Free Rings and Rings Defined by Generators and Relations

Q: How can k x,y contain a free ring on infinitely many generators?

Because free monoids are self-similar. The words xy^0, xy^1, xy^2, … have the property that any concatenation of them can be uniquely decomposed back, so they generate freely. The commutative analogue fails because k[x,y] has transcendence degree 2 and any three elements satisfy a polynomial relation.

Q: Is k x,y a domain?

Yes when k is. Order words by degree and then lexicographically; the leading word of a product is the concatenation of the leading words, with coefficient the product of the leading coefficients, which is nonzero since k is a domain. The same ordering computes the units and gives the noncommutative division algorithm used in Gröbner theory.

Q: Why is the Weyl algebra simple in characteristic zero but not in characteristic p?

Simplicity comes from the commutator identity [,x,y^,n] = n,y^,n-1: bracketing with x or y lowers degree and never annihilates a nonzero element, so any nonzero ideal eventually contains a nonzero scalar. In characteristic p the factor n vanishes when p n, the process stalls, and indeed x^,p and y^,p generate a proper ideal.

Executive Summary

Free rings are the raw material of noncommutative ring theory. Every ring given by a finite list of generators and relations is a quotient of one, so results about free rings become results about presentations, and pathological examples are most cheaply built by choosing relations that force pathology.

Two features distinguish $k ⟨ x, y ⟩$ from $k [x, y]$ sharply. It is not noetherian on either side. And it contains a copy of the free ring on countably many generators — nothing analogous happens for commuting variables, where transcendence degree is an obstruction.

WordsThe

k

-basis

$k ⟨ x, y ⟩$ Contains

k ⟨ z_{0}, z_{1}, \dots ⟩

Not noetherianOn either side

$x y - y x = 1$ One relation gives the Weyl algebra

Overview

Let $k$ be a ring and ${x_{i} : i \in I}$ a set of symbols. The **free $k$ -ring** $k ⟨ x_{i} : i \in I ⟩$ consists of finite $k$ -linear combinations of words — finite strings $x_{i_{1}} x_{i_{2}} \dots x_{i_{m}}$ , including the empty word $1$ — multiplied by concatenation, with coefficients required to commute with every $x_{i}$ .

Equivalently, it is the semigroup ring $k Σ$ where $Σ$ is the free monoid on ${x_{i}}$ . That identification is worth making early: it puts free rings in the same family as group rings and gives their basis for free.

Given a subset $F \subseteq k ⟨ x_{i} ⟩$ , the quotient $k ⟨ x_{i} ⟩ / (F)$ by the two-sided ideal generated by $F$ is *the ring generated over $k$ by the $x_{i}$ subject to the relations $F$ *. The universal property survives: homomorphisms out of the quotient are homomorphisms out of the free ring that kill each relation.

Note that when $k$ is noncommutative the word algebra is a mild abuse — there is no central base — but the construction is unaffected and the terminology is entrenched, particularly for Weyl algebras.

Learning Objectives

Construct $k ⟨ x_{i} : i \in I ⟩$ and identify the words as a $k$ -basis.
State and use the universal property, including the condition that images commute with the image of $k$ .
Prove that the elements $x y^{i}$ generate a free subring of $k ⟨ x, y ⟩$ on countably many generators.
Show that $k ⟨ x, y ⟩$ is neither left nor right noetherian.
Present $ℍ$ , the polynomial ring and the Weyl algebra $A_{1} (k)$ by generators and relations.
Use specialisation to show that $\bar{y} \bar{x} \neq 0$ in $ℤ ⟨ x, y ⟩ / (x y)$ .

Definitions

Definition(1.2)Free

k

-ring

Let $k$ be a ring and $Σ$ the free monoid on ${x_{i} : i \in I}$ . Set $R = k ⟨ x_{i} : i \in I ⟩ = ⨁_{w \in Σ} k w$ , a free left $k$ -module on the words, with multiplication extending concatenation $k$ -bilinearly and subject to $x_{i} c = c x_{i}$ for all $c \in k$ .

