4

I am trying to construct properly the imaginary exponential as an extension of the real exponential. For this, I need to show that $\overline{e^{ix}}=e^{-ix}$.

I know that there is a very simple proof using an infinite sum of powers. But I am really looking for something purely algebraic using mainly if not only the properties of the real exponential extended to complex numbers.

One can easily show that $1=e^{ix}\cdot e^{-ix}$ and therefore that if $z=e^{ix}$, then $e^{-ix}=\lambda\cdot \overline z$ with $\lambda\in\mathbb R$, but how can we show that $\lambda=1$ ?

Once again, I am really looking for a proof that doesn't use any infinite sum.

Bernard
  • 175,478
  • You should start with definitions of the real exponential function and "proper construction" of an extension of real functions into the complex domain. – user Jan 24 '21 at 10:38
  • By proper construction, I mean extend the exponential function from $\mathbb R$ to $\mathbb C$ by keeping the usual properties of powers. – HerculePoivrot Jan 24 '21 at 11:33
  • Please specify what are the "usual" properties of powers. The proof might be very different depending on the axioms you choose. – Emanuele Paolini Jan 24 '21 at 12:07
  • Just the basics : $a^n\cdot a^m=a^{n+m}$, $(a^n)^m=a^{nm}$ – HerculePoivrot Jan 24 '21 at 12:20

4 Answers4

2

Just use Euler's formula: $$\overline{e^{ix}}=\overline{\cos x+i\sin x}=\cos x-i\sin x=\cos(-x)+i\sin(-x)=e^{-ix}.$$

Bernard
  • 175,478
1

How should you define $e^z$ for complex numbers so that it retains the good properties of the real function $e^x$? You take a definition of $e^x$ for real $x$ and modify it to apply to complex values. Then prove all the properties you want.

For each definition of the real case, you get a corresponding definition for the complex case.

One definition is the Taylor series. So for complex $z$, define $$ F(z) = \sum_{n=0}^\infty \frac{z^n}{n!},\quad z\in \mathbb C $$

Another definition is the differential equation. So for complex $z$, define $F(z)$ to be the unique solution $F : \mathbb C \to \mathbb C$ for the differential equation $$ F'(z)=F(z),\quad F(0)=1 $$

If you like another definition, try to use it...


Summary: for your original question, the answer will depend on which definition you used.

GEdgar
  • 111,679
  • I would like a 3rd definition if it makes sense : I want the only function that prolongs the usual properties of the real exponential ($a^n\cdot a^m=a^{m+n}$ and $a^n)^m=a^{mn}$) while being the same as the real exponential on $\mathbb R$. I guess the uniqueness needs to be proven though. – HerculePoivrot Jan 24 '21 at 12:23
1

Write $e^{ix} = a + bi$ where $a, b \in \mathbb{R}$. Clearly $\overline{e^{ix}} = a-bi$. Notice now the following: $$e^{-ix} = \frac{1}{e^{ix}} = \frac{1}{a+bi} = \frac{a-bi}{a^2 + b^2} = \frac{\overline{e^{ix}}}{a^2 + b^2}$$ Hence $(a^2 +b^2)e^{-ix} = \overline{e^{ix}}$. Since $a^2 + b^2 = 1$, the result follows.

Edit: the fact that $|e^{ix}| = a^2 + b^2 = 1$ can be proven without using infinite sums or Euler's formula, as is done here.

pytto
  • 11
  • "Since $a^2 + b^2 = e^{ix} e^{-ix} = 1$, the result follows". This part seems uncorrect to me, $a^2+b^2=e^{ix}\cdot \overline{e^{ix}}$ and you're back at the beginning. – HerculePoivrot Jan 24 '21 at 13:36
  • 1
    Yes indeed. In $a^2 + b^2 = e^{ix}e^{-ix} = 1$ you assume that $e^{ix}$ is conjugate of $e^{-ix}$, which is the property that is to be proven. – V.S.e.H. Jan 24 '21 at 13:41
  • @HerculePoivrot You are correct, I've updated the proof. – pytto Jan 24 '21 at 15:28
1

Let $P(f) = f(x + iy) - f(x-iy)$ and $H(f) = f(x+iy) + f(x-iy)$. These transformations are linear, i.e. $P(\alpha f + \beta g) = \alpha P(f) + \beta P(g)$, and $H(\alpha f + \beta g) = \alpha H(f) + \beta H(g)$.

