## Exponent Combination Laws

For the discussion of math. Duh.

Moderators: gmalivuk, Moderators General, Prelates

### Exponent Combination Laws

Hello everyone.

I'm investigating the different definitions of the exponential function exp x.

I noticed that all the proofs I've seen for exponent combination laws or logarithm laws for irrational indices/bases assumes the definition:

\exp y = x \iff \ln x = y, \ln x = \int_1^x \frac {\mathrm {d}t} t

How does one derive the exponent combination laws or log laws for other definitions of exp x, particularly for exp x = e^x?
gfauxpas

Posts: 97
Joined: Sun Oct 09, 2011 11:24 pm UTC

### Re: Exponent Combination Laws

You can define exp through its power series. Most identities can be easily proven from this definition.
korona

Posts: 116
Joined: Sun Jul 04, 2010 8:40 pm UTC

### Re: Exponent Combination Laws

The limit definition of e can give you both a definition for e, and a direct definition of ex.
One of the painful things about our time is that those who feel certainty are stupid, and those with any imagination and understanding are filled with doubt and indecision - BR

Last edited by JHVH on Fri Oct 23, 4004 BCE 6:17 pm, edited 6 times in total.

Yakk

Posts: 10039
Joined: Sat Jan 27, 2007 7:27 pm UTC
Location: E pur si muove

### Re: Exponent Combination Laws

You usually just show that all your definitions are equivalent and then you can prove the properties you need using any definition you like.
legend

Posts: 42
Joined: Thu Feb 02, 2012 5:42 pm UTC

### Re: Exponent Combination Laws

Korona: can you elaborate on that a bit more, please?

And Yakk, the only way I know how to use the limit definition to go to e^x is to say:

\exp x = \lim_{n \to \infty}(1 + x/n)^n = \lim_{(n/x) \to \infty}((1 + 1/(n/x))^{(n/x)})^x = e^x

but that assumes exponent combination laws already.

Legend, I'm working on showing all the definitions are equivalent, but e^x seems isolated from all others, and I can't get the limit definition to imply any others.
Last edited by gfauxpas on Mon Feb 06, 2012 9:52 pm UTC, edited 1 time in total.
gfauxpas

Posts: 97
Joined: Sun Oct 09, 2011 11:24 pm UTC

### Re: Exponent Combination Laws

You might be interested this:
http://en.wikipedia.org/wiki/Characterizations_of_the_exponential_function
Also sorry, but I don't know what you mean by e^x. Usually for real numbers you define a^b:=exp(ln(a)*b). So using this to define the exp function gives you a circular definition.
legend

Posts: 42
Joined: Thu Feb 02, 2012 5:42 pm UTC

### Re: Exponent Combination Laws

Ok, so in my experience, we normally derive e, and the exponential function by considering power series. In particular we define them and then a radius of convergence, and then we note that this particular power series happens to be everywhere convergent. It is a standard theorem in analysis the power series can be differentiated termwise, which gives us the standard identity d(e^x)/dx = e^x
From this we get e = exp(1) (which happily we know converges). Getting our standard laws are a little work in this, in particular we want
\Sigma \frac{(x+y)^n}{n!} = \Sigma \frac{(x)^n}{n!} \Sigma \frac{(y)^n}{n!}
There are a few ways of proving this (including just writing it out) but in my opinion the most slick way (so long as you have a couple of differentiation rules) is as follows:

Let F(x) = exp(a + b - x) * exp(x) , then F'(x) = exp(a + b - x)*exp(x) - exp(a + b - x)exp(x) = 0,
and so F(x) is constant.
But F(0) = exp(a + b), and F(b) = exp(a)*exp(b)
Done

The line where we state that derivative is 0 implies constant is non-trivial, but doesn't require too much in the way of tools.

I especially like this characterisation, as it follows over into the complex plane quite simply.
G4!!

Grob FTW,

Hello. Smithers. You're. Quite good. At. Turning. Me. On.

OverBored

Posts: 284
Joined: Mon Dec 10, 2007 7:39 pm UTC

### Re: Exponent Combination Laws

legend wrote:You might be interested this:
http://en.wikipedia.org/wiki/Characterizations_of_the_exponential_function
Also sorry, but I don't know what you mean by e^x. Usually for real numbers you define a^b:=exp(ln(a)*b). So using this to define the exp function gives you a circular definition.

