3

Consider the class of algebras of type $<2,2,0,0>$ which are the $(+,*,0,1)$ reducts of commutative rings. What is a finite basis for the identities of that class of algebras? I conjecture that

  1. $x+0=x$
  2. $x+y=y+x$
  3. $(x+y)+z=x+(y+z)$
  4. $x*1=x$
  5. $x*y=y*x$
  6. $(x*y)*z=x*(y*z)$
  7. $x*(y+z)=(x*y)+(x*z)$
  8. $x*0=0$

is sufficient. Also, if this list is indeed sufficient, is the last axiom $x*0=0$ redundant? I know most ring theory books prove it from the remaining axioms, but that is with the help of the subtraction operator. I think, in this case, it is not redundant.

user107952
  • 20,508
  • 4
    Axiom 8 is not a consequence of Axioms 1-7: Let $X$ be a nonempty set and let $A={\mathcal P}(X)$ be the power set of $X$. Equip $A$ with the operations $0=1:=\emptyset$ and $x+y=x\cdot y:=x\cup y$. Then $\langle A; +,\cdot,0,1\rangle$ satisfies Axioms 1-7 and not 8. – Keith Kearnes Apr 08 '21 at 18:19
  • 2
    I don't think the class of $(+,\ast,0,1)$-reducts of commutative rings is an equational class. Do you mean the variety generated by this class? – Eran Apr 08 '21 at 21:11
  • @Eran Yes, the variety generated by the class. – user107952 Apr 08 '21 at 21:46

1 Answers1

5

Axioms 1-8 are a basis for this variety. This fact follows from 2 observations:

(1) Axioms 1-8 are sufficient to put any polynomial into normal form.
(2) If two polynomials $p, q$ written in normal form agree on $\mathbb Q$, then they are identical.

Let me include a few more details. Let $\mathcal A$ be the variety axiomatized by Axioms 1-8. Let $\mathcal C$ be the variety generated by the $(+,\cdot, 0, 1)$-reducts of commutative rings. Since commutative rings satisfy Axioms 1-8, we have ${\mathcal C}\subseteq {\mathcal A}$. The goal is to show equality. For this, it suffices to show that the $\omega$-generated free algebra $\mathbf F_{\mathcal A}(x_0,x_1,\ldots)$ lies in the subvariety $\mathcal C$. There is a natural surjective homomorphism $h\colon {\mathbf F}_{\mathcal A}(x_0,x_1,\ldots)\to {\mathbf F}_{\mathcal C}(x_0,x_1,\ldots)\colon x_i\mapsto x_i$, and I want to argue that $h$ is injective, and this will complete the argument.

Stage 1. Axioms 1-8 suffice to write each element $p\in {\mathbf F}_{\mathcal A}(x_0,x_1,\ldots)$ in normal form. Here, by the normal form of $p$, I mean ``$p = 0$'' or

  • p is a left-associated sum of monomials.
  • each monomial is either 1 or is a left-associated product of $x_i$'s.
  • within a given monomial that is a product of variables, the variables are ordered by increasing subscript.
  • within the sum of monomials, the monomials are ordered by increasing degree, and monomials of the same degree are ordered lexicographically by subscript.

    Stage 2. To obtain a contradiction, assume that $h(p)=h(q)$ where the normal forms of $p$ and $q$ are different. Choose such a pair $(p,q)$ of this type such that the sum of the lengths of the normal forms, written $|p|+|q|$, is least. Argue that neither $p$ nor $q$ can be $0$, nor can they have a constant monomial $1$, nor can they share a monomial. (If they shared any monomial, then cancelling it from both yields a pair $(p',q')$ with the same properties and shorter total length.)

    Stage 3. Now $p-q$ may be considered to be a polynomial with integer coefficients. It cannot be the zero polynomial, since $p$ and $q$ share no monomial, hence there is no cancellation when we form $p-q$. Thus, $p-q$ is a nonzero integer polynomial which vanishes on every commutative ring. To obtain a contradiction, argue by induction on the number of variables appearing that any polynomial that vanishes on $\mathbb Q$ is the zero polynomial.

  • Keith Kearnes
    • 13,798