Orientation

Universal Algebra: Discipline Overview

Group theory, ring theory and lattice theory each rediscover the same theorems in different costumes. Universal algebra is the discipline that proves them once.

Engineering · Mathematics7 min readKV-MATH-0201

Learning objectives

State what an algebra of arbitrary type is, and why the definition omits axioms.
Explain why congruences, not normal subobjects, are the general quotient device.
Distinguish Birkhoff's subdirect representation theorem from the HSP theorem.
Situate lattice theory as the tool by which varieties are classified.
Navigate the nine streams of this collection and choose an entry point.

01The redundancy that motivates the subject

Anyone who has taught a first course in algebra has watched the same argument run three times. The homomorphism theorem for groups, the homomorphism theorem for rings, and the homomorphism theorem for modules are the same proof. So are the correspondence theorems. So are the two isomorphism theorems that follow them. The objects differ; the reasoning does not.

Universal algebra takes that observation seriously. It asks what remains of the argument when the specific operations are removed — when a group is no longer a set with a binary operation satisfying associativity, but simply a set carrying some collection of finitary operations. The surprise is how much survives. The isomorphism theorems survive intact. So does the theory of direct and subdirect decomposition. What does not survive is anything that depended on the particular axioms, and identifying that boundary precisely is much of the subject's content.

Classical algebra

Fixed signature

One operation set, one axiom list. Theorems are proved per structure and re-proved for each new structure.

Universal algebra

Arbitrary type

Operations are a parameter. A theorem proved here holds for every structure whose type it fits, including ones not yet defined.

The payoff is not merely economy. Once the type is a parameter, one can quantify over classes of algebras and ask questions that are meaningless inside a single theory: which classes are definable by equations? which have well-behaved congruence lattices? which have a decidable first-order theory? These are the questions the subject actually studies.

02What an algebra is

An algebra is a non-empty set together with a family of finitary operations on it. That is the whole definition. There are no axioms, because axioms are what distinguish one class of algebras from another, and the theory wants to quantify over classes.

A = ⟨A, F⟩ where each f ∈ F is a map Aⁿ → A for some n ≥ 0

The type (or signature) records the arity of each operation symbol. Two algebras are comparable only when they share a type.

The arity-zero case matters more than it looks. A nullary operation is a distinguished element — a constant. Whether the identity of a group is treated as a constant in the type or merely as an element guaranteed by an axiom changes what counts as a subalgebra, because subalgebras must be closed under the operations of the type. Groups-with-constants have no empty subalgebra and every subalgebra contains the identity; groups-as-semigroups-with-inverse do not behave the same way. This is not pedantry: the choice of type determines the entire lattice of subuniverses.

CautionThe type is part of the data, not a convention

A great many apparent counterexamples in universal algebra dissolve once one checks which operations were actually placed in the type. When a source says rings form a variety, it means rings-with-a-nullary-1 or rings-without-1, and the two classes have different subalgebra lattices, different free algebras and different congruence behaviour. State the type before arguing about the theorem.

03Congruences: the general quotient device

Classical algebra forms quotients by dividing out a distinguished subobject: a normal subgroup, a two-sided ideal, a submodule. That device is an accident of those particular signatures. What is actually happening in every case is that an equivalence relation compatible with the operations is being collapsed, and the distinguished subobject is merely a convenient coordinate for naming the relation.

ProcedureTesting whether an equivalence relation is a congruence

in: A, θ → out: decision θ ∈ Con A

input: algebra A of type F; equivalence relation θ ⊆ A × A
for each operation symbol f ∈ F, of arity n:
for all a₁,…,aₙ, b₁,…,bₙ ∈ A:
if ⟨aᵢ, bᵢ⟩ ∈ θ for every i = 1,…,n:
require ⟨f(a₁,…,aₙ), f(b₁,…,bₙ)⟩ ∈ θ
if every symbol passes: θ ∈ Con A
otherwise: θ is an equivalence relation only, and A/θ is not an algebra

Compatibility is checked symbol by symbol, so the test is finite when the type and the algebra are finite. Correctness: the condition is exactly what is needed for the induced operations on A/θ to be well defined.

Every congruence yields a quotient algebra; every homomorphism yields a congruence as its kernel; and the two constructions are mutually inverse. In groups the congruence lattice is isomorphic to the normal subgroup lattice, which is why the normal-subgroup formulation works there. In semigroups, in lattices, and in most algebras generally, no such coordinatisation exists, and the congruence is the only available object.

04Birkhoff's two theorems

Garrett Birkhoff proved two structural results in the 1930s and 1940s that between them define the shape of the subject. They are frequently confused because both are called Birkhoff's theorem.

