Bourbaki

1. Definition of real numbers

1. The ordered group of rational numbers

We have defined the ordering x≤yx \leq y on the set Q\mathbf{Q} of rational numbers; we have seen that this ordering makes Q\mathbf{Q} a linearly ordered set, and that it is compatible with the additive group structure of Q\mathbf{Q}, i.e. for each z∈Qz \in \mathbf{Q} the relation x≤yx \leq y is equivalent to x+z≤y+zx + z \leq y + z (that is, the ordering is invariant under translations). We recall the notation (which is used in any linearly ordered group)

x+=sup⁡(x,0),x−=sup⁡(−x,0)=(−x)+,∣x∣=sup⁡(x,−x); x^+ = \sup (x, 0), \\ x^- = \sup (-x, 0) = (-x)^+, \\ |x| = \sup (x, -x);

∣x∣|x| is called the absolute value of xx, and we have

x=x+−x−,∣x∣=x++x− x = x^+ - x^-, \qquad |x| = x^+ + x^-

and the triangle inequality

(1)

∣x+y∣≤∣x∣+∣y∣, |x + y| \leq |x| + |y|,
together with the inequality (2)
∣∣x∣−∣y∣∣≤∣x−y∣ ||x| - |y|| \leq |x - y|
which is an immediate consequence of (1); moreover (3)
∣x+−y+∣≤∣x−y∣. |x^+ - y^+| \leq |x - y|.

The relations x≥0, x=x+, x−=0, ∣x∣=xx \geq 0,\ x = x^+,\ x^- = 0,\ |x| = x (resp. x≤0, x=−x−, x+=0, ∣x∣=−xx \leq 0,\ x = -x^-,\ x^+ = 0,\ |x| = -x) are equivalent. The relation ∣x∣=0|x| = 0 is equivalent to x=0x = 0; if a≥0a \geq 0, the relation ∣x∣≤a|x| \leq a is equivalent to −a≤x≤a-a \leq x \leq a, and the relation ∣x∣≥a|x| \geq a is equivalent to "x≥ax \geq a or x≤−ax \leq -a". For all x,yx, y in Q\mathbf{Q}, we have

(4)

sup⁡(x,y)+z=sup⁡(x+z,y+z), \sup (x, y) + z = \sup (x + z, y + z),
(5)
inf⁡(x,y)=−sup⁡(−x,−y), \inf (x, y) = - \sup (-x, -y),

and, as particular cases,

(6)

sup⁡(x,y)=x+(y−x)+=x+(x−y)−, \sup (x, y) = x + (y - x)^+ = x + (x - y)^-,
(7)
inf⁡(x,y)=x−(y−x)−=x−(x−y)+. \inf (x, y) = x - (y - x)^- = x - (x - y)^+.

Finally, let Q+\mathbf{Q}_+ denote the set of rational numbers ≥0\geq 0; we have then

(8)

Q++Q+⊂Q+, \mathbf{Q}_+ + \mathbf{Q}_+ \subset \mathbf{Q}_+,
(9)
Q+∩(−Q+)={0}, \mathbf{Q}_+ \cap (- \mathbf{Q}_+) = \{ 0 \},
(10)
Q+∪(−Q+)=Q. \mathbf{Q}_+ \cup (- \mathbf{Q}_+) = \mathbf{Q}.

The relation x≤yx \leq y is equivalent to y−x∈Q+y - x \in \mathbf{Q}_+.

We shall use this ordering to define a topology on Q\mathbf{Q} compatible with its additive group structure.

2. The rational line

Consider the set F\mathfrak{F} of symmetric open intervals ]−a,+a[]-a, +a[, where aa runs through the set of rational numbers >0> 0; we shall show that F\mathfrak{F} is a fundamental system of neighbourhoods of 00 in a topology compatible with the additive group structure of Q\mathbf{Q}.

The group Q\mathbf{Q} is commutative, and axiom (GVII′)(\mathrm{GV}'_{\mathrm{II}}) is clearly satisfied; it is therefore enough to show that axiom (GVI′)(\mathrm{GV}'_{\mathrm{I}}) is also satisfied, in other words, that for each a>0a > 0 there exists b>0b > 0 such that the conditions ∣x∣<b|x| < b and ∣y∣<b|y| < b together imply ∣x+y∣<a|x + y| < a. The triangle inequality shows that we may take b=a/2b = a/2.

Definition 1

The rational line is the topological space consisting of the set Q\mathbf{Q} together with the additive group topology for which the symmetric open intervals ]−a,+a[]-a, +a[ (a>0)(a > 0) form a fundamental system of neighbourhoods of 00.

The topological group Q\mathbf{Q} thus defined is called the additive group of the rational line.

