nLab
Mal'cev variety

Contents

Definition

A Mal’cev operation on a set X is a ternary operation, a function

t:X×X×XX,(x,y,z)t(x,y,z),t:X\times X\times X\to X,\,\,\,(x,y,z)\mapsto t(x,y,z) ,

which satisfies the identities t(x,x,z)=z and t(x,z,z)=x. An important motivating example is the operation t of a heap, for example the operation on a group defined by t(x,y,z)=xy 1z.

An algebraic theory T is a Mal’cev theory when T contains a Mal’cev operation. An algebraic theory is Mal’cev iff one of the following equivalent statements is true:

  1. in the category of T-algebras, every internal reflexive relation is a congruence;

  2. in the category of T-algebras, the composite (as internal relations) of any two congruences as a congruence;

  3. in the category of T-algebras, the composition of equivalence relations is commutative.

Statement (i) is one of the motivations to introduce the notion of Mal'cev category.

A Mal’cev variety is the category of T-algebras for a Mal’cev theory T, thought of as a variety of algebras.

Proofs of equivalence

If RX×Y is a binary relation on sets, write R(x,y) to say that (x,y)R. If X, Y are T-algebras, then R is an internal relation in T-Alg if the conditions R(x 1,y 1)R(x n,y n), and θ(x 1,,x n)=x, θ(y 1,,y n)=y for any n-ary operation θ of T, jointly imply R(x,y).

The set-theoretic composite of two internal relations in T-Alg is also an internal relation, and the equality relation is always internal, so we may (and will) apply ordinary set-theoretic reasoning in our proofs below.

Proposition 1

If T is a Mal’cev theory, then any internal reflexive relation in T-Alg is an internal equivalence relation.

Proof

If t is a Mal’cev operation and R is any internal reflexive relation on a T-algebra X, then R is transitive because given R(x,y)R(y,z), we infer R(x,y)R(y,y)R(y,z), and this together with t(x,y,y)=x and t(y,y,z)=z gives R(x,z) since R is internal. Also R is symmetric, because if R(x,y), we infer R(x,x)R(x,y)R(y,y), which together with t(x,x,y)=y and t(x,y,y)=x gives R(y,x).

Proposition 2

If every internal reflexive relation is an internal equivalence relation, then the composite of any two internal equivalence relations is also an internal equivalence relation.

Proof

The hypothesis is that internal reflexive relations and internal equivalence relations coincide. But (internal) reflexive relations are clearly closed under composition: Δ=ΔΔRS.

Proposition 3

If internal equivalence relations are closed under composition, then composition of internal equivalence relations is commutative.

Proof

If R and S are equivalence relations and so is SR, then

SR=(SR) op=R opS op=RS,S \circ R = (S \circ R)^{op} = R^{op} \circ S^{op} = R \circ S,

as desired.

Proposition 4

If composition of internal equivalence relations in T-Alg is commutative, then the theory T has a Mal’cev operation t.

Proof

According to the yoga of (Lawvere) algebraic theories, n-ary operations are identified with elements of F(n), the free T-algebra on n generators (more precisely, the Lawvere theory is the category opposite to the category of finitely generated free T-algebras). Thus we must exhibit a suitable element t of F(3).

Let x,y,z be the generators of F(3), and let a,b be the generators of F(2). Let ϕ be the unique algebra map F(3)F(2) taking x and y to a and z to b, and let ψ be the unique algebra map F(3)F(2) taking x to a and y and z to b. An operation tF(3) is Mal’cev precisely when

ϕ(t)=bψ(t)=a\phi(t) = b \qquad \psi(t) = a

Let R be the equivalence relation on F(3) given by the kernel pair of ϕ, and let S be the kernel pair of ψ. Then R(x,y) and S(y,z), so (SR)(x,z). Then, since composition of equivalence relations is assumed commutative, (RS)(x,z). This means there exists t such that S(x,t) and R(t,z), or that ψ(x)=ψ(t) and ϕ(t)=ϕ(z). This completes the proof.

Examples

  • The theory of groups, where t(x,y,z)=xy 1z, is Mal’cev.

  • The theory of Heyting algebras, where

    t(x,y,z)=((zy)x)((xy)z),t(x, y, z) = ((z \Rightarrow y) \Rightarrow x) \wedge ((x \Rightarrow y) \Rightarrow z),

    is Mal’cev.

