Bourbaki

§ 2. Identity element; cancellable elements; invertible elements

1. Identity element

Definition 1

Under a law of composition ⊤\top on a set EE an element ee of EE is called an identity element if, for all x∈Ex \in E, e⊤x=x⊤e=xe \top x = x \top e = x.

Cited by 1

Cited by Definition 1, I § 4

There exists at most one identity element under a given law ⊤\top, for if ee and e′e' are identity elements then e=e⊤e′=e′e = e \top e' = e'. An identity element is permutable with every element: it is a central element.

Definition 2

A magma with an identity element is called a unital magma. If E,E′E, E' are unital magmas, a homomorphism of the magma EE into the magma E′E' which maps the identity element of EE to the identity element of E′E' is called a unital homomorphism (or morphism) of EE into E′E'. An associative unital magma is called a monoid.

If E,E′E, E' are monoids, a unital morphism of EE into E′E' is called a monoid homomorphism or a monoid morphism of EE into E′E'.

Example

(1) In the set N\mathbf{N} of natural numbers 0 is an identity element under addition and 1 is an identity element under multiplication. Each of these two laws gives N\mathbf{N} a commutative monoid structure (Set Theory, III, § 3, no. 3).

(2) In the set of subsets of a set EE, ∅\varnothing is an identity element under the law ∪\cup and EE under the law ∩\cap. More generally, in a lattice the least element, if it exists, is identity element under the law sup; conversely, if there exists an identity element under this law it is the least element of the set. Similarly for the greatest element and the law inf.

(3) The set N\mathbf{N} has no identity element under the law (x,y)↦xy(x, y) \mapsto x^y. Under the law (X,Y)↦X∘Y(X, Y) \mapsto X \circ Y between subsets of E×EE \times E the diagonal Δ\Delta is the identity element. Under the law (f,g)↦f∘g(f, g) \mapsto f \circ g between mappings of EE into EE the identity mapping of EE onto EE is the identity element.

(4) Let EE be a magma and RR an equivalence relation on EE compatible with the law on EE (§ 1, no. 6). If ee is an identity element of EE the canonical image of ee in E/RE/R is an identity element of the magma E/RE/R.

The identity element of a unital magma is a unital homomorphism; the composition of two unital homomorphisms is also one. For a mapping to be a unital magma isomorphism it is necessary and sufficient that it be a bijective unital homomorphism and the inverse mapping is then a unital homomorphism. Let EE and E′E' be unital magmas and e′e' the identity element of E′E'; the constant mapping of EE into E′E' mapping EE to e′e' is a unital homomorphism, called a trivial homomorphism.

The product of a family of unital magmas (resp. monoids) is a unital magma (resp. monoid).

Every quotient magma of a unital magma (resp. monoid) is a unital magma (resp. monoid).

Let EE be a unital magma and ee its identity element. A submagma AA of EE such that e∈Ae \in A is called a unital submagma of EE. Clearly ee is the identity element of the magma AA. Every intersection of unital submagmas of EE is a unital submagma of EE. If XX is a subset of EE then there exists a smallest unital submagma of EE containing XX; it is called the unital submagma of EE generated by XX; it is equal to {e}\{e\} if XX is empty. If EE is a monoid, a unital submagma of EE is called a submonoid of EE.

If FF is a magma without identity element a submagma of FF may possess an identity element. For example, if FF is associative and hh is an idempotent element of FF (I, § 1, no. 4), the set of h↑x↑hh \uparrow x \uparrow h, where xx runs through FF, is a submagma of FF with hh as identity element.

If EE is a magma with identity element ee, a submagma AA of EE such that e∉Ae \notin A may still possess an identity element.

Cites 2

Cites E III § 3. Equipotent sets. Cardinals, I § 1. Laws of composition; associativity; commutativity

Definition 3

Let EE be a unital magma. The identity element of EE is called the composition of the empty family of elements of EE.

Cites 2

Cites Theorem 1 (Associativity theorem), I § 1, Theorem 3, I § 1

If (xα)α∈∅(x_\alpha)_{\alpha \in \varnothing} is the empty family of elements of E its composition ee is also denoted by ∏α∈∅xα\prod_{\alpha \in \varnothing} x_\alpha. For example, we write

∏q≤i≤pxi=e \prod_{q \leq i \leq p} x_i = e
when p<qp < q (p,q∈Np, q \in \mathbf{N}). Similarly we write ∏0x=e\prod^0 x = e for arbitrary xx. With these definitions Theorems 1 and 3 of § 1 remain true if the hypothesis that the sets A and Bi are non-empty is suppressed.* Similarly the formulae ∏m+nx=(∏mx)∏n(∏nx)\prod^{m+n} x = \left( \prod^m x \right) \prod^n \left( \prod^n x \right) and ∏mnx=∏m(∏nx)\prod^{mn} x = \prod^m \left( \prod^n x \right) are then true for m≥0,n≥0m \geq 0, n \geq 0.

Let E be a unital magma whose law is denoted by ⊤\top and ee its identity element. The support of a family (xi)i∈I(x_i)_{i \in I} of elements of E is the set of indices i∈Ii \in I such that xi≠ex_i \neq e. Let (xi)i∈I(x_i)_{i \in I} be a family of elements of E with finite support. We shall define the composition ∏i∈Ixi\prod_{i \in I} x_i in the two following cases:

(a) the set I is totally ordered; (b) E is associative and the xix_i are pairwise permutable.

