# 2.6: Powers with Arbitrary Real Exponents. Irrationals

- Page ID
- 19033

\( \newcommand{\vecs}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} } \)

\( \newcommand{\vecd}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash {#1}}} \)

In *complete* fields, one can define \(a^{r}\) for any \(a>0\) and \(r \in E^{1}\) (for \(r \in N,\) see §§5-6, Example \((\mathrm{f}) ) .\) First of all, we have the following theorem.

Theorem \(\PageIndex{1}\)

*Given* \(a \geq 0\) *in a complete field* \(F,\) *and a natural number* \(n \in E^{1}\), *there always is a unique element* \(p \in F, p \geq 0,\) *such that*

\[p^{n}=a.\]

*It is called the* \(n\) *th root o*f \(a,\) *denoted*

\[\sqrt[n]{a} \text{ or } a^{1 / n}.\]

(*Note that* \(\sqrt[n]{a} \geq 0,\) *by definition*.)

**Proof**-
A direct proof, from the completeness axiom, is sketched in Problems 1 and 2 below. We shall give a simpler proof in Chapter 4,§9, Example (a). At present, we omit it and temporarily take Theorem 1 for granted. Hence we obtain the following result.

Theorem \(\PageIndex{2}\)

*Every complete field* \(F\) (*such as* \(E^{1}\)) *has irrational elements, i.e., elements that are not rational*.

*In particular*, \(\sqrt{2}\) *is irrational*.

**Proof**-
By Theorem 1, \(F\) has the element

\[p=\sqrt{2} \text{ with } p^{2}=2\]

Seeking a contradiction, suppose \(\sqrt{2}\) is rational, i.e.,

\[\sqrt{2}=\frac{m}{n}\]

for some \(m, n \in N\) in lowest terms (see §7, final note).

Then \(m\) and \(n\) are not both even (otherwise, reduction by 2 would yield a smaller \(n ) .\) From \(m / n=\sqrt{2},\) we obtain

\[m^{2}=2 n^{2};\]

so \(m^{2}\) is

*even*.Only even elements have even squares, however. Thus \(m\) itself must be even; i.e., \(m=2 r\) for some \(r \in N .\) It follows that

\[4 r^{2}=m^{2}=2 n^{2}, \text{ i.e., } 2 r^{2}=n^{2}\]

and, by the same argument, \(n\)

*must be even*.This contradicts the fact that \(m\) and \(n\) are not

*both*even, and this contradiction shows that \(\sqrt{2}\) must be irrational. \(\square\)

**Note 1.** Similarly, one can prove the irrationality of \(\sqrt{a}\) where \(a \in N\) and \(a\) is not the square of a natural. See Problem 3 below for a hint.

**Note 2. **Theorem 2 shows that *the field* \(R\) *of all rationals is not complete* (for it contains no irrationals), even though it is *Archimedean* (see Problem 6 ). Thus the *Archimedean property* *does not impleteness* (but see Theorem 1 of §10).

Next, we define \(a^{r}\) for any *rational* number \(r>0\).

Definition

Given \(a \geq 0\) in a complete field \(F,\) and a rational number

\[r=\frac{m}{n} \quad\left(m, n \in N \subseteq E^{1}\right)\]

we define

\[a^{r}=\sqrt[n]{a^{m}}.\]

Here we must clarify two facts.

(1) If \(n=1,\) we have

\[a^{r}=a^{m / 1}=\sqrt[1]{a^{m}}=a^{m}.\]

If \(m=1,\) we get

\[a^{r}=a^{1 / n}=\sqrt[n]{a}.\]

Thus Definition 1 agrees with our previous definitions of \(a^{m}\) and \(\sqrt[n]{a}\) \((m, n \in N) .\)

(2) If \(r\) is written as a fraction *in two different ways*,

\[r=\frac{m}{n}=\frac{p}{q},\]

then, as is easily seen,

\[\sqrt[n]{a^{m}}=\sqrt[q]{a^{p}}=a^{r},\]

and so our definition is *unambiguous* (independent of the particular representation of \(r ) .\)

Indeed,

\[\frac{m}{n}=\frac{p}{q} \text{ implies } m q=n p,\]

whence

\[a^{m q}=a^{p n},\]

i.e.,

\[\left(a^{m}\right)^{q}=\left(a^{p}\right)^{n};\]

cf. §§5-6, Problem 6.

