6

In my analysis class, we covered the properties of the exponential function (as of now we use $\exp$ instead of $e$). One of the properties of $e$ is that $\exp(x_1+x_2)=\exp(x_1)\exp(x_2)$. In high school this was just assumed knowledge but now we have to prove that this statement is indeed true, which is proving to be quite difficult.

I assume that one would need to work out the derivatives of both side and see if they are equal. Then if $f(0)=1$ we can assume that both sides of the equation are true.

Henrik supports the community
  • 7,050
Ski Mask
  • 1,988
  • 13
    How do you define $\text{exp}(x)$? – Olivier Oloa Jan 05 '18 at 12:59
  • 1
    Show that $\displaystyle x \mapsto e^{x+y}-e^{x}e^{y}$ is constant. – Atmos Jan 05 '18 at 13:02
  • 6
    How to prove it depends strongly on the used definition of $\exp$. If that's defined via the power series, a straightforward multiplication of power series is the way to go. If it's defined as the solution to the initial value problem $y' = y,, y(0) = 1$, differentiating is the way to go. – Daniel Fischer Jan 05 '18 at 13:06
  • 3
    @DanielFischer it's the second one. So the second part of the equations's derivative would be found using the product rule and then it should equal the derivative of the left side. – Ski Mask Jan 05 '18 at 13:09
  • Right. Fix $x_1$ arbitrarily, and consider $g \colon x \mapsto \exp(x_1 + x)$ and $h \colon x \mapsto \exp(x_1)\cdot \exp(x)$. – Daniel Fischer Jan 05 '18 at 13:14
  • You can use an arbitrary base $b$, and $b^{x_1 + x_2}$ = $b^{x_1} * b^{x_2}$ just follows from the definition of the exponentiation. ($x$ multiplications of $b$)? – Bernd Strehl Jan 05 '18 at 13:18
  • I'm now interested in proof when exp is defined as power series and limit of (1+x/n)^n, that is as definitions 1 and 4 in https://en.wikipedia.org/wiki/Exponential_function#Formal_definition. But perhaps these can be posted as separate questions. @DanielFischer To me it is not trivial to multiply power series. – croraf Jan 05 '18 at 13:18
  • @Strernd This follows when x1 and x2 are whole numbers (and integers). But not when they are real numbers. – croraf Jan 05 '18 at 13:21
  • 2
    @croraf It's just the Cauchy product. And one uses $\frac{1}{k!n!} = \frac{1}{(k+n)!}\binom{k+n}{n}$ plus the binomial theorem. If one defines it as the limit of $(1+x/n)^n$, one needs to check that $$\biggl(1 + \frac{xy}{n^2\bigl(1 + \frac{x+y}{n}\bigr)}\biggr)^n \to 1,$$ which isn't too hard either. – Daniel Fischer Jan 05 '18 at 13:27
  • How would you work out the derivative of $\exp(x_1+x_2)$? Since $x_1$ and $x_2$ are unknown, would it just be $\exp(x_1+x_2)*(x_1+x_2)'$? – Ski Mask Jan 05 '18 at 13:28
  • @DanielFischer Thanks. I managed the binomial theorem way proof. And I got to your limit. But I don't know how to prove this limit. Any hint? – croraf Jan 05 '18 at 13:52
  • @croraf For large enough $n$, you can sandwich it between $\bigl(1 - \frac{2\lvert xy\rvert}{n^2}\bigr)^n$ and $\bigl(1 + \frac{2\lvert xy\rvert}{n^2}\bigr)^n$. – Daniel Fischer Jan 05 '18 at 13:56
  • And I think you meant "as of now we use $\exp(x) $ instead of $e^x$" and not $\exp$ instead of $e$. – Paramanand Singh Jan 06 '18 at 03:10

