Skip to content

§4 The Completeness Axiom

This section gives the axiom that guarantees R\mathbb{R} has no gaps. Almost every significant result in this book depends on it, and most of them would be false if our world of numbers were only Q\mathbb{Q}.

4.1 Definition.

Let SS be a nonempty subset of R\mathbb{R}.

  1. If SS contains a largest element s0s_0 (that is, s0∈Ss_0 \in S and s≤s0s \le s_0 for all s∈Ss \in S), then s0s_0 is the maximum of SS, written s0=max⁡Ss_0 = \max S.
  2. If SS contains a smallest element, it is the minimum of SS, written min⁡S\min S.
Example 1.
  1. Every finite nonempty subset of R\mathbb{R} has a maximum and a minimum. For instance, max⁡{0,π,−7,e,3,43}=π\max\{0, \pi, -7, e, 3, \frac43\} = \pi and min⁡{0,π,−7,e,3,43}=−7\min\{0, \pi, -7, e, 3, \frac43\} = -7. Also max⁡{n∈Z:−4<n≤100}=100\max\{n \in \mathbb{Z} : -4 < n \le 100\} = 100 and min⁡{n∈Z:−4<n≤100}=−3\min\{n \in \mathbb{Z} : -4 < n \le 100\} = -3.
  2. For real a<ba < b we use the interval notation [a,b]={x:a≤x≤b},(a,b)={x:a<x<b},[a,b)={x:a≤x<b},(a,b]={x:a<x≤b}.[a,b] = \{x : a \le x \le b\},\quad (a,b) = \{x : a < x < b\},\quad [a,b) = \{x : a \le x < b\},\quad (a,b] = \{x : a < x \le b\}. [a,b][a,b] is a closed interval, (a,b)(a,b) an open interval, and the others are half-open. Here max⁡[a,b]=b\max[a,b] = b and min⁡[a,b]=a\min[a,b] = a, while (a,b)(a,b) has neither a maximum nor a minimum, and [a,b)[a,b) has minimum aa but no maximum.
  3. Z\mathbb{Z} and Q\mathbb{Q} have no maximum or minimum. N\mathbb{N} has no maximum, but min⁡N=1\min \mathbb{N} = 1.
  4. {r∈Q:0≤r≤2}\{r \in \mathbb{Q} : 0 \le r \le \sqrt2\} has minimum 00 but no maximum: 2\sqrt2 is not in the set, yet the set contains rationals arbitrarily close to 2\sqrt2.
  5. {n(−1)n:n∈N}={1,2,13,4,15,6,17,… }\{n^{(-1)^n} : n \in \mathbb{N}\} = \{1, 2, \frac13, 4, \frac15, 6, \frac17, \dots\} has no maximum and no minimum.
4.2 Definition.

Let SS be a nonempty subset of R\mathbb{R}.

  1. If M∈RM \in \mathbb{R} satisfies s≤Ms \le M for all s∈Ss \in S, then MM is an upper bound of SS, and SS is bounded above.
  2. If m∈Rm \in \mathbb{R} satisfies m≤sm \le s for all s∈Ss \in S, then mm is a lower bound of SS, and SS is bounded below.
  3. SS is bounded if it is bounded above and below, i.e. S⊆[m,M]S \subseteq [m, M] for some real m,Mm, M.
Example 2.
  1. A maximum is always an upper bound, and a minimum is always a lower bound.
  2. For a<ba < b, the number bb is an upper bound for [a,b][a,b], (a,b)(a,b), [a,b)[a,b) and (a,b](a,b]. So is every number larger than bb, but bb is the least upper bound.
  3. None of Z\mathbb{Z}, Q\mathbb{Q}, N\mathbb{N} is bounded above. N\mathbb{N} is bounded below, by 11 or anything smaller, and 11 is its greatest lower bound.
  4. Every nonpositive number is a lower bound for {r∈Q:0≤r≤2}\{r \in \mathbb{Q} : 0 \le r \le \sqrt2\}, and 00 is the greatest. The least upper bound is 2\sqrt2.
  5. {n(−1)n:n∈N}\{n^{(-1)^n} : n \in \mathbb{N}\} is not bounded above. Its greatest lower bound is 00.
4.3 Definition.

Let SS be a nonempty subset of R\mathbb{R}.

  1. If SS is bounded above and has a least upper bound, we call it the supremum of SS, written sup⁡S\sup S.
  2. If SS is bounded below and has a greatest lower bound, we call it the infimum of SS, written inf⁡S\inf S.

Unlike max⁡S\max S and min⁡S\min S, the numbers sup⁡S\sup S and inf⁡S\inf S need not belong to SS. We use the Latin names because of the standard notations lim sup⁡\limsup and lim inf⁡\liminf that come later; nobody writes “lim lub”.

