0

Just a curiosity in this book of an high school with the title The powers with exponent rational

enter image description here

the $$\large \color{red}{a^{\frac mn}=\sqrt[n]{a^m}}, \quad a\geq 0$$ is given like a definition. I don't remember if there was a proof. But is there one or should we give it as a definition?

Sebastiano
  • 7,649
  • 2
    Questa è una definizione, ma poi si deve dimostrare che è ben definita, e le varie proprietà delle potenze anche per questi esponenti razionali. – Crostul Sep 03 '20 at 16:53
  • @Crostul Hi, from Sicily. I am happy to receive also your answer. – Sebastiano Sep 03 '20 at 19:04
  • 6
  • @Sebastiano There is a post from Andre Nicolas that you might find to be of interest. He states that this can indeed be a definition. However, we can also use the natural logarithm to define exponentiation, which tends to be what is done in practice. – Joe Sep 03 '20 at 20:50
  • @Joe Thank you very much but my question with your link is off-topic. I have asked if $a^{\frac mn}=\sqrt[n]{a^m}$ is a definition or a proof. – Sebastiano Sep 03 '20 at 21:03
  • @Sebastiano Reading the link I posted more carefully, I do agree that some of it is not relevant to your question. However, I think it is worth to bear in mind that often $a^x := \exp(x\log(a))$. This implies that $a^{m/n}:=\exp(m/n \log(a))$. So, using this approach, what you have written would most certainly be a theorem, not a definition. – Joe Sep 03 '20 at 21:09
  • @Joe When I must to start to explain $a^{m/n}$, for my students, I can not explain the concepts of $\exp$ and $\log$ that they are in other chapters. – Sebastiano Sep 03 '20 at 21:14
  • 1
    @Sebastiano Ah, I see—this approach would not be appropriate. Best of luck with your teaching! – Joe Sep 03 '20 at 21:16
  • @Joe :-) :-) Thank you very much however. – Sebastiano Sep 03 '20 at 21:18
  • @downvoter: Can you, please, give me the reason of the downvote? – Sebastiano Sep 04 '20 at 19:17

2 Answers2

3

Let $r$ be a positive rational number, represented as the quotient of two positive integers, $r=m/n$. We will show that two definitions of $a^r$, where $a\ge0$ are equivalent: $a^r=\sqrt[n]{a^m}$, and $a^r=(\sqrt[n]{a})^m$. Furthermore, we shall show that either definition is independent of the representation of $r$ as the quotient of positive integers.


PART $(1)$:

Let $y\ge 0$ be defined

$$y=\sqrt[n]{a^m}=(a^m)^{1/n}\tag1$$

By definition of the $n$'th root, we have from $(1)$

$$y^n=a^m\tag2$$


Now, let $y=z^m$ for some number $z\ge 0$. Then, substituting into $(2)$ reveals

$$(z^m)^n=a^m\tag3$$

Using $(z^m)^n=(z^n)^m$ in $(3)$, we find that $z^n=a$, which implies that

$$y=(a^{1/n})^m\tag4$$

Equating $(1)$ and $(4)$ yields

$$(a^m)^{1/n}=(a^{1/n})^m$$


So, the two definitions of $a^r$, $(a^m)^{1/n}$ and $(a^{1/n})^m$ that are equivalent. That is to say, that

$$\bbox[5px,border:2px solid #C0A000]{\sqrt[n]{a^m}=(\sqrt[n]{a})^m}\tag5$$



PART $(2)$:

Now, we shall show that the definitions of $a^r=a^{m/n}$ in $(5)$ are independent of the representation of $r$.

Suppose $\ell$ is an arbitrary integer. Then, $a^r=a^{m/n}=a^{(\ell m)/(\ell n)}$. We wish to show that

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

We let $w$ be defined as

$$w=(a^{\ell m})^{1/(\ell n)}\tag6$$

so that $a^{\ell m}=w^{\ell n}$, which implies $a^m=w^n$, from which we have

$$w=(a^m)^{1/n}$$

And similarly, we wish to show that

$$(\sqrt[\ell n]{a})^{\ell m}=(\sqrt[n]{a})^{ m}$$

Let $v= (\sqrt[\ell n]{a})^{\ell m}$. But from $(5)$ we have

$$v=\sqrt[\ell n]{a^{\ell m}}$$

which we just showed was equal to $\sqrt[n]{a^m}$. Therefore, we have proven the independence of the representation of $r$ as the quotient of positive integers.

