Skip to content

Notes on Chapter 1

FA

The fundamental axiom of analysis

If \(a_{n} \in \mathbb{R}\) for all \(n \in \mathbb{N}\), \(A \in \mathbb{R}\)

  • \(a_{1}\leq a_{2}\leq a_{3} \leq \dots\)
  • \(a_n < A\) for all \(n \in \mathbb{N}\)

Then there exists \(a \in \mathbb{R}\) such that \(a_n \to a\) as $n \to ∞ $

Theorem 1.2.1 The Axiom of Archimedes

The result is trivial in \(\mathbb{Q}\). The theorem is proving it in \(\mathbb{R}\).

This theorem proves that there is no \(\gimel>0\) in \(\mathbb{R}\) such that,

\(\frac{1}{n} > \gimel\) for all \(n\) \(\iff\) \(\gimel\) is smaller than all strictly positive rationals.

Suppose there exists such a \(0<\gimel< \frac{1}{n}\) for all \(n\).

Then by preservation of non strict inequality (Lemma 1.6 (vii)) \(0<\gimel \leq l = 0\) which contradicts the axiom of Archimedes.

If the theorem was true in \(\mathbb{Q}\) but not in \(\mathbb{R}\) we would have this exotic \(\gimel\).

Exercises

In A0 we proved \(0_{\mathbb{F}} < 1_{\mathbb{F}}\).

Let \(n_{\mathbb{F}} = 1_{\mathbb{F}} + \dots + 1_{\mathbb{F}}\) n times.

In A1 we show by induction that \(n_{\mathbb{F}} > 0_{\mathbb{F}}\) and therefore they are not equal.

This means that any ordered field \(\mathbb{F}\) has characteristic \(∞\).

We will then proceed to show that this characteristic of \(∞\) means we contain a copy of the rationals.

Some obvious facts

F1

\(n_{\mathbb{F}} + m_{\mathbb{F}} = (n+m)_{\mathbb{F}}\) for any \(n,m \in \mathbb{N}^+\)

F2

\(n_{\mathbb{F}} m_{\mathbb{F}} = (nm)_{\mathbb{F}}\) for any \(n,m \in \mathbb{N}^+\) due to distributivity.

F3

For any \(n \in \mathbb{N}^+\) we have $$ -n_{\mathbb{F}} = -(\overbrace{1_{\mathbb{F}} + \dots + 1_{\mathbb{F}} }^{n}) = \overbrace{-1_{\mathbb{F}} + \dots + -1_{\mathbb{F}}}^{n} $$

Hence, for any \(n,m \in \mathbb{N}^+\)

\[ \begin{align*} n_{\mathbb{F}} - m_{\mathbb{F}} &= \overbrace{1_{\mathbb{F}} + \dots + 1_{\mathbb{F}} }^{n} + \overbrace{-1_{\mathbb{F}}+ \dots + -1_{\mathbb{F}}}^{m} \\ &= \begin{cases} 0_{\mathbb{F}} &&\text{if } n=m \\ \overbrace{1_{\mathbb{F}} + \dots + 1_{\mathbb{F}}}^{n-m} &= (n-m)_{\mathbb{F}} &\text{ if } n>m \\ \overbrace{-1_{\mathbb{F}} + \dots + -1_{\mathbb{F}}}^{m-n} &= -(m-n)_{\mathbb{F}} &\text{ if } n < m \end{cases} \end{align*} \]

Exercise A3

Let \(\tau : \mathbb{Z} \to \mathbb{F}\) by,

\[ \tau(n) = \begin{cases} 0_{\mathbb{F}} & \text{if } n = 0 \\ n_{\mathbb{F}} & \text{if } n > 0 \\ -(-n)_{\mathbb{F}} & \text{if } n < 0 \end{cases} \]

i)

Show that the map is well defined, injective and

  • \(\tau \left( n+m \right) = \tau(n) + \tau(m)\)
  • \(\tau \left( nm \right) = \tau(n) \tau(m)\)
  • \(\tau(n) > 0_{\mathbb{F}}, \text{ whenever } n>0.\)

for all \(n,m \in \mathbb{Z}\).

The case definition is already well defined.

Additive

If one or more \(n,m\) equal 0 then additivity is trivial. Therefore assume either both positive OR one positive and one negative OR both negative.