The following characterization is used constantly. If SS is bounded above, then M=sup⁡SM = \sup S if and only if

(i) s≤M for all s∈S,and(ii) whenever M1<M, there is s1∈S with s1>M1.\text{(i) } s \le M \text{ for all } s \in S, \qquad \text{and} \qquad \text{(ii) whenever } M_1 < M, \text{ there is } s_1 \in S \text{ with } s_1 > M_1.

Intuitionsup as a ceiling the set touches

Taking M1=M−εM_1 = M - \varepsilon in (ii) gives the form you will use most often:

M=sup⁡S  ⟺  (i) s≤M for all s∈S, and(ii′) for every ε>0 there is s∈S with s>M−ε.\begin{gathered} M = \sup S \iff \text{(i) } s \le M \text{ for all } s \in S, \text{ and} \\ \text{(ii}'\text{) for every } \varepsilon > 0 \text{ there is } s \in S \text{ with } s > M - \varepsilon. \end{gathered}

Picture MM as a ceiling. Condition (i) says no element pokes through it. Condition (ii′') says the set comes within every distance ε\varepsilon of the ceiling, so the ceiling cannot be lowered at all. Proving sup⁡S=M\sup S = M therefore always takes two separate arguments: one that MM is an upper bound, and one that nothing smaller is.

Drag the blue line MM and the orange line mm. The widget tells you whether MM is an upper bound (and if not, which element of SS is above it) and whether it is the least one. Then compare sup⁡S\sup S with max⁡S\max S.

Keyboard: Tab to the M or m marker, then use the arrow keys to move it.

elements of Sdrag M (upper bound?)drag m (lower bound?)

M = 1.140 is an upper bound, but not the least one: try sliding it left.

m = -0.1400 is a lower bound, but not the greatest one: try sliding it right.

Example 3.
  1. If SS has a maximum, then max⁡S=sup⁡S\max S = \sup S. Similarly for minimum and infimum.
  2. If a<ba < b, then sup⁡[a,b]=sup⁡(a,b)=sup⁡[a,b)=sup⁡(a,b]=b\sup[a,b] = \sup(a,b) = \sup[a,b) = \sup(a,b] = b.
  3. inf⁡N=1\inf \mathbb{N} = 1.
  4. If A={r∈Q:0≤r≤2}A = \{r \in \mathbb{Q} : 0 \le r \le \sqrt2\}, then sup⁡A=2\sup A = \sqrt2 and inf⁡A=0\inf A = 0.
  5. inf⁡{n(−1)n:n∈N}=0\inf\{n^{(-1)^n} : n \in \mathbb{N}\} = 0.
Example 4.
  1. A={1n2:n∈N, n≥3}A = \{\frac1{n^2} : n \in \mathbb{N},\ n \ge 3\} is bounded. It has maximum 19\frac19 but no minimum; sup⁡A=19\sup A = \frac19 and inf⁡A=0\inf A = 0.
  2. B={r∈Q:r3≤7}B = \{r \in \mathbb{Q} : r^3 \le 7\} is bounded above (by 22, say). It has no maximum, because r3≠7r^3 \ne 7 for every rational rr by Theorem 2.2. However, sup⁡B=73\sup B = \sqrt[3]7. BB is not bounded below; think of the graph of y=x3y = x^3.
  3. C={m+n2:m,n∈Z}C = \{m + n\sqrt2 : m, n \in \mathbb{Z}\} is bounded neither above nor below.
  4. D={x∈R:x2<10}=(−10,10)D = \{x \in \mathbb{R} : x^2 < 10\} = (-\sqrt{10}, \sqrt{10}) is bounded, has no maximum or minimum, and inf⁡D=−10\inf D = -\sqrt{10}, sup⁡D=10\sup D = \sqrt{10}.

Common pitfall

  • sup⁡S\sup S need not belong to SS. When it does, it is max⁡S\max S; when it does not, SS has no maximum at all. Compare sup⁡B\sup B in Example 4(b), and inf⁡A\inf A in Example 4(a).
  • Strict inequalities do not survive taking a supremum. If s<Ms < M for every s∈Ss \in S, you may only conclude sup⁡S≤M\sup S \le M: every element of (0,1)(0, 1) is less than 11, and sup⁡(0,1)=1\sup(0, 1) = 1.
  • To show sup⁡S≤M\sup S \le M, show that MM is an upper bound. To show sup⁡S≥M\sup S \ge M, you need elements of SS near MM, or an argument that every upper bound is at least MM. Mixing these up is the most common error in sup proofs.

In every example above, each set that is bounded above has a least upper bound. This is no accident. Otherwise there would be a “gap” between the set and its upper bounds.

