1: \begin{proof}

2: %An inductive argument using successive applications of the identity $(I-AB)^{-1}A = A(I-BA)^{-1}$ proves this lemma. Instead of presenting such rigorous proof, we sketch the conceptual idea. Let the output of the lifted system obtained by sampling the state at timeslot 0 be denoted by $\widetilde{y}^0$ and that obtained by sampling the state at some other timeslot, say $k$, be denoted by $\widetilde{y}^k$. We note that when $t \rightarrow \infty$, the transients die out and $y(t)$ is an N-periodic signal, independent of the initial time $t_0$. Also note that $\widetilde{y}^0(t)$ and $\widetilde{y}^k(t)$ are both vectorized versions of $y(t)$ but with a different permutation. Therefore, $\widetilde{y}^0(t)$ is simply a reordering of $\widetilde{y}^k(t)$ as $t \rightarrow \infty$. This implies that the inequalities defined in \eqref{eq:tightened} would be a re-ordering of the inequalities for a similar set obtained using $\widetilde{y}^k$. This does not change the geometric nature of $\Omega_{ss}$.

3: Let $\widetilde{y}$ and $\widetilde{z}$ be the lifted outputs with the state sampled at timeslot 0 and timeslot $k$, $1\leq k\leq N-1$:

4: $$

5: \widetilde{y} = [y(tN)^\top, \ldots, \, y(tN + N-1)^\top]^\top 

6: $$

7: $$

8: \widetilde{z} = [y(tN+k)^\top, \ldots, \, y(tN + k + N-1)^\top]^\top. 

9: $$

10: Lyapunov stability implies that $\widetilde{y}(t+1)=\widetilde{y}(t)$ as $t \rightarrow \infty$. This observation, together with the definition of $\widetilde{y}$, implies that $y(t)$ is $N$-periodic. Therefore, $\widetilde{z}(t)$ must be a permutation of $\widetilde{y}(t)$. This suggests that lifting the system by sampling the state at a different timeslot leads to a reshuffling of the inequalities in \eqref{eq:tightened}, which does not change the set. 

11: \end{proof}

12: