The Upper Nilradical and the Köthe Conjecture

Q: Why is the upper nilradical called *upper*?

Because it is constructed from above, as the largest ideal with a nilness property, whereas the lower nilradical is constructed from below as the smallest semiprime ideal — equivalently the intersection of the prime ideals. The names record the direction of the construction, not the size, although Nil_* R ⊆ Nil^* R does hold.

Q: Is Köthe's Conjecture equivalent to a statement about a single ring?

No, and this is part of its difficulty. All of the standard formulations quantify over all rings; a counterexample would be a single ring with a nonzero nil one-sided ideal and no nonzero nil ideal, but proving the conjecture requires an argument valid for every ring at once. Krempa's matrix and polynomial reformulations are likewise universally quantified.

Q: What is the relationship with the Jacobson radical?

Every nil one-sided ideal lies in rad R, by (4.11) — that direction is unconditional and easy, since a nilpotent element a makes 1 - a invertible. What is missing is the finer statement that a nil left ideal lies in a nil *ideal*. The Jacobson radical is simply too coarse to detect the difference.

Q: Does the conjecture matter outside pure radical theory?

It controls a genuine gap in how one may argue. Without it, a nil left ideal cannot be enlarged to a nil two-sided ideal, so it cannot be quotiented away; every theorem needing that step must assume a chain condition or a PI hypothesis instead. That is why the Levitzki radical — defined by locally nilpotent ideals, which do behave well one-sidedly — was introduced.

Q: Would a proof of the conjecture simplify the theory much?

Substantially. The upper nilradical would become a genuine radical in the one-sided sense, the equality Nil^* M_n(R) = M_n(Nil^* R) would restore Morita invariance, and Amitsur's description of rad R[t] would sharpen to an explicit formula. A counterexample would be equally informative, and would presumably come from the Golod–Shafarevich circle of constructions.

Executive Summary

Nil ideals behave well as long as they are two-sided. Their sum is nil, so a largest one exists: the upper nilradical ${Nil}^{*} R$ . It sits between the prime radical and the Jacobson radical and coincides with both in the classical cases — commutative rings and left artinian rings.

The moment one-sided ideals are admitted, the theory stops. It is not known whether the sum of two nil left ideals is nil, whether a nil left ideal lies in a nil two-sided ideal, or whether a ring with no nil ideals can carry a nil left ideal. These are the same question, posed by Köthe in 1930 and still open. Every structural statement about nil one-sided ideals in this collection is therefore hedged by a chain condition.

1930Köthe's question

OpenStatus today

${Nil}^{*} R$ Largest nil ideal

$\subseteq rad R$ Always

Overview

A ring has two natural nil-flavoured radicals. The lower nilradical ${Nil}_{*} R$ is built from below, as the intersection of prime ideals, and is the smallest semiprime ideal. The upper nilradical ${Nil}^{*} R$ is built from above, as the sum of all nil ideals, and is the largest nil ideal. They agree in the commutative case and disagree in general.

{Nil}_{*} R \subseteq {Nil}^{*} R \subseteq rad R

(10.27)

The first inclusion holds because ${Nil}_{*} R$ is a nil ideal; the second because every nil one-sided ideal lies in the Jacobson radical, by $(4.11)$ .

The asymmetry is worth naming precisely. Nilpotent one-sided ideals cause no trouble at all: a nilpotent left ideal generates a nilpotent two-sided ideal, and semiprimeness kills them outright. Nil-ness is what breaks, because there is no uniform exponent to work with.

Learning Objectives

Prove $(10.25)$ : a nil left ideal plus a nil ideal is a nil left ideal.
Deduce that ${Nil}^{*} R$ exists and equals ${a \in R : R a R is nil}$ .
Prove ${Nil}_{*} R \subseteq {Nil}^{*} R \subseteq rad R$ and the collapse in the commutative and left artinian cases.
State Köthe's Conjecture in the three forms $(10.28)$ , $(10.28 a)$ , $(10.28 b)$ and prove they are equivalent.
Show that a ring with an involution satisfying $a^{*} a = 0 \Rightarrow a = 0$ has no nonzero nil one-sided ideal.
Name the classes of rings for which the conjecture is known to hold.