Both positive case is trivial:

\[ \tau(n) + \tau(m) = n_{\mathbb{F}} + m_{\mathbb{F}} \overset{F1}{=} (n+m)_{\mathbb{F}} = \tau(n+m) \]

One positive and one negative case.Assume WLOG \(n>0\) and \(m<0\).

\[ \tau(n) + \tau(m) \overset{\text{Def}}{=} n_{\mathbb{F}} + -(-m)_{\mathbb{F}} = (*) \]

We can just apply F3 where \(-m\) takes the role of \(m\).

\[ (*) = \begin{cases} 0_{\mathbb{F}} & & \overset{\text{Def }}{=} \tau(0) &\text{ if } n=-m \\ (n--m)_{\mathbb{F}} &= (n+m)_{\mathbb{F}} & \overset{\text{Def }}{=} \tau(n+m) &\text{ if } n>-m \\ -(-m-n)_{\mathbb{F}} &= -(-(n+m))_{\mathbb{F}} & \overset{\text{Def }}{=} \tau(n+m) &\text{ if } n < -m \end{cases} \]

Multiplicative

If one of \(n,m\) is zero then it's easy to show \(nm=0_{\mathbb{F}} = \tau(n) \tau(m)\). Assume \(n,m \neq 0\), therefore there are cases.

If \(n,m>0\).

\[ \begin{align*} \tau(n) \tau(m) \overset{\text{Def}}{=} n_{\mathbb{F}} m_{\mathbb{F}} \overset{F2}{=} (nm)_{\mathbb{F}} \overset{\text{Def}}{=} \tau(nm) \end{align*} \]

If \(n>0, m<0\),

\[ \begin{align*} \tau(n) \tau(m) &\overset{\text{Def}}{=} n_{\mathbb{F}} \times -(-m)_{\mathbb{F}}\\ [\text{algebra}] &= - (n_{\mathbb{F}} \times (-m)_{\mathbb{F}}) \\ [\text{F2}] &= -(n \times -m)_{\mathbb{F}} = -(-nm)_{\mathbb{F}} \\ [\text{Def}] &= \tau(nm) \end{align*} \]

If \(n,m <0\),

\[ \begin{align*} \tau(n) \tau(m) &\overset{\text{Def}}{=} -(-n)_{\mathbb{F}} \times -(-m)_{\mathbb{F}} = (-n)_{\mathbb{F}} (-m)_{\mathbb{F}} \\ [\text{F2}] &= (-n \times -m)_{\mathbb{F}} = (nm)_{\mathbb{F}} \\ [\text{Def}] &= \tau(nm) \end{align*} \]

Positive Easy since \(\tau(n)=(n)_{\mathbb{F}}>0_{\mathbb{F}}\) by A2.

Injective

Prove via contrapositive. Assume \(n \neq m\), WLOG let \(n \geq m\), therefore \(n > m \overset{\text{F7}}{\implies} \tau(n) > \tau(m)\).

ii)

Define \(\phi: \mathbb{Q} \to \mathbb{F}\) by \(\phi(n/m) = \tau(n) \tau(m)^{-1}\)

This function is well defined and has properties

  • \(\phi (x+y) = \phi(x) + \phi(y)\)
  • \(\phi(xy) = \phi(x) \phi(y)\)
  • \(\phi(x) > 0_{\mathbb{F}}\) whenever \(x>0\)

Proof

Well Defined

Let \(x= \frac{n}{m} = \frac{\lambda n}{\lambda m}\) for some \(\lambda \in \mathbb{N}\). Then,

\[ \begin{align*} \phi\!\left(\frac{\lambda n}{\lambda m}\right) &= \tau(\lambda n) \tau(\lambda m)^{-1} \\ &= \tau(\lambda) \tau(n) \bigl(\tau(\lambda) \tau(m)\bigr)^{-1} \\ &= \tau(\lambda) \tau(n) \tau(\lambda)^{-1} \tau(m)^{-1} \\ &= \tau(n) \tau(m) \\ &= \phi(x) \end{align*} \]

Additive

Let \(x = \frac{n}{m}, y= \frac{p}{q}\) for \(n,m,p,q \in \mathbb{Z}\)

\[ \begin{align*} \phi(x+y) &= \phi\!\left(\frac{nq + pm}{mq}\right) \\ &= \tau(nq + pm) \tau(mq)^{-1} \\ &= \tau(nq + pm) \tau(m)^{-1} \tau(q)^{-1} \\ &= \bigl( \tau(n) \tau(q) + \tau(p) \tau(m) \bigr) \tau(m)^{-1} \tau(q)^{-1} \\ &= \tau(n) \tau(m)^{-1} + \tau(p) \tau(q)^{-1} \\ &= \phi(x) + \phi(y) \end{align*} \]

