The Kolmogorov-Sinai (KS) entropy $h_\mu(T)$ measures the average information generated per iteration of $T$. For a finite partition $\mathcal{P} = \{P_1, \ldots, P_k\}$ of $X$:

$H(\mathcal{P}) = -\sum_{i=1}^k \mu(P_i) \log \mu(P_i)$

is the Shannon entropy. The KS entropy is:

$h_\mu(T) = \sup_{\mathcal{P}} \lim_{n \to \infty} \frac{1}{n} H\left(\bigvee_{j=0}^{n-1} T^{-j}\mathcal{P}\right)$

where $\bigvee$ denotes the refinement (common refinement of partitions). For ergodic systems, this limit exists and equals the supremum over all partitions.

For a measure-preserving transformation $(X, \mu, T)$, define the Koopman operator $U_T: L^2(X, \mu) \to L^2(X, \mu)$ by:

$U_T f = f \circ T$

This is a unitary operator (since $T$ preserves measure). The spectrum of $U_T$ characterizes statistical properties:

• Discrete spectrum: Pure point spectrum indicates quasi-periodic dynamics
• Continuous spectrum: Associated with mixing
• Lebesgue spectrum: Corresponds to Bernoulli systems (strongest mixing)

Spectral analysis provides qualitative and quantitative characterization of ergodic systems.

A system is $k$-mixing if for any sets $A_0, A_1, \ldots, A_k$:

$\lim_{n_1, \ldots, n_k \to \infty} \mu(A_0 \cap T^{-n_1}A_1 \cap \cdots \cap T^{-n_k}A_k) = \prod_{i=0}^k \mu(A_i)$

when $n_{i+1} - n_i \to \infty$. A system is mixing of all orders if it's $k$-mixing for all $k$.

Bernoulli shifts are mixing of all orders, representing the strongest form of stochastic behavior in deterministic systems.