4.4 Completeness Axiom.

Every nonempty subset SS of R\mathbb{R} that is bounded above has a least upper bound. In other words, sup⁡S\sup S exists and is a real number.

The corresponding statement for Q\mathbb{Q} is false. The set A={r∈Q:0≤r and r2≤2}A = \{r \in \mathbb{Q} : 0 \le r \text{ and } r^2 \le 2\} is described entirely in terms of rationals and is bounded above by rationals such as 32\frac32, but no rational number is its least upper bound.

The version for sets bounded below comes for free.

Why this matters

Every axiom in §3 holds in Q\mathbb{Q} as well as in R\mathbb{R}. The Completeness Axiom is the first property that fails in Q\mathbb{Q}, and it is the only thing that separates the two. So whenever a later proof uses a supremum of a set that is not finite, look for the step where the existence of that supremum is essential: that is where the theorem would break down over Q\mathbb{Q}.

4.5 Corollary.

Every nonempty subset SS of R\mathbb{R} that is bounded below has a greatest lower bound inf⁡S\inf S.

Used in Exercise 4.9, Exercise 4.14, §5

Proof

Let −S={−s:s∈S}-S = \{-s : s \in S\}. If m≤sm \le s for all s∈Ss \in S, then −m≥−s-m \ge -s for all s∈Ss \in S, so −S-S is bounded above by −m-m. By the Completeness Axiom, s0=sup⁡(−S)s_0 = \sup(-S) exists. Reflecting the picture suggests inf⁡S=−s0\inf S = -s_0, and to prove it we must show

−s0≤s for all s∈S,(1)-s_0 \le s \text{ for all } s \in S, \tag{1}if t≤s for all s∈S, then t≤−s0.(2)\text{if } t \le s \text{ for all } s \in S, \text{ then } t \le -s_0. \tag{2}

(1) says −s0-s_0 is a lower bound for SS, and (2) says it is the greatest one. The proofs of (1) and (2) are Exercise 4.9.

∎

The Archimedean property and denseness of ℚ

Section titled “The Archimedean property and denseness of ℚ”

It is useful to know that

if a>0, then 1n<a for some n∈N,(∗)\text{if } a > 0, \text{ then } \tfrac1n < a \text{ for some } n \in \mathbb{N}, \tag{$*$} if b>0, then b<n for some n∈N.(∗∗)\text{if } b > 0, \text{ then } b < n \text{ for some } n \in \mathbb{N}. \tag{$**$}

These are not as obvious as they look: there are ordered fields with elements a,b>0a, b > 0 satisfying a<1na < \frac1n and n<bn < b for every nn. Such strange elements cannot exist in R\mathbb{R}, and we must expect to use completeness to prove it.

4.6 Archimedean Property.

If a>0a > 0 and b>0b > 0, then na>bna > b for some positive integer nn.

Even if aa is tiny and bb is huge, some multiple of aa exceeds bb: given enough time, you can empty a large bathtub with a small spoon. Taking b=1b = 1 gives (∗)(*) and taking a=1a = 1 gives (∗∗)(**).

Intuition

The Archimedean property says that R\mathbb{R} contains no infinitely large and no infinitely small positive numbers. Its form (∗)(*) is the step “choose nn with 1n<ε\frac1n < \varepsilon”, which appears in almost every limit proof from §7 on.

Proof

Suppose the property fails: there are a,b>0a, b > 0 with na≤bna \le b for all n∈Nn \in \mathbb{N}. Then bb is an upper bound for S={na:n∈N}S = \{na : n \in \mathbb{N}\}, so s0=sup⁡Ss_0 = \sup S exists. (This is where completeness is used.) Since a>0a > 0, s0−a<s0s_0 - a < s_0, so s0−as_0 - a is not an upper bound for SS: there is n0∈Nn_0 \in \mathbb{N} with s0−a<n0as_0 - a < n_0 a. Then s0<(n0+1)as_0 < (n_0 + 1)a, and (n0+1)a∈S(n_0 + 1)a \in S, contradicting that s0s_0 is an upper bound.

∎
4.7 Denseness of ℚ.

If a,b∈Ra, b \in \mathbb{R} and a<ba < b, then there is a rational r∈Qr \in \mathbb{Q} with a<r<ba < r < b.