We also see

\[ \begin{align*} \phi(xy) &= \phi\!\left(\frac{np}{mq}\right) \\ &= \tau(n) \tau(p) \tau(m)^{-1} \tau(q)^{-1} \\ &= \tau(n) \tau(m)^{-1} \tau(p) \tau(q)^{-1} \\ &= \phi(x) \phi(y) \end{align*} \]

And we can assume \(x>0\) means that \(m,n>0\) since we can multiply both by \(-1\) if not,

\[ \phi(x) = \tau(n) \tau(m)^{-1} >0_{\mathbb{F}} \]

since A2 gives us \(m>0 \implies \tau(m)>0\) which then implies it's inverse is positive (standard result).

More obvious facts

F4

From definition, \(\phi(1) = \tau(1) \tau(1)^{-1} = 1_{\mathbb{F}}\)

This can also be proved as a standard algebra result for homomorphisms.

F5

\(\phi(-x) = -\phi(x)\), literally add them to get \(\phi(0) \overset{\text{Def}}{=}0_{\mathbb{F}}\)

F6

In A2 we proved \(0 < x \in \mathbb{Q} \implies \phi(x) > 0\)

The converse is straightforward by contrapositive. Assume \(x\leq0\).

If \(x=0\), then \(\phi(0) \overset{\text{Def}}{=} 0\).

If \(x<0\) then \(\phi(-x) = -\phi(x)>0\) by forward direction that we proved.

Hence \(\phi(x) \leq 0\).

Hence $$ \phi(x) > 0_{\mathbb{F}} \iff x > 0 $$

F7 Inequalities are equivalent accross \(\phi\)

\[ \begin{align*} \phi(x) > \phi(y) \iff \phi(x) - \phi(y) > 0_{\mathbb{F}} \iff \phi(x-y) > 0_{\mathbb{F}} \overset{\text{F6}}{\iff} x-y > 0 \end{align*} \]

F8 Axiom of Archimedes in general field

Let \(a_n = \phi \left( \frac{1}{n} \right)\). This sequence is bounded and decreasing (F7) and therefore converges by FA, call this limit \(l\).

Let

\[ \begin{align*} b_n &= \phi \left( \frac{1}{2} \right) \phi \left( \frac{1}{n} \right) = \phi \left( \frac{1}{2} \right) a_n \to \phi \left( \frac{1}{2} \right) l \\ & \text{ by Algebra of Limits, Lemma 1.6 } \\ &= \phi \left( \frac{1}{2n} \right) = a_{2n} \to l \\ & \text{by subsequence.} \end{align*} \]

Hence,

\[ \begin{align*} l = \phi \left( \frac{1}{2} \right) l &\iff \left( 1_{\mathbb{F}} - \phi \left( \frac{1}{2} \right) \right)l = 0 \\ &\iff \phi \left( \frac{1}{2} \right) l = 0 \\ &\iff l=0 \end{align*} \]

Basically the same as before!

F9

For any \(x \in \mathbb{F}\) there exists \(n_{\mathbb{F}} \in \phi(\mathbb{Z}) \colon x< n_{\mathbb{F}}\).

i \(x \leq 0_{\mathbb{F}}\) is trivial, \(n=1\) does the job.

ii assume \(x > 0_{\mathbb{F}}\).

By Archimedes there exists \(n \in \mathbb{N}\) such that $$ 0_{\mathbb{F}} < \phi \left( \frac{1}{n} \right) < \frac{1}{x} \iff \phi ( n ) = n_{\mathbb{F}} > x > 0_{\mathbb{F}} $$

F10

For any \(x \in \mathbb{F}\) there exists \(n_{\mathbb{F}}\) such that \(n-1 \leq x_{\mathbb{F}} \leq n_{\mathbb{F}}\)

Apply F8 to \(x\) and \(-x\) to get \(a_{\mathbb{F}} < x < b_{\mathbb{F}}\) for some \(a,b \in \mathbb{Z}\).

Let \(n=b\), if \(x < (n-1)_{\mathbb{F}}\) let \(n \colon = n-1\), repeat...

This terminates eventually since \(x<a_{\mathbb{F}}\) is not possible. Termination occurs when \(x \geq \mathbb{n-1}\)

Exercise A4

Let \(\mathbb{K} = \phi \left( \mathbb{Q} \right) \subseteq \mathbb{F}\).

