math.wiki
Analysis › The real numbers › The Archimedean property and density of the rationals
Sign in

The Archimedean property and density of the rationals

Why can no real number lie above every positive integer, and why does every interval, however short, contain a fraction?

Where you are 4 in The real numbers

Why this exists

Least upper bounds and completeness gave every nonempty set of reals that is bounded above a least upper bound. Its first consequence looks too obvious to need proof: the positive integers are not bounded above. The axioms of The real numbers as an ordered field alone do not prove it, since some ordered fields have an element larger than every integer (see Boundaries). Yet every limit argument rests on it. To make \displaystyle \frac{1}{n} < \varepsilon one needs an integer \displaystyle n > \frac{1}{\varepsilon}, and the cutoff N in every convergence proof of Limits of sequences, epsilon and N is found this way.

The second consequence is about the The real number line. Completeness filled the gaps in \mathbb{Q}; are the rationals now spread thinly among the reals? No. Every open interval, however short, contains a rational number and an irrational one. That is why finite decimals can meet any tolerance.

Definition

Let F be an ordered field. Its positive integers are 1, 1 + 1, 1 + 1 + 1, and so on, and for x \in F and a positive integer n, nx is the sum of n copies of x.

  • F is Archimedean, or has the Archimedean property, if for every x, y \in F with x > 0 there is a positive integer n with nx > y.
  • A set D \subseteq \mathbb{R} is dense in \mathbb{R} if every open interval (x, y) with x < y contains an element of D.
  • For x \in \mathbb{R}, the floor \lfloor x \rfloor is the integer m with m \le x < m + 1 (Result 3 shows there is exactly one), the greatest integer function of algebra.

Real numbers that are not rational are irrational (Irrational and real numbers). The property is named after Archimedes, who took a form of it as an axiom.

Intuition

A step of length x > 0, however short, repeated often enough, passes any point y. Steps from 0 land on x, 2x, 3x, \dots, and the first to land beyond y is nx.

y
0
x
2x
3x
(n-1)x
nx
\cdots
Figure 1Steps of length x > 0 from 0 eventually pass any y. Here n is the first step count with nx > y.

For density, shrink the step instead. If \displaystyle \frac{1}{n} is shorter than y - x, the marks \displaystyle \frac{k}{n}, k an integer, cannot jump over the interval (x, y): the first mark past x lands before y.

x
y
\displaystyle \frac{1}{n}
\displaystyle \frac{m-1}{n}
\displaystyle \frac{m}{n}
\displaystyle \frac{m+1}{n}
Figure 2When the spacing \displaystyle \frac{1}{n} is less than y - x, the first mark \displaystyle \frac{m}{n} past x cannot jump over y.

Results

Results of The real numbers as an ordered field are cited as OF1 to OF12 and those of Least upper bounds and completeness as C1 to C7 (C2 is the approximation property). Throughout, x, y \in \mathbb{R}.

1. The Archimedean property. The positive integers are not bounded above in \mathbb{R}. Equivalently, if x > 0, then for every y there is a positive integer n with nx > y.

2. Small reciprocals. For every \varepsilon > 0 there is a positive integer n with \displaystyle \frac{1}{n} < \varepsilon. Hence \displaystyle \inf\{\frac{1}{n} : n \ge 1\} = 0, and 0 is not a minimum.

3. The integer part. For every x there is exactly one integer m with m \le x < m + 1.

4. Density of \mathbb{Q}. If x < y, there is a rational r with x < r < y. So every open interval contains infinitely many rationals.

5. Density of the irrationals. If x < y, there is an irrational s with x < s < y.

6. Approximation by rationals. x = \sup\{r \in \mathbb{Q} : r < x\} for every x.

Why it is true

Result 1. Proof. Let P be the set of positive integers, and suppose, for contradiction, that P is bounded above. It is nonempty, so b = \sup P exists by the least-upper-bound property. As b - 1 < b, b - 1 is not an upper bound, so some m \in P has m > b - 1. Then m + 1 \in P and m + 1 > b, against b being an upper bound.

