographs 15, Mathematical Association of America, 1968, Chapter 1.
F. W. Anderson and K. R. Fuller, Rings and Categories of Modules, 2nd edition, Graduate Texts in Mathematics 13, Springer-Verlag, 1992, §27.
W. Eberly and M. Giesbrecht, “Efficient decomposition of associative algebras over finite fields”, Journal of Symbolic Computation 29 (2000), 441–458.

AI Suggested Questions

Give a ring with a nil but non-nilpotent ideal and carry out the lifting construction explicitly in it.
Prove that the Newton iteration $a \mapsto 3 a^{2} - 2 a^{3}$ improves $a^{2} - a \in I$ to $b^{2} - b \in I^{2}$ .
Show that idempotents lift modulo the nil radical of a commutative ring, and identify when the lift is unique.
Does $(21.28)$ remain true for one-sided nil ideals, or is two-sidedness essential?
Work out the primitive idempotents of the group algebra of the symmetric group on three letters over $ℤ / 4$ by lifting from characteristic 2.
How does the Köthe conjecture interact with lifting idempotents modulo nil one-sided ideals?
Compare the binomial construction with the idempotent-splitting step in Wedderburn decomposition algorithms.
Prove Hopkins–Levitzki and deduce that every one-sided artinian ring satisfies the hypotheses of $(21.29)$ .

Lifting Modulo Nil Ideals

Executive Summary

Overview

Learning Objectives

Definitions

Core Concepts

Everything happens in a commutative subring

Why cutting at $k = m$ works

Nil is the right hypothesis, not nilpotent

Key Results

Proof Techniques and Method

Restrict to $ℤ [a]$

Split a partition of unity

Overshoot the degree

Worked Example

Lifting $3$ from $ℤ / 6$ to $ℤ / 12$

Non-nil right ideals contain idempotents

A local ring by $(21.29) (1)$

Process and Workflow

Comparison and Classification

Relationship Map

Applications and Industry Use

Artinian and semiprimary rings

Splitting an algebra

Block idempotents

Idempotent generators of cyclic codes

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

05Core Concepts

Everything happens in a commutative subring

Why cutting at k=m works

Nil is the right hypothesis, not nilpotent

06Key Results

07Proof Techniques and Method

Restrict to ℤ[a]

Split a partition of unity

Overshoot the degree

08Worked Example

Lifting 3 from ℤ/6 to ℤ/12

Non-nil right ideals contain idempotents

A local ring by (21.29)(1)

09Process and Workflow

10Comparison and Classification

11Relationship Map

12Applications and Industry Use

Artinian and semiprimary rings

Splitting an algebra

Block idempotents

Idempotent generators of cyclic codes

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

Core Concepts

Why cutting at $k = m$ works

Key Results

Proof Techniques and Method

Restrict to $ℤ [a]$

Worked Example

Lifting $3$ from $ℤ / 6$ to $ℤ / 12$

A local ring by $(21.29) (1)$

Process and Workflow

Comparison and Classification

Relationship Map

Applications and Industry Use

Computational Notes

Failure Modes and Common Mistakes

Best Practices

Quick Reference

Frequently Asked Questions

Related KEVOS Topics

References

AI Suggested Questions