By definition, however,

\[\left(\sqrt[n]{a^{m}}\right)^{n}=a^{m} \text{ and } \left(\sqrt[q]{a^{p}}\right)^{q}=a^{p}.\]

Substituting this in \(\left(a^{m}\right)^{q}=\left(a^{p}\right)^{n},\) we get

\[\left(\sqrt[n]{a^{m}}\right)^{n q}=\left(\sqrt[q]{a^{p}}\right)^{n q},\]

whence

\[\sqrt[n]{a^{m}}=\sqrt[q]{a^{p}}.\]

Thus Definition 1 is valid, indeed.

By using the results of Problems 4 and 6 of §§5-6, the reader will easily obtain analogous formulas for powers with positive *rational* exponents, namely,

\[\begin{aligned} a^{r} a^{s} &=a^{r+s} ;\left(a^{r}\right)^{s}=a^{r s} ;(a b)^{r}=a^{r} b^{r} ; a^{r}<a^{s} \text { if } 0<a<1 \text { and } r>s \\ a &<b \text { iff } a^{r}<b^{r}(a, b, r>0) ; a^{r}>a^{s} \text { if } a>1 \text { and } r>s ; 1^{r}=1 \end{aligned}\]

Henceforth we assume these formulas known, for *rational* \(r, s>0\).

Next, we define \(a^{r}\) for any real \(r>0\) and any element \(a>1\) in a complete field \(F .\)

Let \(A_{a r}\) denote the set of all members of \(F\) of the form \(a^{x},\) with \(x \in R\) and \(0<x \leq r ;\) i.e.,

\[A_{ar}=\{a^{x} | 0<x \leq r, x \text{ rational}\}.\]

By the density of rationals in \(E^{1}\) (Theorem 3 of §10), such rationals \(x\) do exist; thus \(A_{\text {ar}} \neq \emptyset\).

Moreover, \(A_{a r}\) is *right bounded *in \(F .\) Indeed, fix any rational number \(y>r\). By the formulas in \((1),\) we have, for any positive *rational* \(x \leq r\),

\[a^{y}=a^{x+(y-x)}=a^{x} a^{y-x}>a^{x}\]

since \(a>1\) and \(y-x>0\) implies

\[a^{y-x}>1.\]

Thus \(a^{y}\) is an *upper bound* of all \(a^{x}\) in \(A_{a r}\).

Hence, by the assumed completeness of \(F,\) sup \(A_{ar}\) *exists*. So we may define

\[a^{r}=\sup A_{a r}.\]

We also put

\[a^{-r}=\frac{1}{a^{r}}.\]

If \(0<a<1\) (so that \(\frac{1}{a}>1 ),\) we put

\[a^{r}=\left(\frac{1}{a}\right)^{-r} \text{ and } a^{-r}=\frac{1}{a^{r}},\]

where

\[\left(\frac{1}{a}\right)^{r}=\sup A_{1 / a, r},\]

as above.

Summing up, we have the following definitions.

Definition

Given \(a>0\) in a complete field \(F\), and \(r \in E^{1}\), we define the following.(i) If \(r>0\) and \(a>1\), then

\[a^{r}=\sup A_{a r}=\sup \left\{a^{x} | 0<x \leq r, x \text { rational }\right\}\]

(ii) If \(r>0\) and \(0<a<1\), then \(a^{r}=\frac{1}{(1 / a)^{r}}\), also written \((1 / a)^{-r}.\)

(iii) \(a^{-r}=1 / a^{r}\). (This defines powers with negative exponents as well.)

We also define \(0^{r}=0\) for any real \(r>0\), and \(a^{0}=1\) for any \(a \in F\), \(a \neq 0\); \(0^{0}\) remains undefined.

The power \(a^{r}\) is also defined if \(a<0\) and \(r\) is a rational \(\frac{m}{n}\) with \(n\) because \(a^{r}=\sqrt[n]{a^{m}}\) has sense in this case. (Why?) This does not work for other values of \(r\). Therefore, in general, we assume \(a>0\).

Again, it is easy to show that the formulas in (1) remain also valid for powers with real exponents (see Problems } 8-13 below), provided \(F\) is complete.