Now let x > 0 and y \in \mathbb{R}. The number \displaystyle \frac{y}{x} is not an upper bound of P, so some positive integer n has \displaystyle n > \frac{y}{x}, and multiplying by x > 0 (OF5) gives nx > y. Conversely, the second statement with x = 1 says that P is not bounded above. \square

Result 2. Proof. Result 1 with x = \varepsilon and y = 1 gives n\varepsilon > 1, and multiplying by \displaystyle \frac{1}{n} > 0 (OF7, OF5) gives \displaystyle \varepsilon > \frac{1}{n}. Every \displaystyle \frac{1}{n} is positive (OF7), so 0 is a lower bound of the set, and for each \varepsilon > 0 some element lies in [0, \varepsilon). By C2 the infimum is 0, which is not an element. \square

Result 3. Proof. Let S = \{k \in \mathbb{Z} : k > x\}. By Result 1 some positive integer exceeds x, so S \ne \varnothing; and some positive integer p exceeds -x, so every k \in S has k > -p. By the well-ordering principle, in its form for integers bounded below, S has a least element k_0. Then k_0 > x, and k_0 - 1 \notin S gives k_0 - 1 \le x, so m = k_0 - 1 satisfies m \le x < m + 1. If m' does too, then m \le x < m' + 1 and m' \le x < m + 1 give m - m' < 1 and m' - m < 1. The only integer strictly between -1 and 1 is 0, so m = m'. \square

Result 4. Proof. Let x < y. Choose any positive integer n with \displaystyle \frac{1}{n} < y - x (Result 2 gives one), so that 1 < n(y - x) by OF5, and let m = \lfloor nx \rfloor + 1. By Result 3, m - 1 \le nx < m, so \displaystyle x < \frac{m}{n}. Also m \le nx + 1 < nx + n(y - x) = ny, so \displaystyle \frac{m}{n} < y, and \displaystyle r = \frac{m}{n} works. Applying this to (x, r) gives a rational r_2 < r in the interval, then to (x, r_2), and so on; the rationals r > r_2 > r_3 > \cdots are all different. \square

Result 5. Proof. Let x < y and take a rational r with x < r < y (Result 4). The positive real \displaystyle \sqrt{2} exists by C7. Result 2 gives a positive integer n with \displaystyle \frac{1}{n} < \frac{y - r}{\sqrt{2}}, so \displaystyle s = r + \frac{\sqrt{2}}{n} satisfies r < s < y. If s were rational, so would be \displaystyle \sqrt{2} = n(s - r), contradicting Irrational numbers. \square

Result 6. Proof. Let A = \{r \in \mathbb{Q} : r < x\}. Result 4 on (x - 1, x) puts a rational in A, and x is an upper bound of A. For \varepsilon > 0, Result 4 on (x - \varepsilon, x) gives r \in A with r > x - \varepsilon. By C2, \sup A = x. \square

Worked examples

Example 1

A supremum left over from completeness

Find the supremum, infimum, maximum and minimum, where they exist, of \displaystyle E = \{1 - \frac{1}{n} : n \ge 1\}. For \displaystyle \varepsilon = \frac{1}{250}, find the least n whose element exceeds 1 - \varepsilon.

Solution

Every element is less than 1, because \displaystyle \frac{1}{n} > 0, so 1 is an upper bound. Every element is at least 0, because \displaystyle \frac{1}{n} \le 1, and 0 is the element for n = 1. A lower bound in the set is its infimum (C1), so \inf E = \min E = 0.

Given \varepsilon > 0, Result 2 gives n with \displaystyle \frac{1}{n} < \varepsilon, and then \displaystyle 1 - \frac{1}{n} > 1 - \varepsilon. By C2, \sup E = 1. No element equals 1, so E has no maximum.

For \displaystyle \varepsilon = \frac{1}{250}: \displaystyle 1 - \frac{1}{n} > 1 - \frac{1}{250} exactly when \displaystyle \frac{1}{n} < \frac{1}{250}, that is, when n > 250 (OF7). The least such n is 251. This is the supremum Least upper bounds and completeness left to this lesson: without Result 2, nothing rules out a number below 1 lying above every \displaystyle 1 - \frac{1}{n}.

Example 2

A rational just above the square root of 5

Follow the proof of Result 4, with the least admissible n, to find a rational strictly between \displaystyle \sqrt{5} and \displaystyle \frac{9}{4}.