Definitions

Definition(10.26)Upper nilradical

${Nil}^{*} R$ denotes the sum of all nil ideals of $R$ . By $(10.25)$ this sum is itself nil, so ${Nil}^{*} R$ is the largest nil ideal of $R$ , and element-wise ${Nil}^{*} R = {a \in R : R a R is a nil ideal}$ .

Nil: Every element is nilpotent, with no uniform bound on the exponents.
Nilpotent: $𝔄^{n} = 0$ for some fixed $n$ ; strictly stronger than nil.
${Nil}_{*} R$: Lower nilradical: the intersection of all prime ideals, the smallest semiprime ideal.
${Nil}^{*} R$: Upper nilradical: the largest nil ideal.
$L - rad R$: Levitzki radical: the largest locally nilpotent ideal; it sits between the two nilradicals.
$rad R$: Jacobson radical; contains every nil one-sided ideal by $(4.11)$ .

The starred notation is Lam's: the subscript marks the lower, the superscript the upper radical. Other sources write N(R) or Nil(R) for one or both, so check conventions before quoting.

Core Concepts

Why two-sidedness rescues the sum

Let $𝔄$ be a nil left ideal and $𝔅$ a nil ideal. To see that $𝔄 + 𝔅$ is nil, work modulo $𝔅$ : the image of $𝔄 + 𝔅$ in $R / 𝔅$ is the image of $𝔄$ , hence nil, so any $c \in 𝔄 + 𝔅$ has $c^{n} \in 𝔅$ for some $n$ ; nilness of $𝔅$ then finishes the job with a second exponent.

The step *work modulo $𝔅$ * requires $𝔅$ to be two-sided, so that $R / 𝔅$ is a ring. If both ideals are merely left ideals there is no quotient ring to pass to, and the argument evaporates. That single gap is the whole of Köthe's problem.

c \in 𝔄 + 𝔅

\bar{c}

nil in

R / 𝔅

c^{n} \in 𝔅

(c^{n})^{m} = 0

Left and right are interchangeable

For any $x \in R$ , the principal left ideal $R x$ is nil if and only if the principal right ideal $x R$ is nil: if $(r x)^{n} = 0$ then $(x r)^{n + 1} = x (r x)^{n} r = 0$ , and symmetrically. Consequently a ring has no nonzero nil left ideal precisely when it has no nonzero nil right ideal, and one may speak unambiguously of a ring without nonzero nil one-sided ideals.

Key Results

Lemma(10.25)Sums of nil ideals

Let $𝔄$ be a nil left ideal and $𝔅$ a nil two-sided ideal in a ring $R$ . Then $𝔄 + 𝔅$ is a nil left ideal.

Proof

$𝔄 + 𝔅$ is a left ideal since both summands are. Let $c \in 𝔄 + 𝔅$ and pass to $\bar{R} = R / 𝔅$ , a ring because $𝔅$ is two-sided. The image $\bar{𝔄 + 𝔅} = \bar{𝔄}$ is a nil left ideal of $\bar{R}$ , being a homomorphic image of a nil left ideal. Hence ${\bar{c}}^{n} = 0$ for some $n \geq 1$ , i.e. $c^{n} \in 𝔅$ . Since $𝔅$ is nil, $(c^{n})^{m} = 0$ for some $m \geq 1$ , so $c^{n m} = 0$ .

Applying the lemma with both ideals two-sided shows that the sum of any family of nil ideals is nil: an element of the sum lies in a sum of finitely many of them, and induction applies.

Proposition(10.27)Position of the upper nilradical

For any ring $R$ , ${Nil}_{*} R \subseteq {Nil}^{*} R \subseteq rad R$ . If $R$ is commutative, all three coincide with $Nil (R)$ , the set of nilpotent elements. If $R$ is left artinian, all three coincide with $rad R$ , which is then nilpotent.

Proof

${Nil}_{*} R = \sqrt{(0)}$ consists of elements every $m$ -system through which meets $(0)$ , and is in particular a nil ideal; being an ideal it lies in the sum of all nil ideals, giving the first inclusion. The second is $(4.11)$ : every nil one-sided ideal, in particular the nil ideal ${Nil}^{*} R$ , is contained in $rad R$ .

If $R$ is commutative, $Nil (R)$ is an ideal, it is nil, and it is the intersection of the prime ideals; so it is simultaneously the largest nil ideal and the smallest semiprime ideal, forcing ${Nil}_{*} R = {Nil}^{*} R = Nil (R)$ . (The Jacobson radical may be strictly larger — take $R = k [[x]]$ .)

