13
$\begingroup$

I'm wondering if there is a formula for the value of $(1+i)^n+(1-i)^n$?

I calculated the first terms starting with $n=1$ to be, in order, $2$, $0$, $-4$, $-8$, $-8$, $0$, $16$, $\dots$

So it seems to be some sequence of positive and negative powers of $2$ with $0$s thrown in. Is there a more explicit formulation of what $(1+i)^n+(1-i)^n$ is, based on $n$?

With the binomial theorem, I get it equal to $$ \sum_{k=0}^n\binom{n}{k}i^{n-k}(1+(-1)^{n-k}). $$ Can this be made nicer? Thanks.

$\endgroup$
  • 3
    $\begingroup$ Isn't that an exercise in Ahlfors? $\endgroup$ – lhf Jan 17 '12 at 16:46
  • $\begingroup$ Have you tried rewriting your formula using the binomial theorem? $\endgroup$ – Holdsworth88 Jan 17 '12 at 16:48
  • $\begingroup$ @lhf Yep, I checked it out from the suggestion here: math.stackexchange.com/questions/98885/…. It's the fourth part of the first exercise. $\endgroup$ – hagar Jan 17 '12 at 16:48
  • 1
    $\begingroup$ Remember to use $i^2=-1$... $\endgroup$ – lhf Jan 17 '12 at 16:51
  • $\begingroup$ Divide by $(\sqrt{2})^n$, note for example that $(1+i)/\sqrt{2}$ can be expressed in terms of trig functions. But there is an easier way! $\endgroup$ – André Nicolas Jan 17 '12 at 16:52
28
$\begingroup$

Let us avoid the binomial theorem.

  • Since $1-\mathrm i$ is the conjugate of $1+\mathrm i$, the number $\color{red}{x_n=(1+\mathrm i)^n+(1-\mathrm i)^n}$ is twice the real part of $(1+\mathrm i)^n$.

  • Since $\frac{1+\mathrm i}{\sqrt2}=\mathrm e^{\mathrm i\pi/4}$, $(1+\mathrm i)^n$ is $(\sqrt2)^n$ times $\mathrm e^{\mathrm in\pi/4}$.

  • The real part of $\mathrm e^{\mathrm in\pi/4}$ is $\cos(n\pi/4)$.

Thus, $\color{red}{x_n=\ldots}$

$\endgroup$
  • 3
    $\begingroup$ Thanks, so I think $2\Re((1+i)^n)=2(\sqrt{2}^n)\cos(\frac{n\pi}{4})$, which can be evaluated depending on $n\pmod{8}$? $\endgroup$ – hagar Jan 17 '12 at 16:59
  • $\begingroup$ Yes, depending on whether $n=0\pmod 8$ or $n=\pm1\pmod 8$ or $n=\pm2\pmod 8$ or $n=\pm3\pmod 8$ or $n=4\pmod 8$. $\endgroup$ – Did Jan 17 '12 at 17:02
  • 6
    $\begingroup$ This is a neat solution, that generalizes nicely : $a^n + \bar{a}^n = 2 \left|a\right|^n \cos( n \arg (a) )$. $\endgroup$ – Michael Anderson Jan 18 '12 at 7:25
18
$\begingroup$

First, notice that $1+i = \sqrt2(\cos\frac{\pi}{4} + i \sin\frac{\pi}{4}$) and $1-i = \sqrt2(\cos(-\frac{\pi}{4}) + i \sin(-\frac{\pi}{4}))$.

Using de Moivre's formula and the fact that $\sin$ is odd and $\cos$ is even, we get

\begin{align*} (1+i)^n + (1-i)^n &= 2(\sqrt2)^n\cos\frac{n\pi}{4} \end{align*}

$\endgroup$
5
$\begingroup$

$(1+i)^n+(1-i)^n = 2 \Re ((1+i)^n)$. Now expand $(1+i)^n$ using the binomial theorem. The real part is formed by the odd-numbered terms.

Another approach is to note that $(1+i)/\sqrt{2}$ is an 8-th root of unity and so $(1+i)^n$ depends only on $n \bmod 8$ except for a power of $\sqrt{2}$.