Solution

We need \displaystyle \frac{1}{n} < \frac{9}{4} - \sqrt{5}, that is, \displaystyle \sqrt{5} < \frac{9}{4} - \frac{1}{n}. Both sides are positive, so by OF9 this holds exactly when the square of the right side exceeds 5. For n = 71 the right side is \displaystyle \frac{635}{284}, and 635^2 = 403{,}225 < 403{,}280 = 5 \cdot 284^2, so n = 71 fails, and so does every smaller n, since \displaystyle \frac{1}{n} only grows. For n = 72 the right side is \displaystyle \frac{161}{72}, and

161^2 = 25{,}921 > 25{,}920 = 5 \cdot 72^2,

so n = 72 works.

Next, \displaystyle 72\sqrt{5} = \sqrt{25{,}920}, and 160^2 = 25{,}600 \le 25{,}920 < 161^2, so \displaystyle \lfloor 72\sqrt{5} \rfloor = 160 by OF9 and m = 161. The rational is \displaystyle \frac{161}{72}. It exceeds \displaystyle \sqrt{5} by the display, and \displaystyle \frac{161}{72} < \frac{162}{72} = \frac{9}{4}.

The margin is tiny: \displaystyle \frac{161}{72} - \sqrt{5} \approx 0.00004. The proof aims at existence, and simpler rationals such as \displaystyle 2.24 = \frac{56}{25} also lie in the interval.

Example 3

An irrational between two fractions

Use the proofs of Results 4 and 5 to find an irrational number strictly between \displaystyle \frac{3}{7} and \displaystyle \frac{4}{9}.

Solution

Here \displaystyle \frac{4}{9} - \frac{3}{7} = \frac{1}{63}, so \displaystyle \frac{1}{n} < \frac{1}{63} first holds at n = 64. Then \displaystyle 64 \cdot \frac{3}{7} = \frac{192}{7}, and 189 \le 192 < 196 gives \displaystyle \lfloor \frac{192}{7} \rfloor = 27, so m = 28 and \displaystyle r = \frac{28}{64} = \frac{7}{16}. Check: 3 \cdot 16 = 48 < 49 = 7 \cdot 7 and 7 \cdot 9 = 63 < 64 = 4 \cdot 16.

Now \displaystyle \frac{4}{9} - \frac{7}{16} = \frac{1}{144}, and the proof of Result 5 needs \displaystyle \frac{\sqrt{2}}{n} < \frac{1}{144}, that is, \displaystyle n > 144\sqrt{2}. By OF9 this compares n^2 with 2 \cdot 144^2 = 41{,}472, and 203^2 = 41{,}209 < 41{,}472 < 41{,}616 = 204^2. So the least such n is 204, and

s = \frac{7}{16} + \frac{\sqrt{2}}{204}

is irrational with \displaystyle \frac{3}{7} < \frac{7}{16} < s < \frac{4}{9}.

Example 4

Cutting shelves at the lumberyard

Emma cuts shelving at a lumberyard, and she cuts only on a mark of her tape. Her tapes are marked every \displaystyle \frac{1}{2^k} inch, for various k. Owen’s order calls for boards longer than 47.30 inches and shorter than 47.33 inches. (a) Find the least k for which every open window 0.03 inches wide contains a mark. (b) With that tape, which mark does the proof of Result 4 give for Owen’s window?

Solution

(a) The proof of Result 4 puts a mark in every window of width y - x as soon as the spacing is less than y - x. Here that needs \displaystyle 2^k > \frac{100}{3}, first true at k = 6, since 2^5 = 32 and 2^6 = 64. Spacing \displaystyle \frac{1}{32} is not enough: the window from \displaystyle 47\frac{5}{16} = 47.3125 to 47.3425 holds no mark of \displaystyle \frac{1}{32}, because the next one after 47.3125 is 47.34375. Coarser tapes carry only some of those marks, so they fail too. The answer is k = 6: sixty-fourths of an inch.

(b) m = \lfloor 64 \cdot 47.30 \rfloor + 1 = \lfloor 3027.2 \rfloor + 1 = 3028, and \displaystyle \frac{3028}{64} = 47\frac{5}{16} = 47.3125 inches, inside the window. That mark is on the sixteenths tape too: Owen’s window happens to catch one, but part (a) is about every window.

Example 5

The rationals below the square root of 2

Prove that \displaystyle \sup\{r \in \mathbb{Q} : r^2 < 2\} = \sqrt{2} in \mathbb{R}. For \displaystyle \varepsilon = \frac{1}{1000}, give an element above \displaystyle \sqrt{2} - \varepsilon.