If $R$ is left artinian, $rad R$ is a nilpotent ideal by $(4.12)$ . The quotient $R / {Nil}_{*} R$ is semiprime and therefore has no nonzero nilpotent ideal, so the image of $rad R$ there is zero, i.e. $rad R \subseteq {Nil}_{*} R$ . Combined with the chain above, all three radicals agree.

Proposition—Rings with a positive-definite involution

Let $R$ be a ring with an involution $*$ such that $a^{*} a = 0$ implies $a = 0$ . Then $R$ has no nonzero nil one-sided ideal; in particular ${Nil}^{*} R = 0$ and Köthe's Conjecture holds vacuously for $R$ .

Proof

Let $𝔄$ be a nil left ideal and $a \in 𝔄$ . Then $b = a^{*} a \in R 𝔄 \subseteq 𝔄$ is nilpotent and satisfies $b^{*} = b$ . Choose $n \geq 1$ minimal with $b^{n} = 0$ and suppose $n \geq 2$ . Put $c = b^{n - 1}$ ; then $c^{*} c = b^{n - 1} b^{n - 1} = b^{2 n - 2} = 0$ because $2 n - 2 \geq n$ , so the hypothesis gives $c = b^{n - 1} = 0$ , contradicting minimality. Hence $n = 1$ , so $a^{*} a = b = 0$ and therefore $a = 0$ .

The hypothesis holds for any $*$ -closed ring of bounded operators on a complex Hilbert space, since $T^{*} T = 0$ gives $| T v |^{2} = ⟨ T^{*} T v, v ⟩ = 0$ for all $v$ . It also holds for the group ring $ℂ G$ of an arbitrary group under $\sum a_{g} g \mapsto \sum \bar{a_{g}} g^{- 1}$ , since the coefficient of $1$ in $α^{*} α$ is $\sum_{g} | a_{g} |^{2}$ ; this is how $(6.11)$ produces ${Nil}^{*} ℂ G = 0$ .

Conjecture(10.28)Köthe

If ${Nil}^{*} R = 0$ , then $R$ has no nonzero nil one-sided ideal. Equivalently, over all rings:

$(10.28 a)$ every nil left or right ideal of a ring $R$ is contained in ${Nil}^{*} R$ ;
$(10.28 b)$ the sum of two nil left ideals of a ring $R$ is nil (equivalently for right ideals).

ProofEquivalence of the three forms

** $(10.28 a) \Rightarrow (10.28)$ ** is immediate: if ${Nil}^{*} R = 0$ then every nil one-sided ideal is contained in $0$ . Conversely, given $(10.28)$ for all rings, apply it to $\bar{R} = R / {Nil}^{*} R$ , whose upper nilradical vanishes — a nil ideal of $\bar{R}$ pulls back by $(10.25)$ to a nil ideal of $R$ , hence into ${Nil}^{*} R$ . A nil left ideal $𝔄 \subseteq R$ has nil image in $\bar{R}$ , so that image is zero, i.e. $𝔄 \subseteq {Nil}^{*} R$ .

** $(10.28 a) \Rightarrow (10.28 b)$ **: both summands lie in ${Nil}^{*} R$ , whose elements are nilpotent, so the sum is contained in a nil ideal.

** $(10.28 b) \Rightarrow (10.28)$ **: for $x, r \in R$ , $R x$ nil implies $R (x r)$ nil, since $(s x r)^{n + 1} = s x (r s x)^{n} r$ and $r s x \in R x$ . Hence the sum $𝔖$ of all nil left ideals is closed under right multiplication, so it is a two-sided ideal, and it coincides with the sum of all nil right ideals. Under $(10.28 b)$ every finite sum of nil left ideals is nil, so every element of $𝔖$ lies in a nil left ideal and $𝔖$ is nil. Being a nil ideal, $𝔖 \subseteq {Nil}^{*} R$ , which is $(10.28 a)$ and hence $(10.28)$ .