Used in Exercise 17.12

Proof

We need integers mm and n>0n > 0 with a<mn<ba < \frac mn < b, that is,

an<m<bn.(1)an < m < bn. \tag{1}

Since b−a>0b - a > 0, the Archimedean property gives n∈Nn \in \mathbb{N} with

n(b−a)>1,sobn−an>1.(2)n(b - a) > 1, \quad\text{so}\quad bn - an > 1. \tag{2}

It is plausible that an integer mm lies between anan and bnbn, but proving it takes some care. By the Archimedean property again there is an integer k>max⁡{∣an∣,∣bn∣}k > \max\{|an|, |bn|\}, so −k<an<bn<k-k < an < bn < k. The set {j∈Z:−k≤j≤k and an<j}\{j \in \mathbb{Z} : -k \le j \le k \text{ and } an < j\} is finite and contains kk, so it has a minimum mm. Then an<man < m, and since m>−km > -k, the integer m−1m - 1 is at least −k-k but fails an<m−1an < m - 1 by the choice of mm. So m−1≤anm - 1 \le an, and by (2), m≤an+1<bnm \le an + 1 < bn. Thus (1) holds.

∎

Intuitionanother face of completeness: nested intervals

Suppose [a1,b1]⊇[a2,b2]⊇[a3,b3]⊇⋯[a_1, b_1] \supseteq [a_2, b_2] \supseteq [a_3, b_3] \supseteq \cdots are closed bounded intervals. Then some real number lies in all of them.

To see why, let x=sup⁡{an:n∈N}x = \sup\{a_n : n \in \mathbb{N}\}, which exists by completeness since every an≤b1a_n \le b_1. Certainly an≤xa_n \le x for each nn. For any fixed mm and any nn, nesting gives an≤amax⁡{m,n}≤bmax⁡{m,n}≤bma_n \le a_{\max\{m,n\}} \le b_{\max\{m,n\}} \le b_m, so bmb_m is an upper bound for the ana_n’s and therefore x≤bmx \le b_m. Thus x∈[am,bm]x \in [a_m, b_m] for every mm.

Both hypotheses matter. The open intervals (0,1n)(0, \frac1n) are nested with nothing in common. And over Q\mathbb{Q}, the intervals [1,2],[1.4,1.5],[1.41,1.42],…[1, 2], [1.4, 1.5], [1.41, 1.42], \dots that close in on 2\sqrt2 share no rational point. This “nested interval” form of completeness is what makes repeated halving work, as in the bisection method of §18. Theorem 13.10 is the same idea in Rk\mathbb{R}^k.

Reading alongside Abbott

These ideas are in Abbott’s Sections 1.3 and 1.4. Abbott proves the nested interval property first, as the opening consequence of completeness, and uses it to show R\mathbb{R} is uncountable. The Archimedean property and the density of Q\mathbb{Q} follow with essentially the same proofs as here. Abbott also proves that 2\sqrt2 exists as a real number directly from the Axiom of Completeness. Here that fact comes more cheaply in §18, from the Intermediate Value Theorem.

Theorem 4.6 is a proof by contradiction that leans on completeness. Rebuild its argument.

Put the proof in order: Archimedean Property (Theorem 4.6)

These are the steps of the proof, shuffled. Use the buttons to move each step into place, then check.

  1. Argue by contradiction: suppose there are a,b>0a, b > 0 such that na≤bna \le b for every positive integer nn.
  2. Then S={na:n∈N}S = \{na : n \in \mathbb{N}\} is nonempty and bounded above by bb.
  3. By the Completeness Axiom, s0=sup⁡Ss_0 = \sup S exists.
  4. Because a>0a > 0, we have s0−a<s0s_0 - a < s_0, so s0−as_0 - a is too small to be an upper bound of SS.
  5. So some element of SS exceeds it: s0−a<n0as_0 - a < n_0 a for some n0∈Nn_0 \in \mathbb{N}.
  6. Adding aa to both sides gives s0<(n0+1)as_0 < (n_0 + 1)a.
  7. But (n0+1)a(n_0 + 1)a belongs to SS, so s0s_0 is not an upper bound of SS. This contradiction proves the property.

The skeleton of the proof of Theorem 4.7 is given. The gaps are the result it leans on and the integer it picks.

Fill in the gaps: Denseness of ℚ (Theorem 4.7)

This is the skeleton of the proof. In each gap choose the option that makes the argument work, then check.

  1. We want integers mm and n>0n > 0 with a<mn<ba < \frac mn < b. Multiplying by nn, that means an<m<bnan < m < bn.
  2. Since b−a>0b - a > 0, the gives n∈Nn \in \mathbb{N} with .
  3. The interval (an,bn)(an, bn) is longer than 11, so it should contain an integer. Take mm to be .
  4. Then m−1≤anm - 1 \le an, so m≤an+1<bnm \le an + 1 < bn by the choice of nn. Hence an<m<bnan < m < bn, and dividing by n>0n > 0 gives a<mn<ba < \frac mn < b.

    A proof that sounds fine until you ask what it has actually shown.

