This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Characterizing Finitely Based Abelian Mal’cev Varieties

Mateo Muro
Abstract

In this paper, we prove the following characterization: an abelian Mal’cev variety is finitely based if and only it has finite type, its ring of idempotent binary terms is finitely presented, and its module of unary terms is finitely presented.

1 Introduction

In [2, Theorem 14.9], Freese and McKenzie proved that every locally finite abelian Mal’cev variety is finitely based. We also know that for an abelian Mal’cev variety 𝒱\mathcal{V} there exists a ring and module structure on the idempotent binary terms and unary terms in the type of 𝒱\mathcal{V}. We extend these ideas to characterize the finitely based abelian Mal’cev varieties as the abelian Mal’cev varieties with finitely presented rings and modules.

We refer to [3] for notation and background on general algebras. A type σ\sigma is a set of function symbols ff, each associated with a non-negative integer called the arity of ff. An algebra 𝐀\mathbf{A} of type σ\sigma is an ordered pair A,(f𝐀|fσ)\langle A,(f^{\mathbf{A}}\ |\ f\in\sigma)\rangle. Here, AA is a non-empty set, called the universe of 𝐀\mathbf{A}. Each nn-ary fσf\in\sigma is interpreted as an nn-ary basic operation f𝐀:AnAf^{\mathbf{A}}\colon A^{n}\to A. To simplify the notation, we will write ff for f𝐀f^{\mathbf{A}} if the meaning is clear from the context.

More generally, any term tt over σ\sigma in variables x1,,xnx_{1},\ldots,x_{n} induces an nn-ary term operation t𝐀t^{\mathbf{A}} on AA. An equation over σ\sigma is an expression of the form sts\approx t where s,ts,t are terms over σ.\sigma. We say 𝐀\mathbf{A} satisfies sts\approx t (written 𝐀st\mathbf{A}\models s\approx t) if s𝐀=t𝐀s^{\mathbf{A}}=t^{\mathbf{A}}.

A variety 𝒱\mathcal{V} is a class of algebras over a fixed type σ\sigma that is defined by a set of equations Σ\Sigma. Then Σ\Sigma is a basis for 𝒱\mathcal{V}. We say that 𝒱\mathcal{V} is finitely based if there exists a finite basis for 𝒱\mathcal{V}.

We say that a variety 𝒱\mathcal{V} is Mal’cev if there exists a term m(x,y,z)m(x,y,z) in the type of 𝒱\mathcal{V} such that the following equations hold in 𝒱\mathcal{V}:

m(x,x,y)ym(y,x,x)\displaystyle m(x,x,y)\approx y\approx m(y,x,x)

Then, we refer to m(x,y,z)m(x,y,z) as the Mal’cev term of 𝒱\mathcal{V}.

We refer to [2] for definitions and background of commutators of congruences. We say that an algebra 𝐀\mathbf{A} is abelian if [1𝐀,1𝐀]=0𝐀[1_{\mathbf{A}},1_{\mathbf{A}}]=0_{\mathbf{A}} where 0𝐀,1𝐀0_{\mathbf{A}},1_{\mathbf{A}} are the trivial and total congruences of 𝐀\mathbf{A}, respectively. A variety 𝒱\mathcal{V} is abelian if every 𝐀𝒱\mathbf{A}\in\mathcal{V} is abelian.

Let 𝐀\mathbf{A} be an algebra of type σ\sigma and let s(x1,,xn),t(x1,,xk)s(x_{1},\ldots,x_{n}),t(x_{1},\ldots,x_{k}) be terms over σ\sigma. We say ss and tt commute in 𝐀\mathbf{A} if and only if

𝐀s(\displaystyle\mathbf{A}\models s\bigl{(} t(x1,1,x1,2,,tx1,k),,t(xn,1,xn,2,,xn,k))\displaystyle t(x_{1,1},x_{1,2},\ldots,t_{x_{1},k}),\ldots,t(x_{n,1},x_{n,2},\ldots,x_{n,k})\bigl{)}
t(s(x1,1,x2,1,,xk,1),,s(x1,k,x2,k,,xn,k)).\displaystyle\approx t\bigl{(}s(x_{1,1},x_{2,1},\ldots,x_{k,1}),\ldots,s(x_{1,k},x_{2,k},\ldots,x_{n,k})\bigl{)}.
Proposition 1.

[2, Proposition 5.7] An algebra 𝐀\mathbf{A} with Mal’cev term mm is abelian if and only if mm commutes with every basic operation of 𝐀\mathbf{A}.

For a variety 𝒱\mathcal{V} and kk\in\mathbb{N}, let 𝐅𝒱(x1,,xk,z)\mathbf{F}_{\mathcal{V}}(x_{1},\ldots,x_{k},z) be the free algebra in 𝒱\mathcal{V} over variables x1,,xk,zx_{1},\ldots,x_{k},z. Let F𝒱(x1,,xk,z)F_{\mathcal{V}}(x_{1},\ldots,x_{k},z) denote the universe of 𝐅𝒱(x1,,xk,z)\mathbf{F}_{\mathcal{V}}(x_{1},\ldots,x_{k},z). Then F𝒱(x1,,xk,z)F_{\mathcal{V}}(x_{1},\ldots,x_{k},z) represents the (k+1)(k+1)-ary term functions over 𝒱\mathcal{V} and

F𝒱id(x1,,xk,z):={tF𝒱(x1,,xk,z)|t(z,,z,z)=z}\displaystyle F_{\mathcal{V}}^{{id}}(x_{1},\ldots,x_{k},z):=\{t\in F_{\mathcal{V}}(x_{1},\ldots,x_{k},z)\ |\ t(z,\ldots,z,z)=z\}

represents the (k+1)(k+1)-ary idempotent term functions over 𝒱\mathcal{V}.

For an abelian Mal’cev variety 𝒱\mathcal{V}, we define a ring structure on F𝒱id(x,z)F_{\mathcal{V}}^{id}(x,z) as on [2, p. 82]. Later we will prove the following lemma in Section 3:

Lemma 2.

[2, p.82] Let 𝒱\mathcal{V} be an abelian variety with Mal’cev term mm. For s,tF𝒱(x,z)s,t\in F_{\mathcal{V}}(x,z) define

s+t\displaystyle s+t :=m(s(x,z),z,t(x,z)),\displaystyle:=m\bigl{(}s(x,z),z,t(x,z)\bigl{)},
st\displaystyle s\cdot t :=s(t(x,z),z)\displaystyle:=s(t(x,z),z)
s\displaystyle-s :=m(z,s(x,z),z).\displaystyle:=m(z,s(x,z),z).
  1. 1.

    𝐑𝒱:=F𝒱id(x,z),+,,\mathbf{R}_{\mathcal{V}}:=\langle F_{\mathcal{V}}^{id}(x,z),+,-,\cdot\rangle is a ring with identity xx and zero zz.

  2. 2.

    𝐌𝒱:=F𝒱(z),+,,R𝒱\mathbf{M}_{\mathcal{V}}:=\langle F_{\mathcal{V}}(z),+,-,R_{\mathcal{V}}\rangle is an 𝐑𝒱\mathbf{R}_{\mathcal{V}}-module with zero zz.

  3. 3.

    F𝒱(x,z),+,,R𝒱\langle F_{\mathcal{V}}(x,z),+,-,R_{\mathcal{V}}\rangle is an 𝐑𝒱\mathbf{R}_{\mathcal{V}}-module isomorphic to the direct sum of the regular 𝐑𝒱\mathbf{R}_{\mathcal{V}}-module and 𝐌𝒱\mathbf{M}_{\mathcal{V}}.

  4. 4.

    Let a,b,cF𝒱(x,z)a,b,c\in F_{\mathcal{V}}(x,z) and rF𝒱id(x,z)r\in F_{\mathcal{V}}^{id}(x,z). Then

    m(a,b,c)=ab+cm(a,b,c)=a-b+c

    and

    r(a,b)=ra+(xr)b.r(a,b)=r\cdot a+(x-r)\cdot b.

With that, we have enough to state the main theorem.

Theorem 3.

Let 𝒱\mathcal{V} be an abelian Mal’cev variety. Then 𝒱\mathcal{V} is finitely based if and only if

  1. 1.

    𝒱\mathcal{V} has finite type,

  2. 2.

    the ring 𝐑𝒱\mathbf{R}_{\mathcal{V}} of binary idempotent terms is finitely presented,

  3. 3.

    and the 𝐑𝒱\mathbf{R}_{\mathcal{V}}-module 𝐌𝒱\mathbf{M}_{\mathcal{V}} of unary terms is finitely presented.

For example, the variety of abelian groups has a ring of binary idempotent terms isomorphic to \mathbb{Z} and has \mathbb{Z}-module of unary terms isomorphic to \mathbb{Z}, both of which are finitely presented, in fact free. Hence, our result proves the already known fact that the variety of abelian groups is finitely based [4]. In general, if 𝐑\mathbf{R} is a ring and 𝒱\mathcal{V} is the variety of 𝐑\mathbf{R}-modules, the ring 𝐑𝒱\mathbf{R}_{\mathcal{V}} of idempotent binary terms will be isomorphic to 𝐑\mathbf{R}. The result also gives examples of non-finitely based varieties. Consider the ring 𝐑=[x,y]\mathbf{R}=\mathbb{Z}[x,y] with non-commuting variables x,yx,y. Take the ideal II generated by {xynx|n}\{xy^{n}x\ |\ n\in\mathbb{N}\}. Then II is not finitely generated and 𝐑/I\mathbf{R}/I is not finitely presented. The variety of 𝐑/I\mathbf{R}/I-modules is equivalent to the variety 𝒱\mathcal{V} with basic operations +,,0+,-,0 and scaling by xx and yy. Since 𝐑𝒱\mathbf{R}_{\mathcal{V}} is isomorphic to 𝐑/I\mathbf{R}/I, Theorem 3 yields that 𝒱\mathcal{V} (and the variety of 𝐑/I\mathbf{R}/I-modules) is non-finitely based.

We will prove Theorem 3 in Section 5.

2 The Type of an Abelian Variety

In this section we show that abelian Mal’cev varieties are essentialy determind by their binary term functions. This will be used later in the paper.

We begin with a lemma that simplifies proving identities in an abelian Mal’cev variety.

Lemma 4.

Let 𝒱\mathcal{V} be an abelian variety of type σ\sigma with Mal’cev term mm. Let s(x1,,xk,z),t(x1,,xk,z)s(x_{1},\ldots,x_{k},z),t(x_{1},\ldots,x_{k},z) be (k+1)(k+1)-ary terms over σ\sigma. Define s0(z):=s(z,,z).s_{0}(z):=s(z,\ldots,z). For 1ik1\leq i\leq k, define si(x,z):=s(z,,z,x,z,,z),s_{i}(x,z):=s(z,\ldots,z,x,z,\ldots,z), where xx appears in the ii-th position.

  1. 1.

    Then

    𝒱\displaystyle\mathcal{V} t(x1,,xk,z)\displaystyle\models t(x_{1},\ldots,x_{k},z)\approx
    m(m(m(t1(x1,z),t0(z),t2(x2,z)),t0(z),t3(x3,z)),,t0(z),tk(xk,z))\displaystyle m\biggl{(}\ldots m\Bigl{(}m\bigl{(}t_{1}(x_{1},z),t_{0}(z),t_{2}(x_{2},z)\bigl{)},t_{0}(z),t_{3}(x_{3},z)\Bigl{)},\ldots,t_{0}(z),t_{k}(x_{k},z)\biggl{)}
  2. 2.