Prove that if \(x \in \mathbb{F}\) then there exists sequence \(x_n \in \mathbb{K}\) such that \(x_1 \leq x_2 \leq \dots\) such that \(x_n \to x\) as \(n \to \infty\).

Proof By F9 there exists \(a,b \in \phi(\mathbb{Z})\) such that \(a \leq x<b\).

Let \(a_0, b_0 = a,b\). Then let \(c_i = \phi \left( \frac{1}{2} \right) \left( a_{i-1} + b_{i-1} \right)\)

\[ \begin{cases} a_i &= c_i & \text{ and } b_i = b_{i-1} & \text{ if } c_i \leq x \\ a_i &= a_{i-1} & \text{ and } b_i = c_i & \text{ if } x_i > x \end{cases} \]

Then \(a_n \leq x \leq b_n\) and \(a_n \leq a_{n+1}\) for all \(n \in \mathbb{N}\). Hence by FA \(a_n \to l\) for some \(l \in \mathbb{F}\).

Also, \(b_i - a_i = \phi\left( \frac{1}{2} \right)^n \left( b_0 - a_0 \right) \to 0_{\mathbb{F}}\).

Then \(0 \leq x - a_n \leq b_n - a_n\) implies \(0\leq x -l \leq 0\)

F11 Convergence in \(\mathbb{Q}\) implies convergence in \(\phi(\mathbb{Q})\)

Let \(a_n \in \mathbb{Q}\) and \(a \in \mathbb{Q}\). If \(a_n \to a\), then \(\phi(a_n) \to \phi(a)\) in \(\mathbb{F}\).

Proof Let \(\eta > 0_{\mathbb{F}}\). By F8 Axiom of Archimedes there exists \(m \in \mathbb{N}\) such that \(\phi \left( \frac{1}{m} \right) < \eta\). It is therefore sufficient to show

\[ \lvert \phi \left( a_n - a \right) \rvert = \lvert \phi(a_n) - \phi(a) \rvert < \phi \left( \frac{1}{m} \right) < \eta \]

By F7 this is equivalent to

\[ \lvert a_n - a \rvert < \frac{1}{m} \]

Which hold for sufficiently large \(N\) since \(a_n \to a\) in \(\mathbb{Q}\).

Hence \(\phi(a_n) \to \phi(a)\). \(\blacksquare\)

F12 Convergence in \(\phi(\mathbb{Q})\) implies convergence in \(\mathbb{Q}\)

Let \(q_n \in \phi(\mathbb{Q})\) and \(q \in \phi(\mathbb{Q})\). If \(q_n \to q\) in \(\mathbb{F}\), then \(\phi^{-1}(q_n) \to \phi^{-1}(q)\) in \(\mathbb{Q}\).

Proof Let \(\varepsilon > 0\) in \(\mathbb{Q}\). Then \(\phi(\varepsilon) > 0_{\mathbb{F}}\) by F6. Since \(q_n \to q\), there exists \(N\) such that for all \(n \geq N\),

\[ \left|q_n-q\right| < \phi(\varepsilon). \]

Because

\[ q_n-q = \phi\bigl(\phi^{-1}(q_n)-\phi^{-1}(q)\bigr), \]

F7 gives

\[ \left|\phi^{-1}(q_n)-\phi^{-1}(q)\right| < \varepsilon. \]

Hence \(\phi^{-1}(q_n) \to \phi^{-1}(q)\). \(\blacksquare\)

F13 Different limits in Q implies different limits in \(\phi(\mathbb{Q})\)

Let \(x_n,y_n \in \mathbb{Q}\) with \(x_n \to x\) and \(y_n \to y\) in \(\mathbb{R}\). If \(x \neq y\), then

\[ \lim_{n \to \infty} \phi(x_n) \neq \lim_{n \to \infty} \phi(y_n). \]

Proof Suppose \(\lim_{n \to \infty} \phi(x_n) = \lim_{n \to \infty} \phi(y_n).\)

Then \(\phi(x_n-y_n) = \phi(x_n)-\phi(y_n) \to 0_{\mathbb{F}}.\)

By F12, \(x_n-y_n \to 0\) in \(\mathbb{Q}\). Therefore,

\(x-y = \lim_{n \to \infty} (x_n-y_n) = 0\),

which contradicts \(x \neq y\). Hence the two limits are distinct. \(\blacksquare\)

Exercise A5

(i)

Let \(x \in \mathbb{R}\). By A4 with \(\mathbb{F}=\mathbb{R}\), there exists a sequence \(x_n \in \mathbb{Q}\) such that

\[ x_1 \leq x_2 \leq \dots \quad\text{and}\quad x_n \to x. \]

Convergence of \(x_n\) implies that it is bounded. Therefore by F7 \(\phi(x_n)\) is also bounded increasing and therefore converges by FA to some limit \(a \in \mathbb{F}\).