In these two cases let S be the support of the family (xi)(x_i). If J is a finite subset of I containing S, then ∏i∈Jxi=∏i∈Sxi\prod_{i \in J} x_i = \prod_{i \in S} x_i, as is seen by induction on the number of elements in J, applying Theorem 1 of § 1 in case (a) and Theorem 3 of § 1 in case (b). Let ∏i∈Ixi\prod_{i \in I} x_i denote the common value of the compositions ∏i∈Jxi\prod_{i \in J} x_i for all finite subsets of I containing S. When I is the interval [p,→][p, \to] of N\mathbf{N}, we also write ∏i=p∞xi\prod_{i = p}^\infty x_i.

With these definitions and notation, Theorems 1 and 3 of § 1 and the remarks following Theorem 3 extend to families with finite support.

The identity element of a law written additively is usually denoted by 0 and called zero or the null element (or sometimes the origin). Under a law written multiplicatively it is usually denoted by 1 and called the unit element (or unit).

2. Cancelable elements

Definition 4

Given a law of composition ⊤\top on a set E, the mapping x↦a⊤xx \mapsto a \top x (resp. x↦x⊤ax \mapsto x \top a) of E into itself is called left translation (resp. right translation) by an element a∈Ea \in E.

On passing to the opposite law, left translations become right translations and conversely.

Let γa,δa\gamma_a, \delta_a (or γ(a),δ(a)\gamma(a), \delta(a)) denote the left and right translations by a∈Ea \in E; then

γa(x)=a⊤x,δa(x)=x⊤a. \gamma_a(x) = a \top x, \qquad \delta_a(x) = x \top a.

Proposition 1

If the law* T\mathsf{T} is associative, then for all x∈Ex \in E and y∈Ey \in E

Cited by 2

Cited by Proposition 2, Proposition 3

γxTy=γx∘γy,δxTy=δy∘δx. \gamma_{x \mathsf{T} y} = \gamma_x \circ \gamma_y, \qquad \delta_{x \mathsf{T} y} = \delta_y \circ \delta_x.

For all z∈Ez \in E:

γxTy(z)=(xTy)Tz=xT(yTz)=γx(γy(z)) \gamma_{x \mathsf{T} y}(z) = (x \mathsf{T} y) \mathsf{T} z = x \mathsf{T} (y \mathsf{T} z) = \gamma_x(\gamma_y(z))
δxTy(z)=zT(xTy)=(zTx)Ty=δy(δx(z)) \delta_{x \mathsf{T} y}(z) = z \mathsf{T} (x \mathsf{T} y) = (z \mathsf{T} x) \mathsf{T} y = \delta_y(\delta_x(z))

In other words, the mapping x↦γxx \mapsto \gamma_x is a homomorphism from the magma EE to the set EEE^E of mappings of EE into itself with the law (f,g)↦f∘g(f, g) \mapsto f \circ g; the mapping x↦δxx \mapsto \delta_x is a homomorphism of EE into the set EEE^E with the opposite law. If EE is a monoid, these homomorphisms are unital.

Definition 5

An element* aa of a magma EE is called left (resp. right) cancellable (or regular) if left (resp. right) translation by aa *is injective. A left and right cancellable element is called a cancellable (or regular) element.

Cited by 1

Cited by Proposition 1 (« addition of inequalities »), VI § 1

In other words, for aa to be cancellable under the law T\mathsf{T}, it is necessary and sufficient that each of the relations aTx=aTy,xTa=yTaa \mathsf{T} x = a \mathsf{T} y, x \mathsf{T} a = y \mathsf{T} a imply x=yx = y (it is said that "a can be cancelled" from each of these equalities). If there exists an identity element ee under the law T\mathsf{T}, it is cancellable under this law: the translations γe\gamma_e and δe\delta_e are then the identity mapping of EE onto itself.

Example

(1) Every natural number is cancellable under addition; every natural number ≠0\neq 0 is cancellable under multiplication.

(2) In a lattice there can be no cancellable element under the law sup other than the identity element (least element) if it exists; similarly for inf. In particular, in the set of subsets of a set EE, ∅\varnothing is the only cancellable element under the law ∪\cup and EE the only cancellable element under the law ∩\cap.

Proposition 2

The set of cancellable (resp. left cancellable, resp. right cancellable) elements of an associative magma is a submagma.

Cites 1 · Cited by 2

Cites Proposition 1

Cited by Proposition 6, Proposition 1 (« addition of inequalities »), VI § 1

If γx\gamma_x and γy\gamma_y are injective so is γxTy=γx∘γy\gamma_{x \mathsf{T} y} = \gamma_x \circ \gamma_y (Proposition 1). Similarly for δxTy\delta_{x \mathsf{T} y}.

3. Invertible elements

Definition 6

Let* EE be a unital magma, T\mathsf{T} its law of composition, e its identity element and xx and x′x' two elements of EE. x′x' is called a left inverse (resp. right inverse, resp. inverse) of xx if x′Tx=ex' \mathsf{T} x = e *(resp. xTx′=ex \mathsf{T} x' = e, resp. x′Tx=xTx′=ex' \mathsf{T} x = x \mathsf{T} x' = e).

Cited by 1

Cited by I § 4. Groups and groups with operators

An element xx of EE is called left invertible (resp. right invertible, resp. invertible) if it has a left inverse (resp. right inverse, resp. inverse).

A monoid all of whose elements are invertible is called a group.

Symmetric and symmetrizable are sometimes used instead of inverse and invertible. When the law on E is written additively, we generally say negative instead of inverse.