Solution

Proof. Let A = \{r \in \mathbb{Q} : r^2 < 2\}; then 1 \in A. If some r \in A had \displaystyle r > \sqrt{2}, then r > 0 and OF9 would give r^2 > 2; so \displaystyle \sqrt{2} is an upper bound. Let \varepsilon > 0 and \displaystyle a = \max\{\sqrt{2} - \varepsilon, 0\}, so \displaystyle 0 \le a < \sqrt{2}. Result 4 gives a rational r with \displaystyle a < r < \sqrt{2}. Then \displaystyle 0 < r < \sqrt{2}, so r^2 < 2 by OF9: r \in A and \displaystyle r > \sqrt{2} - \varepsilon. By C2, \displaystyle \sup A = \sqrt{2}. \square

So the set with no supremum in \mathbb{Q} (C6) has one in \mathbb{R}, the irrational number the gap was missing. For \displaystyle \varepsilon = \frac{1}{1000}, r = 1.414 works: 1.414^2 = 1.999396 < 2, and 1.415^2 = 2.002225 > 2 gives \displaystyle \sqrt{2} < 1.415, so \displaystyle \sqrt{2} - 1.414 < 0.001.

Example 6

Powers of a number above 1

Prove that if c > 1, the set \{c^n : n \ge 1\} is not bounded above. For c = 1.01, compare the n the proof gives for c^n > 1000 with the least such n.

Solution

Proof. Let h = c - 1 > 0. First, Bernoulli’s inequality: (1 + h)^n \ge 1 + nh for every n \ge 1, by induction. It is an equality at n = 1. If it holds for n, multiply it by 1 + h > 0 (OF5) and use nh^2 \ge 0 (OF6):

\begin{aligned} (1 + h)^{n + 1} &\ge (1 + nh)(1 + h) \\ &= 1 + (n + 1)h + nh^2 \\ &\ge 1 + (n + 1)h. \end{aligned}

Now let M \in \mathbb{R}. Result 1 gives a positive integer n with nh > M, and then c^n \ge 1 + nh > M. So no M is an upper bound. \square

For c = 1.01 and M = 1000, the bound \displaystyle 1 + \frac{n}{100} > 1000 first holds at n = 99{,}901, so 1.01^{99{,}901} > 1000. The least n is far smaller: 1.01^{694} \approx 997.8 and 1.01^{695} \approx 1007.8, and the powers increase, so it is 695. The Archimedean property delivers an n, not a good one, and an existence proof needs no more.

Boundaries

Not every ordered field is Archimedean. In the field of rational functions \displaystyle \frac{p(t)}{q(t)} with real coefficients, call an element positive when the leading coefficients of p and q have the same sign. This makes it an ordered field (stated without proof; checking the axioms is routine). There t - n is positive for every positive integer n, so t bounds the positive integers and \displaystyle \frac{1}{t} is a positive element below every \displaystyle \frac{1}{n} (Exercise 20). So Result 1 does not follow from the axioms of The real numbers as an ordered field; its proof needs completeness.

The Archimedean property does not give completeness. \mathbb{Q} is Archimedean (Exercise 19), yet \{x \in \mathbb{Q} : x^2 < 2\} has no supremum in \mathbb{Q}.

Dense is not adjacent. No rational is the next one after x: between x and any rational r > x lies another. \mathbb{Z} is not dense: (0, 1) contains no integer.

Dense is not large. \mathbb{Q} meets every open interval, yet Countable and uncountable sets shows that \mathbb{Q} is countable and \mathbb{R} is not.

Existence, not economy. The proofs give some n or \displaystyle \frac{m}{n}, rarely the best one: \displaystyle \frac{161}{72} beside \displaystyle \frac{56}{25}, and 99{,}901 beside 695.

Where this goes next

Countable and uncountable sets sets the density of \mathbb{Q} beside its countability, and proves \mathbb{R} uncountable with nested intervals. In Limits of sequences, epsilon and N, Result 2 says exactly that \displaystyle \frac{1}{n} \to 0, every cutoff N is an integer the Archimedean property supplies, and the powers example gives q^n \to 0 for 0 < q < 1. The monotone convergence theorem and the The Riemann integral, defined lean on it whenever a quantity must be made smaller than \varepsilon by taking n large.