Theorem(10.30)Known cases

Right noetherian rings. Levitzki's Theorem: in a right noetherian ring every nil one-sided ideal is nilpotent, and ${Nil}_{*} R = {Nil}^{*} R$ is the largest nilpotent one-sided ideal. Left artinian rings are covered a fortiori.
Rings with ACC on right annihilators. Utumi's Lemma $(10.29)$ places every nil one-sided ideal inside ${Nil}_{*} R$ , which is stronger than the conjecture demands.
**Algebras with $rad$ nil.** If $R$ is an algebra over a field $k$ that is algebraic over $k$ , or satisfies $\dim_{k} R < | k |$ , then $rad R$ is nil by $(4.19)$ and $(4.20)$ , so ${Nil}^{*} R = rad R$ and every nil one-sided ideal — being inside $rad R$ — is inside ${Nil}^{*} R$ .
PI-algebras over an arbitrary field; for finitely generated PI-algebras Braun's theorem gives the stronger conclusion that ${Nil}_{*} R = {Nil}^{*} R = rad R$ is nilpotent.

RemarkExercise 25Krempa's reformulations

Köthe's Conjecture is also equivalent to each of the following, quantified over all rings $R$ and all $n$ : (a) if $I$ is a nil ideal of $R$ then $M_{n} (I)$ is a nil ideal of $M_{n} (R)$ ; (b) ${Nil}^{*} (M_{n} (R)) = M_{n} ({Nil}^{*} R)$ ; (c) if $I$ is a nil ideal then $I [t] \subseteq rad R [t]$ ; (d) $rad (R [t]) = ({Nil}^{*} R) [t]$ .

Form (a) already fails to be obvious for $n = 2$ : it asks whether a $2 \times 2$ matrix over a nil ring is nilpotent, and the companion-matrix argument of Krempa and Amitsur is the only known route from (a) to (c). Form (d), if true, would sharpen Amitsur's description of $rad R [t]$ considerably.

Proof Techniques and Method

The reusable moves behind the proofs above.

Move 1

Quotient by the two-sided piece

To handle a sum $𝔄 + 𝔅$ , kill $𝔅$ and work in $R / 𝔅$ . Two exponents, one from each stage, combine multiplicatively. This works only when the ideal being killed is two-sided.

Move 2

Shift a product cyclically

$(x r)^{n + 1} = x (r x)^{n} r$ converts a statement about $R x$ into one about $x R$ at the cost of one factor. It is the reason nil-ness is side-neutral for principal one-sided ideals even though the conjecture is open.

Move 3

Minimise a nilpotency index

Given $b^{n} = 0$ with $n$ minimal, test $b^{n - 1}$ : the exponent $2 n - 2$ already exceeds $n$ . Used here with an involution to force $n = 1$ , and on the semiprime pages to collapse nilpotent ideals.

Move 3 is available for nilpotent objects and unavailable for nil ones — there is no index to minimise when each element has its own exponent. Almost every open problem in this area sits exactly at that boundary.

Worked Example

A nil radical that is not nilpotent

Let $k$ be a field and set $R = k [x_{1}, x_{2}, x_{3}, \dots] / (x_{i}^{i + 1} : i \geq 1)$ , a commutative ring. Write $𝔪 = (x_{1}, x_{2}, \dots)$ for the ideal of elements with zero constant term.

$𝔪$ is nil: any $a \in 𝔪$ involves finitely many variables $x_{1}, \dots, x_{n}$ , and every monomial of total degree exceeding $\sum_{i \leq n} (i + 1)$ vanishes, so $a$ is nilpotent.
$𝔪$ is not nilpotent: $x_{n}^{m} \neq 0$ whenever $m \leq n$ , so $𝔪^{m} \neq 0$ for every $m$ .
$R$ is local with maximal ideal $𝔪$ , since an element with nonzero constant term is a unit plus a nilpotent.

{Nil}_{*} R = L - rad R = {Nil}^{*} R = rad R = 𝔪, 𝔪^{m} \neq 0 \forall m .

(E.1)

All four radicals coincide because $R$ is commutative and local — but the common value is nil without being nilpotent.

A ring where the upper nilradical is strictly smaller than the Jacobson radical

Let $R = k [[x]]$ , or equally $R = ℤ_{(p)}$ . Both are commutative domains, so ${Nil}_{*} R = {Nil}^{*} R = 0$ , while $rad R$ is the unique maximal ideal $(x)$ , respectively $p ℤ_{(p)}$ . The last inclusion of $(10.27)$ is therefore strict in the strongest possible way: the two ends of the chain are $0$ and a nonzero ideal.