Example

(1) An identity element is its own inverse.

(2) In the set of mappings of E into E an element f is left invertible (resp. right invertible) if f is a surjection (resp. injection). The left inverses (resp. right inverses) are then the retractions (resp. sections) associated with f (Set Theory, II, § 3, no. 8, Definition 11). For f to be invertible it is necessary and sufficient that f be a bijection. Its unique inverse is then the inverse bijection of f.

Let E and F be two unital magmas and f a unital homomorphism of E into F. If x' is the inverse of x in E, f(x') is the inverse of f(x) in F. Hence, if x is an invertible element of E, f(x) is an invertible element of F.

In particular, if R is an equivalence relation compatible with the law on a unital magma E, the canonical image of an invertible element of E is invertible in E/R.

Cites 1

Cites Definition 11, E II § 3

Proposition 3

Let E be a monoid and x an element of E.

Cites 1 · Cited by 2

Cites Proposition 1

Cited by Exercise 6, § 2, Theorem 1 (Jacobson), VIII § 9

(i) For x to be left (resp. right) invertible it is necessary and sufficient that the right (resp. left) translation by x be surjective.

(ii) For x to be invertible it is necessary and sufficient that it be left and right invertible. In that case x has a unique inverse, which is also its unique left (resp. right) inverse.

If x' is a left inverse of x then (Proposition 1)

δx∘δx′=δx′⊤x=δe=IdE \delta_x \circ \delta_{x'} = \delta_{x' \top x} = \delta_e = \mathrm{Id}_E

and δx\delta_x is surjective. Conversely, if δx\delta_x is surjective, there exists an element x′x' of E such that δx(x′)=e\delta_x(x') = e and x′x' is the left inverse of xx. The other assertions of (i) follow similarly.

If x′x' (resp. x′′x'') is a left (resp. right) inverse of xx, then

x′=x′⊤e=x′⊤(x⊤x′′)=(x′⊤x)⊤x′′=e⊤x′′=x′′, x' = x' \top e = x' \top (x \top x'') = (x' \top x) \top x'' = e \top x'' = x'',

whence (ii).

Remark

Let E be a monoid and x an element of E. If x is left invertible it is left cancellable; for, if x′x' is a left inverse of xx, then

γx′∘γx=γx′⊤x=γe=IdE \gamma_{x'} \circ \gamma_x = \gamma_{x' \top x} = \gamma_e = \mathrm{Id}_E

and γx\gamma_x is injective. In particular, if x is left invertible, the left and right translations by x are bijective. Conversely, suppose that γx\gamma_x is bijective; there exists x′∈Ex' \in E such that xx′=γx(x′)=exx' = \gamma_x(x') = e; then γx(x′x)=(xx′)x=x=γx(e)\gamma_x(x'x) = (xx')x = x = \gamma_x(e) and hence x′x=ex'x = e, so that x is invertible. We see similarly that if δx\delta_x is bijective, x is invertible.

Proposition 4

Let E be a monoid and x and y invertible elements of E with inverses x' and y' respectively. Then y' ⊤\top x' is the inverse of x ⊤\top y.

Cited by 7

Cited by § 2. Identity element; cancellable elements; invertible elements, Lemma 2, Remark, Corollary 1, Proposition 6, Theorem 1, Example, I § 4

This follows from the relation

(y′⊤x′)⊤(x⊤y)=y′⊤(x′⊤x)⊤y=y′⊤y=e (y' \top x') \top (x \top y) = y' \top (x' \top x) \top y = y' \top y = e
and the analogous calculations for (x ⊤\top y) ⊤\top (y' ⊤\top x').

Corollary 1

Let E be a monoid; if each of the elements xαx_\alpha of an ordered sequence (xα)α∈A(x_\alpha)_{\alpha \in A} of elements of E has an inverse xα′x'_\alpha, the composition ∏α∈Axα\prod_{\alpha \in A} x_\alpha has inverse ∏α∈A′xα′\prod_{\alpha \in A'} x'_\alpha, where A' is the totally ordered set derived from A by replacing the order on A by the opposite order.

Cites 1 · Cited by 1

Cites Proposition 4

Cited by Corollary, I § 7

This corollary follows from Proposition 4 by induction on the number of elements in A.

In particular, if x and x' are inverses, ∏nx\prod^n x and ∏nx′\prod^n x' are inverses for every integer n≥0n \geq 0.

Corollary 2

In a monoid the set of invertible elements is stable.

Proposition 5

If in a monoid x and x' are inverses and x commutes with y, so does x'.

Cited by 3

Cited by Theorem 1, Corollary 2, I § 4, Proposition 1, I § 9

From x⊤y=y⊤xx \top y = y \top x, we deduce x′⊤(x⊤y)⊤x′=x′⊤(y⊤x)⊤x′x' \top (x \top y) \top x' = x' \top (y \top x) \top x' and hence (x′⊤x)⊤(y⊤x′)=(x′⊤y)⊤(x⊤x′)(x' \top x) \top (y \top x') = (x' \top y) \top (x \top x'), that is y⊤x′=x′⊤yy \top x' = x' \top y.

Corollary 1

Let E be a monoid, X a subset of E and X' the centralizer of X. The inverse of every invertible element of X' belongs to X'.

Cited by 1

Cited by Proposition 2, I § 4

Corollary 2

In a monoid the inverse of a central invertible element is a central element.