If aa is any rational number >0> 0, there is an integer n>0n > 0 such that 1/n<a1/n < a; hence the open intervals ]−1n,+1n[]-\frac{1}{n}, +\frac{1}{n}[ (n=1,2,…)(n = 1, 2, \ldots) form a fundamental system of neighbourhoods of 00 on the rational line.

We obtain a fundamental system of neighbourhoods of any point x∈Qx \in \mathbf{Q} by taking the open intervals ]x−a,x+a[]x - a, x + a[, where aa runs through the set of rational numbers >0> 0 (or the set of numbers 1/n1/n).

Definition 1 is therefore equivalent to that given in Chapter I, § 1, no. 2.

For each pair of rational numbers (a,b)(a, b) such that a<ba < b, there exists c∈Qc \in \mathbf{Q} such that a<c<ba < c < b [for example c=(a+b)/2c = (a + b)/2]; it follows that the rational line is a non-discrete Hausdorff space.

For each a>0a > 0, let UaU_a be the set of pairs (x,y)(x, y) in Q×Q\mathbf{Q} \times \mathbf{Q} such that ∣x−y∣<a|x - y| < a. As aa runs through the set of rational numbers >0> 0 (or just the set of numbers 1/n1/n), the sets UaU_a form a fundamental system of entourages of the uniformity of the additive group Q\mathbf{Q} of the rational line. Relations (2) and (3) show that ∣x∣,x+|x|, x^+ and x−x^- are uniformly continuous on Q\mathbf{Q}. It follows that the functions sup⁡(x,y)\sup(x, y) and inf⁡(x,y)\inf(x, y) are uniformly continuous on Q×Q\mathbf{Q} \times \mathbf{Q}.

3. The real line and real numbers

Definition 2

Let R\mathbf{R} denote the topological group which is the completion of the additive group Q\mathbf{Q} of the rational line. The elements of R\mathbf{R} are called real numbers; as a topological space, R\mathbf{R} is called the real line; as a topological group, R\mathbf{R} is called the additive group of the real line.

We shall always identify Q\mathbf{Q} with the dense subgroup of R\mathbf{R} to which it is canonically isomorphic. With this convention, every rational number is a real number. Every real number which is not rational is said to be irrational; we have seen in Chapter II, § 3, no. 3 that such numbers exist (we shall show this in another way in § 3, no. 3 of this chapter; see also Exercise 2 to § 2); hence (Chapter III, § 2, no. 1) the set CQ\mathbf{CQ} of irrational numbers is dense in R\mathbf{R}.

We shall show that the order structure of Q\mathbf{Q} can be extended to R\mathbf{R} in such a way that the extended ordering is still compatible with the additive group structure of R\mathbf{R}:

Proposition 1

The relation y−x∈Q‾+y - x \in \overline{\mathbf{Q}}_+ is an ordering on R\mathbf{R} which makes R\mathbf{R} into a linearly ordered set; is compatible with the additive group structure on R\mathbf{R}, and induces the ordering x≤yx \leq y on Q\mathbf{Q}.

We begin by showing that the relations y−x∈Q‾+y - x \in \overline{\mathbf{Q}}_+ and z−y∈Q‾+z - y \in \overline{\mathbf{Q}}_+ imply z−x∈Q‾+z - x \in \overline{\mathbf{Q}}_+. Indeed, the function x+yx + y is continuous on R×R\mathbf{R} \times \mathbf{R}, and therefore by (8) we have Q‾++Q‾+⊂Q‾+\overline{\mathbf{Q}}_+ + \overline{\mathbf{Q}}_+ \subset \overline{\mathbf{Q}}_+ (Chapter I, § 2, no. 1, Theorem 1). Next, we shall show that the relations y−x∈Q‾+y - x \in \overline{\mathbf{Q}}_+ and x−y∈Q‾+x - y \in \overline{\mathbf{Q}}_+ imply x=yx = y; this will establish that y−x∈Q‾+y - x \in \overline{\mathbf{Q}}_+ is an ordering on R\mathbf{R}. It is enough to show that Q‾+∩(−Q‾+)={0}\overline{\mathbf{Q}}_+ \cap (-\overline{\mathbf{Q}}_+) = \{0\}. Now the functions x→x+x \to x^+ and x→x−x \to x^- are uniformly continuous on Q\mathbf{Q} and can therefore be extended by continuity to R\mathbf{R} (Chapter II, , no. 6, Theorem 2); let ff and gg be their respective extensions. By extension we have x=f(x)−g(x)x = f(x) - g(x) for all x∈Rx \in \mathbf{R}; if x∈Q‾+x \in \overline{\mathbf{Q}}_+ then g(x)=0g(x) = 0, and since −Q‾+-\overline{\mathbf{Q}}_+ is the closure of −Q+-\mathbf{Q}_+ by the continuity of −x-x, it follows that if x∈−Q‾+x \in -\overline{\mathbf{Q}}_+ then f(x)=0f(x) = 0. Hence if x∈Q‾+∩(−Q‾+)x \in \overline{\mathbf{Q}}_+ \cap (-\overline{\mathbf{Q}}_+) we have f(x)=g(x)=0f(x) = g(x) = 0 and therefore x=0x = 0