Decomposition

Subdirect representation

Every algebra is a subdirect product of subdirectly irreducible algebras. This reduces structure theory to understanding the irreducibles — an analogue of decomposition into simple pieces that requires no chain conditions whatsoever.

Definability

The HSP theorem

A class of algebras is defined by a set of equations if and only if it is closed under homomorphic images, subalgebras and direct products. Syntax and closure properties coincide exactly.

Consequence

Varieties as the unit of study

Because equationally defined classes are precisely the HSP-closed ones, the variety becomes the natural object. Most later theory classifies varieties rather than individual algebras.

The HSP theorem is the more celebrated, and deservedly: it is a completeness result linking a syntactic notion (satisfying equations) to a purely algebraic one (closure under three operators). But the subdirect representation theorem does more day-to-day work. It is what makes it sensible to ask for the subdirectly irreducible members of a variety and to expect that answer to determine the variety.

05Why lattices come first

This collection opens with six pages on lattice theory before defining an algebra at all. That ordering is deliberate and follows the source. The reason is that the principal invariants attached to an algebra — its subuniverses and its congruences — are not merely sets but complete lattices, and specifically algebraic lattices. Classifying a variety almost always means constraining the congruence lattices of its members.

Closure operators produce complete lattices
The subuniverse generated by a set and the congruence generated by a relation are both closure operators; their closed sets form complete lattices automatically.
Algebraicity is the finiteness residue
Because operations are finitary, these closure operators are finitary too, which makes the resulting lattices algebraic. This is the only trace of finiteness the general theory retains.
Lattice identities classify varieties
Congruence-permutable, congruence-distributive and congruence-modular varieties are defined by conditions on Con A, and each is characterised by the existence of certain terms — the Mal'cev conditions.
Distributivity buys structure theorems
Jónsson's lemma, the discriminator variety theory and most of the Boolean product machinery all require congruence distributivity as a hypothesis.

06How this collection is organised

Nine streams. The first six follow the source closely; the last carries material that postdates it and is flagged accordingly.

Streams in this collection
Stream	Pages	Covers
Orientation	2	This page and the notational preliminaries.
Lattice Theory Foundations	6	Posets, distributivity, algebraic lattices, closure operators.
Core Universal Algebra	10	Algebras, subuniverses, congruences, homomorphisms, products.
Terms, Free Algebras and Equational Logic	8	Term algebras, HSP, completeness, Mal'cev conditions, the centre.
Combinatorial and Automata Applications	5	Steiner systems, Latin squares, Euler's conjecture, finite acceptors.
Boolean Algebras and Stone Duality	5	Boolean algebras and rings, filters, Stone duality.
Boolean Constructions and Discriminator Varieties	10	Boolean powers and products, primality, discriminator theory.
Model-Theoretic Connections	10	First-order logic, ultraproducts, compactness, finite basis, undecidability.
Research Frontier and Sourcing	6	The 1981 frontier, the open problems, and post-source developments.

NoteA note on the edition

The Millennium Edition is a corrected re-typesetting of the 1981 text, not a revision. Its closing survey of recent developments and its list of open problems are a 1981 snapshot, and several of those problems have since been settled. The Research Frontier stream addresses that gap explicitly rather than letting it pass silently.

Frequently asked

Is universal algebra the same thing as category theory?

No, though they overlap in ambition. Category theory abstracts the arrows between objects and treats the objects as opaque. Universal algebra keeps the elements and abstracts the operations. The practical consequence is that universal algebra can state and prove things about free algebras, congruence lattices and equational definability that are invisible categorially, while category theory handles constructions across wildly different kinds of structure that universal algebra cannot reach. They are complementary rather than competing.

Do I need model theory before starting?

Not for the first four streams. Chapters I through III of the source, which correspond to the Lattice, Core and Equational streams here, use no logic beyond equations. The Model-Theoretic stream develops first-order languages, satisfaction and ultraproducts from scratch, so it is self-contained — but it is genuinely harder, and it is where the subject stops being elementary.

Why are Boolean algebras given two whole streams?

Because they are the engine of the structure theory rather than merely one example among many. Stone duality converts Boolean algebras into topological spaces; Boolean powers and Boolean products then transport that machinery into arbitrary varieties, and discriminator varieties — arguably the best-behaved non-trivial class in the subject — are defined by it. The source devotes its longest chapter to this and the weighting here follows suit.

Sources and further reading

S. Burris and H. P. Sankappanavar, A Course in Universal Algebra, Millennium Edition (a corrected re-typesetting of Springer GTM 78, 1981).
G. Grätzer, Universal Algebra, 2nd edition, Springer.
R. McKenzie, G. McNulty and W. Taylor, Algebras, Lattices, Varieties, Volume I.

Original KEVOS® explanatory article. Written from the topic map of the cited works; no text is reproduced from them.