Exercises

Finding the integer

  1. Find the least positive integer n with 0.07n > 50.
  2. Find the least positive integer n with \displaystyle \frac{3}{2n + 1} < \frac{1}{400}.
  3. Find the supremum, infimum, maximum and minimum, where they exist, of \displaystyle \left\{\frac{n + 2}{3n} : n \ge 1\right\}, and the least n whose element is within \displaystyle \frac{1}{100} of the infimum.
  4. Find \displaystyle \lfloor -\frac{17}{5} \rfloor, \displaystyle \lfloor \frac{23}{7} \rfloor and \displaystyle \lfloor -\sqrt{10} \rfloor, justifying each with m \le x < m + 1.
  5. Use Bernoulli’s inequality as in the last example to find an n with 0.9^n < 0.001. Then find the least such n.

Rationals and irrationals in an interval

  1. Follow the proof of Result 4, with the least admissible n, to find a rational strictly between \displaystyle \sqrt{3} and \displaystyle \frac{7}{4}.
  2. Do the same for \displaystyle \frac{5}{13} and \displaystyle \frac{7}{18}. Then find a rational between them with a smaller denominator.
  3. Find every positive decimal r with exactly three decimal places and 3 < r^2 < 3.01, and show that no positive decimal with fewer places works.
  4. Find the least positive integer k with \displaystyle \frac{\sqrt{7}}{k} < \frac{1}{200}, and explain why \displaystyle \frac{\sqrt{7}}{k} is irrational.
  5. Owen’s next order calls for boards longer than 35.71 inches and shorter than 35.75 inches. Show that a tape marked in sixteenths has no mark in this window. Find the least k for which a \displaystyle \frac{1}{2^k} tape has a mark in every open window 0.04 inches wide, and the mark the proof of Result 4 gives here.
  6. Find \displaystyle \lfloor 25\sqrt{11} \rfloor, and check that \displaystyle \frac{\lfloor 25\sqrt{11} \rfloor}{25} lies within \displaystyle \frac{1}{25} below \displaystyle \sqrt{11}.