Definition(1.2)Universal property

Let $ϕ_{0} : k \to k'$ be a ring homomorphism and let ${a_{i} : i \in I} \subseteq k'$ be elements such that each $a_{i}$ commutes with every element of $ϕ_{0} (k)$ . Then there is a unique ring homomorphism

ϕ : k ⟨ x_{i} : i \in I ⟩ ⟶ k' with ϕ |_{k} = ϕ_{0} and ϕ (x_{i}) = a_{i} .

(1.2a)

The commuting hypothesis is essential and is exactly the relation $x_{i} c = c x_{i}$ imposed in the construction; without it no extension need exist.

Word: An element of the free monoid $Σ$ ; the empty word is the identity. Words form a $k$ -basis of the free ring.
$(F)$: The two-sided ideal generated by a subset $F$ : all finite sums $\sum g_{j} f_{j} h_{j}$ with $f_{j} \in F$ .
$A_{1} (k)$: The first Weyl algebra $k ⟨ x, y ⟩ / (x y - y x - 1)$ .
$A_{n} (k)$: The $n$ th Weyl algebra, defined inductively by $A_{n} (k) = A_{1} (A_{n - 1} (k))$ .
Generic relation: A relation imposed on free generators so that nothing beyond its consequences holds; e.g. $x y = 0$ makes $x$ a left zero-divisor and nothing more.
Total degree: The length of a word; it makes the free ring graded, with the degree- $n$ part free of rank $| I |^{n}$ when $I$ is finite.

For $| I | = 1$ the free ring is the ordinary polynomial ring $k [x]$ , since there is only one word of each length.

Core Concepts

Grading, degree and the domain property

The free ring is graded by word length, $R = ⨁_{n \geq 0} R_{n}$ with $R_{n}$ free on the $| I |^{n}$ words of length $n$ . If $k$ is a domain then so is $k ⟨ x_{i} ⟩$ : order the words by degree and then lexicographically, and observe that the leading words of $f$ and $g$ concatenate to the leading word of $f g$ with coefficient the product of the leading coefficients.

That argument also determines the units. If $f g = 1$ with $k$ a domain, comparing degrees gives $\deg f = \deg g = 0$ , so $U (k ⟨ x_{i} ⟩) = U (k)$ . The domain hypothesis is not decorative — see the failure recorded in Failure Modes.

Free rings are enormous

Two variables already generate everything. Setting $z_{i} = x y^{i}$ for $i \geq 0$ , the subring of $k ⟨ x, y ⟩$ generated over $k$ by ${z_{i}}$ is free on those countably many generators, because a product $z_{i_{1}} \dots z_{i_{m}}$ spells out the word $x y^{i_{1}} x y^{i_{2}} \dots x y^{i_{m}}$ , from which the exponents can be read back uniquely.

k ⟨ x, y ⟩

k ⟨ z_{0}, \dots, z_{n} ⟩

k ⟨ z_{0}, z_{1}, \dots ⟩

Nothing like this holds for $k [x, y]$ : a polynomial subring of $k [x, y]$ has at most two algebraically independent generators. The distinction is that concatenation of words is free, so there is unlimited room inside a two-letter alphabet.

Presentations and what they can hide

Writing $R = k ⟨ X ⟩ / (F)$ is easy; deciding what $R$ looks like is not. Two elements of $k ⟨ X ⟩$ may or may not become equal in $R$ , and there is no algorithm that decides this for arbitrary finite $F$ . The practical response is twofold: find a normal form for words modulo $F$ — the diamond lemma — or map $R$ into a concrete ring and compute there.

Key Results

Proposition(1.2)A free subring on countably many generators

Let $k$ be a ring and $R = k ⟨ x, y ⟩$ . For $i \geq 0$ put $z_{i} = x y^{i}$ . Then the $k$ -subring of $R$ generated by ${z_{i} : i \geq 0}$ is a free $k$ -ring on the $z_{i}$ ; likewise ${z_{0}, \dots, z_{n}}$ generates a free $k$ -ring on $n + 1$ generators.

Proof

Each $z_{i}$ commutes with $k$ , so by the universal property there is a $k$ -algebra homomorphism $ψ$ from the free ring $k ⟨ Z_{0}, Z_{1}, \dots ⟩$ to $R$ with $ψ (Z_{i}) = z_{i}$ , and its image is the subring in question. It suffices to prove $ψ$ injective, and since $ψ$ carries the $k$ -basis of words in the $Z_{i}$ to the elements

ψ (Z_{i_{1}} Z_{i_{2}} \dots Z_{i_{m}}) = x y^{i_{1}} x y^{i_{2}} \dots x y^{i_{m}},

it suffices to show these words are pairwise distinct — distinct words are $k$ -linearly independent in $R$ , so injectivity on a basis gives injectivity.

