Bourbaki

Theorem 1 (Associativity theorem)

Tag 055J, permanent. Cite it as [Bourbaki, Tag 055J].

Algebra, chapter I, § 1. Laws of composition; associativity; commutativity

Let E be an associative magma whose law is denoted by ⊤\top. Let A be a totally ordered non-empty finite set, which is the union of an ordered sequence of non-empty subsets (Bi)i∈I(B_i)_{i \in I} such that the relations α∈Bi,β∈Bj,i<j\alpha \in B_i, \beta \in B_j, i < j imply α<β\alpha < \beta; let (xα)α∈A(x_\alpha)_{\alpha \in A} be an ordered sequence of elements in E with A as indexing set. Then

∏α∈Axα=∏i∈I(∏α∈Bixα) \prod_{\alpha \in A} x_\alpha = \prod_{i \in I} \left( \prod_{\alpha \in B_i} x_\alpha \right)

† The use of this terminology and the notation ∏α∈Axα\prod_{\alpha \in A} x_\alpha must be avoided if there is any risk of confusion with the product of the sets xαx_\alpha defined in the theory of sets (Set Theory, II, § 5, no. 3). However, if the xαx_\alpha are cardinals and addition (resp. multiplication) is the cardinal sum (resp. the cardinal product), the cardinal denoted by ∑α∈Axα\sum_{\alpha \in A} x_\alpha (resp. ∏α∈Axα\prod_{\alpha \in A} x_\alpha) in the above notation is the cardinal sum (resp. cardinal product) of the family (xα)α∈A(x_\alpha)_{\alpha \in A} (Set Theory, III, § 3, no. 3).

§ 1.3

We prove the theorem by induction on the cardinal nn of AA. Let pp be the cardinal of II and hh its least element; let J=I−{h}J = I - \{h\}. If n=1n = 1, as the BiB_i are non-empty, of necessity p=1p = 1 and the theorem is obvious. Otherwise, assuming the theorem holds for an indexing set with at most n−1n - 1 elements, we distinguish two cases:

(a) BhB_h has a single element β\beta. Let C=⋃i∈JBiC = \bigcup_{i \in J} B_i. The left-hand side of (3) is equal, by definition, to xβ⊤(∏α∈Cxα)x_\beta \top \left( \prod_{\alpha \in C} x_\alpha \right); the right-hand side is equal, by definition, to

xβ⊤(∏i∈J(∏α∈Bixα)); x_\beta \top \left( \prod_{i \in J} \left( \prod_{\alpha \in B_i} x_\alpha \right) \right);
the result follows from the fact that the theorem is assumed true for CC and (Bi)i∈J(B_i)_{i \in J}.

(b) Otherwise, let β\beta be the least element of AA (and hence of BhB_h); let A′=A−{β}A' = A - \{\beta\} and let Bi′=A′∩BiB'_i = A' \cap B_i for i∈Ii \in I; then Bi′=BiB'_i = B_i for i∈Ji \in J. The set A′A' has n−1n - 1 elements and the conditions of the theorem hold for A′A' and its subsets Bi′B'_i; hence by hypothesis:

∏α∈A′xα=(∏α∈Bh′xα)⊤(∏i∈J(∏α∈Bixα)). \prod_{\alpha \in A'} x_\alpha = \left( \prod_{\alpha \in B'_h} x_\alpha \right) \top \left( \prod_{i \in J} \left( \prod_{\alpha \in B_i} x_\alpha \right) \right).
Forming the composition of xβx_\beta with each side, we have on the left-hand side side, by definition, ∏α∈Axα\prod_{\alpha \in A} x_\alpha and on the right-hand side, using the associativity,
(xβ⊤(∏α∈Bh′xα))⊤(∏i∈J(∏α∈Bixα)) \left( x_\beta \top \left( \prod_{\alpha \in B'_h} x_\alpha \right) \right) \top \left( \prod_{i \in J} \left( \prod_{\alpha \in B_i} x_\alpha \right) \right)
which is equal, by Definition 3, to the right-hand side of formula (3).

For an associative law denoted by ⊤\top the composition ∏p≤i≤qxi\prod_{p \leq i \leq q} x_i of a sequence (xi)i∈[p,q](x_i)_{i \in [p, q]} is also denoted (since no confusion can arise) by

xp⊤⋯⊤xq. x_p \top \cdots \top x_q.
A particular case of Theorem 1 is the formula
x0⊤x1⊤⋯⊤xn=(x0⊤x1⊤⋯⊤xn−1)⊤xn. x_0 \top x_1 \top \cdots \top x_n = (x_0 \top x_1 \top \cdots \top x_{n-1}) \top x_n.
Consider an ordered sequence of nn terms all of whose terms are equal to the same element x∈Ex \in E. The composition of this sequence is denoted by ∏nx\prod^n x for a law denoted by ⊤\top, ∏nx\prod^n x for a law denoted by ⊥\perp. For a law written multiplicatively the composition is denoted by xnx^n and called the nn-th power of xx. For a law written additively the composition is usually denoted by nxnx. The associativity theorem applied to an ordered sequence all of whose terms are equal gives the equation

n1+n2+⋯+npTx=(n1Tx)T(n2Tx)T⋯T(npTx). \frac{n_1 + n_2 + \cdots + n_p}{T} x = \left( \frac{n_1}{T} x \right) T \left( \frac{n_2}{T} x \right) T \cdots T \left( \frac{n_p}{T} x \right).

In particular, if p=2p = 2,

m+nTx=(mTx)T(nTx) \frac{m+n}{T} x = \left( \frac{m}{T} x \right) T \left( \frac{n}{T} x \right)
and if n1=n2=⋯=np=mn_1 = n_2 = \cdots = n_p = m,
pmTx=pT(mTx). \frac{pm}{T} x = \frac{p}{T} \left( \frac{m}{T} x \right).

If XX is a subset of EE, we sometimes denote, in conformity with the above notation, by pTX\frac{p}{T} X the set X1TX2T⋯TXpX_1 T X_2 T \cdots T X_p, where

X1=X2=⋯=Xp=X; X_1 = X_2 = \cdots = X_p = X;
it is thus the set of all compositions x1Tx2T⋯Txpx_1 T x_2 T \cdots T x_p with x1∈X,x2∈X,…,xp∈Xx_1 \in X, x_2 \in X, \ldots, x_p \in X.

It is important not to confuse this set with the set of pTx\frac{p}{T} x, where xx runs through XX.

Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.

Cited by

Cites

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