    Spot the error: the supremum of a set of fractions

    One line of this argument is wrong. Press the button on the line you think is the mistake.

    1. Let S={12,34,78,… }S = \left\{\frac12, \frac34, \frac78, \dots\right\}, the set of numbers 1−2−n1 - 2^{-n} for n∈Nn \in \mathbb{N}. We claim sup⁡S=2\sup S = 2.
    2. Every element of SS is less than 11, hence less than 22.
    3. So 22 is an upper bound of SS, and therefore sup⁡S=2\sup S = 2.

    Take the quiz (3 questions)

    Let S={1n:n∈N}S = \{\frac1n : n \in \mathbb{N}\}. Which is right?

    Let SS be nonempty with upper bound MM. Which must be true?

    Which subset of Q\mathbb{Q} has no least upper bound in Q\mathbb{Q}?

    Exercise 4.1

    For each set below that is bounded above, list three upper bounds for the set. Otherwise write “NOT BOUNDED ABOVE” or “NBA.” (An integer p≥2p \ge 2 is prime if its only positive factors are 11 and pp.)

    1. [0,1][0, 1]
    2. (0,1)(0, 1)
    3. {2,7}\{2, 7\}
    4. {π,e}\{\pi, e\}
    5. {1n:n∈N}\{\frac1n : n \in \mathbb{N}\}
    6. {0}\{0\}
    7. [0,1]∪[2,3][0,1] \cup [2,3]
    8. ⋃n=1∞[2n,2n+1]\bigcup_{n=1}^\infty [2n, 2n+1]
    9. ⋂n=1∞[−1n,1+1n]\bigcap_{n=1}^\infty [-\frac1n, 1 + \frac1n]
    10. {1−13n:n∈N}\{1 - \frac1{3^n} : n \in \mathbb{N}\}
    11. {n+(−1)nn:n∈N}\{n + \frac{(-1)^n}{n} : n \in \mathbb{N}\}
    12. {r∈Q:r<2}\{r \in \mathbb{Q} : r < 2\}
    13. {r∈Q:r2<4}\{r \in \mathbb{Q} : r^2 < 4\}
    14. {r∈Q:r2<2}\{r \in \mathbb{Q} : r^2 < 2\}
    15. {x∈R:x<0}\{x \in \mathbb{R} : x < 0\}
    16. {1,π3,π2,10}\{1, \frac\pi3, \pi^2, 10\}
    17. {0,1,2,4,8,16}\{0, 1, 2, 4, 8, 16\}
    18. ⋂n=1∞(1−1n,1+1n)\bigcap_{n=1}^\infty (1 - \frac1n, 1 + \frac1n)
    19. {1n:n∈N, n prime}\{\frac1n : n \in \mathbb{N},\ n \text{ prime}\}
    20. {x∈R:x3<8}\{x \in \mathbb{R} : x^3 < 8\}
    21. {x2:x∈R}\{x^2 : x \in \mathbb{R}\}
    22. {cos⁡nπ3:n∈N}\{\cos\frac{n\pi}{3} : n \in \mathbb{N}\}
    23. {sin⁡nπ3:n∈N}\{\sin\frac{n\pi}{3} : n \in \mathbb{N}\}
    Solution

    Any number at least as large as the supremum (Exercise 4.3) is an upper bound. For example, for (a) use 1,2,31, 2, 3. The sets that are not bounded above are (h), (k), (u). For (r), the intersection is {1}\{1\}, and for (i) it is [0,1][0, 1].

    Tags: Warm-upComputationSup & inf

    Used in Exercise 4.2, Exercise 4.3, Exercise 5.3

    Exercise 4.2

    Repeat Exercise 4.1 for lower bounds.

    Solution

    The sets that are not bounded below are (l), (o) and (t). Every other set is bounded below by its infimum (Exercise 4.4) and anything smaller.

    Tags: Warm-upComputationSup & inf

    Exercise 4.3

    For each set in Exercise 4.1, give its supremum if it has one. Otherwise write “NO sup.”

    Solution

    (a) 1 (b) 1 (c) 7 (d) π\pi (e) 1 (f) 0 (g) 3 (h) NO sup (i) 1 (j) 1 (k) NO sup (l) 2 (m) 2 (n) 2\sqrt2 (o) 0 (p) 10 (q) 16 (r) 1 (s) 12\frac12 (t) 2 (u) NO sup (v) 1 (w) 32\frac{\sqrt3}{2}

    Tags: Warm-upComputationSup & inf

    Used in Exercise 4.1, Exercise 4.4

    Exercise 4.4

    Repeat Exercise 4.3 for infima (the plural of infimum).