  • If T is Mal’cev, and if TT is a morphism of algebraic theories, then T is Mal’cev. From this point of view, the theory of groups is Mal’cev because the theory of heaps is Mal’cev, and the theory of Heyting algebras is Mal’cev because the theory of cartesian closed meet-semilattices is Mal’cev.

See also Mal'cev category.

The lattice of congruences Equiv(X)

Equiv(X) is a modular lattice

In any finitely complete category, the intersection of two congruences (equivalence relations) on an object X is a congruence, so that the set of equivalence relations Equiv(X) is a meet-semilattice.

In a regular category such as a variety of algebras, where there is a sensible calculus of relations and relational composition, it is a simple matter to prove that if Equiv(X) is closed under relational composition, then RS is the join RS in Equiv(X). For, if R,SEquiv(X), then

R=RΔRS,S=ΔSRSR = R \circ \Delta \subseteq R \circ S, \qquad S = \Delta \circ S \subseteq R \circ S

while if R,ST in Equiv(X), then

RSTTT.R \circ S \subseteq T \circ T \subseteq T.
Proposition

In a regular category, if Equiv(X) is closed under relational composition (equivalently, if composition of equivalence relations is commutative), then Equiv(X) is a modular lattice.

Proof

The (poset-enriched) category of relations in a regular category is an allegory, and hence satisfies Freyd’s modular law

R(ST)S((S opR)T)R \wedge (S \circ T) \subseteq S \circ ((S^{op} \circ R) \wedge T)

whenever T:XY, S:YZ, R:XZ are relations. As we have just seen, the hypothesis implies that joins in Equiv(X) are given by composition (so Equiv(X) is a lattice), and so for R,S,TEquiv(X) we have

R(ST)S((SR)T).R \wedge (S \vee T) \subseteq S \vee ((S \vee R) \wedge T).

Therefore, if SR, we have both

R(ST)S(RT)R \wedge (S \vee T) \subseteq S \vee (R \wedge T)

and also S(RT)R(ST). Thus SR implies R(TS)=(RT)S: the modular law is satisfied in Equiv(X).

Corollary

If T is a Mal’cev theory, then the lattice of congruences Equiv(X) on any T-algebra X is a modular lattice.

Equiv(X) is a Desarguesian lattice

A similar argument shows that congruence lattices for T-algebras X, for T a Mal’cev theory, satisfy the following property (stronger than the modular property):

  • Desarguesian property?: if R i,S i,T iEquiv(X) for i=1,2, then

    (R 1R 2)(S 1S 2))T 1T 2implies(R 1S 1)(R 2S 2)((R 1T 1)(R 2T 2))((S 1T 1)(S 2T 2))(R_1 \vee R_2) \wedge (S_1 \vee S_2)) \subseteq T_1 \vee T_2 \qquad implies \qquad (R_1 \vee S_1) \wedge (R_2 \vee S_2) \subseteq ((R_1 \vee T_1) \wedge (R_2 \vee T_2)) \vee ((S_1 \vee T_1) \wedge (S_2 \vee T_2))

Freyd-Scedrov’s Categories, Allegories (2.157, pp. 206-207) gives the following argument: given relations R 1,S 1,T 1:XY, R 2,S 2,T 2:YZ between sets, it is “easily verified” that

R 2R 1S 2S 2T 2T 1impliesS 1R 1 opS 2 opR 2(S 1T 1 opS 2 opT 2)(T 1R 1 opT 2 opR 2)R_2 R_1 \cap S_2 S_2 \subseteq T_2 T_1 \qquad implies \qquad S_1 R_{1}^{op} \cap S_{2}^{op} R_2 \subseteq (S_1 T_{1}^{op} \wedge S_{2}^{op} T_2)(T_1 R_{1}^{op} \cap T_{2}^{op}R_2)

Then, under the assumption that equivalence relations internal to T-Alg commute (so that the join of equivalence relations R,S on X is their relational composite RS=RS), the Desarguesian axiom follows immediately.

References

See the monograph Borceux-Bourn.

Spelling

The original is ‘Мальцев’; besides ‘Malʹcev’, this has also been transliterated ‘Malcev’ and ‘Maltsev’.

Revised on August 22, 2012 19:06:56 by Todd Trimble (67.81.93.25)