    Also, 𝒱s(x1,,xk,z)t(x1,,xk,z)\mathcal{V}\models s(x_{1},\ldots,x_{k},z)\approx t(x_{1},\ldots,x_{k},z) if and only if 𝒱siti\mathcal{V}\models s_{i}\approx t_{i} for all 0ik0\leq i\leq k.

Proof.

We prove the first item by induction on kk. The base case of k=0k=0 is trivial. Now assume the claim has been proven for nn-ary terms. Since mm is a Mal’cev term and commutes with every term operation by Proposition 1, the variety 𝒱\mathcal{V} satisfies

t(x1,,xn,z)\displaystyle t(x_{1},\ldots,x_{n},z) t(m(x1,z,z),m(x2,z,z),,m(xn1,z,z),m(z,z,xn),m(z,z,z))\displaystyle\approx t\bigl{(}m(x_{1},z,z),m(x_{2},z,z),\ldots,m(x_{n-1},z,z),m(z,z,x_{n}),m(z,z,z)\bigl{)}
m(t(x1,,xn1,z,z)t(z,,z,z),t(z,z,,z,xn,z))\displaystyle\approx m\bigl{(}t(x_{1},\ldots,x_{n-1},z,z)t(z,\ldots,z,z),t(z,z,\ldots,z,x_{n},z)\bigl{)}
=m(t(x1,,xn1,z,z),t0(z),tn(xn,z))\displaystyle=m\bigl{(}t(x_{1},\ldots,x_{n-1},z,z),t_{0}(z),t_{n}(x_{n},z)\bigl{)}
m(m(m(t1(x1,z),t0(z),t2(x2,z)),t0(z),t3(x3,z)),,t0(z),tn(xn,z)).\displaystyle\approx m\biggl{(}\ldots m\Bigl{(}m\bigl{(}t_{1}(x_{1},z),t_{0}(z),t_{2}(x_{2},z)\bigl{)},t_{0}(z),t_{3}(x_{3},z)\Bigl{)},\ldots,t_{0}(z),t_{n}(x_{n},z)\biggl{)}.

The last equation is by the inductive hypothesis. So we have proven the first item.

The second item follows immediately. ∎

We use the definition of equivalent varieties from [3]. Let 𝒱\mathcal{V} and 𝒲\mathcal{W} be varieties of respective types σ\sigma and ρ\rho. By an interpretation of 𝒱\mathcal{V} in 𝒲\mathcal{W} is meant a mapping DD from σ\sigma to the set of terms over ρ\rho such that:

  1. 1.

    For fσf\in\sigma, if ff is an nn-ary basic operation for n>0n>0, then D(f)=:fDD(f)=:f_{D} is an nn-ary ρ\rho-term.

  2. 2.

    For fσf\in\sigma, if ff is a 0-ary basic operation, then D(f)=:fDD(f)=:f_{D} is a 1-ary ρ\rho-term such that the equation fD(x1)fD(x2)f_{D}(x_{1})\approx f_{D}(x_{2}) holds in 𝒲\mathcal{W}.

  3. 3.

    For every algebra 𝐀𝒲\mathbf{A}\in\mathcal{W}, the algebra 𝐀D:=A,fD𝐀(fσ)\mathbf{A}^{D}:=\langle A,f^{\mathbf{A}}_{D}(f\in\sigma)\rangle is in 𝒱\mathcal{V}.

We say two varieties 𝒱,𝒲\mathcal{V},\mathcal{W} are equivalent if there exists two interpretations, DD of 𝒱\mathcal{V} in 𝒲\mathcal{W} and EE of 𝒲\mathcal{W} in 𝒱\mathcal{V} such that 𝐀DE=𝐀\mathbf{A}^{DE}=\mathbf{A} for all A𝒲A\in\mathcal{W} and 𝐁ED=𝐁\mathbf{B}^{ED}=\mathbf{B} for all 𝐁𝒱\mathbf{B}\in\mathcal{V}.

It will be convenient to find normal forms for terms in an abelian Mal’cev variety. Then we can consider the equivalent variety over the normal forms instead. We first define a variety 𝒰\mathcal{U} such that every abelian Mal’cev variety of finite type is equivalent to a subvariety of 𝒰\mathcal{U}.

Definition 5.

Let 𝒰\mathcal{U} be the variety of type {u1,,u,r1,,rn,m}\{u_{1},\ldots,u_{\ell},r_{1},\ldots,r_{n},m\} where n\ell\leq n, u1,,uu_{1},\ldots,u_{\ell} are unary operations, r1,,rnr_{1},\ldots,r_{n} are binary operations, and mm is a ternary operation defined by

  1. 1.

    m(x,z,z)m(z,z,x)xm(x,z,z)\approx m(z,z,x)\approx x,

  2. 2.

    m(m(x1,x2,x3),m(y1,y2,y3),m(z1,z2,z3))m(m(x1,y1,z1),m(x2,y2,z2),m(x3,y3,z3))m\bigl{(}m(x_{1},x_{2},x_{3}),m(y_{1},y_{2},y_{3}),m(z_{1},z_{2},z_{3})\bigl{)}\approx m\bigl{(}m(x_{1},y_{1},z_{1}),m(x_{2},y_{2},z_{2}),m(x_{3},y_{3},z_{3})\bigl{)}

  3. 3.

    ri(z,z)zr_{i}(z,z)\approx z for all ini\leq n,

  4. 4.

    m(ri(x1,x2),ri(y1,y2),ri(z1,z2))ri(m(x1,y1,z1),m(x2,y2,z2))m\bigl{(}r_{i}(x_{1},x_{2}),r_{i}(y_{1},y_{2}),r_{i}(z_{1},z_{2})\bigl{)}\approx r_{i}\bigl{(}m(x_{1},y_{1},z_{1}),m(x_{2},y_{2},z_{2})\bigl{)} for all ini\leq n,

  5. 5.

    m(ui(x),ui(y),ui(z))ui(m(x,y,z))m\bigl{(}u_{i}(x),u_{i}(y),u_{i}(z)\bigl{)}\approx u_{i}\bigl{(}m(x,y,z)\bigl{)} for all i,i\leq\ell,

  6. 6.

    m(ui(x),ui(z),z)ri(x,z)m(u_{i}(x),u_{i}(z),z)\approx r_{i}(x,z) for all i.i\leq\ell.

Now we prove that every term in an abelian Mal’cev variety decomposes into a composition of the Mal’cev term, binary idempotent terms, and unary terms. Said decomposition gives an equivalent variety 𝒲\mathcal{W} contained in 𝒰\mathcal{U} when the type is finite.

Lemma 6.

Let 𝒱\mathcal{V} be an abelian variety of type σ\sigma with Mal’cev term mm.

  1. 1.

    Then 𝒱\mathcal{V} is equivalent to a variety 𝒲\mathcal{W} of type F𝒱(z)F𝒱id(x,z){m}.F_{\mathcal{V}}(z)\cup F_{\mathcal{V}}^{id}(x,z)\cup\{m\}.

  2. 2.

    If σ\sigma is infinite, then 𝒱\mathcal{V} cannot be finitely based.

  3. 3.

    If σ\sigma is finite, then there exists n\ell\leq n\in\mathbb{N} and u1,,uF𝒱(z)u_{1},\ldots,u_{\ell}\in F_{\mathcal{V}}(z), r1,,rnF𝒱id(x,z)r_{1},\ldots,r_{n}\in F_{\mathcal{V}}^{id}(x,z) such that 𝒱\mathcal{V} is equivalent to a subvariety 𝒲\mathcal{W} of 𝒰\mathcal{U} of type {u1,,u,r1,,rn,m}\{u_{1},\ldots,u_{\ell},r_{1},\ldots,r_{n},m\} as in Definition 5.

Proof.

1. Let 𝒲\mathcal{W} be a variety of type ρ:=F𝒱(z)F𝒱id(x,z){m}.\rho:=F_{\mathcal{V}}(z)\cup F_{\mathcal{V}}^{id}(x,z)\cup\{m\}. Define EE to be an interpretation of 𝒲\mathcal{W} in 𝒱\mathcal{V} as follows: For fρf\in\rho, let E(f)E(f) be a term over σ\sigma that induces ff in 𝐅𝒱(x,y,z).\mathbf{F}_{\mathcal{V}}(x,y,z). For 𝐀𝒱\mathbf{A}\in\mathcal{V}, let 𝐀E\mathbf{A}^{E} be the algebra with universe AA, unary operations u𝐀(z)u^{\mathbf{A}}(z) for uF𝒱(z),u\in F_{\mathcal{V}}(z), binary operations r𝐀(x,z)r^{\mathbf{A}}(x,z) for rF𝒱id(x,z),r\in F_{\mathcal{V}}^{id}(x,z), and ternary operation m𝐀(x,y,z)m^{\mathbf{A}}(x,y,z) for the Mal’cev term mm of 𝒱\mathcal{V}. Here we slightly abuse notation when writing u𝐀u^{\mathbf{A}} for the term function on 𝐀\mathbf{A} that is induced by the term uu over σ\sigma which induces the element u(z)F𝒱(z)u(z)\in F_{\mathcal{V}}(z). Similarly for r𝐀r^{\mathbf{A}}. Then EE is an interpretation of 𝒱\mathcal{V} in 𝒲:={𝐀E|𝐀𝒱}\mathcal{W}:=\{\mathbf{A}^{E}\ |\ \mathbf{A}\in\mathcal{V}\}.

Conversely, define DD to be an interpretation of 𝒱\mathcal{V} in 𝒲\mathcal{W} as follows: Let fσf\in\sigma. In the case that ff is 0-ary, let t(z)F𝒱(z)t(z)\in F_{\mathcal{V}}(z) be any term such that t(z)=ft(z)=f in F𝒱(z)F_{\mathcal{V}}(z). In the case that f(x1,,xk,z)f(x_{1},\ldots,x_{k},z) is (k+1)(k+1)-ary, define f0(z):=f(z,,z)f_{0}(z):=f(z,\ldots,z) and fi(x):=f(z,,z,x,z,,z)f_{i}(x):=f(z,\ldots,z,x,z,\ldots,z), where xx appears in the ii-th position. Then define

D(f):=m(m(m(t1(x1,z),t0(z),t2(x2,z)),t0(z),t3(x3,z)),,t0(z),tk(xk,z)).D(f):=m\biggl{(}\ldots m\Bigl{(}m\bigl{(}t_{1}(x_{1},z),t_{0}(z),t_{2}(x_{2},z)\bigl{)},t_{0}(z),t_{3}(x_{3},z)\Bigl{)},\ldots,t_{0}(z),t_{k}(x_{k},z)\biggl{)}.

Then DD is an interpretation of 𝒲\mathcal{W} in 𝒱\mathcal{V} by Lemma 4.1. Furthermore, for any algebra 𝐀E𝒲\mathbf{A}^{E}\in\mathcal{W}, the algebra 𝐀ED=𝐀\mathbf{A}^{ED}=\mathbf{A} and hence in 𝒱\mathcal{V}.

It follows that for all 𝐀𝒱\mathbf{A}\in\mathcal{V}, we have 𝐀ED=𝐀𝒱\mathbf{A}^{ED}=\mathbf{A}\in\mathcal{V} and 𝐀EDE=𝐀E𝒲\mathbf{A}^{EDE}=\mathbf{A}^{E}\in\mathcal{W}. The latter means that 𝐁DE=𝐁\mathbf{B}^{DE}=\mathbf{B} for all 𝐁𝒲.\mathbf{B}\in\mathcal{W}. Thus 𝒱\mathcal{V} and 𝒲\mathcal{W} are equivalent.