A collapse

For $R = T_{2} (k)$ , finite-dimensional over $k$ hence left artinian, $(10.27)$ forces ${Nil}_{*} R = {Nil}^{*} R = rad R = 𝔑$ , the strictly upper triangular matrices, and $𝔑^{2} = 0$ . Left artinian rings are exactly where all the distinctions on this page disappear — which is why the interesting examples are all infinite-dimensional.

Comparison and Classification

Where Köthe's Conjecture is known
	Conjecture known?	Reason	Stronger conclusion available?
Left artinian	yes	all radicals equal $rad R$ , nilpotent	yes
Right noetherian	yes	Levitzki's Theorem $(10.30)$	yes — nil one-sided ideals are nilpotent
ACC on right annihilators	yes	Utumi's Lemma $(10.29)$	yes — nil one-sided ideals sit in ${Nil}_{*} R$
Algebraic algebra over a field	yes	$rad R$ is nil, by $(4.19)$	no
PI-algebra over a field	yes	Second Course; Braun for the f.g. case	yes in the finitely generated case
Commutative	yes	one-sided ideals are two-sided	yes
General ring	open	no method known	no

Where Köthe's Conjecture is known

The four radicals on the examples of this page
Ring	${Nil}_{*}$	${Nil}^{*}$	$rad$
$k [[x]]$	$0$	$0$	$(x)$
$ℤ_{(p)}$	$0$	$0$	$p ℤ_{(p)}$
$k [x_{i}] / (x_{i}^{i + 1})$	$𝔪$	$𝔪$	$𝔪$ (nil, not nilpotent)
$T_{2} (k)$	$𝔑$	$𝔑$	$𝔑$ (nilpotent)
$M_{n} (D)$	$0$	$0$	$0$
$*$ -closed operator algebra	$0$	$0$	may be nonzero

Relationship Map

{Nil}_{*} R

L - rad R

{Nil}^{*} R

rad R

All three inclusions are strict for suitable rings; the middle one requires Golod's finitely generated nil algebra that is not nilpotent, hence not locally nilpotent. The full comparison is laid out in The Radicals of a Ring Compared and The Levitzki Radical and Locally Nilpotent Ideals.

You have a nil one-sided ideal $𝔄$ . What can you conclude?

$𝔄$ is two-sided

𝔄 \subseteq {Nil}^{*} R \subseteq rad R

, unconditionally.

$𝔄$ is one-sided, $R$ arbitraryOnly

𝔄 \subseteq rad R

, by

(4.11)

. Whether

𝔄 \subseteq {Nil}^{*} R

is Köthe's Conjecture.

$𝔄$ one-sided, $R$ has ACC on right annihilators

𝔄 \subseteq {Nil}_{*} R

by Utumi's Lemma — the strongest possible conclusion.

$𝔄$ one-sided, $R$ right noetherian

𝔄

is nilpotent, by Levitzki's Theorem.

Computational Notes

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

For a finite-dimensional algebra over a field, ${Nil}^{*} A = rad A$ , so the radical routines in GAP, Magma and Sage compute the upper nilradical directly; nothing about Köthe is visible in that regime.
For a finitely presented associative algebra there is no algorithm: the word problem is already undecidable, so deciding whether a given element is nilpotent — let alone whether an ideal is nil — is out of reach in general.
Golod's algebras are given by explicit generators and relations chosen so that the Golod–Shafarevich inequality forces infinite dimension; they are computable examples, and truncating them at a fixed degree is a standard way to experiment.
Krempa's form (a) suggests a finite test: is a $2 \times 2$ matrix over a nil ring nilpotent? Concretely one can search for counterexamples in nilpotent-by-degree truncations, but no truncation can refute the conjecture, since nil-ness is not a finitary condition.

Failure Modes and Common Mistakes

Do not assume the sum of two nil left ideals is a nil left ideal; that is the conjecture, not a lemma.
Do not read $(10.25)$ as symmetric in its two arguments — one of the ideals must be two-sided for the quotient argument to exist.
Do not conclude ${Nil}^{*} R = rad R$ without a hypothesis; it needs left artinian, or an Amitsur-type condition making $rad R$ nil.
Do not confuse the two stars: ${Nil}_{*}$ is the smallest semiprime ideal, ${Nil}^{*}$ the largest nil ideal, and the inclusion runs from subscript to superscript.