4. Monoid of fractions of a commutative monoid

In this no., e will denote the identity element of a monoid and x* the inverse of an invertible element x of E.

Let E be a commutative monoid, S a subset of E and S' the submonoid of E generated by S.

Lemma 1

In E ×\times S' the relation RR\mathscr{R}x, yR\mathscr{R} defined by:

Cited by 1

Cited by Definition 7

"there exist a,b∈Ea, b \in E and p,q,s∈S′p, q, s \in S' such that x=(a,p),y=(b,q)x = (a, p), y = (b, q) and aqs=bpsaqs = bps"

is an equivalence relation compatible with the law on the product monoid E ×\times S'.

It is immediate that R is reflexive and symmetric. Let x=(a,p),y=(b,q)x = (a, p), y = (b, q) and z=(c,r)z = (c, r) be elements of E ×\times S' such that RR\mathscr{R}x, yR\mathscr{R} and RR\mathscr{R}y, zR\mathscr{R} hold. Then there exist two elements s and t of S' such that

aqs=bps,brt=cqt, aq s = bps, \quad br t = cqt,
whence it follows that
ar(stq)=bpsrt=cp(stq) ar(stq) = bpsrt = cp(stq)

and hence R{x,z}R \{ x, z \} holds, for stqstq belongs to S′S'. The relation RR is therefore transitive.

Further, let x=(a,p),y=(b,q),x′=(a′,p′),x = (a, p), y = (b, q), x' = (a', p'), and y′=(b′,q′)y' = (b', q') be elements of E×S′E \times S' such that R{x,y}R \{ x, y \} and R{x′,y′}R \{ x', y' \} hold. There exist ss and s′s' in S′S' such that

aqs=bps,a′q′s′=b′p′s′, aqs = bps, \qquad a'q's' = b'p's',
whence it follows that (aa′)(qq′)(ss′)=(bb′)(pp′)(ss′)(aa')(qq')(ss') = (bb')(pp')(ss') and hence R{xx′,yy′}R \{ xx', yy' \} for ss′∈S′ss' \in S'. The equivalence relation RR is therefore compatible with the law of composition on E×S′E \times S'.

The quotient magma (E×S′)/R(E \times S')/R is a commutative monoid.

Definition 7

Let EE be a commutative monoid, SS a subset of EE and S′S' the submonoid of EE generated by SS. The quotient monoid (E×S′)/R(E \times S')/R, where the equivalence relation RR is as described in Lemma 1, is denoted by ESE_S and is called the monoid of fractions† of EE associated with SS (or with denominators in SS).

Cites 1 · Cited by 1

Cites Lemma 1

Cited by Proposition 6

For a∈Ea \in E and p∈S′p \in S' the class of (a,p)(a, p) modulo RR is in general denoted by a/pa/p and called the fraction with numerator aa and denominator pp. Then by definition (a/p).(a′/∣p′)=aa′/∣pp′(a/p).(a'/|p') = aa'/|pp'. The fractions a/pa/p and a′/∣p′a'/|p' are equal if and only if there exists ss in S′S' with spa′=sp′aspa' = sp'a; if so, there exist σ\sigma and σ′\sigma' in S′S' with aσ=a′σ′a\sigma = a'\sigma' and pσ=p′σ′p\sigma = p'\sigma'. In particular, a/p=sa/spa/p = sa/sp for a∈Ea \in E and s,ps, p in S′S'. The identity element of ESE_S is the fraction e/ee/e.

Let a/e=ε(a)a/e = \varepsilon(a) for all a∈Ea \in E. The above shows that ε\varepsilon is a homomorphism of EE into ESE_S, called canonical. For all p∈S′p \in S', (p/e).(e/p)=e/e(p/e).(e/p) = e/e and hence e/pe/p is the inverse of ε(p)=p/e\varepsilon(p) = p/e; every element of ε(S′)\varepsilon(S') is therefore invertible. Then a/p=(a/e).(e/p)a/p = (a/e).(e/p), whence

a/p=ε(a)ε(p)∗ a/p = \varepsilon(a)\varepsilon(p)^*
for a∈Ea \in E and p∈S′p \in S', the monoid ESE_S is therefore generated by ε(E)∪ε(S)∗\varepsilon(E) \cup \varepsilon(S)^*.

Proposition 6

The notation is that of Definition 7 and ε\varepsilon denotes the canonical homomorphism of EE into ESE_S. (i) Let aa and bb be in EE; in order that ε(a)=ε(b)\varepsilon(a) = \varepsilon(b), it is necessary and sufficient that there exist s∈S′s \in S' with sa=sbsa = sb. (ii) For ε\varepsilon to be injective it is necessary and sufficient that every element of SS be cancellable. (iii) For ε\varepsilon to be bijective it is necessary and sufficient that every element of SS be invertible.

Cites 3 · Cited by 4

Cites Definition 7, Proposition 2, Proposition 4

Cited by Remark, Definition 8, I § 8, VIII § 11. Grothendieck Groups, Proposition 6, VIII § 11

Assertion (i) is clear and shows that ε\varepsilon is injective if and only if every element of S′S' is cancellable; but since the set of cancellable elements of EE is a submonoid of EE (no. 2, Proposition 2), it amounts to the same to say that every element of SS is cancellable.

† It is also called monoid of differences if the law on EE is written additively.