2. Suppose that 𝒱\mathcal{V} has infinite type and suppose by way of contradiction that BB is a finite basis for 𝒱\mathcal{V}. Then there exists a basic operation fσf\in\sigma such that ff does not occur in BB. Say ff has arity kk. Since 𝒱\mathcal{V} is abelian Mal’cev, by Proposition 1, we know that

𝒱\displaystyle\mathcal{V}\models m(f(x1,,xk),f(y1,,yk),f(z1,,zk))\displaystyle\ m\bigl{(}f(x_{1},\ldots,x_{k}),f(y_{1},\ldots,y_{k}),f(z_{1},\ldots,z_{k})\bigl{)}
f(m(x1,y1,z1),,m(xk,yk,zk)).\displaystyle\approx f\bigl{(}m(x_{1},y_{1},z_{1}),\ldots,m(x_{k},y_{k},z_{k})\bigl{)}.

However, BB cannot prove this fact since ff never occurs in BB, contradicting BB being a basis.

3. If σ\sigma is finite, you need finitely many unary terms and idempotent binary terms with mm to interpret every basic operation in σ\sigma. Let us denote said unary terms as u1,,uu_{1},\ldots,u_{\ell}, where \ell\in\mathbb{N}. Now define ri(x,z)=m(ui(x),ui(z),z)r_{i}(x,z)=m(u_{i}(x),u_{i}(z),z) for ii\leq\ell. Let ri(x,z)r_{i}(x,z) be the rest of the needed idempotent binary terms for <in\ell<i\leq n, where nn\geq\ell. Then we need to show that the terms u1,,u,r1,,rnu_{1},\ldots,u_{\ell},r_{1},\ldots,r_{n} satisfy Definition 5.1-6. That would complete the proof.

Equation 1 is satisfied because mm is assumed to be a Mal’cev term. Equations 2,4, and 5 are satisfied by Proposition 1. Equation 3 is satisfied since riF𝒱id(x,z)r_{i}\in F_{\mathcal{V}}^{id}(x,z) for ini\leq n. Lastly, Equation 6 holds by construction. ∎

3 Ring and Module Structure

We now turn our attention to the ring and module structures in abelian Mal’cev varieties. We first prove Lemma 2.

Proof of Lemma 2.

We prove that 𝐑𝒱\mathbf{R}_{\mathcal{V}} is a ring and that 𝐌𝒱\mathbf{M}_{\mathcal{V}} is an 𝐑𝒱\mathbf{R}_{\mathcal{V}}-module by collecting proofs of properties of +,,+,-,\cdot. [2, Lemma 5.6] already shows that F𝒱(x,z),+,,z\langle F_{\mathcal{V}}(x,z),+,-,z\rangle is an abelian group. Let a,b,cF𝒱(x,z)a,b,c\in F_{\mathcal{V}}(x,z) and r,sF𝒱id(x,z).r,s\in F_{\mathcal{V}}^{id}(x,z).

Now we show that F𝒱id(x,z),+,\langle F_{\mathcal{V}}^{id}(x,z),+,-\rangle is a subgroup. Clearly, zF𝒱id(x,z)z\in F_{\mathcal{V}}^{id}(x,z). We have

r(z,z)s(z,z)=zz=z\displaystyle r(z,z)-s(z,z)=z-z=z

So rsF𝒱id(x,z)r-s\in F_{\mathcal{V}}^{id}(x,z) and hence F𝒱id(x,z),+,,z\langle F_{\mathcal{V}}^{id}(x,z),+,-,z\rangle is an abelian group.

Multiplication is associative since

(ab)c=a(b(c(x,z),z))=a(bc).\displaystyle(a\cdot b)\cdot c=a\Bigl{(}b\bigl{(}c(x,z),z\bigl{)}\Bigl{)}=a\cdot(b\cdot c).

We show now that multiplication distributes over addition if the left factor is in F𝒱id(x,z)F_{\mathcal{V}}^{id}(x,z). We have, since rr is idempotent and commutes with mm by Proposition 1, that

r(a+b)\displaystyle r(a+b) =r(m(a,z,b),z)\displaystyle=r\bigl{(}m(a,z,b),z\bigl{)}
=r(m(a,z,b),m(z,z,z))\displaystyle=r\bigl{(}m(a,z,b),m(z,z,z)\bigl{)}
=m(r(a,z),r(z,z),r(b,z))\displaystyle=m\bigl{(}r(a,z),r(z,z),r(b,z)\bigl{)}
=m(r(a,z),z,r(b,z))\displaystyle=m\bigl{(}r(a,z),z,r(b,z)\bigl{)}
=ra+rb.\displaystyle=ra+rb.

Now we show right distributivity. We have

(a+b)c\displaystyle(a+b)\cdot c =m(a(c(x,z),z),z,b(c(x,z),z))\displaystyle=m\Bigl{(}a\bigl{(}c(x,z),z\bigl{)},z,b\bigl{(}c(x,z),z\bigl{)}\Bigl{)}
=ac+bc.\displaystyle=a\cdot c+b\cdot c.

Clearly 1(x,z):=x1(x,z):=x is a multiplicative identity. So 𝐑𝒱\mathbf{R}_{\mathcal{V}} is a ring and F𝒱(x,z)F_{\mathcal{V}}(x,z) forms an 𝐑𝒱\mathbf{R}_{\mathcal{V}}-module with submodules F𝒱id(x,z)F_{\mathcal{V}}^{id}(x,z) and F𝒱(z)F_{\mathcal{V}}(z). To see 3, note that every element aa in F𝒱(x,z)F_{\mathcal{V}}(x,z) can be written uniquely as a sum of an idempotent term and a unary term in the following way

a(x,z)\displaystyle a(x,z) =(a(x,z)a(z,z))+a(z,z).\displaystyle=\bigl{(}a(x,z)-a(z,z)\bigl{)}+a(z,z).

4. We have by [2, Lemma 5.6] that m(a,b,c)=ab+c.m(a,b,c)=a-b+c. Let a,bF𝒱(x,z)a,b\in F_{\mathcal{V}}(x,z) and rF𝒱id(x,z)r\in F_{\mathcal{V}}^{id}(x,z). Then, we have, using the idempotence of rr,

r(a,b)\displaystyle r(a,b) =r(m(a,b,b),m(z,z,b))\displaystyle=r\bigl{(}m(a,b,b),m(z,z,b)\bigl{)}
=m(r(a,z),r(b,z),r(b,b))\displaystyle=m\bigl{(}r(a,z),r(b,z),r(b,b)\bigl{)}
=m(r(a,z),r(b,z),b)\displaystyle=m\bigl{(}r(a,z),r(b,z),b\bigl{)}
=rarb+b\displaystyle=r\cdot a-r\cdot b+b
=ra+(xr)b.\displaystyle=r\cdot a+(x-r)\cdot b.

Lemma 7.

Let 𝒱\mathcal{V} be an abelian Mal’cev variety with Mal’cev terms mm and mm^{\prime}. Then 𝒱m(x,y,z)m(x,y,z)\mathcal{V}\models m(x,y,z)\approx m^{\prime}(x,y,z).

Proof.

Using that mm and mm^{\prime} commute by Proposition 1, we have

𝒱m(x,y,z)\displaystyle\mathcal{V}\models m(x,y,z) m(m(x,y,y),m(y,y,y),m(y,y,z))\displaystyle\approx m\bigl{(}m^{\prime}(x,y,y),m^{\prime}(y,y,y),m^{\prime}(y,y,z)\bigl{)}
m(m(x,y,y),m(y,y,y),m(y,y,z))\displaystyle\approx m^{\prime}\bigl{(}m(x,y,y),m(y,y,y),m(y,y,z)\bigl{)}
m(x,y,z).\displaystyle\approx m^{\prime}(x,y,z).

Hence, the two terms induce the same term operation in 𝒱\mathcal{V}. ∎

We show that for the variety 𝒰\mathcal{U} as in Definition 5, the ring 𝐑𝒰\mathbf{R}_{\mathcal{U}} and the 𝐑𝒰\mathbf{R}_{\mathcal{U}}-module 𝐌𝒰\mathbf{M}_{\mathcal{U}} are free. Further, we characterize finitely generated fully invariant congruences of 𝐅𝒰(x,z)\mathbf{F}_{\mathcal{U}}(x,z) in terms of ideals and submodules of 𝐌𝒰\mathbf{M}_{\mathcal{U}}.

Lemma 8.

Let 𝒰\mathcal{U} be as in Definition 5.

  1. 1.

    𝒰\mathcal{U} is an abelian Mal’cev variety.

  2. 2.

    The ring 𝐑𝒰\mathbf{R}_{\mathcal{U}} is free over generators r1,,rn.r_{1},\ldots,r_{n}.

  3. 3.

    The 𝐑𝒰\mathbf{R}_{\mathcal{U}}-module 𝐌U\mathbf{M}_{U} is free over generators u1,,u_{1},\ldots, u.u_{\ell}.

  4. 4.

    A congruence θ\theta on 𝐅𝒰(x,z)\mathbf{F}_{\mathcal{U}}(x,z) is fully invariant if and only if

    1. (a)

      I:=z/θR𝒰I:=z/\theta\cap R_{\mathcal{U}} is an ideal of 𝐑𝒰\mathbf{R}_{\mathcal{U}},

    2. (b)

      N:=z/θM𝒰N:=z/\theta\cap M_{\mathcal{U}} is an 𝐑𝒰\mathbf{R}_{\mathcal{U}}-submodule of 𝐌𝒰\mathbf{M}_{\mathcal{U}},

    3. (c)

      z/θ=I+Nz/\theta=I+N,

    4. (d)

      IM𝒰NIM_{\mathcal{U}}\subseteq N,

    5. (e)

      and f(x)f(z)xIf(x)-f(z)-x\in I for all f(z)Nf(z)\in N.

  5. 5.

    A fully invariant congruence θ\theta on 𝐅𝒰(x,z)\mathbf{F}_{\mathcal{U}}(x,z) is finitely generated if and only if

    1. (a)

      the ideal I:=z/θR𝒰I:=z/\theta\cap R_{\mathcal{U}} of 𝐑𝒰\mathbf{R}_{\mathcal{U}} is finitely generated and

    2. (b)

      for N:=z/θM𝒰N:=z/\theta\cap M_{\mathcal{U}} the 𝐑𝒰\mathbf{R}_{\mathcal{U}}-module N/IM𝒰N/IM_{\mathcal{U}} is finitely generated.

Proof.

1. Every basic operation commutes with the Mal’cev term mm. So 𝒰\mathcal{U} is an abelian Mal’cev variety by Proposition 1.

2. Let 𝒰\mathcal{U}^{\prime} be the variety of type {r1,,rn,m}\{r_{1},\ldots,r_{n},m\} satisfying identities 1-4 of 𝒰\mathcal{U}. Note that this makes 𝒰\mathcal{U}^{\prime} an idempotent reduct of 𝒰\mathcal{U} and +,+,- are still term operations of 𝒰\mathcal{U}^{\prime} and zz is still the additive identity. Also, 𝒰\mathcal{U} and 𝒰\mathcal{U}^{\prime} are abelian varieties. The unique Mal’cev term is m(x,y,z)=xy+zm(x,y,z)=x-y+z by Lemma 7. It commutes with every basic operation by Proposition 1.