    Solution

    (a) 0 (b) 0 (c) 2 (d) ee (e) 0 (f) 0 (g) 0 (h) 2 (i) 0 (j) 23\frac23 (k) 0 (l) NO inf (m) −2-2 (n) −2-\sqrt2 (o) NO inf (p) π3\frac\pi3 (q) 0 (r) 1 (s) 0 (t) NO inf (u) 0 (v) −1-1 (w) −32-\frac{\sqrt3}2

    Tags: Warm-upComputationSup & inf

    Used in Exercise 4.2

    Exercise 4.5

    Let SS be a nonempty subset of R\mathbb{R} that is bounded above. Prove that if sup⁡S\sup S belongs to SS, then sup⁡S=max⁡S\sup S = \max S. (Your proof should be very short.)

    Solution

    sup⁡S\sup S is an upper bound for SS and belongs to SS, so it is the largest element of SS.

    Tags: Warm-upSup & inf

    Exercise 4.6

    Let SS be a nonempty bounded subset of R\mathbb{R}.

    1. Prove inf⁡S≤sup⁡S\inf S \le \sup S. (This is almost obvious; your proof should be short.)
    2. What can you say about SS if inf⁡S=sup⁡S\inf S = \sup S?
    Solution

    (a) Pick s∈Ss \in S. Then inf⁡S≤s≤sup⁡S\inf S \le s \le \sup S. (b) Every s∈Ss \in S satisfies inf⁡S≤s≤sup⁡S=inf⁡S\inf S \le s \le \sup S = \inf S, so SS consists of exactly one point.

    Tags: StandardSup & inf

    Used in Exercise 5.5

    Exercise 4.7

    Let SS and TT be nonempty bounded subsets of R\mathbb{R}.

    1. Prove that if S⊆TS \subseteq T, then inf⁡T≤inf⁡S≤sup⁡S≤sup⁡T\inf T \le \inf S \le \sup S \le \sup T.
    2. Prove sup⁡(S∪T)=max⁡{sup⁡S,sup⁡T}\sup(S \cup T) = \max\{\sup S, \sup T\}. (Do not assume S⊆TS \subseteq T.)
    Hint

    (a) sup⁡T\sup T is an upper bound for TT, hence for SS. (b) Show max⁡{sup⁡S,sup⁡T}\max\{\sup S, \sup T\} is an upper bound for S∪TS \cup T, then use (a) for “≥\ge”.

    Solution
    1. sup⁡T\sup T is an upper bound of TT, hence of S⊆TS \subseteq T, so the least upper bound satisfies sup⁡S≤sup⁡T\sup S \le \sup T. Likewise inf⁡T\inf T is a lower bound of SS, so inf⁡T≤inf⁡S\inf T \le \inf S. Finally, pick s∈Ss \in S; then inf⁡S≤s≤sup⁡S\inf S \le s \le \sup S.
    2. Let m=max⁡{sup⁡S,sup⁡T}m = \max\{\sup S, \sup T\}. Every x∈S∪Tx \in S \cup T lies in SS or in TT, so x≤sup⁡S≤mx \le \sup S \le m or x≤sup⁡T≤mx \le \sup T \le m. Thus mm is an upper bound of S∪TS \cup T, which is nonempty and bounded, and sup⁡(S∪T)≤m\sup(S \cup T) \le m. Conversely S⊆S∪TS \subseteq S \cup T and T⊆S∪TT \subseteq S \cup T, so by (a) sup⁡S≤sup⁡(S∪T)\sup S \le \sup(S \cup T) and sup⁡T≤sup⁡(S∪T)\sup T \le \sup(S \cup T), whence m≤sup⁡(S∪T)m \le \sup(S \cup T).

    Tags: StandardSup & inf

    Used in Exercise 5.6, §10

    Exercise 4.8

    Let SS and TT be nonempty subsets of R\mathbb{R} such that s≤ts \le t for all s∈Ss \in S and t∈Tt \in T.

    1. Observe SS is bounded above and TT is bounded below.
    2. Prove sup⁡S≤inf⁡T\sup S \le \inf T.
    3. Give an example of such sets SS and TT where S∩TS \cap T is nonempty.
    4. Give an example of sets SS and TT where sup⁡S=inf⁡T\sup S = \inf T and S∩TS \cap T is empty.
    Solution

    (b) Each t∈Tt \in T is an upper bound for SS, so sup⁡S≤t\sup S \le t. Thus sup⁡S\sup S is a lower bound for TT, and sup⁡S≤inf⁡T\sup S \le \inf T. (c) S=[0,1]S = [0, 1], T=[1,2]T = [1, 2]. (d) S=[0,1)S = [0, 1), T=[1,2]T = [1, 2].