Okay, let me try to explain.

e^x for natural number indices: e multiplied by itself x times. Product of no numbers is defined as 1.
Negative integers: e^x = 1/e^(-x)
rational numbers: e^(p/n) = (the nth root of e)^p = the nth root of (e^p)
irrational numbers: .... ??

Is there a way to continue this pattern using, I dunno, limits? sequences?
gfauxpas

Posts: 97
Joined: Sun Oct 09, 2011 11:24 pm UTC

### Re: Exponent Combination Laws

Yes. Non-rational numbers can be represented as a cauchy sequence of rationals. To generate the cauchy sequence that represents f(x), where f is defined on the rationals, and where x={a_i} is a cauchy sequence, simply calculate f(a_i), possibly dropping some "head" of the cauchy sequence for which f() is not always defined.

This works for defining e^x for x non-rational reasonably well. (Note that it requires certain things of f that I didn't mention -- continuity will do, I think.)
One of the painful things about our time is that those who feel certainty are stupid, and those with any imagination and understanding are filled with doubt and indecision - BR

Last edited by JHVH on Fri Oct 23, 4004 BCE 6:17 pm, edited 6 times in total.

Yakk

Posts: 10039
Joined: Sat Jan 27, 2007 7:27 pm UTC
Location: E pur si muove

### Re: Exponent Combination Laws

Yakk I don't know enough analysis to prove continuity like that, but thanks for that insight. I imagine it's straightforward to prove exponent combination laws for integer indices, but can you prove it for rational indices without recourse to the existence of the logarithm? Would then taking the limit of the sequences prove that the properties hold for all real indices?
gfauxpas

Posts: 97
Joined: Sun Oct 09, 2011 11:24 pm UTC

### Re: Exponent Combination Laws

e^x is continuous so if x is a real number, (x_1, x_2, x_3, ...) is a rational sequence converging to x then (e^x_1, e^x_2, e^x_3, ...) converges to e^x.
But if you define e^x that way then you have to show that the sequence converges and does not depend on the choice of the x_i.
EDIT: @yakk You cannot say "that holds if f is continuous" as you cannot really define continuity for a function defined on the rationals. In general not every function from Q can be extended to R that way.
Last edited by korona on Mon Feb 06, 2012 9:49 pm UTC, edited 1 time in total.
korona

Posts: 116
Joined: Sun Jul 04, 2010 8:40 pm UTC

### Re: Exponent Combination Laws

korona wrote:e^x is continuous so if x is a real number, (x_1, x_2, x_3, ...) is a rational sequence converging to x then (e^x_1, e^x_2, e^x_3, ...) converges to e^x.
But if you define e^x that way then you have to show that the sequence converges and does not depend on the choice of the x_i.

I hear what you're saying, thanks.

I see how you can jump from most of the definitions to the other, using the FToC or theorems for the existence of power series or whatever. But the definition as the limit of a sequence (1 + x/n)^n seems to be in the middle of nowhere. How does that one imply any other definition?
gfauxpas

Posts: 97
Joined: Sun Oct 09, 2011 11:24 pm UTC

### Re: Exponent Combination Laws

gfauxpas wrote:Okay, let me try to explain.

e^x for natural number indices: e multiplied by itself x times. Product of no numbers is defined as 1.
Negative integers: e^x = 1/e^(-x)
rational numbers: e^(p/n) = (the nth root of e)^p = the nth root of (e^p)
irrational numbers: .... ??

Is there a way to continue this pattern using, I dunno, limits? sequences?

Ok, I see where this is going. But you're defining e as a limes and your real number x as a limes of rationals. You're finally ending up with a really ugly definition and I really don't see what it would be good for.
The usual approach is to show that exp(ln(a)*b)=a^b holds for rational b and use this to define a^b for rationals.
legend

Posts: 42
Joined: Thu Feb 02, 2012 5:42 pm UTC

### Re: Exponent Combination Laws

korona wrote:EDIT: @yakk You cannot say "that holds if f is continuous" as you cannot really define continuity for a function defined on the rationals. In general not every function from Q can be extended to R that way.

Actually, you can define continuity on the rational numbers, because the rationals are dense, from which it follows that between any two rational numbers there is another rational number. All you have to do is tweak the epsilon delta definition slightly so that the only x values allowed are rational numbers.
I have discovered a truly marvelous proof of this, which this margin is too narrow to contain.
tomtom2357

Posts: 429
Joined: Tue Jul 27, 2010 8:48 am UTC

### Re: Exponent Combination Laws

gfauxpas wrote:I see how you can jump from most of the definitions to the other, using the FToC or theorems for the existence of power series or whatever. But the definition as the limit of a sequence (1 + x/n)^n seems to be in the middle of nowhere. How does that one imply any other definition?