Given the word $x y^{i_{1}} x y^{i_{2}} \dots x y^{i_{m}}$ , the number of occurrences of $x$ recovers $m$ , and the number of $y$ 's between the $r$ th and $(r + 1)$ st occurrence of $x$ recovers $i_{r}$ , with $i_{m}$ read off from the terminal run of $y$ 's. So the tuple $(i_{1}, \dots, i_{m})$ is determined, and distinct tuples give distinct words.

Proposition—Free rings on at least two generators are not noetherian

Let $k$ be a nonzero ring and $R = k ⟨ x, y ⟩$ . The left ideal $𝔞 = \sum_{i \geq 0} R x y^{i}$ is not finitely generated, so $R$ is not left noetherian. By the symmetric argument with $y^{i} x$ , it is not right noetherian either.

Proof

Suppose $𝔞$ were generated by $g_{1}, \dots, g_{m}$ . Each $g_{j}$ is a finite sum $\sum_{i} f_{j i} x y^{i}$ , so there is $N$ with all $g_{j} \in \sum_{i \leq N} R x y^{i}$ ; the latter is a left ideal containing the generators, hence equals $𝔞$ . In particular $x y^{N + 1} = \sum_{i \leq N} f_{i} x y^{i}$ for some $f_{i} \in R$ .

Compare coefficients of words. Each product $w x y^{i}$ , for $w$ a word, is the word $w x y^{i}$ , and $w x y^{i} = x y^{N + 1}$ forces $w$ empty and $i = N + 1$ . Since $i \leq N$ on the right-hand side, the coefficient of $x y^{N + 1}$ there is $0$ , while on the left it is $1$ — contradiction, as $k \neq 0$ .

Proposition(1.3)(d)A generic left zero-divisor is not a right zero-divisor

Let $R = ℤ ⟨ x, y ⟩ / (x y)$ and write $\bar{x}, \bar{y}$ for the images of $x, y$ . Then $\bar{x} \bar{y} = 0$ but $\bar{y} \bar{x} \neq 0$ . Similarly, in $ℤ ⟨ x, y ⟩ / (x y - 1)$ one has $\bar{x} \bar{y} = 1$ and $\bar{y} \bar{x} \neq 1$ .

Proof

First claim. Let $T = (\begin{matrix} ℤ & ℤ / 2 ℤ \\ 0 & ℤ \end{matrix})$ be the triangular ring of §1, and set

a = (\begin{matrix} 2 & 0 \\ 0 & 1 \end{matrix}), b = (\begin{matrix} 0 & 1 \\ 0 & 0 \end{matrix}), a b = (\begin{matrix} 0 & 2 \cdot 1 \\ 0 & 0 \end{matrix}) = 0 .

Since $ℤ$ is initial among rings, the universal property gives a homomorphism $ℤ ⟨ x, y ⟩ \to T$ with $x \mapsto a$ , $y \mapsto b$ . It kills $x y$ , hence kills the ideal $(x y)$ , hence factors through $R$ . But the image of $y x$ is

b a = (\begin{matrix} 0 & 1 \\ 0 & 0 \end{matrix}) (\begin{matrix} 2 & 0 \\ 0 & 1 \end{matrix}) = (\begin{matrix} 0 & 1 \\ 0 & 0 \end{matrix}) \neq 0,

so $\bar{y} \bar{x} \neq 0$ in $R$ .

Second claim. Let $V = ⨁_{i \geq 1} k e_{i}$ over a field $k$ and take the shift operators in ${End}_{k} (V)$ : $b (e_{i}) = e_{i + 1}$ and $a (e_{1}) = 0$ , $a (e_{i}) = e_{i - 1}$ for $i \geq 2$ . Then $a b = 1$ , so $x \mapsto a$ , $y \mapsto b$ factors through $ℤ ⟨ x, y ⟩ / (x y - 1)$ , while $(b a) (e_{1}) = 0 \neq e_{1}$ shows $b a \neq 1$ . Hence $\bar{y} \bar{x} \neq 1$ .