First, we claim

F𝒰(x,z)=F𝒰id(x,z).\displaystyle F_{\mathcal{U}^{\prime}}(x,z)=F_{\mathcal{U}}^{id}(x,z). (1)

Clearly, F𝒰(x,z)F𝒰id(x,z)F_{\mathcal{U}^{\prime}}(x,z)\subseteq F_{\mathcal{U}}^{id}(x,z). All that is left is to show F𝒰(x,z)F𝒰id(x,z)F_{\mathcal{U}^{\prime}}(x,z)\supseteq F_{\mathcal{U}}^{id}(x,z). Note that every sF𝒰id(x,z)s\in F_{\mathcal{U}}^{id}(x,z) has the form

t(x,z)t(z,z)t(x,z)-t(z,z)

for some tF𝒰(x,z)t\in F_{\mathcal{U}}(x,z). We then induct on tt to show that sF𝒰(x,z)s\in F_{\mathcal{U}^{\prime}}(x,z).

For the first base case, suppose t=xt=x. Then s=xz=xF𝒰(x,z)s=x-z=x\in F_{\mathcal{U}^{\prime}}(x,z). Now, as a second base case, suppose t=zt=z. Then s=zz=zF𝒰(x,z)s=z-z=z\in F_{\mathcal{U}^{\prime}}(x,z).

For the induction step, we consider several cases. First, suppose that t=ui(v(x,z))t=u_{i}(v(x,z)) for some ii\leq\ell and some vF𝒰(x,z)v\in F_{\mathcal{U}}(x,z). We have

s(x,z)\displaystyle s(x,z) =ui(v(x,z))ui(v(z,z))\displaystyle=u_{i}\bigl{(}v(x,z)\bigl{)}-u_{i}\bigl{(}v(z,z)\bigl{)}
=ui(v(x,z))ui(v(z,z))+ui(z)ui(z)\displaystyle=u_{i}\bigl{(}v(x,z)\bigl{)}-u_{i}\bigl{(}v(z,z)\bigl{)}+u_{i}(z)-u_{i}(z)
=ui(v(x,z)v(z,z)+z)ui(z)\displaystyle=u_{i}\bigl{(}v(x,z)-v(z,z)+z\bigl{)}-u_{i}(z)
=ri(v(x,z)v(z,z)),\displaystyle=r_{i}\bigl{(}v(x,z)-v(z,z)\bigl{)},

which is in F𝒰(x,z)F_{\mathcal{U}^{\prime}}(x,z) since v(x,z)v(z,z)F𝒰(x,z)v(x,z)-v(z,z)\in F_{\mathcal{U}^{\prime}}(x,z) by the inductive hypothesis.

For another induction case, suppose that t=ri(v,w)t=r_{i}(v,w) for some ini\leq n and some v,wF𝒰(x,z)v,w\in F_{\mathcal{U}}(x,z). Then

s(x,z)\displaystyle s(x,z) =ri(v(x,z),w(x,z))ri(v(z,z),w(z,z))\displaystyle=r_{i}\bigl{(}v(x,z),w(x,z)\bigl{)}-r_{i}\bigl{(}v(z,z),w(z,z)\bigl{)}
=ri(v(x,z),w(x,z))ri(v(z,z),w(z,z))+ri(z,z) since ri is idempotent\displaystyle=r_{i}\bigl{(}v(x,z),w(x,z)\bigl{)}-r_{i}\bigl{(}v(z,z),w(z,z)\bigl{)}+r_{i}(z,z)\text{ since }r_{i}\text{ is idempotent}
=ri(v(x,z)v(z,z)+z,w(x,z)w(z,z)+z),\displaystyle=r_{i}\bigl{(}v(x,z)-v(z,z)+z,w(x,z)-w(z,z)+z\bigl{)},

which is in F𝒰(x,z)F_{\mathcal{U}^{\prime}}(x,z) since v(x,z)v(z,z),w(x,z)w(z,z)F𝒰(x,z)v(x,z)-v(z,z),w(x,z)-w(z,z)\in F_{\mathcal{U}^{\prime}}(x,z) by the inductive hypothesis.

For the last induction step, suppose that t=m(v,w,y)t=m(v,w,y) for some v,w,yF𝒰(x,z)v,w,y\in F_{\mathcal{U}}(x,z). Using Lemma 2.4, we have

s(x,z)\displaystyle s(x,z) =(v(x,z)w(x,z)+y(x,z))(v(z,z)w(z,z)+y(z,z))\displaystyle=\bigl{(}v(x,z)-w(x,z)+y(x,z)\bigl{)}-\bigl{(}v(z,z)-w(z,z)+y(z,z)\bigl{)}
=(v(x,z)v(z,z))(w(x,z)w(z,z))+(y(x,z)y(z,z))\displaystyle=\bigl{(}v(x,z)-v(z,z)\bigl{)}-\bigl{(}w(x,z)-w(z,z)\bigl{)}+\bigl{(}y(x,z)-y(z,z)\bigl{)}
=m(v(x,z)v(z,z),w(x,z)w(z,z),y(x,z)y(z,z)),\displaystyle=m\bigl{(}v(x,z)-v(z,z),w(x,z)-w(z,z),y(x,z)-y(z,z)\bigl{)},

which is in F𝒰(x,z)F_{\mathcal{U}^{\prime}}(x,z) since v(x,z)v(z,z),w(x,z)w(z,z),y(x,z)y(z,z)F𝒰(x,z)v(x,z)-v(z,z),w(x,z)-w(z,z),y(x,z)-y(z,z)\in F_{\mathcal{U}^{\prime}}(x,z) by inductive hypothesis.

So we have shown F𝒰(x,z)=F𝒰id(x,z)F_{\mathcal{U}^{\prime}}(x,z)=F_{\mathcal{U}}^{id}(x,z). It follows that 𝐑𝒰=𝐑𝒰.\mathbf{R}_{\mathcal{U}}=\mathbf{R}_{\mathcal{U}^{\prime}}.

Now we claim

𝐑𝒰T,+,, the free ring over generators 𝐫1,,𝐫n.\displaystyle\mathbf{R}_{\mathcal{U}^{\prime}}\cong\langle T,+,\cdot\rangle,\text{ the free ring over generators }\mathbf{r}_{1},\ldots,\mathbf{r}_{n}. (2)

We will write elements of TT in boldface to better distinguish them from elements of F𝒰(x,z).F_{\mathcal{U}^{\prime}}(x,z).

Clearly,

ψ:TF𝒰(x,z),𝐫iri for in,\psi\colon T\to F_{\mathcal{U}^{\prime}}(x,z),\ \mathbf{r}_{i}\mapsto r_{i}\text{ for }i\leq n,

defines a ring homomorphism. To show ψ\psi is bijective, we will make TT an algebra in 𝒰\mathcal{U}^{\prime}. Then we will construct a homomorphism

τ:F𝒰(x,z)T\tau\colon F_{\mathcal{U}^{\prime}}(x,z)\to T

such that τψ\tau\circ\psi is the identity on TT and ψτ\psi\circ\tau is the identity on F𝒰(x,z)F_{\mathcal{U}^{\prime}}(x,z). We first have to define operations r1𝐓,,rn𝐓,m𝐓r_{1}^{\mathbf{T}},\ldots,r_{n}^{\mathbf{T}},m^{\mathbf{T}} on TT such that 𝐓:=T,r1𝐓,,rn𝐓,m𝐓𝒰\mathbf{T}:=\langle T,r_{1}^{\mathbf{T}},\ldots,r_{n}^{\mathbf{T}},m^{\mathbf{T}}\rangle\in\mathcal{U}^{\prime}. For 𝐬1,𝐬2,𝐬3T\mathbf{s}_{1},\mathbf{s}_{2},\mathbf{s}_{3}\in T and ini\leq n define

ri𝐓(𝐬1,𝐬2):\displaystyle r_{i}^{\mathbf{T}}(\mathbf{s}_{1},\mathbf{s}_{2}): =𝐫i𝐬1+(𝟏𝐫i)𝐬2,\displaystyle=\mathbf{r}_{i}\mathbf{s}_{1}+(\mathbf{1}-\mathbf{r}_{i})\mathbf{s}_{2},
m𝐓(𝐬1,𝐬2,𝐬3):\displaystyle m^{\mathbf{T}}(\mathbf{s}_{1},\mathbf{s}_{2},\mathbf{s}_{3}): =𝐬1𝐬2+𝐬3.\displaystyle=\mathbf{s}_{1}-\mathbf{s}_{2}+\mathbf{s}_{3}.

From the definition, it is clear that m𝐓m^{\mathbf{T}} is a Mal’cev operation that commutes with itself and the idempotent affine operations ri𝐓r_{i}^{\mathbf{T}} for all ini\leq n. Hence 𝐓\mathbf{T} is in 𝒰\mathcal{U}^{\prime}.

Now we show that 𝐓\mathbf{T} and 𝐅𝒰(x,z)\mathbf{F}_{\mathcal{U}^{\prime}}(x,z) are isomorphic. Let τ:𝐅𝒰(x,z)𝐓\tau\colon\mathbf{F}_{\mathcal{U}^{\prime}}(x,z)\to\mathbf{T} be the homomorphism of algebras in 𝒰\mathcal{U}^{\prime} defined by x𝟏x\mapsto\mathbf{1} and z𝟎z\mapsto\mathbf{0}. We show that τψ\tau\circ\psi is the identity map on TT and that ψτ\psi\circ\tau is the identity map on F𝒰(x,z).F_{\mathcal{U}^{\prime}}(x,z).

We show that

τ(ψ(𝐬))=𝐬 for all 𝐬T\tau(\psi(\mathbf{s}))=\mathbf{s}\text{ for all }\mathbf{s}\in T (3)

via induction on 𝐬\mathbf{s}. For the base case, we have τ(ψ(𝟎))=τ(z)=𝟎\tau(\psi(\mathbf{0}))=\tau(z)=\mathbf{0} and τ(ψ(𝟏))=τ(x)=𝟏\tau(\psi(\mathbf{1}))=\tau(x)=\mathbf{1}. For ini\leq n, we have

τ(ψ(𝐫i))\displaystyle\tau(\psi(\mathbf{r}_{i})) =τ(ri(x,z))\displaystyle=\tau(r_{i}(x,z))
=ri𝐓(τ(x),τ(z))\displaystyle=r_{i}^{\mathbf{T}}(\tau(x),\tau(z))
=ri𝐓(𝟏,𝟎)\displaystyle=r_{i}^{\mathbf{T}}(\mathbf{1},\mathbf{0})
=𝐫i𝟏+(𝟏𝐫i)𝟎\displaystyle=\mathbf{r}_{i}\mathbf{1}+(\mathbf{1}-\mathbf{r}_{i})\mathbf{0}
=𝐫i.\displaystyle=\mathbf{r}_{i}.

For the induction step, let 𝐬1,𝐬2T\mathbf{s}_{1},\mathbf{s}_{2}\in T such that τ(ψ(𝐬i))=𝐬i\tau(\psi(\mathbf{s}_{i}))=\mathbf{s}_{i} for i2i\leq 2. Then

τ(ψ(𝐬1+𝐬2))\displaystyle\tau\bigl{(}\psi(\mathbf{s}_{1}+\mathbf{s}_{2})\bigl{)} =τ(ψ(𝐬1)+ψ(𝐬2))\displaystyle=\tau\bigl{(}\psi(\mathbf{s}_{1})+\psi(\mathbf{s}_{2})\bigl{)}
=τ(m(ψ(𝐬1),z,ψ(𝐬2)))\displaystyle=\tau\Bigl{(}m\bigl{(}\psi(\mathbf{s}_{1}),z,\psi(\mathbf{s}_{2})\bigl{)}\Bigl{)}
=m𝐓(τ(ψ(𝐬1)),τ(z),τ(ψ(𝐬2)))\displaystyle=m^{\mathbf{T}}\Bigl{(}\tau\bigl{(}\psi(\mathbf{s}_{1})\bigl{)},\tau(z),\tau\bigl{(}\psi(\mathbf{s}_{2})\bigl{)}\Bigl{)}
=m𝐓(𝐬1,𝟎,𝐬2)\displaystyle=m^{\mathbf{T}}(\mathbf{s}_{1},\mathbf{0},\mathbf{s}_{2})
=𝐬1𝟎+𝐬2\displaystyle=\mathbf{s}_{1}-\mathbf{0}+\mathbf{s}_{2}
=𝐬1+𝐬2.\displaystyle=\mathbf{s}_{1}+\mathbf{s}_{2}.

