Prove that if $a > b > 0, p > 0$, then $a^p > b^p.$

As I was reading baby Rudin, this fact was a step that Rudin skipped (Theorem 3.20a), but it is not obvious to me how to prove this.

Thanks in advance.

EDIT (relevant definitions and results from exercise 6, chapter 1):

EDIT 2 - Presume $b > 1$.

Let $r = m/n, n>0$, where $m$ and $n$ are integers. Then $b^r = (b^m)^{1/n}$. It is proved that $b^{r+s} = b^rb^s$. If $x$ is real and if we let $$B(x) = \{ b^r | r \in \mathbb{Q}, r\leq x \}$$ then $b^r = sup B(r)$ and we define $b^x = sup B(x) $ for any real $x$. Also it is proved that $b^{r+s} = b^xb^y$, for real $x$ and $y$.

Hopefully that helps, my bad for not including it the first time around.

  • 2
    $\begingroup$ Firstly, you need to ask how is $a^p$ defined. $\endgroup$ – Danny Pak-Keung Chan Dec 24 '18 at 20:25
  • $\begingroup$ If you know the derivative of $x \mapsto x^p$, then it's very easy. But as Danny points out, this involves knowing the definition of this function. $\endgroup$ – parsiad Dec 24 '18 at 20:28
  • 1
    $\begingroup$ Then Steven, please include exercise 6, chapter 1, in an edit to your question post. $\endgroup$ – Namaste Dec 24 '18 at 20:34
  • $\begingroup$ @amWhy you're right, should have included this. Editing the question now. $\endgroup$ – Steven Wagter Dec 24 '18 at 20:39
  • $\begingroup$ I do not have the book "Rudin". For the sake of completeness, it is better for you to type the definition. Moreover, $a^p$, being a real number, is obviously NOT the set $\{a^x \mid x\mbox{ is rational}\}$. $\endgroup$ – Danny Pak-Keung Chan Dec 24 '18 at 20:51

Let $a>b>0$. We go to prove that $a^{p}>b^{p}$ for any $p\in\mathbb{Q}\cap(0,\infty)$.

Firstly, we prove that $a^{n}>b^{n}$ for any $n\in\mathbb{N}$. This can be proved easily by induction. For, the formula is obviously true for $n=1$. Suppose that the formula is true for $n=k$, i.e., $a^{k}>b^{k}$, then $a^{k+1}=a\cdot a^{k}>a\cdot b^{k}>b\cdot b^{k}=b^{k+1}$. By mathematical induction, the formula is true for all $n\in\mathbb{N}$.

Next, we show that $a^{\frac{1}{n}}>b^{\frac{1}{n}}$ for any $n\in\mathbb{N}$ (Here, we assume that for any $x>0$, $n\in\mathbb{N}$, there exists $y>0$ such that $y^{n}=x$. Note that $y$ can be shown unique and is denoted by $x^{\frac{1}{n}}.$). Prove by contradiction, suppose the contrary that there exists $n\in\mathbb{N}$ such that $a^{\frac{1}{n}}\leq b^{\frac{1}{n}}$. If $a^{\frac{1}{n}}=b^{\frac{1}{n}},$ we have $a=\left(a^{\frac{1}{n}}\right)^{n}=\left(b^{\frac{1}{n}}\right)^{n}=b$, which is a contradiction. If $a^{\frac{1}{n}}<b^{\frac{1}{n}}$, then by the previous result, $\left(a^{\frac{1}{n}}\right)^{n}<\left(b^{\frac{1}{n}}\right)^{n}$. That is, $a<b$, which is also a contradiction.

Finally, let $p\in\mathbb{Q}\cap(0,\infty)$. Choose $m,n\in\mathbb{N}$ such that $p=\frac{m}{n}$. Then by the first part, $a^{m}>b^{m}$. By the second part, $\left(a^{m}\right)^{\frac{1}{n}}>\left(b^{m}\right)^{\frac{1}{n}}$. Hence $a^{p}>b^{p}$.

  • $\begingroup$ For the case that $p\in\mathbb{R}\cap(0,\infty)$, we need the precise definition of $a^p$. (Notice that there are many different but equivalent ways to define $a^p$). As I do not have the book nor the precise definition of $a^p$, so I skip the proof for the most general case. $\endgroup$ – Danny Pak-Keung Chan Dec 24 '18 at 20:40
  • $\begingroup$ The asker included the definition from Rudin, in a comment above. $\endgroup$ – Namaste Dec 24 '18 at 20:42

We know that $f^\prime(x)=px^{p-1}$. And we know that both $\dfrac{1}{x}>0$ and $x^p>0$ so it follows that $x^{p-1}>0$. So $f^\prime(x)=px^{p-1}>0$.