$\endgroup$
5
$\begingroup$

This is not a particularly elegant solution, but an alternative route is to simply note that $(1 \pm i)^2 = \pm 2i$. I will show a lot of steps, but this method involves only very easy calculations. If $n$ is even, then

$$ (1\pm i)^n = \left((1+i)^2\right)^{ \frac{n}{2}}= (\pm 2i)^{n/2} = \left\{ \begin{array}{ccl} 2^{n/2} & & n = 8k \\ \pm i \cdot 2^{n/2}& & n = 8k +2 \\ i \cdot 2^{n/2} & & n = 8k + 4 \\ \mp i 2^{n/2} & & n = 8k + 6 \end{array} \right. $$

Hence,

$$(1+i)^n + (1-i)^n = \left\{ \begin{array}{ccl} 2^{\frac{n}{2}+1} & & n = 8k \\ 0 & & n = 8k+2 \text{ or } n = 8k+6 \\ i \cdot 2^{\frac{n}{2}+1} & & n = 8k + 4 \end{array} \right. $$

This is all you need since if $n$ is odd, then $n-1$ is even. For example, when $n = 8k + 1$, we have $$\begin{align} (1+i)^n + (1 - i)^n &= (1+i)^{n-1}(1+i) + (1 -i)^{n-1}(1-i) \\ &= 2^{\frac{n-1}{2}}(1+i) + 2^{\frac{n-1}{2}} (1-i) \\ &= 2^{\frac{n-1}{2}} + i 2^{\frac{n-1}{2}} + 2 ^{\frac{n-1}{2}} - i 2^{\frac{n-1}{2}} \\ &= 2\cdot 2^{\frac{n-1}{2}} \\ &= 2^{\frac{n+1}{2}}. \end{align}$$

The cases $n = 8k+3$, $n=8k+5$, and $n = 8k + 7$ follow similarly.

$\endgroup$
4
$\begingroup$

Calculate!

$(1+i)^0=1$, so $(1-i)^0=1$, so $(1+i)^0+(1-i)^0=2$.

$(1+i)^1=1+i$, so $(1-i)^1=1-i$, so $(1+i)^1+(1-i)^1=2$ .

$(1+i)^2=2i$, so $(1-i)^2=-2i$, so $(1+i)^2+(1-i)^2=0$.

$(1+i)^3=-2+2i$, so $(1-i)^3=-2-2i$, so $(1+i)^3+(1-i)^3=-4$.

$(1+i)^4=-4$, so $(1-i)^4=-4$, so $(1+i)^4+(1-i)^4=-8$.

Now the game starts all over again. The pattern of the first four entries continues forever, except that every time $n$ is incremented by $4$, we multiply by $-4$, for the simple reason that $(1+i)^4=(1-i)^4=-4$.

Let $n=4k+r$, where $r=0$, $1$, $2$, or $3$. Then

$(1+i)^n+(1-i)^n=2(-4)^k=(-1)^k 2^{2k+1}$ if $\;r=0\;$ or $\;r=1$.

$(1+i)^n+(1-i)^n=0$ if $\;r=2$.

$(1+i)^n+(1-i)^n=(-4)^{k+1}=(-1)^{k+1}2^{2k+2}$ if $\;r=3$.

Comment: More briefly, since $(1+i)^4=(1-i)^4=-4$, we have $$(1+i)^{4k+r}=(-4)^k (1+i)^r\qquad\text{and}\qquad (1-i)^{4k+r}=(-4)^k (1-i)^r,$$ and therefore $$(1+i)^{4k+r}+(1-i)^{4k+r}=(-4)^k\left((1+i)^r+ (1-i)^r \right).$$ Note that the "cases" expression for $(1+i)^n+(1-i)^n$ can be made into a single expression in various ways.

$\endgroup$
4
$\begingroup$

Method somewhat similar to others already posted, but trying to be a little more rigorous while combining ideas from several approaches:

$$(1+i)^n + (1-i)^n$$

$$= \left(\sqrt{2} \exp{\frac{\imath\pi}{4}}\right)^n + \left(\sqrt{2} \exp{\frac{-\imath\pi}{4}}\right)^n$$

$$= 2^{n/2} \left(\exp{\frac{\imath \pi n}{4}} + \exp{ \frac{-\imath \pi n}{4}}\right)$$

$$= 2^{\frac{n+2}{2}} \left( \frac{\exp{\dfrac{\imath \pi n}{4}} + \exp{ \frac{-\imath \pi n}{4}}}{2} \right)$$

$$= 2^{\frac{n+2}{2}} \cosh\dfrac{i\pi n}{4}$$

$$= 2^{\frac{n+2}{2}} \cos\dfrac{\pi n}{4}$$

OP did request that the solution be made "nicer".

$\endgroup$
1
$\begingroup$

For variety, here's a sketch of another approach. Let $x_n = (1+i)^n + (1-i)^n$. Then,

$$\begin{align}x_n^2 &= \left( (1+i)^2 \right)^n + \left( (1-i)^2 \right)^n + 2 \left( (1+i)(1-i) \right)^n \\ &= (2i)^n + (-2i)^n + 2 \cdot 2^n \\ &= 2^n (2 + i^n + (-i)^n) \end{align} $$

at which point it's pretty easy to work out the pattern; the only thing left is to work out the sign on the non-zero terms: i.e. when is the real part of $(1+i)^n$ positive?

$\endgroup$

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.