Also,

τ(ψ(𝐬1𝐬2))\displaystyle\tau\bigl{(}\psi(\mathbf{s}_{1}\mathbf{s}_{2})\bigl{)} =τ(ψ(𝐬1)ψ(𝐬2))\displaystyle=\tau\bigl{(}\psi(\mathbf{s}_{1})\psi(\mathbf{s}_{2})\bigl{)}
=τ(ψ(𝐬1)(ψ(𝐬2),z))\displaystyle=\tau\bigl{(}\psi(\mathbf{s}_{1})(\psi(\mathbf{s}_{2}),z)\bigl{)}
=τ(ψ(𝐬1))(τ(ψ(𝐬2)),τ(z))\displaystyle=\tau\bigl{(}\psi(\mathbf{s}_{1})\bigl{)}\Bigl{(}\tau\bigl{(}\psi(\mathbf{s}_{2})\bigl{)},\tau(z)\Bigl{)}
=𝐬1(𝐬2,𝟎)\displaystyle=\mathbf{s}_{1}(\mathbf{s}_{2},\mathbf{0})
=𝐬1𝐬2+(𝟏𝐬1)𝟎\displaystyle=\mathbf{s}_{1}\mathbf{s}_{2}+(\mathbf{1}-\mathbf{s}_{1})\mathbf{0}
=𝐬1𝐬2.\displaystyle=\mathbf{s}_{1}\mathbf{s}_{2}.

So we have shown that τψ\tau\circ\psi is the identity map on TT.

Conversely, we show that

ψ(τ(s))=s for all sF𝒰(x,z)\psi(\tau(s))=s\text{ for all }s\in F_{\mathcal{U}^{\prime}}(x,z) (4)

via induction on ss. For the base cases, we have ψ(τ(x))=ψ(𝟏)=x\psi(\tau(x))=\psi(\mathbf{1})=x and ψ(τ(z))=ψ(𝟎)=z\psi(\tau(z))=\psi(\mathbf{0})=z.

Let s1,s2,s3F𝒰(x,z)s_{1},s_{2},s_{3}\in F_{\mathcal{U}^{\prime}}(x,z) such that ψ(τ(sj))=sj\psi(\tau(s_{j}))=s_{j} for j3.j\leq 3. Then

ψ(τ(m(s1,s2,s3)))\displaystyle\psi\Bigl{(}\tau\bigl{(}m(s_{1},s_{2},s_{3})\bigl{)}\Bigl{)} =ψ(m𝐓(τ(s1),τ(s2),τ(s3)))\displaystyle=\psi\Bigl{(}m^{\mathbf{T}}\bigl{(}\tau(s_{1}),\tau(s_{2}),\tau(s_{3})\bigl{)}\Bigl{)}
=ψ(τ(s1)τ(s2)+τ(s3))\displaystyle=\psi\bigl{(}\tau(s_{1})-\tau(s_{2})+\tau(s_{3})\bigl{)}
=ψ(τ(s1))ψ(τ(s2))+ψ(τ(s3))\displaystyle=\psi\bigl{(}\tau(s_{1})\bigl{)}-\psi\bigl{(}\tau(s_{2})\bigl{)}+\psi\bigl{(}\tau(s_{3})\bigl{)}
=s1s2+s3\displaystyle=s_{1}-s_{2}+s_{3}
=m(s1,s2,s3) by Lemma 2.4.\displaystyle=m(s_{1},s_{2},s_{3})\text{ by Lemma }\ref{small fring}.\ref{ops are ops}.

Also, for ini\leq n,

ψ(τ(ri(s1,s2)))\displaystyle\psi\Bigl{(}\tau\bigl{(}r_{i}(s_{1},s_{2})\bigl{)}\Bigl{)} =ψ(ri𝐓(τ(s1),τ(s2)))\displaystyle=\psi\Bigl{(}r_{i}^{\mathbf{T}}\bigl{(}\tau(s_{1}),\tau(s_{2})\bigl{)}\Bigl{)}
=ψ(𝐫iτ(s1)+(𝟏𝐫i)τ(s2))\displaystyle=\psi(\mathbf{r}_{i}\tau(s_{1})+(\mathbf{1}-\mathbf{r}_{i})\tau(s_{2}))
=ψ(𝐫i)ψ(τ(s1))+ψ(𝟏𝐫i)ψ(τ(s2))\displaystyle=\psi(\mathbf{r}_{i})\psi\bigl{(}\tau(s_{1})\bigl{)}+\psi(\mathbf{1}-\mathbf{r}_{i})\psi\bigl{(}\tau(s_{2})\bigl{)}
=ris1+(1ri)s2\displaystyle=r_{i}s_{1}+(1-r_{i})s_{2}
=ri(s1,s2) by Lemma 2.4.\displaystyle=r_{i}(s_{1},s_{2})\text{ by Lemma }\ref{small fring}.\ref{ops are ops}.

So we have shown that ψτ\psi\circ\tau is the identity on F𝒰(x,z)F_{\mathcal{U}^{\prime}}(x,z). Together with the fact that τψ\tau\circ\psi is the identity on TT, we have shown (2). This, along with (1), proves Lemma 8.2.

3. We want to show

𝐌𝒰V,+,R𝒰, the free 𝐑𝒰-module over generators 𝐮1,,𝐮.\mathbf{M}_{\mathcal{U}}\cong\langle V,+,R_{\mathcal{U}}\rangle,\text{ the free }\mathbf{R}_{\mathcal{U}}\text{-module over generators }\mathbf{u}_{1},\ldots,\mathbf{u}_{\ell}. (5)

We write elements of VV in boldface to better distinguish them from elements in F𝒰(z)F_{\mathcal{U}}(z). Clearly, ψ:VF𝒰(z)\psi\colon V\to F_{\mathcal{U}}(z), 𝐮iui\mathbf{u}_{i}\mapsto u_{i} for ii\leq\ell, defines an 𝐑𝒰\mathbf{R}_{\mathcal{U}}-module homomorphism. To show that ψ\psi is bijective, we will make VV an algebra in 𝒰\mathcal{U}. Then we will construct a homomorphism τ:F𝒰(z)V\tau\colon F_{\mathcal{U}}(z)\to V such that τψ\tau\circ\psi is the identity map on VV and ψτ\psi\circ\tau is the identity map on F𝒰(z)F_{\mathcal{U}}(z). We then have to define operations u1𝐕,,u𝐕,r1𝐕,,rn𝐕,m𝐕u_{1}^{\mathbf{V}},\ldots,u_{\ell}^{\mathbf{V}},r_{1}^{\mathbf{V}},\ldots,r_{n}^{\mathbf{V}},m^{\mathbf{V}} on VV such that 𝐕:=V,u1𝐕,,u𝐕,r1𝐕,,rn𝐕,m𝐕\mathbf{V}:=\langle V,u_{1}^{\mathbf{V}},\ldots,u_{\ell}^{\mathbf{V}},r_{1}^{\mathbf{V}},\ldots,r_{n}^{\mathbf{V}},m^{\mathbf{V}}\rangle is in 𝒰\mathcal{U}. For 𝐯1,𝐯2,𝐯3V\mathbf{v}_{1},\mathbf{v}_{2},\mathbf{v}_{3}\in V, ii\leq\ell and jnj\leq n define

ui𝐕(𝐯𝟏)\displaystyle u_{i}^{\mathbf{V}}(\mathbf{v_{1}}) :=ri𝐯𝟏+𝐮i,\displaystyle:=r_{i}\mathbf{v_{1}}+\mathbf{u}_{i},
rj𝐕(𝐯1,𝐯2):\displaystyle r_{j}^{\mathbf{V}}(\mathbf{v}_{1},\mathbf{v}_{2}): =rj𝐯1+(1rj)𝐯2,\displaystyle=r_{j}\mathbf{v}_{1}+(1-r_{j})\mathbf{v}_{2},
m𝐕(𝐯1,𝐯2,𝐯3):\displaystyle m^{\mathbf{V}}(\mathbf{v}_{1},\mathbf{v}_{2},\mathbf{v}_{3}): =𝐯1𝐯2+𝐯3.\displaystyle=\mathbf{v}_{1}-\mathbf{v}_{2}+\mathbf{v}_{3}.

The definition clearly shows that m𝐕m^{\mathbf{V}} is a Mal’cev operation. It is also clear that m𝐕m^{\mathbf{V}} commutes with itself and the idempotent affine operations ri𝐕r_{i}^{\mathbf{V}} for all ini\leq n and the unary operations ui𝐕u_{i}^{\mathbf{V}}. Hence, 𝐕𝒰.\mathbf{V}\in\mathcal{U}.

Now we show that 𝐕\mathbf{V} and 𝐅𝒰(z)\mathbf{F}_{\mathcal{U}}(z) are isomorphic. Let τ:𝐅𝒰(z)𝐕\tau\colon\mathbf{F}_{\mathcal{U}}(z)\to\mathbf{V} be the homomorphism of algebras in 𝒰\mathcal{U} defined by z𝟎z\mapsto\mathbf{0}. We show that τψ\tau\circ\psi is the identity on VV and that ψτ\psi\circ\tau is the identity on F𝒰(z).F_{\mathcal{U}}(z).

We show that

τ(ψ(𝐯))=𝐯 for all 𝐯V\tau(\psi(\mathbf{v}))=\mathbf{v}\text{ for all }\mathbf{v}\in V (6)

via induction on 𝐯\mathbf{v}. For the first base case, we have τ(ψ(𝟎))=τ(z)=𝟎\tau(\psi(\mathbf{0}))=\tau(z)=\mathbf{0}. For ii\leq\ell we have

τ(ψ(𝐮i)\displaystyle\tau(\psi(\mathbf{u}_{i}) =τ(ui(z))=ui𝐕(τ(z))\displaystyle=\tau(u_{i}(z))=u_{i}^{\mathbf{V}}(\tau(z))
=ui𝐕(𝟎)\displaystyle=u_{i}^{\mathbf{V}}(\mathbf{0})
=ri𝟎+𝐮i\displaystyle=r_{i}\mathbf{0}+\mathbf{u}_{i}
=𝐮i.\displaystyle=\mathbf{u}_{i}.

For the induction step, let 𝐯,𝐰V\mathbf{v},\mathbf{w}\in V such that τ(ψ(𝐯))=𝐯\tau(\psi(\mathbf{v}))=\mathbf{v} and τ(ψ(𝐰))=𝐰\tau(\psi(\mathbf{w}))=\mathbf{w}. Then

τ(ψ(𝐯+𝐰))\displaystyle\tau\bigl{(}\psi(\mathbf{v}+\mathbf{w})\bigl{)} =τ(ψ(𝐯)+ψ(𝐰))\displaystyle=\tau\bigl{(}\psi(\mathbf{v})+\psi(\mathbf{w})\bigl{)}
=m𝐕(τ(ψ(𝐯)),τ(z),τ(ψ(𝐰)))\displaystyle=m^{\mathbf{V}}\Bigl{(}\tau\bigl{(}\psi(\mathbf{v})\bigl{)},\tau(z),\tau\bigl{(}\psi(\mathbf{w})\bigl{)}\Bigl{)}
=m𝐕(𝐯,𝟎,𝐰)\displaystyle=m^{\mathbf{V}}(\mathbf{v},\mathbf{0},\mathbf{w})
=𝐯+𝐰.\displaystyle=\mathbf{v}+\mathbf{w}.