Theorem(1.3)(c)The Weyl algebra as differential operators

Let $k$ be a field of characteristic $0$ , $P = k [t]$ , and let $D, L \in {End}_{k} (P)$ be $D (f) = d f / d t$ and $L (f) = t f$ . Then $D L - L D = {id}_{P}$ , and the induced homomorphism $A_{1} (k) = k ⟨ x, y ⟩ / (x y - y x - 1) \to {End}_{k} (P)$ sending $\bar{x} \mapsto D$ , $\bar{y} \mapsto L$ is injective with image the ring of differential operators $\sum_{i} a_{i} (t) D^{i}$ , $a_{i} \in k [t]$ .

Moreover ${{\bar{y}}^{j} {\bar{x}}^{i} : i, j \geq 0}$ is a $k$ -basis of $A_{1} (k)$ , as is ${{\bar{x}}^{i} {\bar{y}}^{j}}$ .

Corollary—No finite-dimensional representations in characteristic zero

If $char k = 0$ then $A_{1} (k)$ admits no nonzero ring homomorphism into $M_{n} (k)$ for any finite $n$ . Consequently $A_{1} (k)$ is not artinian and has no finite-dimensional simple modules.

Proof

Suppose $ρ : A_{1} (k) \to M_{n} (k)$ is a ring homomorphism, so $ρ (1) = I_{n}$ . Applying the trace to the defining relation,

tr (ρ (\bar{x}) ρ (\bar{y}) - ρ (\bar{y}) ρ (\bar{x})) = 0, tr (I_{n}) = n,

using $tr (A B) = tr (B A)$ . The relation forces $n = 0$ in $k$ , impossible in characteristic $0$ for $n \geq 1$ . In characteristic $p$ the argument collapses, and indeed $A_{1} (k)$ then has $p$ -dimensional representations.

Remark(3.17)Characteristic

p

changes everything

If $char k = p > 0$ , then ${\bar{x}}^{p}$ and ${\bar{y}}^{p}$ are central in $A_{1} (k)$ , the centre is $k [{\bar{x}}^{p}, {\bar{y}}^{p}]$ , and $A_{1} (k)$ is a free module of rank $p^{2}$ over it. In particular $A_{1} (k)$ is no longer simple. The differential-operator model also degenerates: on $k [t]$ one has $D^{p} = 0$ , because $p$ consecutive integers always include a multiple of $p$ .

Proof Techniques and Method

How these proofs work, and which move to reuse.

Move 1

Read the word backwards

To prove that a family of elements of a free ring is free, show that distinct products spell distinct words. Uniqueness of the spelling is the whole argument.

Move 2

Specialise to a small ring

To prove something is nonzero in a presented ring, map the presentation into a concrete ring satisfying the relations and compute there. Triangular rings and shift operators handle most cases.

Move 3

Take a trace

A relation of the form $[u, v] = 1$ is incompatible with any finite-dimensional representation in characteristic zero, because commutators have trace zero and the identity does not.

Move 2 has a limitation worth naming. Specialisation proves elements are nonzero; it never proves they are zero, since a single specialisation can lose information. Proving an element vanishes requires a normal form, which is what the diamond lemma provides.

Worked Example

The first Weyl algebra, computed

Take $k$ of characteristic $0$ and $P = k [t]$ . With $D = d / d t$ and $L$ = multiplication by $t$ , the product rule gives, for any $f \in P$ ,

(D L) (f) = \frac{d}{d t} (t f) = f + t \frac{d f}{d t}, (L D) (f) = t \frac{d f}{d t},

(E.1)

so $(D L - L D) (f) = f$ and $D L - L D = id$ . By the universal property there is a homomorphism $A_{1} (k) \to {End}_{k} (P)$ with $\bar{x} \mapsto D$ , $\bar{y} \mapsto L$ .

Its image is exactly ${\sum_{i} a_{i} (t) D^{i}}$ : the image is spanned by products of $D$ s and $L$ s, and the relation $D L = L D + 1$ lets every such product be rewritten with all $L$ s to the left, which is the assertion that ${{\bar{y}}^{j} {\bar{x}}^{i}}$ spans $A_{1} (k)$ .