If ε\varepsilon is bijective every element of S is invertible, for ε(S)\varepsilon(S) is composed of invertible elements of ESE_S. Conversely, suppose every element of S is invertible; then every element of S′S' is invertible (no. 3, Corollary 2 to Proposition 4) and hence cancellable. Then ε\varepsilon is injective by (ii) and a/p=ε(a.p∗)a/p = \varepsilon(a.p^*) by (1), hence ε\varepsilon is surjective.

Theorem 1

Let E be a commutative monoid, S a subset of E, ESE_S the monoid of fractions associated with S and ε:E→ES\varepsilon : E \to E_S the canonical homomorphism. Further let f be a homomorphism of E into a monoid F (not necessarily commutative) such that every element of f(S)f(S) is invertible in F. There exists one and only one homomorphism fˉ\bar{f} of ESE_S into F such that f=fˉ∘εf = \bar{f} \circ \varepsilon.

Cites 2 · Cited by 6

Cites Proposition 4, Proposition 5

Cited by Exercise 16, § 2, Exercise 17, § 2, Lemma 2, Remark, Theorem 5, I § 8, Proposition 6, VIII § 11

If fˉ\bar{f} is a homomorphism of ESE_S into F such that f=fˉ∘εf = \bar{f} \circ \varepsilon, then

fˉ(a/p)=fˉ(ε(a)ε(p)∗)=fˉ(ε(a))fˉ(ε(p))∗=f(a)f(p)∗ \bar{f}(a/p) = \bar{f}(\varepsilon(a)\varepsilon(p)^*) = \bar{f}(\varepsilon(a))\bar{f}(\varepsilon(p))^* = f(a)f(p)^*
for a∈Ea \in E and p∈S′p \in S', whence the uniqueness of fˉ\bar{f}.

Let g be the mapping of E×S′E \times S' into F defined by g(a,p)=f(a)f(p)∗g(a,p) = f(a)f(p)^*. We show that g is a homomorphism of E×S′E \times S' into F. First of all,

g(e,e)=f(e)f(e)∗=e. g(e,e) = f(e)f(e)^* = e.
Let (a,p)(a,p) and (a′,p′)(a',p') be two elements of E×S′E \times S'; as a′a' and pp commute in E, f(a′)f(a') and f(p)f(p) commute in F, whence f(a′)f(p)∗=f(p)∗f(a′)f(a')f(p)^* = f(p)^*f(a') by no. 3, Proposition 5. Moreover f(pp′)∗=f(p′p)∗=(f(p′)f(p))∗=f(p)∗f(p′)∗f(pp')^* = f(p'p)^* = (f(p')f(p))^* = f(p)^*f(p')^* by no. 3, Proposition 4, whence
g(aa′,pp′)=f(aa′)f(pp′)∗=f(a)f(a′)f(p)∗f(p′)∗=f(a)f(p)∗f(a′)f(p′)∗=g(a,p)g(a′,p′). g(aa',pp') = f(aa')f(pp')^* = f(a)f(a')f(p)^*f(p')^* = f(a)f(p)^*f(a')f(p')^* = g(a,p)g(a',p').
We show that g is compatible with the equivalence relation R on E×S′E \times S': if (a,p)(a,p) and (a′,p′)(a',p') are congruent mod. R, there exists s∈S′s \in S' with spa′=sap′spa' = sap', whence f(s)f(p)f(a′)=f(s)f(a)f(p′)f(s)f(p)f(a') = f(s)f(a)f(p'). As f(s)f(s) is invertible, it follows that f(p)f(a′)=f(a)f(p′)f(p)f(a') = f(a)f(p') and then by left multiplication by f(p)∗f(p)^* and right multiplication by f(p′)∗f(p')^*
g(a′,p′)=f(a′)f(p′)∗=f(p)∗f(a)=f(a)f(p)∗=g(a,p). g(a',p') = f(a')f(p')^* = f(p)^*f(a) = f(a)f(p)^* = g(a,p).
Hence there exists a homomorphism fˉ\bar{f} of ESE_S into F such that fˉ(a/p)=g(a,p)\bar{f}(a/p) = g(a,p), whence fˉ(ε(a))=fˉ(a/e)=f(a)f(e)∗=f(a)\bar{f}(\varepsilon(a)) = \bar{f}(a/e) = f(a)f(e)^* = f(a). Hence fˉ∘ε=f\bar{f} \circ \varepsilon = f.

Corollary

Let E and F be two commutative monoids, S and T subsets of E and F respectively, f a homomorphism of E into F such that f(S)⊂Tf(S) \subset T and ε:E→ES,η:F→FT\varepsilon : E \to E_S, \eta : F \to F_T the canonical homomorphisms. There exists one and only one homomorphism g:ES→FTg : E_S \to F_T such that g∘ε=η∘fg \circ \varepsilon = \eta \circ f.

The homomorphism η∘f\eta \circ f of E into FTF_T maps every element of S to an invertible element of FTF_T.

Remark

(1) Theorem 1 can also be expressed by saying that (ES,ε)(E_S, \varepsilon) is the solution of the universal mapping problem for E, relative to monoids, monoid homomorphisms and homomorphisms of E into monoids which map the elements of S to invertible elements (Set Theory, IV, , no. 1). It follows (loc. cit.) that every other solution of this problem is isomorphic in a unique way to (ES,ε)(E_S, \varepsilon)

(2) For the existence of a solution to the above universal mapping problem it is unnecessary to assume that the monoid E is commutative, as follows from Set Theory, IV, § 3, no. 2 (cf. Exercise 17).

We mention two important special cases of monoids of fractions.