Let rF𝒰id(x,z)r\in F_{\mathcal{U}}^{id}(x,z). Then

τ(ψ(r𝐯))\displaystyle\tau\bigl{(}\psi(r\mathbf{v})\bigl{)} =τ(rψ(𝐯))\displaystyle=\tau\bigl{(}r\psi(\mathbf{v})\bigl{)}
=r𝐕(τ(ψ(𝐯)),τ(z))\displaystyle=r^{\mathbf{V}}\Bigl{(}\tau\bigl{(}\psi(\mathbf{v})\bigl{)},\tau(z)\Bigl{)}
=r𝐕(𝐯,𝟎)\displaystyle=r^{\mathbf{V}}(\mathbf{v},\mathbf{0})
=r𝐯+(1r)𝟎\displaystyle=r\mathbf{v}+(1-r)\mathbf{0}
=r𝐯.\displaystyle=r\mathbf{v}.

So we have shown that τψ\tau\circ\psi is the identity map on VV.

Conversely, we show that

ψ(τ(t)=t for all tF𝒰(z)\psi(\tau(t)=t\text{ for all }t\in F_{\mathcal{U}}(z) (7)

via induction on t(z)t(z). For the base case we have ψ(τ(z))=ψ(𝟎)=z.\psi(\tau(z))=\psi(\mathbf{0})=z.

For the induction steps, let t1,t2,t3F𝒰(z)t_{1},t_{2},t_{3}\in F_{\mathcal{U}}(z) such that ψ(τ(tj))=tj\psi(\tau(t_{j}))=t_{j} for j3j\leq 3. Then

ψ(τ(m(t1,t2,t3)))\displaystyle\psi\Bigl{(}\tau\bigl{(}m(t_{1},t_{2},t_{3})\bigl{)}\Bigl{)} =ψ(m𝐕(τ(t1),τ(t2),τ(t3)))\displaystyle=\psi\Bigl{(}m^{\mathbf{V}}\bigl{(}\tau(t_{1}),\tau(t_{2}),\tau(t_{3})\bigl{)}\Bigl{)}
=ψ(τ(t1)τ(t2)+τ(t3))\displaystyle=\psi\bigl{(}\tau(t_{1})-\tau(t_{2})+\tau(t_{3})\bigl{)}
=ψ(τ(t1))ψ(τ(t2))+ψ(τ(t3))\displaystyle=\psi\bigl{(}\tau(t_{1})\bigl{)}-\psi\bigl{(}\tau(t_{2})\bigl{)}+\psi\bigl{(}\tau(t_{3})\bigl{)}
=t1t2+t3\displaystyle=t_{1}-t_{2}+t_{3}
=m(t1,t2,t3).\displaystyle=m(t_{1},t_{2},t_{3}).

Suppose ini\leq n. Then

ψ(τ(ri(t1,t2)))\displaystyle\psi\Bigl{(}\tau\bigl{(}r_{i}(t_{1},t_{2})\bigl{)}\Bigl{)} =ψ(ri𝐕(τ(t1),τ(t2)))\displaystyle=\psi\Bigl{(}r_{i}^{\mathbf{V}}\bigl{(}\tau(t_{1}),\tau(t_{2})\bigl{)}\Bigl{)}
=ψ(riτ(t1)+(1ri)τ(t2))\displaystyle=\psi\bigl{(}r_{i}\tau(t_{1})+(1-r_{i})\tau(t_{2})\bigl{)}
=riψ(τ(t1))+(1ri)ψ(τ(t2))\displaystyle=r_{i}\psi\bigl{(}\tau(t_{1})\bigl{)}+(1-r_{i})\psi\bigl{(}\tau(t_{2}))
=rit1+(1ri)t2\displaystyle=r_{i}t_{1}+(1-r_{i})t_{2}
=ri(t1,t2) by Lemma 2.4.\displaystyle=r_{i}(t_{1},t_{2})\text{ by Lemma }\ref{small fring}.\ref{ops are ops}.

Suppose that ii\leq\ell. We have

ψ(τ(ui(t)))\displaystyle\psi\Bigl{(}\tau\bigl{(}u_{i}(t)\bigl{)}\Bigl{)} =ψ(ui𝐕(τ(t)))\displaystyle=\psi\Bigl{(}u_{i}^{\mathbf{V}}\bigl{(}\tau(t)\bigl{)}\Bigl{)}
=ψ(riτ(t)+𝐮i)\displaystyle=\psi\bigl{(}r_{i}\tau(t)+\mathbf{u}_{i}\bigl{)}
=riψ(τ(t))+ψ(𝐮i)\displaystyle=r_{i}\psi\bigl{(}\tau(t)\bigl{)}+\psi(\mathbf{u}_{i})
=rit+ui(z)\displaystyle=r_{i}t+u_{i}(z)
=ui(t)ui(z)+ui(z)\displaystyle=u_{i}(t)-u_{i}(z)+u_{i}(z)
=ui(t).\displaystyle=u_{i}(t).

So we have shown ψτ\psi\circ\tau is the identity map on M𝒰M_{\mathcal{U}}. Together with τψ\tau\circ\psi being the identity map on VV, this shows (5). This proves Lemma 8.3.

4. We first prove the forward direction. Assume that θ\theta is a fully invariant congruence of 𝐅𝒰(x,z)\mathbf{F}_{\mathcal{U}}(x,z). We prove 4a. That is, we prove that z/θR𝒰=Iz/\theta\cap R_{\mathcal{U}}=I is an ideal of 𝐑𝒰\mathbf{R}_{\mathcal{U}}. II is clearly an additive subgroup of 𝐑𝒰\mathbf{R}_{\mathcal{U}}. To prove II is multiplicatively closed, consider sR𝒰s\in R_{\mathcal{U}} and tIt\in I. To see st,tsR𝒰s\cdot t,t\cdot s\in R_{\mathcal{U}}, note that 𝐑𝒰\mathbf{R}_{\mathcal{U}} is a ring and thus multiplicatively closed.

To prove that st,tsz/θs\cdot t,t\cdot s\in z/\theta, note that because θ\theta is a congruence, we have

s(t(x,z),z)θs(z,z)=z.\displaystyle s(t(x,z),z)\ \theta\ s(z,z)=z.

Since θ\theta is a fully invariant congruence, we have that t(x,z)θzt(x,z)\ \theta\ z implies

t(s(x,z),z)θz.\displaystyle t(s(x,z),z)\ \theta\ z.

Hence st,tsIs\cdot t,t\cdot s\in I and II is an ideal. This proves 4a.

Now we show 4b. That is, we show that z/θM𝒰=Nz/\theta\cap M_{\mathcal{U}}=N is an 𝐑𝒰\mathbf{R}_{\mathcal{U}}-submodule of 𝐌𝒰\mathbf{M}_{\mathcal{U}}. Clearly, NN is an additive subgroup. To show NN is closed under the action by 𝐑𝒰\mathbf{R}_{\mathcal{U}}, consider sR𝒰s\in R_{\mathcal{U}}, vNv\in N. Since θ\theta is a congruence,

s(v,z)θs(z,z)=z.\displaystyle s(v,z)\ \theta\ s(z,z)=z.

Hence svNs\cdot v\in N and NN is an 𝐑𝒰\mathbf{R}_{\mathcal{U}}-submodule of 𝐌\mathbf{M}. This proves 4b.

We now show 4c. For any s(x,z)θzs(x,z)\ \theta\ z, note that s(x,z)s(z,z)Is(x,z)-s(z,z)\in I and s(z,z)Ns(z,z)\in N. Hence z/θ=I+N.z/\theta=I+N. This proves 4c.

We now show 4d, that IM𝒰N.IM_{\mathcal{U}}\subseteq N. Let sIs\in I and vM𝒰v\in M_{\mathcal{U}}. Then

s(v,z)\displaystyle s(v,z) θz since θ is fully invariant.\displaystyle\ \theta\ z\text{ since }\theta\text{ is fully invariant.}

Since s(v,z)M𝒰s(v,z)\in M_{\mathcal{U}} and s(v,z)z/θs(v,z)\in z/\theta, we have that s(v,z)Ns(v,z)\in N. So IM𝒰NIM_{\mathcal{U}}\subseteq N. This proves 4d.

We now show 4e. That is, we show that f(x)f(z)xIf(x)-f(z)-x\in I for all f(z)Nf(z)\in N. Suppose that f(z)Nf(z)\in N. Then f(x)θxf(x)\ \theta\ x because θ\theta is fully invariant. Also f(x)f(z)θxzf(x)-f(z)\ \theta\ x-z, and f(x)f(z)xθzf(x)-f(z)-x\ \theta\ z. Also, f(x)f(z)xR𝒰f(x)-f(z)-x\in R_{\mathcal{U}} so f(x)f(z)xIf(x)-f(z)-x\in I. This proves 4e. We have proven Properties 4a-4e.

Now we prove the converse. Suppose that θ,I,N\theta,I,N satisfy properties 4a-4e. We introduce some notation. For fF𝒰(x,z)f\in F_{\mathcal{U}}(x,z), define fid(x,z):=f(x,z)f(z,z)f^{id}(x,z):=f(x,z)-f(z,z) and fu(z):=f(z,z)f^{u}(z):=f(z,z). Note that f(x,z)=fid(x,z)+fu(z).f(x,z)=f^{id}(x,z)+f^{u}(z).

Now we show that θ\theta is fully invariant. Let t(x,z)z/θt(x,z)\in z/\theta and v,wF𝒰(x,z)v,w\in F_{\mathcal{U}}(x,z). We need to show that t(v,w)θwt(v,w)\ \theta\ w. Since t(v,w)=m(tid(v,w),w,tu(w))t(v,w)=m(t^{id}(v,w),w,t^{u}(w)) and m(w,w,w)=wm(w,w,w)=w, it will suffice to show that tid(v,w)θwθtu(w)t^{id}(v,w)\ \theta\ w\ \theta\ t^{u}(w).

Note that for rIr\in I and sF𝒰(x,z)s\in F_{\mathcal{U}}(x,z), we have

rs\displaystyle r\cdot s =r(sid+su)\displaystyle=r\cdot(s^{id}+s^{u})
=rsidI+rsuIM𝒰N\displaystyle=\underbrace{r\cdot s^{id}}_{\in I}+\underbrace{r\cdot s^{u}}_{\in IM_{\mathcal{U}}\subseteq N}
θz.\displaystyle\ \theta\ z.

Since tz/θt\in z/\theta, we have tidz/θR𝒰=It^{id}\in z/\theta\cap R_{\mathcal{U}}=I and tuz/θM𝒰=Nt^{u}\in z/\theta\cap M_{\mathcal{U}}=N.

We have

tid(v,w)\displaystyle t^{id}(v,w) =tidv+(xtid)w\displaystyle=t^{id}\cdot v+(x-t^{id})\cdot w
=tidv+wtidw\displaystyle=t^{id}\cdot v+w-t^{id}\cdot w
θw.\displaystyle\ \theta\ w.

Since tu(z)Nt^{u}(z)\in N, by 4e, tu(x)tu(z)xIt^{u}(x)-t^{u}(z)-x\in I. So we have that

tu(w)zw\displaystyle t^{u}(w)-z-w θ(tu(x)tu(z)x)w\displaystyle\ \theta\ (t^{u}(x)-t^{u}(z)-x)\cdot w
θz.\displaystyle\ \theta\ z.