Checking linear independence

Suppose $Θ = \sum_{i, j} c_{i j} t^{j} D^{i}$ acts as zero on $P$ , with not all $c_{i j}$ zero, and let $m$ be the least $i$ for which some $c_{i j} \neq 0$ . Then $Θ = \sum_{i \geq m} \sum_{j} c_{i j} t^{j} D^{i}$ . Apply it to $t^{m}$ : every term with $i > m$ vanishes, since $D^{i} (t^{m}) = 0$ there, while $D^{m} (t^{m}) = m!$ . Hence

0 = Θ (t^{m}) = m! \sum_{j} c_{m j} t^{j},

(E.1a)

and $m! \neq 0$ in characteristic $0$ , so every $c_{m j} = 0$ — contradicting the choice of $m$ . Hence the representation is faithful and ${{\bar{y}}^{j} {\bar{x}}^{i}}$ is a basis of $A_{1} (k)$ .

A commutator computation

From $\bar{x} \bar{y} - \bar{y} \bar{x} = 1$ one gets by induction

\bar{x} {\bar{y}}^{n} - {\bar{y}}^{n} \bar{x} = n {\bar{y}}^{n - 1} (n \geq 1),

(E.2)

the algebraic form of $\frac{d}{d t} (t^{n} f) - t^{n} \frac{d f}{d t} = n t^{n - 1} f$ .

In characteristic $p$ this gives $\bar{x} {\bar{y}}^{p} = {\bar{y}}^{p} \bar{x}$ , which is the centrality of ${\bar{y}}^{p}$ recorded above; the same computation with the roles reversed gives centrality of ${\bar{x}}^{p}$ .

The quaternions as a presented ring

Set $R = ℝ ⟨ x, y ⟩ / (x^{2} + 1, y^{2} + 1, x y + y x)$ . The images $\bar{x}, \bar{y}$ satisfy exactly Hamilton's relations $i^{2} = j^{2} = - 1$ , $i j = - j i$ , so there is a surjection $R ↠ ℍ$ . Conversely the relations let every word be rewritten in the form $\pm 1, \pm \bar{x}, \pm \bar{y}, \pm \bar{x} \bar{y}$ times a real, so $\dim_{ℝ} R \leq 4$ ; since $\dim_{ℝ} ℍ = 4$ the surjection is an isomorphism, $R ≅ ℍ$ .

Frameworks and Models

Almost every algebra in Lam §1 is a quotient of a tensor algebra, which for a free module is a free ring. The relations imposed classify the outcome.

$T (V) = k ⟨ e_{1}, \dots, e_{n} ⟩$ — tensor algebra on an $n$ -dimensional space
- quotient by $u \otimes v - v \otimes u$
  - symmetric algebra $S (V) = k [e_{1}, \dots, e_{n}]$ , dimension infinite, commutative
- quotient by $v \otimes v$
  - exterior algebra $Λ (V)$ , dimension $2^{n}$ ; a local ring with residue field $k$
- quotient by $v \otimes v - q (v)$
  - Clifford algebra $C (V, q)$ , dimension $2^{n}$ when $char k \neq 2$ ; recovers $Λ (V)$ when $q = 0$
- quotient by $u \otimes v - v \otimes u - [u, v]$
  - universal enveloping algebra $U (𝔤)$ , with the Poincaré–Birkhoff–Witt basis
- quotient by $x y - y x - 1$
  - Weyl algebra $A_{1} (k)$ ; a simple noetherian domain in characteristic $0$

The relationship between the last two is not accidental: the enveloping algebra of the $(2 n + 1)$ -dimensional Heisenberg Lie algebra, with its central element set equal to $1$ , is precisely $A_{n} (k)$ .

Process and Workflow

Choose generatorsFix the alphabet

X

and the coefficient ring

k

; decide whether

k

is required to be commutative.

Write the relationsList

F \subseteq k ⟨ X ⟩

. Every relation must be written as an element that is to become zero, not as an equation between words.

Find a spanning setUse the relations as rewriting rules to push words into a normal form; this bounds the size of

k ⟨ X ⟩ / (F)

from above.

Find a modelConstruct a concrete ring satisfying

F

in which the normal-form elements are visibly independent; this bounds the size from below.

