Bourbaki

2. Theorems

From now on, if AA is a relation, we shall write not(AA) instead of ¬A\neg A. If AA and BB are relations, we shall write “(AA) or (BB)” instead of ∨AB\vee AB, and (AA) ⇒\Rightarrow (BB) instead of ⇒AB\Rightarrow AB. Sometimes we shall leave out the brackets. In each case the reader will be able to determine without difficulty the assembly under consideration.

1. Axioms

We have already seen that the specific signs determine the terms and the relations in a theory T\mathscr{T}. To construct T\mathscr{T}, we proceed as follows :

(1) First we write down a certain number of relations in T\mathscr{T}; these are called the explicit axioms of T\mathscr{T}. The letters which appear in the explicit axioms are called the constants of T\mathscr{T}.

(2) We lay down one or more rules [^1], called the schemes of T\mathscr{T}, which must have the following properties : (a) the application of such a rule R\mathscr{R} furnishes a relation in T\mathscr{T}; (b) if TT is a term in T\mathscr{T}, if xx is a letter, and if RR is a relation in T\mathscr{T} constructed by applying the scheme R\mathscr{R}, then the relation (T∣x)R(T|x)R can also be constructed by applying R\mathscr{R}.

In all the cases we envisage, the verification of these conditions is always easy.

Every relation contructed by applying a scheme of T\mathscr{T} is called an implicit axiom of T\mathscr{T}.

Intuitively, the axioms represent either self-evident assertions or else hypotheses from which one wishes to draw consequences. The constants represent well-defined objects for which the properties expressed by the explicit axioms are supposed to be true. On the other hand, if the letter xx is not a constant, it represents a completely undetermined object; if a property of the object xx is assumed to be true by means of an axiom, then this axiom is necessarily implicit, so that the property remains true for any object TT.

2. Proofs

A demonstrative text in a theory T\mathscr{T} comprises :

(1) An auxiliary formative construction of relations and terms in T\mathscr{T}.

(2) A proof in T\mathscr{T}, that is to say a sequence of relations in T\mathscr{T} which appear in the auxiliary formative construction, such that for every relation R\mathbf{R} in the sequence at least one of the following conditions is satisfied :

(a1_1) R\mathbf{R} is an explicit axiom of T\mathscr{T}.

(a2_2) R\mathbf{R} results from the application of a scheme of T\mathscr{T} to terms or relations which appear in the auxiliary formative construction.

(b) there are two relations S\mathbf{S}, T\mathbf{T} in the sequence which precede R\mathbf{R}, such that T\mathbf{T} is S⇒R\mathbf{S} \Rightarrow \mathbf{R}.

¶ A theorem in T\mathscr{T} is a relation which appears in a proof in T\mathscr{T}.

This notion is therefore essentially dependent on the state of the theory under consideration, at the time when it is being described. A relation in a theory T\mathscr{T} becomes a theorem in T\mathscr{T} when one succeeds in inserting it into a proof in T\mathscr{T}. To say that a relation in T\mathscr{T} "is not a theorem in T\mathscr{T}" cannot have any meaning without reference to the stage of development of the theory T\mathscr{T}.

A theorem in T\mathscr{T} is also called a "*true* relation in T\mathscr{T}" (or "proposition", "lemma", "corollary", etc.). Let R\mathbf{R} be a relation in T\mathscr{T}, let x\mathbf{x} be a letter and T\mathbf{T} a term in T\mathscr{T}; if (T∣x)R(\mathbf{T}|\mathbf{x})\mathbf{R} is a theorem in T\mathscr{T}, T\mathbf{T} is said to satisfy the relation R\mathbf{R} in T\mathscr{T} (or to be a solution of R\mathbf{R}), when R\mathbf{R} is considered as a relation in x\mathbf{x}.

¶ A relation is said to be false in T\mathscr{T} if its negation is a theorem in T\mathscr{T}. A theory T\mathscr{T} is said to be contradictory when one has written a relation which is both true and false in T\mathscr{T}.

Here again, we are dealing with a notion that depends on the particular state of development of a theory. The reader should beware of the confusion (unfortunately suggested by the intuitive meaning of the word "false") which consists in believing that, once one has proved that a relation R\mathbf{R} is false in T\mathscr{T}, one has thereby established that R\mathbf{R} "is not true" in T\mathscr{T} (strictly speaking, this phrase has no precise meaning in mathematics, as we have remarked above).