This implies that tu(w)θwt^{u}(w)\ \theta\ w and we are done.

5. We prove the forward direction first. Suppose that θ\theta is a finitely generated fully invariant congruence of 𝐅𝒰(x,z).\mathbf{F}_{\mathcal{U}}(x,z). Since (s,t)θ(s,t)\in\theta if and only if (m(s,t,z),z)θ(m(s,t,z),z)\in\theta, we can assume that the generating set for θ\theta has the form

{(t1,z),,(tk,z)}\{(t_{1},z),\ldots,(t_{k},z)\}

for some kk\in\mathbb{N} and tiF𝒰(x,z)t_{i}\in F_{\mathcal{U}}(x,z) for iki\leq k. Using Lemma 2.3, we can further assume the generating set for θ\theta has the form

{(s1,z),,(sk,z),(v1,z),,(vk,z)}\{(s_{1},z),\ldots,(s_{k},z),(v_{1},z),\ldots,(v_{k},z)\}

for siR𝒰s_{i}\in R_{\mathcal{U}} and viM𝒰v_{i}\in M_{\mathcal{U}} for iki\leq k.

Let II be the ideal of 𝐑𝒰\mathbf{R}_{\mathcal{U}} generated by s1,,sks_{1},\ldots,s_{k}. Let NN be the 𝐑𝒰\mathbf{R}_{\mathcal{U}}-submodule of 𝐌\mathbf{M} generated by v1,,vkv_{1},\ldots,v_{k} and IM𝒰IM_{\mathcal{U}}. Let ρ\rho be the congruence of the 𝐑𝒰\mathbf{R}_{\mathcal{U}}-module F𝒰(x,z),+,R𝒰\langle F_{\mathcal{U}}(x,z),+,R_{\mathcal{U}}\rangle induced by the submodule I+NI+N. We want to show that θ=ρ\theta=\rho, so that z/θ=I+Nz/\theta=I+N, proving that I=z/θR𝒰I=z/\theta\cap R_{\mathcal{U}} and N=z/θM𝒰N=z/\theta\cap M_{\mathcal{U}}.

To see that ρθ\rho\subseteq\theta, notice that z/ρ=I+Nz/θz/\rho=I+N\subseteq z/\theta. Since 𝐅𝒰(x,z)\mathbf{F}_{\mathcal{U}}(x,z) is an abelian Mal’cev algebra, that implies ρθ\rho\subseteq\theta by [2, Corollary 7.7].

The last thing to prove is that θρ\theta\subseteq\rho. Since ρ\rho contains a generating set for θ\theta, it suffices to show that ρ\rho is a fully invariant congruence of 𝐅𝒰(x,z)\mathbf{F}_{\mathcal{U}}(x,z). For this, we use the characterization of 4.

By construction, 4a-4d hold for ρ\rho.

To see that property (e) holds, suppose that f(z)Nf(z)\in N. Then

f(z)\displaystyle f(z) =i=1piui(z)+j=1kqjvj(z)\displaystyle=\sum_{i=1}^{\ell}p_{i}\cdot u_{i}(z)+\sum_{j=1}^{k}q_{j}\cdot v_{j}(z)

for some piIp_{i}\in I and qjR𝒰q_{j}\in R_{\mathcal{U}}. We have

f(x)f(z)x\displaystyle f(x)-f(z)-x =i=1pi(ui(x),x)+j=1kqj(vj(x),x)(k+1)xf(z)x\displaystyle=\sum_{i=1}^{\ell}p_{i}(u_{i}(x),x)+\sum_{j=1}^{k}q_{j}(v_{j}(x),x)-(k+\ell-1)x-f(z)-x
=i=1(piui(x)+(1pi)x)+j=1k(qjvj(x)+(1qj)x)(k+)xf(z)\displaystyle=\sum_{i=1}^{\ell}\bigl{(}p_{i}u_{i}(x)+(1-p_{i})x\bigl{)}+\sum_{j=1}^{k}\bigl{(}q_{j}v_{j}(x)+(1-q_{j})x\bigl{)}-(k+\ell)x-f(z)
=i=1pi(ui(x)x)+j=1kqj(vj(x)x)i=1piui(z)j=1kqjvj(z)\displaystyle=\sum_{i=1}^{\ell}p_{i}(u_{i}(x)-x)+\sum_{j=1}^{k}q_{j}(v_{j}(x)-x)-\sum_{i=1}^{\ell}p_{i}u_{i}(z)-\sum_{j=1}^{k}q_{j}v_{j}(z)
=i=1piI(ui(x)ui(z)x)R𝒰+j=1kqjR𝒰(vj(x)vj(z)x)I.\displaystyle=\sum_{i=1}^{\ell}\underbrace{p_{i}}_{\in I}\underbrace{(u_{i}(x)-u_{i}(z)-x)}_{\in R_{\mathcal{U}}}+\sum_{j=1}^{k}\underbrace{q_{j}}_{\in R_{\mathcal{U}}}\underbrace{(v_{j}(x)-v_{j}(z)-x)}_{\in I}.

Hence f(x)f(z)xIf(x)-f(z)-x\in I. This proves condition (e).

Properties 4a-4e hold. So ρ\rho is a fully invariant congruence and θρ\theta\subseteq\rho. This completes the proof that ρ=θ\rho=\theta. That means that z/θR𝒰=Iz/\theta\cap R_{\mathcal{U}}=I, a finitely generated 𝐑𝒰\mathbf{R}_{\mathcal{U}} ideal, and z/θM𝒰=Nz/\theta\cap M_{\mathcal{U}}=N, a finitely generated 𝐑𝒰\mathbf{R}_{\mathcal{U}}-submodule of 𝐌\mathbf{M}. We have proven the forward direction of 5.

Now we show the converse. Assume that θ\theta is a fully invariant congruence. For oo\in\mathbb{N}, let s1,,soR𝒰s_{1},\ldots,s_{o}\in R_{\mathcal{U}} generate the ideal I=z/θR𝒰I=z/\theta\cap R_{\mathcal{U}} of 𝐑𝒰\mathbf{R}_{\mathcal{U}}. For kk\in\mathbb{N}, let v1,,vkM𝒰v_{1},\ldots,v_{k}\in M_{\mathcal{U}} such that v1+IM𝒰,,vk+IM𝒰v_{1}+IM_{\mathcal{U}},\ldots,v_{k}+IM_{\mathcal{U}} generate the R𝒰R_{\mathcal{U}}-module N/IM𝒰N/IM_{\mathcal{U}} for N=z/θM𝒰N=z/\theta\cap M_{\mathcal{U}}.

Let ρ\rho be the fully invariant congruence of 𝐅𝒰(x,z)\mathbf{F}_{\mathcal{U}}(x,z) generated by

{(s1,z),,(so,z),(v1,z),,(vk,z)}.\{(s_{1},z),\ldots,(s_{o},z),(v_{1},z),\ldots,(v_{k},z)\}.

The claim is that z/ρ=I+N=z/θz/\rho=I+N=z/\theta. If this is true, then ρ=θ\rho=\theta by [2, Corollary 7.7] since 𝐅𝒰(x,z)\mathbf{F}_{\mathcal{U}}(x,z) is abelian.

Since the generators of ρ\rho are in θ\theta, we have ρθ\rho\subseteq\theta and hence z/ρI+N.z/\rho\subseteq I+N.

Then what is left to show is that I+Nz/ρI+N\subseteq z/\rho. First we show Iz/ρI\subseteq z/\rho. Let tIt\in I. We want tρzt\ \rho\ z. We have for some rr\in\mathbb{N} some i1,,ir{1,,o}i_{1},\ldots,i_{r}\in\{1,\ldots,o\} and p1,,pr,q1,,qrRp_{1},\ldots,p_{r},q_{1},\ldots,q_{r}\in R such that

t=j=1rpjsijqj.\displaystyle t=\sum_{j=1}^{r}p_{j}s_{i_{j}}q_{j}.

We have

pjsijqj\displaystyle p_{j}s_{i_{j}}q_{j}\ ρpj(z,z) because ρ is fully invariant\displaystyle\ \rho\ p_{j}(z,z)\text{ because }\rho\text{ is fully invariant}
=z,\displaystyle=z,

for all jrj\leq r. Since addition is defined using mm and ρ\rho is a congruence, we have that tρzt\ \rho\ z. So we have that Iz/ρI\subseteq z/\rho.

Now we show Nz/ρN\subseteq z/\rho. Suppose that wNw\in N. Then it must be that

w(z)\displaystyle w(z) =i=1piui+j=1kqjvj\displaystyle=\sum_{i=1}^{\ell}p_{i}\cdot u_{i}+\sum_{j=1}^{k}q_{j}\cdot v_{j}
=i=1pi(ui,z)+j=1kqj(vj(z),z)\displaystyle=\sum_{i=1}^{\ell}p_{i}(u_{i},z)+\sum_{j=1}^{k}q_{j}(v_{j}(z),z)

for some piIp_{i}\in I and qjR𝒰.q_{j}\in R_{\mathcal{U}}. We have

qj(vj(z),z)\displaystyle q_{j}(v_{j}(z),z) ρqj(z,z) because ρ is a congruence\displaystyle\ \rho\ q_{j}(z,z)\text{ because }\rho\text{ is a congruence }
=z\displaystyle=z

for all jkj\leq k. We also have that

pi(ui(z),z)ρz\displaystyle p_{i}(u_{i}(z),z)\ \rho\ z\text{ }

since piIp_{i}\in I and piρzp_{i}\ \rho\ z and ρ\rho is fully invariant. Since addition is defined using mm and ρ\rho is a congruence, we have that wρzw\ \rho\ z. So we have that Nz/ρN\subseteq z/\rho.

We have shown that Iz/ρI\subseteq z/\rho and Nz/ρN\subseteq z/\rho to prove that I+Nz/ρI+N\subseteq z/\rho. Together with z/ρI+Nz/\rho\subseteq I+N we have that I+N=z/ρI+N=z/\rho. Since 𝐅𝒰(x,z)\mathbf{F}_{\mathcal{U}}(x,z) is an abelian Mal’cev variety, this proves that θ=ρ\theta=\rho. So θ\theta is finitely generated as a fully invariant congruence and we are done with the proof. ∎

4 Equational Theories and Fully Invariant Congruences

In this section, we establish the relationship between an abelian Mal’cev variety’s equational theory and a fully invariant congruence of 𝐅𝒰(x,z)\mathbf{F}_{\mathcal{U}}(x,z). We first give some needed definitions and background. Then, we state a well-known result relating the equational theories of arbitrary varieties and the fully invariant congruences of their term algebras.

For a given variety 𝒱\mathcal{V}, we use Eq(𝒱)\mathrm{Eq}(\mathcal{V}) to denote the equational theory of 𝒱\mathcal{V}. Formally,

Eq(𝒱):={st|𝒱st}.\mathrm{Eq}(\mathcal{V}):=\{s\approx t\ |\ \mathcal{V}\models s\approx t\}.

For XX a set of variables and σ\sigma a type, let 𝐓(X)\mathbf{T}(X) denote the term algebra of type σ\sigma over XX. Let Σ\Sigma be a set of identities over a type σ\sigma. For σ\sigma terms s,ts,t we write Σst\Sigma\models s\approx t if whenever an algebra 𝐀\mathbf{A} satisfies every equation in Σ\Sigma, then 𝐀st\mathbf{A}\models s\approx t.