Best Practices

State which radical you mean by name and symbol on first use; four candidates are in play and the notation varies across the literature.
When a proof uses a nil one-sided ideal, record whether it also needs the ideal to be two-sided — that single distinction decides whether the argument is unconditional.
Prefer hypotheses that force $rad R$ to be nil (algebraicity, PI, chain conditions); they make all the nil questions collapse and are usually checkable.
When quoting a result known under Köthe's Conjecture, mark it as conditional; several results in the literature are stated without that qualification.

Historical Notes and Lessons Learned

1930Köthe poses the questionIn his paper on rings whose quotient by the radical is completely reducible, Köthe asks whether a ring with no nonzero nil ideal can have a nonzero nil one-sided ideal.
1939 / 1950Levitzki settles the noetherian caseLevitzki proves that nil one-sided ideals of a right noetherian ring are nilpotent. The proof was delayed by the war and appeared only in 1950, with a flaw that persisted into the standard references.
1956Amitsur on polynomial radicalsAmitsur's description of $rad R [t]$ as $N [t]$ with $N$ nil supplies the link between the conjecture and polynomial rings that Krempa later exploits.
1964Golod's nil algebrasGolod, using the Golod–Shafarevich inequality, constructs finitely generated infinite-dimensional nil algebras. These separate nil from locally nilpotent and show that no local-finiteness argument can settle the conjecture.
1972Krempa's equivalencesKrempa reduces the conjecture to matrix and polynomial statements: nil-ness of $M_{n} (I)$ , and $rad R [t] = ({Nil}^{*} R) [t]$ . The problem acquires several independent-looking faces.
2000Smoktunowicz's nil ringSmoktunowicz constructs a nil ring $N$ with $N [x]$ not nil, refuting Amitsur's related conjecture. Köthe's Conjecture survives, but the example shows how badly nil-ness behaves under polynomial extension.

The lesson is that every plausible strengthening has been tested and several have failed. What remains open is unusually narrow: not whether nil rings are well behaved — they are not — but whether one-sided nil-ness can escape a two-sided nil ideal.

Quick Reference

Definition

{Nil}^{*} R =

sum of all nil ideals

=

largest nil ideal

Element form

a \in {Nil}^{*} R iff R a R

is nil

Chain

{Nil}_{*} R \subseteq L - rad R \subseteq {Nil}^{*} R \subseteq rad R

Commutative

{Nil}_{*} R = {Nil}^{*} R = Nil (R)

Left artinianall four radicals equal

rad R

, which is nilpotent

Köthe

{Nil}^{*} R = 0 \Rightarrow

no nonzero nil one-sided ideal

Equivalentsum of two nil left ideals is nil;

{Nil}^{*} M_{n} (R) = M_{n} ({Nil}^{*} R)

Side neutrality

R x

nil

iff x R

nil, for every

x

Statements, status, source
Statement	Status	Reference
Sum of nil ideals is nil	theorem	(10.25)
${Nil}^{*} R$ is the largest nil ideal	theorem	(10.26)
${Nil}_{} \subseteq {Nil}^{} \subseteq rad$	theorem	(10.27)
Sum of two nil left ideals is nil	open	(10.28b)
$rad R [t] = ({Nil}^{*} R) [t]$	open, equivalent to Köthe	Exercise 25(d)
$R$ nil $\Rightarrow R [x]$ nil	false	Smoktunowicz 2000

Frequently Asked Questions

Why is the upper nilradical called upper?

Because it is constructed from above, as the largest ideal with a nilness property, whereas the lower nilradical is constructed from below as the smallest semiprime ideal — equivalently the intersection of the prime ideals. The names record the direction of the construction, not the size, although ${Nil}_{*} R \subseteq {Nil}^{*} R$ does hold.

Is Köthe's Conjecture equivalent to a statement about a single ring?

No, and this is part of its difficulty. All of the standard formulations quantify over all rings; a counterexample would be a single ring with a nonzero nil one-sided ideal and no nonzero nil ideal, but proving the conjecture requires an argument valid for every ring at once. Krempa's matrix and polynomial reformulations are likewise universally quantified.

What is the relationship with the Jacobson radical?

Every nil one-sided ideal lies in $rad R$ , by $(4.11)$ — that direction is unconditional and easy, since a nilpotent element $a$ makes $1 - a$ invertible. What is missing is the finer statement that a nil left ideal lies in a nil ideal. The Jacobson radical is simply too coarse to detect the difference.