ConcludeWhen the bounds agree, the normal forms are a basis and the presentation is resolved.

You need to know whether an element $u$ is zero in $k ⟨ X ⟩ / (F)$ . What do you do?

Show it is nonzeroSpecialise: map into a concrete ring satisfying

F

and evaluate. One successful specialisation settles it.

Show it is zeroRewrite

u

using the relations until it reduces to

0

. A single reduction sequence suffices, but finding it may require a confluent rewriting system.

Neither works quicklyAttempt a noncommutative Gröbner basis for

(F)

and reduce

u

against it. The computation may not terminate.

It still will not settleExpect no algorithm: the word problem for finitely presented rings is undecidable, so some instances are genuinely hopeless.

Comparison and Classification

Free ring against polynomial ring
Property	$k ⟨ x, y ⟩$	$k [x, y]$
Basis over $k$	all words in $x, y$	monomials $x^{a} y^{b}$
Dimension in degree $n$	$2^{n}$	$n + 1$
Noetherian	neither side	yes, by the Hilbert basis theorem
Contains a free ring on $ℵ_{0}$ generators	yes	no
Units ( $k$ a domain)	$U (k)$	$U (k)$
Domain ( $k$ a domain)	yes	yes
One-sided ideals	all free, of unique rank when $k$ is a division ring	not free in general
Embeds in a division ring	yes, by Cohn's construction	yes, the rational function field

Presented rings from §1 and their properties
	Domain	Noetherian	Simple	Finite-dimensional over $k$
$k ⟨ x, y ⟩$	yes	no	no	no
$k [x, y]$	yes	yes	no	no
$A_{1} (k)$ , $char k = 0$	yes	yes	yes	no
$A_{1} (k)$ , $char k = p$	yes	yes	no	no
$ℍ = ℝ ⟨ x, y ⟩ / (x^{2} + 1, y^{2} + 1, x y + y x)$	yes	yes	yes	yes, dimension 4
$Λ (V)$ , $\dim V = n$	no	yes	no	yes, dimension $2^{n}$
$ℤ ⟨ x, y ⟩ / (x y)$	no	no	no	no

Presented rings from §1 and their properties

Applications and Industry Use

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

Symbolic computation

D-modules and integration

The Weyl algebra is the ambient ring for algorithmic integration and for holonomic-function methods. Macaulay2's D-modules package, Singular:Plural and the HolonomicFunctions package all compute Gröbner bases in $A_{n} (k)$ to evaluate integrals and prove identities.

Quantum algebra

Quantum groups and quantum planes

Quantum groups are presented algebras: the quantum plane is $k ⟨ x, y ⟩ / (x y - q y x)$ , and the quantised enveloping algebras are free algebras modulo the Serre relations. Everything computational about them is presentation-driven.

Automata and formal languages

Rational and algebraic series

Formal power series over a free monoid are the natural home of weighted automata; rationality of a series corresponds to recognisability, and the free ring is the polynomial part of that theory.

Control and systems

Noncommutative realisation theory

Transfer functions of bilinear and nonlinear systems are series over a free monoid; minimal realisation is a rank condition on a Hankel matrix indexed by words.

Free probability

Noncommutative polynomials in random matrices

Evaluating elements of a free algebra on large random matrices is the basic operation of free probability, which underpins asymptotic results in wireless communications and random matrix theory.

Ring theory internally

Manufacturing counterexamples

Most pathologies in this collection start as a free ring modulo the minimum relations needed. Being able to prove that nothing further collapses is exactly the specialisation technique above.

Computational Notes

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

Normal forms. Bergman's diamond lemma gives a criterion: a rewriting system derived from the relations yields unique normal forms exactly when all overlap and inclusion ambiguities resolve. Verifying finitely many ambiguities settles a presentation completely.
Noncommutative Gröbner bases. Mora's algorithm computes them, but a finitely generated ideal of $k ⟨ X ⟩$ need not have a finite Gröbner basis, so the procedure is only a semi-decision method — it terminates when it can, and a degree bound must be imposed otherwise.
Undecidability. There is no algorithm that decides, for arbitrary finite presentations, whether two words are equal; the word problem for finitely presented associative algebras is undecidable, inherited from the corresponding result for semigroups.
Weyl algebras are better behaved. $A_{n} (k)$ is noetherian and admits a genuine Gröbner theory with terminating algorithms, because the commutator of two generators has lower order than their product — this is the solvable-type or PBW-algebra condition.
Implementations. GAP's GBNP package and Singular's Letterplace handle free algebras; Singular:Plural, Macaulay2 and Magma handle PBW-type algebras including $A_{n} (k)$ .