(a) Let E‾=EE\overline{E} = E_E. As the monoid E‾\overline{E} is generated by the set ε(E)∪ε(E)∗\varepsilon(E) \cup \varepsilon(E)^*, which is composed of invertible elements, every element of E‾\overline{E} is invertible (no. 3, Corollary 2 to Proposition 4). In other words, E‾\overline{E} is a commutative group. Moreover, by Theorem 1 every homomorphism ff of E into a group G can be uniquely factorized in the form f=fˉ∘εf = \bar{f} \circ \varepsilon, where fˉ:E‾→G\bar{f}: \overline{E} \to G is a homomorphism. E‾\overline{E} is called the group of fractions of E (or group of differences of E in the case of additive notation).

(b) Let Φ=EΣ\Phi = E_{\Sigma}, where Σ\Sigma consists of the cancellable elements of E. By Proposition 6, (ii), the canonical homomorphism of E into Φ\Phi is injective; it will be profitable to identify E with its image in Φ\Phi. Hence E is a submonoid of Φ\Phi, every cancellable element of E has an inverse in Φ\Phi and every element of Φ\Phi is of the form a/p=a.p∗a/p = a.p^* with a∈Ea \in E and p∈Σp \in \Sigma; then a/p=a′/p′a/p = a'/p' if and only if ap′=pa′ap' = pa'. It is easily seen that the invertible elements of Φ\Phi are the fractions a/pa/p, where aa and pp are cancellable, and p/ap/a is the inverse of a/pa/p.

Now let S be a set of cancellable elements of E and S′S' be the submonoid of E generated by S. If a/pa/p and a′/p′a'/p' are two elements of ESE_S, then a/p=a′/p′a/p = a'/p' if and only if ap′=pa′ap' = pa' (for sap′=spa′sap' = spa' implies ap′=pa′ap' = pa' for all s∈S′s \in S'). ESE_S may therefore be identified with the submonoid of Φ\Phi generated by E∪S∗E \cup S^*.

If every element of E is cancellable, then Φ=E‾\Phi = \overline{E} and E is a submonoid of the commutative group Φ\Phi. Conversely, if E is isomorphic to a submonoid of a group, every element of E is cancellable.

Cites 5

Cites Theorem 1, Exercise 17, § 2, E IV § 3. Universal mappings, Proposition 4, Proposition 6

5. Applications: I. Rational integers

Consider the commutative monoid N\mathbf{N} of natural numbers with law of composition addition; all the elements of N\mathbf{N} are cancellable under this law (Set Theory, III, § 5, no. 2, Corollary 3). The group of differences of N\mathbf{N} is denoted by Z\mathbf{Z}; its elements are called the rational integers; its law is called addition of rational integers and also denoted by +. The canonical homomorphism from N\mathbf{N} to Z\mathbf{Z} is injective and we shall identify each element of N\mathbf{N} with its image in Z\mathbf{Z}. The elements of Z\mathbf{Z} are by definition the equivalence classes determined in N×N\mathbf{N} \times \mathbf{N} by the relation between (m1,n1)(m_1, n_1) and (m2,n2)(m_2, n_2) which is written m1+n2=m2+n1m_1 + n_2 = m_2 + n_1; an element mm of N\mathbf{N} is identified with the class consisting of the elements (m+n,n)(m + n, n), where n∈Nn \in \mathbf{N}; it admits as negative in Z\mathbf{Z} the class of elements (n,m+n)(n, m + n). Every element (p,q)(p, q) of N×N\mathbf{N} \times \mathbf{N} may be written in the form (m+n,n)(m + n, n) if p≥qp \geq q or in the form (n,m+n)(n, m + n) if p≤qp \leq q; it follows that Z\mathbf{Z} is the union of N\mathbf{N} and the set of negatives of the elements of N\mathbf{N}. The identity element 0 is the only element of N\mathbf{N} whose negative belongs to N\mathbf{N}.

For every natural number mm, −m-m denotes the negative rational integer of mm and −N-\mathbf{N} denotes the set of elements −m-m for m∈Nm \in \mathbf{N}. Then

Z=N∪(−N)andN∩(−N)={0}. \mathbf{Z} = \mathbf{N} \cup (-\mathbf{N}) \quad \text{and} \quad \mathbf{N} \cap (-\mathbf{N}) = \{0\}.

for m∈Nm \in \mathbf{N}, m=−mm = -m if and only if m=0m = 0.

Let mm and nn be two natural numbers;

(a) if m≥nm \geq n, then m+(−n)=pm + (-n) = p, where pp is the element of N\mathbf{N} such that m=n+pm = n + p; (b) if m≤nm \leq n, then m+(−n)=−pm + (-n) = -p, where pp is the element of N\mathbf{N} such that m+p=nm + p = n; (c) (−m)+(−n)=−(m+n)(-m) + (-n) = -(m + n).

Properties (b) and (c) follow from no. 3, Proposition 4; as Z=N∪(−N)\mathbf{Z} = \mathbf{N} \cup (-\mathbf{N}), addition in N\mathbf{N} and properties (a), (b) and (c) describe completely addition in Z\mathbf{Z}.

More generally −x-x is used to denote the negative of an arbitrary rational integer xx; the composition x+(−y)x + (-y) is abbreviated to x−yx - y (cf. no. 8).