Proofs

  1. Prove that if a \ge 0 and \displaystyle a \le \frac{1}{n} for every positive integer n, then a = 0.
  2. Prove that for every real x there is exactly one integer k with k - 1 < x \le k.
  3. Prove that for every real x and positive integer n there is an integer m with \displaystyle \left\lvert x - \frac{m}{n} \right\rvert \le \frac{1}{2n}.
  4. Prove that x = \inf\{r \in \mathbb{Q} : r > x\} for every real x.
  5. Prove that every open interval (x, y) with x < y contains infinitely many irrational numbers.
  6. Prove that the numbers \displaystyle \frac{m}{2^k}, with m an integer and k \ge 1, are dense in \mathbb{R}.
  7. Prove that if 0 < q < 1, then \inf\{q^n : n \ge 1\} = 0.
  8. Prove, without the least-upper-bound property, that \mathbb{Q} is Archimedean.
  9. In the field of rational functions ordered as in Boundaries, prove that t > n and \displaystyle 0 < \frac{1}{t} < \frac{1}{n} for every positive integer n.
  10. Let F be an ordered field in which every interval (x, y) with x < y contains an element \displaystyle \frac{m}{n}, with m an integer and n a positive integer of F. Prove that F is Archimedean.
Answers
  1. 0.07n > 50 exactly when \displaystyle n > \frac{50}{0.07} = 714.28\ldots. Indeed 0.07 \cdot 714 = 49.98 and 0.07 \cdot 715 = 50.05, so n = 715.
  2. Both denominators are positive, so the inequality holds exactly when 2n + 1 > 1200 (OF5), that is, n > 599.5. The least n is 600.
  3. \displaystyle \frac{n + 2}{3n} = \frac{1}{3} + \frac{2}{3n}, which decreases in n. The maximum and supremum are 1, at n = 1. Every element exceeds \displaystyle \frac{1}{3}, and \displaystyle \frac{2}{3n} < \varepsilon once \displaystyle n > \frac{2}{3\varepsilon} (Result 1), so the infimum is \displaystyle \frac{1}{3} by C2, with no minimum. Within \displaystyle \frac{1}{100}: \displaystyle \frac{2}{3n} < \frac{1}{100} exactly when \displaystyle n > \frac{200}{3}, so the least n is 67.
  4. -4 \le -3.4 < -3, so \displaystyle \lfloor -\frac{17}{5} \rfloor = -4. 21 \le 23 < 28 gives \displaystyle 3 \le \frac{23}{7} < 4, so \displaystyle \lfloor \frac{23}{7} \rfloor = 3. 9 < 10 < 16 gives \displaystyle 3 < \sqrt{10} < 4 (OF9), so \displaystyle -4 < -\sqrt{10} < -3 and \displaystyle \lfloor -\sqrt{10} \rfloor = -4.
  5. With \displaystyle h = \frac{1}{9}, \displaystyle \left(\frac{10}{9}\right)^n \ge 1 + \frac{n}{9} > 1000 once n > 8991, so n = 8992 gives \displaystyle \left(\frac{10}{9}\right)^n > 1000, that is, 0.9^n < 0.001 (OF7). The powers 0.9^n decrease, 0.9^{65} \approx 0.00106 and 0.9^{66} \approx 0.000955, so the least n is 66.
  6. \displaystyle \frac{1}{n} < \frac{7}{4} - \sqrt{3} means \displaystyle \sqrt{3} < \frac{7}{4} - \frac{1}{n}. For n = 55 the right side is \displaystyle \frac{381}{220}, and 381^2 = 145{,}161 < 145{,}200 = 3 \cdot 220^2, so n = 55 fails, and every smaller n with it. For n = 56 it is \displaystyle \frac{97}{56}, and 97^2 = 9409 > 9408 = 3 \cdot 56^2, so n = 56. Since 96^2 = 9216 \le 9408 < 97^2, \displaystyle \lfloor 56\sqrt{3} \rfloor = 96 and m = 97: the rational is \displaystyle \frac{97}{56}, below \displaystyle \frac{98}{56} = \frac{7}{4}.
  7. \displaystyle \frac{7}{18} - \frac{5}{13} = \frac{1}{234}, so n = 235. Then \displaystyle \frac{235 \cdot 5}{13} = \frac{1175}{13} has floor 90, so m = 91 and the rational is \displaystyle \frac{91}{235}: 5 \cdot 235 = 1175 < 1183 = 13 \cdot 91 and 18 \cdot 91 = 1638 < 1645 = 7 \cdot 235. A smaller denominator: \displaystyle \frac{12}{31}, since 5 \cdot 31 = 155 < 156 = 13 \cdot 12 and 18 \cdot 12 = 216 < 217 = 7 \cdot 31.
  8. For r > 0, r^2 increases with r (OF9). With three places, 1.732^2 = 2.999824, 1.733^2 = 3.003289, 1.734^2 = 3.006756 and 1.735^2 = 3.010225, so r = 1.733 and r = 1.734. A decimal with at most two places is either at most 1.73, with 1.73^2 = 2.9929 < 3, or at least 1.74, with 1.74^2 = 3.0276 > 3.01.
  9. \displaystyle k > 200\sqrt{7} means k^2 > 7 \cdot 200^2 = 280{,}000 (OF9), and 529^2 = 279{,}841 < 280{,}000 < 280{,}900 = 530^2, so k = 530. If \displaystyle \frac{\sqrt{7}}{530} were a rational q, then \displaystyle \sqrt{7} = 530q would be rational; but 7 is not a perfect square, so \displaystyle \sqrt{7} is irrational (Irrational numbers).