Failure Modes and Common Mistakes

Do not expect the Hilbert basis theorem. $k ⟨ x, y ⟩$ is not noetherian, so finitely generated ideals are the exception and finitely generated modules can have non-finitely-generated submodules.
Do not conclude equality from a single specialisation. Specialisation is one-directional evidence: it can only prove things are different.
Do not assume the Weyl algebra behaves uniformly in characteristic. Simplicity, the centre and the faithfulness of the differential model all change at $char k = p$ .
Do not identify $A_{1} (k)$ with $k [t] ⟨ \partial ⟩$ without saying which side the coefficients sit on; the ring of differential operators is a skew polynomial ring $k [t] [\partial; δ]$ with $δ = d / d t$ .

Quick Reference

Free ring

k ⟨ x_{i} : i \in I ⟩

, basis the words in the

x_{i}

As a semigroup ring

k Σ

for

Σ

the free monoid on

{x_{i}}

Universal propertyany

ϕ_{0} : k \to k'

plus images commuting with

ϕ_{0} (k)

extends uniquely

Units

U (k)

when

k

is a domain; larger otherwise

Chain conditions

k ⟨ x, y ⟩

is neither left nor right noetherian

Self-embedding

z_{i} = x y^{i}

generate a free subring on

ℵ_{0}

generators

Presentation

k ⟨ X ⟩ / (F)

, with

(F)

the two-sided ideal

Weyl algebra

A_{1} (k) = k ⟨ x, y ⟩ / (x y - y x - 1)

; basis

y^{j} x^{i}

Standard presentations
Ring	Presentation	Reference
Polynomial ring	$k ⟨ x_{i} ⟩ / (x_{i} x_{j} - x_{j} x_{i})$	(1.3)(a)
Real quaternions	$ℝ ⟨ x, y ⟩ / (x^{2} + 1, y^{2} + 1, x y + y x)$	(1.3)(b)
First Weyl algebra	$k ⟨ x, y ⟩ / (x y - y x - 1)$	(1.3)(c)
$n$ th Weyl algebra	$A_{n} (k) = A_{1} (A_{n - 1} (k))$	(1.3)(c)
Generic left zero-divisor	$ℤ ⟨ x, y ⟩ / (x y)$	(1.3)(d)
Generic one-sided inverse	$ℤ ⟨ x, y ⟩ / (x y - 1)$	(1.3)(d)
Exterior algebra	$T (V) / (v \otimes v)$	(1.10)

Frequently Asked Questions

Why must the coefficients commute with the variables?

Because otherwise the universal property has no clean statement and the words fail to be a basis. The construction deliberately builds the commutation $x_{i} c = c x_{i}$ into the multiplication. If you want the variable to act on the coefficients, you want a skew polynomial ring $k [x; σ]$ or a differential polynomial ring $k [x; δ]$ instead.

How can

k ⟨ x, y ⟩

contain a free ring on infinitely many generators?

Because free monoids are self-similar. The words $x y^{0}, x y^{1}, x y^{2}, \dots$ have the property that any concatenation of them can be uniquely decomposed back, so they generate freely. The commutative analogue fails because $k [x, y]$ has transcendence degree $2$ and any three elements satisfy a polynomial relation.

k ⟨ x, y ⟩

a domain?

Yes when $k$ is. Order words by degree and then lexicographically; the leading word of a product is the concatenation of the leading words, with coefficient the product of the leading coefficients, which is nonzero since $k$ is a domain. The same ordering computes the units and gives the noncommutative division algorithm used in Gröbner theory.