You can do what Newton himself did: use binomial expansion on (1 + x/n)^n, which will give you the power series. It's not exactly rigorous, but hey.

tomtom2357 wrote:Actually, you can define continuity on the rational numbers, because the rationals are dense, from which it follows that between any two rational numbers there is another rational number. All you have to do is tweak the epsilon delta definition slightly so that the only x values allowed are rational numbers.

* kicks tomtom2357 *

PM 2Ring

Posts: 2591
Joined: Mon Jan 26, 2009 3:19 pm UTC
Location: Mid north coast, NSW, Australia

### Re: Exponent Combination Laws

Let f: Q -> Q; f(x) = 0 if x^2 < 2 and f(x) = 1 otherwise. This function is continuous in Q as the pre-image of every open set is open:
The sets f^-1( (-epsilon, epsilon) ) = ( (-sqrt(2), sqrt(2)) intersected with Q)
and f^-1( (1 - epsilon, 1 + epsilon) ) = ( (-infinity, -sqrt(2)) union (sqrt(2), infinity) intersected with Q)
are open in Q and form a partition of Q. But clearly this function cannot be extended to R.
EDIT: Uniform continuity would be sufficient but e^x is not uniform continuous on Q.
Last edited by korona on Tue Feb 07, 2012 10:27 am UTC, edited 1 time in total.
korona

Posts: 116
Joined: Sun Jul 04, 2010 8:40 pm UTC

### Re: Exponent Combination Laws

PM 2Ring wrote:
gfauxpas wrote:I see how you can jump from most of the definitions to the other, using the FToC or theorems for the existence of power series or whatever. But the definition as the limit of a sequence (1 + x/n)^n seems to be in the middle of nowhere. How does that one imply any other definition?

You can do what Newton himself did: use binomial expansion on (1 + x/n)^n, which will give you the power series.

Another approach: let f(x)= lim (1 + x/n)^n. Prove convergence for all x (don't have a simple argument for that atm, sorry). Note that f(a+b)<=f(a)f(b)<=f(a+b+e) for all positive e, and conclude that f(a+b) = f(a)f(b). You also have monotonicity (perhaps prove this after convergence) and expanding the brackets will give you a value for f(1). That there is only one function that satisfies these properties is the final result you need.
ameretrifle wrote:Magic space feudalism is therefore a viable idea.

jestingrabbit

Posts: 5187
Joined: Tue Nov 28, 2006 9:50 pm UTC
Location: Sydney

### Re: Exponent Combination Laws

gfauxpas wrote:I see how you can jump from most of the definitions to the other, using the FToC or theorems for the existence of power series or whatever. But the definition as the limit of a sequence (1 + x/n)^n seems to be in the middle of nowhere. How does that one imply any other definition?

If you already know some rules about exp, then exp(x+1/n)=c*exp(x) and therefore exp(0+n/n)=c^(n)*exp(0) => exp(1)=c^n
But the taylor expansion of the first equation gives you c=(1+1/n+O(n^2))... so after you found out that you can neglect the higher orders in the limit, you have exp(1)=lim((1+1/n)^n,n->inf). It is possible to do the same with an additional x, just substitute stuff.
mfb

Posts: 803
Joined: Thu Jan 08, 2009 7:48 pm UTC

### Re: Exponent Combination Laws

Sorry, why does \exp (x + \frac 1 n) = c \exp x?

PM 2Ring wrote:You can do what Newton himself did: use binomial expansion on (1 + x/n)^n, which will give you the power series. It's not exactly rigorous, but hey.

Oh whoa the binomial theorem works for all real powers? Cool, I'll look into that
gfauxpas

Posts: 97
Joined: Sun Oct 09, 2011 11:24 pm UTC

### Re: Exponent Combination Laws

gfauxpas wrote:
PM 2Ring wrote:You can do what Newton himself did: use binomial expansion on (1 + x/n)^n, which will give you the power series. It's not exactly rigorous, but hey.