    Tags: StandardSup & inf

    Exercise 4.9

    Complete the proof that inf⁡S=−sup⁡(−S)\inf S = -\sup(-S) in Corollary 4.5 by proving (1) and (2).

    Hint

    (1): for s∈Ss \in S, −s∈−S-s \in -S, so −s≤s0-s \le s_0. (2): if t≤st \le s for all s∈Ss \in S, then −t-t is an upper bound for −S-S, so s0≤−ts_0 \le -t.

    Solution

    Let s0=sup⁡(−S)s_0 = \sup(-S).

    (1) If s∈Ss \in S then −s∈−S-s \in -S, so −s≤s0-s \le s_0, i.e. −s0≤s-s_0 \le s.

    (2) Suppose t≤st \le s for all s∈Ss \in S. Then −t≥−s-t \ge -s for all s∈Ss \in S, so −t-t is an upper bound of −S-S (every element of −S-S has the form −s-s). Hence s0≤−ts_0 \le -t, i.e. t≤−s0t \le -s_0.

    Together, (1) says −s0-s_0 is a lower bound of SS and (2) says it is the greatest one, so inf⁡S=−s0=−sup⁡(−S)\inf S = -s_0 = -\sup(-S).

    Tags: StandardSup & inf

    Used in Corollary 4.5, Exercise 5.4

    Exercise 4.10

    Prove that if a>0a > 0, then there exists n∈Nn \in \mathbb{N} such that 1n<a<n\frac1n < a < n.

    Hint

    Use (∗)(*) and (∗∗)(**) to get n1n_1 and n2n_2, then take n=max⁡{n1,n2}n = \max\{n_1, n_2\}.

    Solution

    By (∗)(*) applied to aa there is n1∈Nn_1 \in \mathbb{N} with 1n1<a\frac1{n_1} < a, and by (∗∗)(**) there is n2n_2 with a<n2a < n_2. Let n=max⁡{n1,n2}n = \max\{n_1, n_2\}. Then 1n≤1n1<a<n2≤n\frac1n \le \frac1{n_1} < a < n_2 \le n.

    Tags: StandardSup & inf

    Exercise 4.11

    Consider a,b∈Ra, b \in \mathbb{R} with a<ba < b. Use Denseness of Q\mathbb{Q} 4.7 to show there are infinitely many rationals between aa and bb.

    Hint

    Get r1∈(a,b)r_1 \in (a, b), then r2∈(a,r1)r_2 \in (a, r_1), then r3∈(a,r2)r_3 \in (a, r_2), and so on.

    Solution

    By Denseness of Q\mathbb{Q} choose a rational r1r_1 with a<r1<ba < r_1 < b. Having chosen rk∈(a,b)r_k \in (a,b), apply denseness to a<rka < r_k to get a rational rk+1r_{k+1} with a<rk+1<rka < r_{k+1} < r_k. Then r1>r2>r3>⋯r_1 > r_2 > r_3 > \cdots are distinct rationals, all in (a,b)(a, b). So there are infinitely many.

    Tags: StandardSup & inf

    Used in §11

    Exercise 4.12

    Let I\mathbb{I} be the set of real numbers that are not rational; these are the irrational numbers. Prove that if a<ba < b, then there exists x∈Ix \in \mathbb{I} with a<x<ba < x < b.

    Hint

    First show {r+2:r∈Q}⊆I\{r + \sqrt2 : r \in \mathbb{Q}\} \subseteq \mathbb{I}. Then apply 4.7 to a−2<b−2a - \sqrt2 < b - \sqrt2.

    Solution

    2\sqrt2 is irrational (Corollary 2.3). For rational rr, r+2r + \sqrt2 is irrational: if r+2=q∈Qr + \sqrt2 = q \in \mathbb{Q} then 2=q−r∈Q\sqrt2 = q - r \in \mathbb{Q}, a contradiction. Now a−2<b−2a - \sqrt2 < b - \sqrt2, so Denseness of Q\mathbb{Q} gives a rational rr with a−2<r<b−2a - \sqrt2 < r < b - \sqrt2. Then x=r+2x = r + \sqrt2 is irrational and a<x<ba < x < b.

    Tags: StandardSup & infContradiction

    Used in Exercise 13.10

    Exercise 4.13

    Prove the following are equivalent for real numbers a,b,ca, b, c (either all of them hold or none does):

    1. ∣a−b∣<c|a - b| < c,
    2. b−c<a<b+cb - c < a < b + c,
    3. a∈(b−c,b+c)a \in (b - c, b + c).
    Hint

    Use Exercise 3.7(b).