Why is the Weyl algebra simple in characteristic zero but not in characteristic

p

Simplicity comes from the commutator identity $[\bar{x}, {\bar{y}}^{n}] = n {\bar{y}}^{n - 1}$ : bracketing with $\bar{x}$ or $\bar{y}$ lowers degree and never annihilates a nonzero element, so any nonzero ideal eventually contains a nonzero scalar. In characteristic $p$ the factor $n$ vanishes when $p ∣ n$ , the process stalls, and indeed ${\bar{x}}^{p}$ and ${\bar{y}}^{p}$ generate a proper ideal.

What exactly does a specialisation argument prove?

That a specified element of a presented ring is nonzero, and nothing more. A homomorphism into a concrete ring can only detect what survives; it cannot certify that two elements are equal upstairs, since the homomorphism may have a large kernel. To prove an element is zero you must exhibit a reduction using the relations.

Can every ring be presented by generators and relations?

Every ring is a quotient of a free ring — take the underlying set as the alphabet — so yes in principle, but the presentation is usually infinite and useless. The content of a good presentation is that it is finite, or at least that it has a manageable normal form.

References

T. Y. Lam, A First Course in Noncommutative Rings, Graduate Texts in Mathematics 131, Springer-Verlag, 1991, §1, Examples (1.2)–(1.3) and (1.10), pp. 6–14.
P. M. Cohn, Free Ideal Rings and Localization in General Rings, New Mathematical Monographs 3, Cambridge University Press, 2006, Chapters 0–2.
G. M. Bergman, “The diamond lemma for ring theory”, Advances in Mathematics 29 (1978), 178–218.
J. C. McConnell and J. C. Robson, Noncommutative Noetherian Rings, revised edition, Graduate Studies in Mathematics 30, American Mathematical Society, 2001, Chapters 1 and 8 (Weyl algebras).
S. C. Coutinho, A Primer of Algebraic D-modules, London Mathematical Society Student Texts 33, Cambridge University Press, 1995.
L. H. Rowen, Ring Theory, Volume I, Academic Press, 1988, §1.3 and §1.6.

AI Suggested Questions

State Bergman's diamond lemma precisely and apply it to the quaternion presentation.
Prove that the first Weyl algebra is simple in characteristic zero.
Sketch Cohn's proof that a free algebra over a division ring embeds in a division ring.
Why do free algebras have no finite Gröbner basis for some finitely generated ideals?
Give a finite presentation whose ring is nonzero but whose triviality is not obvious, and explain how to detect it.
How does the Poincaré–Birkhoff–Witt theorem generalise the basis computation for the Weyl algebra?
Compare the free algebra with the free group algebra and explain which properties transfer.

Free Rings and Presentations

Executive Summary

Overview

Learning Objectives

Definitions

Core Concepts

Grading, degree and the domain property

Free rings are enormous

Presentations and what they can hide

Key Results

Proof Techniques and Method

Read the word backwards

Specialise to a small ring

Take a trace

Worked Example

The first Weyl algebra, computed

Checking linear independence

A commutator computation

The quaternions as a presented ring

Frameworks and Models

Process and Workflow

Comparison and Classification

Applications and Industry Use

D-modules and integration

Quantum groups and quantum planes

Rational and algebraic series

Noncommutative realisation theory

Noncommutative polynomials in random matrices

Manufacturing counterexamples

Computational Notes

Failure Modes and Common Mistakes

Quick Reference

Frequently Asked Questions

References

AI Suggested Questions

Continue learning

01Executive Summary

02Overview

03Learning Objectives

04Definitions

05Core Concepts

Grading, degree and the domain property

Free rings are enormous

Presentations and what they can hide

06Key Results

07Proof Techniques and Method

Read the word backwards

Specialise to a small ring

Take a trace

08Worked Example

The first Weyl algebra, computed

Checking linear independence

A commutator computation

The quaternions as a presented ring

09Frameworks and Models

10Process and Workflow

11Comparison and Classification

12Applications and Industry Use

D-modules and integration

Quantum groups and quantum planes

Rational and algebraic series

Noncommutative realisation theory

Noncommutative polynomials in random matrices

Manufacturing counterexamples

13Computational Notes

14Failure Modes and Common Mistakes

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

Frameworks and Models

Process and Workflow

Comparison and Classification

Applications and Industry Use

Computational Notes

Failure Modes and Common Mistakes

Quick Reference

Frequently Asked Questions

Related KEVOS Topics

References

AI Suggested Questions