Dynkin Pi-Lambda Theorem

definition: a system of sets of a set X is a collection of subsets of X. A \pi-system is a system of sets that is closed under intersection. A \lambda-system is a system of sets that contains X, is closed under monotone set difference and upper countable limits. A \sigma-algebra is a system contaning X, closed under complementation and countable union.

lemma: H is a \pi,\lambda-system iff H is a \sigma-algebra.

proof: Suppose H is a \pi,\lambda-system. Then it contains X. Let a be any set of H. Then a^C = X-a is a monotone set difference of sets of H. Hence H is closed under complement. Since H is a \pi-system that is closed under complement, it is closed under finite unions. Let \left< A_i \right> be a sequence of sets of H. Then the sequence B defined by B_i = \cup_{j = 1..i} A_j is a non-decreasing sequence of sets of H and therefore contains the limit \cup_i B_i = \cup_i A_i. Hence H is a \sigma-algebra.

Conversely, suppose H is a \sigma-algebra. Since it is closed under countable union and contains the \emptyset = X^C, it follows that it is closed under finite union. Since it is also closed under complementation, it is closed under finite intersection and therefore is a \pi-system. Let a,b \in H be such that a \subseteq b. Then b - a = b \cap a^C \in H and so H is closed under monotone set difference. Let A be a non-decreasing sequence of sets of H. Then the union \cup_i A_i is an element of H and so H is a \lambda-system.

The following theorem an an important abstract result used to prove a number of results in measure theory and is called Dynkin’s \pi,\lambda theorem.

theorem: If H is a \lambda-system containing the \pi-system P, then H contains \lambda(P).

proof: Let G(a) be the system of sets such that b \in G(a) iff a \cap b \in \lambda(P). When a \in \lambda(P) then G(a) is a \lambda-system: X \in G(a) since X \cap a = a \in \lambda(P), when c,d \in G(a) such that c \subseteq d, (d - c) \cap a = (d \cap a) - (c \cap a) and d \cap a, c \cap a \in \lambda(P) and c \cap a \subseteq d \cap a it follows that (d - a) \cap (c-a) \in \lambda(P). Finally it B is a non-decreasing sequence of sets of G(a) then the sequence C such that C_i = B_i \cap a is a non-decreasing sequence of sets of \lambda(P) and so \cup_i C_i \in \lambda(P). But \cup_i C_i = (\cup_i B_i) \cap a and so \cup_i B_i \in G(a).

If a \in P, then when b \in P, b \in G(a) since P is a \pi-system and P is a subsystem of \lambda(P). But since a \in \lambda(P), G(a) is a \lambda-system that contains P, it follows that G(a) contains \lambda(P). But this implies when a \in \lambda P, G(a) contains P and so G(a) contains \lambda P. Hence \lambda(P) is closed under intersection and so is a \pi-system.

Navigation

About

Raedwulf ….