3 Answers3

7

From your comments it appears that you are using the definition of $\exp(x) $ as the unique solution to $f'(x) =f(x), f(0)=1$. It can be proved that such a solution must be non-zero for all $x$ (see second part of this answer). Thus $\exp(x) \neq 0$ for all $x$. Let $a$ be any arbitrary real number and consider the function $g$ defined by $g(x) =\exp(x+a) /\exp(x) $. We have $$g'(x) =\frac{\exp (x) \exp(x+a) - \exp(x+a) \exp(x) } {(\exp(x)) ^2}=0$$ and thus $g$ is a constant. It follows that $g(x) =g(0)=\exp(a)$ and thus $$\exp(x+a) =\exp(x) \exp(a) $$ Now replace $x$ by $x_1$ and $a$ by $x_2$.

The above technique does not work if one chooses $$g(x) =\exp(x+a) - \exp(x) \exp(a) $$ because one can't see that $g'(x) =0$ in very obvious manner. But this can be done with some effort. Using above definition of $g$ one gets $g'(x) =g(x),g(0)=0$. Ideally when one studies the definition of exponential function as a solution of differential equation $f'(x) =f(x), f(0)=1$ then the first step is to show that if the solution exists then it must be unique. And that uniqueness is shown by the following:

Theorem: If function $g:\mathbb{R} \to\mathbb{R} $ satisfies $g'(x) =g(x), g(0)=0$ then $g(x) =0$ for all values of $x$.

On the contrary assume that there is some $a$ such that $g(a) \neq 0$ and consider $$h(x) =g(a+x) g(a-x) $$ then $$h'(x) =g(a+x) g(a-x) - g(a+x) g(a-x) =0$$ so that $h$ is constant and $h(x) =h(0)=g(a)g(a)>0$ or in other words $g(a+x) g(a-x) >0$. Putting $x=a$ and noting that $g(0)=0$ we see that $0>0$ and this contradiction shows that $g(x) =0$ for all $x$ and we are done. The same proof can be used to show that $\exp(x) \neq 0$. We just need to consider $h(x) =\exp(x) \exp(-x) $.

Paramanand Singh
  • 92,128
5

if you are familiar with the Cauchy-product $CP$ you can proof this like follows:

$$exp(z) \cdot exp(w)\sum_{n=0}^{\infty} \frac{z^n}{n!}\sum_{n=0}^{\infty} \frac{w^n}{n!} \overset{CP}{=}\\ \sum_{n=0}^{\infty} \left(\sum_{k=0}^{\infty} \frac{z^k}{k!} \frac{w^{n-k}}{(n-k)!} \right)=\sum_{n=0}^{\infty}\left(\frac{1}{n!}\sum_{k=0}^{\infty} \frac{n!}{k!(n-k)!}z^kw^{n-k} \right)= \\ \sum_{n=0}^{\infty}\left(\frac{1}{n!}\sum_{k=0}^{\infty}\binom{n}{k} z^k w^{n-k} \right) = \sum_{n=0}^{\infty} \frac{1}{n!}(z+w)^n = exp(z+w)$$

McBotto.t
  • 462
0

This is one of the basic properties of an exponential function. Assuming that you know the basic connection between it and natural logarithm, this is that $ e^x $ is the inverse of $ \ln{x} $. With that being said, we can write the following

$$ \ln{e^{x+y}} = x + y = \ln{e^x} + \ln{e^y} = \ln{e^x \cdot e^y}$$

Ultimately, since natural logarithm and exponential function are fundamentally related (trivially speaking one-to-one), we conclude that

$$ e^{x+y} = e^x \cdot e^y $$

Be aware that such proof does not use derivatives. I hope that helps.

Pero Alex
  • 49
  • 2
    And how do you prove $\ln a+\ln b=\ln a\cdot b$, that you use for the last step? At the level the OP describes it probably safe to assume that one should not use a property of the natural logarithm that is basically equivalent to what is asked, but that the definition should be used. – Henrik supports the community Jan 05 '18 at 13:21
  • I can only now see the problem in my answer, thank you for pointing it out. I suggest this post. – Pero Alex Jan 05 '18 at 13:55