¶ In what follows we shall give metamathematical criteria, called deductive criteria, which will allow us to shorten proofs. These criteria will be denoted by the letter C followed by a number.

C1 (Syllogism). Let A\mathbf{A} and B\mathbf{B} be relations in a theory T\mathscr{T}. If A\mathbf{A} and A⇒B\mathbf{A} \Rightarrow \mathbf{B} are theorems in T\mathscr{T}, then B\mathbf{B} is a theorem in T\mathscr{T}.

Let R1R_1, R2R_2, ..., RnR_n be a proof in T\mathscr{T} in which AA appears, and let S1S_1, S2S_2, ..., SpS_p be a proof in T\mathscr{T} in which A⇒BA \Rightarrow B appears. Clearly R1R_1, R2R_2, ..., RnR_n, S1S_1, S2S_2, ..., SpS_p is a proof in T\mathscr{T} in which both AA and A⇒BA \Rightarrow B appear. Hence

R1,  R2,  …,  Rn,  S1,  S2,  …,  Sp,  BR_1, \; R_2, \; \ldots, \; R_n, \; S_1, \; S_2, \; \ldots, \; S_p, \; B

is a proof in T\mathscr{T}, and therefore BB is a theorem in T\mathscr{T}.

3. Substitutions in a theory

Let T\mathscr{T} be a theory, let A1A_1, A2A_2, ..., AnA_n be its explicit axioms, let xx be a letter and TT a term of T\mathscr{T}. Let (T∣x)T(T|x)\mathscr{T} be the theory whose signs and schemes are the same as those of T\mathscr{T}, and whose explicit axioms are (T∣x)A1(T|x)A_1, (T∣x)A2(T|x)A_2, ..., (T∣x)An(T|x)A_n.

C2. Let AA be a theorem in a theory T\mathscr{T}, let TT be a term of T\mathscr{T}, and let xx be a letter. Then (T∣x)A(T|x)A is a theorem in the theory (T∣x)T(T|x)\mathscr{T}.

Let R1R_1, R2R_2, ..., RnR_n be a proof in T\mathscr{T} in which AA appears. Consider the sequence (T∣x)R1(T|x)R_1, (T∣x)R2(T|x)R_2, ..., (T∣x)Rn(T|x)R_n, which is a sequence of relations in T\mathscr{T} by reason of CF8 (§ 1, no. 4). We shall show that this sequence is a proof in the theory (T∣x)T(T|x)\mathscr{T}; this will establish the criterion. If RkR_k is an implicit axiom of T\mathscr{T}, then (T∣x)Rk(T|x)R_k is again an implicit axiom of T\mathscr{T} (no. 1) and therefore of (T/x)T(T/x)\mathscr{T}. If RkR_k is an explicit axiom of T\mathscr{T}, then (T∣x)Rk(T|x)R_k is an explicit axiom of (T∣x)T(T|x)\mathscr{T}. Finally, if RkR_k is preceded by relations RiR_i and RjR_j, where RjR_j is Ri⇒RkR_i \Rightarrow R_k, then (T∣x)Rk(T|x)R_k is preceded by (T∣x)Ri(T|x)R_i and (T∣x)Rj(T|x)R_j, and the latter is identical with (T∣x)Ri⇒(T∣x)Rk(T|x)R_i \Rightarrow (T|x)R_k (criterion CS5).

C3. Let AA be a theorem of a theory T\mathscr{T}, let TT be a term of T\mathscr{T}, and let xx be a letter which is not a constant of T\mathscr{T}. Then (T∣x)A(T|x)A is a theorem of T\mathscr{T}.

This follows immediately from C2, because xx does not feature in the explicit axioms of T\mathscr{T}. More particularly, if T\mathscr{T} contains no explicit axioms, or if the explicit axioms contain no letters, then the criterion C3 applies without restriction on the letter xx.

4. Comparison of theories

A theory T′\mathscr{T}' is said to be stronger than a theory T\mathscr{T} if all the signs of T\mathscr{T} are signs of T′\mathscr{T}', all the explicit axioms of T\mathscr{T} are theorems in T′\mathscr{T}', and the schemes of T\mathscr{T} are schemes of T′\mathscr{T}'.

C4. If a theory T′\mathscr{T}' is stronger than a theory T\mathscr{T}, then all the theorems of T\mathscr{T} are theorems of T′\mathscr{T}'.