Does the conjecture matter outside pure radical theory?

It controls a genuine gap in how one may argue. Without it, a nil left ideal cannot be enlarged to a nil two-sided ideal, so it cannot be quotiented away; every theorem needing that step must assume a chain condition or a PI hypothesis instead. That is why the Levitzki radical — defined by locally nilpotent ideals, which do behave well one-sidedly — was introduced.

Would a proof of the conjecture simplify the theory much?

Substantially. The upper nilradical would become a genuine radical in the one-sided sense, the equality ${Nil}^{*} M_{n} (R) = M_{n} ({Nil}^{*} R)$ would restore Morita invariance, and Amitsur's description of $rad R [t]$ would sharpen to an explicit formula. A counterexample would be equally informative, and would presumably come from the Golod–Shafarevich circle of constructions.

References

T. Y. Lam, A First Course in Noncommutative Rings, Graduate Texts in Mathematics 131, Springer-Verlag, 1991, §10, (10.25)–(10.28) and Exercise 25.
G. Köthe, “Die Struktur der Ringe, deren Restklassenring nach dem Radikal vollständig reduzibel ist”, Mathematische Zeitschrift 32 (1930), 161–186.
J. Krempa, “Logical connections between some open problems concerning nil rings”, Fundamenta Mathematicae 76 (1972), 121–130.
E. S. Golod, “On nil-algebras and finitely approximable p-groups”, Izvestiya Akademii Nauk SSSR, Seriya Matematicheskaya 28 (1964), 273–276.
A. Smoktunowicz, “Polynomial rings over nil rings need not be nil”, Journal of Algebra 233 (2000), 427–436.
N. J. Divinsky, Rings and Radicals, University of Toronto Press, 1965.

AI Suggested Questions

Sketch Golod's construction of a finitely generated nil algebra that is not nilpotent.
Explain the companion-matrix argument of Krempa and Amitsur linking $M_{n} (I)$ nil to $I [t] \subseteq rad R [t]$ .
For which classes of group rings is ${Nil}^{*} k G = 0$ known, and what role does the involution $g \mapsto g^{- 1}$ play?
How does Smoktunowicz's example fail to contradict Köthe's Conjecture?
Give a ring in which ${Nil}_{*} R ⊊ {Nil}^{*} R$ and identify both radicals explicitly.
What is known about Köthe's Conjecture for algebras over uncountable fields?

Upper Nilradical and Köthe’s Conjecture

Executive Summary

Overview

Learning Objectives

Definitions

Core Concepts

Why two-sidedness rescues the sum

Left and right are interchangeable

Key Results

Proof Techniques and Method

Quotient by the two-sided piece

Shift a product cyclically

Minimise a nilpotency index

Worked Example

A nil radical that is not nilpotent

A ring where the upper nilradical is strictly smaller than the Jacobson radical

A collapse

Comparison and Classification

Relationship Map

Computational Notes

Failure Modes and Common Mistakes

Best Practices

Historical Notes and Lessons Learned

Quick Reference

Frequently Asked Questions

References

AI Suggested Questions

Continue learning

01Executive Summary

02Overview

03Learning Objectives

04Definitions

05Core Concepts

Why two-sidedness rescues the sum

Left and right are interchangeable

06Key Results

07Proof Techniques and Method

Quotient by the two-sided piece

Shift a product cyclically

Minimise a nilpotency index

08Worked Example

A nil radical that is not nilpotent

A ring where the upper nilradical is strictly smaller than the Jacobson radical

A collapse

09Comparison and Classification

10Relationship Map

11Computational Notes

12Failure Modes and Common Mistakes

13Best Practices

14Historical Notes and Lessons Learned

15Quick Reference

16Frequently Asked Questions

17Related KEVOS Topics

18References

19AI Suggested Questions

Continue learning

Executive Summary

Overview

Learning Objectives

Definitions

Core Concepts

Key Results

Proof Techniques and Method

Worked Example

Comparison and Classification

Relationship Map

Computational Notes

Failure Modes and Common Mistakes

Best Practices

Historical Notes and Lessons Learned

Quick Reference

Frequently Asked Questions

Related KEVOS Topics

References

AI Suggested Questions