By (10), we have Q‾+∪(−Q‾+)=R\overline{\mathbf{Q}}_+ \cup (-\overline{\mathbf{Q}}_+) = \mathbf{R}, and hence R\mathbf{R} is linearly ordered by the ordering y−x∈Q‾+y - x \in \overline{\mathbf{Q}}_+.

Furthermore, since the relations y−x∈Q‾+y - x \in \overline{\mathbf{Q}}_+ and (y+z)−(x+z)∈Q‾+(y + z) - (x + z) \in \overline{\mathbf{Q}}_+ are equivalent, the ordering y−x∈Q‾+y - x \in \overline{\mathbf{Q}}_+ is compatible with the additive group structure of R\mathbf{R}.

Finally, if xx and yy belong to Q\mathbf{Q} the relations y−x∈Q‾+y - x \in \overline{\mathbf{Q}}_+ and y−x∈Q+y - x \in \mathbf{Q}_+ are equivalent, and therefore the relation y−x∈Q‾+y - x \in \overline{\mathbf{Q}}_+ induces the relation x≤yx \leq y on Q\mathbf{Q}. This completes the proof.

The relation y−x∈Q‾+y - x \in \overline{\mathbf{Q}}_+ is again denoted by x≤yx \leq y. The set Q‾+\overline{\mathbf{Q}}_+ is the set of all x≥0x \geq 0 in R\mathbf{R} and is denoted by R+\mathbf{R}_+; it is a closed set. The set of all x>0x > 0 is denoted by R+∗\mathbf{R}_+^*; it is the complement of −R+-\mathbf{R}_+ and is therefore open in R\mathbf{R}.

4. Properties of intervals in R\mathbf{R}

Proposition 2

Every closed (resp. open) interval in R\mathbf{R} is a closed (resp. open) set in R\mathbf{R}.

The sets [a,→[=a+R+[a, \to[ = a + \mathbf{R}_+ and ]←,a]=a−R+]\leftarrow, a] = a - \mathbf{R}_+ are obtained by translation from R+\mathbf{R}_+ and −R+-\mathbf{R}_+ respectively and are therefore closed (Chapter III, § 1, no. 1); the sets ]←,a[]\leftarrow, a[ and ]a,→[,]a, \to[, which are their complements, are open; finally, the closed interval [a,b][a, b] (resp. the open interval ]a,b[]a, b[) is the intersection of [a,→[[a, \to[ and ]←,b]]\leftarrow, b] (resp. of ]a,→[]a, \to[ and ]←,b[]\leftarrow, b[) and is therefore a closed (resp. open) set.

The closed intervals [−a,+a][-a, +a] (a>0a > 0) in R\mathbf{R} are therefore neighbourhoods of 0. Let us show that they form a fundamental system of neighbourhoods of 0 as aa runs through R+∗\mathbf{R}_+^*. For this it is enough to establish the following proposition:

Proposition 3

As rr runs through the set of rational numbers >0> 0, the intervals sr=[−r,+r]s_r = [-r, +r] in R\mathbf{R} form a fundamental system of neighbourhoods of 0.

By Proposition 7 of Chapter III, § 3, no. 4 we obtain a fundamental system of neighbourhoods of 0 in R\mathbf{R} by taking the closures in R\mathbf{R} of the intervals

S_r \cap \mathbf{Q} = [-r, +r] \textit{of} \mathbf{Q}. The proof will be complete if we show that S_r is the closure of S_r \cap \mathbf{Q}. Now S_r is closed in \mathbf{R}, and we need therefore only prove that, if x is a real number such that -r < x < r, then x is in the closure of S_r \cap \mathbf{Q}. The interval ]-r, +r[ is an open set in \mathbf{R} and therefore for all sufficiently small neighbourhoods V of o in \mathbf{R} we have x + V \subset ]-r, +r[; but \mathbf{Q} being dense in \mathbf{R}, there is a rational number r' \in x + V, so that -r < r' < r and therefore r' \in S_r \cap \mathbf{Q}.

Corollary

Every point of the real line has a countable fundamental system of neighbourhoods.