Let R1R_1, R2R_2, ..., RnR_n be a proof in T\mathscr{T}. We shall show, step by step, that each RiR_i is a theorem in T′\mathscr{T}'. Suppose that this is true for the relations preceding RkR_k. If RkR_k is an axiom of T\mathscr{T}, it is a theorem in T′\mathscr{T}' by hypothesis. If RkR_k is preceded by relations RiR_i and Ri⇒RkR_i \Rightarrow R_k, we know already that RiR_i and Ri⇒RkR_i \Rightarrow R_k are theorems in T′\mathscr{T}', and therefore RkR_k is a theorem in T′\mathscr{T}' by virtue of C1. Hence, in every case, RkR_k is a theorem in T′\mathscr{T}', and the proof is complete.

If each of two theories T\mathscr{T} and T′\mathscr{T}' is stronger than the other, T\mathscr{T} and T′\mathscr{T}' are said to be equivalent. Then every theorem of T\mathscr{T} is a theorem of T′\mathscr{T}', and vice versa.

C5. Let T\mathscr{T} be a theory, let A1A_1, A2A_2, ..., AnA_n be its explicit axioms, a1a_1, a2a_2, ..., aha_h its constants, and let T1T_1, T2T_2, ..., ThT_h be terms in T\mathscr{T}. Suppose that

(T1∣a1) (T2∣a2) … (Th∣ah)Ai(fori=1,2,…,n)(T_1|a_1) \, (T_2|a_2) \, \ldots \, (T_h|a_h)A_i \qquad (\text{for} \quad i = 1, 2, \ldots, n)

are theorems in a theory T′\mathscr{T}', that the signs of T\mathscr{T} are signs of T′\mathscr{T}', and that the schemes of T\mathscr{T} are schemes of T′\mathscr{T}'. Then, if AA is a theorem in T\mathscr{T},

(T1∣a1) … (Th∣ah)A(T_1|a_1) \, \ldots \, (T_h|a_h)A

is a theorem in T′\mathscr{T}'.

For T′\mathscr{T}' is stronger than the theory (T1∣a1)…(Tn∣an)T(T_1|a_1) \ldots (T_n|a_n) \mathscr{T}, and we can apply C2 and C4.

When we use this procedure to deduce a theorem in T′\mathscr{T}' from a theorem in T\mathscr{T}, we say that we are applying in T′\mathscr{T}' the results of T\mathscr{T}. Intuitively, the axioms of T\mathscr{T} express properties of a1,a2,...,aha_1, a_2, ..., a_h, and AA expresses a property which is a consequence of these axioms. If the objects T1,T2,...,ThT_1, T_2, ..., T_h in T′\mathscr{T}' have the properties expressed by the axioms of T\mathscr{T}, then they also have the property AA.

\*For example, in the theory of groups T\mathscr{T}, the explicit axioms contain two constants G and μ\mu (the group and the law of composition). In the theory of sets T′\mathscr{T}', we define two terms : the real line and addition of real numbers. If we substitute these terms for G and μ\mu respectively in the explicit axioms of T\mathscr{T}, we obtain theorems in T′\mathscr{T}'. Moreover, the schemes and signs of T\mathscr{T} and T′\mathscr{T}' are the same. We may therefore "apply the results of group theory to the additive group of real numbers". We say that we have constructed a model for group theory in the theory of sets. (Note that since the theory of groups is stronger than the theory of sets, we can also apply the results of the theory of sets to the theory of groups.)\*

Remark 03G7

Under the hypotheses of C5, if the theory T\mathscr{T} turns out to be contradictory, the same will be true of T′\mathscr{T}'. For if AA and "not AA" are theorems in T\mathscr{T}, then (T1∣a1)…(Th∣ah)A(T_1|a_1) \ldots (T_h|a_h)A and not(T1∣a1)…(Th∣ah)A\mathrm{not}(T_1|a_1) \ldots (T_h|a_h)A are theorems in T′\mathscr{T}'. \* For example, if the theory of groups were contradictory, the theory of sets would also be contradictory. \

Exercises

See the exercises for § 2.

[^1]: For the sake of brevity, these rules are expressed by using the symbols mentioned in § 1, no. 1 (and especially bold-faced italic letters); but it would be easy to avoid the use of these symbols completely in the formulation of the rules (see § 3, no. 1, note (*) on p. 28).

Transcribed from the printed edition, 2004, Springer. Not translated.