Mark Viola
  • 179,405
1

$a^{\frac mn}$ can be defined as $\sqrt[n]{a^m}$, in which case there is no point of talking about a proof of $a^{\frac mn} = \sqrt[n]{a^m}$. What you should prove is that this is well defined, i.e. if $\frac mn = \frac{m'}{n'}$ then $\sqrt[n]{a^m} = \sqrt[n']{a^{m'}}$. Indeed \begin{align*} &&mn' &= m'n\\ \Rightarrow &&a^{mn'} &= a^{m'n}\\ \Rightarrow &&(a^m)^{n'} &= (a^{m'})^n\\ \Rightarrow &&((a^m)^{n'})^{1/nn'} &= ((a^{m'})^n)^{1/nn'}\\ \stackrel{(*)}{\Rightarrow} &&(((a^m)^{n'})^{1/n'})^{1/n} &= (((a^{m'})^n)^{1/n})^{1/n'}\\ \Rightarrow &&(a^m)^{1/n} &= (a^{m'})^{1/n'}. \end{align*} $(*)$ Here we make use of the property $b^{1/\kappa\lambda}=(b^{1/\kappa})^{1/\lambda}$ for $b \geq 0$ and $\kappa,\lambda \in \mathbb{Z}_{>0}$. This follows from the fact that both LHS and RHS are non-negative solutions of the equation $x^{\kappa\lambda}=b$.

You could also talk about the motivation behind the definition: We would like $(a^x)^y = a^{x\cdot y}$ to hold for rational exponents, as it holds when $x,y \in \mathbb{N}$. So we would like to define $a^{\frac mn}$ in a way so that $(a^{\frac mn})^n = a^m$.

posilon
  • 2,253
  • 2
    That's not quite true. One needs to show that $(a^m)^{1/n}=(a^{1/n})^m$. I've posted a solution accordingly. – Mark Viola Sep 03 '20 at 17:05
  • What is not true? And what's wrong with defining $a^{m/n} := \sqrt[n]{a^m}$ without first showing the equality you suggest? – posilon Sep 03 '20 at 19:06
  • You need to show that $\sqrt[n]{a^m}=(\sqrt[n]{a})^m$. Have you done that? If not, then you cannot define $a^{m/n}$ uniquely. Is it $(a^m)^{1/n}$ or is it $(a^{1/n})^m$? – Mark Viola Sep 03 '20 at 19:44
  • What's wrong with making an arbitrary choice between the two (for example $(a^m)^{1/n}$ as I suggested)? – posilon Sep 03 '20 at 19:48
  • Suppose that $r$ is a rational number which can be represented by $r=m/n$ for integers $m$ and $n$. Then, $r$ is also given by $r=(qm)/(qn)$ for any integer $q$. So, what is $a^r$? Is it $(a^m)^{1/n}$? Is it $(a^{2m})^{1/(2n)}$? How can you be sure these are equal? – Mark Viola Sep 03 '20 at 20:02
  • Yes, you do need to prove that these two are equal, and I have added this to my answer. But I don't see how this is the same with showing that $(a^m)^{1/n} = (a^{1/n})^m$. – posilon Sep 03 '20 at 20:04
  • What you should prove is that this is well defined, i.e. if $\frac mn=\frac{m'}{n'}$, $\sqrt[n]{a^m}=\sqrt[n']{a^{m'}}$." And how does one show this?

    – Mark Viola Sep 03 '20 at 20:09
  • I do agree that one can define $a^r=a^{m/n}$ as either $(a^{m})^{1/n}$ or as $(a^{1/n})^m$. But just seems a bit hollow and incomplete without showing the equivalence of the two definitions. More importantly, instead of writing "is defined as," one could write "can be defined as." – Mark Viola Sep 03 '20 at 20:20
  • I added a proof. Could you please explain how this follows from $(a^m)^{1/n} = (a^{1/n})^m$ instead? – posilon Sep 03 '20 at 20:47
  • You used $b^{1/(k\lambda)}=(b^{1/k})^{1/\lambda}$. This seems analogous. I've also edited. In Part $(2)$, I do use the result of Part $(1)$ to prove the independence of the representation of $r=m/n=m'/n'$. In any case, I have given your solution a (+1). – Mark Viola Sep 03 '20 at 21:08
  • I see, thank you. (+1) from me for your updated answer as well! – posilon Sep 04 '20 at 03:59
  • You're welcome! My pleasure. – Mark Viola Sep 04 '20 at 04:02