The order relation ≤\leq between natural numbers is characterized by the following property: m≤nm \leq n if and only if there exists an integer p∈Np \in \mathbf{N} such that m+p=nm + p = n (Set Theory, III, § 3, no. 6, Proposition 13 and § 5, no. 2, Proposition 2). The relation y−x∈Ny - x \in \mathbf{N} between rational integers xx and yy is a total order relation on Z\mathbf{Z} which extends the order relation ≤\leq on N\mathbf{N}. For, for all x∈Zx \in \mathbf{Z}, x−x=0∈Nx - x = 0 \in \mathbf{N}; if y−x∈Ny - x \in \mathbf{N} and z−y∈Nz - y \in \mathbf{N}, then

z−x=(z−y)+(y−x)∈N, z - x = (z - y) + (y - x) \in \mathbf{N},

for N\mathbf{N} is stable under addition; if y−x∈Ny - x \in \mathbf{N} and x−y∈Nx - y \in \mathbf{N}, then y−x=0y - x = 0, for 0 is the only element of N\mathbf{N} whose negative belongs to N\mathbf{N}; for arbitrary rational integers xx and yy, y−x∈Ny - x \in \mathbf{N} or x−y∈Nx - y \in \mathbf{N}, for Z=N∪(−N)\mathbf{Z} = \mathbf{N} \cup (-\mathbf{N}); finally, if xx and yy are natural numbers, then y−x∈Ny - x \in \mathbf{N} if and only if there exists p∈Np \in \mathbf{N} such that x+p=yx + p = y. This order relation is also denoted by ≤\leq.

Henceforth when Z\mathbf{Z} is considered as an ordered set, it will always be, unless otherwise mentioned, with the ordering that has just been defined, the natural numbers are identified with the integers ≥0\geq 0; they are also called positive integers; the integers ≤0\leq 0, negatives of the positive integers, are called negative integers; the integers >0> 0 (resp. <0< 0) are called strictly positive (resp. strictly negative); the set of integers >0> 0 is sometimes denoted by N∗\mathbf{N}^*.

Let xx, yy and zz be three rational integers; then x≤yx \leq y if and only if x + z \leq y + z. For x - y = (x + z) - (y + z). This property is expressed by saying that the order relation on Z\mathbf{Z} is invariant under translation.

6. Applications: II. Multiplication of rational integers

Lemma 2

Let E be a monoid and x an element of E. (i) There exists a unique homomorphism f of N\mathbf{N} into E with f(1)=xf(1) = x and f(n)=n1xf(n) = \frac{n}{1} x for all n∈Nn \in \mathbf{N}. (ii) If x is invertible, there exists a unique homomorphism g of Z\mathbf{Z} into E such that g(1)=xg(1) = x and g coincides with f on N\mathbf{N}.

Cites 5 · Cited by 1

Cites Proposition 1, I § 1, Proposition 4, Theorem 1, E III § 3. Equipotent sets. Cardinals, Corollary, E III § 3

Cited by § 2. Identity element; cancellable elements; invertible elements

Writing f(n)=n1xf(n) = \frac{n}{1} x for all n∈Nn \in \mathbf{N}, the formulae

01x=eand(m1x)⋅(n1x)=m+n1x \frac{0}{1} x = e \quad \text{and} \quad \left( \frac{m}{1} x \right) \cdot \left( \frac{n}{1} x \right) = \frac{m+n}{1} x
(no. 1) express the fact that f is a homomorphism of N\mathbf{N} into E and obviously f(1)=xf(1) = x. If f′f' is a homomorphism of N\mathbf{N} into E such that f′(1)=xf'(1) = x, then f=f′f = f', by § 1, no. 4, Proposition 1, (iv).

Suppose now that x is invertible. By no. 3, Corollary 2 to Proposition 4, f(n)=n1xf(n) = \frac{n}{1} x is invertible for every integer n≥0n \geq 0. By construction, Z\mathbf{Z} is the group of differences of N\mathbf{N} and hence (no. 4, Theorem 1) f extends uniquely to a homomorphism g of Z\mathbf{Z} into E. If g′g' is a homomorphism of Z\mathbf{Z} into E with g′(1)=xg'(1) = x, the restriction f′f' of g′g' to N\mathbf{N} is a homomorphism of N\mathbf{N} into E with f′(1)=xf'(1) = x. Hence f′=ff' = f, whence g′=gg' = g.

We shall apply Lemma 2 to the case where the monoid E is Z\mathbf{Z}; for every integer m∈Zm \in \mathbf{Z} there therefore exists an endomorphism fmf_m of Z\mathbf{Z} characterized by fm(1)=mf_m(1) = m. If m is in N\mathbf{N}, the mapping n↦mnn \mapsto mn of N\mathbf{N} into N\mathbf{N} is an endomorphism of the magma N\mathbf{N} (Set Theory, III, § 3, no. 3, Corollary to Proposition 5); hence fm(n)=mnf_m(n) = mn for all m,nm, n in N\mathbf{N}.