(ii)

Let \(x_n' \in \mathbb{Q}\) with \(x_n' \to x\). We claim that \(\phi(x_n') \to a\) as well.

Since \(x_n-x_n' \to 0\) in \(\mathbb{Q}\), F11 gives \(\phi(x_n -x_n') \to 0_{\mathbb{F}}\)

So, \(\phi(x_n) \to a\) and \(\phi(x_n) - \phi(x_n') \to 0_{\mathbb{F}}\) implies (by AOL) that \(\phi(x_n') \to a\).

(iii)

Define \(\Theta : \mathbb{R} \to \mathbb{F}\) by

\[ \Theta(x) = \lim_{n \to \infty} \phi(q_n), \]

where \((q_n)\) is any rational sequence converging to \(x\).

Part A4 gives us such a \(q_n\) exists. A5 (i) gives us convergnce. A5 (ii) gives us that this limit is the same regardless of rational sequence chose.

If \(x \in \mathbb{Q}\), taking the constant sequence \(q_n=x\) shows that \(\Theta(x)=\phi(x)\), so \(\Theta\) extends \(\phi\).

(iv) Injectivity

If \(x \neq y\), choose rational sequences \(x_n \to x\) and \(y_n \to y\). Then F13 gives

\[ \Theta(x) = \lim_{n \to \infty} \phi(x_n) \neq \lim_{n \to \infty} \phi(y_n) = \Theta(y), \]

so \(\Theta\) is injective.

(iv) Surjectivity

Let \(q \in \mathbb{F}\). By A4 there exists an increasing sequence \(q_n \in \phi(\mathbb{Q})\) such that \(q_n \to q\).

Convergence of \(q_n\) implies it's bounded. F7 gives us that \(\phi(q_n)^{-1}\) is therefore bounded and increasing in \(\mathbb{Q}\).

So by FA in \(\mathbb{R}\) there exists \(\ell \in \mathbb{R}\) such that \(\phi^{-1}(q_n) \to \ell.\)

Now \(\Theta(\ell)\) is the \(\phi\)-limit of any rational sequence converging to \(\ell\), so using the particular sequence \(\phi^{-1}(q_n)\),

\[ \Theta(\ell) = \lim_{n \to \infty} \phi\bigl(\phi^{-1}(q_n)\bigr) = \lim_{n \to \infty} q_n = q. \]

Hence \(\Theta\) is surjective.

(v)

Let \(x,y \in \mathbb{R}\), and choose rational sequences \(x_n \to x\) and \(y_n \to y\).

Then \(x_n+y_n \to x+y\), so by definition of \(\Theta\),

\[ \Theta(x+y) = \lim_{n \to \infty} \phi(x_n+y_n) = \lim_{n \to \infty} \bigl(\phi(x_n)+\phi(y_n)\bigr) = \Theta(x)+\Theta(y). \]

Also \(x_n y_n \to xy\), so similarly

\[ \Theta(xy) = \lim_{n \to \infty} \phi(x_n y_n) = \lim_{n \to \infty} \bigl(\phi(x_n)\phi(y_n)\bigr) = \Theta(x)\Theta(y). \]

Finally, if \(x>0\), choose \(m \in \mathbb{N}\) such that \(\frac{1}{m} < x\). Since \(x_n \to x\), we have \(x_n > \frac{1}{m}\) for all sufficiently large \(n\). Hence by F7,

\[ \phi(x_n) > \phi\left(\frac{1}{m}\right) > 0_{\mathbb{F}} \]

eventually, and therefore

\[ \Theta(x) = \lim_{n \to \infty} \phi(x_n) \geq \phi\left(\frac{1}{m}\right) > 0_{\mathbb{F}}. \]

So \(\Theta\) preserves positivity, and therefore order as well. Hence \(\Theta : \mathbb{R} \to \mathbb{F}\) is an isomorphism of ordered fields extending \(\phi\). \(\blacksquare\)