In the excellent paper "Step-by-Step HHL Algorithm Walkthough..., https://arxiv.org/abs/2108.09004, it makes a derivation that is not clear to me at all.
In Equation (15) it combines the effect of phase estimation and the inverse QFT into this close form: $$ \frac{1}{2^n} |b\rangle \sum_{y=0}^{2^n-1} \sum_{k=0}^{2^n-1} e^{2 \pi i k(\phi - y/N)} |y\rangle |0\rangle $$ So far so good, this just combines the closed form of QPE with $QFT^{-1}$. Then it says that only $|y\rangle$ for which is $\phi - y/N = 0$ will have an amplitude, all other amplitudes are 0. I know this is correct, because that's how it works during QPE, but I have difficulties deriving this result in this closed form.
In the next step (Equation 16), it further arrives at $|b\rangle |N \phi\rangle |0\rangle$, which is even more unclear to me. I'm not sure I understand what $|N \phi\rangle$ is supposed to mean here?
Any help with a precise derivation (actually step-by-step ;-) would be greatly appreciated.