Oh whoa the binomial theorem works for all real powers? Cool, I'll look into that
\
It does, but you can evaluate the limit using any sequence that goes to infinity, in particular with just integers. (This doesn't prove the limit exists, but proves that if it does, then it's equal to what you get)
addams wrote:This forum has some very well educated people typing away in loops with Sourmilk. He is a lucky Sourmilk.
mike-l

Posts: 2513
Joined: Tue Sep 04, 2007 2:16 am UTC

### Re: Exponent Combination Laws

mike-l wrote:It does, but you can evaluate the limit using any sequence that goes to infinity, in particular with just integers. (This doesn't prove the limit exists, but proves that if it does, then it's equal to what you get)

Oh, right.

I can prove the limit exists if I'm allowed to use logarithm laws. But, um, yeah.
gfauxpas

Posts: 97
Joined: Sun Oct 09, 2011 11:24 pm UTC

### Re: Exponent Combination Laws

gfauxpas wrote:I can prove the limit exists if I'm allowed to use logarithm laws. But, um, yeah.

Well, I trust that you can show that the limit is monotonically increasing for positive x, so all you need is an upper bound, so I would suggest that that is where to focus your energy. In particular, you know that 3^x is an upper bound, or 4^x, or those after you round x up etc etc. I suspect that if you pick a large enough upper bound for the sequence the proof will be doable.

Note that you might need to assume the rule is true for integer powers of integers, but that's not a real problem imo. To some extent, this result should come out of that other one.
ameretrifle wrote:Magic space feudalism is therefore a viable idea.

jestingrabbit

Posts: 5187
Joined: Tue Nov 28, 2006 9:50 pm UTC
Location: Sydney

### Re: Exponent Combination Laws

jestingrabbit wrote:
gfauxpas wrote:I can prove the limit exists if I'm allowed to use logarithm laws. But, um, yeah.

Well, I trust that you can show that the limit is monotonically increasing for positive x, so all you need is an upper bound, so I would suggest that that is where to focus your energy. In particular, you know that 3^x is an upper bound, or 4^x, or those after you round x up etc etc. I suspect that if you pick a large enough upper bound for the sequence the proof will be doable.

Note that you might need to assume the rule is true for integer powers of integers, but that's not a real problem imo. To some extent, this result should come out of that other one.

I'm willing to assume the power rules for integer powers even not of integer bases. Though I don't know the proofs I'm quite sure they're not Calculus proofs.

I'd have to think about how to prove something is an upper bound, I've never done that before for a case similar to this one. Can I have a hint?
gfauxpas

Posts: 97
Joined: Sun Oct 09, 2011 11:24 pm UTC

### Re: Exponent Combination Laws

gfauxpas wrote:Sorry, why does \exp (x + \frac 1 n) = c \exp x?

Follows directly from exp(a+b)=exp(a)*exp(b) with a=x and b=1/n. Define c as exp(1/n) and look at its series expansion.
mfb

Posts: 803
Joined: Thu Jan 08, 2009 7:48 pm UTC

### Re: Exponent Combination Laws

Oh, okay. But I need to know exponent combination laws for that, which goes back to my original problem
gfauxpas

Posts: 97
Joined: Sun Oct 09, 2011 11:24 pm UTC

### Re: Exponent Combination Laws

Well, you just need a good order in your proofs.

IIRC, my prof introduced it like this:
ln(x) gets defined as an integral over 1/x (->derivative obvious, power series calculatable)
exp(x) is the inverse function of ln(x) (derivative can be calculated -> power series follows -> convergence radius -> it is equivalent to power series definition of exp(x))
e can be defined as exp(1) which is 1/0!+1/1!+1/2!+...
ln(a*b)=ln(a)+ln(b) can be derived from the integral definition of ln(x), this can be translated to exp(a+b)=exp(a)*exp(b)
Now there might be some magic involved, but I think you can extend stuff like exp(ln(a)+ln(a))=a*a=a^2 to identify exp(ln(a)*b) with a^b for rational b (which was defined earlier using multiplication and integer roots). Define a^c for irrational c as the limit of rational b->c, proof that it is well-defined. This also justifies the notation exp(x)=e^x.
mfb

Posts: 803
Joined: Thu Jan 08, 2009 7:48 pm UTC

### Re: Exponent Combination Laws

OverBored wrote:Ok, so in my experience, we normally derive e, and the exponential function by considering power series. In particular we define them and then a radius of convergence, and then we note that this particular power series happens to be everywhere convergent. It is a standard theorem in analysis the power series can be differentiated termwise, which gives us the standard identity d(e^x)/dx = e^x
From this we get e = exp(1) (which happily we know converges). Getting our standard laws are a little work in this, in particular we want
\Sigma \frac{(x+y)^n}{n!} = \Sigma \frac{(x)^n}{n!} \Sigma \frac{(y)^n}{n!}
There are a few ways of proving this (including just writing it out) but in my opinion the most slick way (so long as you have a couple of differentiation rules) is as follows:

Let F(x) = exp(a + b - x) * exp(x) , then F'(x) = exp(a + b - x)*exp(x) - exp(a + b - x)exp(x) = 0,
and so F(x) is constant.
But F(0) = exp(a + b), and F(b) = exp(a)*exp(b)
Done

The line where we state that derivative is 0 implies constant is non-trivial, but doesn't require too much in the way of tools.

