5

I'm looking to learn Real Analysis on my own. Am reading Elements of Real Analysis by Bartle. I came across this project which defines the powers of real numbers i.e. exponentiation.

Firstly I am unsure about my solutions and they are too long to be posted here. Furthermore, the author restricts the development to bases that are strictly greater than $1$. I wanted to bypass this section for now and return later, but the theorem:

$$\forall \ \ a, b \gt 0: a^x \gt b^x \iff a \gt b \;\; \forall \ x \gt 0 $$

is something that is used extensively and I feel like I'm cheating when I take it for granted without seeing a proof.

Can anyone direct me to a reference which defines $a^x$ for real $x$ so that I can move forward. I am only just beginning to finish off the Topology chapters so my knowledge on sequences isn't all that great either.

Any advice is appreciated. Thanks in advance.

Ishfaaq
  • 10,230
  • 5
    The usual definition is $a^x=e^{x\ln(a)}=\sum_{k=0}^\infty \frac{(x\ln(a))^k}{k!}$. – J.R. Feb 09 '14 at 01:16
  • 4
    You probably want $x>0$ also, since $x=-1$ screws up your assertion. – Ian Coley Feb 09 '14 at 01:16
  • @Ian Coley: you're right. Corrected.. – Ishfaaq Feb 09 '14 at 01:26
  • @TooOldForMath: That's my problem. My knowledge on sequences is confined to the basic definition of the limit of a convergent sequence and the Cauchy criterion for convergence. So I doubt if I could prove the above statement using this definition. Any advice? – Ishfaaq Feb 09 '14 at 01:28
  • 1
    Then you can also use an alternative definition: for rational $x=p/q$ you define $a^x=\sqrt[q]{a^p}$. Then for real $x$ you choose a sequence of approximating rationals $(p_n/q_n)n$ and define $a^x=\lim{n\rightarrow\infty} \sqrt[q_n]{a^{p_n}}$. – J.R. Feb 09 '14 at 01:31
  • So $(\frac {p_n} {q_n})$ converges to $x$, hence $\lim {(a^{p_n})^{\frac 1 {q_n}} } $ would converge to $a^x$?? Lemme see if my inept brains can make some sense of this.. – Ishfaaq Feb 09 '14 at 01:39
  • @TooOldForMath: Nope.. Not making much progress. I think I can go from $a \gt b$ to $(a^{p_n})^{\frac 1 {q_n}} \gt (b^{p_n})^{\frac 1 {q_n}} $ But then I can only tell myself intuitively that the inequality holds for the limit. No rigour... – Ishfaaq Feb 09 '14 at 01:53
  • 1
    some problems are better with logarithms. Do you want to build a house with a hammer or a nail gun. Choose the right tool. – James S. Cook Feb 09 '14 at 02:35
  • @James S. Cook: Doesn't matter.. Don't think I have any nails to use either way.. What do you think people? Should I keep pursuing this or should I abandon this for now and come back to it when I know more about series and sequences? Any advice for a poor suffering soul?? – Ishfaaq Feb 09 '14 at 02:38
  • I would say go on with your studies while admitting there is a gap. When you study sequences and series in a bit the gap will be filled with much less effort. – James S. Cook Feb 09 '14 at 02:49
  • @James S. Cook: Alright Sir.. I shall take your advice. Thanks everyone for your input. – Ishfaaq Feb 09 '14 at 02:52

1 Answers1

3

You should proceed with the definition $a^{x} = \lim_{n \to \infty}a^{x_{n}}$ where $x_{n}$ is a sequence of rationals tending to $x$ and $a > 0$. This route is bit difficult compared to the standard route of using logarithms via integral. But providing a rigorous justification of all the usual properties of exponents using the above definition turns out to be a good exercise in real analysis.

The first thing one needs to do is to show that the limit $\lim_{n \to \infty}a^{x_{n}}$ exists and the definition is unambiguous i.e. if $x_{n}, y_{n}$ are sequences of rationals both tending to $x$ then $\lim_{n \to \infty}a^{x_{n}} = \lim_{n \to \infty}a^{y_{n}}$. The algebraic properties like $a^{x + y} = a^{x}a^{y}$ don't pose a major challenge.

About your inequalities let us say we have $a > b > 0$ and $x_{n}$ is a sequence of rationals tending to $x$. We need to show $a^{x} > b^{x}$. Clearly by the rules of rational exponents we have $a^{x_{n}} > b^{x_{n}}$ but taking limits as $n \to \infty$ weakens the inequality to $\geq $. So what we need are the following inequalities $$ra^{r - 1}(a - b) > a^{r} - b^{r} > rb^{r - 1}(a - b)$$ and $$sa^{s - 1}(a - b) < a^{s} - b^{s} < sb^{s - 1}(a - b)$$ where $r, s$ are rationals with $0 < s < 1 < r$. These are established in this answer.

Let $x > 1$ then we can suppose after a certain value of $n$, $x_{n} > 1$. Clearly then we have $$a^{x_{n}} - b^{x_{n}} > x_{n}b^{x_{n} - 1}(a - b)$$ Taking limits as $n \to \infty$ we get $$a^{x} - b^{x} \geq xb^{x - 1}(a - b) > 0$$ so that $a^{x} > b^{x}$. Similarly we can use other inequality (dealing with $s$) to handle the case when $0 < x < 1$. For negative values of $x$ the inequality is reversed i.e. $a^{x} < b^{x}$. This can be easily done via the rule that $a^{-x} = 1/a^{x}$.

  • Thanks loads for the input. But I think we've established that I need to make much more progress before I can grasp all of this. Like I said my knowledge of the limits of sequences is very minimal. I'm gonna keep studying and then come back to this soon.. Thanks again though. Much appreciated.. – Ishfaaq Feb 09 '14 at 05:15