Now, for $x$ real, we have: $$ \overline{e^x} = e^{\overline{x}}. $$ Applying $P$ on the LHS yields $$ P[\overline{e^x}] = \overline{P[e^x]} = \overline{e^{x+iy}} - \overline{e^{x-iy}}, $$ and similarly on the RHS $$ P[e^{\overline{x}}] = e^{\overline{x+iy}} - e^{\overline{x-iy}} = e^{x-iy} - e^{x+iy}. $$

Thus: $$ \overline{e^{x+iy}} - \overline{e^{x-iy}} = e^{x-iy} - e^{x+iy} \implies \overline{e^{x+iy}} + e^{x+iy} = \overline{e^{x-iy}} + e^{x-iy} \implies \Re{e^{x+iy}} = \Re{e^{x-iy}} $$

Applying $H$ on the LHS we get $$ H[\overline{e^x}] = \overline{H[e^x]} = \overline{e^{x+iy}} + \overline{e^{x-iy}}, $$ and similarly on the RHS $$ H[e^{\overline{x}}] = e^{\overline{x+iy}} + e^{\overline{x-iy}} = e^{x-iy} + e^{x+iy}, $$ so $$ \overline{e^{x+iy}} + \overline{e^{x-iy}} = e^{x-iy} + e^{x+iy}\implies e^{x+iy} - \overline{e^{x+iy}} = \overline{e^{x-iy}} - e^{x-iy} \implies \Im{e^{x-iy}} = -\Im{e^{x+iy}}. $$

Therefore for $z = x+iy$, $e^z$ is a conjugate of $e^{\overline{z}}$, i.e. $\overline{e^z} = e^{\overline{z}}$. The result $\overline{e^{ix}}=e^{-ix}$ follows trivially.

V.S.e.H.
  • 2,724
  • That seems nice. I'm just a bit perplex about the way you define $P$ and $H$. They seem to take a function as argument, then somehow $x$ and $y$ appear but you use it on a purely real function. Can you clarify that please ?

    But it will anyway be too hard for my students to understand. I was looking for a clean way to extend the exponants, a bit the like how you do it from $\mathbb N$ to $\mathbb Q$ and then $\mathbb R$, but it's probably not that trivial. I guess I will have to compromise and just tell them that $e^{ix}$ is a notation.

    – HerculePoivrot Jan 24 '21 at 15:25
  • Well, I first assume that $e^x$ has a complex extension. The transformations then just maps a one variable function to a two variable function. Steele suggested a similar 'polarization transformation' in his book 'The Cauchy-Schwarz Master Class'. – V.S.e.H. Jan 24 '21 at 15:41
  • Still, doesn't the exponential need to be holomorphic for $P[e^\overline x]=\overline{P[e^x]}$ to be true ? Maybe my question is silly, forgive me if so. I used to be better at maths but I am getting rusty with time – HerculePoivrot Jan 24 '21 at 15:53
  • @HerculePoivrot Your question is not silly at all, and is well justified! Can't say that I'm an expert at math either, and my answer is most certainly just an attempt to be verified by the experts. Is it not the case that the exponential is holomorphic, assuming that its complex extension exists? – V.S.e.H. Jan 24 '21 at 16:00
  • My complex analysis courses are far away, but doesn't any holomorphic function $f$ verifiy $f(\overline z)=\overline{f(z)}$. If so, the fact it's holomorphic is sufficient to demonstrate what I need, but on the other hand, this would heavily rely on the Taylor development. – HerculePoivrot Jan 24 '21 at 16:15
  • Indeed! I think there is no easy proof without a sufficient development of the theory, and a well defined extension. The idea of my proof was to try to show that, without knowing how this extension is defined, but assuming that it exists (which it does), then you can still prove the desired equality. Indeed also, assuming $f$ satisfies $f(\overline{x}) = \overline{f(x)}$ for $x$ real, then this same technique can be used to prove this for complex numbers. – V.S.e.H. Jan 24 '21 at 16:47
  • 1
    Yeah I guess you're probably right. Thanks for the help, will tell them that the Euler formula is a notation ;-) – HerculePoivrot Jan 24 '21 at 16:54