I especially like this characterisation, as it follows over into the complex plane quite simply.

What do you mean by writing it out? I tried playing with various expansions but couldn't get any to work.

Your second suggestion assumes the knowledge that D_x exp x = exp x though, and I'm pretending not to know that yet.
gfauxpas

Posts: 97
Joined: Sun Oct 09, 2011 11:24 pm UTC

### Re: Exponent Combination Laws

Okay, I will write it out explicitly, but I am not sure how useful it is really.

Spoiler:
e^x e^y =(\sum_0^\infty {\frac{x^n}{n!}})(\sum_0^\infty {\frac{y^n}{n!}}) = (1 + x + \frac{x^2}{2!} + \frac{x^3}{3!} + \cdots)(1 + y + \frac{y^2}{2!} + \frac{y^3}{3!} + \cdots)

We find the constant term, then the term linear in x and y, quadratic in x and y, etc
= 1 + x + y + \frac{x^2}{2!} + xy + \frac{y^2}{2!} + \frac{x^3}{3!} + \frac{x^2 y}{2!} + \frac{x y^2}{2!} + \frac{y^3}{3!} \cdots = 1 + (x+y) + \frac{(x+y)^2}{2!} + \frac{(x+y)^3}{3!} \cdots

or a little more carefully
(\sum_0^\infty {\frac{x^n}{n!}})(\sum_0^\infty {\frac{y^n}{n!}}) = \sum_{n=0}^\infty{\sum_{k=0}^n{\frac{x^k y^{n-k}}{k! (n-k)!}}}= \sum_{n=0}^\infty{\frac{(x+y)^n}{n!}}

The inner sum is just the binomial coefficients, and we get them just from adding the terms.

The other method does assume the derivative property, yes. If you don't want to, that is fair enough, but proving it can be useful. The way I was shown it was to prove that power series can be differentiated termwise (which actually looking back was a fairly long proof) and then just applying this to our definition of exp(x). It isn't the most elegant way to go about defining exp and e, but once you have done the work it gives you a nice solid definition, and quite a few properties.
G4!!

Grob FTW,

Hello. Smithers. You're. Quite good. At. Turning. Me. On.

OverBored

Posts: 284
Joined: Mon Dec 10, 2007 7:39 pm UTC

### Re: Exponent Combination Laws

Cool, thanks for showing me that.

I'm thinking that, using the limit definition of exp, if I can show:

\left({1 + \frac{x + y}{n}} + \frac {xy}{n^2}\right)^n \to \left({1 + \frac{x + y}{n}}\right)^n as n \to +\infty and similar things, I can prove the exponent combination laws that way. My friend is examining this with me to see if it's feasible, feel free to chime in.

I appreciate everyone's help btw.

edit: oh, the reason I'm trying to avoid that knowledge is because f = Df, f(0) = 1 is one of the definitions, I'm trying to keep the definitions seperate.
gfauxpas

Posts: 97
Joined: Sun Oct 09, 2011 11:24 pm UTC

### Re: Exponent Combination Laws

\left({1 + \frac{x + y}{n}} + \frac {xy}{n^2}\right)^n = 1 + (x+y) + \frac{n-1}{n}(x+y) + ... + \frac {xy}{n} + c \cdot \frac{(x+y)^2}{n} + O(1/n^2)
Well, write it down in a strict mathematical way, and you get the usual exponential power series with (x+y).
mfb

Posts: 803
Joined: Thu Jan 08, 2009 7:48 pm UTC

### Re: Exponent Combination Laws

I'm bull-headed so I insisted on proving one law directly from the limit of a series definition (with a lot of help from an online friend). It's here if anyone wants to see it, it's proof 2.

edit: on further investigation, that is the hardest one to prove, I think.
gfauxpas

Posts: 97
Joined: Sun Oct 09, 2011 11:24 pm UTC