Let 𝐀\mathbf{A} be an algebra and SA2.S\subseteq A^{2}. Then let Cgfi(S)\mathrm{Cg}_{\mathrm{fi}}(S) denote the smallest fully invariant congruence of 𝐀\mathbf{A} containing SS. We call this the fully invariant congruence generated by SS. Let 𝐂𝐨𝐧fi(𝐀)\mathbf{Con}_{\mathrm{fi}}(\mathbf{A}) denote the lattice of fully invariant congruences of 𝐀\mathbf{A}.

Theorem 9.

[1, Chapter II, Theorem 14.12] If Σ\Sigma is a set of identities over XX and pqp\approx q an identity over XX, then

Σpq(p,q)Cgfi({(s,t)|stΣ}).\displaystyle\Sigma\models p\approx q\iff(p,q)\in\mathrm{Cg}_{\mathrm{fi}}\bigl{(}\{(s,t)\ |\ s\approx t\in\Sigma\}\bigl{)}.

For two varieties 𝒱,𝒲\mathcal{V},\mathcal{W}, we write 𝒱𝒲\mathcal{V}\leq\mathcal{W} to mean 𝒱\mathcal{V} is a subvariety of 𝒲\mathcal{W}.

We now relativize Theorem 9 to varieties 𝒲\mathcal{W} contained in 𝒰\mathcal{U}. Instead of looking at the fully invariant congruences of a term algebra, we will be looking at the fully invariant congruences of 𝐅𝒰(x1,x2,)\mathbf{F}_{\mathcal{U}}(x_{1},x_{2},\ldots) and then 𝐅𝒰(x,z)\mathbf{F}_{\mathcal{U}}(x,z). The latter one will be important since we have already related the ring ideals and submodules of 𝐑𝒰\mathbf{R}_{\mathcal{U}} and 𝐌𝒰\mathbf{M}_{\mathcal{U}} to fully invariant congruences of 𝐅𝒰(x,z)\mathbf{F}_{\mathcal{U}}(x,z) in Lemma 8. This brings us one step closer to proving Theorem 3.

Corollary 10.

Let XX be a countable set of variables and let 𝒰\mathcal{U} be a variety. Then

τ:{Eq(𝒲)|𝒲𝒰}𝐂𝐨𝐧fi(𝐅𝒰(X)),Eq(𝒲){(s𝒰,t𝒰)|𝒲st},\tau\colon\{\mathrm{Eq}(\mathcal{W})\ |\ \mathcal{W}\subseteq\mathcal{U}\}\to\mathbf{Con}_{\mathrm{fi}}(\mathbf{F}_{\mathcal{U}}(X)),\;\mathrm{Eq}(\mathcal{W})\mapsto\{(s^{\mathcal{U}},t^{\mathcal{U}})\ |\ \mathcal{W}\models s\approx t\},

is a lattice isomorphism. Also, this isomorphism restricts to a bijection between the equational theories that are finitely based relative to 𝒰\mathcal{U} and the finitely generated fully invariant congruences of 𝐅𝒰(X)\mathbf{F}_{\mathcal{U}}(X).

Proof.

Let ψ:𝐓(X)𝐅𝒰(X),ss𝒰\psi\colon\mathbf{T}(X)\to\mathbf{F}_{\mathcal{U}}(X),\ s\mapsto s^{\mathcal{U}}. By the Isomorphism Theorem, we know that 𝐓(X)/kerψ𝐅𝒰(X)\mathbf{T}(X)/\ker\psi\cong\mathbf{F}_{\mathcal{U}}(X). Also, there exists a lattice isomorphism between the fully invariant congruences of 𝐓(X)\mathbf{T}(X) that contain kerψ\ker\psi and the fully invariant congruences of 𝐅𝒰(X)\mathbf{F}_{\mathcal{U}}(X).

By Theorem 9, there exists a lattice isomorphism between the fully invariant congruences of 𝐓(X)\mathbf{T}(X) containing kerψ\ker\psi and {Eq(𝒲)|𝒲𝒰}.\{\mathrm{Eq}(\mathcal{W})\ |\ \mathcal{W}\leq\mathcal{U}\}. This is because the equational theory of 𝒲\mathcal{W} contains the equational theory of 𝒰\mathcal{U}.

Hence, there exists a lattice isomorphism between the fully invariant congruences of 𝐅𝒰(X)\mathbf{F}_{\mathcal{U}}(X) and the equational theories of subvarieties of 𝒰\mathcal{U}. It is easy to check that τ\tau is this lattice isomorphism.

Now τ\tau will map compact elements to compact elements. That is, τ\tau will map finitely based (relative to 𝒰\mathcal{U}) equational theories to finitely generated fully invariant congruences of 𝐅𝒰(X)\mathbf{F}_{\mathcal{U}}(X). ∎

Lemma 11.

Let 𝒰\mathcal{U} be as in Definition 5 and 𝒲\mathcal{W} be a subvariety of 𝒰\mathcal{U}. Let ϕ:𝐅𝒰(x1,z)𝐅𝒲(x1,z),\phi\colon\mathbf{F}_{\mathcal{U}}(x_{1},z)\to\mathbf{F}_{\mathcal{W}}(x_{1},z), s𝒰s𝒲.s^{\mathcal{U}}\mapsto s^{\mathcal{W}}. Then 𝒲\mathcal{W} is finitely based over 𝒰\mathcal{U} if and only if θ:=kerϕ\theta:=\ker\phi is finitely generated as a fully invariant congruence of 𝐅𝒰(x1,z)\mathbf{F}_{\mathcal{U}}(x_{1},z).

Proof.

Let X={z,x1,x2,}X=\{z,x_{1},x_{2},\ldots\}. Let θ𝒲:=τ(Eq(𝒲)),\theta_{\mathcal{W}}:=\tau(\mathrm{Eq}(\mathcal{W})), where τ\tau is the lattice isomorphism from Corollary 10. The claim is that θ\theta is finitely generated as a fully invariant congruence of 𝐅𝒰(x,z)\mathbf{F}_{\mathcal{U}}(x,z) if and only if θ𝒲\theta_{\mathcal{W}} is finitely generated as a fully invariant congruence of 𝐅𝒰(X)\mathbf{F}_{\mathcal{U}}(X). The latter is equivalent to 𝒲\mathcal{W} being finitely based over 𝒰\mathcal{U} by Corollary 10. Thus, we will have proven the lemma if we can prove the claim.

For a term t(x1,,xk,z)t(x_{1},\ldots,x_{k},z), let t0(z):=t(z,,z)t_{0}(z):=t(z,\ldots,z) and ti(x,z):=t(z,,z,x,z,,z)t_{i}(x,z):=t(z,\ldots,z,x,z,\ldots,z) with xx in position ii for 1ik1\leq i\leq k. For terms s,ts,t we have

𝒲s(x1,,xk,z)t(x1,,xk,z)\mathcal{W}\models s(x_{1},\ldots,x_{k},z)\approx t(x_{1},\ldots,x_{k},z)

if and only if, by Lemma 4,

𝒲siti, for all 0ik.\mathcal{W}\models s_{i}\approx t_{i},\text{ for all }0\leq i\leq k.

Since θ=θ𝒲F𝒰(x1,z)2\theta=\theta_{\mathcal{W}}\cap F_{\mathcal{U}}(x_{1},z)^{2}, it follows that θ𝒲=Cgfi(θ).\theta_{\mathcal{W}}=\mathrm{Cg}_{\mathrm{fi}}(\theta).

So if θ\theta is finitely generated as a fully invariant congruence, then so is θ𝒲\theta_{\mathcal{W}}. Conversely, if θ𝒲\theta_{\mathcal{W}} is finitely generated, then θ𝒲\theta_{\mathcal{W}} is generated by a finite subset BB of θ\theta. Then BB is also a finite generating set of θ\theta.

We have shown that θ\theta is finitely generated as a fully invariant congruence if and only if θ𝒲\theta_{\mathcal{W}} is finitely generated as a fully invariant congruence. The latter is equivalent to 𝒲\mathcal{W} being finitely based over 𝒰\mathcal{U} by Corollary 10. So we have proven the lemma. ∎

5 Proof of Main Theorem

We are now ready to prove Theorem 3

Proof of Theorem  3.

Let 𝒱\mathcal{V} be an abelian Mal’cev variety. By Lemma 6, it suffices to consider the variety 𝒲𝒰\mathcal{W}\leq\mathcal{U}.

Let 𝐑𝒰:=F𝒰id(x,z),+,\mathbf{R}_{\mathcal{U}}:=\langle F_{\mathcal{U}}^{id}(x,z),+,\cdot\rangle be the ring described in Lemma 2. By Lemma 8.2, 𝐑𝒰\mathbf{R}_{\mathcal{U}} is a free ring with generators r1,,rn.r_{1},\ldots,r_{n}. Let ϕ:𝐅𝒰(x,z)𝐅𝒲(x,z)\phi\colon\mathbf{F}_{\mathcal{U}}(x,z)\to\mathbf{F}_{\mathcal{W}}(x,z) be the homomorphism defined by t𝒰(x,z)t𝒲(x,z)t^{\mathcal{U}}(x,z)\mapsto t^{\mathcal{W}}(x,z) and let θ:=kerϕ\theta:=\ker\phi. Let I:=z/θF𝒰id(x,z)I:=z/\theta\cap F_{\mathcal{U}}^{id}(x,z).

Let 𝐌𝒰:=F𝒰(z),+,R𝒰\mathbf{M}_{\mathcal{U}}:=\langle F_{\mathcal{U}}(z),+,R_{\mathcal{U}}\rangle be the 𝐑𝒰\mathbf{R}_{\mathcal{U}}-module described in Lemma 2. By Lemma 8.3, 𝐌𝒰\mathbf{M}_{\mathcal{U}} is a free 𝐑𝒰\mathbf{R}_{\mathcal{U}}-module over u1,,u.u_{1},\ldots,u_{\ell}. Let N:=z/θF𝒰(z)N:=z/\theta\cap F_{\mathcal{U}}(z).

We have that 𝒲\mathcal{W} is finitely based if and only if θ\theta is finitely generated as a fully invariant congruence of 𝐅𝒰(x,z)\mathbf{F}_{\mathcal{U}}(x,z) by Lemma 11. By Lemma 8.5, the latter is equivalent to II being a finitely generated ideal of 𝐑𝒰\mathbf{R}_{\mathcal{U}} and N/IM𝒰N/IM_{\mathcal{U}} being a finitely generated 𝐑𝒰\mathbf{R}_{\mathcal{U}}-module. This is further equivalent to 𝐑𝒲𝐑𝒰/I\mathbf{R}_{\mathcal{W}}\cong\mathbf{R}_{\mathcal{U}}/I being a finitely presented ring and 𝐌𝒲\mathbf{M}_{\mathcal{W}} being a finitely presented 𝐑𝒰\mathbf{R}_{\mathcal{U}}-module, equivalently a finitely presented module over 𝐑𝒲𝐑𝒰/I\mathbf{R}_{\mathcal{W}}\cong\mathbf{R}_{\mathcal{U}}/I. ∎

References

  • [1] Stanley Burris and H. P. Sankappanavar. A course in universal algebra, volume 78 of Graduate Texts in Mathematics. Springer-Verlag, New York-Berlin, 1981.
  • [2] Ralph Freese and Ralph McKenzie. Commutator theory for congruence modular varieties, volume 125 of London Mathematical Society Lecture Note Series. Cambridge University Press, Cambridge, 1987.
  • [3] Ralph N. McKenzie, George F. McNulty, and Walter F. Taylor. Algebras, lattices, varieties. Vol. 1. AMS Chelsea Publishing/American Mathematical Society, Providence, RI, 2018. Reprint of [MR0883644], ©1969.
  • [4] Hanna Neumann. Varieties of groups. Springer-Verlag New York, Inc., New York, 1967.