Multiplication on N\mathbf{N} can therefore be extended to multiplication on Z\mathbf{Z} by the formula mn=fm(n)mn = f_m(n) for m,nm, n in Z\mathbf{Z}. We shall establish the formulae:

(2)

xy=yx xy = yx
(3)
(xy)z=x(yz) (xy)z = x(yz)
(4)
x(y+z)=xy+xz x(y+z) = xy + xz
(5)
(x+y)z=xz+yz (x+y)z = xz + yz
(6)
0.x=x.0=0 0.x = x.0 = 0
(7)
1.x=x.1=x 1.x = x.1 = x
(8)
(−1).x=x.(−1)=−x (-1).x = x.(-1) = -x

for x,y,zx, y, z in Z\mathbf{Z}. (In other words, Z\mathbf{Z} is a commutative ring.) The formulae x(y+z)=xy+xzx(y+z) = xy + xz and x.0=0x.0 = 0 express the fact that fxf_x is an endomorphism of the additive monoid Z\mathbf{Z} and fx(1)=xf_x(1) = x may be written x.1=xx.1 = x. The endomorphism fx∘fyf_x \circ f_y of Z\mathbf{Z} maps 1 to xyxy and hence is equal to fxyf_{xy}, whence (3). Now fx(−y)=−fx(y)f_x(-y) = -f_x(y), that is x(−y)=−xyx(-y) = -xy; similarly, the endomorphism y↦−xyy \mapsto -xy of Z\mathbf{Z} maps 1 to −x-x, whence (−x).y=−xy(-x).y = -xy and therefore

(−x)(−y)=−(x(−y))=−(−xy)=xy. (-x)(-y) = -(x(-y)) = -(-xy) = xy.
For m,nm, n in N\mathbf{N}, mn=nmmn = nm (Set Theory, III, § 3, no. 3, Corollary to Proposition 5), whence (−m).n=n(−m)(-m).n = n(-m) and (−m)(−n)=(−n)(−m)(-m)(-n) = (-n)(-m); as Z=N∪(−N)\mathbf{Z} = \mathbf{N} \cup (-\mathbf{N}), xy=yxxy = yx for x,yx, y in Z\mathbf{Z}; and this formula means that (5) follows from (4) and completes the proof of formula (6) to (8).

7. Applications: III. Generalized powers

Let E be a monoid with identity element ee and law of composition denoted by ⊤\top. If xx is invertible in E, let gxg_x be the homomorphism of Z\mathbf{Z} into E mapping 1 to xx. Let gx(n)=n⊤xg_x(n) = \frac{n}{\top} x for all n∈Zn \in \mathbf{Z}; by Lemma 2 this notation is compatible for n∈Nn \in \mathbf{N} with the notation introduced earlier. Then

(9)m+n⊤x=(m⊤x)⊤(n⊤x)(10)0⊤x=e(11)1⊤x=x \begin{align*} \text{(9)} \quad & \frac{m+n}{\top} x = \left( \frac{m}{\top} x \right) \top \left( \frac{n}{\top} x \right) \\ \text{(10)} \quad & \frac{0}{\top} x = e \\ \text{(11)} \quad & \frac{1}{\top} x = x \end{align*}
for xx invertible in E and m,nm, n in Z\mathbf{Z}. Further, if y=m⊤xy = \frac{m}{\top} x, the mapping n↦gx(mn)n \mapsto g_x(mn) of Z\mathbf{Z} into E is a homomorphism mapping 1 to yy, whence gx(mn)=gy(n)g_x(mn) = g_y(n), that is
(12)mn⊤x=n⊤(m⊤x). \text{(12)} \quad \frac{mn}{\top} x = \frac{n}{\top} \left( \frac{m}{\top} x \right).
As −1-1 is the negative of 1 in Z\mathbf{Z}, −1⊤x\frac{-1}{\top} x is the inverse of x=1⊤xx = \frac{1}{\top} x in E. If we write n=−mn = -m in (9), it is seen that −m⊤x\frac{-m}{\top} x is the inverse of m⊤x\frac{m}{\top} x.

8. Notation

(a) As a general rule the law of a commutative monoid is written additively. It is then a convention that −x-x denotes the negative of xx. The notation x+(−y)x + (-y) is abbreviated to x−yx - y and similarly

x+y−z,x−y−z,x−y+z−t,etc. . . . x + y - z, \quad x - y - z, \quad x - y + z - t, \quad \text{etc. . . .}
represent respectively
x+y+(−z),x+(−y)+(−z),x+(−y)+z+(−t),etc. . . . x + y + (-z), \quad x + (-y) + (-z), \quad x + (-y) + z + (-t), \quad \text{etc. . . .}

For n∈Zn \in \mathbf{Z} the notation ⊤nx\overset{n}{\top} x is replaced by nxnx. Formulae (9) to (12) then become

(13)

(m+n).x=m.x+n.x (m + n).x = m.x + n.x
(14)
0.x=0 0.x = 0
(15)
1.x=x 1.x = x
(16)
m.(n.x)=(mn).x m.(n.x) = (mn).x

where mm and nn belong to N\mathbf{N} or even to Z\mathbf{Z} if xx admits a negative. Also in the latter case the relation (−1).x=−x(-1).x = -x holds. We also note the formula

(17)

n.(x+y)=n.x+n.y. n.(x + y) = n.x + n.y.

(b) Let E be a monoid written multiplicatively. For n∈Zn \in \mathbf{Z} the notation ⊤nx\overset{n}{\top} x is replaced by xnx^n. We have the relations

xm+n=xm.xnx0=1x1=x(xm)n=xmn x^{m+n} = x^m.x^n \\ x^0 = 1 \\ x^1 = x \\ (x^m)^n = x^{mn}

and also (xy)n=xnyn(xy)^n = x^ny^n if xx and yy commute.

When xx has an inverse, this is precisely x−1x^{-1}. The notation 1x\frac{1}{x} is also used instead of x−1x^{-1}. Finally, when the monoid E is commutative, xy\frac{x}{y} or x/yx/y is also used for xy−1xy^{-1}.

Exercises

See the exercises for § 2.

Transcribed from the printed edition, 1998, Springer. Not translated.