  10. The sixteenths marks nearby are \displaystyle 35\frac{11}{16} = 35.6875 and \displaystyle 35\frac{12}{16} = 35.75, and neither is strictly inside. A \displaystyle \frac{1}{2^k} tape is guaranteed a mark when 2^k > 25, first at k = 5; k = 4 fails in this window, and coarser tapes carry only some of its marks. Then m = \lfloor 32 \cdot 35.71 \rfloor + 1 = \lfloor 1142.72 \rfloor + 1 = 1143, and \displaystyle \frac{1143}{32} = 35\frac{23}{32} = 35.71875 inches.
  11. \displaystyle 25\sqrt{11} = \sqrt{6875}, and 82^2 = 6724 \le 6875 < 6889 = 83^2, so \displaystyle \lfloor 25\sqrt{11} \rfloor = 82. From \displaystyle 82 \le 25\sqrt{11} < 83, \displaystyle 3.28 \le \sqrt{11} < 3.32, so \displaystyle \frac{82}{25} = 3.28 is within \displaystyle \frac{1}{25} below \displaystyle \sqrt{11} (in fact \displaystyle \sqrt{11} - 3.28 \approx 0.0366).
  12. Proof. If a > 0, Result 2 gives n with \displaystyle \frac{1}{n} < a, against \displaystyle a \le \frac{1}{n}. So a \le 0, and with a \ge 0, a = 0. \square
  13. Proof. k - 1 < x \le k is the same as -k \le -x < -k + 1 (OF4). Result 3 for -x gives exactly one integer j with j \le -x < j + 1, so k = -j is the one integer that works. \square
  14. Proof. Let \displaystyle m = \lfloor nx + \frac{1}{2} \rfloor. Then \displaystyle m \le nx + \frac{1}{2} < m + 1, so \displaystyle -\frac{1}{2} \le nx - m < \frac{1}{2} and \displaystyle \lvert nx - m \rvert \le \frac{1}{2} (OF10). Dividing by n > 0 gives \displaystyle \left\lvert x - \frac{m}{n} \right\rvert \le \frac{1}{2n}. \square
  15. Proof. Let B = \{r \in \mathbb{Q} : r > x\}. Result 4 on (x, x + 1) puts a rational in B, and x is a lower bound. For \varepsilon > 0, Result 4 on (x, x + \varepsilon) gives r \in B with r < x + \varepsilon. By the infimum half of C2, \inf B = x. \square
  16. Proof. Result 5 gives an irrational s_1 \in (x, y). Given s_j, Result 5 on (x, s_j) gives an irrational s_{j + 1} with x < s_{j + 1} < s_j. The irrationals s_1 > s_2 > s_3 > \cdots all lie in (x, y) and are all different. \square
  17. Proof. Bernoulli’s inequality with h = 1 gives 2^k \ge 1 + k > k for k \ge 1. Let x < y. Result 2 gives k \ge 1 with \displaystyle \frac{1}{k} < y - x, so \displaystyle \frac{1}{2^k} < \frac{1}{k} < y - x (OF7). The proof of Result 4 works for every positive integer n with \displaystyle \frac{1}{n} < y - x; with n = 2^k it gives m = \lfloor 2^k x \rfloor + 1 and \displaystyle x < \frac{m}{2^k} < y. \square
  18. Proof. Let \displaystyle c = \frac{1}{q}, so c > 1 (OF7). Every q^n > 0, so 0 is a lower bound. Given \varepsilon > 0, the powers example gives n with \displaystyle c^n > \frac{1}{\varepsilon}, so \displaystyle q^n = \frac{1}{c^n} < \varepsilon (OF7). By the infimum half of C2, the infimum is 0. \square
  19. Proof. Let x, y \in \mathbb{Q} with x > 0. Write \displaystyle x = \frac{a}{b} with a, b positive integers and \displaystyle y = \frac{c}{d} with c an integer and d a positive integer. Let n = b(\lvert c \rvert + 1). Then \displaystyle nx = a(\lvert c \rvert + 1) \ge \lvert c \rvert + 1 > \lvert c \rvert \ge \frac{\lvert c \rvert}{d} \ge \frac{c}{d} = y, using a \ge 1, d \ge 1 and OF10. \square
  20. Proof. \displaystyle t - n = \frac{t - n}{1} has leading coefficients 1 and 1, so it is positive and t > n. \displaystyle \frac{1}{t} has leading coefficients 1 and 1, so \displaystyle \frac{1}{t} > 0. And \displaystyle \frac{1}{n} - \frac{1}{t} = \frac{t - n}{nt} has leading coefficients 1 and n, both positive, so \displaystyle \frac{1}{t} < \frac{1}{n}. \square
  21. Proof. Let x > 0 and y \in F. The interval (y, y + 1) contains some \displaystyle \frac{p}{q} with q \ge 1, and \displaystyle \frac{p}{q} \le \frac{\lvert p \rvert}{q} \le \lvert p \rvert (OF10, OF5, OF7), so the positive integer N = \lvert p \rvert + 1 exceeds y. The interval (0, x) contains some \displaystyle \frac{a}{b} with b \ge 1; as \displaystyle \frac{a}{b} > 0, a is a positive integer, so \displaystyle \frac{1}{b} \le \frac{a}{b} < x. With n = Nb, \displaystyle nx > Nb \cdot \frac{1}{b} = N > y. \square
math.wiki · updated 2026-11-17 · CC-BY-SA-4.0