[guided]We now show that the system $(X, T)$ constructed above is not minimal, and then combine both examples to conclude.
To prove non-minimality, we exhibit a non-empty proper closed $T$-invariant subset of $X$. Take the singleton
\begin{align*}
M := \{ 0 \} \subseteq X.
\end{align*}
We verify each of the four required properties:
- *Non-empty*: $0 \in M$.
- *Closed*: $X$ is a metric space and hence Hausdorff, so every singleton is closed.
- *Proper*: $1 \in X$ but $1 \notin M$, so $M \subsetneq X$.
- *$T$-invariant*: $T(0) = 0 \in M$, so $T(M) = \{0\} \subseteq M$.
We now invoke the standard equivalence in topological dynamics: a topological dynamical system $(Y, S)$ on a non-empty compact Hausdorff space $Y$ is minimal if and only if $Y$ contains no non-empty proper closed $S$-invariant subset. The argument is short. Suppose first that every $S$-orbit is dense. If $F \subseteq Y$ is a non-empty closed $S$-invariant subset, pick any $p \in F$; then $\{S^n p : n \geq 0\} \subseteq F$ by invariance, and taking closures gives $Y = \overline{\{S^n p : n \geq 0\}} \subseteq F$, so $F = Y$, meaning $F$ is not proper. Conversely, suppose every non-empty closed $S$-invariant subset of $Y$ equals $Y$. For any $p \in Y$, the orbit closure $\overline{\{S^n p : n \geq 0\}}$ is non-empty, closed, and $S$-invariant, so it equals $Y$; that is, the orbit of $p$ is dense, so $(Y, S)$ is minimal. A textbook reference is P. Walters, *An Introduction to Ergodic Theory*, Graduate Texts in Mathematics **79** (Springer, 1982), Theorem 5.9.
Since $M = \{0\}$ is a non-empty proper closed $T$-invariant subset of $X$, the system $(X, T)$ is not minimal. Combined with the previous step, where we proved that $(X, T)$ is uniquely ergodic, this gives a uniquely ergodic topological dynamical system that is not minimal, proving part (ii) of the theorem.
Finally, we put the two halves of the proof together. Part (i) was established in the first step by citing Furstenberg's skew-product homeomorphism $(\mathbb{T}^2, S)$, which is minimal but admits two distinct invariant Borel probability measures and hence is not uniquely ergodic. Part (ii) has just been established with the explicit system $(X, T)$, which is uniquely ergodic with unique invariant measure $\delta_0$ but admits the non-empty proper closed invariant subset $\{0\}$ and hence is not minimal. Together these examples show that, in the category of topological dynamical systems on non-empty compact metrizable spaces with continuous dynamics, neither of the two properties — minimality and unique ergodicity — implies the other. This is precisely the statement that minimality and unique ergodicity are logically independent, and the proof is complete.[/guided]