    Solution

    (i) ⇔\Leftrightarrow (ii) is Exercise 3.7(b). (ii) ⇔\Leftrightarrow (iii) holds because (b−c,b+c)(b - c, b + c) is, by definition, the set of aa with b−c<a<b+cb - c < a < b + c. So the three statements are equivalent.

    Tags: Warm-upInequalities

    Exercise 4.14

    Let AA and BB be nonempty bounded subsets of R\mathbb{R}, and let A+B={a+b:a∈A, b∈B}A + B = \{a + b : a \in A,\ b \in B\}.

    1. Prove sup⁡(A+B)=sup⁡A+sup⁡B\sup(A + B) = \sup A + \sup B.
    2. Prove inf⁡(A+B)=inf⁡A+inf⁡B\inf(A + B) = \inf A + \inf B.
    Hint

    For sup⁡A+sup⁡B≤sup⁡(A+B)\sup A + \sup B \le \sup(A+B): for each b∈Bb \in B, sup⁡(A+B)−b\sup(A+B) - b is an upper bound for AA, so sup⁡A≤sup⁡(A+B)−b\sup A \le \sup(A+B) - b. Then sup⁡(A+B)−sup⁡A\sup(A+B) - \sup A is an upper bound for BB.

    Solution
    1. ≤\le: for a∈Aa \in A, b∈Bb \in B we have a+b≤sup⁡A+sup⁡Ba + b \le \sup A + \sup B, so sup⁡(A+B)≤sup⁡A+sup⁡B\sup(A+B) \le \sup A + \sup B. ≥\ge: let s=sup⁡(A+B)s = \sup(A+B). For fixed b∈Bb \in B and any a∈Aa \in A, a+b≤sa + b \le s, so a≤s−ba \le s - b. So s−bs - b is an upper bound of AA, and sup⁡A≤s−b\sup A \le s - b, i.e. b≤s−sup⁡Ab \le s - \sup A. This holds for every b∈Bb \in B, so s−sup⁡As - \sup A is an upper bound of BB and sup⁡B≤s−sup⁡A\sup B \le s - \sup A. Thus sup⁡A+sup⁡B≤s\sup A + \sup B \le s.
    2. Note −(A+B)=(−A)+(−B)-(A + B) = (-A) + (-B). By Corollary 4.5 and part (a), inf⁡(A+B)=−sup⁡((−A)+(−B))=−[sup⁡(−A)+sup⁡(−B)]=inf⁡A+inf⁡B.\inf(A+B) = -\sup\bigl((-A) + (-B)\bigr) = -\bigl[\sup(-A) + \sup(-B)\bigr] = \inf A + \inf B .

    Tags: ChallengeSup & inf

    Used in §5

    Exercise 4.15

    Let a,b∈Ra, b \in \mathbb{R}. Show that if a≤b+1na \le b + \frac1n for all n∈Nn \in \mathbb{N}, then a≤ba \le b. Compare Exercise 3.8.

    Hint

    If a>ba > b, use the Archimedean property to find nn with 1n<a−b\frac1n < a - b.

    Solution

    Suppose instead a>ba > b. By the Archimedean property (∗)(*) there is nn with 1n<a−b\frac1n < a - b, i.e. b+1n<ab + \frac1n < a. That contradicts the hypothesis a≤b+1na \le b + \frac1n. Hence a≤ba \le b.

    Tags: StandardInequalitiesContradiction

    Exercise 4.16

    Show sup⁡{r∈Q:r<a}=a\sup\{r \in \mathbb{Q} : r < a\} = a for each a∈Ra \in \mathbb{R}.

    Hint

    aa is clearly an upper bound. If M1<aM_1 < a, Denseness of Q\mathbb{Q} gives a rational rr with M1<r<aM_1 < r < a.

    Solution

    Let T={r∈Q:r<a}T = \{r \in \mathbb{Q} : r < a\}. It is nonempty (by denseness there is a rational in (a−1,a)(a-1, a)) and bounded above by aa, so sup⁡T≤a\sup T \le a. If sup⁡T<a\sup T < a, denseness gives a rational rr with sup⁡T<r<a\sup T < r < a; then r∈Tr \in T but r>sup⁡Tr > \sup T, which is impossible. Hence sup⁡T=a\sup T = a.

    Tags: StandardSup & inf

    Used in §6

    Keyboard shortcuts

    Single-key shortcuts
    KeyAction
    nNext section
    pPrevious section
    jNext exercise
    kPrevious exercise
    hShow or hide all hints
    sShow or hide all solutions
    /Search
    ?Show this list
    Ctrl K or ⌘ KSearch (built in)
    EscClose a dialog

    Shortcuts never fire while you type in a text field or hold Ctrl, Alt or ⌘. Screen reader users in browse mode can press the keys after switching to focus mode, or turn them off here.