Prove that if μ is a finite measure on C and A⊆C is a μ-measurable set
such that μ(A)>0, then for every ϵ>0, there is a non-null
finitely-determined set U⊆C with
μ(U)μ(A△U)≤ϵ.
proof
definition
For each finite string p∈{0,1}∗, let Up denote the set of strings with
prefix p. Recall that {Up}p∈{0,1}∗ forms a basis for the topology of
C.
Fix ϵ>0. Let δ>0 be small enough so that δ(2+δ)≤ϵ. By regularity of
measure, there exists an open set V so that A⊆V and
μ(V)≤(1+δ)μ(A).
Write V as a union of the basis elements; that is, let P⊆{0,1}∗ so that
V=⋃p∈PUp. Assume without loss of generality that there do not exist p,q∈P
with p a proper prefix of q.
aside
To see why we can assume this, let
P~={q∈P∣∄p:p is proper prefix of q} and observe that
for every q∈P∖P~, since there exists p a proper prefix of q we
already have Up⊃Uq in the union, so ⋃p∈P~Up=⋃p∈PUp.
Then notice that for all p,q∈P, since neither is a proper prefix of the other
they must differ at a common index, so Up∩Uq=∅. Therefore
μ(V)=∑p∈Pμ(Up). This is either just a finite sum (if P itself is finite)
or a sum converging to a finite positive value; either way, there exists a
finite P′⊆P so that letting U=⋃p∈P′Up we have
μ(U)≥(1−δ)μ(V).
Note that U⊆V and also that U is finitely-determined.