Proposition 4

If (x, y) is any pair of real numbers such that x < y, there is a rational number r such that x < r < y.

Since \mathbf{Q} is dense in \mathbf{R}, it is enough to show that ]x, y[ is not empty; by translation we may assume x = 0 and y > 0. Now \mathbf{R} is a Hausdorff space and therefore, by Proposition 3, there is a rational number r > 0 such that y \notin ]-r, +r[, and this implies that 0 < r < y.

Proposition 5

Let I be any interval in \mathbf{R}. Then the topology induced on I by the topology of \mathbf{R} is generated by the open intervals of I (where I is considered as linearly ordered by the relation x \leq y).

Every open interval of I is the trace on I of an open interval of \mathbf{R}. This is clear for a bounded interval, and the unbounded interval ]a, \to[ of I is the trace of the unbounded interval ]a, \to[ of \mathbf{R}. We may therefore restrict ourselves to the case I = \mathbf{R}; but in this case the result follows from Proposition 3, since every neighbourhood of a point x \in \mathbf{R} contains an open interval ]x - a, x + a[.

Remark

If A is a dense subset of \mathbf{R}, the topology of \mathbf{R} is generated by the open intervals whose end-points belong to A. For if

]x−a,x+a[ ]x - a, x + a[

is an open interval containing x, there exist two points y, z of A such that x - a < y < x and x < z < x + a; hence ]y, z[ contains x and is contained in ]x - a, x + a[. This proof shows, moreover, that the intervals under consideration form a base (Chapter I, § 1, no. 3) of the topology of \mathbf{R}. In particular, if we take A = \mathbf{Q}, we see that the topology of \mathbf{R} has a countable base.

5. Length of an interval

Definition 3

The length of a bounded interval with end-points a and b (a \leq b) is defined to be b - a.

Every bounded interval which contains more than one point therefore has length > 0. If a≤ba \leq b, the four intervals [a,b],]a,b[,[a,b[[a, b], ]a, b[, [a, b[ and ]a,b[]a, b[ all have the same length. An interval with end-points a+ca + c and b+cb + c has the same length as an interval with end-points aa and bb; in other words, the length of an interval is invariant under translation.

If a≤c≤d≤ba \leq c \leq d \leq b we have d−c≤b−ad - c \leq b - a. Hence if a bounded interval II is contained in a bounded interval I′I', the length of II is less than or equal to the length of I′I'.

If nn mutually disjoint open intervals I1,I2,…,InI_1, I_2, \ldots, I_n are contained in the interval [a,b][a, b] (a<ba < b) it is easily seen by induction on nn that, if Ik=]ck,dk[I_k = ]c_k, d_k[, there is a permutation σ\sigma of the indices kk (1≤k≤n1 \leq k \leq n) such that dσ(k)≤cσ(k+1)d_{\sigma(k)} \leq c_{\sigma(k+1)} for 1≤k≤n−11 \leq k \leq n - 1. It follows immediately that the sum of the lengths of the intervals IkI_k is at most equal to the length of [a,b][a, b], and that equality holds if and only if cσ(1)=a,dσ(n)=bc_{\sigma(1)} = a, d_{\sigma(n)} = b and dσ(k)=cσ(k+1)d_{\sigma(k)} = c_{\sigma(k+1)} for 1≤k≤n−11 \leq k \leq n - 1.

6. Additive uniformity of R\mathbf{R}

Since the group R\mathbf{R} is linearly ordered, the functions x+,x−x^+, x^- and ∣x∣|x| are defined on R\mathbf{R} in the same way as on Q\mathbf{Q} and satisfy all the relations listed above for Q\mathbf{Q}, notably relations (1) to (7). Let aa be a real number > 0 and let UaU_a be the set of all pairs (x,y)∈R×R(x, y) \in \mathbf{R} \times \mathbf{R} such that ∣x−y∣<a|x - y| < a; as aa runs through the set of real numbers > 0 (or the set of numbers 1/n1/n), the sets UaU_a form a fundamental system of entourages of the uniformity of the additive group R\mathbf{R} of the real line (called the additive uniformity of the real line).

The functions ∣x∣|x|, x+x^+ and x−x^- are uniformly continuous on R\mathbf{R}, and the functions sup⁡(x,y)\sup(x, y) and inf⁡(x,y)\inf(x, y) are uniformly continuous on R×R\mathbf{R} \times \mathbf{R}; these functions therefore coincide with those obtained by extending by continuity the corresponding functions defined on Q\mathbf{Q} and Q×Q\mathbf{Q} \times \mathbf{Q}, respectively.

Exercises

See the exercises for § 1.

Transcribed from the printed volume. Not translated.