While working on a separate problem, I stumbled upon a beautiful fact that I simply couldn't prove.. I suspect it requires some tools/tactics I've not encountered yet. The statement is the following: $$\sum_{k=0}^n\binom{n}{k}\cos\left(\frac{k\pi}{2} - \frac{n\pi}{4}\right) = \sqrt{2^n}$$
From my humble experience, I've got two suspicions for a successful approach:
- Complex number manipulation and working with the real part;
- Some kind of Fourier transform... this really just reminds me of it.
I couldn't come up with anything nice using complex numbers and I'm not that familiar with Fourier transforms yet, so I don't have the right intuition there.
Furthermore, this sum is just begging to employ Newton's binomial at some point, but I couldn't transform it to a form where it is applicable.
Any advice is welcome! Thanks in advance!
P. S. I don't opt for induction proofs, I'm rather interested in evaluating the sum on the LHS and arriving at the RHS.
You first idea is correct.
$$\begin{align} \sum_{k=0}^n\binom nk\cos\left(\frac{k\pi}2-\frac{n\pi}4\right)&= \sum_{k=0}^n\binom nk \Re\left(\exp\left(\frac{k\pi i}2-\frac{n\pi i}4\right)\right)\\ &=\Re\left(\left(e^{-n\pi i/4}\right)\sum_{k=0}^n\binom nk \left(e^{\pi i/2}\right)^k\right)\\ &=\Re\left(\left(e^{-n\pi i/4}\right)\sum_{k=0}^n\binom nk i^k\right)\\ &=\Re\left(\left(e^{-n\pi i/4}\right)(1+i)^n\right)\\ &=\Re\left(\left(e^{-n\pi i/4}\right)\left(\sqrt2e^{\pi i/4}\right)^n\right)\\ &=\sqrt{2^n} \end{align}$$