So $f$ is increasing on the interval $(0,\infty)$.

So a power function $f(x)=x^p$ is an increasing function when $p>0$.

So by the definition of an increasing function on $(0,\infty)$ $a>b$ if and only if $a^p>b^p$.

  • $\begingroup$ I know that, but isn't that a restatement of the result I am trying to prove? $\endgroup$ – Steven Wagter Dec 24 '18 at 20:29
  • $\begingroup$ I will add more to my answer. $\endgroup$ – John Wayland Bales Dec 24 '18 at 20:36
  • $\begingroup$ Does the proposition say anything about $f(x) = x^p$ being an increasing function? That's an observation you need to argue that $a> b \iff a^p >b^p$. For a non-increasing function, we can't conclude that. You need to use information about the function you defined in comments. $\endgroup$ – Namaste Dec 24 '18 at 20:37
  • $\begingroup$ However, you don't need to prove all of "if and only if"; you need only prove $a>b \implies a^p\gt b^p$ for $a>b>0, p>0$. $\endgroup$ – Namaste Dec 24 '18 at 20:39
  • $\begingroup$ I'm sorry, John Wayland Bales, my second to last comment above was addressing the asker, not you or your answer. $\endgroup$ – Namaste Dec 24 '18 at 20:45

You are using Rudin's "Principals of Mathematical Induction" and you are doing Chapter 1, Excercise 6. Which relies very heavily on the Theorem 1.21 and the proof thereof that;

For any $b > 1$ and $n \in \mathbb N$ there is a unique positive $c$ so that $c^n =b$. We call such a $c:= b^{\frac 1n}$.

The proof makes use of the least upper bound property and the archimedian principal and the fact that for all $c < b$ then $c^n < b*c^{n-1} < b^2*c^{n-2} < ....< b^n$. We then consider $C= \sup \{c|c^n < b\}$ and... the proof writes itself.

But HERE's the thing. In the process of doing this we have established that for all $c < b$ than so that $d= c^n < b$ that $c < b^{\frac 1n} = \sup \{c|c^n < b\}$. Thus for $d < b$ we have $c = d^{\frac 1n} < b^{\frac 1n}$.

And if that WASN't immediately clear, it'd have to be by contradiction:

If $d^{\frac 1n} \ge b^{\frac 1n}$ then $d=(d^{\frac 1n})^n \ge (b^{\frac 1n})^n = b$ which is a contradiction.

So if we show that it is consistent to define for $p =\frac mn$ that $b^p = (b^{\frac 1n})^m$ we would have $0 < a < b \iff 0 < a^{\frac 1n} < b^{\frac 1n} \iff a^{\frac mn} < b^{\frac mn}$.

And it'd only take a line to extend that result to $a^x = \sup \{a^q|q< x; q\in \mathbb Q\}< \sup\{a^q|q< x; q\in \mathbb Q\} = b^x$.

Which is why Rudin "took it for granted".

====in recap ==

For natural numbers it's clear by induction.

If $a^n > b^n > 0$ then $a^{n+1} =a^n*a > a^n*b > b^n*b = b^{n+1}$.

For $p = \frac 1n; n\in \mathbb N$ it's clear by contradiction.

If $a^{\frac 1n}\le b^{\frac 1n}$ we'd have $a = (a^{\frac 1n})^n \le (b^{\frac 1n})^n = b$.

So for rational $p = \frac nm; n,m\in \mathbb n$ then $a^p = (a^n)^{\frac 1m} > (b^n)^{\frac 1m} = b^p$.

ANd for any real $x>0$ we have $a^x = \sup\{a^q|q < x\} > \sup \{b^q|q < x\}$ [admittedly that step would need a sentence or two but it'd be straight forward] $= b^x$.


If $p\in \mathbb{N}$ induction mathematical.

  • 1
    $\begingroup$ I am seeking a proof where p can be any real number. $\endgroup$ – Steven Wagter Dec 24 '18 at 20:28
  • $\begingroup$ If $a^n \leq b^n \Rightarrow \ln a \leq \ln b$ Is a contradiction, since $a>b \Rightarrow \ln a> \ln b$ $\endgroup$ – Julio Trujillo Gonzalez Dec 24 '18 at 20:41
  • $\begingroup$ natural logarithm is a function monotonically increasing $\endgroup$ – Julio Trujillo Gonzalez Dec 24 '18 at 20:48

Here is another potential route through this.

Since $a\gt b\gt 0$ we have $\frac ab\gt 1$ and we might be in a position to say that $\frac ab=1+r$ with $r\gt 0$ and $\left(\frac ab\right)^p=(1+r)^p\gt 1$.

For example we can show that $(1+r)^n\gt 1^m$ for integer $n, m$ so we can do this for $p$ a positive rational.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.