I seem to have the right idea here, but have a little trouble following through with it.
So here's the original problem:\[\binom{n}{3} + \binom{n}{7} + \binom{n}{11}+ \cdots \]
Now here's what I thought: I should plug something in which has its powers repeating periodically with a period of four. Oh hey, why not the fourth roots of unity?
So I decided to get on with it! Yeaaaah! The fourth roots of unity, as we know them, are \(1\), \(-1\), \(i\) and \(-i\). If we plug them in to the expression for \((1+x)^n\) and kinda mess around with the thing... maybe adding and subtracting... we'll get the answer.
I'll denote \(\binom{n}{r}\) as \(C_r\) for short.\[(1+i)^n = C_0 + iC_1 - C_2 - iC_3 + \cdots \]\[(1-i)^n =C_0 -iC_1 - C_2 + iC_3 + \cdots \]\[(1-1)^n = C_0 - C_1 + C_2 - C_3 + \cdots \]\[(1+1)^n = C_0 + C_1 + C_2 + C_3 + \cdots \]Kinda thinking of it in terms of constructive and destructive interference.
you're on the right track. this is basically the same thing you use when defining even and odd parts of a function \(f\) as follows: $$f_e(x)=\frac12(f(x)+f(-x))\\f_o(x)=\frac12(f(x)-f(-x))\\f(x)=f_e(x)+f_o(x)=\frac12(2\cdot f(x))=f(x)$$
except now instead of looking at mere even, odd terms, we're doing a more general discrete fourier transform to extract the terms corresponding to a different period; the generalization is from \((\pm1)^2=1\) to roots of unity as you noted \((e^{2\pi ik/n})^n=1\) for \(k\in\{0,1,2,\dots,n-1\}\)
Now\[\frac{(1-i)^n - (1+i)^n}2 = -iC_1 + i C_3 -iC_5 + \cdots \]And\[i\frac{(1+1)^n - (1-1)^n}{2}=iC_1 + iC_3 + iC_5 + \cdots \]I add both and it's done, I guess.
so just like we define \(\cosh x\) and \(\sinh x\) as the even and odd parts of \(e^x\): $$\cosh x=\frac12(e^x+e^{-x})\\\sinh x=\frac12(e^x-e^{-x})$$or, likewise, \(\cos x,\sin x\) as something *like* the even and odd parts of \(e^{ix}\): $$\cos x=\frac12(e^{ix}+e^{-ix})\\\sin x=\frac1{2i}(e^{ix}-e^{-ix})$$
Oh, yeah. That totally makes sense. Nice.
in this case, notice the binomial theorem tells us: $$(1+x)^n=\sum_{k=0}^n\binom{n}k x^k$$ if we use \(x=e^{2\pi i/4}\), so that \(x=i,x^2=-1,x^3=-i,x^4=i\) we see: $$(1+i)^n=\sum_{k=0}^n\binom{n}k i^{k}=\sum_{k=0}^{\lfloor n/2\rfloor}(-1)^k\binom{n}{2k}+i\sum_{k=0}^{\lfloor n/2\rfloor}(-1)^k\binom{n}{2k+1}$$and similarly $$(1-i)^n=\sum_{k=0}^n(-1)^k\binom{n}k i^k=\sum_{k=0}^{\lfloor n/2\rfloor}(-1)^{k}\binom{n}{2k}-i\sum_{k=0}^{\lfloor n/2\rfloor}(-1)^k\binom{n}{2k+1}\\(1+1)^n=\sum_{k=0}^n\binom{n}k\\(1-1)^n=\sum_{k=0}^n(-1)^k\binom{n}k$$so observe that \(\frac1{2i}\left[(1+i)^n-(1-i)^n\right]\) annihilates the even terms and gives alternating odd terms, while \(\frac12\left[(1+1)^n-(1-1)^n\right]\) annihilates even terms and gives non-alternating odd terms, and combining these to give: $$\frac14\left[{(1+1)^n+(1+i)^n-(1-1)^n-(1-i)^n}\right]=\sum_{k=0}^{\lfloor n/3\rfloor}\binom{n}{4k+3}$$
Oh boy, you didn't need to write all that.
oops, forgot some \(i\)s in the final expression, but that's besides the point :-p
I'm already done with the problem - and thank you for providing that analogy and insight. :)
$$\frac14\left[(1+1)^n-i(1+i)^n-(1-1)^n+i(1-i)^n\right]$$ and note for \(f(x)=(1+x)^n\) this is precisely $$\frac14\sum_{k=0}^4 f(e^{2\pi ik/4})\cdot e^{-2\pi ik/4}$$
Join our real-time social learning platform and learn together with your friends!